Answers to Selected Exercises

Worked solutions to the daggered (†) and odd-numbered exercises from each chapter. Try every problem before reading its solution.

Chapter 1

Exercise 1.4

Undercapitalization is a stock problem: you never had enough money in the first place. The business opened with insufficient capital to reach the point where revenue covers costs, so the shortfall was baked in on day one and the only question was when it would surface.

Cash timing is a flow problem: you have enough money across a year, but not in a particular week. Obligations (payroll every two weeks, rent on the first, sales-tax remittance monthly, insurance annually) arrive on a schedule that does not match revenue, which is seasonal. A profitable business can miss payroll in February.

The distinction is practical, because the remedies differ. Undercapitalization is fixed only by raising more money — before opening, if you are lucky, or on unfavorable terms afterward. Cash timing is fixed by forecasting (the thirteen-week forecast in Chapter 33), by a line of credit, and by managing payables.

Exercise 1.8

The essential difference is that full service pays a table staff and owns the guest for roughly ninety minutes, which caps volume and raises labor. Counter models trade that relationship for throughput. Fast casual exists because it captured full-service perceived quality on a quick-service labor model — which, expressed in prime-cost terms, is exactly the arbitrage you would predict someone would find.

Exercise 1.9

$$\text{Prime cost} = \$25{,}200 + \$28{,}600 = \$53{,}800$$ $$\text{Prime cost \%} = \$53{,}800 \div \$84{,}000 = \mathbf{64.0\%}$$

Characterization: in the 60–65% band — workable, but tight. Profit is thin and there is very little cushion. This business would not comfortably absorb a rent escalation, an equipment failure, or a soft month. The immediate diagnostic question is which half is carrying the excess: COGS at 30.0% is unremarkable, but labor at 34.0% is toward the high end, so the schedule is the first place to look.

Exercise 1.12

Sales $960,000.

Check: $624{,}000 + 86{,}400 + 153{,}600 + 38{,}400 = \$902{,}400$; $960{,}000 - 902{,}400 = \$57{,}600$. ✓

Note that occupancy at 9% is on the high side, which is why a 65% prime cost leaves only 6% here rather than the 7% in the chapter's Figure 1.2. Prime cost must always be read alongside rent.

Exercise 1.14

Food sales are 76% of $960,000 = **$729,600**.

A 3-point improvement in food cost = $0.03 \times \$729{,}600 = \mathbf{\$21{,}888}$.

New operating profit: $\$57{,}600 + \$21{,}888 = \mathbf{\$79{,}488}$, or $79{,}488 \div 960{,}000 = \mathbf{8.3\%}$ of sales.

Interpretation: three points of food cost — the difference between counting weekly and not — moved operating profit by 38%. This is the leverage argument from §1.7 in a single calculation.

Exercise 1.16

B has the better position by 3 points, or $30,000 a year.

The likely operating choices: A is producing from scratch — whole-animal butchery, house-made stocks, pastas, breads, and pastry — which buys a very low product cost with a great deal of skilled prep labor. B is buying portioned proteins and some prepared components, accepting eight points of product cost to save eleven points of labor.

When A's approach is nonetheless correct: when the scratch production is why guests come and why they pay a premium. If A's average check is meaningfully higher than B's because the house-made program is the concept, then A is not overpaying for labor — it is investing in the differentiator that supports its price. Prime cost alone cannot tell you that; you need the check average, the repeat-visit rate, and an honest view of whether guests can actually taste the difference. (Chapter 12 gives you the tool for this: contribution margin per item, not cost percentage.)

Exercise 1.18

$$\text{Average check} = \$18{,}400 \div 392 = \mathbf{\$46.94}$$

To reach $50.00 with covers unchanged: $392 \times \$50.00 = \$19{,}600$.

Additional weekly revenue: $\$19{,}600 - \$18{,}400 = \mathbf{\$1{,}200}$. Annually: $\$1{,}200 \times 52 = \mathbf{\$62{,}400}$.

Interpretation: a $3.06 movement in average check — roughly one additional glass of wine on every third table, or a dessert on every fifth — is worth $62,400 a year to a business whose entire annual profit at 6% would be around $57,000. This is why Chapter 23 treats the table touch as a financial instrument and Chapter 24 treats the check average as a managed variable rather than a result.

Exercise 1.20

Revenue. 95 covers × $22 = $2,090 per service; × 52 = $108,680 annually.

Contribution per service.

Annually: $683 \times 52 = \mathbf{\$35{,}516}$.

Does it contribute positively? Yes — but note carefully what this calculation does and does not include. It covers product and the direct labor floor, leaving $683 per service toward fixed costs that are being paid anyway (rent, insurance, salaried management). On that basis the service is clearly worth running: you already pay the rent on Sunday whether you open or not.

What it excludes: incremental utilities, additional supplies, laundry, credit-card fees (roughly 2.5–3% of sales, about $60 a service), any additional management time, and — the largest hidden item — the effect on staff fatigue and turnover of adding a seventh operating day. It also assumes 95 covers materializes, which for a new service is an assumption, not a fact.

A defensible recommendation: run it for a defined trial period, measure actual covers and actual labor rather than planned, and set a specific closure threshold in advance (say, below 70 covers averaged over eight weeks).

Exercise 1.22

Dinner. 68 seats × 1.4 turns = 95 covers; × $46 = $4,370 per service; × 5 = $21,850/week. Brunch. 110 covers × $24 = $2,640 per service; × 2 = $5,280/week. Total: $27,130/week**; × 52 = **$1,410,760 annually.

**Gap to $1,550,000:** $\$1{,}550{,}000 - \$1{,}410{,}760 = \$139{,}240$, or $\$139{,}240 \div 52 = \mathbf{\$2{,}678 \text{ per week}}$.

Two plausible sources:

1. Turns. Moving dinner from 1.4 to 1.55 adds about 10 covers a night → $460 per service → $2,300 a week. Chapters 22 and 24 are about exactly this: pacing, table mix, and reservation policy.
2. Beverage attachment. Moving the dinner average check from $46 to $48.50 through a stronger cocktail and by-the-glass program adds $237 per service → $1,187 a week. Chapters 15 and 16.

Either alone gets most of the way; together they overshoot, which is a useful thing to notice — the $1.55M target is reachable, but only if both of the assumptions underlying it are optimistic simultaneously. That is precisely the kind of observation an assumptions register (Chapter 4) exists to surface.

Exercise 1.23

The three most likely explanations:

1. The reported food cost is wrong. "Running about 29%" almost always means invoices divided by sales, which ignores inventory movement. Actual food cost could easily be several points higher.
2. Labor is the problem, not food. A 29% food cost with a 40% labor cost is a 69% prime cost. The operator is watching the half that looks good.
3. It is a cash-timing problem, not a profitability problem. The business may be genuinely profitable while cash is consumed by debt principal (which never appears on the P&L), owner draws, sales-tax remittance, or inventory build.

The single question to ask first: "What is your prime cost?" — because it separates explanation 2 from the others in one number, and because the operator's ability to answer it tells you immediately whether explanation 1 is also in play.

Exercise 1.26

Seven possible causes of a 34.5% food cost against a 30% target:

1. Prices set against costs that have since moved (no re-costing)
2. Over-portioning
3. Waste and spoilage
4. Theft
5. Poor purchasing (no specs, no bid comparison, no par levels)
6. Uncosted specials
7. Menu mix drift — guests shifting toward lower-margin items

Ranked by speed of investigation with information the restaurant already has:

The ordering itself is the lesson: the causes people reach for first (theft) are the slowest to establish and the least likely, while the causes nobody suspects (uncosted specials, stale pricing) are both common and diagnosable before lunch.

Exercise 1.28

What the 4% increase does if covers hold. On a $1,000,000 business, sales rise to $1,040,000. COGS and most other costs are unchanged in dollars, so essentially the whole $40,000 falls to the bottom line. Prime cost falls from 67% to roughly $\$670{,}000 \div \$1{,}040{,}000 = 64.4\%$ — a 2.6-point improvement. That is a real and substantial gain.

The risk. "If covers hold" is doing enormous work in that sentence. A 4% across-the-board increase is applied without regard to which items can absorb it, and price elasticity is not uniform: a $2 increase on a $12 appetizer is far more visible than the same increase on a $34 entrée. If covers fall even 3%, the gain largely evaporates and the operator has spent goodwill for nothing.

What to do first instead.

1. Find out whether this is a price problem at all. A 67% prime cost with the split unexamined might be a labor problem. Compute both halves before touching the menu.
2. Cost the menu properly (Chapter 11) and run a menu-engineering analysis (Chapter 12). Almost always this reveals that three or four items are badly underpriced and the rest are fine — a targeted correction that raises the same revenue with a fraction of the guest-perception cost.
3. Only then price, item by item, using contribution margin rather than a uniform percentage.

Across-the-board price increases are what operators do when they have not costed their menu. They work, and they are the bluntest available instrument.

Exercise 1.30

(A 200-word note to a prospective investor who has cited the 90% figure.)

The 90% figure is not supported by any study anyone has been able to produce. The best empirical work — Parsa and colleagues, published through Cornell — found first-year failure around 26%, with roughly 60% closing or changing ownership within three years.

I want to be careful here, because the correction is not good news and I am not offering it as reassurance.

A 90% year-one failure rate would describe a random outcome — lightning. It would mean skill is not the operative variable, and there would be no point in due diligence, because there would be nothing to diligence.

The real numbers describe something harder to dismiss. Most failures occur in years two and three, which means most failed restaurants worked for a while and then declined. That is a slow financial deterioration, and it means the difference between the survivors and the casualties is largely operational discipline — cost control, labor management, and cash.

So the correction raises the bar on this conversation rather than lowering it. Do not ask me whether I can beat a coin flip. Ask me what my prime cost target is, how often I will compute it, and what happens in month twenty-nine when the lease escalates.

Exercise 1.32

The five open questions from the Chapter 1 checkpoint:

A sixth question the checkpoint does not raise but should:

What is the downside case, and what happens to the partners personally if it occurs? The plan assumes an SBA loan, and SBA lending to a new restaurant will require a personal guarantee. That means the failure scenario is not "the business closes" but "the business closes and the partners personally owe the balance." Nothing in the Chapter 1 checkpoint acknowledges this, and it is arguably the single most consequential fact about the entire venture. Chapters 5 and 39 address it; a good plan raises it in the executive summary rather than burying it.

Chapter 2

Exercise 2.2

Model answer; the exercise is a skill, not a fact. A strong response produces three sentences of the form:

1. Who and when. "We serve [specific guest] within [specific geography], on [specific occasion], about [n] times a month."
2. What and what for. "They get [specific experience] at [specific per-person price], in about [n] minutes."
3. Why us. "They choose us over [named alternative] because [reason the guest would recognize]."

The second half of the exercise is the graded half. For each sentence, the student must name the evidence that would show it false. Acceptable answers look like: sentence one is false if the reservation book shows party sizes averaging 3.8 rather than 2.1; sentence two is false if the average check settles at \$34 rather than \$44; sentence three is false if the competitor adds the same item and our covers on that occasion fall.

A response in which no sentence can be falsified has failed the exercise, and that failure is the lesson: it means the concept as stated cannot be managed, because nothing about it can be measured.

Exercise 2.5

Trade area — the geographic area from which a restaurant draws the large majority of its guests.

Two reasons drive time beats a ring. First, guests experience travel time, not distance: a three-mile ring around a riverfront or industrial site may be 40% water, rail yard, or land nobody lives on, so the ring counts population that will never come. Second, a ring assumes the street network is symmetrical, and it never is — one direction may be a highway on-ramp and the opposite a sequence of stoplights.

Three barriers: a river with limited bridges (and bridges that back up at a specific hour); a rail corridor with few at-grade crossings; a multi-lane arterial with no pedestrian crossing for a long stretch, which is permeable to cars and impermeable to walk-in trade. Others that earn credit: highways, one-way pairs, a large park or campus with no through streets, a hill, a toll, and — the most under-rated — the absence of parking on one side of a district.

Exercise 2.8

Positioning is the place a concept intends to occupy in a guest's mental map relative to the alternatives, stated as a claim about occasion, price, and reason-to-choose. It is testable on a positioning map against the real competitive set.

Concept-market fit is the demonstrated condition in which a defined guest, in a defined trade area, chooses the concept at its price often enough to produce the forecast cover count.

What a plan can and cannot establish: a business plan can establish that the claim is coherent and quantified — a defined guest, a measured trade area, a stated capture rate, a real competitive set, and arithmetic that closes. It cannot establish fit itself, because fit is a behavior, and only paying guests returning repeatedly demonstrate it. A plan that says it has "proven" concept-market fit before opening has misunderstood the term.

Exercise 2.9

$(0.50 \times 38) + (0.32 \times 54) + (0.18 \times 29) = 19.00 + 17.28 + 5.22 = \mathbf{\$41.50}$

Check: the shares sum to 1.00, and the blended figure sits between the lowest and highest PPA — two sanity tests worth running every time.

Exercise 2.11

Food subtotal required: $0.75 \times \$38.00 = \$28.50$.

Solving for dessert: entrée and starters give $22.80 + 4.60 = \$27.40$. The dessert line must supply $28.50 - 27.40 = \$1.10$, so the required attachment is $\$1.10 \div \$8.80 = \mathbf{0.125}$ — one dessert per eight covers.

Verification: food \$28.50 ÷ \$38.00 = 75.0%; beverage \$9.50 ÷ \$38.00 = 25.0%. ✓

The point of the exercise: 0.125 is a forecast of guest behavior, and one dessert per eight covers is a demanding assumption for a concept with a short time budget. If the true rate is 0.08, the food subtotal falls to \$28.10 and the check to \$37.60 — which is what happens when a concept sets a price target and back-solves without asking whether the resulting behavior is plausible.

Exercise 2.13

Per cover: $(0.50 - 0.38) \times \$14.00 = 0.12 \times \$14.00 = \mathbf{\$1.68}$.

Across 24,700 dinner covers: $\$1.68 \times 24{,}700 = \mathbf{\$41{,}496}$ of additional annual revenue.

The two other plan numbers that move:

1. The food/beverage mix. The check becomes \$47.68, of which beverage is $\$12.88 + \$1.68 = \$14.56$. That is $14.56 \div 47.68 = \mathbf{30.5\%}$ beverage on dinner, against a plan mix of 28%. Any plan line built on 72/28 — food sales dollars, beverage sales dollars — must be restated.
2. Blended COGS. At 72/28 with a 30% food cost and a 22% pour cost, blended COGS is $(0.72 \times 30) + (0.28 \times 22) = 21.6 + 6.16 = 27.76\%$, which is the plan's 27.8%. At 69.5/30.5 it becomes $(0.695 \times 30) + (0.305 \times 22) = 20.85 + 6.71 = \mathbf{27.56\%}$ — about two-tenths of a point better, because beverage carries the lower cost.

Should the operator be pleased? On margin, yes: more revenue at a lower blended cost of goods. But three caveats belong in the answer. The plan's stated mix is now wrong and must be corrected rather than quietly enjoyed (Chapter 4's assumptions register exists for this). A rising alcohol attachment shifts labor toward the bar and raises over-service and dram-shop exposure, which is a training and compliance obligation, not a windfall (Chapters 8 and 18). And an operator who sees this happen without having predicted it does not have a better bar program — they have a plan they were not watching.

Exercise 2.16

• Covers per night: $80 \times 1.25 = \mathbf{100}$
• Covers per week: $100 \times 6 = \mathbf{600}$
• Covers per year: $600 \times 52 = \mathbf{31{,}200}$

Exercise 2.18

$31{,}200 \div 3 = 10{,}400$; $\div 4 = 7{,}800$; $\div 5 = 6{,}240$; $\div 8 = 3{,}900$. Each divided by 95,000 gives the capture rate.

What the pattern implies. The required guest base falls faster than frequency rises at the low end — going from three visits to four removes 2,600 guests from the requirement, while going from five to eight removes 2,340 across three times the frequency change. In plain terms: the cheapest guest to acquire is the one you already have. Marketing money spent on reasons to return (a changing menu, recognition of regulars, a bar worth stopping at, an email list) reduces the capture requirement more efficiently than money spent on reach — and it does so permanently, whereas acquisition spending has to be repeated. Chapters 23 and 27 make this operational.

Exercise 2.20

Field exercise; no single answer. A complete response contains: a hand-drawn map with water, rail, highways, six-lane arterials, one-way pairs, and bridges marked; four directional estimates of the ten-minute reach measured at 6:30 p.m. on a weekday (not from a mapping app's off-peak default); and a stated surprise.

The surprises that recur, and which earn full credit when noticed: that the ten-minute polygon is strongly asymmetric; that a bridge or a single left turn removes a large population; that walk-in trade stops dead at an arterial even where the drive time barely changes; and that the direction with the largest population is often the direction with the worst evening travel time, because both are caused by the same thing — density feeding a constrained crossing.

Exercise 2.22

The arithmetic is correct: $90 \times 2.0 = 180$ covers per night; $180 \times \$52 = \$9{,}360$ per service; $\times 7 = \$65{,}520$ per week; $\times 52 = \mathbf{\$3{,}407{,}040}$.

Sales per seat:

• This plan: $\$3{,}407{,}040 \div 90 = \mathbf{\$37{,}856}$ per seat per year.
• Bellwether: $\$1{,}410{,}760 \div 68 = \mathbf{\$20{,}746}$ per seat per year.

The plan claims 82% more revenue per seat than Bellwether's already-optimistic estimate ($37{,}856 \div 20{,}746 = 1.82$).

The three sentences. "Your arithmetic is right and your assumptions are not. Two turns at a \$52 check means every seat in the room is used twice every night — including Monday, including February — on a check that implies at least ninety minutes at the table, which needs a four-hour dinner service running at full capacity with perfect pacing and no gaps. Rebuild it with a different check and turn figure for each night of the week, and show me the Monday."

Additional credit for noticing: seven nights assumes no closure, no holidays, and no seasonality; the model has no daypart variation; and the founder has almost certainly applied peak-night performance across the whole year, which is the single most common error in restaurant forecasting.

Exercise 2.23

What it tells you. There is no such thing as "our competitors." There are four different lists, and the establishments that appear on the highest-frequency list (F and B, for the twenty-to-forty-visit bar-first guest) are the two that a conventional "similar restaurants" analysis would have excluded entirely. A competitive set built per restaurant rather than per occasion systematically misses the competitors taking your most frequent guests.

Exercise 2.25

Field exercise; no single answer. A complete response contains a survey table of at least five establishments with format, approximate seats, dinner per-person average taken from the posted menu, dayparts, and the occasion each one owns; a price ladder; and a positioning map with the empty space marked.

The graded paragraph is the last one. Strong answers apply Case Study 1's distinction: is this space empty because no delivery model has existed to occupy it — in which case the student should be able to name the mechanism that would — or because the arithmetic does not close, meaning the check that occasion will bear cannot carry the cost structure the product requires. Answers that treat "no one is doing it" as evidence of opportunity have not done the exercise. A useful test to apply: has anyone tried it here before, and what happened to them?

Exercise 2.28

Standing inventory (working capital, not an expense):

$$5 \text{ SKUs} \times \$52 = \mathbf{\$260} \text{ tied up permanently}$$

Spoilage:

$$\$1{,}450 \times 0.14 = \$203 \text{ per month} \times 12 = \mathbf{\$2{,}436} \text{ per year}$$

Added prep labor: 35 minutes is $35 \div 60 = 7/12$ of an hour.

$$\tfrac{7}{12} \times \$21.00 = \$12.25 \text{ per day} \times 310 \text{ days} = \mathbf{\$3{,}797.50} \text{ per year}$$

Total annual cost: $\$2{,}436 + \$3{,}797.50 = \mathbf{\$6{,}233.50}$, plus \$260 of working capital.

Plates required to justify it. Gross margin per plate is $\$28.00 - \$8.40 = \$19.60$.

$$\$6{,}233.50 \div \$19.60 = 318.04 \rightarrow \mathbf{319 \text{ plates a year}}$$

$$319 \div 52 = 6.1 \rightarrow \textbf{about six plates a week}$$

Interpretation. Six plates a week across three dishes — two each — is a low bar, and most concepts will clear it. That is the honest finding, and it should be stated: this calculation usually supports the signature dishes. Its value is that it converts "should we add these?" into a number a kitchen can be held to, and it identifies the case where the answer is no — a dish selling one or two a week is costing more than it returns and should be cut or reworked (Chapter 12 does this systematically). Note also what the calculation excludes: training time, menu-printing complexity, the ticket-time cost of an extra station step, and the opportunity cost of the menu space.

Exercise 2.30

Target: $\$1{,}200{,}000 \div 52 = \$23{,}076.92$ per week.

Service values:

• Dinner: $70 \times 1.3 = 91$ covers $\times \$42 = \$3{,}822$
• Brunch: $90 \times \$22 = \$1{,}980$
• Lunch: $55 \times \$19 = \$1{,}045$

A week that reaches the target — five dinners plus two brunches:

$$(5 \times \$3{,}822) + (2 \times \$1{,}980) = \$19{,}110 + \$3{,}960 = \$23{,}070 \text{ per week}$$ $$\$23{,}070 \times 52 = \mathbf{\$1{,}199{,}640} \text{ — within \$360 of the target}$$

Alternatives, and why they are worse:

The fixed labor floor commitment. Every daypart-day opened commits a manager, a chef or lead cook, an opening prep cook, and a closing dishwasher — paid whether guests arrive or not — plus utilities, a prep cycle, and a cleaning cycle. The five-dinner-plus-two-brunch week reaches the target with seven daypart-days. The lunch route reaches a similar number with ten, adding three full days of fixed floor to capture \$1,045 of revenue per service. At even \$400 of fixed floor per lunch service, three extra lunch days cost roughly \$1,200 a week against \$3,135 of revenue — and that is before food cost, before variable labor, and before asking whether a 55-cover lunch is realistic in the first place.

The lesson to state explicitly: revenue per daypart-day is the wrong question. The right one is contribution per daypart-day, and a service that clears its fixed floor by a small margin is consuming management attention that the other services need. Chapter 32 formalizes this as break-even by daypart.

Exercise 2.32

Writing exercise; no single answer. A complete response follows the 2a–2g structure, stays under 1,200 words, and contains at minimum: three falsifiable sentences; at least two occasion segments with all seven fields; trade-area figures with a stated source and a stated geography (drive time, not a ring); a capture-rate calculation with the visit-frequency assumption made explicit; a competitive-set table including at least one non-restaurant competitor; and a positioning statement that says whether it is gap-filling or share-taking.

The second half — handing it to an unsympathetic reader and recording the first attack — is the part that teaches. In practice the first attack is almost always one of three things: the visit-frequency assumption, the price point relative to the nearest competitor, or the absence of any evidence behind the capture rate. Students should note that all three are attacks on assumptions, not on the concept, which is precisely why Chapter 4 makes the assumptions register a required section.

Exercise 2.34

Model answer. The honest finding, which the exercise is designed to produce, is that no two of the four candidates close the gap.

The best available pair leaves \$25,200 unaccounted for. The smallest sufficient combination is three: patio + Fri/Sat turns + private events = \$147,528.

A strong answer picks the two it would commit to and treats the residual honestly. For each, it supplies:

• (a) The arithmetic. Patio: 14 incremental covers × 5 nights × 22 weeks × \$46 = \$70,840. Private events: 18 events × \$2,400 = \$43,200.
• (b) The single assumption. Patio: that 22 weeks of usable outdoor weather materialize and that patio covers are incremental rather than cannibalized from the dining room — the same incremental-versus-cannibalized question Chapter 28 asks about delivery. Events: that inbound demand for a 68-seat room's private dining exists at all, in a district with a 140-seat steakhouse twelve minutes away.
• (c) How you would know within ninety days. Patio: compare dining-room covers on patio nights against matched non-patio nights; if the room's covers fall as the patio fills, the revenue is not incremental. Events: count inbound inquiries, not bookings, in the first sixty days — inquiries lead bookings by weeks and are the earlier signal.
• (d) What you would do. Patio not incremental: stop treating it as growth and re-forecast; it is a weather hedge, not a revenue line. Events not materializing: decide by month four whether to fund outbound selling (a real cost with its own return) or to remove the line from the forecast and find the money in check average instead.

Why the chapter refuses to close the gap. Because four hypotheses that sum to more than the gap are not revenue, and treating a list of possibilities as a forecast is exactly the error the book exists to prevent. The gap belongs to Chapter 24, where it is settled with covers by daypart and a revenue model rather than with optimism.

Chapter 3

Exercise 3.1

Brand is the accumulated experience of a restaurant as held in a guest's mind — what they expect will happen to them tonight. Brand identity is the set of designed, controllable inputs: name, mark, typography, palette, materials, plateware, uniform, voice.

You control identity directly. You control brand only indirectly, and only over time, by making identity and behavior consistent long enough that guests form a stable expectation.

Exercise 3.2

The six jobs: sayable, spellable, findable, clearable, fitting, durable.

A name people cannot spell after hearing it fails spellable directly and findable consequentially — the second is a knock-on of the first.

Findable costs more, and by a wide margin. Spellability failures produce a moment of mild friction. Findability failures compound: a guest who cannot type your name cannot reach your booking page, a writer who publishes the wrong spelling propagates it into other people's search behavior permanently, reviews and map listings fragment across two spellings, and every marketing dollar you spend for the rest of the restaurant's life leaks a fraction of its traffic. The remedy list in Exercise 3.20 exists because the defect is permanent but its consequences are largely purchasable away for a few hundred dollars.

Exercise 3.3

Exercise 3.5

A touchpoint is a single moment of contact in the guest journey at which a guest forms or revises a judgment.

Three from Figure 3.5 at $0: the first eight seconds at the door (#6), the walk to the table (#7), and the goodbye (#15). Others qualify — the search result, the booking call, the check.

"Free" is misleading because those touchpoints are behaviors, and behaviors revert. A $14,000 acoustic ceiling stays installed whether or not anyone thinks about it. An eight-second greeting has to be re-purchased every single shift through hiring, training, staffing, and management attention — which is recurring spend in labor, the half of prime cost that moves this week. The capital items have a one-time cost and an annual carry; the free items have no capital cost and an entirely recurring one.

Exercise 3.7

The menu as brand artifact is the menu treated as a physical object that communicates price point, formality, competence, and confidence before a word of it is read — through its size, format, stock and finish, typography, length, voice, and what it omits.

Chapter 10 covers the menu as a commercial instrument: which items go on it, how they are sequenced, how the eye moves across a panel, how prices are presented, and how those choices move mix and margin. The split is deliberate. Chapter 3 owns what the object says; Chapter 10 owns what the document does.

Exercise 3.8

Brand drift is the slow degradation of a restaurant's atmosphere and identity through individually trivial decisions that aggregate into a room guests describe as "not what it used to be."

Its analogue is Chapter 1's cost drift.

The shared structural feature: both are made of individually defensible decisions that no single report shows. Cost drift is a protein price that moved, a portion that grew an ounce, a special nobody costed — none of which appears as a line item called "drift." Brand drift is four lamps at the wrong color temperature, a replacement plate from a different supplier, a reverted playlist — none of which appears anywhere at all. In both cases the aggregate is material and the detection mechanism must be deliberately constructed: a weekly flash report for one, a 4:45 walk and a quarterly review coding for the other.

Exercise 3.9

Dining floor 1,240 sq ft.

Difference between tightest and most generous: $82 - 53 = \mathbf{29}$ seats.

Value at one additional turn a night, two nights a week, $48 check:

$$29 \times \$48 = \$1{,}392 \text{ per night}$$ $$\$1{,}392 \times 2 = \$2{,}784 \text{ per week}$$ $$\$2{,}784 \times 52 = \mathbf{\$144{,}768} \text{ per year}$$

Read it carefully before you act on it. That is a capacity figure, not a revenue figure. It assumes both nights fill at the denser layout and that the check holds at $48 in a room that is now substantially louder and tighter — which is precisely the assumption §3.7 says you must test rather than assume. It also ignores the three demand-constrained nights, on which the 29 seats are worth exactly nothing. The number's real job is to establish that density is a six-figure decision, so that it gets made deliberately rather than by whoever draws the floor plan.

Exercise 3.11

48 seats, 240-minute peak window, 85% utilization.

Seat-minutes available: $48 \times 240 = 11{,}520$.

(a) Revenue per constrained night

The loud room wins a constrained night by $111.50.

(b) Weekly net

• Two constrained nights: $-\$111.50 \times 2 = -\$223.00$
• Three unconstrained nights, covers unchanged at 54: $54 \times 3 = 162$ covers, gaining the $3.50 check lift only: $162 \times \$3.50 = +\$567.00$

$$\text{Net} = \$567.00 - \$223.00 = \mathbf{+\$344.00 \text{ per week}}$$

Annualized: $\$344 \times 52 = \mathbf{\$17{,}888}$.

(c) Payback

$$\$16{,}000 \div \$17{,}888 = 0.89 \text{ years} \approx \mathbf{11 \text{ months}}$$

Note what carries the result: the gain comes almost entirely from the three nights where nothing about capacity changed. That is the whole lesson. Turn time only converts to money when seats are the binding constraint, and for most independents that is two nights out of five.

Exercise 3.13

Fix costs $3,450. Two seats at a $52 average check produce $2 \times \$52 = \$104$ per night when sold.

$$\$3{,}450 \div \$104 = 33.2 \rightarrow \mathbf{34 \text{ nights}}$$

Is it a good deal? On a five-night week across a five-month winter — roughly 100 services — you need to sell those seats on about a third of them. If the restaurant is capacity-constrained even two nights a week in winter, that is 40+ opportunities and the fix pays back inside the first season.

What it depends on: whether the room is ever actually full in winter. If February runs at 55% occupancy every night, those two seats were never going to be sold, the revenue calculation returns zero, and the honest justification has to come from somewhere else — the complaints, the reviews, the guests who were seated there and did not return, and the staff time spent apologizing. Which is a real justification, just not this one. Do not dress a service argument up as a capacity argument; if you cannot make the capacity case, make the service case honestly.

Exercise 3.15

Opening package $14,200; breakage and loss 38% annually.

$$\text{Annual replacement} = \$14{,}200 \times 0.38 = \mathbf{\$5{,}396}$$ $$\text{Five-year total} = \$5{,}396 \times 5 = \mathbf{\$26{,}980}$$ $$\$26{,}980 \div \$14{,}200 = \mathbf{1.90\times} \text{ the original package}$$

The implication: you are not buying a set of plates, you are entering a five-year supply relationship in which you will buy roughly two more sets. That reframes the specification decision entirely. The variable that matters most is not how the plate looks in the showroom; it is whether the manufacturer will still make it in year four. Specify from ranges that have been in continuous production for a decade, buy long-lead items deeper than feels comfortable at opening, and treat a beautiful plate from a small producer with no distribution as what it is — a decision to accept a mismatched pass in year three, or to re-buy the entire tabletop.

Exercise 3.17

Which check would have caught it: the common-law market check — check 4. The unregistered restaurant has rights arising from actual use in commerce in its geographic market, and it will appear in no registry search. This is why a clean USPTO result is a necessary and wildly insufficient condition.

What to do, in order:

1. Stop the spend. Do not order the sign, do not print, do not submit permit applications in the name. Sunk cost is much cheaper at week minus four than at week plus four.
2. Establish the facts. How long have they used it, in what geography, for what services, and is there any registration or pending application anywhere? Have they expanded, franchised, or licensed?
3. Get real advice. This is a trademark question with facts in it, and forty miles is exactly the distance at which the answer is genuinely uncertain — close enough to plausibly share a market, far enough to plausibly not. Use an attorney who does this routinely; the fee will be a fraction of Exercise 3.18's number.
4. Weigh the three options honestly. Proceed and accept the risk; proceed with a modifier that creates real distinction (rarely as protective as founders hope, and it dilutes the name); or go back to your Gate 4 survivors, of which you kept two for exactly this reason.
5. Take the lesson forward. Run the market check before Gate 4, not after. It is a search engine, a map search of your metro and the two adjacent ones, and thirty minutes.

Exercise 3.19

For a small format, Gate 2 (findability) and Gate 4 (durability) both become more important, for opposite reasons.

Gate 2, because a truck or ghost-kitchen brand has almost no passive discovery. A brick-and-mortar restaurant is found by people walking past it; that channel is worth real money and it costs nothing. A truck's location changes, and a virtual brand has no street presence at all, so essentially 100% of discovery runs through search, maps, platform listings, and social. A name that is hard to find is not inconvenient for a small format — it is close to fatal.

Gate 4, because small formats are frequently a step rather than a destination (Chapter 30). A name that fits a taco truck and cannot survive on a brick-and-mortar awning, a catering contract, or a retail label imposes a rebrand at exactly the moment you can least afford one. The founders who graduate cleanly are usually the ones who named for the business they wanted rather than the vehicle they could afford.

Gate 3 (clearance) does not get easier, and small-format operators skip it more often, because the stakes feel lower. They are not lower; they are merely deferred.

Exercise 3.21

A defensible fast-casual journey — the useful part of this exercise is the differences, so:

Steps that disappear: the host stand and the seating decision (#6–7 collapse into a queue), first contact with a server (#10), the table touch, and the check as a service moment (#14 becomes a pre-meal transaction).

Steps that change owner: the greeting moves from a host to whoever is at the register or the line; pacing moves from a server to the kitchen and the queue; the goodbye usually has no owner at all.

Steps that appear: the queue itself (its length, its legibility, whether you can see the menu board from the back of it), the menu board as a decision surface under time pressure, the order handoff, the condiment and drink station, the self-bussing moment — which is a genuine hospitality touchpoint that most operators treat as a maintenance task — and, in most modern fast casual, the digital ordering journey running in parallel.

The important observation: the counter-service format removes most of the touchpoints that cost labor and keeps most of the ones that cost capital. That is the whole prime-cost logic of the category, and it is also why atmosphere carries proportionally more weight in fast casual: with fewer human moments, the room does more of the work.

Exercise 3.22

Print run: $1,150 for 400 menus.

Weekly contribution lost by not repricing:

$$190 \times \$0.85 = \$161.50 \text{ per week}$$

$$\$1{,}150 \div \$161.50 = 7.1 \rightarrow \mathbf{7 \text{ weeks}}$$

What it tells you: print cost and food cost are connected, and the connection runs the wrong way. The menu did not raise the plate cost; it made lowering it inconvenient, which produces the same result more slowly — which is precisely Chapter 1's definition of drift.

Note also that the real-world delay is rarely a deliberate seven-week calculation. It is "we'll do it at the next menu change," which in a seasonal restaurant is ten to twelve weeks, by which point the delay has cost $1,600 to $1,900 to avoid an $1,150 expense. Own the file, print small, print often.

Exercise 3.23

Two acceptable appetizers, in the stated voice:

• Grilled bread, cultured butter, sea salt
• Chicory salad, walnut, aged sheep's cheese, cider vinegar

Three entrées:

• Half chicken from the hearth, roasted roots, salsa verde
• Whole trout, embers, brown butter, lemon
• Cannellini and greens, garlic, chili, hearth toast

One dessert:

• Buttermilk pudding, honey, black pepper

The badly rewritten versions ("our famous farmhouse chicken, lovingly kissed by the flames of our signature artisanal hearth and nestled on a bed of chef's seasonal root vegetables") change three things at once. They move the claim from the ingredient to the restaurant, which reads as insecurity. They use words no server would say out loud, which means the printed voice and the spoken voice now disagree — a consistency failure a guest registers without naming. And they promise a level of production the plate cannot deliver, which converts a good dish into a small disappointment. Note that the first version also happens to be shorter, cheaper to typeset, and easier to change.

Exercise 3.24

$$\text{Annualized} = \$298.25 \times 52 = \mathbf{\$15{,}509}$$

Over the actual fourteen months (about 60.7 weeks): $\$298.25 \times 60.7 \approx \mathbf{\$18{,}100}$.

The design decision that most plausibly caused it: a menu that is expensive or slow to change — an outsourced designer who must be engaged for every edit, a large minimum print run, a laminated or letterpress format, or a layout so tightly typeset that changing one price re-flows the page. Any of these converts a five-minute decision into a project, and projects get deferred.

Note the shape of the loss: item B has the largest cost increase but item A and item C together account for $175 of the $298. Volume matters as much as the increase, which is why a re-costing review should be sorted by units sold, not by the size of the price move.

Exercise 3.25

Stations that change: the sound check moves from "volume at the mark" to a zoned schedule that steps up through the night; the light check acquires a late scene; the walk itself has to happen twice, because a room at 4:45 tells you nothing about a room at 10:30.

Stations added: the back bar and speed rail (facing, fruit, ice), the bar top itself (sticky surfaces are the bar's version of a burnt bulb), the bathroom at a much higher frequency, the door/queue area, the ID-check station and lighting at the door — you cannot check identification in a room you cannot see — and glassware polish and par, which is the bar's plateware problem and drifts the same way.

At 10:30 p.m. you would check things that do not exist at 4:45: whether the room is over-serving, whether the volume has crept up (it always creeps up, because staff acclimate), whether the door control is working, whether the bathrooms have degraded, whether the ice is holding, and whether the back bar still looks like a display or has become a work surface. A late-night walk is also a compliance walk — see Chapter 8 on dram shop obligations and Chapter 25 on the food-safety end of a late close.

Exercise 3.27

Q1 — cost to buy: $46,000 of capital. Straightforward, and it should be held against the whole atmosphere schedule, not considered in isolation: on the running project's $116,920 total, this single line would be 39% of everything.

Q2 — cost to keep: this is where the proposal is most likely to fail. Custom pottery from a small producer means: a higher unit cost on replacement; a breakage rate that is probably worse than commodity plateware, because handmade pieces are heavier, less uniform, and often not dish-machine optimized; a lead time measured in months rather than days, which forces you to carry a deeper par; and an existential dependency on one small business continuing to exist and continuing to make this piece. At a 38–40% annual breakage rate, $46,000 of tabletop implies something like $17,000–$18,000 a year of replacement, at lead times that could leave you short mid-season.

Q3 — what does it change? Only two mechanisms are plausible. Check average, if the plate genuinely reads as more valuable and supports a higher price — testable by moving prices on two comparable items and watching mix and covers. Return visits, if it is distinctive enough to be remembered and photographed — partially testable by coding reviews and social mentions before and after (Figure 3.6) for whether anyone mentions the plate at all.

The likely honest answer: almost nobody notices plateware consciously, everybody notices four mismatched whites on a pass, and the second risk is much larger than the first opportunity. If the partner wants it, the disciplined version is a small custom run for two signature items on a commodity base, which caps both the capital and the dependency.

Exercise 3.28

Percentage changes

As a percentage of sales

$$\text{Year 1: } \$20{,}731 \div \$1{,}550{,}000 = \mathbf{1.34\%}$$ $$\text{Year 2: } \$26{,}845 \div \$1{,}720{,}000 = \mathbf{1.56\%}$$

Sales grew 11.0%. The atmosphere carry grew 29.5%. Four of the seven lines grew at roughly 3–4%, which is ordinary inflation and price escalation and is not a leak.

Sizing the two overruns. If tabletop replacement had grown with sales it would be $\$3{,}920 \times 1.11 = \$4{,}351$; the excess is $4,829. If uniforms had grown with sales they would be $\$2{,}600 \times 1.11 = \$2{,}886$; the excess is $1,534**. Together **$6,363 — more than the entire $6,114 increase, because the third leak masked part of it.

The three leaks

1. Tabletop replacement, +134%. Breakage does not double because volume rose 11%. The candidates are a dish-machine or racking problem, a bussing-station layout that breaks glassware, a staffing change (new bussers, high turnover), a discontinued SKU forcing wholesale replacement rather than piecemeal, or theft.
2. Uniforms, +70%. Uniform spend is a turnover proxy. A 70% increase in a year almost certainly means departures rose sharply — which, if true, is also sitting in the tabletop line, in training hours, and in the guest experience. Chapters 17 and 21.
3. Menu print, −55%, and this is the alarming one. Spending less on menu printing is not a saving; it is evidence that the menu stopped changing. In a restaurant with a seasonal menu, a halved print budget means either the seasonal changes were dropped or the reprint discipline was abandoned — and a menu that is not being reprinted is a menu that is not being repriced. Exercise 3.24 shows what that costs, and it is an order of magnitude larger than the $1,441 "saved."

Monday's three questions to the general manager:

1. Show me the breakage log for both years, broken out by category, and tell me what changed in the dish pit or the bussing station.
2. How many new hires did we onboard last year by position, and what is our turnover rate against the prior year?
3. When did the menu last change, what has moved in plate cost since, and who owns the re-costing cycle?

Exercise 3.29

The proposal is almost certainly wrong, and the diagnosis is the interesting part.

A rebrand changes brand identity — the name, mark, print, and uniforms. Figure 3.1 says identity is an input; brand is the output. If the restaurant "isn't landing," the failure is nearly always downstream of identity: the concept did not fit the market (Chapter 2), the room fails at sound, light, temperature, or density (§3.3), the guest journey has unowned touchpoints (§3.4), execution is inconsistent (§3.6), or the price and the experience do not match. A new mark fixes none of those and costs the Exercise 3.18 number.

What evidence would make a rebrand right? Genuinely: a name that is legally compromised; a name that has been permanently associated with a public failure such as an outbreak; a name that is actively unfindable or confusable with a stronger local business; or a deliberate concept pivot (Chapter 39) where the restaurant is becoming a materially different business and the old name promises the wrong thing.

The more likely diagnosis is available for the price of ninety minutes: code the last quarter's reviews by theme (Figure 3.6), measure the room's sound and light, walk the sixteen touchpoints as a guest, and read the menu-mix and check-average trend. In my experience the answer is in there four times out of five, and it costs a fraction of a rebrand to fix.

The honest caution: occasionally the answer really is that the concept was wrong, and a founder who reaches for a rebrand is sometimes half-right — they have correctly sensed a problem and misdiagnosed its layer. Take the sensing seriously and refuse the prescription.

Exercise 3.30

The case for 84 dBA. Energy is a real product. Loud rooms read as busy and desirable, which is itself a demand signal; a full-sounding room makes a half-full room feel occupied. The competitive set is louder, and a quiet room in a loud category can read as empty or as stuffy, both of which are worse than loud. Louder rooms turn faster, which is money on constrained nights. And the hard, reflective finishes producing the noise are frequently the cheap, durable, and visually correct ones for the concept.

The case against. It costs check average — descriptions do not land, wine conversations get abandoned, dessert and coffee are conversation courses. It costs a meaningful segment of your market: older guests, guests with hearing loss, and anyone who came to talk, which for a neighborhood restaurant is most of them. It generates the single most common negative review theme in the industry. And, decisively, it is a workplace. A server or bartender in that room is exposed for five hours a shift, five shifts a week, for years — and occupational exposure limits published by OSHA and NIOSH begin to bite in exactly this range. Your guests self-select into two hours of it; your staff cannot.

What I would do. Design toward the low-to-mid seventies — lively, not loud — and buy the difference with treatment that does not read visually: ceiling baffles above the sightline, upholstered banquette backs, felt behind slats, acoustic plaster. Then measure, monthly, and log it. The point is not that 84 is forbidden; it is that a room at 84 should be there because someone chose it with the staff-exposure question answered, not because nobody priced the ceiling.

What I would measure. dBA mid-room at peak, weekly. Average check and beverage attachment by night. Dessert and coffee incidence. Review themes, coded quarterly. And staff turnover in the positions that stand in the loudest part of the room — which is the measurement almost nobody takes and the one most likely to settle the argument.

Exercise 3.31

(a) The schedule

Note that Saturday carries two host shifts, not one long one — 10.5 hours across brunch and dinner. That is a scheduling fact worth seeing early, because it is the day the position is hardest to fill and the day it matters most.

(b) Cost

$$37.5 \text{ h} \times \$17.00 = \mathbf{\$637.50 \text{ per week}}$$ $$\$637.50 \times 52 = \mathbf{\$33{,}150 \text{ per year}}$$

(c) As a percentage of sales

$$\$33{,}150 \div \$1{,}550{,}000 = \mathbf{2.1\%}$$

(d) Share of the labor line

$$\$33{,}150 \div \$500{,}000 = \mathbf{6.6\%}$$

One position, never in the kitchen, never carrying a plate, consuming one dollar in fifteen of the entire labor budget. That is the correct frame for the argument in part (e).

(e) Cutting the host at 8:30. Saving: roughly 1.75 hours a service × 7 = 12.25 hours a week × $17 = $208.25 a week, or $10,829 a year — about a third of the position.

What degrades, from Figure 3.5: #6 the first eight seconds (a server mid-service cannot own the door, so late-arriving guests wait unacknowledged); #7 the walk to the table and seating decision (table assignment stops being managed, which pushes parties into whatever is open rather than what balances stations and pacing — a revenue problem Chapters 22 and 24 will quantify); and #15 the goodbye, which in most restaurants is the host's, and which disappears entirely.

What to measure before and after: seat-to-greet time, average turn time and its variance after 8:30, covers seated after 8:30 versus before the change, late-service check average, and review mentions of waiting or greeting. Run it for a month, not a week. If the room is genuinely quiet after 8:30, the cut is correct and the arithmetic supports it; if it is not, you have traded $10,829 for a degraded last hour on your two best nights, which is a bad trade at any price.

Exercise 3.33

A defensible answer uses the nine-row format from the checkpoint — materials, palette, light, type, voice, plateware, uniform, music, and "the one thing we will not do" — with every row stated as a rule someone could be held to rather than an adjective.

The row that separates a real document from an aspirational one is the last. "We will not have televisions," "we will not use a laminated menu," "we will not play music with lyrics before 7:00," "we will not seat a party of two at a four-top before 7:30" — each of these costs something real, and that is what makes it a constraint rather than a preference. A "one thing we will not do" that costs nothing has not been written yet.

Exercise 3.34

The landlord version sells a tenant: neighborhood fit, the kind of guest it draws, hours and dayparts, what it does for the rest of the property, and the credibility of the operators. It is written in adjectives and outcomes because that is what a leasing agent underwrites.

The contractor version is a specification: 56 dining seats at 18–20 sq ft each, roughly 1,060 sq ft of dining floor; 2700K throughout on five dimming scenes; a target of 63–70 dBA at peak with the treatment strategy named; 68–72°F at every table including the entry zone; the material palette with maintenance expectations stated.

Where they say the same thing in different units: "warm and intimate" becomes 2700K and five scenes. "A room where people linger" becomes 18–20 sq ft per seat and a target dining time. "You can have a conversation" becomes 63–70 dBA. "Comfortable year-round" becomes 68–72°F at every table.

Why the second is the useful document: it can be bid, built, inspected, and enforced. The first can only be agreed with. A founder who has only written the first will get whatever the contractor already knows how to build — and will discover the gap on opening night, when it is a change order rather than a bid item.

Exercise 3.35

The strongest submissions do three things: (1) they carve atmosphere costs out of the budget lines they normally hide in — construction, furniture, smallwares, pre-opening, operating supplies — rather than inventing a new budget; (2) they carry an annual column for every capital line, including the lines where the honest answer is a small one; and (3) they express the totals three ways: capital as a percentage of project cost, carry as a percentage of planned sales, and carry per cover.

The cut-first question is the real one. A good answer names a line, states what is lost, and distinguishes lines that can be deferred (a second tabletop par, patio dressing, the sound system's patio zone) from lines that cannot be added later without re-opening the room (acoustic treatment above a finished ceiling, dimming infrastructure, banquette framing, anything behind drywall). The general rule: cut what you can buy in month nine; never cut what the ceiling has to come down for.

Exercise 3.36

The three assumptions most likely to be wrong, and what settles each:

1. The $3.50 check-average lift attributed to the quieter room. It is the load-bearing assumption in the largest atmospheric investment, and it is a belief, not a measurement. Settled by: average check and beverage attachment by night, measured against a comparable period, plus dessert and coffee incidence — Chapters 23 and 24. If it comes back against you: the treatment was still probably worth doing for the review effect and the staff-exposure question, but you have learned that the room does not buy its way out through the check, which means the revenue model has to come from covers instead — a much harder problem, and one to surface immediately rather than at year end.

2. That a space exists in the trade area with roughly 1,060 sq ft of usable dining floor at an occupancy cost the plan can carry. The brand brief writes a requirement; the market may not have one. Settled by: the site search itself — Chapters 6 and 7. If it comes back against you: you are choosing between a denser room (which changes the check assumption above and the concept underneath it) and a higher rent, and the plan should say in advance which way it will go.

3. That the atmosphere schedule survives contact with real bids. Every number in Figure 3.7 was estimated before a single bid was received, and construction and signage in particular vary by an order of magnitude with city and landlord. Settled by: the bid process — Chapters 6 and 7. If it comes back against you: the answer is the deferral discipline in Exercise 3.35, not a proportional trim across every line, because the lines are not equally reversible.

Strong answers may nominate a fourth: that guests in this trade area want this room at all, which is Chapter 2's question and is not fully settled by anything until the restaurant opens.

Exercise 3.37

The memo has to do four things, and the fourth is what makes it survive.

What gets checked — the ten stations, listed, with the standard for each stated as a pass/fail rather than a judgment ("every lamp lit" not "lighting looks good").

Who signs — a named position per shift, and the signature is on the log, not on a habit. Make it the same log every time and keep it where the closing paperwork lives.

What happens when a defect is found — fix it now if it takes under two minutes; log it with a named owner and a date if it does not; escalate anything that takes a seat out of service or that recurs twice.

What happens when the walk is skipped — this is the paragraph most memos omit and the only one that matters. Say plainly that the walk is the first thing that goes on a short-staffed night, that nothing bad will happen the night it is skipped, and that this is exactly why it has to be logged rather than remembered. Then say what you will actually do: review the log weekly, ask about gaps, and treat a pattern of missed walks as a management performance issue rather than a scolding for whoever was expediting at 5:30.

A memo that acknowledges the failure mode honestly will survive a skeptical manager. One that asserts the walk is important will not.

Chapter 4

Exercise 4.4

A pro forma profit-and-loss statement and an actual profit-and-loss statement have the same rows, the same subtotals, and the same arithmetic. The difference is entirely in where the numbers came from.

An actual P&L reports measured history: the food cost line is beginning inventory plus purchases minus ending inventory, counted by somebody, against sales recorded by a point-of-sale system. Every figure is a fact about a period that happened, and it can be audited.

A pro forma reports beliefs about a period that has not happened. Its food cost line is a target percentage applied to a forecast, and its forecast is an assumption applied to a physical constraint. Nothing in it can be audited, because there is nothing yet to audit.

What that obliges you to attach is the assumptions register. Since the statement itself cannot be checked, the only thing a reader can check is your reasoning — and the register is the reasoning, written down. A pro forma delivered without its assumptions asks the reader to accept a conclusion with the argument removed, which is precisely the request a competent reader refuses.

The practical test: for every number in a pro forma, you should be able to point at either a formula or a register row. If a number is neither, you typed it and forgot.

Exercise 4.8

What it is for. Sensitivity analysis exists to answer one question: which of my assumptions matters most in dollars? You move one at a time, hold everything else at plan, and record the effect on operating profit.

Why the output is a ranking rather than a range. A range — "revenue will be between $1.2M and $1.7M" — is close to useless operationally, for two reasons.

First, a range implies a probability distribution nobody has. To state a range honestly you would need to know how likely each outcome is, and for a restaurant that has never opened, nobody does.

Second, and more importantly, a range does not tell you what to do. A ranking does. If turns are worth $60,087 and pour cost is worth $13,020, you now know that the reservation policy deserves five times the management attention that the jigger does, that the turns assumption is the one to defend hardest in the loan meeting, and that the first metric to instrument after opening is covers per service, not bottle yield.

The ranking also does something a range cannot: it identifies what doesn't matter. Learning that three full points of pour cost costs $13,020 — less than half the cost of losing two operating weeks — reallocates attention, and reallocated attention is the actual deliverable.

Exercise 4.9

Dinner. 92 seats × 1.5 turns = 138 covers per service. 138 × $38 = **$5,244 per service. $5,244 × 6 services = **$31,464 per week. $31,464 × 52 = **$1,636,128** annually.

Two notes worth making even though the problem didn't ask. First, 1.5 turns on 92 seats six nights a week is a busy room; before believing the figure you would want to know the average table time and the length of the service window. Second, the 52 is doing unexamined work — at $31,464 a week, one closed week costs $31,464 of revenue and roughly $15,700 of operating profit.

Exercise 4.11

Dinner. 54 seats × 1.6 turns = 86.4 → 86 covers per service (round down; you cannot seat four-tenths of a guest, and rounding down is the conservative direction).

Brunch. 96 covers × $21.00 = **$2,016 per service × 1 service = $2,016 per week**.

Weekly revenue: $16,770 + $2,016 = $18,786. Annual revenue: $18,786 × 52 = **$976,872.**

Sanity check the turn rate implied by brunch: 96 covers ÷ 54 seats = 1.78 turns. That is fast, and it is the first number in this forecast I would interrogate. A single weekly brunch running 1.78 turns while dinner runs 1.6 is possible — brunch tables turn faster — but it should be observed, not assumed.

Exercise 4.13

The gap.

$$\$1{,}200{,}000 - \$1{,}092{,}400 = \mathbf{\$107{,}600}$$ $$\$107{,}600 \div 52 = \mathbf{\$2{,}069 \text{ per week}}$$ $$\$107{,}600 \div \$1{,}200{,}000 = \mathbf{9.0\%} \text{ of the headline}$$

Three named bridge items (constructed to the same standard as Figure 4.3 — each one a separate claim with its own arithmetic):

And that residual is the point of the exercise. Three plausible bridge items get you to $86,880 — 81% of the gap — and leave $20,720 with nowhere to come from. You now have exactly three honest options: find a fourth named source and defend it, revise the headline down to $1,179,280, or disclose the residual as unallocated. What you may not do is quietly raise the turns assumption until the arithmetic closes.

Which would I trust least? The takeout line, at $36,400 — the largest single item and the one most likely to be cannibalized rather than incremental. A guest who orders takeout on a Tuesday may be a guest who would otherwise have sat in the dining room at a higher check, in which case the channel adds packaging cost and kitchen load and very little revenue. Chapter 28 tests exactly this. The events line is the most trustworthy of the three, because fourteen events either get booked or they don't, and you find out inside a quarter.

Exercise 4.15

| | Base case ($1,410,760) | Plan ($1,550,000) | |---|---|---| | Sales per seat (68 seats) | $20,746 | $22,794 | | Sales per square foot (2,800 sq ft) | $503.84 | $553.57 |

$\$1{,}550{,}000 \div 68 = \$22{,}794$ · $\$1{,}550{,}000 \div 2{,}800 = \$553.57$ $\$1{,}410{,}760 \div 68 = \$20{,}746$ · $\$1{,}410{,}760 \div 2{,}800 = \$503.84$

What the comparison tells a reader that the annual figure does not: it converts the forecast into a unit that can be compared against something outside the plan. "$1,550,000" is only meaningful relative to this restaurant's own cost base. "$22,794 per seat" can be held against four competitors you actually visited, against the operator's own previous room, or against a broker's comparables — and the gap between the two columns, about $2,050 a seat, is a compact statement of exactly how much of the plan rests on the revenue bridge.

Note the honest limit: a cross-check is only as good as the comparison you can source. Neither figure means anything on its own, and a benchmark quoted from an article you cannot produce is not a comparison. Go get four menus and a seat count.

Exercise 4.16

Revenue $1,320,000; mix 70% food / 30% beverage.

The footing check, which is the actual exercise. $\$286{,}440 + \$83{,}160 = \$369{,}600$ ✓ $\$369{,}600 + \$435{,}600 = \$805{,}200$; $\$805{,}200 \div \$1{,}320{,}000 = 61.0\%$ ✓ $\$805{,}200 + \$102{,}000 + \$191{,}400 + \$39{,}600 = \$1{,}138{,}200$ $\$1{,}320{,}000 - \$1{,}138{,}200 = \mathbf{\$181{,}800}$ ✓ Percentages: $28.0 + 33.0 + 7.7 + 14.5 + 3.0 = 86.2$, leaving $13.8\%$ ✓

Reading it: a 61.0% prime cost is one point above the full-service benchmark from Chapter 1 — workable but with no cushion — and occupancy at 7.7% is on the high side, which is why the operating profit lands at 13.8% rather than higher. Note that this restaurant looks worse than Bellwether's plan on every controllable line and still shows a double-digit operating profit, which should tell you something about how aspirational plan-stage pro formas generally are.

Exercise 4.19

Ranking by dollars: labor and other operating tie at $13,200, then food at $9,240, then beverage at $3,960. A point of labor is worth 3.3 times a point of beverage cost.

Why this is not the same as the ranking by difficulty. Labor is worth the most and is also the fastest to move — you can change a schedule this week. Beverage cost is worth the least and is among the easiest to control precisely, because a standard pour is a mechanical procedure. Other operating is worth as much as labor and is the hardest of the four to move, because it is a dozen small contracts, subscriptions, and card-processing rates that each require a separate negotiation.

The practical consequence: rank by dollars to decide where to look, then rank by difficulty to decide what to do this month. An operator who only ranks by dollars will spend a quarter fighting the other-operating line for gains a schedule change could have delivered in a week.

Exercise 4.21

The likely mechanism: operating leverage. Revenue grows faster than fixed costs, so the fixed base is spread over a larger number and the percentage falls without anyone becoming better at anything.

A numerical illustration that produces exactly the stated relationship. Take year one at $1,000,000 of revenue, $300,000 of annual fixed cost, and 50¢ of contribution on each sales dollar:

• Year 1: $(\$1{,}000{,}000 \times 0.50) - \$300{,}000 = \mathbf{\$200{,}000}$
• Year 2 at +11% revenue and +5% fixed cost: $(\$1{,}110{,}000 \times 0.50) - \$315{,}000 = \mathbf{\$240{,}000}$

$\$240{,}000 \div \$200{,}000 = 1.20$ — an 11% revenue increase producing a 20% profit increase, with the contribution rate completely unchanged. Nobody bought better, portioned better, or scheduled better. The rent simply got divided by a bigger number.

What would make me doubt it. One specific thing: cost percentages that improve materially in the same column. If year two also shows food cost dropping two points and labor dropping two points, then the plan is claiming leverage and operational improvement simultaneously — and since the two are being asserted together, the reader cannot tell how much of the profit growth depends on the operator suddenly getting better at their job. That is the version to distrust.

The diagnostic question: hold every cost percentage flat and recompute year two. If most of the profit growth survives, it is leverage and it is credible. If it collapses, the plan is a promise about behavior, and behavior is the thing nobody can underwrite.

Exercise 4.24

Fixed costs $460,000; each marginal sales dollar carries 51¢ of variable cost and leaves 49¢.

At $1,300,000 of sales: $$\$1{,}300{,}000 \times 0.49 = \$637{,}000 \text{ of contribution}$$ $$\$637{,}000 - \$460{,}000 = \mathbf{\$177{,}000} \text{ operating profit } (13.6\%)$$

At $1,150,000 of sales: $$\$1{,}150{,}000 \times 0.49 = \$563{,}500$$ $$\$563{,}500 - \$460{,}000 = \mathbf{\$103{,}500} \text{ operating profit } (9.0\%)$$

**The difference: $73,500** — which is exactly 49% of the $150,000 revenue difference, as it must be.

What the relationship tells you. Revenue fell 11.5% ($150,000 ÷ $1,300,000). Operating profit fell 41.5% ($73,500 ÷ $177,000). That is a leverage factor of about 3.6×: every one percent of revenue miss costs roughly three and a half percent of profit.

This is the single most important structural fact about a restaurant's cost base and it is why the entire chapter exists. It also runs in both directions, which is the part operators forget in a good quarter — an 11.5% revenue beat would have produced a 41.5% profit increase, and it is exactly that experience that persuades an operator to sign a second lease. Chapter 32 formalizes this.

Exercise 4.25

$$\text{Sales required} = \frac{\$460{,}000}{0.49} = \mathbf{\$938{,}776}$$

Check: $\$938{,}776 \times 0.49 = \$460{,}000$ ✓

Across 52 weeks that is $\$938{,}776 \div 52 = \mathbf{\$18{,}053}$ of average weekly sales.

The weekly figure is the useful one, because it is the only form of this number a manager can actually see. "We must average $18,053 a week" is an instruction. "$938,776 of annual sales" is a fact about a year you cannot observe until it is over. Chapter 32 takes this one step further and converts it into covers per night, which is the form a floor manager can act on at 6:15 on a Tuesday.

Exercise 4.26

Working from the plan: food sales $1,116,000 · beverage sales $434,000 · total sales $1,550,000 · 68 seats · five dinner services at a $46 check · 50.24¢ of contribution on each marginal revenue dollar.

(a) Food cost 30% → 32% of food sales. $$0.02 \times \$1{,}116{,}000 = \mathbf{-\$22{,}320}$$

(b) Dinner turns 1.40 → 1.30. $68 \times 1.30 = 88.4 \rightarrow 88$ covers, against 95 on plan — a loss of 7 covers per service. $$7 \times \$46 = \$322 \text{ per service} \times 5 = \$1{,}610 \text{ per week} \times 52 = \$83{,}720 \text{ of revenue}$$ $$\$83{,}720 \times 0.5024 = \mathbf{-\$42{,}061}$$

(c) Labor 32.3% → 33.8% of sales. $$0.015 \times \$1{,}550{,}000 = \mathbf{-\$23{,}250}$$

(d) Pour cost 22% → 24% of beverage sales. $$0.02 \times \$434{,}000 = \mathbf{-\$8{,}680}$$

The ranking:

Three observations worth more than the ranking itself.

A tenth of a turn — one table on a busy night — costs nearly twice what two full points of food cost do. Two points of food cost is a serious operational failure; a tenth of a turn is a reservation policy and a pacing decision.

Ranks 2 and 3 are within $930 of each other, which means a one-and-a-half-point labor problem and a two-point food problem are, to this plan, the same problem. An operator who intuitively treats food cost as the "real" number and labor as the soft one has the relationship backwards.

And note that the four moves are not equally likely. This ranking measures consequence, not probability. A two-point food-cost overrun in a first year is close to routine; a tenth of a turn is a genuine unknown. A complete risk view multiplies consequence by likelihood — which is a judgment, which is why the register has a confidence column as well as a dollar column.

Exercise 4.28

The argument. An 11.4% operating profit is a good result for an independent full-service restaurant. Chapter 1 put the typical range at three to ten percent. So the combined downside case — turns missing by a tenth and a half, the check missing by three dollars, and labor missing by three full points, all at once — still lands above the middle of the range that ordinary operating restaurants actually achieve.

That cannot both be a genuine downside case and a plausible one. Either the plan has found a structurally advantaged business, or the downside case is not a downside case: it is the base case with three of fifteen assumptions moved, while the other twelve are still sitting at the favorable end of their ranges. The second reading is more likely, because the arithmetic of the pro forma shows that Bellwether is better than Chapter 1's generic full-service dollar on all five cost lines simultaneously — 2.2 points on COGS, 1.7 on labor, 1.9 on occupancy, 3.0 on other operating, and 1.0 on G&A. A sensitivity analysis that stresses three of those and leaves the rest untouched has tested the plan's arithmetic, not its optimism.

The two lines I would examine first.

Other operating, at 14.0% ($217,000). It is a category estimate with no vendor quotes behind it, and it contains a dozen separate costs — utilities in a building with a wood-fired hearth, credit-card processing on nearly all of $1.55M of sales, the technology stack, insurance, repairs, marketing, laundry, smallwares replacement. Every one is a separate number nobody has yet obtained, and a plan that gets this line three points wrong loses $46,500 — as much as the labor scenario.

The fixed/variable split itself. The sensitivity work rests on $517,700 of fixed cost and 49.76¢ of variable cost per sales dollar, and that split is a planning estimate, not a built-up figure. If the true fixed base is larger — a bigger salaried management floor, more of the other-operating line non-flexible than assumed — then every downside scenario in the chapter is understated, because more of the revenue miss falls straight through to profit. Chapter 19 builds the labor half properly and Chapter 32 builds the whole structure. Until then, the downside case is itself an estimate resting on an estimate, and the register should say so.

Exercise 4.30

The original: "Labor cost may be higher than projected."

That sentence contains no movement, no dollars, no consequence, and no action. It is a hedge — it exists so that the writer can say later that they mentioned it. A reader gains nothing from it, and worse, a reader who sees several sentences like it concludes that the writer does not know which risks are real.

A version that does the job:

Risk: the year-one labor line. The plan carries all-in labor at 32.3% of sales ($500,000), which is a trained-team number. A first-year operation with a new kitchen, a new floor, overlapping training hours, and no demand history to schedule against will plausibly run three points higher. At 35.3%, labor is $547,150 — an increase of **$47,150 — prime cost rises from 60.0% to 63.1%, and operating profit falls from $261,020 to **$213,870, or 13.8% of sales. Debt service of $69,500 remains covered.

What we would do. The staffing guide in Section 7 sets the cover count at which each shift comes off the floor: the second server at 55 covers on the books by 5:00 p.m., the second prep shift at a four-week average below 85 dinner covers. The trigger is a four-week rolling labor average above 34.0%, reviewed in the weekly flash report from week one. At that point we cut the Tuesday prep shift and move the third server to on-call, which recovers approximately 1.8 points.

What we would not do. We would not reduce the training hours in the first ninety days. The training ramp is the reason the 32.3% is reachable at all, and cutting it buys two points this quarter and costs four next year in turnover and waste.

What makes the second version work. It states the movement (three points), the dollar consequence ($47,150), the effect on the book's central benchmark (prime cost to 63.1%), the effect on the bottom line ($213,870), the specific action, the trigger for taking it, the expected recovery, and — the detail that reads as genuinely professional — the thing they would refuse to do and why. A reader finishes that paragraph knowing the writer has thought about the failure and would recognize it in week six.

Exercise 4.32

Three additional register rows, each with the arithmetic that produced its exposure. (Any three defensible rows are acceptable; these are the three I would add first.)

A17 — card processing. If 92% of $1,550,000 is settled on cards, that is $1,426,000 of card sales. At an effective 2.6% all-in rate — interchange plus assessments plus the processor's markup — the cost is $37,076 a year, buried inside the 14.0% other-operating line and almost never modeled separately. Each quarter-point of effective rate is $3,565. That is more than a quarter-point of pour cost and it is negotiable, which is a combination worth noticing.

A18 — the opening ramp. The plan's average week is $1,550,000 ÷ 52 = $29,808. If the first twelve weeks run 15% below that — $25,337, a shortfall of $4,471 a week — the year is short 12 × $4,471 = $53,652 of revenue, which at 50.24¢ on the dollar is **$26,955** of operating profit. This row matters more than its dollar value suggests, because the shortfall lands in the exact weeks when the working-capital reserve is the only thing between the business and a missed payroll. Chapter 33 is where it becomes a cash question rather than a profit one.

A19 — utilities. A wood-fired hearth carries a make-up air load, and make-up air is conditioned air being pushed into a building and immediately exhausted. A $800-a-month miss is $9,600 a year, which is more than two points of pour cost. Nobody has looked at a utility bill for this space.

The general lesson of the exercise: the chapter's sixteen-row register looked thorough and it was not. Other operating at 14.0% was a single row covering at least eight separate assumptions, any one of which is worth more than the pour-cost line. A register's weakest point is always the line item that summarizes a category — and the discipline is to decompose it until each row is a number somebody could actually go and get a quote for.

Exercise 4.34

Bellwether's occupancy of $95,200 decomposes as: base rent 2,800 sq ft × $28 = $78,400, plus NNN charges 2,800 × $6 = **$16,800**. Check: $78,400 + $16,800 = $95,200 ✓

Escalating the base rent 3% annually, NNN held flat:

($78,400 \times 1.03 = \$80,752$; $\$80,752 \times 1.03 = \$83,174.56 \rightarrow \$83,175$.)

Restated operating profit:

Total three-year cost of the escalation: $0 + $2,352 + $4,775 = $7,127.

Why that figure badly understates the real problem — three reasons.

One: the lease is ten years, not three, and escalation compounds. By year ten the base rent is $78,400 × 1.03⁹ = **$102,294 — an increase of $23,894, or 30.5% over the opening figure. Total occupancy would be $119,094 against $95,200. On the plan's year-three revenue of $1,850,000 that is 6.4%** of sales rather than 5.1%, and the plan's single best structural advantage has largely evaporated. This is precisely the mechanism Chapter 1 described: the restaurant in Figure 1.3 was killed by a rent escalation arriving in year three against a cushion that had already been spent.

Two: NNN charges do not hold flat, and they frequently escalate faster than base rent. Common area maintenance, property tax, and insurance are typically reconciled against actual landlord cost, which means the tenant absorbs increases without any negotiated cap unless one was negotiated. Holding NNN at $16,800 for ten years is not conservative; it is fictional. Chapter 6 covers what to negotiate here, and a cap on controllable CAM is worth real money.

Three: the escalation is worst when revenue is weakest. The modeled cost assumes revenue grows to $1,720,000 and $1,850,000 on schedule. If it doesn't — if the $139,240 bridge fails, or turns come in at 1.25 — then a rising fixed rent divides into a flat revenue number, and occupancy percentage rises from both directions at once. The escalation and the revenue miss are not independent risks; they compound.

The register entry should therefore not read "escalation costs $7,127 over three years." It should read: occupancy is modeled flat, actual terms unknown; a 3% base escalation costs $7,127 across the plan's three-year window and roughly $24,000 a year by the end of the initial term; NNN is unmodeled and uncapped. Resolve in Chapter 6 before funding.

Chapter 5

Exercise 5.3

The SBA guaranty is a promise from the Small Business Administration to a participating lender that it will cover a portion of the lender's loss if the loan defaults, provided the lender followed the program's rules. It protects the lender, and it protects the lender from the borrower's default.

A personal guarantee is a promise from an individual to the lender that they will repay the business's debt personally if the business does not. It protects the lender, and it protects the lender from the borrower as well.

Both instruments run in the same direction. Nothing in the SBA program protects the borrower. The guaranty exists to make lenders willing to lend into the collateral gap that Figure 5.3 describes; the personal guarantee fills whatever the federal guaranty and the collateral do not.

The practical consequence: an operator who believes "the SBA is backing me" has the relationship exactly inverted, and will be surprised at closing.

Exercise 5.8

Owner injection at 20% of \$410,000 = **\$82,000**.

$$\text{SBA} = \$410{,}000 - \$82{,}000 - \$50{,}000 - \$45{,}000 = \$233{,}000$$

Uses total \$190,000 + \$120,000 + \$30,000 + \$40,000 + \$30,000 = **\$410,000**. Sources equal uses, so the statement foots.

As a check on reasonableness, the loan is 56.8% of project cost, which is in the same neighborhood as Bellwether's 54.0% — unsurprising, since both projects have a similar mix of leasehold, equipment, and soft cost.

Exercise 5.10

Apply the principle: outside money funds what the provider could repossess or already owns. The TI allowance can only fund improvements to the landlord's building. The lessor funds only equipment. The SBA loan is flexible but is happiest against hard assets. Whatever remains falls to the owner.

Rows foot to the use totals; columns foot to the source totals; the grand total is \$410,000 both ways.

Uses funded entirely or almost entirely by the owner: pre-opening (\$40,000) and the working-capital reserve (\$30,000), plus \$12,000 of smallwares. That is \$82,000 — the entire injection. Nobody else will fund training, opening payroll, licensing, or a reserve, because none of it can be repossessed and sold.

Worth noting against Bellwether: here the 7(a) loan does fund \$18,000 of smallwares, which 7(a) permits. Bellwether's structure has the loan funding only construction and equipment. Both are legitimate; the more soft cost a loan funds, the thinner the collateral behind it becomes, which is a lender-by-lender judgment rather than a rule.

Exercise 5.12

Option A — \$200,000 at 9%, seven years, \$3,218 a month

Option B — \$200,000 at 10.5%, ten years, \$2,699 a month

The trade: Option A costs \$53,568 less in total interest** and **\$6,228 more every year in cash.

Which to take, for a restaurant build-out: Option B, in almost every case. A start-up restaurant's binding constraint is not lifetime interest cost, it is cash in the first two years — the exact period Chapter 1 identifies as the fragile one. \$6,228 a year is \$519 a month, which in a bad February is the difference between an uncomfortable week and a call to the landlord. The extra \$53,568 is real money, but it is spread over a decade and paid out of a business that by then either works or does not. Term drives cash; rate drives cost. Buy the cash.

The answer flips if you have an unusually strong reserve, a seasonal business with predictable surpluses, or a genuine intention to prepay — in which case confirm first that no prepayment charge applies.

Exercise 5.14

Face value.

• Four months of free rent on \$110,000 annual occupancy: \$110,000 × 4 ÷ 12 = \$36,667.
• Additional TI allowance: \$30,000.

Free rent wins on face by \$6,667.

Which is actually worth more. Free rent, and by more than the \$6,667 gap suggests, for three reasons. First, timing: rent abatement lands during build-out and ramp-up, when you have zero revenue and maximum outflow, whereas TI is typically reimbursed after completion and lien release — so the TI dollars require you to spend them before you have them, and the abated dollars require nothing. Second, certainty: abatement is automatic, while a TI reimbursement can be delayed by an inspection, a missing lien waiver, or a disputed invoice. Third, cash-cost equivalence: \$36,667 of rent you do not pay is \$36,667 you do not have to borrow, which at Bellwether-like rates is worth several thousand dollars of avoided interest as well.

The provision that could reverse the answer: the definition of what the abatement covers. Free rent that abates base rent only while you remain liable for NNN charges, taxes, insurance, and CAM during those four months is worth substantially less than the headline figure. Read the abatement clause for the word "base," and price it on what is actually abated. (Chapter 6 takes this apart properly.)

Exercise 5.15

$$\text{Total debt service} = \$41{,}000 + \$12{,}500 + \$9{,}600 = \$63{,}100$$

$$\text{DSCR} = \frac{\$148{,}000}{\$63{,}100} = \mathbf{2.35}$$

Characterization: above the 1.35–2.00 "comfortable" band and into the 2.00+ range. On an operating business this is a genuinely strong position — the restaurant produces more than twice its required payments and would absorb an ordinary bad quarter without a conversation.

On a projection, read it as a prompt rather than a reassurance. A 2.35 forecast coverage on a start-up is an invitation to check the revenue assumption, the labor line, and whether the owners are paid — which is exactly what Exercise 5.17 does.

Exercise 5.17

As presented:

$$\text{DSCR} = \frac{\$210{,}000}{\$96{,}000} = 2.19$$

With the owners paid a market salary:

$$\text{DSCR} = \frac{\$210{,}000 - \$135{,}000}{\$96{,}000} = \frac{\$75{,}000}{\$96{,}000} = \mathbf{0.78}$$

The business does not generate its own payments. The apparent 2.19 was entirely an artifact of two people working for nothing.

What it demonstrates: the numerator of a coverage ratio is a judgment, not a fact, and the single largest judgment in an owner-operated business is how the operators' own labor is treated. A plan that omits owner compensation is not describing a business; it is describing a business plus an undisclosed subsidy from two people who will eventually need to pay rent. Lenders adjust for this routinely, which is why §5.4 tells you to ask how a lender computes the numerator. Apply the same skepticism to your own plan first — it is much cheaper to find at your kitchen table than across a desk.

Exercise 5.19

Required operating profit at a DSCR of 1.35:

$$1.35 \times \$69{,}500 = \$93{,}825$$

Required revenue, using the model in §5.4 (34.0¢ of every sales dollar survives the costs that move with volume; \$265,980 of costs do not move):

$$R = \frac{\$93{,}825 + \$265{,}980}{0.34} = \frac{\$359{,}805}{0.34} = \mathbf{\$1{,}058{,}250}$$

As a share of plan: \$1,058,250 ÷ \$1,550,000 = 68.3%.

Why the true break point is higher. The model treats all labor as variable, and labor is not. The fixed labor floor from Chapter 1 — salaried managers, the chef, the opening prep cook, the closing dishwasher — does not shrink by a third when revenue does. At roughly two-thirds of plan revenue the labor percentage rises, so operating profit falls faster than this model predicts, and the revenue needed to hold a 1.35 coverage is therefore above \$1,058,250. Chapter 32 separates fixed from variable properly; its answer will be less flattering than this one.

Exercise 5.22

$$\text{Monthly payment} = \$85{,}000 \times 0.0243 = \mathbf{\$2{,}065.50}$$

Against Bellwether's SBA note at 16.2¢: this lease costs about 80% more per dollar per year, almost entirely because it amortizes over four years rather than ten. The finance cost of \$14,144 on \$85,000 over four years is not outrageous as a rate; the burden is, relative to a ten-year note.

The operational reading: an equipment package financed this way consumes a great deal of cash in the exact years when a new restaurant has the least of it. That is sometimes the right trade — the equipment is what generates the revenue — but it has to appear in the cash forecast (Chapter 33), not just in the sources-and-uses statement.

Exercise 5.24

$$\$22 \times 6 \text{ kegs} \times 52 \text{ weeks} = \mathbf{\$6{,}864 \text{ per year}}$$

$$\$6{,}864 \times 5 \text{ years} = \mathbf{\$34{,}320}$$

You pay \$34,320** across the term for a system worth **\$9,000 installed — roughly 3.8 times its value — and at the end you do not own it.

Two circumstances under which you would take it anyway:

1. The premium is not really a premium. If the alternative supplier's price is not genuinely comparable — different products, worse delivery, no draft service, minimum orders you cannot hit — then \$22 is not the cost of the equipment, it is partly the cost of the beer. Price the same product from both, or you are comparing nothing.
2. You do not have \$9,000, and you do not have credit. At opening, cash preserved has a value far above its face amount, and a draft system that is installed and serviced on somebody else's dime is revenue you can start earning immediately. Taking expensive money is a legitimate decision; taking it without knowing it is expensive is not.

A third consideration worth raising in class: negotiate the term. A five-year commitment on a \$9,000 asset is the problem. The same deal over two years costs \$13,728 — still expensive, and survivable. Ask.

Exercise 5.27

The loan itself.

Effect on a plan already carrying \$69,500 of debt service against \$261,020 of operating profit:

$$\text{New total debt service} = \$69{,}500 + \$12{,}276 = \$81{,}776$$

$$\text{DSCR} = \frac{\$261{,}020}{\$81{,}776} = \mathbf{3.19} \quad (\text{down from } 3.76)$$

Coverage falls by 0.57 and remains strong. But do not stop at the ratio. Two things matter more than the number. First, the senior lender's negative covenants very likely restrict additional debt and additional liens; a family loan taken without disclosure can be a technical default on a much larger obligation. Disclose it, and expect the lender to want it subordinated — repayment behind theirs — which is normal and is exactly what a subordination agreement is for. Second, \$12,276 a year is \$1,023 a month of real cash leaving in months when the restaurant may have none, owed to somebody you will see at a wedding.

Exercise 5.30

$$4 \text{ investors} \times 3 \text{ visits} \times 12 \text{ months} = 144 \text{ visits per year}$$

A workable policy, and the justification.

Investor dining benefit: each investor receives a standing 25% discount on food and beverage, up to twelve visits per calendar year, excluding Friday and Saturday after 6:00 p.m. and excluding private events and holidays. Reservations required. The benefit is personal and not transferable.

Two sentences of justification. First, a discount rather than a comp keeps the restaurant's product cost covered on every visit and preserves the feeling of a benefit without the arithmetic of a gift — at 25% off, a \$95 tab still returns \$71.25 against \$26.41 of product cost. Second, the blackout protects the seat-hours the business actually needs to sell, which matters far more than the food: Chapter 24 will show you that a Saturday 7:30 four-top is the most perishable inventory you own, and giving it away is a much larger cost than the \$3,803 of food.

Exercise 5.32

A model answer. Roughly 250 words, written to inform rather than persuade.

Two ways you could do this, and they are genuinely different.

You could lend me the money. Say \$50,000 at 6% over five years. I would pay you about \$967 a month, every month, and by the end you would have received about \$58,000 — your \$50,000 back plus roughly \$8,000. The payments start whether or not the restaurant is doing well, which is the risk to me. The upside is capped at that \$8,000, which is the cost to you: if the restaurant does extraordinarily well, you still get \$58,000. If it fails, you are an unsecured creditor behind the bank, and you would probably not be repaid.

Or you could buy a piece of it. If \$50,000 bought 20%, we would be saying the restaurant is worth \$250,000 the day it opens — a number neither of us can defend, because it has no revenue and no history. There are no payments, which protects me in a bad year. But if the business matures into \$150,000 a year of distributable cash, your 20% is \$30,000 a year, indefinitely — a 60% annual return on your \$50,000. You would earn several times the loan's total in a few years. If it fails, you get nothing, and you get nothing before the lender gets nothing.

What I want you to notice: the loan is finite and the equity is not. The loan risks my cash flow; the equity risks how we both feel about this in year six. There is no correct answer. There is only the one we write down before any money moves, with a lawyer, and then honor.

Exercise 5.34

Sizing a six-week reserve.

On plan, Bellwether's total operating cost is \$1,550,000 − \$261,020 = \$1,288,980 a year, or \$1,288,980 ÷ 52 = **\$24,788 a week**.

$$6 \times \$24{,}788 = \mathbf{\$148{,}728}$$

$$\$148{,}728 - \$45{,}000 = \mathbf{\$103{,}728 \text{ of additional capital}}$$

Could each source supply it, and what would it demand?

(Check: \$103,728 × 0.162 ≈ \$16,804. \$69,500 + \$16,800 = \$86,300. \$261,020 ÷ \$86,300 = 3.02.)

The change I would actually make. Increase the equipment lease and the loan together rather than either alone — lease a further \$25,000 of the equipment package (about \$6,300 a year) and add roughly \$79,000 to the note (about \$12,800 a year), for total additional debt service near \$19,100 and a coverage ratio still above 2.8. That funds a six-week reserve without asking the partners for money they do not have.

What it costs: about \$19,100 a year of cash, forever until the paper is retired, and a materially weaker collateral position on the loan. Whether that trade is right is precisely the question Chapter 33 exists to answer, and it is the honest reason the reserve is carried as an open question rather than settled here.

Chapter 6

Exercise 6.1

Second-generation space is a space previously built out for restaurant use, in which some restaurant-specific infrastructure may be reusable. The word doing the work is may.

Three systems you typically inherit, with the hazard attached to each:

Also acceptable: floor drains and slab penetrations (moving one means cutting concrete), restrooms (accessibility compliance under an older code vintage), and base-building HVAC (age and the question of who owns replacement).

Exercise 6.3

A natural breakpoint is the sales level at which the percentage rent exactly equals the base rent:

$$\text{Natural breakpoint} = \frac{\text{annual base rent}}{\text{percentage rate}}$$

An artificial breakpoint is any negotiated figure, and in practice it is almost always lower than the natural one, which means the landlord begins collecting percentage rent before the base rent has been "earned back."

The natural breakpoint favors the tenant, and by a large margin. Worked example from Exercise 6.26's abstract: base rent \$74,400 at a 6% rate gives a natural breakpoint of \$1,240,000. The abstract sets an artificial breakpoint of \$900,000. At \$1,300,000 of sales:

• Natural: $(\$1{,}300{,}000 - \$1{,}240{,}000) \times 0.06 = \$3{,}600$
• Artificial: $(\$1{,}300{,}000 - \$900{,}000) \times 0.06 = \$24{,}000$

A difference of \$20,400 a year, from one number in one clause.

The circumstance in which a tenant would prefer an artificial breakpoint is when it is negotiated higher than natural — usually in exchange for a lower base rent. That structure lowers fixed cost and raises variable cost, which is genuinely attractive for a concept with uncertain volume: you pay less in a bad year. It is the same logic as the percentage-rent conversions used in the 2020 settlements described in Case Study 1.

Exercise 6.5

A change order is a written amendment to the construction contract that alters the scope, the contract sum, the schedule, or all three.

The five categories:

1. Concealed conditions — what was behind the wall or under the slab.
2. Authority-required — the plans examiner, fire marshal, or health inspector requires something not on the drawings.
3. Design errors and omissions — the drawings were wrong or incomplete.
4. Allowance reconciliation — the placeholder figure was wrong.
5. Owner-requested changes — you changed your mind.

Only category 5 is fully under the operator's control, and it is the only optional one. Note the practical implication: categories 3 and 4 are reduced almost to zero by finishing the drawings before bidding, which means the operator actually controls three of the five, just not at the moment the change order appears.

Exercise 6.6

The rent commencement date is the date the obligation to pay rent begins — frequently neither the lease commencement date nor the date you open.

It can be worth more than a dollar a foot because a dollar a foot is a stream and a delayed opening is a lump, and the lump lands in the months when you have no revenue at all. On Bellwether's numbers, a dollar a foot is \$2,800 a year; a two-month construction overrun under a delivery-of-possession trigger is \$15,867 of rent paid on a closed building, all of it in one quarter with zero sales against it.

The three formulations, most to least tenant-friendly:

1. The earlier of opening for business or a date certain far enough out to be comfortable.
2. A fixed number of days after delivery of possession — the landlord's preference, because it starts a clock the landlord controls the beginning of and the tenant controls the end of.
3. A calendar date, negotiated when everybody was optimistic.

Case Study 2 is the whole argument: partners who traded formulation 1 for a dollar a foot lost \$35,750 in a single stretch to save \$2,600 a year.

Exercise 6.7

A punch list is the list of incomplete or defective items identified at substantial completion, which the contractor must correct before final payment.

Retainage — commonly 5–10% withheld from each draw — is connected to it because retainage is the money you have not yet paid, and therefore the only leverage you have left. Once it is released, your only remedies are the warranty, the contract, and goodwill, in ascending order of unreliability.

Releasing retainage against a promise costs you leverage at precisely the moment the contractor's crew has moved to another job and their attention has moved with it. Walk the punch list yourself, twice — once in daylight and once at night with the lights set the way they will be in service — and release retainage against completed work, not against intentions.

Exercise 6.9

A co-tenancy clause conditions the tenant's rent obligation, or obligation to remain open, on the presence of specified other tenants or a minimum occupancy level in the property.

It is most standard for in-line tenants in enclosed malls and large shopping centers, where an anchor tenant genuinely determines whether anyone walks past your door.

The usual remedies, in escalating order: reduced rent (often a switch to percentage-only rent for the period), then, if the condition persists beyond a stated cure period, a right to terminate.

The practical rule from §6.5: ask for a co-tenancy remedy whenever the landlord is selling you a story about the property — a food-and-beverage cluster, an anchor arriving next year, a redevelopment. If the story is part of why you are paying this rent, ask for a remedy if the story does not happen.

Exercise 6.10

$$\text{All-in rate} = \$24.00 + \$8.25 = \$32.25 \text{ per square foot}$$ $$3{,}400 \times \$32.25 = \mathbf{\$109{,}650 \text{ per year}}$$ $$\$109{,}650 \div 12 = \mathbf{\$9{,}137.50 \text{ per month}}$$ $$\$109{,}650 \div \$1{,}380{,}000 = \mathbf{7.9\%}$$

Check: base $3{,}400 \times \$24 = \$81{,}600$; NNN $3{,}400 \times \$8.25 = \$28{,}050$; $\$81{,}600 + \$28{,}050 = \$109{,}650$. ✓

Reading: 7.9% is inside the 6–10% full-service range but at the top of the working band, with almost no room for the forecast to disappoint. Note also that the NNN is \$8.25 — more than a quarter of the total occupancy cost — which makes the CAM cap, exclusions, and audit right unusually valuable here.

Exercise 6.11

$$\text{Required revenue} = \frac{\text{annual occupancy cost}}{\text{target occupancy \%}}$$

At 6.5%: $\$109{,}650 \div 0.065 = \mathbf{\$1{,}686{,}923}$

At 9.0%: $\$109{,}650 \div 0.09 = \mathbf{\$1{,}218{,}333}$

What the second figure tells an operator: \$1,218,333 is the floor beneath which this lease becomes a structural problem rather than a line item. The restaurant can lose 12% of its forecast revenue and still be at 9% occupancy — which is survivable but leaves essentially nothing for the rest of the cost structure once the other fixed costs are added. Any forecast within about 12% of the plan should be treated as the operating case, not the downside case.

Exercise 6.12

3% annual escalation, 2,600 sq ft starting at \$30.00:

**\$1.25 step every two years**, starting at \$30.00:

Check: $2 \times (\$78{,}000 + \$81{,}250 + \$84{,}500 + \$87{,}750 + \$91{,}000) = 2 \times \$422{,}500 = \$845{,}000$. ✓

**The step schedule is cheaper by \$49,182 over ten years.** It is also \$10,772 a year cheaper in year ten (\$101,772 versus \$91,000), which matters more than the total because the year-ten rate is usually the base from which renewal-option rent is calculated.

And the non-financial benefit is real: you can state your rent for every year of the term to the dollar, which makes the break-even model in Chapter 32 an instrument rather than an estimate.

Exercise 6.13

$$\text{Natural breakpoint} = \$96{,}000 \div 0.07 = \$1{,}371{,}429$$

$$\text{Sales over breakpoint} = \$1{,}640{,}000 - \$1{,}371{,}429 = \$268{,}571$$

$$\text{Percentage rent} = \$268{,}571 \times 0.07 = \mathbf{\$18{,}800}$$

$$\text{Total rent} = \$96{,}000 + \$18{,}800 = \mathbf{\$114{,}800}$$

Note the structure. Above the breakpoint, the tenant pays 7 cents of every incremental sales dollar to the landlord — which means every improvement this book teaches (a two-dollar increase in average check, one more turn on a Saturday, a successful private-events program) is taxed at 7%, forever, with the landlord bearing none of the cost of producing it.

Exercise 6.14

$$\text{Estimated: } 2{,}900 \times \$4.75 = \$13{,}775 \text{ per year} = \$1{,}147.92 \text{ per month}$$ $$\text{Actual: } 2{,}900 \times \$5.90 = \$17{,}110 \text{ per year} = \$1{,}425.83 \text{ per month}$$ $$\text{True-up invoice} = \$17{,}110 - \$13{,}775 = \mathbf{\$3{,}335}$$

The new monthly estimate is **\$1,425.83**, an increase of \$277.92 a month going forward.

Where it hits and when. The expense economically belongs to the year just closed — an accrual accountant would say it should have been recognized month by month. But the bank account is hit in a single lump in the month the reconciliation arrives, which is typically two to four months after the landlord's year-end, and the restaurant almost certainly has not accrued for it. So an operator on cash-basis reporting sees a one-month occupancy spike of \$3,335 in, say, month fourteen or fifteen, in addition to a permanently higher monthly charge.

This is Chapter 33's argument in miniature: profit is an opinion about a period; cash is a fact about a Thursday. The disciplined countermeasure is to accrue a CAM reserve monthly against an expected overage, and to negotiate the cap and audit right that limit how large the surprise can be.

Exercise 6.15

Base and NNN: $\$95{,}200 \div 4 = \mathbf{\$23{,}800}$ (three months is one quarter of the annual figure).

Base only: $\$78{,}400 \div 4 = \mathbf{\$19{,}600}$.

Difference: \$4,200, which is exactly one quarter of the annual NNN ($\$16{,}800 \div 4 = \$4{,}200$).

Where it comes from: "free rent" in a landlord's first draft very frequently means abatement of base rent only, with the triple-net charges payable throughout the abatement period. This is not a trick — it is the market default in many places, because the landlord is passing through actual costs and does not want to absorb them. But it is a \$4,200 difference on a single word, and the ask ("abatement of base rent and all additional rent") takes one sentence.

Exercise 6.16

$$2{,}800 - 2{,}660 = 140 \text{ phantom square feet}$$ $$140 \times \$34.00 = \mathbf{\$4{,}760 \text{ per year}}$$ $$\$4{,}760 \times 10 = \mathbf{\$47{,}600 \text{ over a ten-year term at a flat rate}}$$

(On Bellwether's actual step schedule the figure is larger still, because the rate rises: 140 sq ft across a ten-year base-rent total of \$300 per square foot is \$42,000 of base rent alone, plus the NNN component.)

What should have prevented it: have the space measured by your own architect during the due-diligence period, state the measured area in the lease as a fact rather than as a landlord representation, and — in a multi-tenant building — get the load factor and the method of measurement stated explicitly. If the landlord will not agree to a measured figure, negotiate a re-measurement right with a rent adjustment.

The general principle: you are paying per unit for a quantity written in a document by somebody who benefits from it being larger. Count it yourself. This is the same discipline as counting your own walk-in (Chapter 13), applied to real estate.

Exercise 6.17

$$240 \text{ covers} \times 50 \text{ weeks} = 12{,}000 \text{ covers}$$ $$\$118{,}000 \div 12{,}000 = \mathbf{\$9.83 \text{ per cover}}$$

At 15% fewer covers:

$$12{,}000 \times 0.85 = 10{,}200 \text{ covers}$$ $$\$118{,}000 \div 10{,}200 = \mathbf{\$11.57 \text{ per cover}}$$

An increase of \$1.74 per guest in the rent each guest must carry.

What the operator can do: raise the average check (menu engineering, Chapter 12; beverage attachment, Chapter 15; coursing and pacing, Chapter 24), or recover covers (marketing, Chapter 27; turns and table mix, Chapter 24; a second daypart or channel, Chapters 28 and 29).

What the operator cannot do: change the \$118,000. It is contractual. And they cannot reprice the menu quickly enough or often enough to track a covers shortfall in real time — the menu is printed, the guests have expectations, and price increases have their own demand consequences.

That asymmetry is the entire lesson of §6.4: the numerator is fixed and the denominator is the only thing you can work on, so the denominator had better have been forecast honestly.

Exercise 6.19

$$\text{Contingency} = \$318{,}000 \times 0.12 = \$38{,}160$$ $$\text{Total} = \$318{,}000 + \$41{,}000 + \$38{,}160 = \mathbf{\$397{,}160}$$ $$\text{Contingency as \% of total} = \$38{,}160 \div \$397{,}160 = \mathbf{9.6\%}$$

Why the two percentages differ: the 12% is computed on hard cost only (\$318,000), while the 9.6% is computed on a total that also includes \$41,000 of soft costs and the contingency itself. Same dollars, different denominators.

This matters practically, and it is a common way people talk past each other. When a contractor, lender, or partner says "we have a 10% contingency," ask ten percent of what. A 10% contingency on hard cost is roughly a 7.8% contingency on the total project construction line, and if somebody computed it on the whole \$620,000 project they have a number that means nothing at all, because equipment and pre-opening do not generate construction change orders.

Exercise 6.21

Check: $6{,}400 + 2{,}150 + 9{,}300 - 4{,}000 + 5{,}850 + 3{,}100 + 7{,}600 + 1{,}250 = \$31{,}650$. ✓

The contingency is exhausted at CO 007, in week 10, on the fire marshal's rated enclosure at the exhaust shaft. Final overrun: \$3,650.

Two observations worth making. First, the log was in trouble long before week 10 — after CO 003 in week 4, more than 60% of the contingency was gone on a job that was barely started, and that is the week a disciplined operator escalates rather than the week the number goes negative. Second, note that CO 004 is a credit: value engineering is a live tool throughout a job, not a decision made once at bid time.

Exercise 6.23

$$3{,}100 \times \$142 = \mathbf{\$440{,}200}$$

Cash required during construction: the full \$440,200.

The \$60,000 tenant-improvement allowance is paid on completion, against paid invoices, unconditional lien waivers, and (in most leases) the certificate of occupancy and sometimes the opening of business. It is money that arrives after every dollar of it has already been spent by the tenant.

Why "funded" and "liquid" are different questions. On a sources-and-uses statement (Chapter 5), this project is funded: \$440,200 of uses against \$440,200 of sources, one of which is the TI allowance. That statement is true and it is not the operative one. The operative question is sequencing: the tenant must be able to write \$440,200 of checks before receiving \$60,000 of it back. If the capital stack was assembled assuming the TI reduces the cash requirement, the project is \$60,000 short of liquidity for several months at exactly the moment the contractor expects to be paid.

The general principle is theme five of this book: cash is not profit, and a funded project is not a liquid one. Landlord money, equipment financing draws, and loan disbursements all arrive on schedules that are somebody else's, and the schedule is as much a term as the amount.

Exercise 6.25

$$\text{Mechanical: } \$52{,}000 - \$18{,}000 = \$34{,}000$$ $$\text{Plumbing: } \$19{,}000 - \$8{,}000 = \$11{,}000$$ $$\text{Total increase} = \mathbf{\$45{,}000}$$

Against a \$9,000 contingency:

$$\$45{,}000 - \$9{,}000 = \mathbf{\$36{,}000 \text{ that must be found inside the \$310{,}000 line}}$$

Which means \$36,000 of currently-budgeted scope has to come out. Note carefully what this does not mean: it does not mean the project is over budget. The \$310,000 is fixed by the project budget, so the arithmetic is forced — the budget did not go up; the scope went down. The operator's job is to decide which \$36,000 of scope, and that is a design conversation, not an accounting one.

Note also that this is a constructed what-if for practice, not a forecast. Chapter 7 does the actual sizing and pricing. The point of the exercise is the structure of the problem: an allowance that re-bids high consumes the contingency first and then eats scope, and neither event produces a bigger budget.

Exercise 6.26

$$2{,}400 \times (\$31.00 + \$9.00) = 2{,}400 \times \$40.00 = \mathbf{\$96{,}000 \text{ per year}} = \$8{,}000 \text{ per month}$$

The five clauses that will cost this tenant the most:

1. Assignment: "landlord consent in landlord's sole and absolute discretion." This is the most expensive line in the abstract, and it costs nothing today. It means the tenant cannot sell the business without the landlord's unreviewable permission — and since what a buyer is really purchasing is a built-out, permitted, operating restaurant at a known rent, the landlord holds a veto over the entire enterprise value. Combined with clause 2 below, this tenant has built an asset they may not be able to convert into money.

2. Ten-year term with no renewal options, plus an unlimited personal guaranty. The two halves of the same problem. The tenant carries all the downside (personally guaranteed for the full term, no good-guy limitation) and none of the upside (at year ten they have no right to stay, and every dollar of build-out reverts to the landlord). If the restaurant succeeds, the landlord holds all the leverage in the year-ten negotiation and knows exactly how much the tenant cannot afford to move.

3. The relocation clause. A right to move a restaurant with a six-figure build-out embedded in its floors and walls, on 90 days' notice. For a restaurant this is close to a termination right in the landlord's hands, and there is no stated obligation to fund the rebuild.

4. Uncapped NNN at \$9.00 per square foot, with no audit right and no exclusions. NNN is 22.5% of the all-in cost here (\$21,600 of \$96,000), and the tenant has no cap on controllable increases, no protection against capital expenditures being loaded into the operating pool, and no ability to verify the reconciliation. A single dollar of NNN growth is \$2,400 a year, uncontestable.

5. Rent commencement 45 days after delivery, with two months of base-only free rent. These compound: the abatement is worth only \$14,400 (2,400 × \$31 ÷ 6) rather than \$16,000, the NNN is payable throughout, and the clock starts before the permit is likely to have issued. Case Study 2 is what this looks like when construction runs long.

Honorable mentions, any of which could displace one of the above depending on the concept: the artificial breakpoint of \$900,000 against a natural breakpoint of \$1,240,000 (worth \$20,400 a year at \$1.3M of sales — see Exercise 6.3); the 4% compounding escalation; the seven-day continuous- operation covenant; the narrow "full-service Italian restaurant" use clause; and the complete absence of an exclusive.

Exercise 6.27

4% annual escalation, 2,400 sq ft at \$31.00:

Year-1 base rent is $2{,}400 \times \$31.00 = \$74{,}400$. Over ten years with 4% compounding, the multiplier is $(1.04^{10} - 1) \div 0.04 = 12.0061$.

$$\$74{,}400 \times 12.0061 = \mathbf{\$893{,}254}$$

Flat \$36.00 per square foot:

$$2{,}400 \times \$36.00 = \$86{,}400 \text{ per year} \times 10 = \mathbf{\$864{,}000}$$

The flat \$36 deal is cheaper by \$29,254 over the ten years — despite starting \$5 a foot higher, which is the point of the exercise and the reason escalation clauses get signed without argument.

And the comparison at the end of the term is more dramatic than the total. The escalating lease reaches $\$31.00 \times 1.04^{9} = \$44.12$ per square foot in year ten — \$105,894 against \$86,400. If renewal-option rent is set from the year-ten rate, that \$19,494 gap compounds forward through both option periods.

Which would I prefer? The flat \$36, decisively, for three reasons: it is cheaper in total, it is dramatically cheaper in the years when a restaurant is most likely to be tired and most vulnerable, and it is knowable — a rent you can state to the dollar for ten years makes every forecast in Chapters 31 through 33 a better instrument.

Exercise 6.29

The diagnosis: the operator has computed a real numerator against a hypothetical denominator, and then drawn a conclusion about risk from it. The 5.9% is arithmetically correct ($\$106{,}000 \div \$1{,}796{,}610 = 5.9\%$) and analytically meaningless, because \$1,796,610 is not money anyone has collected.

What occupancy actually was last year:

$$\text{Last year's revenue} = \$1{,}796{,}610 \div 1.14 = \$1{,}575{,}974$$ $$\$106{,}000 \div \$1{,}575{,}974 = \mathbf{6.7\%}$$

And if next year comes in flat rather than up 14%, occupancy stays at 6.7%. If it comes in 10% below last year (\$1,418,377), occupancy is 7.5%.

The one question to ask next: "What was your occupancy percentage last year, on money you actually banked — and what is it if next year is flat?"

That question does two things at once. It forces the honest number onto the table, and it exposes whether the 14% growth assumption has any support behind it. In a plan, an occupancy figure computed against a growth forecast should always be shown alongside the same figure computed against trailing actuals; if the operator will not show both, the second one is the interesting one.

Exercise 6.31

A model policy. Yours should be one page, signed at the pre-construction meeting by the owner, the general contractor, and the architect.

CHANGE ORDER CONTROL POLICY — [Project]

1. Written authorization required. No change to the scope, contract sum, or schedule shall be performed without a written change order signed by the Owner in advance. Work performed without a signed change order is performed at the Contractor's risk and will not be paid.

2. Pricing precedes performance. Every proposed change shall be submitted with a written scope description, a line-item price (labor, material, subcontractor, markup stated separately), and the schedule impact in calendar days. Where an emergency prevents advance pricing, the Contractor shall notify the Owner in writing within 24 hours and submit pricing within 3 business days.

3. Authority levels. Changes under \$1,000: either partner may approve. \$1,000–\$5,000: both partners. Over \$5,000: both partners plus written notice to the lender if required by the loan documents. No individual may split a change to stay under a threshold.

4. The log. The Owner maintains the change-order log. It is updated within 48 hours of any signed change order and shows, for each entry: number, date, description, category (concealed / authority / design / allowance / owner), amount, running total, and contingency remaining. The log is reviewed at the weekly job meeting with the Contractor and the Architect present, and the contingency position is stated aloud at every meeting.

5. Owner-change freeze. Owner-requested changes are frozen as of [DATE]. After that date, an owner change requires a written business justification and both partners' signatures.

6. Allowance reconciliation. Each allowance shall be reconciled by change order upon award of the underlying subcontract, with three quotes provided where practicable. Allowances are reconciled at actual cost plus the contract markup, with no additional fee.

7. Escalation trigger. If the running change-order total reaches 50% of the contingency, the Owner, Contractor, and Architect shall meet within 5 business days to review remaining scope and identify value-engineering opportunities. This meeting occurs regardless of how much work remains.

The seventh clause is the one most policies omit and the one Exercise 6.21 argues for: the trouble in that log was visible in week 4 and did not become a crisis until week 10.

Exercise 6.32

A model counter-proposal. Business language, one paragraph per term, each with a rationale a landlord could accept.

Escalation. We would like to replace the proposed 3% annual escalation with fixed steps of \$1.00 per square foot at the beginning of years 3, 5, 7, and 9. Over the initial term this produces a total very close to your proposal in the early years and gives us a rent we can state exactly for ten years, which is what our lender will be underwriting against. Predictability is worth more to us than the difference, and a tenant who can model their fixed costs precisely is a tenant who does not surprise you in year six.

Operating expenses. We accept the \$6.00 per square foot estimate and ask for three standard protections: a 5% annual cap on controllable expenses (taxes and insurance excluded, since neither of us controls those); exclusion of capital expenditures from the operating pool, or their amortization over useful life; and the right, on 30 days' notice and at our cost, to review the books supporting the annual reconciliation, with your reimbursement of the review cost only if an error exceeds 5%. None of these change what you collect when the building is run normally. They give us the ability to budget and to explain the number to a lender.

Rent commencement. We ask that rent commence on the earlier of our opening for business or 30 days after issuance of the certificate of occupancy. We are taking the entire construction risk and funding the entire build-out in advance of your improvement allowance; permit and plan-review timing is the one part of the schedule neither of us controls. This structure costs you nothing if the project runs on time, and it is the difference between a tenant who opens with working capital and one who does not.

Note what each paragraph does: states the ask, gives a reason the landlord can act on, and — crucially — names what the landlord keeps. A counter-proposal that only asks reads as a demand; one that explains the tenant's constraint and protects the landlord's economics reads as a deal.

Exercise 6.33

(a) The "maintenance permit" shortcut.

You decline, and you decline in writing, and you do it in a way that does not make an enemy of the contractor you are about to spend eight months with.

Substantively: demolition performed outside the permitted scope is unpermitted work. If it is discovered — and building departments do discover it, sometimes because a neighbor calls — the consequences run from a stop-work order (which costs you far more than three weeks) through fines to a soured relationship with the office that has to sign every subsequent inspection on your project. It can also complicate the certificate of occupancy, which is the gate on your opening date, and it sits badly with an insurer if anything goes wrong during the work.

There is a second-order problem that matters more. This is week zero of a relationship in which you will be asked to accept a dozen judgment calls you cannot personally verify. The precedent you set now is the precedent that governs all of them. What you say is something like: "I know it's common and I know you'd do it safely. I'm not going to start this project outside the permit — if it goes sideways it costs me the schedule I'm trying to protect. What else can we pull forward while we wait? Can we get the long-lead equipment released, or run the shop drawings?" Redirect the energy rather than just refusing it.

(b) The listing broker and the good-guy clause.

The advice is not malicious and it is not disinterested. The listing broker represents the landlord and is paid by the landlord, typically as a percentage of lease value, when the deal closes. Their economic interest is in a signed lease at a high rent, and a protracted negotiation over a guaranty limitation serves neither. That does not make them dishonest; it makes them a party with a position, which you should have established at the first meeting by asking "who do you represent, and how are you paid?"

On the substance: the good-guy clause is the highest-value term in the lease that costs the landlord no current income, and the "souring the relationship" framing is the oldest argument in commercial leasing. Landlords negotiate leases constantly and do not take markups personally. What you say to the broker is roughly: "I understand, and I'm not trying to grind anybody. This is the one term where I'm asking for something, and I'm prepared to trade for it — a bigger deposit, a longer notice period, six months instead of three. Can you take it to them that way?"

If the answer comes back no — which it does in Bellwether's case — that is a legitimate outcome. What is not legitimate is failing to ask because somebody who is paid by the other side told you it would be awkward.

Exercise 6.34

Site & Lease summary — Bellwether (constructed teaching example)

Required-revenue inversion

Check: each row is $\$95{,}200 \div \text{target}$. ✓

The honest caption:

At the plan's \$1,550,000 forecast, occupancy is 6.1% — a genuinely strong figure, and one of the best features of this deal. But the rent is contractual and the forecast is not: this lease commits the business to producing at least \$1,190,000 of revenue every year for ten years simply to hold occupancy inside 8%, and \$952,000 is the level below which one dollar in every ten through the register goes to the landlord.

Exercise 6.35

Risk paragraph 1 — the personal guarantee.

Both partners have executed a full personal guaranty of the lease, jointly and severally, with no good-guy limitation and no burn-down, for the entire ten-year initial term. The guaranteed obligation is approximately **\$1,032,600**: \$840,000 of contractual base rent under the agreed step schedule, plus roughly \$192,600 of triple-net charges assuming operating costs grow about 3% a year. That estimate is sensitive to the NNN assumption, which is an estimate rather than a term. A landlord generally has some duty to mitigate by re-letting the premises, and the strength of that duty varies by jurisdiction, so the enforceable figure in a default would likely be lower — but the plan treats the stated number as the exposure, because that is the only conservative way to hold it. The guarantee would be called if the operating entity failed to pay rent and the partners did not cure. It is the largest single obligation in this plan and it appears on no financial statement in it. Both partners have been advised to obtain independent legal counsel and to execute a written contribution agreement between themselves before opening.

Risk paragraph 2 — the construction contingency.

The construction line of \$310,000 comprises a \$264,000 stipulated-sum contract, \$37,000 of soft costs, and \$9,000 of contingency. That contingency is 3.4% of the contract sum, against a commonly used rule of thumb of 10–15%, which would have indicated \$26,400 to \$39,600. The plan acknowledges this as its weakest construction assumption. Two allowances totaling \$26,000 — for the exhaust hood and make-up air, and for the grease interceptor — were set on the assumption that the existing equipment could be modified and reused, and due diligence established that it cannot; the revised scope is being priced (Chapter 7) and the \$310,000 line will be re-cut against it, which necessarily means reducing scope elsewhere. The partners' response is threefold: complete the drawings to 100% and re-bid the mechanical and plumbing scope as specified work rather than as allowances before signing the construction contract; operate a written change-order control policy with a weekly contingency review and a 50%-consumed escalation trigger; and hold the \$45,000 working-capital reserve as untouchable, treating any construction shortfall as a scope decision rather than a reserve draw.

Chapter 7

Exercise 7.1

Front of house (FOH) is every part of the restaurant a guest may occupy or see from a seat: entry and vestibule, host stand and waiting area, dining room, bar, service stations, and guest restrooms. Back of house (BOH) is everything else: the cooking line, prep, dish, walk-in and dry storage, receiving, office, and staff areas.

The rule-of-thumb split for full service is 60–70% FOH / 30–40% BOH (Tier 2 — trade convention).

Bellwether runs 1,700 / 900 / 200 on 2,800 square feet, which is 60.7% / 32.1% / 7.1% — and if storage and office are grouped with BOH, 60.7% front against 39.3% back. That is the kitchen-heavy end of the band.

What it bought: the ability to produce 50 entrées in the peak hour of a 135-cover Saturday from a scratch kitchen with a wood-fired hearth, house butchery, and two prep tables. The seats given up to get it are the ones that would not have filled on a Tuesday, and that the kitchen could not have fed on a Saturday.

Exercise 7.2

A Type I hood is a listed, grease-rated exhaust hood required over equipment producing grease-laden vapors. It carries grease-extraction baffles, grease collection, fire suppression, and grease-rated duct to above the roof.

A Type II hood handles heat and condensate only and has none of that apparatus.

(Any three of the Type I list and two of the Type II list are acceptable. Local classification varies and some ovens fall either way depending on use.)

What happens if you install the wrong one: a Type II over grease-producing equipment fails inspection and will not be approved, and if it were somehow built it is a fire risk — there is no grease extraction, no listed suppression, and no grease-rated duct. It is not a value-engineering option; it is a life-safety failure. The reverse error (Type I where Type II would do) is merely expensive.

Exercise 7.5

An equipment schedule is the numbered, itemized list of every fixed and major movable piece of equipment, with tag, description, quantity, dimensions and clearances, utility requirements, and who supplies and installs.

Three documents, with a consequence each:

1. A purchase order. It is what you buy from and what a lender or lessor finances against. Consequence: an item omitted from the schedule is an item nobody ordered — discovered when the line is set and the ice machine has no home.
2. A permit document. The health authority's plan review and the mechanical, electrical, and plumbing engineers all design from it: the hood from the appliance list, the gas meter from the BTU column, the panel from the amps column. Consequence: change the schedule after the drawings are stamped and you have changed the mechanical design — and possibly the permit.
3. A capacity statement. Station ceilings are computed from it. Consequence: the schedule is the actual answer to "how many plates can you send," so an equipment cut is a revenue cut.

Exercise 7.9

$$44 \times 26 = \mathbf{1{,}144 \text{ sq ft}}$$

**For a \$62 average check, choose 19** (or something between 18 and 20). A \$62 check is fine-dining or high-end upscale-casual territory, and the occasion that produces it — celebration, business, special-occasion dining — is precisely the occasion a dense room destroys. The lost 28 seats against the 13-sq-ft option look expensive until you price the alternative: if crowding drops the check from \$62 to \$52 across the year, ten dollars a cover on even 25,000 covers is \$250,000, which no additional seating recovers.

The transferable rule from §7.1: decide density against the check average, not against the square footage.

Exercise 7.11

$$\text{FOH} = 2{,}100 \div 3{,}400 = \mathbf{61.8\%}$$ $$\text{BOH} = 1{,}000 \div 3{,}400 = \mathbf{29.4\%}$$ $$\text{Storage} = 300 \div 3{,}400 = \mathbf{8.8\%}$$

Check: 61.8 + 29.4 + 8.8 = 100.0%. ✓

$$\text{Seat-to-square-foot ratio} = 3{,}400 \div 84 = \mathbf{40.5 \text{ total sq ft per seat}}$$

Against the rules of thumb: the FOH share (61.8%) is at the low end of the 60–70% band and the BOH share (29.4%) is just below the 30–40% band — but adding the 300 sq ft of storage to the back puts the non-guest side at 38.2%, comfortably inside it. The 40.5 total sq ft per seat is close to Bellwether's 41.2.

What the numbers suggest: a full-service restaurant with meaningful on-premise production and a generous room — not a high-turn casual concept, which would run denser seating and a smaller kitchen. If the operator told you this was a fast-casual build, the numbers would be arguing with them.

Exercise 7.13

Peak-hour entrées at the 37% planning rule:

At 50% routing, hearth load is half of each figure: 13 · 18 · 23 · 25 · 30.

Against a 28-item sustained ceiling, the hearth is fine through 135 covers (25, or 89%) and breaks at 160 covers, where it is asked for 30 against 28 — 107% of capacity.

Solving for the exact cover count at which hearth load equals 28:

$$\text{covers} \times 0.37 \times 0.50 = 28 \;\Rightarrow\; \text{covers} = \frac{28}{0.185} = \mathbf{151}$$

So the design ceiling of this kitchen, on this menu routing, is about 151 covers a night. Bellwether's design night of 135 sits 11% under it, which is thin but real headroom — and it is why the menu constraint in §7.3 is a constraint rather than a preference.

Exercise 7.15

$$90 \text{ seats} \times 1.6 \text{ turns} = \mathbf{144 \text{ covers a night}}$$ $$144 \times \$38 = \mathbf{\$5{,}472 \text{ per service}}$$ $$\$5{,}472 \times 6 = \mathbf{\$32{,}832 \text{ a week}}$$ $$\$32{,}832 \times 52 = \mathbf{\$1{,}707{,}264 \text{ a year}}$$

Peak-hour entrées: $144 \times 0.37 = 53.3 \rightarrow \mathbf{53}$.

The question to ask the chef: "Show me the station that sends the fifty-third plate." Fifty-three entrées in an hour on a six-night-a-week operation is a real kitchen — it needs the station count, the equipment, and the prep to support it, and 1.6 turns at a \$38 check implies a fast, disciplined service. The forecast is not implausible; it is unproven until somebody computes the binding station's ceiling the way Figure 7.5 does.

A good second question: "What is Tuesday?" A 1.6 average across six nights hides a Tuesday that may be 0.8 and a Saturday that may be 2.2 — and it is Saturday that has to be built for.

Exercise 7.17

The added-seat upside. Compressing 890 sq ft from 15.9 to 12.0 sq ft per seat:

$$890 \div 12 = 74 \text{ seats}, \text{ an increase of } 74 - 56 = 18$$ $$18 \times 1.4 \text{ turns} = 25.2 \rightarrow 25 \text{ more covers a night}$$ $$25 \times \$46 = \$1{,}150 \text{ per service} \;\;(\text{text uses } \$1{,}159 \text{ at } 25.2)$$ $$25.2 \times \$46 \times 5 \times 52 = \mathbf{\$301{,}392 \text{ a year}}$$

The check-average downside.

$$\$4 \times 36{,}140 \text{ covers} = \mathbf{\$144{,}560 \text{ a year}}$$

Which is bigger? On its face the upside — \$301,392 against \$144,560. But the upside is hypothetical and constrained three ways and the downside is immediate and unconstrained:

1. The kitchen cannot produce it. At 74 seats and 1.4 turns the design night rises to about 104 covers, and in patio season 116 — which Exercise 7.13 shows is inside the 151-cover ceiling, but only if nothing else changes; the real limit arrives on Saturday, where 74 × 1.81 = 134 covers plus 12 patio = 146, essentially at the ceiling with no slack at all.
2. Chapter 2 established a supplied market. The seats do not create demand; at 1.4 turns on 74 seats you need 104 covers to look as busy as 95 looks on 68.
3. The \$144,560 is certain the moment the room gets tight, and it applies to every cover, not just the marginal ones.

What it tells you about density decisions: the upside of density is a revenue number that depends on demand and production you do not control; the downside is a price effect you cannot escape. Asym- metric risk of that shape should be resolved conservatively — which is what "density is a pricing decision" means in practice.

Exercise 7.18

(Model answer using Figure 7.3; a student's own room will differ.)

$$\text{Parties} = 95 \text{ covers} \div 2.4 \text{ per party} = 39.6 \rightarrow \mathbf{40}$$ $$\text{Round trips} = 40 \times 12 = \mathbf{480 \text{ a service}}$$ $$\text{Average round trip} = 30 \text{ paces} \times 2.5 \text{ ft} = \mathbf{75 \text{ ft}}$$ $$\text{Total} = 480 \times 75 = \mathbf{36{,}000 \text{ feet}}$$ $$36{,}000 \div 5{,}280 = \mathbf{6.8 \text{ miles per dinner service}}$$

Across 260 dinner services that is about 1,772 miles a year walked by the floor team — before brunch, and before anyone runs food.

The instruction the number produces: an audit like this takes an hour with a notebook and a floor plan, and it is the only way a layout argument ever gets settled with arithmetic instead of taste.

Exercise 7.20

A dish return path that crosses the hot line, four costs:

The design answer is one sentence: soiled ware travels on its own path, on the opposite side, and never crosses in front of the range — which costs nothing at the drawing stage and is not retrofittable.

Exercise 7.24

$$\text{Real scope} = \$52{,}000 + \$6{,}500 + \$16{,}500 = \mathbf{\$75{,}000}$$ $$\text{Allowances carried} = \$18{,}000 + \$8{,}000 = \mathbf{\$26{,}000}$$ $$\text{Gap} = \$75{,}000 - \$26{,}000 = \mathbf{\$49{,}000}$$ $$\text{Plus additional soft cost} = \$49{,}000 + \$4{,}500 = \mathbf{\$53{,}500}$$

\$53,500 must be found inside the frozen \$310,000 construction line, and — this is the part students miss — not from the \$9,000 contingency, because the hood is now identified, drawn, and priced. Known scope is funded from scope; a contingency funds the unidentified.

Note also where the increase lands: across three trade lines (mechanical +\$34,000, plumbing +\$8,500, fire protection +\$6,500), and the fire-protection trade carried no allowance at all.

Exercise 7.26

The chef-owner's sentence — "we have \$35,000: the \$26,000 of allowances plus the \$9,000 contingency" — is wrong because it misuses the definition of a construction contingency, which Chapter 6 established as money reserved for costs that are certain to occur but cannot yet be identified. The hood, the make-up air unit, the fire-protection increment, and the grease interceptor are no longer unidentified: they have been re-drawn, re-scoped, and re-bid, and they have a price. A cost with a price on a drawing is scope, and scope is funded by cutting other scope. Spending the contingency here does not solve the problem; it converts a thin cushion into no cushion at week zero, which means the next surprise — and there will be one — has nowhere to land except the equipment budget or the working-capital reserve, both of which Chapter 6 warned about explicitly.

The four unknowns the \$9,000 exists for, all named in the chapter:

1. The gas service and meter, sized for a café against roughly 1.3 million BTU/hr of connected load — and a utility upgrade is a queue as well as a price.
2. The waste line to the main, 4 inches, interior condition never scoped.
3. Two rooftop HVAC units, about eleven years old, being reused, where a triple-net lease may make replacement the tenant's problem.
4. Abandoned floor drains under the slab, which in Chapter 6's own change-order log already cost \$3,850 to cap, backfill, and re-pour.

Against those four, \$9,000 is already thin at 3.5% of the revised contract sum where the rule of thumb is 10–15% — \$25,950 to \$38,925. The honest plan says so rather than spending it.

Exercise 7.28

Annual labor cost of the deferral:

$$11 \text{ minutes} \times 7 \text{ bar shifts} \times 52 \text{ weeks} = 4{,}004 \text{ minutes}$$ $$4{,}004 \div 60 = 66.7 \rightarrow \mathbf{67 \text{ hours a year}}$$ $$67 \times \$19 = \mathbf{\$1{,}273 \text{ a year}}$$

Payback on the deferred purchase:

$$\$7{,}400 \div \$1{,}273 = \mathbf{5.8 \text{ years}}$$

On the labor arithmetic alone the deferral is defensible — a five-and-a-half to six-year payback is a weak capital case, and it is the right item to be the last one on the list.

Two costs the calculation does not capture:

1. Glass breakage in transit. Racks of glassware carried across a restaurant break at a rate that an under-counter machine at the bar does not produce. It is small per incident, recurring, and entirely absent from the payback.
2. The bartender off the bar during service. Eleven minutes a shift is not eleven minutes of nothing; it is eleven minutes when nobody is making drinks, reading the bar, or greeting the guest who just sat down. Chapter 23 would price that higher than \$1,273, and Chapter 24 would note that the missing drink is also the missing check.

The teaching point: the deferral is defensible because it was computed, not because it was the cheapest thing on the schedule — and the two uncaptured costs belong in the plan next to the number.

Exercise 7.30

What the queue costs when rent commences at the certificate of occupancy.

The construction job is 19 weeks. The utility queue is 14 weeks and runs in parallel with construction if it is ordered on day one — so if the operator applies at permit issuance, the service lands in week 14 of a 19-week job and costs nothing in rent. The exposure appears only if the application is late: every week the utility work extends past week 19 delays the certificate of occupancy, and each of those weeks delays opening.

Suppose the application is made at week 8 instead of week 1. The service now lands in week 22, three weeks past substantial completion. With rent tied to the certificate of occupancy, those three weeks cost \$0 in rent — but they cost three weeks of a hired management team against no revenue (Chapter 9's \$35,000 pre-opening budget) and three weeks of the season.

What the same three weeks cost under a delivery-of-possession trigger. Rent would already be running. Three weeks at \$7,933 a month is:

$$\$7{,}933 \div 4.33 \times 3 = \mathbf{\$5{,}496}$$

And if the whole 14-week queue had been serialized after construction rather than run in parallel, the delay is 14 weeks:

$$\$7{,}933 \div 4.33 \times 14 = \mathbf{\$25{,}646 \text{ of rent on a closed building}}$$

The lesson, which is Chapter 6's and Chapter 7's at once: the utility queue is a schedule risk, not a price risk, and the instrument that converts a schedule risk into a zero is the rent-commencement clause. Order long-lead utility work the day the permit issues, and never later.

Exercise 7.31

(Model memo — students' wording will differ; grade on honesty, completeness, and whether the arithmetic foots.)

To: the front-of-house partner Re: the mechanical re-scope, and what it costs us

The mechanical contractor's re-bid came back. The real number for the hood, the duct and roof work, the fan, the make-up air unit, the controls, the added fire suppression, and the in-ground grease interceptor is \$75,000**. We carried \$26,000 in allowances. Re-drawing and re-permitting adds \$4,500. So we are **\$53,500 short inside a \$310,000 line that does not grow.

I want to be clear about what I did not do. I did not spend the \$9,000 contingency. It is for the things we still cannot name — the gas service, the waste line, the eleven-year-old rooftop units, whatever is under the slab — and we would be spending it in week zero.

Here is what came out, in the order I cut it. Nothing that a code requires, nothing that produces a plate, and nothing that turns a one-time cost into a permanent one.

The dining floor is now polished and sealed concrete with rugs under the banquettes instead of reclaimed plank (\$11,900). The bar top and chef's counter are engineered quartz instead of stone (\$3,400). The two service stations and the back-bar shelving are modular stainless instead of custom millwork (\$9,800). The banquettes are straight runs with no returns and one grade down on upholstery (\$4,400). The restrooms get a durable, plain finish package — no dimension, clearance, fixture height, or grab bar changed (\$5,500). The dining-room deck stays open and painted instead of a hung ceiling, and the acoustic treatment stays in full (\$7,400). The decorative lighting package is reduced, with the service upgrade and every kitchen circuit untouched (\$4,700). We do the non-structural demo ourselves over two weekends (\$1,900). And the office and dry-storage built-ins become open shelving and a stock desk (\$4,500).

That is \$53,500. The line lands on \$310,000 and the cost per square foot is exactly what it was.

Two things I want you to push back on. The patio was on my first list and I took it off — those sixteen seats carry \$64,800, which is 46.5% of the revenue bridge, and cutting \$8,000 to endanger that is not a saving. And the room is going to look different from what we showed people: concrete floor, open ceiling, quartz. I think it still reads as a \$46 room. If you don't, tell me now, while it is still a drawing.

Exercise 7.34

(Model answer — two paragraphs for the plan's budget page.)

Paragraph one — what it cost and how the line held. The existing Type I hood, make-up air unit, and grease interceptor were sized for the previous tenant's café line and cannot serve a wood-fired hearth, which is a solid-fuel cooking appliance and a separate regulatory category. Re-drawn and re-bid as specified scope, the package totals \$75,000**: \$52,000 of mechanical (hood, grease-rated duct and shaft, roof curb and framing, upsized exhaust fan, tempered make-up air unit, controls and balancing), \$6,500 of fire protection (enlarged wet-chemical suppression plus solid-fuel provisions, a trade that carried no allowance at all), and \$16,500 of plumbing (in-ground grease interceptor, excavation, and sanitary tie-in). The construction contract carried \$26,000 of allowances against it, and re-drawing and re-permitting added \$4,500 of soft cost. The **\$310,000 construction line was not increased. \$53,500 was removed from other scope through nine value-engineering items, none of which touched an accessibility dimension, a code requirement, a capacity item, or anything that converts capital into a permanent operating cost. The revised line is \$259,500 contract sum + \$41,500 soft costs + \$9,000 contingency = \$310,000, unchanged at \$110.71 per rentable square foot. The equipment line remains **\$185,000**: \$7,400 of modular stainless moved onto it from the construction contract, offset by deferring the bar's under-counter glasswasher to year two at a computed labor cost of \$1,273 a year.

Paragraph two — what was given up, and what is still open. Holding the line cost the specified reclaimed-plank dining floor, the stone bar top, custom millwork at the service stations and back bar, a hung dining-room ceiling, the restroom finish package, part of the decorative lighting package, the office and storage built-ins, two weekends of the owners' own labor, and one piece of bar equipment. The room that gets built is a coherent design and it is not the room that was drawn, and the plan's brand section should confirm it still supports a \$46 check. The **contingency remains \$9,000 — 3.5% of the revised contract sum against a 10–15% rule of thumb, a shortfall of \$16,950 to \$29,925** — and it now stands in front of four named unknowns: a gas service and meter sized for a café against roughly 1.3 million BTU/hr of connected load, a waste line whose interior has not been scoped, two rooftop HVAC units approximately eleven years old being reused, and abandoned floor drains under the slab. Separately, the re-scope lengthened the schedule: the hood, the exhaust fan, and the make-up air unit are long-lead items, and a hood redesigned after the original drawings does not ship on the original date. Both the contingency and the schedule are carried in the risk register.

Chapter 8 — Worked Solutions

Solutions to the daggered (†) and odd-numbered exercises. Every figure is illustrative and constructed; nothing here is legal advice, and the specifics vary by jurisdiction.

Exercise 8.1

A certificate of occupancy is issued by the building department and certifies that the space complies with applicable codes and may lawfully be occupied for a stated use at a stated occupant load. A food service establishment permit is issued by the health authority (usually county or city) and is the permission to operate a food business at that address.

They are different permissions from different agencies on different clocks. A restaurant can therefore hold a certificate of occupancy — the building is signed off, the trades are done, the fire marshal is satisfied — and still be legally unable to cook, because the health authority has not yet conducted its pre-opening inspection or issued the establishment permit.

Exercise 8.2 †

An LLC is an entity. It is formed by filing articles with a state, and it creates the liability separation: the company holds the contracts, incurs the debts, and is the defendant.

An S-corp election is a federal tax status. It is available to an eligible corporation or LLC, and it determines how income is reported and how owner compensation is split between wages and distributions.

So: the LLC describes liability; the S-corp election describes tax. The confusion matters because operators routinely say "I'm an S-corp" when asked about protection, which answers a question nobody asked and leaves the actual liability question — what entity, respected how, guaranteeing what — unexamined. It also leads people to believe a tax election changed their exposure. It did not.

Exercise 8.3

Exercise 8.5 †

Dram shop liability is the legal exposure of a seller of alcohol for harm caused by a person it served — most commonly where that person was visibly intoxicated or under the legal drinking age.

The two exposures:

• Civil. A private suit for damages by the injured third party (and in some states by others). This is what liquor liability insurance responds to, up to its limits — and note the general liability policy will not, because of the liquor liability exclusion.
• Administrative. Action by the licensing authority against the license itself: fine, attached conditions, suspension, or revocation. No insurance policy buys this back. At Bellwether's figures a thirty-day suspension costs roughly \$27,900 of beverage contribution plus the food covers that come with the bar, plus a record that follows the licensee.

In some circumstances there is also criminal exposure, and in a number of states the individual server can be named in the civil suit.

Exercise 8.7

Business interruption coverage requires a covered physical loss to the property before the period of restoration begins. It reimburses lost net income and continuing expenses while you cannot trade because the property was damaged.

That requirement is why most pandemic-era claims failed. A government order closing indoor dining is not physical damage to the building, so in the great majority of cases across many jurisdictions there was no triggering loss and the coverage never engaged. Some policies have since addressed communicable disease explicitly — usually by excluding it. Read your own form.

Exercise 8.8 †

• New construction must be built to the applicable accessibility standards.
• Alterations must make the altered area accessible, plus an obligation regarding the path of travel to that area (entrance, route, restrooms), subject to a proportionality limit on cost.
• Existing facilities carry a continuing duty to remove architectural barriers where doing so is readily achievable — accomplishable without much difficulty or expense, judged against available resources.

A \$310,000 second-generation conversion is an alteration, which is why Chapter 6 warned that substantial alteration triggers current code. The obligation that never ends is readily achievable barrier removal: a barrier that was not readily achievable to remove in year one may be readily achievable in year four, and the duty re-attaches.

Exercise 8.9

An auto-renewal (evergreen) clause extends the contract term automatically unless one party gives notice of non-renewal. The notice window is the defined period before expiration during which that notice must be given — commonly something like "not more than 90 and not fewer than 60 days before the end of the term."

Two calendar entries, not one:

1. A decision reminder ahead of the window opening (say 120 days out), which gives you time to evaluate, get competing quotes, and negotiate.
2. The last day of the window itself, as a hard deadline.

One entry alone gives you a deadline with no decision time, which in practice means you renew.

Exercise 8.10 †

As a share of the \$48,000 pre-opening budget: \$17,752 ÷ \$48,000 = 37.0%.

Remaining for training payroll and opening inventory: \$48,000 − \$17,752 = \$30,248. For a 90-seat restaurant with a bar, opening food and beverage inventory alone will take a large share of that, which means pre-opening training hours are being funded by whatever is left — the same collision Bellwether faces at \$14,550 inside \$35,000, and the reason Chapter 9's countdown has to be built with the licensing number already in it rather than added afterward.

Exercise 8.11

• As a percentage of a \$46,500 G&A line: \$8,545 ÷ \$46,500 = 18.4%.
• As a percentage of \$1,600,000 of sales: \$8,545 ÷ \$1,600,000 = 0.53%.

The interpretation: half a point of sales, and nearly a fifth of the G&A line, is spent every year on permission. It is not a large number, but it is a recurring one with hard dates attached, and it is the reason the compliance calendar exists.

Exercise 8.13 †

Rent commences 30 days after the certificate of occupancy: March 3 + 30 days = April 2.

The license issues May 28. Days from April 2 to May 28:

• April 2 → April 30 = 28 days
• May 1 → May 28 = 28 days
• Total = 56 days

At \$7,933 a month, the daily rate is \$7,933 ÷ 30 = \$264.43.

$$56 \times \$264.43 = \mathbf{\$14{,}808}$$

That is rent paid on a building that is lawfully occupied and legally unable to serve alcohol — and it does not include the payroll, utilities, or debt service running alongside it. Note also what does not help: Chapter 6's rent-commencement clause is excellent protection against construction delay and no protection at all against licensing delay, because the C of O is the trigger and the C of O does not care what the alcohol authority is doing.

Exercise 8.15 †

• Beverage revenue: \$1,850,000 × 0.26 = **\$481,000**
• Beverage cost: \$481,000 × 0.21 = **\$101,010**
• Annual beverage contribution: \$481,000 − \$101,010 = \$379,990
• Weekly beverage contribution: \$379,990 ÷ 52 = **\$7,307.50**
• Ten weeks open without a license: 10 × \$7,307.50 = **\$73,075**

Ten weeks is not a catastrophe scenario; it is an ordinary licensing slip. It costs this restaurant \$73,075 of contribution while every fixed cost runs at full rate.

Exercise 8.17 †

• Total acquisition cost: \$88,000 + \$11,500 = \$99,500
• Added monthly payment: \$99,500 × \$0.0134935 = \$1,342.60
• Added annual debt service: \$1,342.60 × 12 = **\$16,111**

Two observations worth stating in a plan. First, this is \$16,111 a year of new fixed obligation incurred before the restaurant has served a guest, and it does not go away in a slow February. Second, what has been purchased is an intangible whose value rests on a policy that a legislature can change — so the borrower is, in a real sense, financing a policy position. That is not a reason never to do it. It is a reason to say so in the risk section.

Exercise 8.19 †

A model email. Note that every question is answerable in writing, is specific to one address, and costs the authority almost nothing to answer.

Subject: License eligibility inquiry — 1412 Mill Street, Rivermill District

I am evaluating the property at 1412 Mill Street for a 68-seat full-service restaurant with an on-premise bar, and I would be grateful for written guidance on the following before I proceed with a lease. I am happy to submit this on a form or to pay any inquiry fee.

1. Is an on-premise license of the type I would need available at this address — or is this jurisdiction quota-limited, dry, or currently subject to a moratorium?
2. What distance restrictions apply relative to schools, places of worship, parks, or other licensed premises, and by what method is the distance measured?
3. Does the municipality require a separate conditional-use or special-use permit for this use, and does that process include a public hearing?
4. What is the current realistic processing time, filing to issuance, for an application at this address?
5. What posting, publication, or neighbor-notice requirements apply, and how long is the protest period?
6. What conditions are typically attached to a license of this type — hours, outdoor service, entertainment, noise, or parking?
7. What are the ownership-disclosure and background-check requirements, at what ownership threshold, and what is the standard for source-of-funds documentation?
8. Is a temporary or conditional permit available while an application is pending, and is its issuance discretionary?

I would also appreciate confirmation of which items must be complete at filing and which may be supplemented afterward. Thank you for your time.

Exercise 8.21

(a) The margin of safety. The requirement is food ≥ 55% of gross receipts. The plan is 72% food. The margin is 17 percentage points, which is comfortable.

(b) At 41% beverage, food is 59% — still above 55%, but the cushion has fallen from 17 points to 4. Two more good quarters at the bar and the restaurant is out of compliance with a condition on its own license while congratulating itself on its pour cost. The problem is not the bar's success; it is that nobody was measuring the ratio the license actually cares about. What you needed in year one was a reporting standard: food and beverage receipts captured separately, reported by period, reviewed on the same cadence as prime cost, and archived — because the licensee, not the regulator, carries the burden of proving the ratio.

(c) Chapter 26 — the point-of-sale is the system of record, and this is one of the specific things it must be configured to produce on demand. Chapter 31's weekly flash report is the natural place to surface it.

Exercise 8.23 †

21 days = 3 weeks. At \$6,510 of weekly beverage contribution:

$$3 \times \$6{,}510 = \mathbf{\$19{,}530}$$

Against the \$4,900 annual liquor liability premium: \$19,530 ÷ \$4,900 = approximately 4.0×. One three-week suspension costs four years of the premium for the coverage that would not have applied to it anyway.

Two costs the arithmetic does not capture:

1. The food covers that come for the bar. A neighborhood restaurant that cannot serve a drink loses dinner parties, not just drink sales, and the loss is invisible in a beverage line.
2. The record. A disciplinary history follows the licensee, affects the treatment of any future violation, can affect renewal and transfer, and is frequently public. It also affects the liquor liability premium at the next renewal.

Exercise 8.25 †

Every intervention below is a system, not a person being careful.

• At the door, on a high-volume night. Card at the door so the bartender is not the checkpoint. The wave-through happens because one person is checking IDs and pouring at the same time while a line builds. Removing that conflict removes the failure.
• At the bar. Staff a barback so the bartender's hands and eyes are free. Make the point-of-sale drink count visible at the well.
• At the handoff. A required verbal handoff when a guest moves bar-to-table or when a shift changes, so the incoming server inherits the drink count rather than starting from zero.
• On the floor. A manager working the rail during the crush rather than the office, whose actual job in that hour is looking at people.
• In the policy. A written refusal protocol, and the rule that a manager never overrides a server's refusal in front of a guest — which is what makes a server willing to call it at drink four.
• In the calendar. Certification for every server, tracked, with recertification dates.
• In the pricing. A promotion strategy that does not make its money on volume drinking.

If you can install only one: the visible drink count at the bar. It is nearly free, it converts a fragmented memory problem into a number, and in the composite scenario in Case Study 2 it is the single control that would have surfaced "nine" to a human being before 11:20. Everything else depends on somebody noticing; this one does the noticing for them.

Exercise 8.27 †

• Other-operating subtotal against a \$245,000 line: \$32,200 ÷ \$245,000 = 13.1%
• Other-operating subtotal against sales: \$32,200 ÷ \$1,750,000 = 1.84%
• Grand total against sales: \$46,942 ÷ \$1,750,000 = 2.68%

The point of splitting the table this way is the last row. Nearly a third of the total cost of risk is sitting inside the labor line, which means it is inside prime cost — and an operator reading only the "insurance" line in other operating will conclude they spend \$32,200 on insurance when they spend \$46,942.

Exercise 8.29 †

• Required limit: \$850,000 × 0.80 = **\$680,000**
• Carrier payment: (\$600,000 ÷ \$680,000) × \$140,000 = 0.882353 × \$140,000 = \$123,529
• After the \$10,000 deductible: **\$113,529**
• Shortfall against the \$140,000 loss: **\$26,471**

The ordinary, blameless event that created the gap: the build-out finished and the equipment was delivered, and nobody re-measured the limit. The policy was bound against an estimate made before construction, and the value grew past it through change orders, a hood upgrade, and smallwares. Nothing was neglected in any dramatic sense — a number simply went stale, and a coinsurance clause converted a stale number into a 19% reduction on every covered loss.

The fix is one calendar entry: re-file the property values against final construction and equipment invoices before binding, and again at each renewal.

Exercise 8.31 †

Four problems, in descending order of seriousness:

1. The policy period expired four months ago. This certificate evidences nothing. Whatever coverage it described may or may not exist today.
2. Your restaurant is a certificate holder, not an additional insured. A certificate holder merely receives the document. Additional insured status is what makes the contractor's policy respond to a claim arising from their work at your premises. Without it, you have a piece of paper and no protection.
3. Workers' compensation is shown as "coverage in place" with no carrier, no policy number, and no limits. That is not evidence of anything. If this contractor's employee is injured on your roof and there is no coverage, the question of who is responsible becomes expensive quickly.
4. No umbrella. A hood-cleaning contractor works with degreasers and open flame risk directly above your kitchen. A serious fire is not a \$1M event, and there is nothing above the primary limit.

A fifth, arguably: no carrier names or policy numbers appear anywhere, which makes every line on the document unverifiable.

What you do: do not let them on the roof. Ask for a current certificate with your entity named as additional insured, with carrier names, policy numbers, and workers' compensation limits shown, and put the expiration date on the compliance calendar.

Exercise 8.33

(a) Opening dry and letting guests bring wine. Do not do it, and say so immediately and without negotiation. Whether a restaurant may permit consumption of alcohol it did not sell is itself a licensed question in most jurisdictions — "we're not selling it" is a folk theory, not a defense — and doing it while an application is pending puts the application itself at risk in front of the authority you are asking to trust you. The correct move is to call the licensing authority and your licensing attorney and ask what, if anything, is permitted at this address before a license issues, and to get the answer in writing. What you say to your partner: "I understand the pressure, and the downside here isn't a fine — it's the application. We ask, in writing, and we do exactly what they say."

(b) Understating alcohol as 15% of receipts. Refuse. The insurance application is a document you sign, and a material misstatement can give the carrier grounds to deny a claim or rescind the policy — which means the \$1,400 saving buys you a policy that fails on the single day you need it, in the exact category the misstatement concerned. It is also worth noticing the incentive: the broker is paid on placement. Get the accurate quote, and if the premium is uncomfortable, take the discomfort as information about your risk rather than as a problem to be drafted around.

(c) Fixing two barriers quietly and not responding. Two errors. Not responding to a demand letter is never the right call; send it to counsel and to your broker the same day, because some policies provide defense for these claims and notice provisions are strict. But also: fix the barriers regardless of the claim's merits, and document the fix with dated invoices and photographs. The barrier is what produces the next letter from the next plaintiff, and a documented remediation history is the difference between a defensible position and a story. Disputing the third item is legitimate — do it through counsel.

Exercise 8.34 †

A model memo. Note what it contains and, more importantly, what it does not: no speculation about cause, no admission, no conclusion about fault, no characterization of the guest.

INCIDENT REPORT — to file Date/time of incident: Friday, 8:52 p.m. Report written: same night, 11:40 p.m. Location: dining room, floor adjacent to the service station between tables 14 and 16. Reported by: manager on duty.

What happened. At approximately 8:52 p.m. a guest seated at table 16 fell while walking toward the restrooms. Two servers and the manager on duty responded within seconds. The guest reported pain in the left wrist and declined to stand unaided. Emergency services were called at 8:55 p.m. and arrived at 9:07 p.m. The guest was transported at 9:19 p.m., accompanied by a member of their party.

What was observed at the scene. The floor in that area was wet over an area of roughly two feet. A wet-floor sign was not present at the time of the fall; one was placed at 8:54 p.m. Lighting and flooring in the area were in their normal condition. Photographs were taken at 8:57 p.m. before the area was dried and are attached.

What was done. The area was cordoned, dried, and re-inspected at 9:10 p.m. The guest's party's check was closed with no charge for the table; no statement was made to the party about responsibility. Contact details for the guest and for one member of the party were recorded with their permission. Staff on the floor were asked to write their own accounts before leaving; three did, and are attached.

Who was notified. Owner-operators, by phone, 9:30 p.m. Insurance broker, by email, 11:35 p.m.

Attachments. Four photographs; three staff statements; the shift's floor plan and station assignments; the cleaning log for the evening.

Who receives it and why. The owners, because they need to know the same night and will be asked about it. The insurance broker or carrier, because notice provisions are strict and late notice can prejudice coverage. And the file — kept per counsel's retention advice — because a year from now the only reliable account of that evening will be the one written that evening.

Exercise 8.35

The table below is the Bellwether version; a reader working their own concept substitutes their own quotes from their own authorities.

The honest caption. All figures are illustrative and must be replaced with quotes from the actual issuing authorities; the jurisdiction is unnamed and every substantive number in this section depends on it. The one-time column consumes 41.6% of the \$35,000 pre-opening budget before a single training hour or case of wine is paid for, and Chapter 9 must resolve that collision rather than absorb it.

Chapter 9

Exercise 9.1

A pre-opening budget is the money a restaurant spends between the point at which the space stops being a construction project and the point at which a paying guest first walks in.

The four categories the Bellwether project budget names are labor, training, licensing, and opening inventory.

Three costs that fall inside a real pre-opening period and appear in none of them (any three):

In Bellwether's build these total \$11,696 on their own — a third of the entire frozen line.

Exercise 9.2 †

A friends-and-family service is an invitation-only, non-revenue service for people connected to the restaurant and its staff, run at low volume to test whether the building, the kitchen, and the floor can produce a meal at all. A soft open is a limited-capacity, limited-publicity period of real service before the announced opening, run to find and fix operational failures at a volume the restaurant can survive.

What each tests that the other does not. Friends and family tests capability — can a ticket be fired, can the point of sale ring it, do the stations talk to each other — with expectations low enough that failure is safe. The soft open tests repeatability under load — can the same thing happen four tables in a row at pace, in front of people who will talk about it. Capability is a yes/no question answered in one service. Repeatability is only visible across a sequence.

A legal obligation that applies identically to both: essentially all of them, but the cleanest answer is that both are food services under the applicable food code. There is no exemption for guests who did not pay. Permits, certified-manager coverage, temperature logs, handwashing, and allergen protocol apply to service one. So does the alcohol licensing rule: pouring without a license is a licensing offense in essentially every American jurisdiction, and giving it away is commonly treated the same as selling it.

Exercise 9.3 †

A menu rehearsal is the systematic production of every item on the opening menu, to the standardized recipe, under service conditions, before any guest is served — to verify the recipe, the yield, the plating standard, the station assignment, and the fire time.

The six artifacts:

1. A confirmed cost card for every item, with yields verified in your kitchen.
2. A plating photograph for every item, taken from the guest's angle, printed, hung on the line.
3. A weighed portion standard for every component portioned by eye, with the scoop or ladle chosen.
4. A station map — which station touches which item.
5. A measured fire time for every item, taken with a timer.
6. An 86 plan — what the kitchen does and what the floor says when an item runs out.

The cheapest consistency control is the plating photograph. It costs a phone and a printer, and it converts a plating standard from something in the chef's head into something a new cook can match at four in the afternoon when the chef is at the bank. Almost nobody produces it before opening because in the final fortnight the plate keeps changing — which is itself the argument for freezing the menu in week −4 (§9.3).

Exercise 9.5 †

The honeymoon period is the opening weeks during which a new restaurant experiences both demand and guest tolerance above its sustainable level, driven by novelty, concentrated attention, and a guest population disproportionately composed of friends, neighbors, and industry.

Effect one: it inflates revenue. A meaningful share of opening-month covers are one-time novelty visits that will not repeat on any forecastable schedule.

Effect two: it biases the feedback sample. The guests in the room in week two want you to succeed. They do not send the fish back, they do not write the one-star review, and they forgive the forty-minute ticket. So the comp rate looks good, the reviews look good, and service failures go unreported.

The second is more dangerous, and the reason is that the first effect distorts a number while the second effect disables an instrument. You can correct for inflated revenue by comparing to plan — the number is still being measured accurately, it is just temporarily high. You cannot correct for a feedback channel that has stopped reporting failures, because the absence of complaints reads exactly like the absence of problems. An operator who trusts guest feedback during the honeymoon concludes the restaurant is in better shape than it is, and stops looking.

Exercise 9.7 †

The first ninety days is the period from opening through roughly week thirteen, during which a restaurant's operating habits, cost structure, and guest expectations are established, and after which changing any of them costs several times what changing them now would.

The three reasons §9.7 treats it as distinct:

1. The numbers are not yet meaningful, because there is no baseline to compare them to. A 71% prime cost in week two is neither good nor bad; it is the first reading.
2. The habits are not yet set, which makes them unusually cheap to change. A cut order established in week three is what the restaurant will do for years.
3. The honeymoon is running, which makes the numbers look better than the business is and simultaneously removes the feedback that would normally correct you.

The management contradiction that follows: manage tightly using numbers you cannot yet trust. The resolution is to measure everything and decide almost nothing until a reading repeats.

Exercise 9.9 †

Employer burden at 15%: $\$15{,}077 \times 0.15 = \$2{,}261.55$, carried at \$2,262.

Pre-opening payroll, all in: $\$15{,}077 + \$2{,}262 = \mathbf{\$17{,}339}$.

Exercise 9.11 †

Weekly beverage usage: $\$6{,}900 \times 0.21 = \$1{,}449$.

Multiple: $\$6{,}415 \div \$1{,}449 = \mathbf{4.4 \text{ weeks of usage}}$.

That multiple is the point. A bar buys depth, not turnover — a back bar must be complete on night one even though a given bottle may take months to sell, and a wine list you cannot sell from is a wine list nobody ordered off. Compare a food inventory at roughly 1.2 weeks of usage, where almost everything on the shelf turns inside a week.

Exercise 9.13 †

The gap. $\$59{,}400 - \$28{,}000 = \mathbf{\$31{,}400}$, which is $\$31{,}400 \div \$28{,}000 = \mathbf{112.1\%}$ of the budgeted line — the honest cost is more than twice what was allocated.

The reserve after. $\$38{,}000 - \$31{,}400 = \$6{,}600$.

In days. Monthly fixed obligations of \$41,600 are $\$41{,}600 \div 30 = \$1{,}386.67$ a day.

$$\$6{,}600 \div \$1{,}386.67 = \mathbf{4.8 \text{ days}}$$

Four and a half days is not a working-capital reserve. It is enough to survive a bad weekend and nothing else, and it is the exact condition Chapter 1 §1.4 named as undercapitalization. Note also that the full \$38,000 reserve was only 27.4 days — so this operator was thinly capitalized before the gap and catastrophically so after it.

Exercise 9.15 †

$$2 \times 46 \text{ hours} \times \$15.00 = \$1{,}380$$ $$\$1{,}380 \times 0.14 = \$193 \text{ of burden}$$ $$\textbf{Added cost} = \$1{,}573$$

The argument, in four sentences with a number. The industry runs roughly 75% annual turnover and opening attrition is front-loaded, so on eight opening servers you should expect to lose two or three inside ninety days. Replacing one mid-ramp costs recruiting plus 46 hours of training at \$15 plus burden — about \$790 in direct cost alone — and it lands in the weeks when the trainer is a manager who should be on the floor. The \$1,573 pre-buys two of those replacements and removes the specific failure mode of running a Saturday section short in week three, when your permanent public record is being written. Chapter 17 prices turnover properly; here it is simply the cheapest insurance in the chapter, and the rule is to over-hire the floor, not the kitchen, because a surplus cook has nothing to do and a surplus server has a section.

Exercise 9.17 †

Rent commences on the earlier of opening or thirty days after the certificate of occupancy. The certificate issues 3 March, so rent commences 2 April. Opening is 28 April.

Days of rent running before a guest arrives: 26 days, or $26 \div 7 = 3.714$ weeks.

$$3.714 \times \$1{,}831 = \mathbf{\$6{,}801 \text{ of abatement consumed}}$$

$$\$23{,}800 - \$6{,}801 = \mathbf{\$16{,}999 \text{ remaining for the operating ramp}}$$

So 29% of the abatement was spent on an empty dining room. The operator did nothing obviously wrong — eight weeks from certificate to opening feels careful — and it still cost \$6,801, because the clause gave thirty days and the plan used fifty-six.

Exercise 9.19 †

$$\text{Payroll carried} = 5 \times \$5{,}900 = \$29{,}500$$ $$\text{Abatement burned} = 5 \times \$1{,}780 = \$8{,}900$$ $$\textbf{Total cost of the slip} = \$38{,}400$$

As a share of a \$40,000 pre-opening budget: $\$38{,}400 \div \$40{,}000 = \mathbf{96.0\%}$ — the entire budget, spent again, on nothing.

Two costs the calculation omits (any two):

• Attrition. Staff with signed offers and twice-deferred start dates take other jobs. Each replacement costs recruiting, re-training, and a green body on opening night.
• The abatement's changed job. The \$8,900 burned was supposed to cover the operating ramp. Its loss is not just \$8,900 of cash; it is \$8,900 of cushion removed from precisely the weeks where revenue is least predictable.
• Re-announcement. Reservations already taken, marketing already placed, a press list already told, and the goodwill cost of a second slipped date.
• Perishable opening inventory already ordered against the original date, and deposits paid.
• Owner opportunity cost — five more weeks of two people not earning.

Exercise 9.21 †

Beverage contribution at risk. $$\$24{,}000 \times 0.26 = \$6{,}240 \text{ of weekly beverage sales}$$ $$\$6{,}240 \times (1 - 0.20) = \$4{,}992 \text{ of weekly contribution}$$ $$6 \times \$4{,}992 = \mathbf{\$29{,}952 \text{ over six weeks}}$$

Cost of a six-week delay. $$6 \times \$5{,}400 = \$32{,}400 \text{ of carried payroll}$$ $$6 \times \$1{,}700 = \$10{,}200 \text{ of burned abatement}$$ $$\textbf{Total} = \$42{,}600$$

Opening dry is cheaper by $\$42{,}600 - \$29{,}952 = \$12{,}648$.

The answer you should actually give. Neither, because the honest observation is that you have already lost. The \$29,952 figure assumes you hold every cover with no bar, which a concept planning a 26% beverage mix will not — covers fall too, and the shortfall is unknowable in advance. And the permanent public record of your opening weeks is being written either way.

If forced to choose on these facts, opening dry is arithmetically cheaper and can be managed honestly: state the date the bar opens, price the menu accordingly, and treat it as a phased opening rather than a defect. But the real answer is the structural one from §9.2 — never place opening day on the critical path of an item whose date you do not control. Set opening day after the license is issued and build the countdown backward from that. This decision should not have existed.

Exercise 9.23 †

$$110 \times 0.38 = 41.8 \text{ covers in the peak hour}$$ $$41.8 \times 1.6 = 66.88 \approx \mathbf{67 \text{ plates in sixty minutes}}$$ $$60 \div 66.88 = 0.90 \text{ minutes} = \mathbf{\text{one plate every } 54 \text{ seconds}}$$

If measured capacity is 45 plates an hour, you are 22 plates short — a 33% shortfall — and no amount of exhortation closes it. Four honest responses, roughly in order of cost:

1. Re-engineer the menu to move items off the bottleneck station. Cheapest and most effective, and only available now. If four of nine entrées come off one station, that is a menu problem.
2. Pre-portion or par-cook a component so the à-la-minute work shrinks. Chapter 14 covers the batch-size and waste tradeoff; note that this usually moves cost from labor into food cost, which is a prime-cost trade you should compute rather than assume.
3. Add a body to the bottleneck station. Real labor cost, every night, forever — price it against the covers it unlocks before committing.
4. Pace the book. The host stand controls arrivals (Chapters 22 and 24). Spreading the peak hour reduces the plate rate but also reduces turns, which is revenue.

What you do not do is plan to cook faster. A measured capacity is a measured capacity.

Exercise 9.25 †

$$\$230{,}534 \div \$325{,}000 = 70.9\%$$

Excess against the 61% target: $$\$325{,}000 \times 0.61 = \$198{,}250$$ $$\$230{,}534 - \$198{,}250 = \mathbf{\$32{,}284}$$

What weeks 14–52 must do: $$\text{Annual prime } \$ = \$1{,}300{,}000 \times 0.61 = \$793{,}000$$ $$\text{Remaining revenue} = \$1{,}300{,}000 - \$325{,}000 = \$975{,}000$$ $$\text{Remaining prime } \$ = \$793{,}000 - \$230{,}534 = \$562{,}466$$ $$\$562{,}466 \div \$975{,}000 = \mathbf{57.7\%}$$

Do I believe it? No, not without a named plan. Two things have to happen at once. Prime cost must fall 13.2 points, from 70.9% to 57.7%. And revenue must rise: \$975,000 across 39 weeks is \$25,000 a week against a week-13 exit rate of \$23,800 — a 5.0% climb that must then hold. Either one alone is a hard year's work. Both simultaneously, from a business that has been getting worse every four weeks, is a claim requiring evidence: which points come from food cost, which from labor, which channel supplies the revenue, and on what dates. Absent that, the honest statement is that the annual plan is no longer achievable and should be restated.

Exercise 9.27 †

The real problem is Tuesday, and it is not visible to a manager watching the weekend.

What fixed labor costs on a 38-cover Tuesday. At a \$46 check, 38 covers is \$1,748 of sales. A minimum viable crew still has to exist:

\$583 on \$1,748 is 33.4% hourly labor alone, before any allocation of the salaried chef and manager who must also be in the building, and before COGS. Add COGS at 28% (\$489) and you are at \$1,072 — 61.3% prime cost before a dollar of salaried labor is counted. Tuesday does not pay for itself.

Three options, none of which is "market it harder":

1. Close Tuesday and redeploy the week. Check the lease's continuous-operation clause first (Chapter 6), and be honest that closing removes a night of fixed-cost absorption as well as a night of loss — Chapter 32 shows you how to test which effect dominates.
2. Change the Tuesday offer so the fixed floor shrinks: a bar-forward or single-station menu, a prix fixe, a shorter service window. You are not trying to fill 95 seats; you are trying to make 38 covers profitable.
3. Restructure Tuesday's labor. One salaried person covering both roles, cross-trained staff (Chapter 18), a later open, a cut order written for 38 rather than improvised at nine o'clock.

Exercise 9.29

A strong answer concedes, prices, and proposes. Model shape:

(a) What is true in the proposal. The gap is real, cash is the binding constraint, and the reserve is finite. A partner who proposes cuts is doing the right thing by refusing to pretend the gap away.

(b) Price each cut.

(c) What to do instead. Hold training and inventory. Cut where the guest cannot feel it: reduce pre-opening marketing from \$3,000 to \$1,200 (a restaurant that executes gets found; a restaurant that is found and does not execute has paid to accelerate bad word of mouth), negotiate 30-day terms with two vendors to shift several thousand dollars of opening inventory into month one's cash, and then take the balance from the working-capital reserve explicitly, in writing, with the restated reserve figure published in the plan — including what it is worth in days of fixed cost.

Exercise 9.31 †

There is no single right answer; the grading is on whether the build is bottom-up, whether it foots, and whether the student states the gap honestly. A worked model for a 60-seat, full-service, beer-and-wine, dinner-six-nights concept:

Labor before revenue

Burden at 15% = \$2,450. **Labor subtotal \$18,785.**

The rest

$$\$18{,}785 + \$5{,}200 + \$6{,}150 + \$3{,}790 + \$3{,}200 + \$9{,}200 = \mathbf{\$46{,}325}$$

The point of the exercise is the last paragraph the student writes. Almost everyone allocates \$20,000–\$30,000 before building it and lands somewhere near \$46,000 after. Note also what dropping spirits did: beverage inventory falls from Bellwether's \$8,120 to \$3,790, which is the single largest structural saving available to a beer-and-wine concept and is worth naming in the plan.

Chapter 10

Exercise 10.1

The five jobs: sales (what a guest is asked to choose from, and at what price), production (what the kitchen must make, and in what order), purchasing (what arrives on the dock, from whom, how often), labor (how many hands, at what skill, at which station), and brand (what the guest believes about you before the food arrives).

Purchasing and labor together are prime cost. Purchasing sets cost of goods sold; the labor model sets the labor line. The menu therefore fixes both halves of the number Chapter 1 said predicts survival, months before anyone is hired or anything is bought — which is why it is written before the staffing guide (Chapter 19) and before the order guide (Chapter 13).

Exercise 10.3

Item count is the number of distinct items offered on a menu.

Three costs it drives that never appear as their own P&L line:

1. Prep labor — every item is a task on a prep list, and the tasks compete for the same morning. It surfaces as part of the labor line, indistinguishable from anything else.
2. Spoilage — more items means lower velocity per ingredient, which means more product sitting, which means more discarded. It surfaces inside food cost, and inventory usage cannot tell you it was caused by item count.
3. Slower ticket times at peak — more items means more stations under load and more component pickups per plate. It surfaces as covers you did not turn, which appears on no report at all.

Also acceptable: training hours, order-guide and count-sheet lines, walk-in space, and the noisier sales signal that makes menu analysis take longer to say anything.

Exercise 10.4 †

A standardized recipe tells the kitchen what to make: components in exact quantities, method, plating, hold and reheat instructions, portion size, a ticket-time target, and a description of what correct looks like. Its job is consistency — the same dish leaving the pass the same way on the four hundredth repetition as on the first, executed by a cook who is not the chef.

A recipe cost card tells the business what the dish costs: the same components, priced, plus a waste allowance, producing a plate cost. Its job is pricing and control. That is Chapter 11's document.

What they share is the quantities, and that is exactly why the recipe must be written in weights and fluid ounces rather than in "a handful." A cost card is arithmetic performed on the recipe's numbers. "A handful" cannot be multiplied by a price per pound; six ounces can. A kitchen that writes recipes in adjectives has made its own costing impossible, and it will not discover that until it tries.

Second-order point worth making: portions expressed in weights are also the only enforceable ones. "Generous" drifts; six ounces can be checked on a \$40 scale.

Exercise 10.5

A menu panel is a bounded visual region of the menu — a column, a boxed section, one side of a folded card — that the eye takes in as a unit.

The thing it genuinely controls, regardless of any psychological claim, is what is compared with what. Items inside a panel are evaluated against each other; items in different panels are not, or much less so. Bellwether splits its mains into "from the hearth" and "from the stove" precisely so that a guest weighs the chicken against the pork chop rather than against the pasta — a real design lever that requires no theory of eye movement to work.

Exercise 10.7

Exercise 10.8 †

Note the pattern, which is the real answer: only one of the four was rejected for being a bad dish. Everyone in the room could cook all four.

Exercise 10.9

Concept fit. A whole roasted branzino is a share-and-linger item with a theatrical element. Ask whether the 60-seat neighborhood room's dominant occasion — most likely a weeknight two-top on a compressed timeline — will order it, and at what price. If the concept's check average is in the thirties, a whole fish priced to cover a whole fish will sit at the top of the ladder and sell four a week. It can still be a signature if the concept is celebration-weighted; it cannot be one if the restaurant is a Tuesday restaurant.

Skill fit. Whole-fish cookery is a real skill with a narrow window between underdone at the spine and dry at the tail, and it is unforgiving at volume. If it is presented whole, somebody in the dining room also has to bone it at the table — which makes it a front-of-house training obligation too (Chapter 18).

Equipment fit. This kitchen has a flat-top, a six-burner, a convection oven, and a fryer. The convection oven will cook the fish. It will not produce the crisp, blistered skin the dish is usually bought for, and there is no salamander or broiler in the list to finish it. This is the decisive filter.

Verdict: reject as a signature. Two honest alternatives: run it as a limited special (§10.8), where a low count and a per-order price are acceptable; or reformulate as a portioned fillet that the flat-top can genuinely execute — a different dish, honestly named, that the equipment supports.

Exercise 10.11 †

Affected item: the chicken liver mousse. It is built entirely from a component that arrived attached to the birds.

What has to change on the purchasing side: livers become a purchased line of their own. That means a new product specification (Chapter 13), a supplier, a delivery cadence, and — because livers are highly perishable — a short shelf life and a real spoilage risk on a component that previously had none. It also means a minimum order quantity that may not match the roughly seventeen portions a week the item sells.

Where the change shows up first: on the order guide and the spec, which is why the chapter says to ask the question before you write the item. Financially it appears next on an invoice, and then on the cost card in Chapter 11 as a line that used to be free. If the purchased cost is material, the item becomes a candidate for reconsideration in Chapter 12's analysis — a menu item whose entire economic logic was "it came with something else."

The transferable rule: any item justified by a by-product needs the by-product written into the purchasing spec. A cross-utilization map is only as good as the specification that guarantees the parts arrive.

Exercise 10.13

The risk: a meal that reads as repetitive even though each individual dish is good. The guest does not order a dish; they order a meal, and the meal is the product. Case Study 2 prices what this costs.

Two ways to cross-utilize the component without the guest noticing:

1. Vary the finish, not the base. Keep the one vinaigrette in the walk-in and split it at the point of use — more acid and chile on one item, brown butter folded through it on another, emulsified into a dressing on a third. Same purchased box, four plates that do not taste like each other.
2. Vary the role and separate the placements. A component used as a dressing on a salad, as a drizzle on a vegetable, and as a marinade that is cooked and therefore transformed reads as three different things. Then place them so they are unlikely to land on the same check — one in a category a guest rarely doubles up on, one on an item that is a two-top's second choice.

Third answer, also correct: count the placement rate. Four of how many? If it is four of twelve, you have a problem; four of thirty is unlikely to be noticed.

Exercise 10.14 †

The reasoning in prime-cost terms. Breaking down primals in house lowers food cost — you pay less per pound for a primal than for portion-cut steaks, and you gain trim. It raises labor — skilled butchery hours, plus the cost of a mis-cut on an expensive piece of beef. Chapter 1 named this trade explicitly and said it runs in both directions; the only way to know whether it was a good trade is to watch prime cost, not food cost.

Why Bellwether declines it. Two reasons, both structural. First, volume: the ribeye is 6% of entrées, five or six a night, roughly thirty a week. That is far too little to amortize the labor of breaking and aging primals. Second, the trim has no home: the burger grind is already fed by hanger trim, chuck, and brisket, so ribeye trim would not reach another menu item and much of the cross-utilization benefit evaporates.

What you would measure to know.

1. The per-pound price difference between the primal and portion-cut steaks, from real quotes.
2. Your own usable yield off the primal — not the supplier's number (Chapter 11 runs the test).
3. Pounds required per week at your actual sales volume.
4. Butchery hours per week, at a fully loaded wage.
5. An allowance for cutting error, which is not zero on a first-time butcher and is expensive on ribeye.

Then compare: annual food-cost saving against annual labor cost plus error allowance. If the saving does not exceed the labor by a comfortable margin, buy it cut — and note that "comfortable margin" is doing real work, because the labor number is certain and the saving is an estimate.

Exercise 10.15

A restaurant's map assumes three things a truck does not have: volume (enough throughput for a trim stream to be worth capturing rather than spoiling), storage (a walk-in that can hold multiple programs at different stages), and stations (somewhere to do the butchery, the rendering, and the stock).

A truck's map should look different in four ways:

1. Fewer programs, used harder. Two or three purchased families feeding all eight items, rather than eight families feeding thirty-four lines.
2. Cross-utilization aimed at reducing volume held, not just line count. The binding constraint is cubic feet, not SKU count. An ingredient that appears in six items but must be held in three different prepped states has not helped.
3. Components that survive a day, not a service. Trucks often prep at a commissary and run for several days, so shelf life matters more than it does in a restaurant with daily deliveries (Chapter 30).
4. No dependence on trim streams that require scale. If you sell forty portions a week, the stock program is a fantasy and the livers spoil. Buy the component or leave the item off.

Exercise 10.16 †

$$0.85 \times \$27.00 = \$22.95$$ $$0.50 \times \$11.00 = \$5.50$$ $$0.18 \times \$9.50 = \$1.71$$

$$\$22.95 + \$5.50 + \$1.71 = \mathbf{\$30.16} \text{ of food per cover}$$

Exercise 10.17

Target is \$32.00; the menu produces \$30.16. The shortfall is \$1.84 per cover.

Three levers, with the arithmetic each one requires:

1. Raise the weighted average entrée price. Since entrées carry an 0.85 attachment, $\$1.84 \div 0.85 = \$2.16$ — the weighted average entrée price must rise from \$27.00 to \$29.16. That is either a price increase or a mix shift toward the top of the ladder.
2. Raise starter attachment. $\$1.84 \div \$11.00 = 0.167$, so attachment would have to go from 0.50 to 0.667. That is a very large behavioral change and almost certainly not achievable by itself; it is a service and menu-design problem, not a pricing one.
3. Raise dessert attachment. $\$1.84 \div \$9.50 = 0.194$, taking attachment from 0.18 to 0.374 — more than doubling it. Implausible alone.

The honest answer, and the one an operator should give: no single lever closes it. A realistic plan combines a \$1.00 move in the entrée average (\$0.85), a starter attachment lift to 0.55 (\$0.55), and a dessert lift to 0.23 (\$0.475) — \$1.875 in total, which clears \$1.84. Note that two of the three are service changes, not menu changes, which is a useful thing to discover about your own check.

Exercise 10.19 †

The six new points are drawn proportionally, so the original five items scale by 0.94.

Original weighted average across those five was \$25.76 (Exercise 10.18). Scaled:

$$\$25.76 \times 0.94 = \$24.2144$$

New item: $\$44 \times 0.06 = \$2.64$.

$$\$24.2144 + \$2.64 = \mathbf{\$26.85} \text{ (to the cent, } \$26.8544\text{)}$$

The weighted average entrée price rises \$1.09. At an entrée attachment of 0.88:

$$0.88 \times \$1.0944 = \mathbf{\$0.96} \text{ more food per cover}$$

Interpretation. One item at six percent of the mix moved the check by nearly a dollar without a single existing price changing. That is worth saying out loud, because it is the mechanism most operators reach for last and it is often the cheapest: you can move a check average by changing what people order rather than by changing what things cost. It is also the mechanism that runs in reverse — a cheap item that over-performs pulls the average down exactly as hard.

Exercise 10.21

Target check \$38.00 at a 70/30 mix, so food per cover is $0.70 \times \$38.00 = \$26.60$.

Starters: $0.40 \times \$10.00 = \$4.00$. Desserts: $0.12 \times \$8.00 = \$0.96$. Together \$4.96.

Entrées must contribute $\$26.60 - \$4.96 = \$21.64$, so:

$$\$21.64 \div 0.92 = \$23.52 \rightarrow \text{round to } \mathbf{\$24}$$

Is it plausible? Yes — a weighted average entrée price of \$24 with \$11.40 of beverage per cover describes a normal American casual full-service restaurant. A defensible ladder around that average might run \$16, \$19, \$22, \$24, \$26, \$29, \$34. Note that rounding *up* to \$24 pushes food per cover to \$27.04 and the check to \$38.63 — which is fine, and is worth noticing rather than discovering later.

Exercise 10.22 †

One valid answer among many. Prices: \$18, \$22, \$24, \$26, \$28, \$33, \$42. Three items fall between \$23 and \$28 (\$24, \$26, \$28); the low is \$18 and the high is \$42; no price repeats.

$1.80 + 3.52 + 4.56 + 5.72 + 4.76 + 3.96 + 1.68 = \mathbf{\$26.00}$ exactly.

What to check in your own answer, beyond the arithmetic:

• The high-to-low ratio here is $42 \div 18 = 2.33$, inside the usual working range of about 2.5×.
• The median is \$26 and the weighted average is \$26.00 — they agree, which means no single item is distorting the menu.
• The two largest shares sit at \$24 and \$26, in the tight middle of the ladder. If your heaviest share sits on your cheapest item, your ladder is fine and your mix is a problem — a distinction Chapter 12 exists to make.

Exercise 10.23

Replacing the ribeye (\$46) with a hanger steak (\$36) removes \$10 from the night's food total:

$$\$828 - \$10 = \$818 \qquad \frac{\$818}{25} = \mathbf{\$32.72}$$

The average falls \$0.40 a cover — about 1.2% — from a single substitution at a single table.

What it tells you about reading one service: almost nothing is stable at n = 25. One guest choosing differently moved the headline number by forty cents, which annualized across Bellwether's 24,700 dinner covers would be \$9,880 of revenue. An operator who "reads the check average" off one night is reading noise, and will make a menu decision from it. This is the same discipline Chapter 12 insists on before drawing a matrix, and the same one Chapter 1 applied to a single month's food cost.

Exercise 10.24 †

Units sold: $12 \times 52 = 624$ a year.

$$\frac{\$5{,}915}{624} = \mathbf{\$9.48 \text{ per plate}}$$

Interpretation. Nine dollars and forty-eight cents attaches to every plate of this dish before a single ingredient is bought for it — nearly half again the \$6.45 figure in §10.4, driven mostly by the forty-minute prep and the six-day week. For most casual full-service price points that is a substantial fraction of the menu price, and it is invisible on every report the restaurant produces.

Do not report this as the item's cost. It is the non-ingredient cost. Chapter 11 adds the ingredients, and only then can Chapter 12 say whether the item earns its place.

Exercise 10.25

Question one: what does it contribute in dollars, not units? "It sells" is a statement about popularity, and popularity is only half the picture — Chapter 12's entire argument. Twelve a week at a thin margin is worse than four a week at a fat one. I would want the item's margin in dollars per plate against the \$9.48 of non-ingredient cost computed in Exercise 10.24.

Question two: is the sale incremental or transferred? If the item vanished, what would those twelve guests order? If eleven of them would order something else on the menu and stay, the item is producing almost no incremental revenue and pure incremental cost. If some fraction of those twelve are guests who come for this item and would otherwise not come, that changes the answer completely.

What would change my mind: a genuinely high contribution in dollars, plus credible evidence of incrementality — a repeat pattern in the point-of-sale data showing the same guests returning for it, or a dietary or occasion segment the rest of the menu does not serve. Absent both, the item is a candidate for the specials board (§10.8) rather than the printed menu, which keeps it available without keeping it permanent.

A third question worth asking: what does it do to peak throughput? An item that fires from the busiest station at 7:45 costs more than the same item fired from a station with slack.

Exercise 10.26 †

The allocation. With four bodies against 84 fire actions, 9 plancha items, 11 pastas plus every sauce, and 36 cold pickups and desserts:

The plancha rides with sauté because nine burgers a night is easily absorbed and the two are adjacent; splitting them would waste a body.

The fifth pair of hands, for three hours: the hearth. The arithmetic makes the case. The fire carries 84 of roughly 141 plates — about 60% of everything the kitchen sends — and it is the only station running near its practical ceiling, at 68% utilization at peak with real upside variance in the flatbread. A second person on the fire, even one doing nothing but flatbreads and vegetable finishes off a separate zone, removes roughly 20 of the 84 actions from the primary cook and takes peak utilization down toward the low fifties.

The defensible alternative is garde manger, and it is worth taking seriously: 36 plates is the largest count outside the hearth, and cold pickups gate the start of a table's meal, so a slow starter delays everything downstream and is felt by the guest first. If your restaurant's complaint pattern is "we waited a long time for the first thing," staff garde manger. If it is "the entrées came out slow," staff the fire. The numbers here favor the fire; the guest data would settle it (Chapter 23).

Exercise 10.27

Eighty-eight percent at peak is not a working number. It leaves no room for a re-fire, a mis-called ticket, a large party ordering four ribeyes, or a cook having a bad ten minutes — and every one of those happens weekly.

The first thing I would change on the menu: cap or pull the hearth flatbread during the peak window. It is worth roughly 5.8 of those actions, it is the only item that is both discretionary and directly colliding with entrée fire, and 86ing it at peak costs the restaurant a \$16 sale rather than a \$29 one.

The second change, if capping is not enough: move the ember trout off the fire. At 13% of entrées and a nine-minute dwell it is the largest single block of fire time that could plausibly be reformulated to a plancha-and-oven method — roughly 7.8 actions and a disproportionate share of the dwell.

And the change that is not the menu's to make: adding hearth positions. That is an equipment decision and belongs to Chapter 7, but note the direction of the dependency — the menu states the requirement, and if the equipment cannot meet it, one of the two documents has to change.

Exercise 10.29

Open-ended. Grade against six criteria, all of which appear in Figure 10.5:

1. A station is named. If the card does not say where the dish fires from, it is a home recipe.
2. Every quantity is a weight or a fluid measure. Any "handful," "drizzle," "to taste" on a costed component is a failure — Chapter 11 cannot cost it and no manager can enforce it.
3. Prep and service methods are separated, with any hold, cool, or reheat step stated explicitly and any required temperature named.
4. The plating instruction is specific enough to be wrong. "Plate attractively" is not an instruction. "Sauce over the leg only, never over the skin" is.
5. There is a 'what good looks like' paragraph written so a cook in week two can self-check without asking anyone. This is the field students most often omit and the one that does the most work.
6. There are no prices anywhere on the card.

Strong submissions also carry a ticket-time target and an 86 threshold.

Exercise 10.30 †

A model answer. The policy should be one page and should specify five things.

SPECIALS POLICY

1. Who. The chef writes the special. The sous costs it. If the sous is off, the chef costs it. The manager on duty verifies that a card exists before the special is entered in the point-of-sale system. No exceptions and no verbal approvals.

2. What document. Every special gets a cost card on the standard template, with components, quantities, and the same waste allowance the printed menu uses. It takes about eight minutes once the component prices are already in the system.

3. By when. The card is complete and the price is set before pre-shift, not before service. A special without a card at pre-shift does not run that night.

4. What gets communicated. The special goes on the pre-shift board with its price and its description, both written out. Servers taste it. A special that servers have not tasted is a special that gets described wrong, and a special described wrong is a comp.

5. What happens when there is not time. This is the clause that makes the policy real, because "there was not time" is how every control dies. The answer is: the special does not run. The item goes into the walk-in for tomorrow, gets used in family meal, or gets folded into an existing menu item. The kitchen does not sell an unpriced plate.

6. What gets counted. Every special's unit sales and price are recorded at close. At each seasonal rebuild, the specials log is reviewed and the top performers are candidates for the printed menu. This turns the specials board into a research pipeline rather than a leak.

Grading note: the two clauses students most often omit are #5 and #6. #5 is the enforcement mechanism; a policy with no stated failure mode is a suggestion. #6 is what converts the policy from a cost control into a competitive advantage.

Exercise 10.31

A model answer, in three sentences:

"The menu is the only document in this building that carries prices, and we change prices — a letterpress run at \$1,100 means that the next time a protein contract moves, the cheapest thing to do is nothing, and doing nothing costs us more than the printing. What we need from you is a beautiful template we can print in-house every Tuesday on good uncoated stock, plus one expensive object that genuinely does not change — a cover, a card, or a jacket that holds the insert. Design the thing that lasts; let us own the file for the thing that moves."

The reasoning to look for: the student has separated what is permanent (identity, cover, format) from what carries a price, and has understood Chapter 3's arithmetic — that a \$900 run pays for itself in delay after fewer than six weeks, and that nobody delays six weeks, they delay until the next natural menu change.

Exercise 10.33

(a) The three filters. Concept fit: strong. The item is already the signature; the request is for the signature, which means the guest wants exactly what the restaurant is for. Turning away a table twice a week is a concept-fit argument for accommodating. Skill fit: moderate. A second sauce batch is not difficult; the difficulty is procedural rather than culinary — labeling, separate utensils, separate storage, and a handoff that survives a busy Saturday and a new cook. Equipment fit: fine. Nothing new is required.

(b) The obligations. This is allergen and dietary-restriction management, and Chapter 25 owns it properly. The structural requirements: a documented procedure for the modification; separate storage and labeling of the alternate sauce and bread; prevention of cross-contact at the point of assembly, not just at prep; a trained and specified path for the ticket to reach the kitchen as an allergen ticket rather than a note; and staff training for everyone who touches the plate — the server who takes the order, the expediter, the cook who fires it, and the runner. Requirements vary by state and locality; verify locally. Note also the honest distinction between "gluten-free" as a claim and "made without gluten-containing ingredients" as a description — the two are not the same in a kitchen with a flour-dusted bread station, and the language matters.

(c) What I would do. Accommodate, with the procedure written down first. The volume is real (roughly a hundred tables a year), the item is the signature, and the change is one sauce and one bread. But the accommodation goes live only after the procedure exists and the whole staff has been trained — not the Friday the question is asked.

Two sentences for the training notes:

"The chicken can be made without gluten-containing ingredients: substitute the reserved sauce from the labeled blue container and serve with the rice-flour crisp instead of grilled bread. Tell the guest exactly that — 'made without gluten-containing ingredients, in a kitchen that also handles flour' — and never say 'safe,' because we cannot promise a dedicated space."

(d) If the modification is not reliably safe. The right answer is a specific, honest sentence that respects the guest and does not read as a disclaimer:

"I'm sorry — I can't do that one safely. We bake bread here and there's flour in the air all morning, so I won't tell you it's clean when it isn't. What I can do is the trout or the roasted roots, and I'll walk it through the kitchen myself so nothing touches it."

Grading note. Look for two things: whether the student distinguishes an accommodation from a claim, and whether the refusal answer offers something rather than merely declining. The second is the hospitality point — a "no" that ends with a better "yes" is a different experience from a "no."

Chapter 11

Exercise 11.1

$$\text{Food cost \%} = \frac{\text{beginning inventory} + \text{purchases} - \text{ending inventory}}{\text{food sales}}$$

The invoice-over-sales shortcut computes purchases ÷ food sales, which is a purchasing ratio: how much you spent relative to what you sold. It is not food cost, because it ignores the fact that inventory levels move — buy heavy and it spikes, run the walk-in down and it flatters you.

Exercise 11.2

As-purchased (AP) cost is the price of an ingredient in the condition it was delivered, including peel, bone, stem, fat, water, and outer leaves — what the invoice says. Edible-portion (EP) cost is the cost of the usable product after everything you cannot serve has been removed.

$$\text{EP cost} = \frac{\text{AP cost}}{\text{Yield \%}}$$

The conversion raises cost because you paid for the whole thing and can only sell part of it. The whole purchase price must be recovered across a smaller usable weight, so the per-usable-pound figure is necessarily higher. A 75% yield means the ingredient costs 1 ÷ 0.75 = 133% of the invoice price.

Exercise 11.3

A waste allowance is a percentage added to the sum of costed components to cover what the recipe cannot see: unmeasured trim, spillage, a broken sauce, a plate returned, the last of a batch that dried out. Common practice runs roughly 2% to 5%, higher where there is heavy scratch production or an expensive protein.

The single piece of evidence that tells you whether yours is set correctly is the waste log (Chapter 13). Everything else is assertion.

Exercise 11.5

Ideal (theoretical) food cost is what the period should have cost: the sum, across the menu, of units sold × plate cost. Actual food cost is what it did cost: beginning inventory + purchases − ending inventory.

Six things that live in the gap: over-portioning; waste and spoilage; uncosted items (specials, off-menu requests, new dishes); comps, voids, and staff meals; receiving errors (short deliveries, uncaught price increases, substitutions accepted without a credit memo); and counting or pricing errors on the inventory sheet. Theft is a seventh, and it is investigated last.

Exercise 11.7

A "twelve" is a #12 disher, holding about 2.7 fluid ounces. The disher number is scoops per quart, so the volume is 32 fl oz ÷ 12 = 2.67 fl oz.

Exercise 11.8

Components:

\$6.80 + \$0.72 + \$0.88 + \$0.54 + \$0.21 + \$0.11 = \$9.26

Waste allowance at 3%: \$9.26 × 0.03 = \$0.2778 → \$0.28

Plate cost: \$9.26 + \$0.28 = \$9.54

Price at a 30% target: \$9.54 ÷ 0.30 = **\$31.80**

In practice you would not print \$31.80. You would price it at \$32 (30.0% food cost — \$9.54 ÷ \$32 = 29.8%) or wherever the menu's price ladder puts it, and note that the formula's answer was \$31.80 so you know how much you moved and why.

Exercise 11.9

As the recipe claims: \$34.20 ÷ 52 fl oz = \$0.6577 per fl oz. Portion of 2.5 fl oz = \$1.64.

As measured: \$34.20 ÷ 46 fl oz = \$0.7435 per fl oz. Portion of 2.5 fl oz = \$1.86.

Difference: \$0.22 per plate.

Annually: 140 plates × 52 weeks = 7,280 plates × \$0.22 = **\$1,601.60**.

Sixteen hundred dollars a year on one component of one dish, because the recipe's stated yield was never checked against a scale. This is the most common single error in a cost-card binder.

Exercise 11.10

The Q factor:

As points of food cost: \$0.42 × 28,000 covers = **\$11,760 a year. Against \$840,000 of food sales: \$11,760 ÷ \$840,000 = 0.014 = 1.4 points of food cost**.

Note that the Q factor applies per cover, not per entrée — a guest who orders only an appetizer still eats the bread — which is why it must be computed against covers and applied to the whole food-cost build rather than buried in an entrée card.

Exercise 11.11

Components: \$5.60 + \$0.95 + \$1.05 + \$0.42 + \$0.18 + \$0.15 = \$8.35 ✓ Waste allowance: \$8.35 × 0.02 = \$0.167 → \$0.17 ✓ Plate cost: \$8.35 + \$0.17 = \$8.52 ✓

Price for exactly a 28.0% food cost: \$8.52 ÷ 0.28 = **\$30.43**.

Worth noticing: moving from a 29.4% food cost to a 28.0% one requires a \$1.43 price increase on a \$29.00 item — nearly 5% — for 1.4 points of a ratio. That is the arithmetic behind §11.8's warning about chasing percentages.

Exercise 11.12

| Allowance | Allowance \$ | Plate cost | Food cost at \$29.00 | |---|---|---|---| | 2% | \$0.17 | **\$8.52 | 29.4% | | 4% | \$0.33 | **\$8.68 | 29.9% | | 5% | \$0.42 | **\$8.77 | 30.2%** |

(\$8.35 × 0.02 = \$0.167; × 0.04 = \$0.334; × 0.05 = \$0.4175.)

What the range implies: the allowance alone swings the reported food cost of this item by 0.8 points — from comfortably under the 30% plan target to over it — without a single thing changing in the kitchen. An allowance is therefore not a rounding convention. It is a substantive assumption that deserves evidence, and the evidence is the waste log.

Exercise 11.13

Using the Hearth Chicken as the model, the "oil, salt, misc" and "garnish" lines are \$0.18 + \$0.15 = **\$0.33**, which is 3.95% of the \$8.35 of components.

A card omitting them would carry \$8.02 of components; with the 2% allowance, a plate cost of \$8.02 × 1.02 = **\$8.18 against the true \$8.52** — an understatement of **\$0.34, or 4.0%**.

Across a menu of thirty items with a similar proportion, that is roughly 4% of food cost. On a restaurant running 30% food cost, about 1.2 points — and it will show up as a persistent, unexplained ideal-to-actual variance that the operator will spend months chasing in the kitchen, where it is not.

Exercise 11.15

$$\text{Yield \%} = \frac{26.4}{40.0} = \mathbf{66.0\%}$$ $$\text{EP cost} = \frac{\$1.28}{0.66} = \mathbf{\$1.94 \text{ per pound}}$$

Second route: 40 lb × \$1.28 = \$51.20 for the case; \$51.20 ÷ 26.4 lb usable = **\$1.94**. The two agree, which is the check.

The invoice says \$1.28. The plate costs \$1.94 — 52% more than the invoice price, on a product nobody thinks of as expensive.

Exercise 11.16

(a) Reconciliation. 3.4 + 2.8 + 16.4 = 22.6 lb ✓ — the streams account for the whole sub-primal, so nothing left the station unweighed.

(b) Yield to the steaks. 16.4 ÷ 22.6 = 72.6%

(c) EP cost per pound.

Total AP cost: 22.6 lb × \$9.85 = **\$222.61**

• Without credit: \$222.61 ÷ 16.4 lb = **\$13.57 per pound**
• With credit: grind trim 2.8 lb × \$5.60 = **\$15.68; \$222.61 − \$15.68 = \$206.93; \$206.93 ÷ 16.4 lb = **\$12.62 per pound

(d) 8 oz portion (0.5 lb) at the credited EP figure: 0.5 × \$12.62 = **\$6.31**

(e) At the AP price: 0.5 × \$9.85 = **\$4.93 — understating the true portion cost by \$1.38**, which is 21.9% of the correct figure. On 200 portions a week that is \$276 of cost the card does not know about, and \$14,352 a year.

Exercise 11.17

Total-out ÷ total-in: 41.6 ÷ 58.0 = 71.7%

Simple average of the three percentages: (72.6 + 66.7 + 73.5) ÷ 3 = 70.9%

The weighted figure (71.7%) is correct. The simple average gives the small twelve-pound test the same weight as the twenty-three-pound one, which overstates the influence of whatever was unusual about it. They differ by 0.8 points here, and 0.8 points of yield is real money on a protein.

The deeper reading: test 2's 66.7% is 6.8 points below test 3. That spread is the finding. Before you cost anything, find out what was different about that delivery — a different size grade, a different cook, a different lot — because the spread is your actual exposure.

Exercise 11.18

• Trim yield: 9.0 ÷ 12.0 = 75.0%
• Cooking yield: 7.6 ÷ 9.0 = 84.4%
• Overall AP-to-as-served: 7.6 ÷ 12.0 = 63.3%

Check the chain: 0.750 × 0.844 = 0.633 ✓

Which belongs on the card? A spec reading "5 oz roasted" is written in as-served units, so the card must carry the overall 63.3% — both losses. Carrying only the trim yield would understate the line by about a third.

The general rule: match the yield to the unit the portion spec is written in, and write on the card which unit that is.

Exercise 11.19

Five questions, in the order you would actually ask them:

1. What spec did their test use, and what spec do we use? If they trimmed to a commodity standard and we trim for presentation, we cut away more by design, and the gap is a plating decision rather than an error.
2. What did they count as yield? If their figure includes frames, collars, or trim that we discard, the two numbers are measuring different things.
3. Who ran our tests, and were all three the same person? A prep cook three weeks into the job and a cook of ten years produce genuinely different numbers on the same product.
4. What size grade did we actually receive? Larger fish and larger primals generally yield better; a spec sheet quotes one grade and a delivery may be another.
5. Was our knife sharp, and did we weigh everything? An unreconciled stream — the pan of trim that went straight into the stockpot — reads as lost yield.

Only after all five would you conclude that anyone is wrong. And you would still cost from your own number, because your yield is a property of your kitchen.

Exercise 11.21

Whole birds, per portion: \$5.60 of product + \$0.79 of butchery and brining labor = \$6.39

Pre-portioned halves, per portion: a half of a 3.5 lb bird is 1.75 lb; 1.75 × \$4.35 = **\$7.61** of product, no butchery labor.

The switch is clearly wrong at these prices. On 4,992 portions a year it adds \$6,090 to prime cost. The break-even price for the pre-portioned product is \$6.39 ÷ 1.75 lb = **\$3.65 a pound** — below that, the trade becomes favorable.

What would change the answer:

• Whether the labor is genuinely eliminated. If the prep cook stays on the clock and does something else, you have not saved \$0.79 — you have saved nothing, and the trade is worse than the table shows. This is the single most common error in convenience-product analysis.
• The lost by-products. Whole birds give you frames and backs for stock. Buying halves means buying stock or bones you were previously getting free, which is an additional cost not in the table.
• Portion consistency. Pre-portioned product removes the over-weight-bird drift worth \$0.32 a plate in §11.6, which claws back a quarter of the gap.
• What the freed hour does. If it enables production that generates revenue, evaluate it against that revenue, not against zero.

Exercise 11.22

(a) \$4.85 ÷ 0.29 = **\$16.72**

(b) At \$19.00: food cost \$4.85 ÷ \$19.00 = **25.5%**; contribution margin \$19.00 − \$4.85 = \$14.15

(c) At the formula price, contribution would be \$16.72 − \$4.85 = \$11.87. The difference is \$14.15 − \$11.87 = \$2.28 per plate**. At 85 plates a week: 85 × 52 = 4,420 plates × \$2.28 = \$10,077.60 a year**.

(d) The item is cheap to make. Target-cost pricing caps the price at plate cost × 3.45, which is a statement about ingredients and not about what the dish is worth to a guest. The guest is buying a composed dish, not a bill of materials.

Exercise 11.23

Allowable plate cost: \$26.00 × 0.31 = **\$8.06**

Your card is \$8.90 — you have negative headroom of \$0.84. The item currently runs \$8.90 ÷ \$26.00 = 34.2% food cost, more than three points above target.

This is the useful direction of the formula: it did not tell you what to charge, it told you the dish is 84 cents too expensive to build for the price the market allows. The options are to find 84 cents in the card — a yield improvement, a cheaper garnish, a smaller protein portion honestly specified, a better purchase price — or to accept the item at 34.2% because its contribution margin (\$17.10) is good and something else on the menu will carry the blend.

Exercise 11.24

The pattern is the lesson. Target-cost pricing under-prices everything cheap to make and over-prices everything expensive to make, and it does so systematically rather than randomly. The two items it would have priced at \$54 and \$49 are the two best contribution-margin items on the list; pricing them to the target would very likely delete them from the menu by pricing them out of the market and cost the restaurant its two strongest dollar-contributors.

Note also that the formula's answer is closest to correct on the mid-priced, mid-cost item — the chicken. That is the zone where target-cost pricing is genuinely useful, and it is a narrow one.

Exercise 11.25

The rebuttal, in substance:

Paragraph one — the arithmetic does not do what the consultant thinks. Overall food cost is a weighted result: it is total cost divided by total sales, weighted by how much of each item sells. Pricing every item to 30% would produce a 30% blended food cost only if you also removed the Q factor, achieved zero variance, and had perfect portion control — none of which pricing affects. In the chapter's five-item extract, individual items range from 20.8% to 31.5% and the blend lands at 28.0%. You do not need every item at target; you need the blend at target, and the blend is a much easier and more forgiving problem.

Paragraph two — the recommendation destroys contribution. On the grain bowl alone, pricing to the target rather than to the market gives away \$5.83 a plate, or **\$18,796 a year. On the ribeye and the whole fish in Exercise 11.24, it would raise prices by \$8 and \$7 respectively, almost certainly collapsing volume on the two best dollar-contributors on the menu. The proposal simultaneously under-charges for cheap dishes and over-charges for expensive ones, which is precisely backwards. The right use of the target is as a screen** — it finds items that are wildly out of line — not as a pricing rule.

Exercise 11.26

(a) Before: \$6.10 ÷ \$23.00 = 26.5%. After: \$7.05 ÷ \$23.00 = 30.7%.

(b) Restore the contribution margin. Original CM = \$23.00 − \$6.10 = \$16.90. New price = \$7.05 + \$16.90 = \$23.95 → in practice **\$24.00. That is a 4.3%** price increase.

(c) Restore the percentage. \$7.05 ÷ 0.2652 = **\$26.58, a 15.6%** price increase.

(d) CM at \$24.00 = **\$16.95. CM at \$26.58 = **\$19.53.

Which to choose: (b). The cost moved 95 cents. Restoring the percentage requires a \$3.58 price move to offset it and hands you \$2.63 a plate more contribution than you had before the increase — which sounds attractive until you notice you are asking the guest for a 15.6% increase to solve a 4.1% cost problem, and that guests compare prices with what they paid last month, not with your ratio.

The general rule: restore the dollars, not the ratio. The percentage will be slightly worse than before, and that is the correct outcome, because your costs are slightly worse than before.

Exercise 11.27

(a) True food cost.

• Period 1 usage: \$16,200 + \$24,800 − \$19,400 = **\$21,600; ÷ \$74,000 = 29.2%**
• Period 2 usage: \$19,400 + \$22,300 − \$16,900 = **\$24,800; ÷ \$76,000 = 32.6%**

(b) Invoices ÷ sales.

• Period 1: \$24,800 ÷ \$74,000 = 33.5%
• Period 2: \$22,300 ÷ \$76,000 = 29.3%

(c) Direction. The shortcut reports an improvement of 4.2 points. The count reports a deterioration of 3.4 points. The two methods do not merely differ in magnitude — they disagree about the sign, and the operator using the shortcut receives a congratulatory signal in the period the business got worse.

(d) The identity.

• Period 1: (\$19,400 − \$16,200) ÷ \$74,000 = \$3,200 ÷ \$74,000 = +4.3 points; 33.5% − 29.2% = +4.3 ✓
• Period 2: (\$16,900 − \$19,400) ÷ \$76,000 = −\$2,500 ÷ \$76,000 = −3.3 points; 29.3% − 32.6% = −3.3 ✓

(e) Combined. Beginning \$16,200 + purchases \$47,100 − ending \$16,900 = **\$46,400 used on \$150,000 of sales = **30.9%** true. Shortcut: \$47,100 ÷ \$150,000 = 31.4%**. The gap is 0.5 points, because the net inventory change across both periods was only \$700 (\$700 ÷ \$150,000 = 0.5 points).

They converge because the error is a change in inventory, and over a longer window the changes partly cancel. That is precisely why an annual figure can look fine while every period inside it was wrong — and why the shortcut fails exactly when a decision is still possible.

Exercise 11.28

(a) Ideal.

Ideal food cost: \$4,961.55 ÷ \$17,546 = 28.3%

(b) Variance. \$5,510.00 − \$4,961.55 = \$548.45**; ÷ \$17,546 = 3.1 points**.

(c) Characterization. Above the two-point threshold — a real problem requiring investigation and action, not monitoring. Note that the actual food cost is \$5,510 ÷ \$17,546 = 31.4%, which on its own might look tolerable against a 30% target. The variance is what reveals that 3.1 of those points are unexplained. Actual alone would not have told you.

Exercise 11.29

What happened: the sales mix moved toward higher-cost items, or menu prices fell (a promotion, a discount), or the cards were correctly updated after a cost increase. All three raise ideal cost legitimately. Because the variance held at 0.6 points, the kitchen is executing to the cards exactly as well as it was before.

What it does not mean: it does not mean the kitchen got worse, and it does not mean the business got worse. A shift toward expensive, high-contribution items can raise food cost percentage and raise profit at the same time — which is the whole of §11.1's second failure mode.

The one number to look at next: total contribution margin dollars for the period. If the dollars rose, this is a good period wearing a bad ratio. If they fell, the mix shift is genuinely hurting you and Chapter 12 is where you go.

Exercise 11.30

Ranked by likelihood × cost to investigate, with the document to pull:

1. Uncosted specials. Eleven daily specials and two menu changes in one period is the single loudest signal here. → The POS item report against the cost-card binder: list every SKU sold that has no card.
2. Comps. The comp report already shows 1.9% against a 0.8% standard — more than double, and comped product consumes inventory while producing no sales. → The comp and void report, costed at plate cost.
3. Spoilage of the forward buy. A full walk-in on a poultry promotion is a classic route to product that ages out. → The waste log, plus a physical look at date labels in the walk-in.
4. Portioning. A new prep cook in week two is a named, dated event that coincides with the period. → Weigh plates and batches against the cards; this is a measurement, not a document.
5. Receiving. Always worth a pass, and cheap. → A week of invoices reconciled against the order guide and the receiving log.
6. Count and math errors. In a period with this many nameable causes it is less likely to be the whole story, but it is the cheapest check of all. → The inventory sheet; recount the two highest-value categories.
7. Theft — last, and only if one through six do not close the gap.

Notice that the four given facts point at causes one through four, and that none of them is theft. That is typical.

Exercise 11.31

34 comped entrées × \$7.90 average plate cost = **\$268.60** of product consumed.

\$268.60 ÷ \$71,000 of food sales = 0.38 points of food cost.

Why comps land in the variance rather than in the ideal: ideal food cost is computed from items sold. A comped entrée generates no sales, so it contributes nothing to the ideal — but the food left the walk-in, so it does appear in actual usage. The entire cost therefore surfaces as unexplained variance.

The fix is a discipline, not a formula: cost your comps weekly at plate cost and carry the figure as a named line in your variance decomposition. Then the variance report says "0.38 points is comps" rather than "0.38 points is a mystery," and you can have an actual conversation about the comp policy. Chapter 31 owns comps and voids on the P&L; Chapter 34 owns the authorization control.

Exercise 11.32

(a) Food cost effect. Bread purchases fall from \$310 to \$88 a week: −\$222 a week.

⚠️ The trap in this problem: the \$0.04 per cover of Q-factor reduction is \$0.04 × 700 = \$28 a week**, and that \$28 is part of the \$222 already counted, expressed per cover rather than per week. Adding it would double-count. The correct food-cost effect is −\$222**, full stop. (A student who adds them and reports −\$250 has made exactly the error that inflates the business case for every in-house-production proposal ever written.)

(b) Labor effect. 10 hours × \$19.00 fully loaded = **+\$190 a week**.

(c) Net prime cost effect. −\$222 + \$190 = −\$32 a week**, or **−\$1,664 a year. Positive, but thin — the proposal saves about thirty dollars a week.

(d) Non-financial factors that could change the decision:

• Whether the ten hours are truly incremental. If they are absorbed into shifts already being paid, the labor cost is near zero and the trade becomes strongly favorable. If they require adding a part-time position with a minimum shift, the real cost exceeds \$190 and the trade goes negative.
• Oven and proofer capacity, and what production the bread displaces during those hours.
• Bread as a brand signal. House-baked bread on the table is one of the cheapest hospitality moves in the business, and its value does not appear anywhere in this arithmetic.
• Consistency and risk. A delivered product arrives the same every day. A house product depends on one person being there.

Exercise 11.33

The arithmetic. With poultry up 15% to \$3.68/lb, a conventionally chilled bird at, say, \$3.30/lb at 3.25 lb gives a half-bird cost of 3.25 × \$3.30 ÷ 2 = **\$5.36. Components become \$5.36 + \$0.95 + \$1.05 + \$0.42 + \$0.18 + \$0.15 = \$8.11; plus 2% = \$0.16; plate cost \$8.27 — below the original \$8.52, at a 28.5%** food cost. The proposal does not merely absorb the increase; it improves the card.

And that is exactly why it is a dilemma. The arithmetic is not the hard part. The arithmetic is seductive.

The guest relationship. Two changes are being made at once, and both are perceptible. The portion falls about 7%, on the restaurant's signature dish, which regulars have ordered repeatedly and can compare against memory. And air-chilled versus water-chilled poultry differs in moisture content and behaves differently on a hearth, so the product itself changes. A guest cannot diagnose either change, but they can register that the dish is not what it was — and per §11.5 and Chapter 23, what they will conclude is not "input costs rose."

The legal exposure. This is the clear part. The menu says "air-chilled." Serving a conventionally chilled bird under that description is a false statement about the product in a consumer transaction, reachable under federal and state consumer-protection law and, in many jurisdictions, under menu-labeling provisions in the health code. Whatever one thinks of the portion question, this half of the proposal is not a judgment call.

What I would do: decline the proposal as written. Then choose openly among the honest options — absorb it, take a modest price increase on the affected items, or change the spec and the menu copy together and tell the floor what changed and why. If the air-chilled claim is not worth defending, remove it from the menu; if it is, pay for it.

What I would tell the staff: exactly what changed, exactly why, and exactly what to say if a guest asks. Servers who know the reason sell through a change; servers who suspect something is being hidden transmit that suspicion to every table.

Exercise 11.34

A strong answer contains, in roughly this order:

1. The number, stated plainly and without blame. "Our ideal-to-actual variance has run above two points for two periods — about \$2,200 a period of food that left the building without being sold." Naming the size of the problem is what makes the response proportionate.
2. What was ruled out first. Say that the counts were re-verified, the invoices reconciled, and the specials costed — so that nobody in the kitchen thinks the first assumption was that they were at fault. This is the paragraph that determines whether the memo lands.
3. The change, concretely. Scales on two stations; portion specs printed and posted at the station, not filed in the office; the first three plates of each service weighed at the pass, out loud, as part of setup.
4. Why a cook should care. Not "food cost." Consistency: the same dish twice, no plate going out heavier than the one before it, no guest at table nine getting less than table four. Cooks care about the plate; write to that.
5. What will be measured, how often, and what happens next. Variance computed weekly and posted where the team can see it. A stated threshold. And an honest statement of what happens if the number does not move — which should be further investigation of other causes, not consequences for the kitchen, because the kitchen is one of six candidate causes.

Mark down any answer whose implicit theory is that people are cheating, and any answer that asks people to "be careful." Portion control is a physical-systems problem; a memo that supplies only exhortation has not supplied a system.

Exercise 11.35

Assessed on five things:

1. Does it foot? Components sum to the stated total; the allowance is computed from the stated total; the plate cost is their sum. Any card that does not foot fails regardless of everything else.
2. Are batch items divided by a measured yield? Look for evidence that a batch was weighed rather than assumed. This is the single most reliable indicator of a card built at a station rather than a desk.
3. Are the small lines present? Fat, salt, oil, garnish. Their absence is worth about 4% (Exercise 11.13).
4. Is the provenance recorded? Each price should carry a source and a date. A card that cannot say where its numbers came from cannot be re-costed and cannot be defended to a lender.
5. Is the price defended rather than derived? The one-sentence defense should name a reason other than the formula — the market, the ladder, the labor content, the item's role on the menu — and should acknowledge the formula's answer so the reader can see how far it was overridden and why.

Exercise 11.36

A complete answer reproduces the checkpoint's structure with the student's own numbers:

$$\text{blended theoretical} + \text{Q factor} + \text{honest variance allowance} = \text{realistic range}$$

For Bellwether that is 28.1% + 1.6 points + 0.5 to 1.0 points = 30.2% to 30.7% against a 30.0% plan line.

The three-line commitment should be specific and testable, not aspirational — for Bellwether: weekly counts from opening week, ideal-versus-actual computed weekly with a 1.0-point action threshold, and every special costed before it is sold.

And the final sentence must contain a real number. For Bellwether: half a point of food cost is \$5,580 a year on \$1,116,000 of food sales — 0.36% of total sales, moving prime cost from 60.0% to 60.4%.

The point of the exercise is the last sentence. A plan that states a target without stating what a miss costs has not made an argument; it has made a wish. Reward students who convert the miss into prime cost, because prime cost is the number a lender and an operator both actually read.

Chapter 12 — Worked Solutions

Solutions to the daggered (†) exercises and the odd-numbered items. Bellwether reference data: 330 units · $8,739.00 entrée sales · $2,447.98 ideal food cost · 28.0% blended · $6,291.02 total CM · weighted CM $19.06 · threshold 14.0%.

Exercise 12.1

Contribution margin is the number of dollars a dish contributes toward everything that is not the food itself — labor, occupancy, other operating cost, and profit. It is menu price minus plate cost.

Exercise 12.3

Star (protect), Plowhorse (re-cost), Puzzle (sell), Dog (investigate — then fix, cut, or keep for a stated reason).

Exercise 12.4 †

Eight items. Expected mix per item = $1 \div 8 = 12.5\%$. Threshold = $0.70 \times 12.5\% = \mathbf{8.75\%}$.

Equivalently, $0.70 \div 8 = 0.0875$. Either route; the second is faster and is what §12.3 uses.

Exercise 12.5

Popularity is a count of guest choices. Measuring mix in sales dollars makes an expensive item look more popular than it is, because price is baked into the number — which smuggles the margin axis into the popularity axis and destroys the independence the whole matrix depends on. Bellwether shows it: the pork chop is second by sales dollars and fourth by units.

Exercise 12.7

The item sells about two-thirds as often as an average item in its category (0.68 of the expected share). Under the conventional 70% rule it falls just below the popularity line — which, if it is that close, means it is unplaced rather than unpopular.

Exercise 12.9 †

Three sentences, with a worked example of the student's own construction. A model answer:

A 20% food cost on a $10 appetizer contributes $8.00; a 35% food cost on a $40 steak contributes $26.00. The percentage says the appetizer is nearly twice as good, and the bank account says the steak is more than three times better. Food cost percentage measures how efficiently you converted product into price, which is a real question — but it deliberately throws away the size of the price tag, and the size of the price tag is what pays your rent.

Exercise 12.11

Both axes break. On the margin axis, an appetizer's contribution margin is structurally lower than an entrée's, so a mixed matrix will place every appetizer below the weighted average and every entrée above it — the quadrants would be reporting the category, not the item. On the popularity axis, guests order roughly one entrée each but appetizers per table, so the unit counts are drawn from different denominators; a $12 appetizer selling 200 times will bury a $31 entrée selling 54, and the threshold (0.70 ÷ n) becomes meaningless when n mixes two different kinds of thing.

Exercise 12.13 †

Round 1: 330 units, total CM $6,291.02, weighted average **$19.06. Burger CM $15.90 → **$3.16 below the line.

Round 2 (trout deleted, no substitution): 285 units, total CM $6,291.02 − $792.00 = $5,499.02. Weighted average = $5{,}499.02 \div 285 = \mathbf{\$19.29}$. Burger CM $15.90 → **$3.39 below** the line.

Nothing about the burger changed — not its price, its plate cost, or its unit count. The average it is measured against rose, because the item that was pulling the average down is gone. This is the relative-axes problem of §12.7 in its smallest form: on the matrix, an item's position is a statement about its neighbors as much as about itself.

Exercise 12.15

Twenty items: threshold = $0.70 \div 20 = \mathbf{3.5\%}$.

It sounds like a lower bar, and it is — but the expected share is only 5.0%, so an item needs just 3.5% of units to clear it. The reason items still fall below constantly is that a twenty-item menu has a long tail by construction: a handful of items take a disproportionate share and the remainder split what is left. On Bellwether's five-item menu every item carries a large share by default (the smallest is 13.6%), so it is genuinely hard to miss a 14.0% line. Broad menus manufacture Dogs; short menus manufacture Stars. Neither fact says anything about the food.

Exercise 12.16 †

Mix percentages:

Popularity indices (expected share = 1 ÷ 5 = 20.0%):

• Hearth Chicken: $29.091 \div 20.0 = 1.45$
• Bellwether burger: $22.121 \div 20.0 = 1.11$
• squash and grains: $18.788 \div 20.0 = 0.94$
• pork chop: $16.364 \div 20.0 = 0.82$
• ember trout: $13.636 \div 20.0 = 0.68$

Sum: $1.45 + 1.11 + 0.94 + 0.82 + 0.68 = \mathbf{5.00}$ — which it always must, on five items, because each index is a share divided by a constant and the shares sum to 1.

Exercise 12.17

(a) $96(20.48) + 73(15.90) + 62(20.05) + 54(20.91) + 45(17.60)$ $= 1{,}966.08 + 1{,}160.70 + 1{,}243.10 + 1{,}129.14 + 792.00 = \mathbf{\$6{,}291.02}$

(b) $\$8{,}739.00 - \$2{,}447.98 = \mathbf{\$6{,}291.02}$

Run both because they fail differently. Route (a) fails if a unit count or a CM is wrong for one item; route (b) fails if the sales or cost totals are wrong in aggregate. Agreement between them means your item-level detail reconciles to your totals — which is exactly the check that catches a missing item, a duplicated line, or a stale price.

Exercise 12.19 †

Half of a 3.5 lb bird = 1.75 lb.

• At $3.20/lb: $\$5.60$. At $3.85/lb: $1.75 \times 3.85 = \mathbf{\$6.74}$ (+$1.14).
• Components: $\$2.75 + \$6.74 = \$9.49$.
• Plus 2% waste: $9.49 \times 1.02 = \mathbf{\$9.68}$ plate cost (was $8.52).
• Food cost: $9.68 \div 29.00 = \mathbf{33.4\%}$ (was 29.4%).
• Contribution margin: $29.00 - 9.68 = \mathbf{\$19.32}$ (was $20.48) — down $1.16.

Annual cost of doing nothing: $96 \times \$1.16 = \$111.36$ a week $\times 52 = \mathbf{\$5{,}791}$ a year.

Note the component check: the Chapter 11 card is chicken $5.60 + all other components $2.75 = $8.35, and $8.35 \times 1.02 = \$8.52$. The card reconciles, which is why we can rebuild it confidently.

Exercise 12.21 †

Gap to the line: $\$20.05 - \$19.06 = \$0.99$. Mix share: 18.8%.

$$d = \frac{\$0.99}{1 - 0.188} = \frac{0.99}{0.812} = \mathbf{\$1.22}$$

Verification. New plate cost $3.95 + 1.22 = \$5.17$; new CM $= 24.00 - 5.17 = \$18.83$. New total CM $= 6{,}291.02 - (62 \times 1.22) = 6{,}291.02 - 75.64 = \$6{,}215.38$. New weighted average $= 6{,}215.38 \div 330 = \$18.8345$.

$\$18.83 < \$18.8345$ — the squash falls below the line by less than half a cent. Star → Plowhorse on a $1.22 movement in one ingredient line.

The reason the divisor is $(1 - \text{mix})$ rather than 1: lowering this item's CM also lowers the menu average, and it lowers it by the item's mix share. The gap therefore closes at only $(1 - 0.188) = 81.2\%$ of each dollar you move. A popular item is harder to knock across the margin line than an unpopular one, because it is a bigger part of the line.

Exercise 12.23 †

Burger from $21.00 to $22.50. New CM $= 22.50 - 5.10 = \mathbf{\$17.40}$.

Current contribution: $73 \times \$15.90 = \$1{,}160.70$. Break-even units: $\$1{,}160.70 \div \$17.40 = \mathbf{66.7} \rightarrow 67$ units.

Volume cushion: $73 - 66.7 = 6.3$ units, or $6.3 \div 73 = \mathbf{8.6\%}$ of burger volume.

If volume holds entirely: $73 \times \$1.50 = \$109.50$ a week = $5,694 a year.

Judgment. An 8.6% cushion is comfortable for a $1.50 move on a $21 item, and the burger is the menu's weakest earner by $3.16, so it is the right item to reprice. Two cautions: do it at a menu change so the whole page moves at once rather than one conspicuous line, and note that the burger is your second most-ordered dish — a great many guests know exactly what it costs. If you are nervous, take $1.00 (a 5.9% cushion, per §12.5) and re-cost the plate as well; the two together are worth about $5,500 with less price visibility.

Exercise 12.25

Model answer. A $4.50 side of hearth-charred greens at a $1.30 plate cost → CM $3.20. At a one-in-four attachment on 73 burgers, that is 18 sides a week: $18 \times \$3.20 = \$57.60$ a week = $2,995 a year, with the burger's own numbers untouched.

The operational reason the rate may be optimistic: the side has to be offered, every time, by a server who is weeded — and attachment rates that depend on a consistent verbal step degrade the moment the floor gets busy, which is exactly when your burger volume peaks. If you want the rate, you build it into the sequence of service (Chapter 22) and measure it, rather than hoping for it.

Exercise 12.26 †

Contribution margins and totals:

Weighted average CM = $\$6{,}704.80 \div 600 = \mathbf{\$11.17}$ Threshold = $0.70 \div 5 = \mathbf{14.0\%}$

Cross-check: sales $= 3{,}456 + 1{,}500 + 1{,}716 + 1{,}976 + 972 = \$9{,}620$; cost $= 950.40 + 360.00 + 631.80 + 600.40 + 372.60 = \$2{,}915.20$; $9{,}620 - 2{,}915.20 = \$6{,}704.80$ ✓

Placement:

The point of this dataset: the fish tacos miss the margin line by seven cents and the grain bowl clears the popularity line by 2.7 points. Both are unplaced under §12.3's step 8. The only item with a genuinely clear position is the soup and half sandwich, which is decisively popular and decisively low-margin — and it is therefore the only item a first-period action should touch.

Exercise 12.27

Model answer, two actions:

1. Re-cost the soup and half sandwich. It is $2.12 below the line and 25.3% of units. Taking $0.35 out of a $3.95 plate — a bread contract, a batch size, a portioning standard — is worth $152 \times \$0.35 = \$53.20$ a week, $2,766 a year, invisible to the guest. Test in four weeks by re-costing the card and confirming the plate cost held.
2. Sell the steak salad. It is the highest CM on the menu at $13.90 and misses popularity by a point. Move it to the panel position and give servers one sentence about it. Twelve more a week is $12 \times \$13.90 = \$166.80$ a week, $8,674 a year, with no price or cost change. Test by pulling units in four weeks and comparing to 78.

Do not touch the fish tacos this period — seven cents below the line is not a finding.

Exercise 12.29 †

What is happening: entrées are being rung through the open-food or miscellaneous button, almost certainly because a modifier was awkward, a special had no button, or the line was moving and the server took the fastest path. Seven percent of a twelve-item menu's units is more than half an average item's worth of volume, sitting in a bucket with no plate cost attached.

What it does to the other eleven items: every real item's mix share is understated, because the denominator (total category units) includes 7% of units that belong to some item and are credited to none. The bias is not uniform — it falls hardest on whichever items are most often rung through the open button, which are typically the ones with complicated modifiers or the ones that ran as specials. Items near the popularity threshold can be pushed below it by this alone.

Two things before running the matrix:

1. Find out what was in it. Pull the open-food line detail — most systems retain the typed description and the amount rung. Reallocate what you can identify to the correct items and rerun.
2. Close the leak. Build buttons for the recurring cases (specials, half portions, the common substitutions) and restrict open-food to manager authorization. Chapter 34 treats this as a control question, which it also is: an unrestricted open-price button is an audit-trail problem as well as a data problem.

Exercise 12.31

Explanation one, about the menu: the category has very few items with an unusually even spread. On a five-item menu the threshold is 14.0% against a 20.0% expected share, so an item has to sell less than seven-tenths as often as average to miss — genuinely possible to avoid. This is a real and unremarkable finding, and Bellwether nearly produces it (four of five clear the line).

Explanation two, about the analysis: the category is too broad, so the item count is high, the threshold ($0.70 \div n$) is tiny, and everything clears a bar set almost at the floor. On a forty-item "all food" category the threshold is 1.75%, which nearly nothing misses.

How to tell them apart: look at the item count and the spread of the shares. Few items with shares in a narrow band is explanation one. Many items with a long tail of small shares, all of which somehow "pass," is explanation two — and the fix is to split the category into things guests actually choose between.

Exercise 12.33

Model memo structure (the graded content is the arithmetic and the trigger, not the prose):

Subject: The ember trout stays through the spring menu.

The trout runs a 37.1% food cost, which is the worst number on our menu, and I understand why it looks like the obvious cut. Here is why it is not.

At $28.00 against a $10.40 plate, the trout contributes $17.60 — a dollar seventy more per plate than the burger, which runs 24.3%. It sells 45 a week, which is $792 a week and $41,184 a year of contribution margin. It misses our popularity line by 1.2 units a week, and the sampling noise on a 45-unit item over a full season is wider than that gap, so the data cannot actually place it.

It is also the only fish on the menu, and I do not know how many tables chose us because of it. That is the risk I am accepting: if the answer is "almost none," we are leaving margin on the table by keeping it.

The plan. We re-cost it before the spring menu — a yield test on whole fish against portioned fillets, which I expect to take about $1.30 out of the plate — and we work the description and the pre-shift for eight weeks. Trigger: at the end of period four, if it is still under 45 units a week after a real selling effort, it comes off the spring menu and we replace it with a fish that costs less. I will bring the numbers to that meeting either way.

Exercise 12.35

The agenda from §12.8, in order: (1) are the cost cards current — if two or more are stale, the meeting becomes a costing meeting and reconvenes; (2) what moved since last period and by how much; (3) which items are within two points of a line, named out loud as unplaced; (4) two changes, each with an expected dollar value and an owner; (5) read back last period's two changes against what was expected.

Attendees: the chef, the person who runs the floor, and whoever owns the numbers — three people, 45 minutes. The item that ends the meeting early is (1).

Exercise 12.37 †

(a) As arithmetic. Suppose the six-ounce protein is $2.40 of a $7.80 plate. Cutting to five ounces saves $0.40, which on a 22% mix item at Bellwether's volume (73 units) is $29.20 a week, $1,518 a year. That is real but modest — less than a $1.00 price increase (§12.5, $3,796) and less than the re-cost plus attachment combination.

(b) As guest experience. This is your most-ordered item. A great many guests know exactly what it looks like, and portion changes on familiar dishes are noticed and discussed in a way that price changes are not — a price change is a fact, a smaller portion feels like something done to you. Chapter 23's second-visit economics are the relevant number, and they are larger than $1,518.

(c) As honesty. The dividing line is the representation. If the menu says "6 oz," changing the portion without changing the menu is a false statement about the product, full stop, and in many jurisdictions a consumer-protection matter as well as an ethical one. If the menu makes no weight claim, the portion is not a representation and reducing it is a legitimate operating decision — but "not a lie" is a low bar for your most-ordered dish, and the disciplined test is whether you would say it out loud: "we made the burger smaller to save forty cents." If that sentence is uncomfortable, the move is probably wrong, or it is right and belongs alongside a change the guest gets something from.

Where the line is: it is not at re-portioning, which is sometimes correct — a portion that drifted upward from the standard should be corrected, and §12.5's order of operations puts re-portioning fifth rather than last for exactly that reason. The line is at re-portioning silently, against a stated claim, on an item guests know, and at reaching for it before the two invisible levers (re-cost, attach) have been tried.

Exercise 12.39

You may not present an undisclosed-threshold label as a finding. The obligation has three parts: find out what threshold the software used (ask the vendor; it is often 70% but sometimes the category average); recompute at least one alternative threshold yourself and see whether the labels survive it; and if you cannot establish the threshold, present the underlying numbers — CM, mix, distance from each line — and no labels at all.

The reason this is an obligation rather than a nicety: §12.2 showed that moving the threshold from 70% to 100% of expected share changes the recommendation for two of Bellwether's five items. A label derived from an unknown parameter is not information, and presenting it to a chef whose dish is about to be cut — or to a partner deciding where to spend — borrows an authority the analysis does not have.

Exercise 12.41

The projected entrée mix produces a 28.0% ideal food cost against the plan's 30% target, which on $1,116,000 of year-one food sales would be worth roughly **$22,320. The plan does not spend it: the 28.0% is ideal cost on a projected mix, and the gap between ideal and actual — waste, comps, over-portioning, and the mix simply landing somewhere else — is exactly what the two points are reserved against.

(Check: $1{,}550{,}000 \times 0.72 = \$1{,}116{,}000$; $\times 0.02 = \$22{,}320$.)

Chapter 13 — Worked Solutions

Solutions to the daggered (†) and odd-numbered exercises. Arithmetic is shown in full; qualitative items get a model answer and the standard a grader should hold students to.

Reference figures: food sales \$1,116,000/yr (\$21,462/wk avg) · food order guide \$6,438/wk (\$6,438 × 52 = \$334,776, which is the \$334,800 annual budget less rounding) · Hearth Chicken frozen card \$8.35 components / \$8.52 plate / \$29.00 price / 29.4% / \$20.48 CM.

Exercise 13.1

Broadline — one supplier, very wide catalog, one truck, one invoice, credit terms. Bellwether's right channel for dry goods, oil, vinegar, spice, chemicals, and paper: undifferentiated product where what you are buying is consolidation.

Specialty — deep in one category, smaller, often better, usually dearer per unit. The right channel for the poultry, the trout, and the beef, because these are the lines the menu is built on and because a specialist can hold a spec (air-chilled, giblets in, 3.25–3.75 lb) that a broadliner's catalog description cannot.

Direct — from the producer. The right channel seasonally for the root box and the herb box: it is where the menu's sourcing claims come from, and the operational cost (irregular delivery, no substitutions, no credit terms) is acceptable on items with a short list of suppliers and a story attached.

Standard: the student must justify each choice by the property of the item, not by price.

Exercise 13.3 †

Which to drop with fifteen minutes: check 6, price. Not because it matters least, but because it is the only one of the seven that can be performed accurately after the truck leaves. The invoice stays on your desk; the driver, the pallet, and your leverage do not. Temperature, count, weight, and spec are all destroyed the moment the door closes. A student who answers "temperature" or "count" has inverted the logic and should be pushed to say what evidence still exists at 9:00 a.m.

Exercise 13.5

A credit memo is the supplier's written acknowledgment that they owe you money or product for something short, rejected, or out of spec. The four rules: (1) get it at the door, in writing, signed by the driver; (2) keep it physically with the invoice so the pair travels to the bookkeeper; (3) verify it actually appears on the next statement; (4) log it — date, vendor, amount, reason, applied Y/N.

Most often skipped: rule 3. And it is the most expensive omission, precisely because it is invisible: you already decided you were owed the money, so nobody goes looking for it. A credit issued and never applied is a loss with no symptom. The credit log in rule 4 exists to make rule 3 mechanical rather than dependent on somebody remembering.

Exercise 13.7

Because a count sheet in shelf order matching a fixed count route is walked once, while an alphabetical sheet is walked many times.

Cost one — time. A single-pass count of Bellwether's 54 lines takes about 75 minutes. An alphabetical sheet in a five-section walk-in plus dry store plus reach-ins turns that into a back-and-forth of well over two hours, most of it walking.

Cost two — accuracy, which is the expensive one. On the eighth pass, attention fails and lines get skipped. A skipped line understates ending inventory; understated ending inventory overstates usage (beginning + purchases − ending); overstated usage inflates food cost and manufactures a variance that does not exist. The manager then spends a week investigating a counting artifact — and, worse, may "find" a cause and act on it.

Exercise 13.9 †

Tuesday par. The Tuesday delivery covers Tuesday, Wednesday, and Thursday dinner. Forecast Hearth Chickens: 12 + 15 + 18 = 45 halves. Safety stock 3 halves — one cushion plate per service, on the three quiet nights. Par = 48 halves = 24 whole birds.

Friday par. The Friday delivery covers Friday and Saturday dinner. Forecast: 24 + 27 = 51 halves. Safety stock 5 halves, because an 86 on the signature dish at 8:40 on a Saturday costs far more than one at 7:15 on a Wednesday. Par = 56 halves = 28 whole birds.

Total: 24 + 28 = 52 birds, exactly Chapter 10's program. The program is unchanged; only its distribution across the week has changed.

The general rule. Chapter 10 was sizing a program — answering "how much chicken does this menu consume in a week?" — which is a question about totals and is correctly answered with an average. Chapter 13 is writing an order — answering "how much do I need on this shelf on this morning?" — which is a question about a distribution, and a distribution cannot be answered with an average. The week is not five identical nights: Saturday is more than twice Tuesday. Averages size programs; forecasts write orders.

Exercise 13.11

The three costs a turnover number cannot see:

1. 86s and their revenue consequences. Running out of an item is invisible on an inventory report and highly visible to a guest. Four days on hand in a restaurant with two protein deliveries a week means a single missed truck empties a station.
2. Emergency replacement at retail. Cash-and-carry runs cost the price difference plus a labor hour plus mileage, and none of the three lands on the inventory report.
3. Lost supplier standing. Ordering below minimum drops triggers small-order fees, forfeits volume commitments, and — less measurably — moves you down the priority list when supply is short.

Evidence to ask for: the 86 log for the last eight weeks; the cash-and-carry line as a separate account; the count of invoices per week and average invoice value (a rising invoice count with a falling average value is a restaurant ordering in panic); and the small-order and expedite fees on the last three statements. If all four are clean, the operator is right and should be congratulated.

Exercise 13.12 †

The mechanism. Safety stock is inventory held against a forecast being wrong. Its cost is the probability that it is never sold multiplied by what it cost. On a line with one outlet, unsold safety stock has one chance to be sold before its shelf life expires, so the expected cost is high and the rational cushion is thin — you accept more stockouts to avoid more waste. On a cross-utilized line, the same physical product has three or four more chances at a different price point, in a different daypart, in a different preparation. The expected cost of holding it falls, and therefore the optimal cushion rises. It is not that mistakes are cheaper to forgive; it is that the item is genuinely less likely to be wasted, so the arithmetic itself changes.

Bellwether's poultry is the clean case: eight spare halves at Saturday close are not a forecasting error to be regretted but roughly \$288 of Sunday hash revenue that Chapter 10 put on the menu on purpose.

Thinnest cushions: the trout program and the herb-and-caper box.

• Trout. Whole fish has the shortest usable life on the guide, and although it reaches three menu items, all three are trout — an unsold fish has no non-trout home, and Chapter 10 explicitly discards the frames. Two fish of cushion, and no more.
• Herb-and-caper box. Fresh herbs have a two-to-three-day usable window. The salsa verde does reach four menu items, but the sauce is the cross-utilized thing, not the herbs; the sauce is made twice a week to order, which puts the cushion in the delivery rhythm rather than on the shelf. Par is one box per delivery, with no safety stock.

Exercise 13.13

No, it is not good news. A waste log with zero entries for four of six weeks is not a kitchen with no waste; it is a log nobody is filling in. Every real kitchen produces trim, spoilage, drops, and re-fires every single week.

The stable food cost makes it more worrying, not less, because there are two ways to produce a stable number: real control, or a measurement that is not sensitive to anything.

What to check first, in order: (1) stand at the bin during one service and watch — that answers it in twenty minutes; (2) confirm the log has a pen, a clipboard, unit costs pre-printed, and a named owner per shift; (3) pull the ideal-versus-actual variance (Chapter 11) for the same six weeks — if variance is normal while the log reads zero, the waste is real and unrecorded; (4) check whether ending inventory is being counted consistently, because a stable food cost across six weeks can also be manufactured by a stable counting error.

Exercise 13.15

Receiving is the only moment because after it, every fact is gone: the pallet is broken down, the product is commingled, the driver has left, and all that remains is the supplier's own document.

Two other moments with the same structure:

• Signing the lease or the LOI (Chapter 6). Total control before signature — every clause is negotiable and costs nothing to change — and essentially none for ten years afterward.
• The offer and the first shift (Chapter 17, Chapter 21). Standards set on day one are free; standards imposed in month seven are a confrontation.

(A third defensible answer: the last ninety seconds of a guest's visit — Chapter 23 — where recovery is possible until they reach the sidewalk and impossible after.)

What they have in common: a short, reversible window of total control followed by a long irreversible period, in which the correct behavior is cheap, unglamorous, and must be performed by somebody under time pressure who would rather be doing something else. That is precisely why all three require a written standard rather than good judgment — good judgment is exactly what fails at 6:52 in the morning.

Exercise 13.16 †

(a) 112.00 lb ÷ 28 birds = 4.00 lb average. The spec is 3.25–3.75 lb, so the delivery is out of spec, over the ceiling by a quarter pound a bird.

(b) 112.00 lb × \$3.20 = **\$358.40.**

(c) 28 × 3.50 lb = 98.00 lb × \$3.20 = **\$313.60.**

(d) \$358.40 − \$313.60 = \$44.80 credit requested at the door, with the exception written on both copies and the driver's signature.

(e) A 4.00 lb bird halves to 2.00 lb. Chicken component = 2.00 × \$3.20 = **\$6.40.**

(\$9.15 × 1.02 = \$9.333 → \$9.33; \$9.33 ÷ \$29.00 = 32.17%; \$29.00 − \$9.33 = \$19.67.)

(f) \$9.33 − \$8.52 = \$0.81 per plate.** × 4,992 plates (96/week × 52) = **\$4,043.52 a year, which is 0.36 points of food cost on \$1,116,000 of food sales.

Credit for the best answer: a student who also notes that the correct operational response is to portion the halves back to 1.75 lb and route the 0.25 lb × 56 = 14.0 lb of trim (\$44.80 of product) to the hash and the stock has understood the chapter. The trim recovery happens to equal the credit requested — which is worth pointing out in class.

Exercise 13.17

At \$3.20/lb.** 52 birds × 3.5 lb = 182 lb × \$3.20 = \$582.40/week**; × 52 = **\$30,284.80/year. As a share of the \$6,438 weekly guide: **9.0%.** As a share of \$1,116,000 of food sales: 2.71%.**

At \$3.55/lb.** 182 lb × \$3.55 = \$646.10/week**; × 52 = **\$33,597.20/year. New weekly guide = \$6,438 − \$582.40 + \$646.10 = **\$6,501.70, so poultry is now 9.9% of the guide and 3.01% of food sales.

Annual increase: \$33,597.20 − \$30,284.80 = \$3,312.40, which is 0.30 points of food cost.

On the plate. Half a bird = 1.75 lb × \$3.55 = \$6.21. Components \$8.96; plate cost \$8.96 × 1.02 = \$9.14**; food cost **31.5%**; contribution margin **\$19.86 (down \$0.62 from \$20.48). Over the 96 chickens sold weekly, \$0.62 × 4,992 = **\$3,095.04/year** — the difference from the \$3,312.40 above is the eight hash halves a week, which cost more too but are not sold as Hearth Chicken. Both numbers are correct for what they measure; students should be able to say why they differ.

What to do about the menu price — including the case for doing nothing. \$29.00 is a whole-number price on the menu's signature Star. Recovering \$0.62 requires roughly \$2 of price (\$31 at a 30% target), which is a 6.9% increase on the item guests most associate with the restaurant, to solve a problem worth 0.3 points of total food cost. The stronger answers, in order: (1) absorb it and hold the price — the item still carries the largest contribution margin on the menu at \$19.86, and Chapter 12's rule is that you protect a Star; (2) recover it elsewhere in the mix, where a \$1 move on a plowhorse is invisible; (3) re-negotiate or re-spec — a slightly smaller bird within the existing range moves the plate more than a price change would; (4) re-cost at contract renewal, which is what Chapter 11's frozen card already instructs. Raising the signature dish's price is the last option, not the first.

Exercise 13.18 †

Walk-in inventory = \$6,025 × 70% ≈ **\$4,200.**

(a) \$4,200 × 2% = **\$84 a week.**

(b) \$84 × 11 weeks = **\$924.**

(c) \$84 × 52 = **\$4,368 a year; ÷ \$1,116,000 = 0.39 points of food cost.**

(d) \$400 ÷ \$84 = 4.76 weeks — payback in under five weeks. First-year return: \$4,368 of avoided loss on \$400 spent = **10.9×**, or \$3,968 net.

Push students on the limit: the 2% is an assumption, not a measurement, and nobody can tell you the real figure for a specific restaurant without a waste log and a variance. What is not an assumption is the direction, the order of magnitude, and the fact that the repair pays back inside two months.

Exercise 13.19

(a) \$142.60 + \$310.15 + \$88.40 + \$61.85 = \$603.00.** ÷ \$34,900 = 1.73% of food sales.**

(b) \$603.00 × 52 = **\$31,356 a year.**

(c) Ranking by expected recovery, not by size:

1. OVERPREP (\$310.15). Largest and the most systematically fixable: it is a production-planning failure with a known method (prep to the covers forecast, batch sizes written on the card). Expect to recover half or more within a month.
2. SPOIL (\$142.60). Second, and diagnostic — spoilage is a par problem, a rotation problem, or a temperature problem, and the three have different fixes. Recovery is high once you know which.
3. TRIM (\$88.40). Third. Real, but the fix is training and yield testing (Chapter 11), which is slower and partly irreducible — some trim is the price of buying whole product.
4. REFIRE/DROP (\$61.85). Last on cost, and it is arguably not a cost problem at all. Eleven entries is a service-pressure signal (Chapter 14) that happens to have a dollar value attached; attacking it as a cost problem will produce a cook who hides a dropped plate rather than one who drops fewer.

Standard: the ranking must be defended on expected recovery per hour of management attention, not on the size of the number. A student who simply sorts descending has not answered the question.

Exercise 13.20 †

(a) Net purchases = \$6,905 − \$185 = \$6,720. Food used = 6,340 + 6,720 − 110 + 25 − 5,780 = \$7,195.

(b) \$7,195 ÷ \$22,410 = 32.1%.

(c) Average inventory = (\$6,340 + \$5,780) ÷ 2 = \$6,060. Weekly turnover = \$7,195 ÷ \$6,060 = 1.19 turns. Annualized = 1.19 × 52 = about 62 turns a year. Days on hand = 7 ÷ 1.187 = 5.9 days.

(d) Gross invoices ÷ sales = \$6,905 ÷ \$22,410 = 30.8%. Net purchases ÷ sales = \$6,720 ÷ \$22,410 = 30.0%.

Both understate, and the reason is arithmetic rather than opinion. Inventory fell \$560 this week (6,340 → 5,780), which means the restaurant used \$560 more product than it bought; \$560 ÷ \$22,410 = 2.50 points. Working the other way, net transfers out to the bar of \$85 removed product from food usage: \$85 ÷ \$22,410 = 0.38 points. Net: 2.50 − 0.38 = 2.12 points, which is exactly the gap between 32.11% and 29.99%.

The transferable point: the shortcut error is not a constant bias you can mentally adjust for. It is the change in inventory divided by sales, and it changes sign whenever your buying rhythm changes.

Exercise 13.21

Perishable lines \$4,600/week × 10% = **\$460 a week of over-order. Of that, 15% ages out: \$460 × 0.15 = **\$69 a week = \$3,588 a year, or 0.32 points of food cost.

Cash tied up: a persistent standing over-order raises the steady-state average inventory by roughly one cycle's excess — call it \$460 permanently on the shelf.

Why the two are different kinds of loss, for a skeptical chef-owner:

The \$3,588 is a P&L loss. It happens again next year, and the year after. It is gone. Nothing recovers it.

The \$460 is a balance-sheet transfer, not a loss. You still own it — it is sitting in the walk-in in the form of product, and it comes back the moment you stop over-ordering. What it costs is narrower: the return you could have earned on it elsewhere, and — the part that actually matters for a restaurant — the fact that it is not available in the week you need it. Chapter 33's whole argument is that a business can be profitable and still fail because money was in the wrong form on the wrong Thursday. Inventory is cash you decided to store as food, and food does not make payroll.

The reason to separate them: the \$3,588 justifies a policy change on its own; the \$460 justifies one only if the operation is cash-constrained. Conflating them overstates the case and gets the argument dismissed.

Exercise 13.22 †

Step 1 — entrées. Covers 58 / 70 / 95 / 126 / 141 = 490. At 0.90 attachment: 52 / 63 / 86 / 113 / 127 = 441 entrées (490 × 0.90 = 441 ✓; Thursday's 85.5 rounded up).

Step 2 — Hearth Chickens at 22.4% of entrées:

(11.65→12, 14.11→14, 19.26→19, 25.31→25, 28.45→28; rounded to whole halves.)

Step 3 — Tuesday par. Covers Tue/Wed/Thu = 12 + 14 + 19 = 45 halves, + 3 safety = 48 halves = 24 birds.

Step 4 — Friday par. Covers Fri/Sat = 25 + 28 = 53 halves, + 5 safety = 58 halves = 29 birds.

Step 5 — weekly order: 24 + 29 = 53 birds. Cost: 53 × 3.5 lb = 185.5 lb × \$3.20 = **\$593.60 a week.**

Step 6 — halves for Sunday hash. Trace it: Tuesday 0 on hand, +48 = 48; usage 45 → Thursday close 3. Friday 3 on hand, +58 = 61; usage 53 → Saturday close 8. Eight halves to the hash, and Tuesday opens at 0 again — the week closes.

Check: 98 sold + 8 to hash = 106 halves = 53 birds = the weekly order. ✓

Rounding stated: entrées rounded to whole (Thursday up); chicken counts rounded to whole halves; the Friday par of 58 halves happens to be exactly 29 birds, so no rounding there. Note that the Tuesday par is unchanged at 24 despite a busier week — the three quiet nights barely moved — while Friday went from 28 to 29. The par is sensitive to exactly the nights that carry the volume, which is the point of building it per-service instead of per-week.

Exercise 13.23

Model spec — Trout, whole [T-01]:

The two harder fields, and what you must learn first:

1. Size / weight range. For chicken, 3.5 lb was handed to us by an existing cost card. For trout you cannot write a range honestly until you have run a yield test (Chapter 11): how much boneless fillet a 1.25 lb fish gives versus a 1.50 lb fish, and where the plate portion lands. A range set before the yield test is a guess that will show up as an unexplained plate cost.
2. Age at receipt / freshness criteria. "Fresh" is not a spec. You have to learn the actual chain — harvest day, days to the distributor, delivery days to you — before you can write a number. Ask the supplier to walk the chain, then write the number their own chain can actually meet, or you have written a spec that guarantees a weekly argument.

(A student who also flags farmed vs. wild as a menu-claim question, and country of origin as a sourcing-claim question that Chapter 38 governs, has read carefully.)

Exercise 13.25 †

Weekly spend on the twelve: roughly \$582 + \$415 + \$535 + \$290 + \$470 + \$215 + \$230 + \$95 + \$245 + \$79 + \$185 + \$215 = about \$3,556**, which is **55% of the \$6,438 guide in twelve lines out of fifty-four.

Time estimate: about twelve minutes, and the defense is the design rather than the speed. Each line is a single count in one known location; the route is fixed (A → B → dry store → E) and runs in one direction; no extensions or pricing happen at the shelf — the sheet carries counts only, and the arithmetic is done afterward at a desk or by the software. Two people make it faster, not slower, because nobody puts down a case to write.

Accept 10–15 minutes. Reject any answer above 25 minutes without a stated reason, and ask what about their route is making it slow — that question is the real lesson.

Exercise 13.27

Model allocation (students' splits will differ; what is graded is the reasoning):

The commit bucket is close to the chapter's ~\$185,000 estimate of undifferentiated spend; a student who lands between roughly \$150,000 and \$200,000 has reasoned correctly.

The two opening sentences to a broadline salesperson:

"We will do about a hundred and sixty-five thousand dollars a year with one distributor on dry goods, dairy, eggs, flour, coffee, pork, and commodity produce — and I can commit to that in writing for twelve months. The proteins and the specialty produce are not on the table, so let's not spend time on them."

Standard: the student must (a) quantify the commitment before asking for anything, (b) exclude the identity lines explicitly rather than leaving them ambiguous, and (c) not open with a request for a discount. Offering to commit spend you were never going to give is how a small account loses credibility in the first ten minutes.

Exercise 13.28 †

(a) Usage = \$11,400 + \$41,850 − \$14,900 = **\$38,350. True food cost = \$38,350 ÷ \$126,000 = 30.4%.**

(b) The owner computed \$41,850 ÷ \$126,000 = 33.2% — invoices over sales. Inventory rose \$3,500** during the period, and \$3,500 ÷ \$126,000 = 2.78 points, which is exactly the gap (33.21% − 30.44% = 2.77%). The restaurant did not use that \$3,500; it still owns it, sitting on a shelf. The chef is right and the owner is looking at the wrong number.**

(c) Most likely causes of a \$3,500 inventory build, ranked:

1. Timing — one or more large deliveries landing in the last days of the period, before the count. The commonest cause by a wide margin and completely benign.
2. Deliberate buy-ahead — a known price increase, a contract change, a holiday week, or a supplier holiday schedule. Also benign, and someone in the building knows about it.
3. Par drift / over-ordering — the genuinely concerning one, and the only one that recurs. Test it by looking at whether inventory rose in the previous period too.

(d) What to ask for before doing anything at all: the count sheets for both period ends; the date and time of each count; the invoice register for the final five days of the period; and confirmation that prepped inventory was counted at both ends by the same method. Three of those four could individually explain the whole thing. Do not have a conversation with the chef until you have all four — an accusation built on the wrong number is expensive in a currency you cannot recover.

Exercise 13.29

The two most likely explanations, neither about food:

1. Inconsistent counting. A storage location or a category — prepped inventory in the walk-in, the dry-store floor, the reach-in behind the line — counted in periods 1 and 3 and missed in period 2. Missing product understates ending inventory, which overstates usage, which is exactly a one-period spike that reverses.
2. Count timing relative to deliveries. A count taken after the truck in one period and before it in the others moves an entire delivery's value across the period boundary — and then moves it back the next period, which is why the pattern is high-then-normal rather than a step change.

The single test that distinguishes them: put the count timestamps next to the invoice dates for all three period ends. If period 2's count is on the wrong side of a delivery, it is a timing problem and you can prove it in ten minutes. If the timestamps are clean, recount the same section twice with two different people and compare — a stable recount rules out omission and pushes you toward a real usage event.

A third acceptable answer: credit memos applied in a period other than the invoice's. Worth mentioning; not worth ranking above the two above.

Exercise 13.31

The sequence, with time and what each proves:

Total: under three hours, and it can be done in one day by one person.

When suspicion of a person becomes reasonable: only after all seven come back clean, and the variance persists across at least two periods, and it is concentrated in specific high-value, portable, easily-resold items rather than spread across the guide. At that point it becomes Chapter 34's subject — controls, separation of duties, surprise counts, and a process that is fair to the people it examines. Note that Chapter 11 put theft seventh in the investigation order for a reason that is practical as much as ethical: an operator who starts there stops looking, and usually stops looking about four steps before the actual answer.

Exercise 13.33 †

Model answer — Bellwether Receiving Policy (staff handbook, one page):

Who. The receiver named on the schedule for that morning. On days without a named receiver, the sous. The person who receives is never the person who placed the order. If neither is present, the delivery waits or is refused — do not accept a delivery you cannot check.

When. 6:30–8:30 a.m. only. No deliveries during service. If a driver arrives outside the window, call the manager before signing anything.

What you do, in order, every time: 1. Temperature. Probe the cold load before it leaves the truck. Refrigerated 41°F or below; frozen solid with no sign of thaw and refreeze. Out of temperature = reject that line. There is no price at which we take it. 2. Count. Open every case. Count the pieces. Compare to the invoice and to the order guide. 3. Weigh. Every catch-weight line on the scale — poultry, whole fish, primals, cheese. 4. Spec. Check the spec sheet for any line with a spec number on it. Grade, size, pack, fabrication, pack date. 5. Price. Compare each line to the quoted or contract price. 6. Sign. Write every exception on both copies. The driver signs. You keep your copy.

What you may reject on your own authority, without calling anyone: anything out of temperature; anything damaged, leaking, or with a compromised package; anything short; anything past its pack-date window; any product not on the order.

What you must escalate before deciding: an unapproved substitution; product outside a spec weight or size range; a price that does not match the quote by more than a few cents; a delivery outside the window. Call the chef. If the chef is unreachable, accept under protest with the exception written and signed, and put it on the chef's desk.

Paperwork. Invoice plus any credit memo, stapled, in the invoice tray by the end of your shift. Every exception goes on the receiving log with the date, vendor, line, and dollar amount. Log the walk-in temperature while you are putting the delivery away.

Then put it away. Label and date everything. New product goes behind old product. Chicken on the bottom shelf of section E, always.

Standard: the policy must be usable on a second shift, must give explicit reject-without-calling authority (or the whole thing collapses into "ask the chef," which fails at 6:52 a.m.), and must say where the paper goes. A policy that lists checks but not authority has failed the assignment.

Exercise 13.35

What you do: open the cases. You do not sign blind.

What you say — keep it about the process, never about the person: "I sign for what I've counted. It's not you — it's the rule here for every vendor. Give me four minutes and I'll have you out." Then be fast, because the second half of that sentence is the part that keeps the relationship.

If you genuinely cannot check it — a second truck is waiting, or the delivery is enormous — sign with a written notation on both copies: "Subject to count. Cases not opened. Exceptions to follow by 10:00 a.m." Photograph the pallet before it is broken down, and then actually count it within the hour and actually call. A conditional signature that is never followed up is worse than no condition, because it teaches the route that your conditions are noise.

What it costs if you are wrong in either direction:

• Sign blind and be short. You eat it — perhaps \$20, perhaps \$200 — and, far more expensively, you have taught the route that this stop does not check. The next short delivery is more likely, not less. This is a reputational fact about your dock, and it travels.
• Refuse rudely and delay a driver on a tight route. You become the stop that gets serviced last, the one where substitutions get made without a phone call, and the one that does not get called when something is short before the truck loads. Driver goodwill is a real asset with real cash value.

The disciplined answer preserves both, and it does so by being fast and impersonal. The operator who gets this wrong is almost always the one who makes it about trust.

Exercise 13.37

What you do not do first: confront the colleague; search anyone or anyone's bag; announce it at pre-shift; repeat the accusation to other staff; or treat the reporting cook as a witness in an investigation. Every one of those converts a fixable system problem into a personnel crisis, and three of them may create legal exposure (Chapter 20).

What you do:

1. Thank the person, protect them, and commit to nothing. "Thanks for telling me. I'll look at it. This conversation stays with me." Do not promise an outcome.
2. Check whether a policy exists and whether it was ever trained. "Everybody does it" almost always means nobody was ever told — that end-of-shift product is not a benefit, what family meal covers, and what happens to Saturday's surplus. If there is no written policy, the failure is management's and you should say so internally.
3. Verify with data, not accusation. Key-item counts on the affected proteins for two weeks; closing procedures; whether the product is leaving as unlogged "waste"; whether the waste log shows a Saturday pattern; whether the variance for those weeks is even abnormal. It may not be.
4. Fix the system first. Write the policy, train it, put it in the handbook, state the consequence, and change the physical conditions — an observed exit route, a locked walk-in after close, receiving and ordering separated, a named closer.
5. Then, and only then, address individuals — consistently, documented, and applied to everyone the same way, per Chapter 21's progressive-discipline framework.

Where this chapter ends and Chapter 34 begins: at the boundary between designing a system and investigating a person. Everything in steps 1–4 is Chapter 13 and Chapter 21 — pars, counts, waste logs, storage, policy, training. The moment you begin building a case about an individual, you are in Chapter 34, and Chapter 34's rules on evidence, fairness, documentation, and when law enforcement is appropriate govern from there. Do not improvise across that line.

Exercise 13.39

The preventive-maintenance schedule:

What it protects. On §13.5's arithmetic, a walk-in drifting from 38°F to 46°F costs \$84 a week, \$4,368 a year** in accelerated spoilage alone. That is a **3.6× return** on the \$1,220 before counting two larger and lumpier risks: an emergency after-hours service call (typically several hundred dollars minimum, and it always happens on a Saturday), and a compressor failing overnight with roughly \$4,200 of product in the box** — a single event that exceeds four years of the entire PM budget.

Where it belongs in the pro forma. Repairs and maintenance, inside Other Operating (14.0% of sales). \$1,220 on \$1,550,000 is 0.08% of sales — small enough that it is routinely cut, which is precisely why it should be a scheduled line rather than a discretionary one. Chapter 31 builds the chart of accounts that gives it a home.

Exercise 13.40 †

(a) Why opening food inventory is deliberately above the 6.4-day steady state. Because a par level is derived from a forecast and there is no forecast yet — only a plan. Three specific reasons: opening pars are guesses that will be wrong item by item even if the total is right; the opening mix is unknown (Chapter 12's menu-mix data does not exist until the POS has run for weeks); and opening-period volume is unusually volatile and unusually consequential, because the guests forming their impression during the honeymoon are the ones who decide whether there is a second visit. \$9,200 is roughly ten days at plan against a 6.4-day steady state — the extra three and a half days are the price of not knowing.

(b) The case for \$7,500.** It frees **\$1,700 of pre-opening cash into a \$35,000 budget that also has to carry pre-opening labor, training, and licensing — and Chapter 1's first killer is undercapitalization, not under-stocking. Operationally it costs: more cash-and-carry runs in weeks one to three (each one a labor hour, mileage, and a retail price premium); a higher probability of an 86 during the period when a first impression is being formed; and a chef spending attention on procurement during the two weeks when the line most needs them. Those costs are real but they are also recoverable and short-lived, which is the strongest form of the argument.

(c) The case for \$11,000. More cushion during the exact period when the cushion has the most value, and a lower probability of running out of the signature dish in front of a first-time guest. The cost is \$1,800 more of the scarcest money in the whole project, plus a genuine spoilage risk: you are holding more of items whose real velocity nobody knows yet, so a share of the extra will age out before the first accurate par is written.

(d) Recommendation: hold \$9,200, but hold it unevenly. Weight the extra days toward items that are (i) hard to replace quickly, (ii) slow to spoil, or (iii) on the critical path of the signature dishes — dry goods, flour, the root box, portioned proteins that tolerate a freezer. Do not carry ten days of greens, herbs, or whole fish, which will simply be thrown away.

What would change my mind: the actual quoted delivery frequency and minimum drops from the chosen suppliers. If the produce house will deliver three times a week with no minimum, the produce share of opening inventory should be cut hard and the money moved. If the poultry specialist requires a seven-day lead in the first month, it should be raised.

Which chapter resolves the tension: Chapter 33. Opening inventory is a working-capital question before it is a purchasing question, and the \$20,700 total sits against both the \$35,000 pre-opening line from Chapter 9 and the \$45,000 working-capital reserve. Chapter 13's job is to produce an honest number and name the conflict; it is not to resolve it.

Chapter 14 — Worked Solutions

Solutions to the daggered (†) and odd-numbered exercises. All Bellwether figures are constructed teaching examples.

Exercise 14.1

Mise en place is the complete set of prepared components, tools, and station set-up a cook needs to execute the menu at service speed — everything portioned, positioned, and within reach before the first ticket lands.

The chapter insists it is a system rather than a virtue because virtues do not scale and do not transfer. If the mise depends on a cook being conscientious, throughput depends on which cook is working, and the restaurant discovers this on the night the conscientious one calls out. A written production sheet tied to a forecast survives a personnel change; a habit does not.

Exercise 14.3

Ticket time is the elapsed interval from a ticket reaching the kitchen to the last plate of that course leaving the pass.

The two boundaries most often set wrongly:

• The start. Measuring from when the guest ordered imports the server's ring-in delay into the kitchen's number. Measure from when the kitchen could have started — arrival at the pass.
• The end. Measuring to plate-up rather than to sold hides food dying under a heat lamp. A plate waiting six minutes for a runner is a six-minute failure that a badly configured KDS records as a triumph.

Exercise 14.4 †

Ordering tells the stations that a table exists and that the long-lead items should start now: "ordering two chicken, one hanger." Nothing short has begun.

Firing tells the stations to start the short items so everything converges: "fire table nine."

A kitchen needs both because the components of one table's meal cook at different speeds and must arrive at one moment. At Bellwether the ribeye is 16 minutes on the fire plus 6 resting, the chicken is 8 from par-roast, the pasta 6, a cold salad 90 seconds. If every station starts when the ticket lands, three plates sit under a lamp for fourteen minutes. The expediter works backward from the longest item, orders it, and fires the rest against it. One word without the other collapses the whole coordination scheme.

Exercise 14.5

All day is the running total of an item across every open ticket on the rail.

Example: three tickets are up — table 12 wants two chickens, table 7 wants one, table 21 wants one. The expediter calls: "Four chicken all day. Two of those are on twelve, going with a hanger and a pasta — hold the hanger until I fire." Every cook now knows the true demand and can batch four birds onto the fire as a group rather than four separate times.

Exercise 14.7

Batch cooking is producing a component in quantity rather than to order.

Every batch-size choice trades prep labor against quality waste. Big batches spread the fixed set-up cost (pulling and washing product, breaking down and cleaning equipment) across more portions but push product further from its quality window. Small batches protect quality but repeat the set-up. The chapter's point is that the direction of the trade is not predictable — at Bellwether's volumes the daily cadence turned out to be both cheaper and fresher — so it must be computed per item, not assumed.

Exercise 14.9

• "44% of entrées go to the hearth" — denominator is entrée orders. It is a menu-mix statement.
• "Peak fire utilization is 67%" — denominator is the hearth's hourly capacity ceiling of 28 items. It is a utilization statement: 18.7 fire actions demanded in the busiest hour of a 95-cover plan night, and $18.7 \div 28 = 66.8\%$.

Different quantities with different denominators. Conflating them destroys the capacity math, because the 67% figure already includes flatbreads and hearth-finished sides, which are not entrées at all.

Exercise 14.11 †

Three sentences, roughly:

The median tells us the kitchen is capable; the 95th percentile tells us how many guests had a bad night, and thirty-four minutes against a fourteen-minute standard is a bad night. Because the median is fine and the tail is not, this is almost certainly a shape problem rather than a capacity problem — so before we touch the schedule, I want the over-22 tickets pulled against the reservation log to see whether they cluster in one band. If they cluster, the fix is at the host stand and it is free; if they are spread across the service, the fix is staffing and it is not.

Exercise 14.13

Two operationally different reasons not to level the stations:

1. Utilization is not the objective; throughput under peak load is. Garde manger's 56% is the kitchen's only shock absorber. It is what the chef raids when the hearth backs up, what absorbs a 40-cover first course plated ahead, and what covers a bathroom break at 8:15. Spending the cushion to make a spreadsheet look even removes the kitchen's ability to recover from anything.

2. The stations are not interchangeable, and moving mix moves the ceiling. Shifting entrées onto garde manger is impossible (they are hot), and shifting starters off the hearth is exactly what the chapter recommends — but shifting them onto sauté, already at ~76%, makes the second-most-loaded station the bottleneck. Any rebalancing has to be re-run through the cover-ceiling arithmetic in §14.3 before it is approved.

Exercise 14.15

The operations objection: the flatbread is a starter that competes for entrée fire in the peak window. Each one is a fire slot; sixteen on a Friday is over half an hour of the hearth's sustained capacity spent on a \$16 item, and five of them land in the 7:00–8:00 hour when the hearth is already over the ceiling.

The revenue argument: \$11.65 of contribution margin per unit, and a bar with no food is a bar that sells fewer drinks and turns fewer waiting guests into diners. Sixteen a night across 260 services is real money — on the order of \$48,000 of contribution margin a year.

A design change that resolves it: move the flatbread off the hearth. Bake it on the plancha or in the convection oven with a finishing char, or hold a par-baked base and finish it — either way it stops consuming entrée fire and keeps every dollar of its margin. That is a menu-development decision (Chapter 10) driven by a kitchen-operations constraint, which is exactly the direction the arrow is supposed to point.

Exercise 14.16 †

Why garde manger broke first, at 6:30, and not the hearth:

Forty guests seated at once means forty first courses landing in roughly a twelve-minute window. At about three minutes of plating work each, that is 120 minutes of work compressed into 12 — about 1,000% of one person's capacity. The hearth's overload, by contrast, was 136% for one hour. Garde manger is the more extreme failure by an order of magnitude, and it arrives forty-five minutes earlier.

Why the fix happened at 4:05 and not 6:25:

The fix is confirm the first course is cold and plate it ahead at 6:15, held at or below 41°F. That converts a production problem into a delivery problem: forty plates already exist, and garde manger's job at 6:52 becomes handing them to runners.

At 4:05 the decision costs fifteen minutes of planning and one conversation with the FOH partner about the event order. At 6:25 there is no time to plate forty covers ahead, and if the event order specified a hot first course, the alternatives at 6:25 are all bad: send it late, send it in waves without telling anyone, or cook it in a window where it will collide with the party's own entrée fire at 7:00. The whole point of the Friday is that the option set shrinks with the clock.

Exercise 14.17

Non-linear because you lose a capability, not just a pair of hands. Three losses stack:

1. Capacity. At Bellwether the hearth cook owns hearth and plancha — 78% of entrée production, not 25% of the line.
2. Situational knowledge. A live fire has moving hot zones and a feeding rhythm; two hours of standing in front of it cannot be handed over in a sentence. So can the station's mise-in-motion — what is resting, holding, two minutes out.
3. The coordination layer. Covering the station pulls the chef off the pass. The specific capability lost is the only view of the whole service — the ability to sequence a dozen tickets, each with its own critical path, and to decide who waits.

Point 3 is the one the chapter emphasizes: ticket times rise on the stations that did not lose anyone.

Exercise 14.19 †

The labor overlay on a Hearth Chicken.

• Hands-on prep per whole bird: 4.5 minutes (butcher, brine, par-roast, chill, in batches).
• Loaded prep wage \$19/hour = $19 \div 60 = \$0.3167$ per minute.
• Labor per bird: $4.5 \times \$0.3167 = \$1.425 \approx \mathbf{\$1.43}$.
• One bird yields two portions: $\$1.425 \div 2 = \mathbf{\$0.71}$ per plate.

Labor-loaded plate cost: $\$8.52 + \$0.71 = \mathbf{\$9.23}$.

**Food cost at a \$29.00 price:** $\$9.23 \div \$29.00 = \mathbf{31.8\%}$ (against 29.4% on the card).

Contribution margin: $\$29.00 - \$9.23 = \mathbf{\$19.77}$ (against \$20.48 on the card).

What changes and what does not. The Chapter 11 cost card does not change: it prices ingredients, it is correct, and the menu price stays at \$29.00. Nothing in the food-cost reporting changes either, since the prep labor is already sitting in the labor line — counting it in both places would double-count it. What changes is comparison between items. A cost card is systematically kind to labor-intensive dishes, so a menu-engineering decision made on cost cards alone is biased toward them. Carry the labor overlay into Chapter 12's matrix as a tiebreaker, not as a re-pricing.

Exercise 14.21 †

The salsa verde batch decision. Weekly need 96 portions; product \$1.05 a portion, so \$100.80 of product a week. Labor \$19/hour = \$0.3167 a minute.

Labor: $150 \times \$0.3167 = \$47.50$; $90 \times \$0.3167 = \$28.50$. Waste: $\$100.80 \times 0.06 = \$6.99 \approx \$7.00$; $\times 0.02 = \$2.33$; $\times 0.01 = \$1.17$.

Ranking: daily (\$24.76) < twice weekly (\$40.02) < once weekly (\$44.05).

Annual difference, best to worst: $\$44.05 - \$24.76 = \$24.83$ a week $\times 52 = \mathbf{\$1,292}$ a year, on one sauce.

The counterintuitive result — that the smallest batches are cheapest — comes from the fixed set-up not scaling with batch size at these volumes: six batches back-to-back is 150 minutes; 1.2 batches a day is 18. The transferable lesson is that batching is not automatically cheaper and the trade must be computed per item.

Exercise 14.23

Costing the flatbread 86 on the Friday.

• Flatbread: \$16.00 menu − \$4.35 plate = \$11.65 CM.
• Substitute cold starter: \$14.00 − \$3.60 = \$10.40 CM.
• Twelve units not sold. Six tables substituted; six ordered nothing extra.

$$6 \times \$11.65 \;+\; 6 \times (\$11.65 - \$10.40) \;=\; \$69.90 + \$7.50 = \mathbf{\$77.40}$$

What it bought: twelve fire slots across the 86 window, five of them in the 7:00–8:00 peak hour. Five slots × 2.14 minutes = 10.7 minutes of hearth time returned at the worst moment of the night. Combined with the ribeye cap, the peak-hour overrun fell from 21 minutes to about 6.

Exercise 14.24 †

Deriving the cover ceiling.

Peak-hour hearth demand for a night of $C$ covers:

• entrées: $C \times 0.90 \times 0.30 \times 0.44 = 0.1188\,C$
• flatbreads and hearth-finished sides: $(C \div 95) \times 24 \times 0.30 = 0.0758\,C$
• total: $0.1946\,C$

Set equal to the ceiling: $0.1946\,C = 28 \;\Rightarrow\; C = 143.9 \approx \mathbf{144\ covers}$.

At a 50% hearth entrée share:

• entrées: $C \times 0.90 \times 0.30 \times 0.50 = 0.1350\,C$
• plus $0.0758\,C$ = $0.2108\,C$
• $0.2108\,C = 28 \;\Rightarrow\; C = 132.8 \approx \mathbf{133\ covers}$

The cost of the difference: $144 - 133 = \mathbf{11\ covers}$ of ceiling. At a \$46 dinner check that is $11 \times \$46 = \mathbf{\$506}$ of sales on every night the room would otherwise have filled.

Note what the derivation assumes and state it whenever you quote the number: a 0.90 entrée attachment, a peak hour carrying 30% of the night's food, a starter-and-side load that scales with covers, and — the one that broke on the Friday — a normal arrival curve.

Exercise 14.25

Recompute the peak hour for 138 covers arriving entirely à la carte on a normal curve.

• Entrées: $138 \times 0.90 = 124.2$.
• Peak-hour entrées: $124.2 \times 0.30 = 37.3$; hearth share $\times 0.44 = 16.4$.
• Flatbreads and hearth sides: $(138 \div 95) \times 24 \times 0.30 = 10.5$.
• Ribeye adjustment, say 3 in the peak hour at +1 slot each: +3.
• Peak-hour demand ≈ 29.9 slots against 28 — about 107%.

So the hearth breaches the ceiling by roughly 2 slots, about 4 minutes — a night that runs a little long at the worst moment and recovers. Compare the actual Friday: 136% and a 21-minute overrun.

What that says about the cause: the failure was not the cover count. 138 covers on a normal curve is a hard but survivable night. The failure was simultaneity — forty covers arriving as one ticket with fourteen hearth items inside it. The reservation book, not the kitchen, produced the outcome, which is why §14.5's decision is a booking rule and why the Friday is a bill sent to Chapter 22.

Exercise 14.27

Why the ribeye costs two slots. The 28-an-hour ceiling is derived from four working positions at an eight-minute average dwell. A ribeye occupies a position for sixteen minutes — twice the average — so while it cooks, the hearth produces one fewer item than the ceiling assumes. It therefore consumes two items' worth of capacity for one plate. (The six-minute rest happens off the fire and costs ticket time, not fire capacity.)

Four in one hour: $4 \times 2 = \mathbf{8\ slots}$ rather than four — an extra four slots, or $4 \times 2.14 = \mathbf{8.6\ minutes}$ of hearth time, out of a 60-minute hour. On a peak hour already at or near the ceiling, four ribeyes are roughly a 14% capacity reduction for that hour, all of which lands on other guests' tickets.

Exercise 14.28 †

Roughly 200 words:

The floor plan's 151 covers is a seating number. It says how many people can sit down, not how many plates the kitchen can produce.

Run the fire. The hearth sustains 28 items an hour: four working positions, an eight-minute average dwell, less an allowance for tending the fire. On our mix, the busiest hour of a night of $C$ covers demands $0.1188\,C$ hearth entrées — that is 90% entrée attachment, 30% of the night in the peak hour, 44% of entrées on the fire — plus $0.0758\,C$ of flatbreads and hearth-finished sides. That totals $0.1946\,C$. Set it equal to 28 and the answer is 144 covers, seven below the floor plan.

So the binding constraint is the fire, not the room, and it is sensitive to menu mix: at a 50% hearth share the ceiling falls to 133. Any capacity number we publish has to be stated as menu-mix-dependent.

The one condition under which you would be right: if we move items off the hearth — the flatbread to the oven, one entrée to the plancha — until peak-hour hearth demand falls to $0.1854\,C$ or lower, the fire stops binding and 151 becomes achievable. That is a menu decision, and it is cheaper than a bigger hearth.

Exercise 14.29

A defensible Wednesday (78 covers) production sheet. The graded column is the right-hand one.

Working: 78 covers × 0.90 = 70 entrées; chicken at ~26% of entrées ≈ 18. Note that the coverage factors differ by item and by day — 1.15 on the two wood par-cooks because they cannot be replaced mid-service, 1.05 on Wednesday flatbread because the bar under-indexes midweek — and that Wednesday runs into a Thursday of 92 covers, so a modest over-par is nearly free.

Exercise 14.30 †

Par-cook policy: pork chop and hanger steak (one page, operating document)

What. Bone-in pork chop and hanger steak are par-cooked over wood during the afternoon prep block and finished on the plancha to order during service. No other item on the menu is wood par-cooked.

When. Prep block 2, beginning no earlier than the hearth reaching working temperature (lit 1:30, ready ~3:00) and completed by 4:15, before line cooks take the station.

Who. The prep cook, with the sous verifying counts and temperatures. The grill cook does not own this task — a task that gates service must not depend on one person's punctuality.

How much. Par = the day's forecast entrée count for that item × 1.15, plus any contracted event quantities in full, from the production sheet. The 15% coverage factor exists because these two items have no replacement path during service: from raw they are 11–14 minutes on live fire against a 14-minute standard.

Cooling. Par-cooked proteins are cooled through the temperature danger zone under the two-stage requirements of the Food Code version this jurisdiction has adopted, using shallow pans and the ice bath or blast chiller as the certified manager's program specifies. This document does not state the time and temperature limits. They come from the adopted local code, the program is designed and verified by the certified food-protection manager, and every batch is entered on the cooling log. See Chapter 25.

Labeling. Every pan is labeled with item, quantity, par-cook date, and use-by, per the program.

Counting. A physical count at 3:00 p.m., against the production sheet, by the sous — not a look, a count, with the sheet in hand.

If the count is short at 3:00. Two hours remain before service, which is enough. In order: (1) par-cook the shortfall immediately if the fire and the raw product allow; (2) if it does not, set a count on the item at the stretch and tell the floor what to sell instead; (3) if the shortfall is large, 86 the item for the peak band rather than mid-service, so the decision is made once, in advance, with the number known. Do not attempt to cook the item from raw during the peak window; it does not fit inside the ticket-time standard and it will take the whole rail down with it.

Exercise 14.31

Twelve checks, each under twenty seconds, tied to a station:

Hearth/plancha — (1) par-cook counts match the 3:00 count sheet; (2) fire at working temperature and wood stocked for the service; (3) resting rack, probe thermometer, and oil/salt at the station. Sauté — (4) sauce pars in the low-boy at quantity; (5) pasta water on and salted; (6) one sauce tasted by the chef. Garde manger — (7) cold starter components at par and covered; (8) dessert components and plates staged; (9) bread and butter set. Pass — (10) 86 board and count board written and legible from the floor; (11) heat lamps on, plates warm and stacked; (12) tonight's event or party sheet posted at the rail.

Exercise 14.33

The derivation is what is graded, not the number. Two worked examples for a hypothetical menu:

Entrées, 12 minutes. The longest routine entrée is a seared duck breast: 7 minutes on the flat-top, 2 minutes rest, 2 minutes plating and garnish, 1 minute of pass slack to hold the table together = 12. Not-to-exceed at 18, which is the same path with one item remade.

Hot starters, 8 minutes. The longest is a fried item: 3.5 minutes in the fryer, 1 minute to drain and season, 1.5 minutes plating, plus 2 minutes of queue allowance because the fryer is shared with a side dish = 8. Not-to-exceed at 12.

The point: each number is the actual critical path of the slowest item in that course on that equipment, plus explicit pass slack — not a benchmark borrowed from another restaurant.

Exercise 14.34 †

The data.

Friday has the second-highest covers and by far the worst times — Saturday does eight more covers with a median six minutes faster. Volume is therefore not the explanation.

Three hypotheses:

1. Shape. Friday's arrivals were compressed — a party, a block of same-time reservations, or a host stand seating heavily in one band. Consistent with a tail that blows out, though note the median also moved, which is more than pure shape usually does.
2. Staffing. Friday was short a cook, or ran a green one in a key station. Consistent with both the median and the tail rising together.
3. Mise or par failure. A par-cook ran short and the kitchen cooked something from raw for part of the service — which raises the median across the whole affected window and produces a long tail.

The single piece of data that discriminates: the timestamps of the over-22 tickets. If they cluster inside one 30-to-45-minute band, it is shape (hypothesis 1). If they are spread evenly across the whole service, it is staffing (hypothesis 2). If they begin at a specific moment and continue to the end of service, it is a mid-service stock-out (hypothesis 3) — and the production sheet will confirm it in thirty seconds.

Exercise 14.35

"Salsa verde — par 6 qt — on hand 6 qt — make 0," and service runs out at 8:05. Three explanations, ranked by how to investigate:

1. The count was wrong (investigate first, because it is the most common and the cheapest to check). Somebody wrote 6 quarts without measuring, or counted containers rather than volume, or counted a partially empty tub as full. Check: is there a physical count discipline, and who performed it?
2. The par was wrong. Six quarts is 96 portions at 2 oz; if the night's chicken count exceeded that — a busy Saturday, or an event added after the sheet was printed — the par was simply too low for the forecast. Check: what forecast was the par derived from, and did the book change after 8:00 a.m.?
3. Portioning drifted. The recipe says 2 oz and the line is putting down 2.75. Six quarts then yields 70 portions, not 96. Check: is there a ladle or spoon spec at the station, and does the plate-up photo match what went out? This one also shows up in food cost (Chapter 11), which is a useful corroborating signal.

Exercise 14.37

Under 200 words, spoken:

Forty-two minutes to service. Here's the night.

One-thirty-eight on the book. Forty of that is the private party, seated six-thirty, one ticket. Their first course is cold and it's already plated — it's on sheet trays in the walk-in, it goes at ten of seven, garde manger you just hand them out. Their entrées fire in two waves, seven at seven o'clock, seven at seven-twelve. Between seven and seven-thirty the hearth belongs to that party and nothing else.

So: flatbread is 86'd from six until eight forty-five. Ribeye is a count of six — tell me at three. Servers, sell the pork and the hanger between seven and eight; they're on the plancha and they're fast, and if somebody orders hearth in that window you quote them thirty minutes before they order, not after.

Stations: sous is on hearth and plancha tonight. I'm on the pass and I'm on sauté when it gets bad. Garde manger, you're the only one with room, so you're my relief.

One thing from last night: two plates went out with cold roots. Touch the pan before you plate.

That's it. Somebody eat something. Let's go.

Exercise 14.39

The argument against sending what is ready: a table that waits twenty-six minutes and eats together has had a slow meal; a table where two people eat while two watch has had a bad meal. The two who were served eat guiltily and fast, the two who weren't have their evening defined by the wait, and the server absorbs a failure they did not cause. It also destroys the pass's only real control: once components are allowed to leave independently, the expediter has no leverage over anything and the whole sequencing scheme collapses into first-come, first-out.

The strongest honest version in favor: there are circumstances where partial service is right, and a good expediter uses them.

• Coursing that the guest expects. Shared starters, or a table that has explicitly asked for food as it comes.
• A very long outlier. If one plate on a six-top is twenty minutes behind because it was dropped and remade, holding five plates to protect one is worse for everyone — send five, tell the table, and comp the sixth's course.
• A bar or counter. Different social contract; guests at a bar expect food when it's ready.

The distinction is not "fast versus slow." It is whether the guest has been told. Partial service that has been offered and accepted is hospitality; partial service that just happens is a failure.

Exercise 14.41 †

The Kitchen Operations section, per the §14 checkpoint. Two pages, condensed here to its structure and load-bearing content:

Station structure. Four positions at peak: hearth (working the plancha as well), sauté/oven/pasta, garde manger/cold, and the chef expediting at the pass and covering the drowning station. The hearth position owns 78% of entrée production; that concentration is a deliberate cost choice and a named single-point-of-failure risk.

Entrée production split. 44% hearth, fired to order · 34% plancha, finished from a wood par-cook · 22% sauté and oven.

Prep and production system. A daily production sheet on which every par is a stated multiple of a stated forecast; a 4:30 mise audit against a written station set-up standard; physical counts at 3:00 on the two items with no mid-service replacement path.

Par-cook policy. Chicken par-roasted (8-minute finish, versus ~35 from raw); pork chop and hanger par-cooked over wood (3-minute plancha finish, versus 11–14 from raw); all under a documented two-stage cooling and logging program designed and verified by the certified food-protection manager against the locally adopted Food Code.

Ticket-time standard. Cold starters 6 / 10 · hot starters 9 / 14 · entrées 14 / 22 · ribeye tickets 24 / 30, tracked separately · desserts 8 / 12. Measured from arrival at the pass to sold. Reported weekly by median, 75th, and 95th percentile alongside comps and remakes.

Capacity — stated as menu-mix-dependent. The hearth sustains 28 items an hour. At the current mix the kitchen's ceiling is about 144 covers against a floor-plan ceiling near 151; at a 50% hearth entrée share it falls to about 133. The plan therefore states capacity as a function of mix, not as a seat count, and commits to re-deriving it whenever the menu changes.

Closing controls. Nightly temperature log, cooling log, walk-in visual check, waste log, counts for tomorrow, and a morning read-and-sign.

What this section does not settle. What the line costs (Chapter 19); whether the dining room can pace to a 28-an-hour fire (Chapter 22); whether a 40-guest event is good business (Chapter 29); whether the par-cook food-safety program is adequate (Chapter 25).

Exercise 14.43

The booking rule, as it would appear in the operations manual:

Simultaneous seating and the hearth. No single reservation, party, or event may place more than ten hearth items on the fire at one time. Because the hearth sustains 28 items an hour, ten items is roughly twenty-one minutes of its entire capacity, which is the most the à la carte side can absorb without pushing tickets past the 22-minute not-to-exceed. A party larger than ten guests on a hearth-heavy menu must therefore do one of three things, chosen at booking and written on the event order: (a) seat in two waves at least twenty minutes apart; (b) take a pre-set menu on which no more than ten covers' worth of entrées come off the hearth, with the balance on the plancha or in the oven; or (c) book on a Tuesday or Wednesday, when the fire has capacity to spare. Any exception requires the chef's approval at the time of booking, not on the day. Events booked without one of these three conditions recorded on the event order are not confirmed.

Chapter 15 — Worked Solutions

Solutions to the daggered (†) exercises and the odd-numbered exercises. All prices are illustrative; students working from their own products will get different dollar answers and the same method.

Exercise 15.1

$$\text{Pour cost \%} = \frac{\text{Beverage COGS}}{\text{Beverage sales}} \qquad \text{Usage} = \text{Beginning inventory} + \text{Purchases} - \text{Ending inventory}$$

The second is required because "beverage COGS" means what you used, not what you bought. Without an ending count, a heavy delivery just before a period close inflates the reported cost and a run-down back bar deflates it — with no change whatever in how anyone poured.

Exercise 15.3

Pour cost is total beverage COGS ÷ total beverage sales, covering spirits, wine, beer, and non-alcoholic. Liquor cost is spirits COGS ÷ spirits sales alone.

Misleading situation: a manager reports "16% liquor cost" to an owner who hears it as pour cost and concludes the whole beverage program is running eight points better than target. At Bellwether the two numbers are 17.0% and 22.0% — a five-point gap created entirely by the wine list.

Exercise 15.4 †

One US fluid ounce = 29.5735 mL.

These are theoretical maximums — zero spillage, zero drip, no heel left in the bottle. Nobody achieves them, which is why the cost card carries a spillage allowance.

Exercise 15.5

Well is the default spirit poured when no brand is named; call is a brand the guest asks for; premium (top shelf) is the high-priced tier, often sipped neat.

Lowest cost percentage: well, by a wide margin — Bellwether's \$10 well vodka soda runs 5.8%. Highest: premium, because you cannot mark up a \$70 bottle six times. Note that the premium pour often carries the best contribution margin in dollars even at the worst percentage, which is Chapter 12's argument in its purest form.

Exercise 15.7

Beer-clean glassware is glass free of fat, protein, and detergent film, on which head forms and laces down the side as the glass empties.

It is a Bellwether-specific concern because Chapter 7 deferred the under-counter glasswasher to year two. All bar glass therefore goes through the kitchen dish machine, whose detergent leaves a film that collapses head on otherwise good draft. In year one, beer-clean is a trained and audited protocol at the three-compartment bar sink — a glass brush, a bar-specific detergent, a rinse — rather than a machine setting somebody selected once.

Exercise 15.9

A beverage transfer is product moving from bar to kitchen (cooking wine, brandy for a sauce, beer for mussels) or the reverse.

You are wrong twice because the product stays in beverage COGS against beverage sales that never happened — overstating pour cost — while never reaching food COGS — understating food cost. At Bellwether an illustrative \$40 a week is \$2,080 a year, or 0.48 pour-cost points, and an equal understatement on the food line.

Exercise 15.11 †

The plan.

$$0.72 \times 30.0\% + 0.28 \times 22.0\% = 21.60\% + 6.16\% = \mathbf{27.76\%} \rightarrow 27.8\%$$

Prime cost: 27.76% + 32.3% labor = 60.06% → 60.1%.

At a 21% beverage share.

$$0.79 \times 30.0\% + 0.21 \times 22.0\% = 23.70\% + 4.62\% = \mathbf{28.32\%}$$

Prime cost: 28.32% + 32.3% = 60.62% → 60.6% — over the 60% benchmark.

The cost. 28.32% − 27.76% = 0.56 points. On \$1,550,000 of sales: **\$8,680 a year.**

Worth noting for discussion: that figure assumes total revenue holds. If the mix falls because guests attach less beverage rather than because they buy more food, revenue falls too, and the loss is far larger — §15.1 works that version at \$46,624.

Exercise 15.13

The dependency is the recipes loaded into the system. A theoretical-cost report compares this period's POS item mix to whatever cost cards are stored, so a bar that changed its house syrup, took a price increase from a distributor, or added three drinks without costing them is comparing reality to fiction — and the report will look reassuringly stable while being meaningless.

The calendar habit: re-cost on a schedule, not on inspiration. Quarterly for the full list, immediately on any contract renewal or menu change, and a standing rule that no drink goes on the menu until its card is built and loaded.

Exercise 15.14 †

In order:

1. Recount. Physically recount the categories carrying the variance, ideally with a different pair of eyes, against the same sheet. Roughly half of all large variances end here.
2. The invoice file. Purchases entered in the wrong period, entered twice, or not entered at all. Check delivery dates against the count date, and check that credit memos were applied.
3. The keg log. A keg tapped and not recorded, or a full keg counted as an empty (or vice versa), moves hundreds of dollars in one line.
4. Transfers. Anything the kitchen took and nobody booked.
5. Comps, spills, and remakes. Product that left without a sale and without a reason code.

Only after all five are clean do you have a control question — and Chapter 34 owns what happens then. The sequence matters: an accusation you have to retract costs more than the variance did.

Exercise 15.15

Wine runs a materially higher cost percentage than spirits — Bellwether plans 28.0% against 17.0% — because bottle pricing is transparent to guests (they can look it up), because inventory sits and ages, and because by-the-glass pouring carries oxidation waste. A program weighted toward wine therefore has a structurally lower beverage margin than one weighted toward cocktails.

It is a legitimate choice because margin percentage is not the only thing a beverage program produces. A serious wine list may be central to the concept, may raise the check average, may attract a guest who returns more often, and may sell bottles that produce excellent contribution dollars at a poor percentage. The mistake is not choosing wine. The mistake is choosing it without doing the arithmetic and then being surprised by the blended pour cost. Chapter 16 works this properly.

Exercise 15.17 †

At \$14.00**: pour cost = 2.79 ÷ 14.00 = **19.9%.** Contribution margin = 14.00 − 2.79 = **\$11.21.

Note the lime: the tested yield of 0.8 oz makes limes \$0.40 an ounce against lemons at \$0.318. Limes yield less and cost more per ounce of juice than students expect, and cold fruit yields less still.

Exercise 15.19

Fully loaded syrup cost: ingredients \$3.84 + labor (11 min ÷ 60 × \$19.00 = \$3.48) = **\$7.32 per 32 oz batch = \$0.229/oz** — nearly double the \$0.120 ingredient-only figure.

Restated Rivermill Sour: the syrup line rises from \$0.09 to 0.75 × 0.229 = **\$0.17. Component subtotal \$3.11 − \$0.09 + \$0.17 = **\$3.19; 3% allowance \$0.10; **total \$3.29. Pour cost at \$15.00 rises from 21.3% to 21.9%.**

The argument for including it: it is a real cost, it makes house-versus-buy decisions honest, and it prevents a chef from "saving money" by making everything in house while the labor line quietly absorbs it — precisely Chapter 1's prime-cost trade.

The argument against: the prep hand is on the clock regardless, so the labor is fixed rather than incremental; loading it onto the card double-counts against the labor line; and cost cards that mix fixed labor into variable product cost stop being comparable across items.

The defensible practice: keep labor off the card, but compute it separately whenever you make a house-versus-buy decision, and say out loud in the beverage plan that the card is ingredients only. What is never defensible is pretending the labor is free.

Exercise 15.20 †

Theoretical glasses = keg oz ÷ 14. Sellable = theoretical × 0.88.

The teaching point is the spread: 31.6% and 17.4% on the same bar, same night, same bartender. Keg format, beer selection, and price set the beer line; execution only moves the loss rate. A tap list of six expensive sixtels priced like domestic lager is a structurally unprofitable beer program no amount of careful pouring can rescue.

Exercise 15.21

The 14-ounce pour wins by \$18.18 a keg. The price went up 22% while the pour went up 29%, so the larger glass is a price decrease per ounce dressed as an increase. To match, you would need \$11.56 — call it \$12.00.

The non-financial override, and it is the one that should decide it: an 18-ounce serving is 29% more alcohol in a single unit. It removes a natural pacing checkpoint, makes it harder for a guest to track their own consumption, and makes it harder for a bartender to track it for them. That is a responsible-service consideration and it outranks \$18.18 a keg.

Exercise 15.22

Ice cost per service: \$2,100 ÷ 52,000 = **\$0.0404 → \$0.04 per drink.**

Annual ice carried by a 15,400-drink cocktail program: 15,400 × \$0.0404 = **\$622** (or \$616 using the rounded four cents). Small, real, and invisible unless someone puts it on the card.

Exercise 15.23 †

Drink cost \$3.85.

**A defensible choice: \$17.00** (a 22.6% pour cost, \$13.15 of contribution). Chapter 12's argument is that you bank dollars, not percentages: the \$19.25 price is theoretically better per unit and will sell fewer, and a drink that sells 60 times a week at \$13.15 (\$789) beats one that sells 35 times at \$15.40 (\$539). The whole-number price also fits the menu's ladder, which Chapter 10 established matters more than the third decimal place of a target.

Full credit for any price in the \$16–\$19 range defended on ladder position, expected volume, and contribution dollars. No credit for a price defended only by "it hits the target."

Exercise 15.25

• Extra cost: 3,815 drinks × \$0.26 = **\$991.90 a year.**
• Extra revenue: 3,815 × \$1.00 = **\$3,815.00.**
• Net gain: \$2,823.10 a year.
• Pour cost before: \$0.58 ÷ \$10.00 = 5.8%. After: \$0.84 ÷ \$11.00 = 7.6%.
• Contribution before: \$9.42.** After: **\$10.16.

Decision: take it. The percentage gets 1.8 points worse and you make \$2,823 more. An operator managing to a pour-cost percentage alone would decline this trade, which is exactly why nobody should manage to a pour-cost percentage alone. The one thing worth verifying first: that the \$1.00 increase survives contact with the market — Chapter 2's trade area, not a spreadsheet, decides that.

Exercise 15.26 †

A 10-SEAT BAR, FOUR TAPS, NO GLASSWASHER              [constructed teaching example]

  BACK BAR
  ┌──────────────────────────────────────────────────────────────┐
  │ TIER 3   premium — what guests read                          │
  │ TIER 2   call brands — what the bartender reaches            │
  │ TIER 1   backups, syrups, glass storage                      │
  │          [ under-bar cooler: beer / wine / mixers ]           │
  └──────────────────────────────────────────────────────────────┘
        ↑ 36 in. aisle, floor clear

  ┌──────────────────────────────────────────────────────────────┐
  │ ┌─────────┐  ┌──────────────────┐  ┌────────┐  ┌───────────┐ │
  │ │ SERVICE │  │  MAIN WELL       │  │ DRAFT  │  │  DIRTY    │ │
  │ │  WELL   │  │  speed rail (6)  │  │ 4 taps │  │  GLASS    │ │
  │ │ ice·gun │  │  ice · 3-comp    │  │ drip   │  │  DROP →   │ │
  │ │ rail(4) │  │  sink · garnish  │  │ tray   │  │  dish pit │ │
  │ └─────────┘  └──────────────────┘  └────────┘  └───────────┘ │
  │  ← servers      ← bartender works here →         ← barback   │
  └──────────────────────────────────────────────────────────────┘
     BAR TOP — 10 seats   ●  ●  ●  ●  ●  ●  ●  ●  ●  ●

  LEGEND:  ● = stool   ← = traffic direction   │ = counter edge

What each placement protects. The service well at the far end keeps servers out of the guests' eighteen inches, so a Friday rush does not force the bartender to be rude to someone through no fault of their own. The speed rail, ice bin, and garnish tray within one step of the working hand protect labor — two extra steps per drink across twelve thousand drinks a year is a real number. The dirty glass drop at the opposite end from the well protects both sanitation and throughput: with no glasswasher, glass has to leave the station immediately and travel one direction only, and the barback station is positioned so that run never crosses the bartender's path.

Exercise 15.27

Three changes, largest lever first:

1. Change keg formats on the two fastest-moving taps. Moving from sixth barrels to half barrels or 50-liter kegs on the beers that turn quickly is by far the biggest lever — Exercise 15.20 shows the same 14-ounce glass costing \$2.84 from a sixtel and \$1.22 from a half barrel. Direction: down. Size: potentially several points on the whole draft line. The constraint is turnover — a half barrel that takes six weeks to blow will pour badly before it empties.
2. Attack the loss rate. Measure yield per keg against POS glass counts for a month. Check walk-in temperature, system balance and applied pressure, faucet technique, and glass rinsing. Moving 18% loss to 12% is roughly two to three points on the draft line. Direction: down. Size: moderate, and free.
3. Re-shape the tap list. Fewer lines that turn faster, at least one lower-cost style carrying volume, and consolidated purchasing for whatever leverage the local distribution structure allows. Direction: down. Size: moderate. Note that the three-tier system limits your options here far more than it does with food; Chapter 16 explains the structure.

Exercise 15.28 †

Step 1 — compute usage.

Sales: 3,180 + 2,780 + 1,290 + 480 = \$7,730. Actual pour cost = 2,345 ÷ 7,730 = 30.3%. Ideal = 1,683 ÷ 7,730 = 21.8%. Variance +8.5 points.

Step 2 — locate it. Spirits carry \$610 of the \$662, or 92% of the variance. Spirits ran 1,150 ÷ 3,180 = 36.2% against a 17.0% ideal. Wine, beer, and non-alcoholic are all within counting noise.

Step 3 — and this is the answer the exercise is actually testing — ask whether over-pour can explain it. \$610 at a blended \$1.20 an ounce is 508 extra ounces. Spirits sales of \$3,180 at a \$13.65 blended price is about 233 drinks. That would require 2.18 ounces of over-pour on every single drink — you would have to more than double every pour, all week, without anyone noticing.

It is not over-pour. It is a measurement failure, and §15.2's rule applies: over 2.0 points, recount before you conclude anything.

Monday: (1) recount the spirits, including the storeroom and any secondary location — the most likely single cause is a shelf, a case, or an entire storage area omitted from the ending count; (2) pull the invoice file and confirm the \$820 is complete and correctly dated; (3) check the transfer log and the comp report. Say nothing to anyone about anyone until those three are clean.

Exercise 15.29

The transfer effect. \$70 ÷ \$8,400 = 0.83 pour-cost points, or \$3,640 a year.

Against a 2.8-point overrun, transfers explain about 30% of it — a meaningful piece and nowhere near all of it. The remaining ~2 points still need an explanation, and the candidates are the ones in §15.7.

The two sentences: "Before we talk about anybody, we should know that the kitchen has added three dishes that cook with beverage inventory and we have never had a transfer sheet — that alone is roughly 0.8 of the 2.8 points, about \$3,600 a year, and it has nothing to do with the bar. Let me put a transfer clipboard by the walk-in this week and recount, and if we still have two points after that I will bring you a real answer instead of a suspicion."

Exercise 15.30

• Annual drinks: 245 × 52 = 12,740.
• Extra spirit: 12,740 × 0.25 oz = 3,185 oz.
• Cost: 3,185 × \$1.20 = **\$3,822 a year.**
• Pour-cost points: 3,822 ÷ 434,000 = 0.88 points.
• Bottles: 3,185 ÷ 25.4 = 125.4 bottles of 750 mL — about two and a half a week.
• Jigger policy cost: 3 seconds × 12,740 = 38,220 sec = 10.6 hours × \$22 = **\$234 a year.**
• Return: about 16 to 1.

(The chapter's \$3,815 figure uses 12,718 drinks derived from revenue rather than 245 × 52; the one-dollar-in-four-hundred difference is rounding and does not change anything.)

Exercise 15.31 †

Glasses required: \$61,200 ÷ \$8.50 = 7,200 glasses. Beer required at 14 oz: 7,200 × 14 = 100,800 sellable ounces.

On yield alone: \$1,320 of benefit against a \$1,680 contract. It does not pay — a \$360 shortfall.

Adding remakes. Cost per sellable glass at 11% loss = \$13,364 ÷ 7,200 = **\$1.856. Two remade beers a service × 364 services = 728 glasses × \$1.856 = **\$1,351.

Total benefit: \$1,320 + \$1,351 = \$2,671** against \$1,680. A 1.59-to-1 return** — it pays.

And the honest coda: the largest item is still not in the table. A guest served a sour or flat beer does not complain; they order something else next time or go elsewhere. That is unquantifiable and it is larger than both computed lines. A control that fails its obvious payback may still be the cheapest thing you do.

Exercise 15.33 †

Model policy.

Comps and buybacks — Bellwether bar. Any bartender may comp up to two drinks per shift at their own discretion, for service recovery, a regular's occasion, or a taste that sells a bottle. Anything beyond two requires a manager. Every comped drink is rung into the POS at full price and then comped with a reason code — RECOVERY, REGULAR, TASTING, SPILL, REMAKE, or STAFF. Product that leaves without a sale and without a code is the only thing on this list that is a problem. The bar manager reviews the comp report every Monday and posts the weekly total on the bar's board. Target: under 1.5% of beverage sales. If we are over it, we talk about why, not about who.

Pre-shift introduction, three sentences.

"Starting Tuesday, everything that leaves the bar without a sale gets rung with a reason code — comps, buybacks, spills, remakes, all of it — and you each have two comps a shift to use however your judgment says to use them. This is not because I think anybody is doing anything wrong; it is because right now when the count comes up short I cannot tell the difference between generosity and a counting error, and that is unfair to you. When it's coded, I can defend the number, and so can you."

Grading note: the pre-shift language must (a) attach a number to the generosity rather than withdrawing it, (b) state the purpose as visibility rather than suspicion, and (c) put the manager's own problem — not the staff's honesty — at the center of the explanation.

Exercise 15.35

The arithmetic. \$15.00 cocktail, \$3.20 cost, CM \$11.80. At 50% off: price \$7.50, CM \$4.30.

$$\text{Multiplier} = \frac{11.80}{4.30} = 2.744 \rightarrow \textbf{+174.4\% units required to break even}$$

Nearly triple the volume, in a window that is not a dead window, from a guest population that is small and finite. On the beverage line alone it will not clear.

The two non-financial risks, and they are the reason to modify rather than approve:

1. Responsible service. Industry guests come off shift, arrive together, arrive tired, and arrive late. A 50%-off mechanic aimed at that population removes the price friction that paces consumption from exactly the group least likely to have eaten. This is the risk that decides it.
2. Positioning and staff fairness. A standing discount for one identifiable group changes who is in the room on Thursdays, and it puts your own staff in the position of adjudicating who qualifies.

A defensible answer: modify. Counter-propose a fixed-price short list — three cocktails, one draft, two wines by the glass at a set price — available all night on Thursday rather than in a window, with a food item included or required, and the same refusal standard stated explicitly in pre-shift. That structure is materially safer, is legal in more jurisdictions (verify locally), and is far cheaper: at a \$10 fixed price the CM is \$6.80 and the multiplier drops from 2.744 to 1.735.

Exercise 15.37 †

The three explanations to rule out first, in order:

1. The shifts, not the person. Are their shifts the ones with the cocktail-heavy crowd, the patio, the Friday rush, or the industry-night list? Category mix moves pour cost independently of anyone's pouring, and §15.2's whole argument is that the blend misleads.
2. The station. Are they working the shifts with no barback, the shifts where the glass runs are longest, the shifts where the jigger ends up under a shaker tin? §15.3's On the Line callout says the station is usually the culprit, not the hands.
3. Training and drift. When did anyone last check them against the standard, with a jigger and a graduated cylinder? A four-count that has quietly become a five-count is the single most common cause and it is invisible from the inside.

The conversation, in order.

First, do nothing publicly and say nothing to anyone else. Second, run the three checks above — pull the shift-level mix, look at the schedule, and do a routine pour test with every bartender, not just this one, so nobody is singled out. Third, if the gap persists, have a normal, private, non-accusatory conversation that starts with the number and not with a theory: "your shifts run about a point and a half heavier on spirits and I can't work out why — walk me through how you're building the sour." Fourth, re-measure in three weeks.

The framing that matters: this person is your best guest-relationship asset in the building, and 1.5 points is roughly \$1,300 a year at Bellwether's spirits volume. The cost of handling it badly — losing them, or teaching the staff that a variance report is an accusation machine — is far larger than \$1,300. Chapter 34 owns what you do if the three checks come back clean and the gap does not close.

Exercise 15.39

Decline it, and the reason is the mechanic rather than the depth. A price that falls on the second and third round is a structure that pays a guest to drink faster. It compresses consumption into a window, it removes the pacing friction that price provides, and it puts a bartender in the position of serving a cheaper drink to the guest they should be slowing down. Legality is not the test — several states ban exactly this format precisely because of the mechanism, and the ones that do not have not thereby made it safe.

The counter-proposal: the same traffic goal, achieved with a fixed price on a defined short list, available for a long window, dine-in, with food available throughout, and a per-guest limit if the volume warrants one. It is cheaper to run than deep rounds, it is legal in more places, it is far easier to train against, and it does not reward speed. If the goal is genuinely traffic, a fixed-price list plus a food value message will produce it; if the goal is volume per guest, the answer should be no regardless of what it earns.

Exercise 15.41 †

Revenue \$1,550,000 throughout; food cost 30.0%, pour cost 22.0%, labor 32.3%.

Which row is the plan: 28%.

Which row should concern you most: 22%. Prime cost lands at 60.5%, over the full-service benchmark Chapter 1 established, and beverage contribution falls **\$72,540** below plan — \$1,395 a week. Note that only the top row breaches the benchmark, and note how small the movements look: six points of mix moves prime cost by 0.72 points and contribution by \$108,810 across the range. That is the leverage Chapter 1 warned about, appearing in the beverage line.

The honest caveat to state in the plan: this table holds total revenue constant, which is the generous version. If beverage falls because guests attach less rather than because they spend more on food, revenue falls with it and the damage roughly quadruples — §15.1 works that case at \$46,624 of lost contribution for a four-point miss.

Exercise 15.43

Model answer. Bellwether's beverage line is not flat across the year. The 16-seat patio is open roughly twenty weeks, and patio covers skew toward beverage — more cocktails, more beer, higher attachment per guest. Layered on that is the general seasonality of a Midwestern dining room.

Expected shape: a June–September plateau meaningfully above the annual average; a soft, warm October; a December bump from private events and holiday drinking that is real but short; and a trough in late January and February, which is both the lowest-revenue period and the point at which the patio contributes nothing. Those two months are the ones to name.

What it implies operationally:

• Purchasing. Par levels must move with the season, not sit at an annual average. Buying summer pars in February leaves capital sitting in a back bar during the weeks you can least afford it, and \$11,800 of beverage inventory turning eight times a year is slow money to begin with.
• Count cadence. Do not relax the weekly count in the slow season — that is precisely when variance is proportionally largest and when the habit is easiest to lose. If anything, count more carefully: a \$100 variance against \$8,400 of sales is 1.2 points; the same \$100 against \$5,600 is 1.8 points.
• The patio's opening and closing weeks should be treated as inventory events, with a full count either side, because the mix moves sharply and a mix shift will otherwise be misread as a discipline problem.

Chapter 33 owns what the February trough does to cash, and it is the more dangerous half of this answer.

Chapter 16 — Worked Solutions

Solutions to the daggered (†) and odd-numbered exercises. Arithmetic shown throughout. Where an exercise asks for judgment, a model answer is given and alternative defensible answers are noted.

Exercise 16.1

Wine list architecture — the deliberate structure of a list: number of selections, how they are grouped, how prices ladder, which are poured by the glass, and what each section is for. BTG yield — saleable glasses per bottle at the house pour: bottle ounces ÷ pour ounces (25.4 ÷ 5 = 5). Bottle-price laddering — pricing with a markup multiple that declines as wholesale cost rises, so gross-profit dollars rise even as the cost percentage worsens.

Exercise 16.3

The 21st Amendment, ratified 1933. The three-tier system was built to dismantle the pre-Prohibition tied house — the arrangement in which a producer owned, financed, or controlled the outlet that sold its product and pushed volume through it. The remedy was to force an independent, separately licensed wholesaler between the maker and the seller.

Exercise 16.5

1 ÷ multiple = cost %. 2.0× = 50.0% · 2.5× = 40.0% · 2.8× = 35.7% · 3.2× = 31.3% · 3.5× = 28.6%.

Exercise 16.6 †

Wine markup is a per-bottle multiple applied to wholesale cost to reach a menu price — the buyer's tool, and what a distributor rep talks in ("you can get three times on this"). Wine cost percentage is a category result: wine COGS ÷ wine sales over a period — the accountant's number, and what appears on the P&L. They are reciprocals on any single bottle but not on a list, because a list is a mix: the category percentage is the mix-weighted average of every bottle's percentage, weighted by dollars sold, not by bottles on the shelf. This is why a list can be laddered from 28.6% to 59.3% and still land at 28% — and why an operator who "sets a 3× markup" has not thereby set a 33.3% wine cost.

Exercise 16.7

(a) Exclusivity — a wine is usually represented by one wholesaler per state; if that house does not call on accounts your size, the wine is unavailable at any price. (b) Price posting — many states require posted wholesale prices and equal treatment of licensees, so there is no negotiating your cost. (c) Quantity-discount restrictions — permitted in some states, restricted or barred in others. (d) Credit terms — several states mandate COD or short terms with a statewide delinquency list, making alcohol the one category you may not be able to float.

Exercise 16.9

Court of Master Sommeliers — service- and blind-tasting-focused, four levels, upper levels famously difficult. WSET — the most widely available academic track, Levels 1–4, Level 4 is the Diploma. Society of Wine Educators (CSW) — knowledge-based, exam-only. Institute of Masters of Wine (MW) — trade and academic rather than floor service.

Exercise 16.11

Because the comparison that matters is gross profit per unit of labor and risk, not cost percentage. A \$14 glass of wine has \$2.80 of product in it (20%) and yields \$11.20 of gross profit for the labor of pulling a cork. A \$14 cocktail has \$2.52 of liquor in it (18%) and yields \$11.48 — 28 cents more — but required juicing, syrup, batching, shaking, straining, garnishing, and washing a tin. Wine's percentage is ten points worse and its contribution per minute of bartender labor is far better. In a business chasing a 60% prime cost, where labor is the half you can move this week, that trade is a good one.

Exercise 16.12 †

Restatement. The cost of a by-the-glass bottle is incurred the instant the cork comes out, not gradually as the wine sits. Once open, the only question is how many of the five glasses sell before the wine stops being sellable. Preservation slows the clock; it never un-opens the bottle. So every dumped glass traces back to a single decision — somebody opened that bottle at that moment — and the diagnostic question is never "how do we store it better" but "why did we open it then?"

The control that follows. Add a date-and-time column to the open-bottle log (Figure 16.4 is criticized for lacking exactly this), and a rule tied to the pattern it exposes. Bellwether's version: no sparkling bottle opens before 7:00 p.m. on a Tuesday unless a guest has already ordered the glass. The log finds the pattern; the rule fixes it. A vacuum stopper does neither.

Exercise 16.13

Why \$11 and not \$9. At a \$46 check average, a \$9 glass of wine signals cheapness that costs more in perception than the \$2 gains in margin — and the guest choosing the least expensive glass is choosing a position on the list, not a price point they have computed. At \$11 the pour cost is 9 ÷ 55 = 16.4%, better than the 20% the "glass = bottle cost" rule produces, and nobody experiences \$11 as expensive.

Why nothing above \$17. Because the same rule that lets you price up at the bottom punishes you at the top. A \$25 glass in this room sells perhaps twice a week; the remaining three glasses oxidize. The rule breaks in both directions, and the ceiling is set by velocity, not by margin — a bottle whose glasses cannot all sell inside its open-bottle window should not be on the by-the-glass list at any price.

Exercise 16.15

The variable that changed is guest willingness to pay for a premium glass. At Bellwether's \$46 check there is no demand for a \$32 glass, so the cabinet can only save waste — \$1,426 of it, net \$946 after gas, a 4.4-year payback. In a room where \$32 glasses sell, the cabinet creates a transaction that could not otherwise exist: a \$60-wholesale bottle poured five ways at \$32 is \$160 of revenue and \$100 of gross profit per bottle, and three glasses a night is roughly \$21,840 a year.

The general lesson: equipment that reduces waste has a payback capped by the waste; equipment that unlocks revenue has no such cap. Before buying anything, decide which kind it is — and never buy the revenue case if the guests who would produce that revenue do not walk in.

Exercise 16.16 †

The test. Hand the list to someone who does not care about wine — a line cook, your accountant, a friend from another industry. Say nothing. Watch.

Passing: their eye lands somewhere, stays, and they ask "what's this one like?" Failing: they scan top to bottom, stop, and hand it back with a slightly apologetic face.

The most common structural reason for failure: the list is organized by a category the reader does not possess — grape variety, region and appellation, or producer alphabetization. All three assume the reader already has a map. A person handed a document they cannot navigate does the safe thing and orders a beer. The fix is to organize by what the wine does (Crisp & Bright, Dark & Structured) and put the grape and origin underneath, where the guest who wants them will still find them.

Exercise 16.17

Target-cost pricing says price = cost ÷ target cost %. At a 28% target: an \$8 bottle prices at \$28.57 and an \$80 bottle prices at \$285.71.

At the bottom it fails by leaving money on the table. The bottom of a list is the least price-sensitive part of it, because the guest is choosing a position rather than computing a number. A \$28 price yields \$20 of gross profit against \$24 at a flat 3× — twenty-five percent more margin that no guest experiences as a difference.

At the top it fails by not being a price. \$286 in a \$46-check neighborhood room is not a high price; it is a refusal. And a bottle that never sells does not run a 28% cost — it runs no cost percentage at all, because both terms are zero. What it runs is \$80 of immobilized capital in a room that hits the mid-seventies in August.

The ladder resolves both by holding the dollar margin as the governed quantity: \$20 at the bottom, \$55 at the top, rising monotonically, with the percentage allowed to deteriorate all the way up. You bank contribution margin. You have never deposited a percentage.

Exercise 16.19

Hospitality reason: refilling is a point of contact — it lets the server read the table, offer the next thing, and control pace. A table pouring for itself has been left alone. Compliance reason: the server is the only person in the building counting. A four-top that pours its own two bottles has had roughly five drinks apiece and nobody knows it. Removing the refill removes your last line of sight before the dram-shop question becomes real.

Exercise 16.21 †

\$21.50 falls in the **\$10–\$22 band, at the top, where the multiple has fallen to 2.8×**.

• 21.50 × 2.8 = \$60.20** → rounded to the nearest dollar: **\$60
• Gross profit: 60.00 − 21.50 = \$38.50
• Cost percentage: 21.50 ÷ 60 = 35.8%

Note the rounding went down here, costing \$0.20 of margin. Round to the nearest dollar as a rule and do not chase the twenty cents; a list of clean numbers is worth more than the rounding.

Exercise 16.23

Bottle \$16.50, glass \$17, 5 oz, 5 glasses.

• Cost per glass: 16.50 ÷ 5 = \$3.30
• Revenue at full yield: 5 × 17 = \$85
• Pour cost 5 of 5: 16.50 ÷ 85 = 19.4%
• Pour cost 4 of 5: 16.50 ÷ 68 = 24.3%
• Pour cost 3 of 5: 16.50 ÷ 51 = 32.4%

Selling three of five instead of five of five costs thirteen points on that bottle.

Exercise 16.24 †

• Yield: 25.4 ÷ 6 = 4.23 → 4 saleable glasses.
• Pour cost at full yield: if the glass is priced at the wholesale bottle cost C, revenue is 4C and cost is C. C ÷ 4C = 25.0% — five points worse than the 20% a five-glass yield produces, and it is worse before any waste.
• Stranded per bottle: 25.4 − (6 × 4) = 1.4 oz.
• Annual value of stranded wine: 1,200 bottles × 1.4 oz = 1,680 oz. At 25.4 oz per bottle that is 1,680 ÷ 25.4 = 66.1 bottles. At \$15 each: **\$992 a year.**

Poured away not through carelessness but through a pour-size decision made once, by someone who never divided 25.4 by 6.

Exercise 16.25

• Average bottle price: \$66,000 ÷ 1,243 = **\$53.10**
• Total bottle COGS: \$66,000 × 0.37 = \$24,420; per bottle: 24,420 ÷ 1,243 = \$19.65
• Average gross profit per bottle: 53.10 − 19.65 = \$33.45

This is the figure the corkage decision is set against in §16.8: \$25 recovers 74.7% of it with no inventory, no capital, and no spoilage.

Exercise 16.26 †

Bottles: 48 + 48 + 20 + 6 = 122. Value: 672 + 864 + 520 + 288 = \$2,344. Average wholesale per bottle: 2,344 ÷ 122 = \$19.21.

Note that the ceiling shelf is 12.5% of selections but only 12.3% of dollars (288 ÷ 2,344) — because it is stocked two deep rather than three. That is §16.8's rule satisfied by construction, and it is the difference between a disciplined list and Bellwether's, which sits at 19.2%.

Exercise 16.27

• Total turns: \$31,000 ÷ \$2,344 = 13.2 turns; days: 365 ÷ 13.2 = 28 days.
• Ceiling shelf alone: \$1,900 ÷ \$288 = 6.6 turns; days: 365 ÷ 6.6 = 55 days.

Better than Bellwether's 2.7 turns / 135 days, entirely because of the two-deep decision. Same list shape, half the depth at the top, twice the velocity of the capital.

Exercise 16.29 †

Applying Figure 16.2's ladder, interpolating within bands:

(7.50 × 3.5 = \$26.25, which yields only \$18.75 of gross profit; the \$20 floor lifts it to \$28.)

Blended at ten bottles each:

• Revenue: 10 × (28 + 35 + 48 + 60 + 81 + 110) = 10 × 362 = \$3,620
• COGS: 10 × (7.50 + 11 + 16 + 24 + 36 + 58) = 10 × 152.50 = \$1,525
• Blended cost: 1,525 ÷ 3,620 = 42.1% · Gross profit: \$2,095

Hold that 42.1% next to Exercise 16.30's answer before concluding anything about it.

Exercise 16.31

Four moves, ranked cheapest first:

1. Train the two questions (§16.7). Free, one pre-shift, works on a fourth-shift server. The finger-on-the-price gesture is the highest-return zero-cost move in the chapter.
2. Reorganize the list by style with grape underneath (§16.2). Cost: one afternoon and a laser printer. Removes the navigation failure that sends guests to beer.
3. Write the SAY THIS line on twelve tasting cards (§16.7). A few hours of the FOH partner's time. Servers use language they have been handed; they do not invent it at the table.
4. Move something genuinely good into the second-cheapest slot (§16.4). Costs about three points of cost percentage on one slot and recruits guests upward for years.

Nothing on this list changes a price, and together they are plausibly worth the fifty cents.

Exercise 16.32 †

Replacement, priced on the ladder: \$15 wholesale sits in the \$10–\$22 band at 3.2×. 15 × 3.2 = \$48.** Gross profit **\$33. Cost percentage 15 ÷ 48 = 31.3%, against the incumbent's 9 ÷ 32 = 28.1% — the slot gets 3.2 points worse on cost percentage.

At 180 bottles a year:

\$1,800 more gross profit, at a cost percentage three points worse. Make the swap.

The honest caveat, and the number that answers it. Will the slot still sell 180 bottles at \$48 when it sold 180 at \$32? Possibly not — that is a sixteen-dollar jump. Compute the break-even volume: \$4,140 ÷ \$33 of gross profit per bottle = 126 bottles. The replacement wins as long as it sells at least 126, which is a 30% volume decline you can absorb. Given that the slot's traffic comes from a position on the list rather than from a price, a 30% decline is unlikely. Swap it, and watch the slot's units for a quarter.

Exercise 16.33

A defensible 24-selection list for a 40-seat, \$38-check sibling:

Approximate inventory: 10 BTG × 6 deep = 60 bottles; 14 bottle-only × 4 deep = 56; 116 bottles at roughly \$16 average = about \$1,850.

Defense of the cuts. The ceiling comes down from \$135 to \$78 because a \$38-check room has no buyer above it and the anchoring benefit does not justify capital that turns twice a year. Dark & Structured drops from 12 to 6 because half of that depth existed to serve a steak-and-lamb menu this restaurant may not have. Round & Textured drops from 7 to 4 because the by-the-glass core does most of that work. Rosé goes to one because a single dry rosé covers the need and two is a seasonal luxury. Ten by the glass rather than twelve, because fewer open bottles in a smaller room is fewer dumped glasses — the waste rate is driven by covers per open bottle, and a 40-seat room has fewer.

Exercise 16.34 †

Totals. Opened 26 · possible 130 · sold 112 · dumped 18. (112 + 18 = 130 ✓)

• Revenue: 112 × \$14 = **\$1,568**
• Wine cost: 26 × \$14 = **\$364**
• Achieved pour cost: 364 ÷ 1,568 = 23.2%
• Full-yield pour cost: 364 ÷ (130 × 14) = 364 ÷ 1,820 = 20.0%
• Dumped wine at cost: 18 glasses × \$2.80 = **\$50.40 in one week — annualized, roughly \$2,600**, or about 186 bottles.

The two selections doing the damage:

• Sparkling — 7 dumped of 20 possible (35% waste). By far the worst, and unsurprising: sparkling has the shortest open-bottle window on any list. Fix: a proper sparkling stopper is non-negotiable and costs about \$8; then a hard opening rule (no bottle opens before a guest has ordered the glass, on slow nights); then, if the waste persists, move sparkling to bottle-only or to a smaller format where the format is available and appropriate.
• Round white — 4 dumped of 20 (20%). Second-worst. Fix: carry one round white by the glass rather than two if the list has two, or drop the price a dollar to move the last glass. If it is a single selection, it is simply over-poured for demand — reduce the par and open later.

Everything else ran between 3% and 15%, which is normal. You do not fix a by-the-glass program in general; you fix two selections.

Exercise 16.35

• Days of cover: 3 bottles ÷ 1 sold × 90 = 270 days.
• Gross profit produced in the quarter: \$113 − \$58 = \$55.
• Capital held: 3 × \$58 = **\$174.**
• The same \$174 on a shelf turning 9.1× at a 35% cost:** annual COGS 174 × 9.1 = \$1,583; revenue 1,583 ÷ 0.35 = \$4,523; **gross profit \$2,940 a year, or about \$735 a quarter — roughly thirteen times the \$55.

Recommendation: reduce from three bottles to one, freeing \$116 and two rack slots while keeping the price point on the page.

The limits — and this is the part that separates a good answer from a naive one. The \$2,940 comparison is not a redeployment plan. The fast shelf is demand-limited; owning more of it does not sell more of it. And the \$113 bottle may be earning its place as a price anchor that makes the \$75 bottle read as reasonable, which no units-sold report can show. Hence reduce, don't cut — one bottle anchors as well as three.

Exercise 16.36 †

Size of the miss: 31% − 28% = 3 points × \$40,000 = **\$1,200 for the quarter, \$4,800 annualized.** That is real money and worth a week of attention.

Investigation order (largest and most likely first):

Note the order is not arbitrary. Bottle sales are stated to be on plan, which rules out mix shift toward the expensive end of the ladder as the cause, and points the investigation at the by-the-glass program — where the waste, the pour drift, and the untracked pours all live.

Exercise 16.37

Model outline for the manual section. Open with the two questions verbatim and the instruction to ask them in order. Then: pronunciation policy (offer the name so the guest can point; never correct, ever). Then the tasting-card protocol — where the cards live, that every server tastes every by-the-glass wine at the quarterly session, and that the SAY THIS line is a starting point rather than a script. Then the opening rule: who may open a new by-the-glass bottle, at what time, and what to check first (is there an open bottle already, and how old is it?). Then the dump standard — would you set this in front of table 12 without saying anything? Then the refusal procedure, in full, with the explicit statement that management will back the server and that the check average will not be discussed afterward. Two pages. Sentences a person can use while carrying plates.

Exercise 16.38 †

Guest-facing, printed at the foot of the list:

Corkage is \$25 per 750 ml bottle, two bottles maximum per table. We'll happily waive one corkage fee for each bottle purchased from our list. We're not able to accept wines that appear on our own list — but we'd love to pour you a glass of ours alongside.

Three sentences: the fee, the waiver, the exclusion, and a hospitality exit from the exclusion.

Internal version for staff, additions:

• The two-bottle limit is per table, not per guest, and is not negotiable at the table — a manager approves any exception, and the reason goes in the log.
• The waiver is one fee per bottle purchased, applied automatically, not on request. A guest who has to ask for a discount has already had a worse evening.
• Check the list before accepting the bottle, at the point of greeting, not at the point of opening. Declining a bottle after it is open is a bad conversation you can prevent entirely.
• Pour a taste for the host and pour it as well as you would pour a \$135 bottle of ours. A guest who brings a bottle has an occasion; the twenty-five dollars is not the transaction.
• If a guest disputes the fee: do not argue and do not defend the policy at the table. Say "let me take care of that for you," remove the fee, and tell a manager afterward. The fee is \$25. A table's goodwill is worth many multiples of \$25, and the argument is one you cannot win in a dining room. The manager decides whether the policy needs to be printed more visibly.

Exercise 16.39

Model memo. Thank them specifically and by wine — you tasted it, you liked it, you are not pretending otherwise. State the constraint honestly and numerically: "we sell about four bottles a month of this style, so ten cases is thirty months of supply. At our cost of capital that's roughly \$310 of financing against a \$264 discount, before I count what thirty months does to it in our storage room." Then place a real order — two cases at list — so the call produced revenue. Then ask for what you actually want: "send me the posted specials the week they drop, and put me on the list for anything in the \$14–\$18 range that would work by the glass." You have declined the deal, demonstrated exactly why, bought something, and told the rep how to sell you next time. Preserve the relationship; you need the next call.

Exercise 16.41

How to decide: you do not decide. You ask the regulator, in writing, before accepting anything. Inducement rules — what a wholesaler may provide to a licensee in the way of printing, glassware, signage, training, staff trips, or equipment — vary enormously by state and are among the most actively enforced provisions in alcohol regulation.

Whom to ask: your state ABC authority, and your attorney if the answer is ambiguous.

Who bears the risk: in most states, the licensee — you — not the wholesaler. That asymmetry is the entire reason a rep's assurance is worthless as a defense.

If you have already accepted: stop using the item, document what was received and when, disclose it to counsel, and ask about the remediation path. Most regulators distinguish sharply between an operator who self-reports a mistake and one who is found out. Do not simply hide the glassware.

Exercise 16.42 †

At the table, immediately: take the bottle back. No diagnosis, no debate about whether it is "actually" faulty, no invitation to the guest to justify themselves. "Let me get that off the table — what would you like instead?" Then bring the replacement quickly and pour it yourself.

Cost of that option: \$52 of wholesale wine, once, for a guest who spends well and now has a story about how you handled it. Against the alternative — a guest who was told the wine was fine — which costs you the guest.

Afterward, and this is the part that matters: taste it. Determine, if you can, whether it is cork taint (a fault the distributor may credit) or heat damage (a fault your storage room caused, which nobody will credit). Log it either way, with the SKU, the date, and your judgment. Then check the rack: if it is heat, the other two bottles of that wine have the same problem and should come off the list before another guest finds one.

Recording: a running "bad bottle" line in the wine log, reconciled monthly against the \$977 loss allowance in §16.6. A restaurant that budgets fifty bottles a year of loss and has never written a single one down does not actually know whether its allowance is right.

The option you do not take: arguing. It costs \$52 to be generous and it costs a customer to be correct.

Exercise 16.43

Possible underlying facts, and what each obligates:

• The wine genuinely is not good. Taste it yourself, immediately, blind if you can. If the server is right, replace the selection. A slot that produces reluctant servers produces no sales, and you have paid for inventory that actively harms you.
• The wine is fine and the server has never tasted it. Obligates you to fix the tasting program, not the server. This is the \$1,980 line item doing exactly what it exists for.
• The wine is fine and the server has absorbed a snobbery about inexpensive wine. Obligates a conversation about what the second-cheapest slot is for, and about the guest who orders it.
• The server is uncomfortable selling anything and this is where it surfaced. Obligates coaching on the two questions, which take the judgment out of it entirely.

In all four cases: thank them for telling you. A server who raises this is doing the program a favor, and how you receive it determines whether anyone raises the next one.

Exercise 16.45 †

Wine at 45% of the beverage line, beverage revenue unchanged at \$434,000:

• Wine revenue: 0.45 × 434,000 = \$195,300
• Wine COGS at 28%: \$54,684
• Remaining for spirits and beer: 434,000 − 195,300 = \$238,700, split in the original 186 : 83 ratio → spirits \$165,043** at 18% = **\$29,708; beer \$73,657** at 19% = **\$13,995
• Total beverage COGS: 54,684 + 29,708 + 13,995 = \$98,387
• Blended beverage pour cost: 98,387 ÷ 434,000 = 22.7%

So the 22% target does not hold. It rises 0.7 points, which on the beverage line is 98,387 − 95,450 = \$2,937 a year of additional cost. That flows straight through to prime cost: 0.19 points of the plan's 27.8% blended COGS.

What would have to change. To restore a 22.0% blend with wine at a 45% share, wine's own cost percentage would have to fall to (95,480 − 29,708 − 13,995) ÷ 195,300 = 51,777 ÷ 195,300 = 26.5% — a point and a half better than the plan. That is achievable, but only one way: shift the mix inside wine toward by-the-glass, which runs 22%, and away from bottles, which run 37%. Going from a 60/40 BTG-to-bottle split to roughly 72/28 does it.

Per-guest wine spend: 45% of the \$12.88 beverage component = **\$5.80** (up from \$4.90).

The lesson to write into the plan: wine and spirits have a ten-point cost gap, so beverage mix is a cost variable even when beverage revenue is unchanged. A plan that states a blended pour cost without stating the mix underneath it has not actually stated anything.

Exercise 16.47

For opening at 28 selections: frees roughly \$1,200 of opening inventory and six weeks of the buyer's attention; eliminates the shelf that turns 2.7 times before you have any data on whether it sells here; puts cash in the working-capital reserve, which is the only asset with a liquidation value of one hundred percent (Case Study 16.2). And you can add the top shelf in month nine knowing exactly what your guests buy.

For opening at 40: the ceiling shelf is marketing inventory (§16.8) — it makes the \$60 bottle read as reasonable, and a list with nothing above \$78 tells a certain guest this is not a place that takes wine seriously. That guest orders a cocktail, and you never see the loss. Opening thin also means opening twice, and a list that changes in month nine loses the servers' fluency at precisely the point it was becoming an asset.

The call: open at 36 — the full architecture, the full price range, but the ceiling shelf at two bottles deep instead of three and three selections instead of five. That preserves every price point on the page (the anchoring argument) while cutting the slowest capital roughly in half. It costs about \$400 of inventory and it costs nothing in what the list says.

What would change my mind: a market analysis (Chapter 2) showing a materially higher-income guest than the \$46 check implies, or a neighborhood with no other serious wine list within walking distance. Either would make the ceiling shelf a differentiator rather than an ornament — and differentiators earn depth.

Chapter 17

Exercise 17.2

The nine components and where each hides:

The point of the table is that no account is named "turnover." Roughly \$550 of a \$2,180 replacement sits inside management salary, and about \$650 sits inside food cost, where Chapter 11's variance analysis will find it and attribute it to the kitchen.

Exercise 17.5

At-will employment means either party may end the employment relationship at any time, for any lawful reason or no reason at all. It is the default in nearly every U.S. state, subject to real exceptions: you may not terminate for an unlawful discriminatory reason, in retaliation for protected activity, or in breach of a contract.

Language that undercuts it: "You will remain employed with us as long as your performance is satisfactory." That sentence creates an implied condition — satisfactory performance — and therefore an implied promise of continued employment while the condition holds. An employee terminated for business reasons rather than performance can point at it. Say instead: "Employment with Bellwether is at will, meaning either you or the restaurant may end it at any time, with or without cause or notice."

Exercise 17.8

Note this lands below the \$700 used for dishwashers in Figure 17.1, because the exercise omits learning-curve waste, reduced output, and the trainer's diverted attention. That gap is the answer to the exercise: a build that counts only the obvious components understates the real figure by roughly 40%, which is exactly why operators think turnover is cheap.

Exercise 17.10

Baseline from Figure 17.1: \$38,070.

Line cooks at 60%: 4 heads × 0.60 = 2.4 separations × \$2,180 = **\$5,232 (was \$8,720). Servers at 50%: 8 heads × 0.50 = 4.0 separations × \$1,500 = **\$6,000 (was \$9,000).

Change: −\$3,488 − \$3,000 = −\$6,488.

New annual total: \$38,070 − \$6,488 = \$31,582.

As a percentage of the \$500,000 labor line, the saving is \$6,488 ÷ \$500,000 = 1.30 points — which, since labor is half of prime cost, is 1.3 points of prime cost recovered without touching a wage rate, a schedule, or a menu price. For context, Chapter 11 priced a 0.5-point food-cost miss at \$5,580; this is larger than that.

Exercise 17.12

Cost of the raise: 4 cooks × 32 hrs/week × 52 weeks = 6,656 hours. At \$1.00 = **\$6,656/year**. (Fully loaded, with payroll taxes and workers' compensation, call it roughly \$7,400 — and note that it lands inside prime cost.)

Break-even on turnover cost alone: \$6,656 ÷ \$2,180 = 3.05 avoided separations.

Current line-cook turnover is 100% on 4 heads = 4.0 separations a year. To pay for itself, the raise must reduce that to fewer than 1 separation a year — a drop from 100% to about 24%.

The verdict, and it matters: on turnover cost alone, the raise almost certainly does not pay. A drop from 100% to 24% is not a realistic single-lever outcome.

But that is the wrong test, and a student who stops there has missed the chapter. The raise must be evaluated against the full consequence set: reduced vacancy (Chapter 14 showed eleven tables past standard when the line runs short), reduced learning-curve food waste, better retention of the trained cooks who currently absorb the coverage, and the recruiting time it frees. It should also be tested against alternatives — \$6,656 buys a referral program, a bench, and paid stages several times over. The disciplined answer is: price the raise, price the alternatives, and choose; do not justify a wage increase on turnover arithmetic alone, because it usually will not carry it.

Exercise 17.14

• Per cover: \$38,070 ÷ 36,140 covers = **\$1.05 a cover.** (For scale: Chapter 6 priced rent at \$2.63 a cover and Chapter 8 priced permission and risk transfer at \$1.35.)
• Percentage of revenue: \$38,070 ÷ \$1,550,000 = 2.46%.
• Points of prime cost: turnover cost sits inside labor and food cost, both of which are prime. \$38,070 ÷ \$1,550,000 = 2.5 points of prime cost — against a plan target of 60.0% with a 60.4% miss already priced by Chapter 11 at half a point.

The comparison worth making out loud: turnover costs more per cover than the liquor license, the insurance schedule, and every permit combined (Chapter 8's \$1.35), and nobody budgets for it.

Exercise 17.16

A rewrite containing all six components:

Line Cook — Sauté and Pasta · Bellwether

The restaurant. Bellwether is a 68-seat neighborhood restaurant in the Rivermill District built around a wood-fired hearth and a short seasonal menu. Dinner Tuesday through Saturday, brunch on weekends. Small room, small team, real cooking.

The work. You will run the sauté and pasta station on a five-night dinner service averaging 95 covers, executing hand-cut pasta, sauces, and the vegetable entrée, plus every brunch egg. You will prep your own station. The line is four people at peak, so you will also cover the plancha when it is buried.

The schedule. Five shifts, Tuesday through Saturday, 2:00 p.m. to close (about 11:30). Weekends required. Sunday and Monday off, every week.

The pay. \$21–24 an hour depending on experience, plus shift meal. Paid four-hour stage.

Required: you can hold a station on a busy night and keep it clean. Not required, and we will teach it: the hearth, our pasta, and anything about wood.

The process: we reply within three days. A conversation, then a paid four-hour stage, then a decision within a week. We will tell you either way.

Note what each element does: the schedule paragraph filters out anyone who cannot work weekends before they apply; the stated range costs you no candidate you wanted; the required/teachable split widens the pool without lowering the bar; and the process paragraph signals a restaurant that runs on systems.

Exercise 17.19

Five behavior-based questions for a bartender, each mapped to a duty:

1. (speed and organization) "Walk me through your set-up for a Friday. What is in your well, and what do you do at 4:30 that makes 8:30 possible?"
2. (accuracy and cost control) "Tell me about a time you realized you were over-pouring. How did you find out and what changed?"
3. (responsible service — the critical one) "Describe the last time you cut someone off. What did you say, and what happened next?"
4. (guest recovery) "Tell me about a drink you sent back out. What was wrong and how did you handle the guest?"
5. (working the pass) "You have a full bar, three service tickets, and a guest waiting to order. Describe the last time that actually happened and what order you did things in."

A 5-out-of-5 on question 3 describes a specific incident, a calm and early intervention rather than a confrontation at the end, an offer of food or water or a ride, notification of a manager, and no defensiveness about having done it. A 1 treats the question as hypothetical or implies they never have.

A 5-out-of-5 on question 2 names a specific mechanism — a variance report, a count, a manager's observation — rather than "I noticed I was heavy." Chapter 15 priced a half-ounce over-pour at \$7,631 a year; a bartender who has never thought about it is telling you something.

Exercise 17.22

A structured server working interview:

Tasks (identical for every candidate): shadow a section for one hour during early service; then take two tables under supervision, from greet through order entry; then run food for thirty minutes; then reset and polish during close.

Three things to watch:

1. The greet and the read. Do they look at the table before speaking? Do they adjust for a couple on a date versus a four-top of regulars? This is Chapter 23's material arriving early.
2. Order entry accuracy under a small load. Two tables is not pressure; errors here predict errors under real pressure.
3. What they do with idle hands. A server standing still during a lull is telling you what they will do on a Tuesday.

The correction: partway through, correct how they carried plates or where they stood at the table. Watch the next ninety seconds — the same test as §17.5, for the same reason.

What each reveals: the greet tests hospitality instinct, which is difficult to train. Accuracy tests care. Idle hands test initiative. The correction tests how coachable they will be for two years.

Exercise 17.25

Why the unpaid stage creates exposure. Under the general FLSA framework, the question is whether the work primarily benefits the employer or the worker. A candidate who spends four hours doing prep that you would otherwise have paid a cook to do has produced value you captured and sold. The label on the shift does not change that analysis, and neither does the candidate's willingness — wage-and-hour rights generally cannot be waived by agreement.

There is a narrow space for genuine observation: a candidate watching service, touching nothing, producing nothing. The moment they pick up a knife, the analysis changes.

The safe practice: pay for the stage. Four hours at Bellwether's fully loaded \$22 = **\$88. That is 4.0%** of the \$2,180 replacement cost, and it purchases both a far better hiring signal and a defensible position. It also tells the candidate something true about how the restaurant treats people — at the exact moment they are deciding whether to work for you.

Exercise 17.28

The business case against blanket criminal-record screening, separate from the legal case:

1. It shrinks the pipeline in a market where the pipeline is the constraint. Case Study 1 showed what happens when supply tightens. An operator who eliminates a substantial candidate population by policy has made a self-inflicted version of the same problem.
2. It screens on something weakly related to the job. The relevant question is whether a specific history bears on a specific role. A conviction unrelated to the duties predicts little about whether someone can hold sauté on a Saturday.
3. It forgoes a retention advantage. Employers who hire from populations with fewer options frequently report longer tenure — and §17.1 showed that tenure is worth \$2,180 a head.
4. It cedes a genuine strength of the industry. Restaurants have long been one of the few reliable routes into stable work for people with interrupted histories. That is a real social good and it is also a real competitive advantage in hiring, and abandoning it costs both.

The disciplined policy is an individualized assessment — which is, not coincidentally, close to what fair-chance law requires: does this offense, of this age, bear on the duties of this job?

Exercise 17.31

A thirty-day onboarding plan for a server, with named owners: