Chapter 4: The Business Plan: Financial Projections, Market Analysis, and Convincing Someone to Fund Your Dream

"Nobody ever declined a loan because the forecast was too low. They decline because they cannot tell where the number came from." — constructed; the only complaint a lender actually has

Overview

You are going to write a document that asks someone to hand you three hundred thirty-five thousand dollars against a business that does not exist, in an industry where roughly a quarter of new entrants do not reach their first anniversary. You have a concept (Chapter 2), you have a brand (Chapter 3), and you have a spreadsheet that says the restaurant will do $1,550,000 in its first year. The spreadsheet is where the trouble starts.

Here is the thing nobody tells you about business plans. A plan is not a prediction. You are not being asked to be right — nobody, including the lender, believes anyone can forecast a restaurant's first year to the dollar. You are being asked to demonstrate that you know which parts of your forecast are load-bearing, what each one rests on, and what you would do on the Monday you discovered it was wrong. A plan is an argument submitted to a skeptical reader, and it wins or loses on whether the reader can trace every number back to something you can defend out loud.

That reframing changes what a good plan looks like. The most common failure I see is not a plan with bad numbers. It is a plan with unattributed numbers — a revenue line that appears from nowhere, a labor percentage lifted from an industry article, a food cost of 30% asserted before a single recipe has been costed. Every one of those numbers might turn out to be right. None of them can be argued with, which means none of them can be believed.

So we are going to do this in a specific order, and the order is the lesson. We build the revenue forecast from the bottom up, out of physical facts — seats, turns, a check average, a calendar. We write down every assumption in a register, with its dollar exposure attached, so that our beliefs are visible enough to be attacked. We push the forecast through a three-year profit-and-loss statement that foots line by line. We break it on purpose, to find out which assumption fails first and what it costs. And only then — last, after everything else exists — do we write the one page that most readers will actually read.

By the end of this chapter the running project has real numbers for the first time. It also has an honest, named, unresolved gap between the number the plan puts on its cover and the number the base arithmetic produces. We are not going to hide that gap. It is the most useful thing in the chapter.

In this chapter, you will learn to:

• Name the five readers of a restaurant plan and write for the one whose objection would kill you.
• Lay out the document's standard anatomy and say, for each section, what question it closes.
• Build a bottom-up sales forecast from seats, turns, check, and days — and cross-check it with two independent top-down measures before you believe it.
• Construct an assumptions register: every belief the plan rests on, its confidence, its dollar exposure per point of movement, and the chapter that tests it.
• Assemble a three-year pro forma whose columns actually add up, and explain why year one is the structurally hardest of the three.
• Run a sensitivity analysis, rank the plan's risks by dollars, and identify what breaks first.
• Write an executive summary that states the ask, the repayment argument, and the three risks you cannot resolve — and understand why naming the risks strengthens rather than weakens it.

Learning Paths

🏗️ Opening — this is your chapter. Do §4.3, §4.4, and §4.6 with your own numbers, not Bellwether's. If you write only one thing from this book, write the assumptions register. 📋 Managing — you will spend your career being measured against assumptions someone else wrote down. §4.4 and §4.6 teach you to read a budget as a set of claims rather than a set of orders, and to find the one line that will get you fired. 🍸 Beverage — the 28% beverage mix in §4.3 and the 22% pour cost in §4.5 are two of the plan's quietest assumptions and two of its cheapest levers. §4.6 shows why a check miss driven by beverage costs more margin than the same miss driven by food. 🚚 Small Format — the document is shorter but the discipline is identical, and the sensitivity is sharper: a smaller fixed-cost base means less cushion between a missed assumption and a closed business. §4.6 matters more to you, not less.

4.1 Who reads a restaurant business plan, and what each reader looks for

Start with the reader, because the reader determines the document.

A business plan is a written argument that a specific business, in a specific place, run by specific people, will generate enough revenue to cover its costs, repay its capital, and survive the things that go wrong — supported by evidence a skeptical outsider can check. That is the whole definition, and every word in it is doing work. Specific excludes the generic. Evidence excludes enthusiasm. Survive the things that go wrong is the part most first plans omit entirely.

Nobody reads all thirty pages. Everybody reads the first one and the spreadsheet.

FIGURE 4.1 — Five readers, five questions                       [constructed teaching example]

  READER                WHAT THEY ACTUALLY ASK               WHAT THEY READ FIRST
  ─────────────────────────────────────────────────────────────────────────────────────
  The lender            "Does this repay me — and what        Executive summary → the
                        happens to my money if it             financials → your résumé →
                        doesn't?"                             your personal financial statement

  The landlord          "Will this pay rent for ten years     Executive summary → the
                        and hand my space back better         capital behind it → the concept
                        than it came?"

  The investor          "What do I get, when do I get it,     Executive summary → the ask →
                        and what exactly am I risking?"       how anyone ever gets out

  The equipment lessor  "Is the asset worth the paper,        Credit application → the
                        and is your credit clean?"            equipment schedule

  You and your partner  "Is this actually a business, and     All of it. Twice. Then again
                        do the two of us agree on what        every quarter for ten years.
                        it is?"
  ─────────────────────────────────────────────────────────────────────────────────────
  Write for the lender. Everyone else is reading a subset of what the lender demands.

Read the figure again and notice the asymmetry. Four of these readers want to know what they get. Only one of them — the lender — is primarily interested in what happens when the plan is wrong. That is why you write for the lender: satisfy the reader who is professionally paid to imagine your failure, and the others are already handled.

The lender

For a startup restaurant in the United States, the realistic debt source is a bank making a loan under the Small Business Administration's 7(a) guarantee program, which Chapter 5 covers properly. What matters here is how that reader thinks.

A lender is not evaluating your restaurant. A lender is evaluating a stream of monthly payments and asking three questions in a fixed order: Can the business service the debt from operations? If it can't, what secures the loan? And if the collateral is worthless — which, for a restaurant, it very nearly is — who else is on the hook?

That third question is why restaurant lending is hard and why the phrase personal guarantee is going to follow you through this book. A used hearth, a leasehold improvement, and a walk-in bolted to somebody else's building are not assets a bank enjoys repossessing. So the loan is underwritten substantially on the projections and on you.

Which means: the credibility of your forecast is your collateral. A lender who cannot trace your revenue line to a physical constraint — seats, hours, a check average grounded in an actual menu — has nothing to lend against but optimism.

The landlord

The landlord is a lender too, in every respect that matters. A ten-year lease at $95,200 a year is a $952,000 obligation before a single escalation, and the landlord is deciding whether to extend you that credit. They read your plan to answer one question: can this tenant pay rent through a bad winter? Secondarily, they want to know that the money to build out their space exists, because a half-finished build-out is a landlord's nightmare — a dark box, no rent, and a demolition bill.

The investor and the friends-and-family round

An equity reader asks a different question than a debt reader: not "am I repaid?" but "what is my share worth, and how would I ever convert it to money?" Restaurants are famously bad at answering the second half. There is rarely a buyer for a minority stake in a single-unit independent.

Chapter 5 handles the mechanics. The relevant point for the document is this: if you are raising money from anyone other than a bank, the plan stops being just a persuasive document and starts having legal weight.

⚖️ Code and Compliance

A projection shown to an investor is not the same thing as a projection shown to a banker.

When you hand a bank a forecast, you are making a credit representation. When you hand a private individual a forecast and ask them to buy a piece of the business, you may be offering a security — and the sale of securities is regulated at both the federal and state level in the United States, with exemptions that have conditions attached.

Practical guardrails, none of which is legal advice:

• Label projections as projections. Never present a forecast as a guarantee, a promise, or an expected return. Write down the assumptions behind it and hand them over with the numbers.
• Disclose the downside in writing. The risk factors section is not boilerplate; it is the part that protects you if the business fails and someone asks why they were not told.
• Do not promise a return, a distribution schedule, or a buyout. Especially not verbally, especially not to family.
• Get a lawyer before you accept a dollar from anyone who is not a bank. The cost of the consultation is trivial next to the cost of an improperly documented raise.

Requirements vary by state, by the number and type of investors, and by how the offer is made. Verify locally, with counsel, before you circulate anything.

Yourself, and your partner

The reader everyone forgets. A plan is the only document in which two partners are forced to write down what they each believe about the same business — how many covers, at what check, with how many people on the floor, and what happens if February is bad.

I have watched more partnerships break on an unstated assumption than on an argument. The chef believed the kitchen would run with four; the front-of-house partner assumed six. Both had been nodding at each other for eight months. The plan is where that discovery is cheap.

⚠️ Where the Money Leaks

The plan you buy instead of writing.

There is an industry of template plans, fill-in-the-blank financial models, and consultants who will produce a bound document in ten days for a few thousand dollars. Some of them are competent. All of them have the same defect: the numbers are not yours, so you cannot defend them.

Watch what that costs. You sit down with a lender and they ask why the labor line is 32.3%. If you wrote it, you say: "It's built from a staffing guide — four servers on a Friday, three on a Tuesday, a salaried chef and sous, and here's the week." If you bought it, you say: "That's the industry benchmark." One of those answers ends the meeting well.

The real cost is not the fee. It is the three to five months a declined application costs you — months of a rising build-out estimate, a landlord's patience, and in the worst case a signed lease generating rent against a dark room. At Bellwether's $95,200 of annual occupancy, four dark months is \$31,733 of rent against zero revenue, and that is before you count the pre-opening payroll you carried to hold your key people.

Buy help with the format if you want it. Never buy the numbers.

4.2 The anatomy of the document, section by section

A restaurant business plan is a conventional document, and the convention is a feature: a lender who reads twenty plans a quarter can find what they need in yours in ninety seconds, and that is worth more to you than originality. Length runs roughly twenty to thirty pages of narrative plus a financial workbook and an appendix. Longer is not better. A forty-page plan signals that the writer could not decide what mattered.

Here is the standard shape, with what each section must actually settle and where this book builds it.

Four of those fifteen are this chapter's: the executive summary, the use of funds, the pro forma, and the two instruments — register and sensitivity — that make the pro forma believable.

Use of funds

Use of funds is the schedule that states exactly what the money will be spent on, line by line, totaling the amount you are asking for. It is the least glamorous page in the plan and the one a lender turns to second, because it answers the question underneath every credit decision: what happens to my money the moment it leaves the bank?

Here is Bellwether's, which is also the plan's total project cost.

Check it: $310,000 + $185,000 + $45,000 + $35,000 + $45,000 = **$620,000**. It foots. That sounds like a trivial observation. It is not — a use-of-funds table that does not sum to the ask is one of the most common reasons a package gets sent back, and the reader who finds it stops trusting every other table in the document.

Two things about this schedule are worth naming now.

The working-capital reserve is a use of funds, not a leftover. Chapter 1 named undercapitalization as the most common first-year killer, and the mechanism is precisely this: operators budget to open and not to operate. Bellwether's plan carries \$45,000 for the months when revenue does not yet cover payroll. Whether $45,000 is enough is a live question this plan will have to defend, and Chapter 33 answers it with a thirteen-week cash forecast rather than a feeling.

Construction is the least reliable line on the page. It is an estimate built from square-footage rules of thumb and equipment list pricing, and there is no contractor bid behind it because there is no signed lease and no drawn plan. Every plan I have read underestimates build-out. Chapters 6 and 7 will test this number hard, and you should expect it to move.

The sources side — where the $620,000 comes from — is the capital stack, and Chapter 5 builds it properly. The plan's ask is **\$335,000** of SBA 7(a) debt inside that $620,000 project.

What goes in the appendix, and why it matters more than the narrative

The appendix is where a plan becomes checkable. It should contain, at minimum: résumés for both partners; a personal financial statement for each; the draft menu with prices; contractor bids or quotes as soon as they exist; the letter of intent or lease term sheet; equipment quotes; and any comparable-market data you actually gathered rather than asserted.

That last item is the one that separates plans. A market claim in the narrative — "the Rivermill District supports a $46 dinner check" — is an assertion. The same claim with four competitor menus in the appendix and a note of the dates you visited is evidence. It costs a Saturday. It is the cheapest credibility in the entire document.

4.3 Building a sales forecast bottom-up: seats × turns × check × days

Now the number everything else depends on.

A bottom-up sales forecast builds revenue from the physical constraints of the business — how many seats you have, how many times each one is occupied, what each guest spends, and how many services you run — rather than from a share of some market total. The opposite approach, the one that sinks plans, is the top-down forecast: the metro area spends $400 million a year on restaurants; if we capture four-tenths of one percent, that is $1.6 million. That sentence is not a forecast. It is a wish with a decimal point in it, and any competent reader will recognize it instantly.

The bottom-up build has four variables, which Chapter 1 introduced:

$$\text{Revenue} = \text{seats} \times \text{turns} \times \text{average check} \times \text{services}$$

Every one of those four is physically bounded. You cannot seat more people than you have chairs. You cannot turn a table faster than the kitchen can fire and the guests can eat. The check average is capped by your own menu prices. And the services are on a calendar. The virtue of a bottom-up forecast is that each variable can be argued about separately, which is exactly what you want, because a reader who disagrees with your turns assumption can tell you so without rejecting the whole document.

The Bellwether build

Bellwether has 68 seats — 56 in the dining room and 12 at the bar — and serves dinner Tuesday through Saturday plus brunch on Saturday and Sunday. That is five dinner services and two brunch services a week.

🧮 Run the Numbers

The base case, built from four variables.

Dinner. 68 seats × 1.4 turns = 95 covers a night. At a $46 average check that is $4,370** per service. Five dinner services a week: **$21,850.

Brunch. 110 covers per service at a $24 average check = **$2,640 per service. Two brunch services: $5,280**.

Weekly revenue: $21,850 + $5,280 = $27,130. Annualized over 52 weeks: $1,410,760.

Note what the brunch line implies. 110 covers on 68 seats is 1.62 turns — faster than dinner, which is correct: brunch tables turn in fifty or sixty minutes against ninety at dinner. If you write a brunch cover count without checking what turn rate it implies, you can very easily assume a number the room physically cannot produce. Always convert covers back into turns and ask whether the room can do it.

Annual covers: dinner 95 × 5 × 52 = 24,700; brunch 110 × 2 × 52 = 11,440; total 36,140. Blended average check across the whole business: $1,410,760 ÷ 36,140 = **$39.04**.

Hold that number: $1,410,760. It is the base case, and it is what the four physical variables produce with nothing else assumed.

The plan's headline revenue for year one is \$1,550,000.

The difference is $139,240** — about **$2,678 a week, or roughly 9.0% of the plan. That gap is real, it is deliberate, and we are going to spend the rest of this chapter refusing to hide it.

🧾 Read the Numbers

```text FIGURE 4.2 — "The forecast worksheet" [the Bellwether plan] THE ARTIFACT The year-one revenue build, as it would appear on a single tab of the plan's financial workbook. Pre-opening; no operating history exists. THE CONTEXT A 68-seat neighborhood American restaurant, dinner Tuesday–Saturday plus Saturday and Sunday brunch, in a gentrifying former warehouse district of a mid-size Midwestern metro.

                 DINNER
                   Seats                                            68
                   Turns per service                               1.40
                   Covers per service                                95
                   Average check                                 $46.00
                   Revenue per service                          $4,370
                   Services per week                                 5
                   Revenue per week                            $21,850

                 BRUNCH
                   Covers per service                               110   (= 1.62 turns)
                   Average check                                 $24.00
                   Revenue per service                           $2,640
                   Services per week                                 2
                   Revenue per week                             $5,280
                 ─────────────────────────────────────────────────────────
                 TOTAL REVENUE PER WEEK                        $27,130
                 × 52 weeks                                 $1,410,760
                 ─────────────────────────────────────────────────────────
                 PLAN, YEAR ONE                              $1,550,000
                 UNEXPLAINED BY THIS WORKSHEET                 $139,240   (9.0%)

WHAT IT SHOWS Four physical variables produce $1,410,760. Each is separately arguable: 68 seats is a fact, 1.4 turns and $46 are beliefs, 52 weeks is a calendar assumption. The worksheet is honest about what it does and does not generate — and it does not generate the plan's headline number. WHAT IT DOESN'T It says nothing about the seasonal patio, private events, takeout, or any revenue outside the dining room on a normal week. It assumes the restaurant operates all 52 weeks. It assumes a flat year with no ramp — no slow opening months, no February. And it contains no evidence at all for 1.4 turns or a $46 check; those are the two numbers a reader will attack first. THE DECISION Do not adjust the worksheet upward to reach $1,550,000. Instead, itemize the $139,240 as named revenue sources with their own arithmetic, put every one of them in the assumptions register, and let a reader accept or reject them individually. THE LESSON A forecast that arrives at a round number is a forecast that was worked backwards. Build the number you can defend, then show your work for the distance between it and the number you want. ```

Closing the gap honestly: the revenue bridge

The plan's $1,550,000 is not arbitrary and it is not fraudulent. It reflects sources of revenue the base worksheet deliberately excludes — the seasonal patio, private events, and an off-premise channel. The correct move is not to nudge turns up until the arithmetic lands on the number you wanted. It is to build a bridge: an itemized reconciliation from the base case to the headline, where every step is a separate claim a reader can accept or reject.

FIGURE 4.3 — From the base case to the plan: the $139,240 bridge      [the Bellwether plan]

  BASE CASE  (68 seats × 1.4 turns × $46 × 5 dinners
              + 110 brunch covers × $24 × 2 services, × 52)      $1,410,760  ██████████████████

  + patio, dinner     20 wks × 5 svc × 12 covers × $46             +$55,200  ██
  + patio, brunch     20 wks × 2 svc × 10 covers × $24              +$9,600  ▌
  + private events    14 events × $3,000                           +$42,000  █▌
  + takeout           52 wks × $600                                +$31,200  █
  ────────────────────────────────────────────────────────────────────────────────────────
  SUBTOTAL                                                        $1,548,760
  + rounding, to reach a headline number                             +$1,240
  ────────────────────────────────────────────────────────────────────────────────────────
  PLAN, YEAR ONE                                                  $1,550,000  ███████████████████

  All four bridge items are assumptions with zero operating history behind them.
  Together they are 9.0% of the plan's revenue — and 27% of its operating profit.

Walk through it. The patio is sixteen seats that exist in the frozen concept and were left out of the base worksheet entirely; the bridge claims twelve incremental dinner covers a service across a twenty-week season, which is deliberately less than the sixteen seats could hold, because some patio guests would have sat inside anyway. That subtraction — the cannibalization haircut — is the sort of thing a reader notices and rewards. Private events claims fourteen buy-outs or large private parties a year at $3,000 net incremental each. **Takeout** claims $600 a week, which is roughly twelve orders a night at a $40 ticket — modest, and Chapter 28 will test whether it is incremental or simply moves dining-room guests to a lower-margin channel.

And then the last line. The bridge lands at $1,548,760**, not $1,550,000. The remaining $1,240** is not a revenue source. It is the plan being rounded up to a tidy headline.

Say so. Put it in the register. A reader who finds a $1,240 rounding item you disclosed thinks you are careful; a reader who finds one you didn't disclose starts checking every other number in the document, and they will be right to.

There is more than one bridge, and this matters. The same $139,240 could come entirely from the four core variables: dinner turns at 1.55 instead of 1.40 adds 10 covers a service — $460 per service, $2,300 a week, **$119,600 a year — and brunch at 118 covers instead of 110 adds $192 a service, $384 a week, **$19,968 a year. That combination totals $139,568, which closes the gap and overshoots it by $328, landing at $1,550,328.

Two different stories, one number. The plan has to pick one and be tested on it. A forecast that cannot say which bets it is making has not actually been made.

👨‍🍳 On the Line

Where a turns number actually comes from.

You cannot look up your seat turns. You have to go get them, and the only way I know is unglamorous.

Pick the four closest restaurants in your competitive set that sit within a few dollars of your intended check. Go to each one twice — once on a Tuesday, once on a Saturday. Sit at the bar where you can see the room. Then count, on paper:

• Covers, not tables. Sweep the room at 6:00, 7:00, 8:00, and 9:00 and count occupied seats.
• One table, start to finish. Note the minute a specific table is seated and the minute the check is dropped, and again when it is bussed. That is your turn time; turns per service is roughly service hours ÷ turn time, adjusted for the fact that no room fills and empties cleanly.
• The shape of the night. A room that is full at 7:30 and empty at 8:45 has one turn. A room that is full at 6:15, full again at 8:00, and half-full at 9:30 has two.

The failure modes, all of which I have committed. One night is not a sample — a rainy Tuesday in March and a Saturday before a holiday will give you numbers 40% apart. You cannot see the reservation book, so you cannot tell a full room from a full room with a two-week wait. You cannot see the bar's covers from the dining room or the dining room's from the bar. And most importantly: their turns are not your turns. They have a menu that fires in eleven minutes and you have a hearth. Use their number as a ceiling and a reality check, never as your forecast.

What this buys you is not precision. It is the ability to say to a lender: "1.4 turns, and here is the notebook." That sentence is worth more than the number in it.

Cross-checking top-down: never build with it, always check with it

You build bottom-up. You check top-down, with two or three independent measures that have nothing to do with each other. If the bottom-up number and the cross-checks disagree badly, one of them is wrong and you need to find out which before a reader does.

| Cross-check | Base case ($1,410,760) | Plan ($1,550,000) | |---|---|---| | Sales per seat, per year | $20,746 | $22,794 | | Sales per square foot (2,800 sq ft) | $503.84 | $553.57 | | Occupancy cost as % of sales ($95,200 fixed) | 6.75% | 6.14% | | Average weekly sales | $27,130 | $29,808 |

Three observations, and each one is a lesson.

Sales per seat is the fastest sanity check in the business. A neighborhood full-service dinner house with weekend brunch landing somewhere in the low twenty-thousands per seat is unremarkable. A plan claiming $40,000 a seat for the same concept is claiming something extraordinary and had better explain it. Note that I have not given you a benchmark range to check against — I do not have one I can stand behind, and neither does most of the industry literature. Source your own comparison. Four competitor menus, a rough seat count from a visit, and a plausible turn estimate will get you a defensible neighborhood figure in an afternoon, and "here is how I derived my comparison" beats a quoted benchmark every time.

Occupancy percentage is not an input — it is an output. This is the observation operators miss most often. Bellwether's rent is $95,200 whether the restaurant does $1.55 million or $1.2 million. At the plan it is 6.1% of sales, which is excellent. At the base case it is 6.75%. At $1,200,000 it would be 7.9%. The rent does not care about your forecast, so every percentage in the plan that divides a fixed dollar cost by revenue is really a restatement of the revenue assumption. When you see "occupancy 6.1%" in a plan, read it as "occupancy $95,200, and we believe the sales number."

Average weekly sales is the number you will actually manage against. Nobody runs a restaurant in annual dollars. The plan says $29,808 a week; the base case says $27,130. That $2,678 difference is the entire bridge, restated in the unit a manager can see on a Monday morning. Chapter 31's weekly flash report is where this becomes a habit.

4.4 The assumptions register: making your beliefs visible and testable

This is the instrument the chapter is really about.

An assumptions register is a single table listing every belief the plan's numbers rest on, with — for each one — the value used, how confident you are, what the value is based on, the dollar consequence of being wrong, and where in the plan or the operation it will be tested. It is the document that converts a forecast from a claim into an argument.

Most plans do not have one. Most plans bury their assumptions in footnotes, in a paragraph of narrative, or nowhere at all — and the effect is that the reader has to reverse-engineer your beliefs from your arithmetic, which is exactly the work that makes a reader hostile. The register does that work for them, in public, on one page.

It is also the single most useful page in the plan for you, and here is why. A forecast is a conclusion. Once written, it stops being questionable — the number sits there in a bold font looking like a fact. The register keeps every input in a state where it can still be argued with, revised, and watched. A plan without a register hardens. A plan with one stays alive.

The Bellwether register

Here it is, at the state of knowledge the plan has reached by Chapter 4. Read the last two columns first; they are the ones that make it an instrument rather than a list.

Now look at what the table did.

It made A6 visible. The largest single risk in this plan is not the food cost, or the labor line, or the rent. It is the $139,240 of revenue that sits above the base case and has no operating history behind it. That is 9.0% of sales and — because roughly half of every marginal revenue dollar survives as margin — about 27% of the plan's entire operating profit. Before this table existed, that exposure was invisible; it lived in the difference between two numbers on different pages. Now it has a row, a dollar value, and three chapters assigned to test it.

It ranked the plan's ignorance. Five rows are marked Low or Lowest. Those five are where the reader's attention should go, and — this is the counterintuitive part — you want the reader's attention there. A plan that marks everything "high confidence" is telling a lender that the writer cannot tell the difference between a fact and a hope, which is precisely the disqualifying quality.

It attached dollars to points. A point of labor is $15,500. A point of food cost is $11,160, because food cost applies to food sales ($1,116,000), not to total sales. A point of pour cost is $4,340, because beverage sales are only $434,000. Those three numbers are not the same size, and an operator who treats "a point is a point" will spend a month of effort on the cheapest one. This is why the register expresses exposure in dollars rather than percentages — dollars are comparable and percentages are not.

⚠️ Where the Money Leaks

The 52-week year.

Look at row A5. The forecast multiplies weekly revenue by 52, which asserts that the restaurant operates every week of the year without exception. It will not.

Count the realistic closures: Thanksgiving and Christmas at minimum. A week in the deep summer or deep winter when the staff is given a break and the hood gets cleaned properly. A compressor that fails, a water main, a snow event that closes the street. A health-department re-inspection that costs you a service. None of these is a disaster. Together they routinely cost an independent one to two weeks of trading.

At Bellwether's $27,130 of base-case weekly revenue, **two lost weeks is $54,260 of sales — and because roughly half of a marginal revenue dollar survives as margin, about \$27,260 of operating profit.** That is more than the entire pour-cost line could lose across three full points.

The disciplined move is not to forecast 50 weeks and look conservative. It is to state the assumption explicitly — "the plan assumes 52 operating weeks; each week lost costs approximately $13,600 of operating profit" — and let the reader price it. Half of what the register buys you is not accuracy. It is the reader's confidence that you know what you assumed.

How to actually build one

Four rules, learned the hard way.

One: write the register while you build the model, not after. Every time you type a number into a cell that is not the result of a formula, that number is an assumption and it gets a row. If you wait until the model is finished, you will not remember which numbers you reasoned toward and which you simply typed.

Two: state the basis in the words you would use out loud. "Competitive-set observation only, no operating history" is a basis. "Industry standard" is not — it is a way of not saying where a number came from. If you cannot finish the sentence "I believe this because…", the confidence column says Low and you have learned something.

Three: put the exposure in dollars, per unit of movement. Not "sensitive" or "high risk." A point. A tenth of a turn. A dollar of check. The register's job is to let you and a reader sort the rows by consequence, and you cannot sort adjectives.

Four: name where it gets tested and by when. An assumption with no test attached is a permanent guess. "Ch. 19" in the last column means the labor line stops being a belief and becomes a built-up number when the staffing guide exists. After you open, the same column becomes a date: food cost verified against the first four weekly inventories; turns verified against the POS at week eight.

🔍 Check Your Understanding

1. Bellwether's food cost is 30% and its pour cost is 22%. Which is worth more to the business: one point of food cost or one point of pour cost, and by how much?
2. The register shows occupancy at 6.1% of sales in the plan and the same $95,200 in dollars. If year-one revenue comes in at $1,410,760 instead, what happens to the occupancy percentage, and what — if anything — has actually changed about the business?
3. Why does the register give assumption A6 the largest exposure in the plan when it is not the largest number in the plan?

(1: A point of food cost is 1% of $1,116,000 of food sales = $11,160; a point of pour cost is 1% of $434,000 = $4,340. Food is worth about 2.6 times as much, because the base it applies to is 2.6 times larger. 2: It rises to 6.75%. Nothing about the business has changed — the rent is identical. The percentage moved because its denominator moved, which is why occupancy percentage is an output of the revenue assumption, not an independent target. 3: Because A6 is entirely unsupported. The other rows are estimates that could be off by a point or two; A6 is $139,240 with no operating history behind any part of it, and it carries about 27% of the plan's operating profit.)

4.5 The pro forma: three years of P&L, and why year one is the hard one

A pro forma is a projected financial statement — a profit-and-loss statement, and usually a cash flow and balance sheet as well, built for periods that have not happened yet. "Pro forma" is Latin for as a matter of form, and the phrase carries the right warning: it is the form of a financial statement filled with beliefs instead of history. Everything in §4.4 exists to make those beliefs inspectable.

Restaurant convention is three years of annual projections, with year one broken out monthly. The monthly detail is not decoration. It is where the ramp, the seasonality, and the cash trouble live, and a lender who receives only annual columns will ask for the months.

Year one, on plan

🧾 Read the Numbers

```text FIGURE 4.4 — "Year one, on plan" [the Bellwether plan] THE ARTIFACT The year-one projected profit-and-loss statement as it appears in the financial section of the plan. Pre-opening; no operating history. THE CONTEXT 68 seats, five dinner services and two brunches, full bar, $95,200 of all-in occupancy on a ten-year lease not yet signed.

                 Revenue                                 $1,550,000    100.0%
                   Food sales      (72%)     $1,116,000
                   Beverage sales  (28%)       $434,000

                 Food cost       (30.0% of food sales)     $334,800
                 Beverage cost   (22.0% of bev sales)       $95,480
                 ──────────────────────────────────────────────────────────
                 TOTAL COGS                                $430,280     27.8%
                 Labor, all-in                             $500,000     32.3%
                 ──────────────────────────────────────────────────────────
                 PRIME COST                                $930,280     60.0%
                 Occupancy                                  $95,200      6.1%
                 Other operating                           $217,000     14.0%
                 General & administrative                   $46,500      3.0%
                 ──────────────────────────────────────────────────────────
                 OPERATING PROFIT                          $261,020     16.8%
                 Debt service (SBA note + equipment lease)  $69,500      4.5%
                 ──────────────────────────────────────────────────────────
                 BEFORE TAX AND DISTRIBUTIONS              $191,520     12.4%

WHAT IT SHOWS A plan that hits its central target: prime cost at 60.0%, the benchmark Chapter 1 established for full service. Every line foots. Occupancy at 6.1% is genuinely advantaged — a former warehouse district eight years into gentrification is exactly where that advantage comes from — and it is the single best decision visible in this plan. WHAT IT DOESN'T It is an annual column. It cannot show the opening ramp, the February trough, or the week payroll and the sales-tax remittance land together; only the monthly detail and Chapter 33's cash forecast do that. It does not show that $139,240 of the revenue line has no operating history behind it. And a 16.8% operating profit sits well above the 3–10% Chapter 1 gave as typical for full-service independents — because every line here is at the favorable end of its range simultaneously, which is what a plan is. THE DECISION Publish it with the register attached and the sensitivity table on the facing page. Do not present the 16.8% as an expectation. Present it as what happens if fifteen assumptions all land, and show what happens when they don't. THE LESSON A pro forma is only as good as the weakest row in its register. The statement is the conclusion; the register is the argument; a reader who is handed the conclusion alone is entitled to disbelieve it. ```

Check the arithmetic yourself, because the book's own rule applies to the book: food $1,116,000 + beverage $434,000 = $1,550,000. COGS $334,800 + $95,480 = $430,280. Prime cost $430,280 + $500,000 = $930,280. Costs below the line: $930,280 + $95,200 + $217,000 + $46,500 = $1,288,980. And $1,550,000 − $1,288,980 = **$261,020**. It foots.

FIGURE 4.5 — Where the dollar goes, Bellwether on plan               [the Bellwether plan]
  food & beverage cost   ███████████                27.8¢   ┐
  labor (all-in)         █████████████              32.3¢   ┘ PRIME COST = 60.0¢
  occupancy              ██                          6.1¢
  other operating        █████                      14.0¢
  general & admin        █                           3.0¢
  ─────────────────────────────────────────────────────────
  = operating profit     ██████                     16.8¢    before debt service and taxes
  − debt service                                     4.5¢    $69,500 on $1,550,000
  ─────────────────────────────────────────────────────────
  = before tax and distributions                    12.3¢

  Rounded components sum to 60.1¢ of prime cost; the exact figure is 60.0%.
  Say so in a footnote rather than letting a reader find it.

Compare that picture with Figure 1.2, the generic full-service dollar: 30¢ of COGS, 34¢ of labor, 8¢ of occupancy, 17¢ of other operating, 4¢ of G&A, leaving 7¢. Bellwether's plan is better on all five lines — 2.2 points on COGS, 1.7 on labor, 1.9 on occupancy, 3.0 on other operating, 1.0 on G&A. Nine and a half points, all at once.

Each of those five is individually defensible. A low rent in a neighborhood on the way up is a real advantage. A hearth-driven menu with a tight item count genuinely can run 30% food cost. A bar carrying 28% of sales at a 22% pour cost really does pull the blended number down. But five favorable assumptions landing simultaneously is not a forecast; it is what a plan looks like, and the honest response is not to discard them. It is to hand the reader §4.6 on the next page.

Years two and three

¹ percent of food sales · ² percent of beverage sales · all other percentages are of total revenue. Percentages are rounded to one decimal and may not sum exactly.

Foot the year-two column: $477,472 + $541,800 + $95,200 + $240,800 + $51,600 = $1,406,872, and $1,720,000 − $1,406,872 = $313,128**. Year three: $504,310 + $573,500 + $95,200 + $259,000 + $55,500 = $1,487,510, and $1,850,000 − $1,487,510 = $362,490**. Both foot.

Now — where does the improvement actually come from? Notice that it is barely coming from cost control at all.

• COGS holds flat for two years and improves half a point in year three. That is honest. A restaurant does not get dramatically better at buying; it gets a little better, and inflation eats most of it.
• Labor improves 1.3 points across three years, which is a modest claim: a trained team wastes fewer hours, but wages rise, so most of the gain is arithmetic — a fixed management floor spread over more sales.
• Occupancy falls from 6.1% to 5.1% without a single dollar changing. The rent is $95,200 in all three columns. The entire improvement is revenue growth dividing a fixed number.
• Other operating and G&A are held at constant percentages, which is a modeling convenience and should be flagged as one.

So roughly two-thirds of the profit improvement across three years is operating leverage — fixed costs spread across a bigger number — rather than anything anyone did well. Chapter 32 formalizes this and shows that leverage runs violently in both directions. It is worth understanding now, because a reader who sees a plan where profit doubles on cost savings will not believe it, and a reader who sees profit rise mostly on leverage will.

One line in that table is a known problem: occupancy is held flat for three years, and it will not be. A ten-year lease contains an escalation clause and NNN charges get reconciled annually. The plan cannot model it yet because the terms are not negotiated — which is exactly why row A12 of the register exists, flagged Medium-Low with "no escalation modeled" written in plain sight. When Chapter 6 produces real lease terms, this row of the pro forma changes, and the register is the reason everyone will know why.

Why year one is the hard one

Three structural reasons, and they compound.

The ramp. The annual column assumes fifty-two identical weeks. Year one is not that. There is an opening spike — three to eight weeks of curiosity traffic that will overstate what the business actually does — followed by a normalization downward that inexperienced operators read as a catastrophe. Then a slow build as the neighborhood forms a habit. A plan that spreads $1,550,000 evenly across twelve months is describing a restaurant in its second year.

The cost lines are worst when you can least afford them. Every cost assumption in the pro forma is a trained number. A 30% food cost assumes cooks who portion correctly and a walk-in that gets counted; in month two you have neither. A 32.3% labor cost assumes a schedule written against a demand forecast you do not yet have, staffed by people who are still learning the menu, with overlapping training hours on the clock. The prime cost that is achievable in month fourteen is routinely three to five points better than what actually happens in month two — and the plan shows one number for the whole year.

Cash and profit diverge hardest in year one. The pro forma says $261,020 of operating profit. It does not say when. Opening inventory is bought before a dollar of it is sold. Deposits go out — utilities, insurance, the liquor license, the POS. Payroll runs from before you open. The working-capital reserve exists precisely to bridge that gap, and whether $45,000 is the right number is a question this document cannot answer. Cash is not profit, and Chapter 33 is where the plan finds out whether it survives February.

The practical consequence: the monthly year-one schedule is not optional. Build it with a real ramp — a spike, a dip, a build — and let the annual column be the sum of twelve honest months rather than one confident division.

4.6 Sensitivity analysis: what breaks first

Sensitivity analysis is the deliberate breaking of your own forecast: you move one assumption at a time, hold everything else constant, and record what happens to the bottom line. Its purpose is not to produce a range of outcomes. It is to rank your assumptions by consequence, so you know which one to watch first, which one to defend hardest, and which one — this is the part people miss — turns out not to matter much at all.

Doing it requires one modeling step, and it is worth understanding rather than accepting.

Splitting the cost base

Costs do not all move with sales. Rent does not. The chef's salary does not. Food cost does, almost exactly. Most restaurant cost lines are somewhere in between, and to flex a forecast you have to say where.

For this chapter's purposes, we split Bellwether's year-one costs like this. The split is a planning estimate — Chapter 19 builds the labor line bottom-up from a staffing guide and Chapter 32 builds the whole structure properly for break-even work. Expect both to revise it.

Check it: fixed $252,000 + $95,200 + $124,000 + $46,500 = $517,700. Variable at plan: 49.76% × $1,550,000 = **$771,280**. Together $1,288,980, which is exactly the cost total in Figure 4.4. And $1,550,000 − $1,288,980 = $261,020. The model reproduces the plan to the dollar.

What the split buys you is a single sentence you can apply to any revenue change: every additional dollar of sales brings 49.76¢ of cost with it and leaves 50.24¢ behind. Move revenue by any amount and about half of it lands on the operating-profit line. (Chapter 32 names this properly and builds the break-even from it.)

The fixed labor floor of $252,000 deserves a sentence of its own, because Chapter 1 named it as a killer. It is the salaried chef and sous, the management, and the minimum crew that has to be in the building whether the room does 60 covers or 110. It does not flex on a Tuesday. It is why a slow service is disproportionately expensive and why the sensitivity below is as sharp as it is.

Three scenarios

🧮 Run the Numbers

Breaking the plan on purpose, one assumption at a time.

Scenario 1 — turns come in at 1.25 instead of 1.40. Dinner covers fall from 95 to 68 × 1.25 = 85. That is 10 covers a service × $46 × 5 services × 52 weeks = $119,600 of lost revenue. (We hold the bridge items constant and flex only the dining room, which is a generous simplification and should be stated as one.) Revenue: $1,430,400. Costs: COGS $397,079 · labor $252,000 + 16% = $480,864 · occupancy $95,200 · other operating $124,000 + 6% = $209,824 · G&A $46,500 = **$1,229,467. Operating profit $200,933 — 14.0% of sales. Prime cost rises to 61.4%, because the fixed labor floor is now spread over less revenue. Profit falls $60,087**.

Scenario 2 — the dinner check comes in at $43 instead of $46. Total dinner covers in the plan: 24,700 in the dining room plus 1,200 on the patio = 25,900. At $3 less each, that is **$77,700 of lost revenue. Revenue: $1,472,300. Costs: COGS $408,710 · labor $487,568 · occupancy $95,200 · other operating $212,338 · G&A $46,500 = $1,250,316. Operating profit $221,984 — 15.1% of sales.** Prime cost **60.9%**. Profit falls **$39,036**.

Scenario 3 — labor lands at 35.3% of sales instead of 32.3%. Revenue unchanged at $1,550,000. Labor becomes 35.3% × $1,550,000 = $547,150, an increase of $47,150** over the plan's $500,000. Prime cost: $430,280 + $547,150 = $977,430 = 63.1% of sales — three points above the full-service benchmark Chapter 1 established. Operating profit $213,870 — 13.8% of sales.** Profit falls **$47,150**, dollar for dollar, because nothing else moved.

Three points of labor is not an exotic scenario. It is what a new team, a new kitchen, and a schedule written without a demand history routinely produce in a first year. The register should say so, and it should say what you would do about it — which shift you would cut, at what covers, and in what week you would decide.

Ranking the damage

Now put every adverse move on one scale, in dollars, and look at the shape.

FIGURE 4.6 — What breaks first: one assumption moving alone            [the Bellwether plan]

  Bridge revenue never materializes (−$139,240)   ████████████████████   −$69,954
  Dinner turns 1.40 → 1.25                        █████████████████      −$60,087
  Labor 32.3% → 35.3% of sales                    █████████████          −$47,150
  Dinner check $46 → $43                          ███████████            −$39,036
  Other operating 14.0% → 15.5%                   ███████                −$23,250
  Food cost 30% → 32% of food sales               ██████                 −$22,320
  Pour cost 22% → 25% of beverage sales           ████                   −$13,020
  ────────────────────────────────────────────────────────────────────────────────
  Plan operating profit: $261,020 (16.8%). Each bar is ONE assumption moving alone,
  with everything else held at plan. Occupancy is not shown: on a signed lease it
  does not move, which is exactly why a good lease is worth so much.

Four things fall out of that picture, and they are the chapter's real payload.

The largest single risk in the plan is the part of the revenue that isn't in the base case. Not food cost. Not labor. The $139,240 bridge — the patio, the events, the takeout, and the rounding — carries $69,954 of operating profit, more than a quarter of the total. If none of it appears, revenue is $1,410,760 and operating profit is **$191,066, or 13.5%**. That is still a working restaurant. But it is a $70,000 swing produced entirely by assumptions that were, until §4.3, sitting invisibly in the gap between two pages.

Revenue assumptions dominate cost assumptions. The top four bars are all revenue or labor; the purchasing lines are at the bottom. This is not an argument against costing your menu — Chapters 11 through 13 exist and they matter enormously once you are open. It is an argument about where the plan's uncertainty lives. Before you open, you know almost nothing about your revenue and you can know quite a lot about your costs.

The pour-cost line is nearly irrelevant to the plan and enormously relevant to the operation. Three full points of pour cost is $13,020 — the smallest bar on the chart, because beverage sales are only $434,000. And yet Chapters 15 and 16 will spend two chapters on it. Both things are true: it is a small share of the plan's risk and a large share of what a bar manager can actually control on a Tuesday. The tornado tells you what to defend in the document, not what to manage in the building.

Not everything is sensitive, and knowing that is worth money. If a two-point food-cost miss costs $22,320 while a fifteen-hundredths miss on turns costs $60,087, then an operator agonizing over protein pricing while ignoring reservation policy has misallocated their attention by a factor of nearly three.

🧮 Run the Numbers

The refinement that most sensitivity tables miss: not all revenue is the same revenue.

Scenario 2 assumed a $3 check shortfall flowing at the blended contribution of 50.24¢. But why the check missed changes the answer.

Suppose the $3 is entirely beverage — guests are not taking the second cocktail. Beverage costs 22¢ on the dollar, not the blended 27.76¢, so more of each lost dollar was margin. The contribution lost per dollar becomes 1 − 0.22 − 0.16 − 0.06 = 56¢, not 50.24¢.

$77,700 × 0.56 = **$43,512 of lost operating profit, against $39,036 in the blended version. Operating profit: $217,508, or 14.8%** of sales — half a point worse than the naive answer.

The lesson generalizes. A dollar of beverage revenue is worth more to the bottom line than a dollar of food revenue, which is why a strong bar quietly subsidizes a kitchen and why beverage attachment is the cheapest lever on the check average. It is also why the 72/28 mix in row A7 of the register is a more consequential assumption than it looks: shift the mix toward food and the blended COGS rises even if every individual cost target is hit exactly.

The combined downside

Single-variable sensitivity is the standard tool and it has a standard defect: assumptions do not miss one at a time. The conditions that produce 1.25 turns — a slow neighborhood, a soft opening, a competitor that opened the same spring — are the same conditions that produce a $43 check and a labor line that will not come down because you cannot cut a floor that is already thin.

So run them together.

The combined revenue is built the same way as the base case, with the misses applied: 85 dinner covers at $43 × 5 × 52 = $950,300; brunch unchanged at $274,560; patio dinner at the lower check $51,600; patio brunch $9,600; events $42,000; takeout $31,200. Total $1,359,260. Costs foot to $1,204,406, leaving **$154,854**.

Two honest readings of that table, and you need both.

The optimistic reading is that the plan survives all three misses at once. Operating profit of 11.4% and $85,354 after debt service is a functioning business — not the one in the pro forma, but one that pays its people, pays its landlord, and pays its note. The reason it survives is almost entirely the occupancy line: $95,200 of rent is a genuinely good deal, and a good lease is the cheapest insurance an operator ever buys. Chapter 6 is where that gets earned.

The pessimistic reading is more useful. An 11.4% operating profit in the downside case is still above the 3–10% Chapter 1 gave as typical for full-service independents. That should make you suspicious of the base case rather than comfortable about the downside. Either this plan has found a genuinely advantaged structure — plausible, and the rent is the reason — or a line that has not been stress-tested here is wrong. The two candidates are the other-operating line at 14.0%, which is a category estimate with no vendor quotes behind it, and the fixed/variable split itself, which Chapter 19 has not yet built. Both are flagged Low in the register, which is the register doing its job.

For completeness: the model's fixed cost base of $517,700 against a 50.24¢ contribution implies the business does not cover its costs until roughly $1,030,000 of annual sales — about 34% below the plan and 27% below the base case. That is a substantial cushion, and it is the single most reassuring number in this chapter. Chapter 32 builds it properly, in covers per night, which is the form an operator can actually use.

4.7 Executive summary last: writing the page that gets read

Write it last. Write it after the forecast, after the register, after the sensitivity table, after the use of funds — because until those exist you do not know what you are summarizing, and a summary written first becomes a promise the rest of the document has to keep.

The executive summary is a one-page statement of the business, the ask, and the argument for repayment, written so that a reader who reads nothing else can repeat your case accurately to someone else. That last clause is the actual specification. A loan officer will present your file to a credit committee that has not read it. Your executive summary is the script they will use. Write it for that room.

FIGURE 4.7 — The one page, in nine moves                        [constructed teaching example]

  1  THE CONCEPT       One sentence. What it is, where it is, who it is for.
  2  THE MARKET        One claim with the evidence in a clause, not a paragraph.
  3  THE TEAM          Who runs it, what they have actually done, and — say it — the gap.
  4  THE MODEL         Seats, services, check, the revenue number and where it comes from.
  5  THE ECONOMICS     Prime cost, occupancy, operating profit. Four numbers, no more.
  6  THE ASK           The amount, the instrument, and what it buys.
  7  THE STACK         Where the rest of the money comes from, including yours.
  8  THE REPAYMENT     How the debt gets serviced, and from what.
  9  THE RISKS         The three you cannot resolve, each with what you would do.
  ──────────────────────────────────────────────────────────────────────────────
  One page. Two at the absolute outside. If it needs three, move something to the
  body — the body is where the reader goes when the summary made them curious.

Move 9 is the one that separates plans, and it is counterintuitive enough to be worth arguing for.

Every instinct tells you to keep the risks out of the summary. The summary is where you sell. Why would you hand a skeptical reader your three worst problems on the first page?

Because the reader is going to find them anyway, and what happens next depends entirely on whether you found them first. A lender who discovers an unnamed risk on page nineteen does not conclude that you missed it. They conclude either that you did not see it — which makes you unqualified — or that you saw it and left it out, which makes you untrustworthy. Both readings discount every other number in the document. A risk you named yourself, with a response attached, does the opposite: it tells the reader that the parts of the plan you didn't flag are probably sound.

There is a craft to it. A named risk needs three things: the risk, its dollar consequence, and the specific action you would take. "Labor cost may run higher than projected" is not a risk statement; it is a hedge. "Labor at 35.3% instead of 32.3% costs $47,150 and takes prime cost to 63.1%; the staffing guide in the operations section shows which shift comes out at which cover count, and we would act in the first week the four-week average holds below plan" is a risk statement. One of those makes a lender more comfortable.

🤝 Hospitality

The second visit is a line in your forecast, whether you wrote it down or not.

A plan is a financial document, and it is easy to write one in which hospitality never appears — where the product is food, the revenue is covers times check, and the guest experience is a paragraph in the concept section that no reader takes seriously.

But look at what the forecast actually assumes. 36,140 covers in year one, in a trade area that does not contain 36,140 different people who will pay $46 for dinner. Bellwether's revenue assumption is, structurally, a repeat-visit assumption. The 1.4 turns in row A1 are not produced by marketing; they are produced by a neighborhood forming a habit, and habits are formed by rooms that people want to be in again.

Chapter 1 made the economic case: a first visit is expensive to acquire and a second costs nothing, so the second visit is where the margin actually lives. Chapter 23 works the arithmetic properly. The point for the plan is narrower and sharper: when you assume 1.4 turns, you are assuming a repeat-visit rate. Say so in the register. A lender who has read enough restaurant plans has seen a hundred forecasts that assumed a full room and none that explained why anyone would come back — and the one that explains it reads as though a professional wrote it.

⚠️ Where the Money Leaks

The summary that hides the number.

Three specific ways executive summaries destroy value, all of them common and all of them free to fix:

• Burying the ask. If a reader has to hunt for the amount, the instrument, and the term, you have told them you are uncomfortable asking. Put it in a sentence in the top half of the page.
• A revenue number with no derivation. "$1,550,000 in year one" on its own is an assertion. "$1,550,000 — 68 seats at 1.4 turns and a $46 check across five dinners and two brunches, plus a seasonal patio and a modest events and takeout channel" is the same number with a spine. It costs you twenty-eight words.
• Omitting the owners' compensation. A plan in which the partners work for nothing produces a profit line that will not survive contact with reality, and an experienced reader will spot it in seconds and reprice everything. Bellwether's \$500,000 labor line includes both partners at market for their roles — which is row A11 of the register, flagged deliberately, because a plan that quietly omits it is overstating operating profit by whatever those two salaries would have been.

🔍 Check Your Understanding

1. Why is the executive summary written last, and what specifically goes wrong when it is written first?
2. A reader asks why your plan discloses a $1,240 rounding adjustment that is 0.08% of revenue. Give the answer in two sentences.
3. The plan's occupancy improves from 6.1% to 5.1% of sales between year one and year three without a dollar of rent changing. Is this a cost improvement? What is it, and what does it imply about what happens if revenue disappoints?

(1: Because the summary is a summary of conclusions the rest of the document has not yet reached; written first, it becomes a target the forecast is then reverse-engineered to hit, which is exactly the failure mode §4.3 warns about. 2: Because the credibility of every other number in the document depends on the reader believing you disclose small things; a rounding item they find themselves costs far more than one you hand them. 3: No — nothing improved. It is operating leverage: a fixed dollar cost divided by a growing revenue number. It implies the reverse runs too — if revenue disappoints, occupancy percentage rises with no decision by anyone, which is why the same lease is a cushion in a good year and a weight in a bad one.)

🍽️ The Business Plan

Checkpoint 4 of 40 — the plan gets numbers.

Chapter 2 gave the plan a market and Chapter 3 gave it a face. This chapter gives it arithmetic, an executive summary, and — most importantly — a written record of everything the arithmetic assumes.

What this chapter adds to the plan.

1. The bottom-up sales forecast (§4.3). Year one: \$1,410,760 from the four physical variables — 68 seats, 1.4 dinner turns, a \$46 dinner check across five services, and 110 brunch covers at \$24 across two — plus a named, itemized bridge of **\$139,240 to the plan's headline of \$1,550,000**. The bridge is four claims: a seasonal patio (\$64,800), fourteen private events (\$42,000), a modest takeout channel (\$31,200), and \$1,240 of disclosed rounding.

2. The assumptions register (§4.4). Sixteen rows, five of them flagged Low or Lowest, each with a dollar exposure and an assigned chapter. It is the plan's most important page and the one the partners will revise most often.

3. The three-year pro forma (§4.5). Year one at \$1,550,000 revenue, 60.0% prime cost, and \$261,020** of operating profit before **\$69,500 of debt service. Years two and three at \$1,720,000 and \$1,850,000, with the improvement coming mostly from operating leverage rather than from cost heroics.

4. The sensitivity analysis (§4.6). Seven single-variable scenarios ranked in dollars, plus a combined downside case at \$1,359,260** of revenue and **\$154,854 of operating profit. The largest single exposure in the plan is the \$139,240 bridge, at **\$69,954** of operating profit.

5. The executive summary. Drafted last, as it should be:

BELLWETHER — Executive Summary (constructed teaching example)

The concept. Bellwether is a 68-seat chef-driven neighborhood American restaurant in the Rivermill District, built around a wood-fired hearth and a short seasonal menu, serving dinner Tuesday through Saturday and brunch on Saturday and Sunday.

The market. The Rivermill District is a former warehouse neighborhood roughly eight years into a residential conversion, in a metro of about 350,000. The competitive analysis in Section 2 surveys eight establishments, with menus and check averages collected on site. It identifies one direct competitor for the core occasion — a two-year-old new-American bistro at a \$44 check — and a wine bar competing for the bar occasion. Bellwether's 68 seats represent roughly a 40% increase in the direct-occasion seat base in this trade area. The plan is share-taking, not gap-filling, and is written on that basis.

The team. A chef with fourteen years of kitchen experience, including six as an executive chef, partnered with a front-of-house manager who has run dining rooms for eleven years. Neither partner has owned a restaurant, and neither has carried P&L responsibility. The plan addresses this directly: an outside accountant engaged before opening, a weekly prime-cost review from week one, and a bookkeeping and reporting calendar included in Section 10.

The model. Year one revenue of **\$1,550,000**: 68 seats at 1.4 dinner turns and a \$46 average check across five dinner services, 110 brunch covers at \$24 across two, a seasonal 16-seat patio, and modest private-event and takeout channels. Sales mix 72% food / 28% beverage.

The economics. Prime cost of 60.0% (COGS 27.8%, labor 32.3%), occupancy of 6.1% on a ten-year lease, and operating profit of \$261,020 — 16.8% of sales before debt service.

The ask. \$335,000** under the SBA 7(a) program, ten-year term, to complete a **\$620,000 project: \$310,000 of construction, \$185,000 of equipment including the hearth, \$45,000 of smallwares and FF&E, \$35,000 of pre-opening cost, and a \$45,000 working-capital reserve.

The stack. Partner injection of \$150,000 — 24.2% of total project cost, from savings and a retirement rollover — a \$75,000 landlord improvement allowance, and \$60,000 of equipment lease financing alongside the requested loan.

Repayment. Annual debt service of approximately \$69,500 (the SBA note plus the equipment lease) against projected year-one operating profit of \$261,020. In the combined downside case in Section 12 — turns at 1.25, a \$43 dinner check, and labor at 35.3% — operating profit is \$154,854 and debt service is still covered.

The three risks we cannot resolve before opening. (i) \$139,240 of the revenue forecast sits above the base case and depends on a seasonal patio, a private-events channel, and takeout, none of which has operating history. If none of it appears, revenue is \$1,410,760 and operating profit is \$191,066. (ii) The 32.3% labor line is a first-year target with a new team. At 35.3%, prime cost is 63.1% and operating profit falls to \$213,870. The staffing guide in Section 7 identifies the shift and the cover count at which hours come out. (iii) The \$620,000 project cost carries no contractor bid. No drawings exist and no lease is signed. The plan will be re-issued with bid-supported construction figures before funding.

What this checkpoint does not settle. Whether any of it is true. Every number above is a belief with a chapter assigned to it, and the plan's own register says so. It also does not settle the sources side of the capital — the stack is stated but not built, and no application has been made.

Open questions carried forward:

1. Does the \$139,240 bridge exist, and in what proportions? (Chapters 24, 28, 29)
2. Can a new team hold 32.3% labor in year one, and what is the trigger for cutting hours if not? (Chapters 17, 19, 21)
3. What does the build-out actually cost once someone bids it? (Chapters 6, 7)
4. What are the real lease terms, and what does the escalation clause do to a flat occupancy line? (Chapter 6)
5. Is a \$45,000 working-capital reserve enough to reach the month the business turns cash-positive? (Chapter 33)
6. Will the ask be granted, on what terms, and against what test? (Chapter 5 builds the package; the answer waits.)

Conclusion

A business plan is an argument, not a prediction. That is the whole chapter, and everything in it follows: you build revenue from physical constraints rather than market shares, you write down every belief in a register so it can be attacked, you push the beliefs through a statement that foots, you break the statement on purpose to find out which belief matters most, and only then do you write the page that tells a stranger what you are asking for and what you are risking.

Bellwether now has numbers. Year one at \$1,550,000, prime cost at 60.0%, operating profit at \$261,020, an ask of \$335,000 inside a \$620,000 project. Every one of those figures is an assertion, and the plan is honest about which are strong and which are guesses in a bold font.

The most useful thing in this chapter is the gap. The four physical variables produce \$1,410,760. The plan says \$1,550,000. The difference — \$139,240, or about \$2,678 a week — is real revenue from real sources, but it has no operating history behind it, and the sensitivity analysis says it carries more of the plan's profit than the food cost, the pour cost, and the general and administrative lines combined. A weaker plan would have nudged turns from 1.4 to 1.57 and made the arithmetic land on the round number. This one names the gap, itemizes it, and hands the reader a page on which to disagree.

That is what makes a plan credible, and credibility is the only collateral a restaurant startup actually has. A used hearth is not security. A forecast a reader can trace, argue with, and check is.

Chapter 5 takes the ask and turns it into a financing structure — where each layer of the money comes from, what it costs, what it demands, and what each source is really testing when it reads the document you just built. The plan says it needs \$335,000. Chapter 5 is about who might supply it, and on what terms.

Key Terms

Business plan — a written argument that a specific business, in a specific place, run by specific people, will generate enough revenue to cover its costs, repay its capital, and survive what goes wrong, supported by evidence a skeptical outsider can check. Not a prediction. (Ch. 4)

Executive summary — the one-page statement of the concept, model, economics, ask, repayment argument, and unresolved risks, written last, and specified so that a reader who reads nothing else can repeat the case accurately to someone who has not read it at all. (Ch. 4)

Pro forma — a projected financial statement built for periods that have not yet occurred; conventionally three years of P&L for a restaurant, with year one broken out monthly. The form of a financial statement filled with beliefs rather than history. (Ch. 4)

Bottom-up sales forecast — a revenue projection built from the physical constraints of the business — seats × turns × average check × services — rather than from a share of a market total. The opposite, and the standard error, is the top-down forecast. (Ch. 4)

Assumptions register — a single table listing every belief the plan's numbers rest on, with each one's value, confidence, basis, dollar exposure per unit of movement, and where it will be tested. The instrument that converts a forecast from a claim into an argument. (Ch. 4)

Use of funds — the schedule stating exactly what the requested capital will be spent on, line by line, totaling the amount asked for; the page a lender reads second, and one that must foot to the ask exactly. (Ch. 4)

Sensitivity analysis — the deliberate breaking of a forecast by moving one assumption at a time while holding the others constant, in order to rank assumptions by dollar consequence rather than by anxiety. (Ch. 4)

Spaced Review

1. Chapter 2 defined the competitive set. Name two specific pieces of evidence you could collect from a competitor in a single Saturday evening that would strengthen a line in the assumptions register, and say which row each one would support.
2. Chapter 3 argued that the room is part of the product. Which row of the Bellwether register is most directly a bet on the servicescape, and what would you have to observe after opening to know whether the bet was landing?
3. From Chapter 1: a restaurant's owner tells you their plan shows a 16.8% operating profit and they are pleased. Given the typical range Chapter 1 gave for full-service independents, what are the two most likely explanations, and which one would you check first?
4. Bellwether's food cost is 30% of \$1,116,000 of food sales and its pour cost is 22% of \$434,000 of beverage sales. Compute the dollar value of one point of each. Then explain why an operator who treats "a point is a point" will systematically misallocate their attention.
5. The recurring question: the partners are considering adding a Sunday dinner service to close the \$139,240 gap. Does this decision move prime cost, and in which direction? Name the two cost lines that would move first, and state when the decision hits the bank account as opposed to the P&L.