Appendix A — Key Formulas, Ratios, and Benchmarks

"Show me the number and I'll tell you what you believe. Show me how you got it and I'll tell you whether you're right." — constructed; the thing an accountant says to an operator exactly once

This is the appendix you print. Tape it inside the office door, next to the count sheets, where the person doing Monday's numbers can see it without opening a laptop. Everything in the book that has an equals sign in it is here, in one place, with the arithmetic done.

A.1 How to use this appendix

Every entry in this appendix has the same four parts, in the same order:

1. The formula, stated cleanly, in the form you would actually type into a cell.
2. The inputs, each one explained operationally — not what the textbook calls it, but what you physically have to count, pull, or ask for to get it.
3. A worked example in dollars, so you can check your own work against a number that resolves.
4. What it cannot tell you. This is the part that matters most and the part every other reference card leaves out. A ratio is a question, not an answer. An operator who knows what a number excludes is worth three who can compute it.

Worked examples are labeled either [the Bellwether plan] — the book's constructed 68-seat neighborhood restaurant, whose Year 1 plan figures are frozen and used throughout — or [constructed teaching example]. All of it is illustrative. None of it is a real business's records, and none of it is a target for your restaurant.

The standing rule: dollars are canonical, percentages are rounded displays

Every figure in this book is stored as a dollar amount. The percentages printed beside those dollars are rounded to one decimal place so a human can read a column quickly. They are for reading. Never reconstruct a line item by multiplying a percentage back against revenue.

Here is why, using the plan you will see fifty more times in this appendix.

Two things are happening in that table and both of them will cost you a Tuesday if you don't know about them.

First, the rebuilt figures are wrong. COGS at 27.8% of \$1,550,000 is \$430,900, which is \$620 more than the actual \$430,280. That \$620 is not a leak. It is rounding. Chase it and you will spend an afternoon recounting a walk-in that is fine.

Second, the displayed percentages don't add the way the dollars do. 27.8% + 32.3% = 60.1%, but the plan's prime cost is 60.0%. Both are correct. The true figures are 27.760% and 32.258%, which sum to 60.018%. Each rounds up on its own; the sum rounds down. There is no error here — there is only a rounding convention, applied consistently, which is exactly what a rounding convention is for.

The four conventions this book holds to

The denominator trap

Food cost percentage is computed against food sales. Pour cost against beverage sales. Labor, prime cost, occupancy, and everything else against total sales. These are four different denominators and they are not interchangeable.

THE SAME DOLLARS, THREE DIFFERENT PERCENTAGES              [the Bellwether plan]

  Food cost      $334,800  /  $1,116,000 food sales     =  30.0%   <- "food cost"
  Food cost      $334,800  /  $1,550,000 total sales    =  21.6%   <- a different fact
  Beverage cost  $ 95,480  /  $  434,000 bev sales      =  22.0%   <- "pour cost"
  Beverage cost  $ 95,480  /  $1,550,000 total sales    =   6.2%   <- a different fact
  ─────────────────────────────────────────────────────────────
  21.6% + 6.2% = 27.8%  =  blended COGS against total sales

Notice that 30% food cost and 22% pour cost do not average to 26%. They blend to 27.8%, because the blend is weighted by the sales mix — 72% food, 28% beverage. If your mix shifts toward food, your blended COGS rises even if neither the kitchen nor the bar did anything differently. Section A.2 covers this properly.

⚠️ Where the Money Leaks

The variance that isn't.

An operator compares last period's blended COGS of 27.8% to this period's 28.4% and calls a meeting. Six-tenths of a point on \$119,200 of period sales is \$715, which is worth a meeting.

Except that the sales mix moved. A cold snap killed patio drinking; beverage fell from 28% of sales to 24%. Run the same 30% food cost and 22% pour cost against the new mix and the blend is 28.1% before anybody did anything wrong. Half the "variance" was mix.

Always decompose a blended number before you act on it. Compute food cost against food sales and pour cost against beverage sales, separately, every single period. If both are on plan and the blend moved, you have a mix story, not a cost story — and mix is a marketing and menu problem, not a purchasing one.

A.2 Cost of goods sold

A.2.1 Usage — the only honest starting point

There is exactly one way to know what you used. You count.

USAGE  =  Beginning inventory  +  Purchases  -  Ending inventory

Worked [the Bellwether plan] — a four-week period, food only:

Check the arithmetic: \$8,400 + \$25,940 = \$34,340, less \$8,600 = \$25,740. And \$25,740 ÷ \$85,800 = 0.300 exactly.

The same period, beverage:

What it cannot tell you. Usage is a total. Theft, spoilage, over-portioning, uncosted specials, and a generous manager comping desserts all look identical inside it — a single number with a dozen possible causes. Usage tells you the wound exists. It never identifies the weapon. That is what the theoretical comparison in A.2.4 is for.

It also cannot tell you it is accurate if your count is sloppy. A \$400 miscount on an \$8,500 inventory moves your food cost by half a point, which on a year of food sales is \$5,580 of imaginary money moving in and out of your reports. The count is the number. Everything downstream inherits its errors.

A.2.2 Food cost, pour cost, and the blend

FOOD COST %      =  food usage      /  food sales
POUR COST %      =  beverage usage  /  beverage sales
BLENDED COGS %   =  (food usage + beverage usage)  /  total sales

Worked [the Bellwether plan] — the same four-week period:

\$25,740 + \$7,348 = \$33,088; \$33,088 ÷ \$119,200 = 27.76%, displayed as 27.8%.

On the annual plan the same relationship holds: food sales \$1,116,000 (72% of revenue) at 30.0% = \$334,800; beverage sales \$434,000 (28%) at 22.0% = \$95,480; total COGS \$430,280 = 27.8% of \$1,550,000.

The mix effect, stated once so you never forget it. Blended COGS moves for three reasons: food cost changed, pour cost changed, or the mix changed. Only two of those are cost problems.

(Each row: food share × 30% + beverage share × 22%. At 20% beverage: 0.80 × 30 + 0.20 × 22 = 28.4%.)

Three-tenths of a point of blended COGS per four points of mix. That is why a strong bar program quietly subsidizes a kitchen, and why a restaurant whose beverage attachment collapses gets a COGS problem it did not cause.

What it cannot tell you. A food cost percentage cannot tell you whether it is good. A 26% food cost bought with three extra prep hours a day is a worse business than a 32% food cost bought with portioned protein — the money moved to the other half of prime cost, where you weren't looking. And a pour cost of 18% may mean disciplined pouring or it may mean short-pouring, which you will read about in a review before you read about it on a statement.

A.2.3 Adjusted versus unadjusted COGS

Raw usage includes food that never appeared on a guest check. If you leave it there, your food cost percentage is punished for costs that belong in other categories — and, more importantly, your variance analysis in A.2.4 becomes meaningless, because you are comparing sold-item theory against total-consumption reality.

ADJUSTED FOOD USAGE  =  raw food usage
                        -  employee meals        (moves to labor / benefits)
                        -  comps and promotions  (moves to marketing, at cost)
                        -  transfers OUT to bar  (citrus, cream, garnish, syrup)
                        +  transfers IN from bar (wine for braising, beer for batter)
                        -  spoilage written off   (if you track it as its own line)

The bar's adjustment is the mirror image: transfers out of the kitchen are transfers in to the bar, and vice versa. Across the two departments, transfers net to zero.

Worked [the Bellwether plan] — same four-week period:

And the whole thing still reconciles: \$23,900 + \$7,508 + \$1,120 (to labor) + \$560 (to marketing) = \$33,088, the raw total. Nothing disappeared. It moved. That is the test of an honest adjustment: every dollar you take out of COGS must land somewhere else on the same statement.

What it cannot tell you. Adjusted food cost is the right number for judging the kitchen. It is the wrong number for judging the business, because the business still bought that food. An operator who reports 27.9% adjusted and forgets that 30.0% is what actually left the bank has optimized a report instead of a restaurant. Report both, every period, side by side.

It also cannot tell you your employee-meal figure is real. Most restaurants estimate it, which means most restaurants have a plug number sitting inside their food cost. Price the staff meal like a recipe or don't claim the adjustment.

A.2.4 Theoretical versus actual — the variance is the whole point

THEORETICAL FOOD COST  =  SUM over all items of ( plate cost  x  units sold )
THEORETICAL %          =  theoretical food cost  /  food sales
VARIANCE ($)           =  actual (adjusted) usage  -  theoretical food cost
VARIANCE (points)      =  actual %  -  theoretical %

Theoretical cost is what the food should have cost if every plate went out exactly as costed. It comes from two places and only two: your cost cards (A.3) and your point-of-sale item-count report. Both have to be current. A cost card priced against last spring's invoices produces a theoretical number that is confidently, precisely wrong.

Worked [the Bellwether plan] — same four-week period:

Annualize it: \$890 × 13 periods = **\$11,570 a year** walking out of the building without a check attached to it.

Reading the variance:

What it cannot tell you. The variance cannot tell you the cause, and it cannot tell you it is real. A one-point variance built on cost cards nobody has touched since opening is not a one-point variance; it is a one-point unknown. Before you accuse a station of over-portioning, re-cost the five items that make up the largest share of your food sales. In most restaurants that fixes half the "variance" on the spot.

It also cannot see anything that was never rung in. If a plate leaves the pass without a ticket, it appears in actual usage and never appears in theoretical — which is precisely why the variance is the tool it is, and precisely why it takes a human being to interpret.

A.3 Recipe and plate costing

Everything in Section A.4 depends on this section being right. A menu price built on a bad cost card is a guess wearing a decimal point.

A.3.1 Plate cost

PLATE COST  =  ( sum of costed components )  x  ( 1 + waste allowance )

Worked [the Bellwether plan] — the Hearth Chicken, the book's costing lab:

Foot it: \$5.60 + \$0.95 + \$1.05 + \$0.42 + \$0.18 + \$0.15 = \$8.35. And \$8.35 × 1.02 = \$8.517, which prints as \$8.52.

What it cannot tell you. A cost card prices ingredients, not the dish. It does not include the labor to butcher and brine, the gas to run the hearth, the four percent of birds that arrive over-weight and blow the portion, or the fact that both halves of the bird have to sell for the \$5.60 to be true. If you sell one half and scrap the other, your chicken component is \$11.20 and your plate cost is \$14.23, not \$8.52 — a 49.1% food cost on a dish you believe runs 29.4%.

A.3.2 Yield percentage, AP, and EP

As purchased (AP) is what you buy and pay for. Edible portion (EP) is what survives peeling, trimming, boning, and cooking loss to reach a plate. You pay AP prices. You serve EP quantities. Cost cards that ignore the gap understate every plate on the menu.

YIELD %        =  EP weight  /  AP weight
EP COST / LB   =  AP cost per lb  /  yield %
EP COST / OZ   =  EP cost per lb  /  16
AP NEEDED      =  EP needed  /  yield %

Worked [constructed teaching example] — a yield test on carrots:

YIELD TEST WORKSHEET
  Item ....................  Carrots, jumbo, loose
  AP weight ...............  10 lb 0 oz
  AP cost per lb ..........  $1.10
  AP total cost ...........  $11.00
  Peel and trim loss ......  2 lb 3.2 oz
  EP weight ...............  7 lb 12.8 oz   (= 7.80 lb)
  ─────────────────────────────────────────────────────
  YIELD %  = 7.80 / 10.00 ..............  78.0%
  EP COST / LB = $11.00 / 7.80 .........  $1.4103  ->  $1.41
  EP COST / OZ = $1.41 / 16 ............  $0.0882
  A 3 oz plated portion ................  $0.26
  AP you must buy for 3 oz EP = 3 / 0.78   3.85 oz

Both routes agree, which is the check: 3.85 oz AP × (\$1.10 ÷ 16 = \$0.06875/oz) = \$0.26.

Common yield ranges, for orientation only — run your own test, in your kitchen, with your cook:

What it cannot tell you. A yield percentage does not travel. It changes with the season, the supplier, the size grade, and who is holding the knife — a new prep cook can cost you six points of yield on the same case. Retest at least twice a year and always after a supplier change. And note that a published yield table is somebody else's kitchen; it is a place to start a test, never a substitute for one.

A.3.3 Waste allowance

WASTE ALLOWANCE  =  a flat % uplift applied ONCE, to the components subtotal

Bellwether uses 2%. A high-waste concept — whole fish, delicate garnishes, a hot line running four deep on a Saturday — may justify 3–5%. A tightly controlled, portion-packed operation may sit at 1%.

Two rules that prevent double-counting:

1. Apply it to the subtotal, once. Never line by line, and never again at the menu level.
2. If your yield percentages already capture trim loss, the waste allowance is for what happens after prep — spills, refires, drops. Do not use it to cover trim twice.

What it cannot tell you. A waste allowance is a plug. It does not reduce waste, it budgets for it. If your actual-to-theoretical variance (A.2.4) is running two points, raising the waste allowance to make the cost card match reality is not costing — it is bookkeeping the problem into permanence.

A.3.4 Q factor

The Q factor is the cost of everything a guest consumes that never appears on a check: bread and butter, the amuse, condiments, oil and vinegar, the ramekin of sauce nobody ordered, coffee refills, salt.

Q FACTOR ($ per cover)  =  total period cost of unpriced items  /  covers in the period
TRUE DELIVERED COST     =  plate cost  +  Q factor

Worked [the Bellwether plan]:

Applied to the Hearth Chicken: true delivered cost = \$8.52 + \$0.89 = \$9.41, which is 32.4% of the \$29.00 price, and the contribution margin falls from \$20.48 to \$19.59.

Applied to the year: 37,740 covers × \$0.89 = **\$33,589, or 2.2% of revenue** — nearly three-quarters of the plan's entire G&A line. Bread is not free.

⚠️ Where the Money Leaks

The Q factor charged twice.

Here is a mistake that produces menu prices nobody will pay. An operator adds the \$0.89 Q factor into every plate cost, prices from the loaded cost, and then subtracts the Q factor again at the menu-engineering stage when computing contribution margin.

On the Hearth Chicken that means costing the plate at \$9.41, pricing at \$9.41 ÷ 0.30 = \$31.37, and then reporting a contribution margin of \$31.37 − \$9.41 − \$0.89 = \$21.07 — which is a number describing no transaction that has ever occurred.

Decide where the Q factor lives and put it there once. The cleaner convention: keep plate costs pure (ingredients only), and carry the Q factor as a single line at the menu or P&L level, where it is visible and can be managed. Then the \$8.52 on the cost card means exactly one thing, and everybody in the building means the same thing by it.

A.3.5 Batch cost and portion cost

Almost nothing on a restaurant menu is made one plate at a time. You cost the batch, test the yield of the batch, and divide.

BATCH COST          =  sum of extended ingredient costs
PORTIONS PER BATCH  =  TESTED batch yield  /  portion size
PORTION COST        =  batch cost  /  portions per batch

Worked [the Bellwether plan] — salsa verde, the Hearth Chicken's \$1.05 component:

Note the word tested. The batch card must state the yield you actually measured after the sauce came out of the processor, not the sum of the input volumes. Scraping a food processor loses two to three percent, and two to three percent of every batch cost in the building is a real number.

What it cannot tell you. Portion cost assumes the portion. A 2 fl oz spoon used by a cook who "eyeballs it about right" is a 2.4 fl oz portion, which makes the component \$1.26 instead of \$1.05 and moves the Hearth Chicken's food cost from 29.4% to 30.1%. Batch costing is only as good as the portioning tool sitting in the container.

A.3.6 The cost card foot-check

Run this on every card before you price from it. It takes ninety seconds and it catches almost everything.

The Hearth Chicken passes all eight: components foot to \$8.35; 2% applied once gives \$8.52; \$8.52 ÷ \$29.00 = 29.4%; \$29.00 − \$8.52 = \$20.48. Three separate figures, one consistent card.

A.4 Menu pricing

Four methods. They are not alternatives to each other so much as four different questions, and a disciplined operator runs all four and then decides.

A.4.1 Food-cost-percentage pricing

MENU PRICE  =  plate cost  /  target food cost %

Worked [the Bellwether plan] — the Hearth Chicken:

\$8.52 ÷ 0.30 = **\$28.40, rounded up to a menu price of \$29.00**, at which the actual food cost is 29.4%.

That rounding is not sloppiness. Menu prices live on a price ladder the guest reads at a glance; \$28.40 reads as an awkward number and buys you nothing. The 60 cents of rounding, at 6,084 plates a year, is \$3,650 of contribution margin created by pressing a key differently.

A.4.2 The multiplier (factor) method

The same arithmetic, restated for people who would rather multiply than divide.

FACTOR      =  1  /  target food cost %
MENU PRICE  =  plate cost  x  factor

\$8.52 × 3.33 = \$28.37 — the same \$28.40, off by three cents because the factor was rounded. This is the "round last" convention from A.1 earning its keep.

What both methods cannot tell you. They cannot tell you what the market will pay. A price computed from a cost is an arithmetic floor wearing the costume of a decision. They also price cheap items too cheap and expensive items too expensive: a \$3.00 side dish at a 30% target prices at \$10.00, which no one will buy, and a \$22 plate of shellfish at the same target prices at \$73, which is a different restaurant. And they ignore labor entirely — the whole other half of prime cost.

A.4.3 Contribution-margin pricing

CONTRIBUTION MARGIN (CM)  =  menu price  -  plate cost
REQUIRED CM PER COVER     =  total fixed cost  /  covers   ( the break-even floor )
MENU PRICE                =  plate cost  +  target CM

This is the method that answers the question the restaurant actually has to answer: does this plate carry its share of the rent?

Worked [the Bellwether plan]:

Total fixed cost is \$437,635 (A.9.1). The plan's cover base is 37,740 — 36,140 base covers plus roughly 1,600 from the seasonal patio (A.8.3).

\$437,635 ÷ 37,740 = **\$11.60 of contribution margin required per cover, just to break even.**

Now look at two items side by side:

You can raise the numerator by adding a target profit — but only if the profit figure you add belongs to the same cost base as the fixed-cost figure. See the warning in A.9.1; mismatching those two is the most expensive arithmetic error available in this appendix.

The cavatelli has a food cost eight points better and contributes \$1.58 less per plate. Sell 396 of them in a period, as Bellwether's menu mix does, and that gap is \$626 a period — \$8,134 a year — invisible to anyone watching percentages.

You bank dollars. You cannot deposit a percentage. That sentence is the reason this method exists.

A.4.4 Prime-cost pricing

The only method that prices the dish rather than the ingredients, by loading direct labor onto the plate.

FULLY BURDENED LABOR RATE  =  hourly wage  x  ( 1 + burden rate )
DIRECT LABOR PER PLATE     =  burdened rate  x  ( minutes of direct labor / 60 )
PLATE PRIME COST           =  plate cost  +  direct labor per plate
MENU PRICE                 =  plate prime cost  /  target plate-prime %

Worked [the Bellwether plan] — the Hearth Chicken:

Run this on the two or three items you suspect are labor traps. Almost every menu has one: a dish with a beautiful food cost and eleven minutes of a cook's hands in it.

What it cannot tell you. Direct-labor minutes are the hardest number in this appendix to measure honestly. Nobody stands at a station with a stopwatch, and prep labor for a batch item spreads across portions in a way that resists clean allocation. Use prime-cost pricing as a comparison tool between dishes, not as a precision instrument.

A.4.5 Why contribution margin beats food cost percentage as a pricing basis

State it plainly, because this is the book's argument and it is worth one clean paragraph.

Food cost percentage is a ratio. Contribution margin is money. You pay rent, payroll, and debt service in money. Two items make the case:

Item A wins on every food-cost report ever printed. Item B contributes \$14.16 more per plate. An operator running a menu toward the best percentage will systematically drive out the items that pay the bills — a phenomenon common enough that it has killed restaurants whose food cost reports looked excellent right up to the end.

The correct use of food cost percentage is as a screen, not a basis: it tells you fast when an item's cost has drifted, and it lets you compare an item to itself over time. It is a terrible tool for comparing an item to a different item, and a worse one for deciding a price.

What contribution margin cannot tell you. Whether the item is worth making. A \$26 CM item that occupies the grill station for eleven minutes on a Friday at 8:15 may be costing you two other tickets and a table's second round. Contribution margin is blind to capacity, and capacity is what you actually run out of. See A.8.

A.5 Menu engineering

Menu engineering plots every item in a category on two axes — how often it sells and how much it contributes — and sorts the menu into four quadrants with four different actions. It is the single highest-return two hours a manager spends in a quarter.

A.5.1 The two thresholds

MENU MIX %            =  item units sold  /  total category units sold
POPULARITY THRESHOLD  =  ( 100%  /  number of items )  x  70%
                      =  70%  /  number of items
CM THRESHOLD          =  total category CM  /  total category units
                         ( the weighted average contribution margin )

The 70% in the popularity threshold is a convention, not a law. The logic: if every item sold equally, each would take 1/N of the mix; an item is "popular" if it achieves at least 70% of that equal share. On a ten-item menu, equal share is 10.0% and the threshold is 7.0%.

A.5.2 A worked menu

Worked [the Bellwether plan] — dinner entrées, one four-week period, 2,400 entrées sold:

A note on that last line, because it looks alarming next to a 30% plan: entrées carry a menu's highest food cost. Appetizers, sides, and desserts run well below it and pull the food-cost blend down to the plan figure. Never benchmark an entrée category against a whole-restaurant target.

A.5.3 The four quadrants

THE MENU ENGINEERING MATRIX                              [the Bellwether plan]
                      LOW popularity        HIGH popularity
                      (below 7.0% mix)      (7.0% mix or above)
                 ┌──────────────────────┬──────────────────────┐
  HIGH margin    │       PUZZLE         │        STAR          │
  (CM >= $19.14) │  Duck breast         │  Hearth Chicken      │
                 │                      │  Wood-grilled steak  │
                 │                      │  Pork chop           │
                 ├──────────────────────┼──────────────────────┤
  LOW margin     │        DOG           │      PLOWHORSE       │
  (CM <  $19.14) │  Winter squash plate │  House cavatelli     │
                 │  Mussels and toast   │  Hearth trout        │
                 │  Lamb ragù           │  Bar burger          │
                 └──────────────────────┴──────────────────────┘

The Hearth Chicken is the plan's clearest Star: 19.5% of the entrée mix at a \$20.48 contribution margin, generating \$9,584.64 of the period's \$45,936.84 of entrée contribution — 20.9% of the category's margin from one dish. Protect it accordingly.

A.5.4 Weighted average contribution margin

WEIGHTED AVERAGE CM  =  SUM ( item CM  x  item units )  /  total units
                     =  total category CM  /  total category units

The weighted average is not the simple average of the CM column. The simple average of the ten CM figures above is \$18.93; the weighted average is \$19.14. The difference is that the high-volume items pull the average toward themselves, which is correct — they are what the restaurant actually sells.

What the thresholds cannot tell you.

• The popularity threshold cannot make a category real. The cavatelli's \$18.90 sits 24 cents below a \$19.14 threshold. It is not a Plowhorse in any meaningful sense; it is a rounding difference from a Star. Treat items within about 5% of either threshold as unclassified and use judgment.
• The CM threshold moves when the menu moves. Cut the bar burger and the weighted average CM rises to \$20.15, which reclassifies the trout, the cavatelli, and the squash plate without any of them changing at all. Rerun the whole analysis after every menu change; never carry a threshold forward.
• Menu mix cannot tell you why. Position on the page, server scripting, the weather, the season, and what the next table ordered all move mix. A Puzzle may be a Puzzle because it is printed at the bottom of the third column.
• The matrix is blind to labor, station load, prep time, and inventory complexity. A Star that saturates one station is a capacity problem dressed as a success — see A.8.3, where Bellwether's hearth caps the kitchen at roughly 132 covers regardless of how well the chicken sells.

A.6 Labor

Food cost changes over a purchasing cycle. Labor changes this shift. It is the half of prime cost you can actually move this week, and the half that most reliably misleads people who read it as a single percentage.

A.6.1 Labor cost percentage

LABOR COST %  =  total all-in labor  /  total sales

The word that matters is all-in. A labor percentage built on gross wages alone understates your true cost by roughly a fifth.

Worked [the Bellwether plan] — the full decomposition:

Check: \$415,000 + \$38,388 + \$12,035 + \$34,577 = \$500,000, and \$500,000 ÷ \$1,550,000 = 32.26%, displayed as 32.3%.

What it cannot tell you. A labor percentage cannot separate discipline from volume. See A.6.7 — the same crew, working exactly as well, produces 36.7% in a soft February week and 30.1% in a strong October week purely because the fixed floor does not move. Judging a manager on the raw percentage punishes them for the calendar.

A.6.2 The burden rate

BURDEN RATE       =  ( all-in labor  -  wages )  /  wages
BURDENED RATE     =  hourly wage  x  ( 1 + burden rate )

Worked [the Bellwether plan]:

(\$500,000 − \$415,000) ÷ \$415,000 = \$85,000 ÷ \$415,000 = 20.5%

This is the number to keep in your head when someone asks for an extra hour, when you are deciding whether to add a shift, and when you are comparing an hourly hire against a service contract. The wage is never the cost.

What it cannot tell you. The burden rate is an average, and it is not stable. Workers' compensation is priced by class code and by your experience modifier — a single serious claim can move that 2.90% materially at renewal. Benefit costs reprice annually. Recompute the burden every year from your actual figures rather than carrying last year's forward.

A.6.3 Sales per labor hour and covers per labor hour

SPLH                   =  sales  /  labor hours worked
COVERS PER LABOR HOUR  =  covers  /  labor hours worked

Worked [the Bellwether plan] — using a working decomposition of the frozen \$415,000 wage line that reconciles to the fixed-labor figures in A.9.1: three salaried positions at \$140,195 of wages (which at the plan's 20.5% burden is the \$168,935 of fixed salaried labor) plus \$274,805 of hourly wages:

Note that \$22,960 of those hourly wages is burdened fixed labor — the open-and-close floor broken out in A.9.1 — and does not flex with volume. Everything above it does.

SPLH is the number a staffing guide is built from: forecast the sales, divide by your target SPLH, and you have the hours you can afford to schedule. At an \$83 target and a forecast Friday of \$5,810, you can schedule 70 hours. Not 74 because it feels busy.

What it cannot tell you. SPLH cannot tell you whether the guest had a good time. Push it high enough and you have built a room where nobody can find a server, which shows up as a revenue problem two quarters later, in a line item that will never be labeled "we cut too hard." It also does not compare across concepts or even across dayparts — brunch SPLH and Saturday-dinner SPLH are different businesses and should carry different targets.

Covers per labor hour has the same limitation plus one more: it is blind to check average. Two covers per labor hour at a \$24 brunch check and two at a \$46 dinner check are not the same productivity.

A.6.4 Full-time equivalents

FTE  =  total labor hours in the period  /  ( 40  x  weeks in the period )

Annual FTE divides by 2,080. Bellwether's hourly staff is 9.0 FTE, plus three salaried positions.

What it cannot tell you. FTE says nothing about scheduling quality. Ten FTE spread across twenty part-timers is a fundamentally different restaurant from ten FTE as ten full-time people — different training cost, different turnover, different consistency on a Friday, different benefits exposure. FTE is a headcount abstraction useful for planning and useless for managing.

A.6.5 Overtime

OVERTIME COST     =  hours over 40 in the workweek  x  regular rate  x  1.5
OVERTIME PREMIUM  =  the 0.5 x portion  -  the amount you paid for being late to schedule

Worked [the Bellwether plan] — a six-hour call-in on a Friday, at a \$16.10 base wage:

One six-hour call-in at overtime costs \$58.20 more than the same six hours scheduled. Twice a week for a year: 104 × \$58.20 = **\$6,053, or 0.4 points of labor cost**. That is a schedule-writing problem with a price tag on it.

⚖️ Code and Compliance

The wage arithmetic is federal, state, and sometimes municipal — all at once.

Under the federal Fair Labor Standards Act (FLSA), overtime is owed at one and one-half times the regular rate for hours over 40 in a workweek. Several points that routinely surprise operators:

• The regular rate is not always the base wage. Nondiscretionary bonuses, shift differentials, and certain service charges must be folded in before the 1.5 multiplier is applied.
• Some states require daily overtime (over 8 hours in a day) in addition to weekly. Some have double-time provisions. Some have mandatory rest and meal-period premiums.
• For a tipped employee, overtime is computed on the full applicable minimum wage before the tip credit is applied — not on the reduced cash wage. Getting this backwards is one of the most common wage-and-hour violations in the industry.
• Predictive-scheduling ordinances in a growing number of cities attach a penalty to schedule changes made inside a notice window.

All of this varies by state, county, and city, and all of it changes. Nothing in this appendix is legal advice. Verify locally, keep the records, and use a payroll provider and an employment attorney who know your jurisdiction.

A.6.6 The tip credit

The tip credit permits an employer, where the jurisdiction allows it, to count a portion of an employee's tips toward the applicable minimum wage. Several states do not allow it at all and require the full minimum in cash before tips.

CASH WAGE  =  applicable minimum wage  -  tip credit claimed
TEST:  cash wage  +  tips actually received  >=  applicable minimum wage
       ( applied per workweek under the federal framing; some jurisdictions test per shift )
MAKE-UP OWED  =  ( applicable minimum  -  ( cash wage + hourly tips ) )  x  hours

Worked [illustrative; every rate below varies by state and city]:

TIP CREDIT ARITHMETIC — a good week and a bad week

  Applicable minimum wage ................  $12.00 / hour
  Maximum tip credit permitted ...........   $5.00 / hour   (statutory; varies)
  Required cash wage from the employer ...   $7.00 / hour

  GOOD WEEK
    Hours worked .........................  24
    Tips received ........................  $312.00  =  $13.00 / hour
    Cash wage + tips .....................  $7.00 + $13.00 = $20.00 / hour
    Test .................................  $20.00 >= $12.00   PASSES
    Employer's wage cost .................  24 x $7.00 = $168.00 (plus burden)

  BAD WEEK  (a February Tuesday-heavy schedule)
    Hours worked .........................  24
    Tips received ........................   $84.00  =   $3.50 / hour
    Cash wage + tips .....................  $7.00 + $3.50  = $10.50 / hour
    Test .................................  $10.50 <  $12.00   FAILS
    MAKE-UP OWED .........................  ($12.00 - $10.50) x 24 = $36.00
    Employer's wage cost .................  $168.00 + $36.00 = $204.00 (plus burden)

What it cannot tell you. A labor percentage computed in a tip-credit jurisdiction is not comparable to one computed in a jurisdiction without a tip credit. Two identical restaurants with identical crews and identical hours can report labor six or seven points apart purely because of where they sit. When you benchmark your labor line against any published range, the first question is which wage regime produced the range.

The tip credit also cannot be claimed on autopilot. Tip-pool composition rules, the treatment of non-tipped side work, and the prohibition on managers and supervisors sharing in a tip pool are active, contested, and jurisdiction-specific areas of law. Verify, document, and re-verify.

A.6.7 Fixed versus variable labor

This is the decomposition that makes a weekly labor number readable.

LABOR FORECAST  =  fixed labor  +  ( variable labor rate  x  forecast sales )

Worked [the Bellwether plan] — at the plan's \$500,000 labor line (Base 1 in A.9.1):

The fixed half — \$168,935 of burdened salaried positions plus a \$22,960 open-and-close hourly floor — is identical under either labor base. Only the variable half moves: at the bottom-up \$570,461 labor line it is \$378,566, a variable rate of 24.42% rather than 19.88%. Break-even work in A.9 uses that higher figure.

Now run it across the year, with the same discipline in both weeks:

⚠️ Where the Money Leaks

Firing a good manager for the weather.

Look at that table again. 6.6 points separate the February week from the October week, and nothing changed except sales. The fixed floor is \$3,690 whether the room does \$22,000 or \$36,000; at the lower number it is 16.8% of sales instead of 10.3%.

An owner who reviews weekly labor percentage without decomposing it will conclude their manager lost control in February and got it back in October. Both conclusions are wrong, and acting on the first one is how restaurants lose the person who was actually running the place well.

Judge variable labor against the variable rate; judge fixed labor once a quarter, as a structural decision. The question in February is never "why is labor 36.7%." It is "is variable labor still 19.88%, and is the fixed floor still the right floor for a February?"

What it cannot tell you. The split is a management convention, not an accounting fact, and the line between fixed and variable is genuinely blurry — a salaried sous is fixed until you cut the position, at which point it was variable all along. Set the split deliberately, write it down, and hold it constant for at least a year so your period-over-period comparisons mean something.

A.7 Prime cost and the P&L

A.7.1 Prime cost

PRIME COST ($)  =  total COGS  +  total all-in labor
PRIME COST %    =  prime cost  /  total sales

Worked [the Bellwether plan]: \$430,280 + \$500,000 = \$930,280, which is 60.0% of \$1,550,000.

Say it in words, because the sentence is the point: out of every dollar through the register, sixty cents is gone before you have paid the rent.

Recall the rounding note from A.1 — the displayed 27.8% and 32.3% sum to 60.1% while the true prime cost is 60.018%. Report prime cost from the dollars.

WHERE THE DOLLAR GOES — Bellwether, Year 1 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¢
  = cash before tax      ████                       12.4¢

What it cannot tell you. Prime cost cannot tell you whether the business is viable, because it says nothing about what sits underneath it. A 58% prime cost under an 11% occupancy is a worse position than a 62% prime cost under a 5% one. Always read prime cost and occupancy as a pair — they are the two structural facts that decide how much room the rest of the statement has.

It also cannot tell you which half moved. A prime cost that holds at 60.0% while food rises two points and labor falls two points is not stability; it is two separate stories that happen to cancel. Decompose every period.

A.7.2 The standard order of a restaurant statement

Formats vary — the Uniform System of Accounts for Restaurants and most accounting packages differ on where occupancy, marketing, and depreciation sit. What does not vary, and what you should insist on, is that prime cost appears as a visible subtotal and that the same format is used every single period.

THE RESTAURANT P&L — the standard order        [the Bellwether plan, Year 1]

  SALES
    Food sales                                    $1,116,000    72.0%
    Beverage sales                                  $434,000    28.0%
  TOTAL SALES                                     $1,550,000   100.0%

  COST OF SALES
    Food cost                (30.0% of food sales)  $334,800
    Beverage cost            (22.0% of bev sales)    $95,480
  TOTAL COST OF SALES                               $430,280    27.8%

  GROSS PROFIT                                    $1,119,720    72.2%

  LABOR
    Wages and salaries                              $415,000
    Payroll taxes           (9.25% of wages)         $38,388
    Workers' compensation   (2.90% of wages)         $12,035
    Benefits, meals, uniforms, training              $34,577
  TOTAL LABOR                                       $500,000    32.3%

  ==============================================================
  PRIME COST                                        $930,280    60.0%
  ==============================================================

  CONTROLLABLE OPERATING EXPENSES
    Other operating (utilities, supplies, marketing,
      repairs, technology, card processing)         $217,000    14.0%
    General and administrative                       $46,500     3.0%
  TOTAL CONTROLLABLE OPERATING                      $263,500    17.0%

  CONTROLLABLE INCOME                               $356,220    23.0%

  NON-CONTROLLABLE
    Occupancy (rent, NNN, property insurance/tax)     $95,200     6.1%

  OPERATING PROFIT  (= EBITDA on this plan)         $261,020    16.8%
    Debt service (principal + interest)               $69,500
  CASH BEFORE TAXES, DISTRIBUTIONS, AND CAPEX       $191,520    12.4%

Every line foots: \$1,116,000 + \$434,000 = \$1,550,000. \$334,800 + \$95,480 = \$430,280. The four labor components sum to \$500,000. \$430,280 + \$500,000 = \$930,280. \$217,000 + \$46,500 = \$263,500. \$1,550,000 − \$930,280 − \$263,500 = \$356,220. − \$95,200 = \$261,020. − \$69,500 = \$191,520. Total costs above operating profit: \$1,288,980.

A.7.3 Controllable income

CONTROLLABLE INCOME  =  total sales
                        -  cost of sales
                        -  total labor
                        -  controllable operating expenses

Controllable income is the restaurant-level result before the costs a manager cannot change this week: rent, property tax, insurance carried in the lease, equipment leases, depreciation, and interest. It is the number to judge a general manager on, because it is bounded by the decisions a general manager actually makes.

Worked [the Bellwether plan]: \$356,220, or 23.0% of sales.

What it cannot tell you. "Controllable" is aspirational. A great deal of what sits above that line is contracted for the year — the linen agreement, the pest contract, the technology subscriptions, the insurance premium. Recheck what is genuinely controllable this period before you hold anyone accountable for it.

A.7.4 EBITDA, and why it is not cash

EBITDA is Earnings Before Interest, Taxes, Depreciation, and Amortization.

EBITDA  =  operating profit  +  depreciation  +  amortization

Bellwether's plan carries no depreciation line, so its \$261,020 of operating profit is EBITDA. On a statement that does carry depreciation and amortization below controllable income, add them back.

EBITDA is a useful number: it strips out financing structure and non-cash charges so two restaurants can be compared on operations alone. It is also, routinely, the most dangerously misread figure in a small business, because it is not cash and nobody's landlord accepts it.

The two adjustments that catch people:

1. Add back depreciation. It is a real cost of a real asset wearing out, but no money left the account this period. It is non-cash.
2. Subtract principal. Your loan payment has two halves. Interest appears on the P&L. Principal does not appear anywhere on the P&L and is 100% cash out the door. EBITDA is blind to it.

Then subtract the two things nobody budgets: cash taxes, and the capital replacement that a restaurant with fryers, a walk-in, and a dining room absolutely will need.

EBITDA IS NOT CASH — the full bridge                 [constructed teaching example]

  Operating profit (statement carries a D&A line) ......  $180,000
  + Depreciation and amortization ......................  $ 62,000   non-cash: ADD BACK
  = EBITDA .............................................  $242,000
  - Interest ...........................................  $ 31,000
  - Principal ..........................................  $ 44,000   cash out; NOT on the P&L
  - Cash income taxes ..................................  $ 18,000
  - Capital replacement (real and recurring) ...........  $ 25,000
  = CASH AVAILABLE TO THE OWNERS .......................  $124,000

  EBITDA said $242,000. The bank account said $124,000.
  The $118,000 difference is not an accounting subtlety. It is the year.

What it cannot tell you. EBITDA cannot tell you the account will clear payroll on the fifteenth. It excludes principal, capital replacement, taxes, working-capital swings, and — above all — timing, which is the thing that actually closes restaurants. A restaurant can post a healthy annual EBITDA and miss rent in February. See A.10.

A.8 Revenue and capacity

Every seat-hour is inventory you cannot store. The 7:00 seat at 6:40 on a Friday is a product with a twenty-minute shelf life; at 7:20 it is worthless and it will never be sold again. An airline sells a seat once per flight. You sell yours one and a half times a night, and three times if you are very good and the room is designed for it. That is the entire reason this section exists.

A.8.1 The four core figures

COVERS         =  guests served ( one guest = one cover; a party of four = four covers )
AVERAGE CHECK  =  sales  /  covers            ( also: per-person average, PPA )
SEAT TURNS     =  covers  /  seats            ( per service )
REVPASH        =  revenue  /  ( available seats  x  hours open )

Worked [the Bellwether plan] — a plan-average dinner:

Annual seat-hours [the Bellwether plan]: dinner, 68 × 5 hours × 260 services = 88,400 seat-hours; brunch, 68 × 4 hours × 104 services = 28,288; total 116,688 seat-hours. Annual RevPASH = \$1,550,000 ÷ 116,688 = **\$13.28.**

A.8.2 RevPASH by hour — where the empty inventory actually is

Annual RevPASH is a planning number. Hourly RevPASH is a management number, because it shows you exactly which seat-hours you are failing to sell.

REVPASH BY HOUR — a strong Saturday dinner              [the Bellwether plan]

  hour       covers   revenue   seat-hours   RevPASH
  5-6 pm         14   $  644         68      $ 9.47    early; bar and two-tops
  6-7 pm         30   $1,380         68      $20.29
  7-8 pm         38   $1,748         68      $25.71    the room is full
  8-9 pm         32   $1,472         68      $21.65
  9-10 pm        18   $  828         68      $12.18    the tail
  ────────────────────────────────────────────────────
  TOTAL         132   $6,072        340      $17.86
  Seat turns: 132 / 68 = 1.94

Read the shape, not the total. The 5:00 and 9:00 hours are running less than half the RevPASH of the 7:00 hour. Those are the hours a reservation policy, an early-seating price, a bar menu, or a different marketing message is aimed at — and they are worth attacking precisely because the fixed cost of being open is already paid.

What RevPASH cannot tell you. It cannot tell you why the empty hour is empty. Weak demand, a reservation system that refuses 5:15, understaffing that made you close a section, or the simple fact that nobody in your neighborhood eats at 5:30 all produce an identical low number. RevPASH finds the hole; it never explains it.

It is also easily gamed. Shorten your posted hours and RevPASH rises, because the denominator fell — which may be exactly the right decision, or may be you giving up on a daypart you could have fixed.

A.8.3 Capacity math

ANNUAL REVENUE     =  seats  x  turns per service  x  average check  x  services per year
THEORETICAL MAX
COVERS PER SERVICE =  seats  x  ( service hours  /  average table time )

Worked [the Bellwether plan] — the base arithmetic:

That base is 695 covers a week — five dinner services plus two brunch services at roughly 110 covers each — which annualizes to 36,140 covers.

The plan projects \$1,550,000**, which is \$139,240 above the base, about 9%. Two things close the gap. The 16-seat seasonal patio adds roughly 1,600 covers, taking the plan's cover base to 37,740. The rest is check: \$1,550,000 across 37,740 covers is a blended average check of \$41.07**, above the base illustration's \$39.03 — which the plan attributes to private events, beverage attachment, and the fact that a plan year is not fifty-two identical weeks.

Name the cover base every time you quote an average check. \$1,550,000 is one number, and it means something different over 36,140 covers than over 37,740. A revenue projection that cannot be reconciled to seats, turns, check, and operating days is a wish with a comma in it.

Now the ceilings:

The binding constraint on a peak Saturday at Bellwether is not the dining room. It is the hearth. At roughly 28 items an hour, and assuming close to one hearth item per cover, the kitchen tops out near 132 covers no matter how many people want a table. That leaves 37 covers of headroom above the 95-cover plan average — which is the difference between a great Saturday and a ticket time that turns into a review.

And in patio season the room is 84 seats, not 68. Run the capacity math twice: once for the season and once for the rest of the year. Most operators run it once, on the good number, and are surprised in January.

What capacity math cannot tell you. It cannot account for the shape of demand. You do not have 227 covers of capacity; you have 68 seats and a ninety-minute window in which nearly everyone wants to be in them. It also cannot see party-size friction — a four-top that cannot seat a five, a two-top nobody wants at 8:30 — which in a real room costs several points of theoretical capacity every night.

And a kitchen ceiling is a peak-night number, never an average. At 132 covers every dinner Bellwether's dinner revenue would be \$1,578,720 — more than the entire plan, before brunch. The plan's \$1,136,200 of dinner revenue is 72% of that ceiling, and the missing 28% is not slack. It is Tuesday.

A.9 Break-even and sensitivity

A.9.1 The variable/fixed split

Break-even arithmetic requires sorting every cost into one of two buckets. Do this once, carefully, write it down, and hold it — the comparison across periods is worth more than the precision of any single split.

The fixed side [the Bellwether plan] — this figure does not change no matter which labor assumption you use:

Two internal checks: the two fixed-labor lines sum to \$191,895, the plan's fixed labor figure; and the two fixed other-operating lines plus \$112,960 of variable other operating sum to \$217,000, the plan's other-operating line.

The variable side — and this is where operators go wrong. Bellwether carries two defensible labor figures, and they produce two different variable-cost bases:

| | Base 1 — the plan's \$500,000 labor line | **Base 2** — the bottom-up \$570,461 labor line | |---|---|---| | COGS | \$430,280 | \$430,280 | | Variable labor | \$308,105 | \$378,566 | | Other operating, variable | \$112,960 | \$112,960 | | Total variable cost | \$851,345** | **\$921,806 | | Variable cost ratio | 54.93% | 59.47% | | Contribution margin ratio | 45.07% | 40.53% | | + Total fixed cost | \$437,635 | \$437,635 | | Total cost | \$1,288,980** | **\$1,359,441 | | Operating profit at \$1,550,000** | **\$261,020 (16.8%) | \$190,559 (12.3%) |

Both columns foot. Base 1 reproduces the plan's P&L in A.7 exactly. Base 2 costs \$70,461 more, which is precisely the gap between the two labor lines.

⚠️ Where the Money Leaks

The two-base error, which understates break-even by about \$174,000.

This book does its break-even work on Base 2, because the bottom-up staffing model is the honest one — it is what it actually costs to staff this restaurant to this service standard. 40.53% is therefore the operative contribution margin ratio.

Here is the trap. Both of these figures are correct and both are published: the CM ratio is 40.53% and the plan's operating profit is \$261,020. They are also not from the same cost base, and pairing them back-solves a fixed-cost figure of about \$367,000 that corresponds to no version of this restaurant. Run break-even off that number and you get roughly \$906,000 instead of \$1,079,815 — a break-even understated by about **\$174,000 of annual sales**, which is around eleven covers a night of false comfort.

The rule: a contribution margin ratio and a profit figure must come from the same cost base. Before you divide anything, write down which labor line you are standing on. If you cannot say, you are not ready to compute a break-even.

A.9.2 Break-even

CM RATIO           =  ( sales - variable costs )  /  sales
BREAK-EVEN SALES   =  fixed costs  /  CM ratio
BREAK-EVEN COVERS  =  break-even sales  /  average check

Worked [the Bellwether plan] — on Base 2, at a 40.53% contribution margin ratio:

One note on precision, and it is the "round last" rule from A.1 doing real work. The CM ratio displays as 40.53%; its unrounded value is 40.528645% (\$628,194 of contribution margin ÷ \$1,550,000). Divide \$437,635 by the displayed 0.4053 and you get \$1,079,780; divide by the unrounded ratio and you get **\$1,079,815**, which is the figure the book carries. The \$35 between them is rounding, not money — but compute break-even off a two-decimal ratio on a larger business and that gap stops being trivial.

A third figure is worth carrying: at the lawfully reclassified labor line — the higher number that results once every hour is classified the way wage-and-hour rules require — break-even rises again, to \$1,317,918, or 81 covers a night. Compliance is not free, and the correct response to that sentence is to plan for it rather than to hope.

The middle column is the one an owner actually cares about. The bank does not accept an operating profit; it accepts a payment. Bellwether needs about 77 covers a night, every night, to stand still.

A useful way to hold the ratio in your head:

1 / 0.4053  =  2.47

  Every $1 of fixed cost you add requires $2.47 of new sales to pay for it.
  A $500-a-month software subscription is $6,000 a year of fixed cost
  and $14,804 a year of sales you did not previously have to make.
  ( Compute from the ratio, not the rounded 2.47 - see "round last" in A.1. )

⚠️ Where the Money Leaks

Break-even covers computed against a revenue base nobody named.

Converting a break-even in dollars into a break-even in covers means dividing by an average check, and an average check means nothing without the cover base it came from. This is not pedantry — it moves the answer.

Divide the plan's revenue by a cover base that excludes brunch, or by dinner covers only, or by a base you never wrote down, and you can manufacture an average check as high as \$51.64 — a figure that describes no guest who has ever sat down. Run break-even against that phantom check and it overstates the covers you need by about seven a night, which is enough to talk yourself out of a viable business.

Always state the base. Bellwether's is 36,140 covers, or 37,740 including the patio, at a blended check of \$41.07. Write it next to every cover figure you publish.

A.9.3 Margin of safety

MARGIN OF SAFETY ($)  =  planned sales  -  break-even sales
MARGIN OF SAFETY (%)  =  margin of safety  /  planned sales

Worked [the Bellwether plan]:

Read it as the sentence it is: sales can fall by 19.3% before the plan stops covering its obligations, and by 15.0% before it stops covering them under a fully compliant labor model. That is a real cushion — and it is on paper, on a plan, for a restaurant that has not opened.

What it cannot tell you. Margin of safety assumes your fixed costs are actually fixed. They are not. Rent escalates on a schedule written into your lease (Bellwether's occupancy steps to \$98,000 by Year 3, adding \$2,800 to the fixed base). Insurance reprices. The salaried sous you added last spring is now part of the floor. Recompute the split annually, and recompute it the day you add a salaried position.

It also cannot tell you when. Break-even is an annual identity; the bank account is a weekly one. A restaurant with a 30% margin of safety over a year can still be short in the third week of February. See A.10.

A.9.4 Operating leverage

DEGREE OF OPERATING LEVERAGE  =  contribution margin  /  operating profit

Worked [the Bellwether plan], on Base 2: \$628,194 ÷ \$190,559 = 3.30.

A 1% change in revenue moves operating profit by about 3.30%. A 10% revenue shortfall — a bad winter, a road closure, a competitor opening — costs \$62,819, or 33.0% of operating profit, without a single cost line going wrong.

Run the grid. Every row is on Base 2, which is why the plan row shows \$190,559 rather than the \$261,020 in A.7 — A.7 states the plan at its \$500,000 labor line, and this grid stands on the bottom-up \$570,461 one. Both are true; they answer different questions, and you must say which you are using.

(For Year 3, when occupancy steps to \$98,000, rerun with fixed costs of \$440,435: at \$1,850,000 of revenue, operating profit becomes \$309,345, or 16.7%.)

Notice how much steeper this is than the plan's own P&L suggests. Below about \$1,200,000 of revenue the business is barely clearing its fixed costs on Base 2, and it is nowhere near covering \$69,500 of debt service. Operating leverage is why a 12% revenue miss is not a 12% problem.

What it cannot tell you. Operating leverage does not have a direction. It amplifies both ways, and the downside arrives faster than the upside because you cannot cut fixed costs on the timescale that revenue falls. High leverage is a description of risk, not of quality.

A.9.5 The one-point move in prime cost

This is the most important sensitivity in the book, so it gets its own subsection.

ONE POINT OF PRIME COST  =  1%  x  total sales

Worked [the Bellwether plan] — one point is \$15,500, measured here against the plan's stated operating profit and post-debt-service cash from A.7 (Base 1):

Three points of drift — a protein price that moved, portions that crept, a schedule written by feel, comps nobody reviewed — takes a quarter of the cash. That is the pattern from Chapter 1, quantified.

🧮 Run the Numbers

What it costs to out-sell a point of prime cost instead of fixing it.

A point of prime cost is \$15,500 of pure margin, gone. Suppose you decide not to fix it and to grow your way out instead. How much additional revenue do you need?

Not \$15,500. Additional sales arrive with additional variable costs attached, so only 40.53 cents of each new dollar reaches the bottom line:

text REVENUE NEEDED = $15,500 / 0.4053 = $38,244

That is a 2.5% increase in annual sales — about 931 additional covers at the plan's \$41.07 blended check over a 37,740-cover base, or roughly 18 more guests every week, all year, to replace one point of cost you could have fixed on a Monday.

A point of cost is two and a half times harder to out-sell than to correct. Every marketing plan should be read next to that sentence. Sales fix revenue problems. They do not fix cost problems, and a restaurant that tries to grow out of a cost problem grows the problem too.

A.10 Cash and working capital

Profit is an accounting result over a period. Cash is what is in the account on the Thursday payroll clears. This section is the arithmetic of the second one.

A.10.1 The cash conversion cycle

DIO  ( days inventory on hand )   =  average inventory  /  ( annual COGS / 365 )
DSO  ( days sales outstanding )   =  average receivables  /  ( annual sales / 365 )
DPO  ( days payable outstanding ) =  average payables  /  ( annual COGS / 365 )

CASH CONVERSION CYCLE  =  DIO  +  DSO  -  DPO

Worked [the Bellwether plan]:

Restaurants have a structurally short — sometimes negative — cash conversion cycle, because guests pay at the table and vendors extend terms. That is a genuine advantage and it is also the trap: negative working capital means growth funds itself while a slowdown starves you immediately, with no receivables to collect on the way down.

What it cannot tell you. Look at that 13.9 combined inventory turn and notice that it describes nothing in the building. Food turns 39 times a year; beverage turns 4. Never manage food and beverage inventory off a single blended number — a 40-bottle wine list and a walk-in of herbs are different asset classes that happen to share a storeroom.

More importantly, the cash conversion cycle measures only the inventory-to-collection loop. It is blind to payroll, which lands every two weeks regardless; to rent, which lands on the first; to the sales tax you collected and have not yet remitted; and to the insurance renewal that always seems to arrive in the slow season. A restaurant with a seven-day cash conversion cycle can absolutely miss payroll in February.

A.10.2 The money that was never yours

SALES TAX HELD  =  taxable sales  x  sales tax rate    ( collected daily, remitted later )

Worked [the Bellwether plan] — at a 7% rate: \$1,550,000 × 0.07 = **\$108,500 a year, about \$9,042 a month**, sitting in your operating account between collection and remittance and making the balance look healthier than it is.

Card processing runs the other direction: 2.81% of \$1,550,000 = **\$43,555 a year** that arrives in the account already netted out, which makes daily deposits look smaller than the day's sales and is the single most common source of "the POS says one thing and the bank says another."

A.10.3 The thirteen-week cash forecast

The forecast is a structure, not a formula. Build it once and roll it forward every Monday.

THIRTEEN-WEEK CASH FORECAST — the structure           [constructed teaching example]

  Week beginning ..............  W1    W2    W3    W4   ...   W13
  OPENING CASH

  CASH IN
    Card settlements, net of processing
    Cash and check sales
    Gift card sales (cash in now, liability later)
    Private event deposits
    Draws on the line of credit
  TOTAL CASH IN

  CASH OUT
    Food vendors            (dated to terms, not to delivery)
    Beverage vendors        (many states: cash on delivery for alcohol)
    Payroll                 (gross wages, by pay DATE)
    Payroll taxes           (by deposit schedule)
    Rent and NNN            (the 1st)
    Sales tax remittance    (the filing date, not the month earned)
    Utilities
    Insurance               (by policy month, in lumps)
    Debt service
    Equipment lease
    Licenses and permits    (by renewal date)
    Marketing
    Repairs and maintenance
    Owner distributions
  TOTAL CASH OUT

  NET CHANGE
  CLOSING CASH
  ─────────────────────────────────────────────────────
  MINIMUM CASH FLOOR   (the line you do not cross)
  HEADROOM = CLOSING CASH - FLOOR

Four rules that make the difference between a forecast and a decoration:

1. Date every outflow to the day it clears the bank, not the day it was incurred.
2. Enter payroll gross, plus the employer taxes. Net payroll is not what leaves the account.
3. Never forecast a revenue week from an annual average. Use last year's same week, adjusted.
4. Roll it forward every Monday, adding a new week 13. A forecast built once in January is a document; a forecast rolled weekly is a control.

A.10.4 Debt service coverage ratio

DSCR  =  net operating income  /  total debt service ( principal + interest )

Worked [constructed teaching example — not Bellwether]: a restaurant with \$148,000 of EBITDA and \$96,000 of annual debt service has a DSCR of \$148,000 ÷ \$96,000 = 1.54. A DSCR of 1.0 means the business generates exactly enough to make its payments and nothing whatever besides.

What it cannot tell you. DSCR is a single annual ratio computed on a definition you did not choose. It cannot tell you whether the cash arrives in the weeks the payments are due, which is the question that actually determines whether you make them. And a DSCR computed from a projection is a projection — lenders discount forecasts heavily and will want trailing figures. Do not present a planned DSCR as though it were a result.

Loan documents specify a minimum coverage level and how and when it is tested. That level varies by lender, by program, by the strength of the rest of the file, and it is negotiated. Read the covenant section of any term sheet with an accountant before you sign it.

A.11 Inventory and purchasing

A.11.1 Par levels and order quantity

PAR LEVEL       =  ( average daily usage  x  days until next delivery )  x  ( 1 + safety factor )
ORDER QUANTITY  =  par level  -  quantity on hand  -  quantity already on order

Worked [the Bellwether plan] — carrots, with deliveries Monday and Thursday:

⚠️ Where the Money Leaks

The permanent overhang from a par that does not match the pack.

A par of 19 lb on an item that only comes in 25 lb sacks produces a standing 6 lb surplus, every order, forever. It sits in the walk-in, ages, and eventually gets thrown away — and because it was never a decision, nobody counts the cost.

You have three honest options and one dishonest one. Change the par to the pack. Find a second use for the overhang and cost it into a batch item. Accept the carrying cost knowingly and write it on the par sheet. The dishonest option is to keep ordering 25 and telling yourself the par is 19.

Run this check across your whole order guide once a year. In a typical independent, a dozen items are quietly carrying a pack-size overhang, and the aggregate is measured in thousands.

What par levels cannot tell you. A par is a snapshot of a demand pattern that has already changed. A par set in October is wrong in February, wrong again when the menu turns, and wrong the week a Star item gets written up. Rebuild pars with every menu change and at least seasonally.

A.11.2 Buying on EP cost, not AP price

The yield test from A.3.2 is a costing tool. It is also a purchasing tool, and this is where it earns the most money.

Worked [constructed teaching example] — two carrot suppliers, per 25 lb case:

Supplier B costs 24 cents more per pound at the door and 36 cents less per pound on the plate. That is exactly the prime-cost trade from Chapter 1: the "cheaper" product was bought with labor you paid for anyway, in the other half of prime cost, where the food-cost report cannot see it.

What this comparison cannot tell you. Whether the quality is the same. Pre-cleaned product has a shorter shelf life, sometimes a different texture, and always a supply-chain dependency you don't control. Run the arithmetic, then taste it, then decide — in that order.

A.11.3 Conversions

Weight:

Volume:

Beverage — the ones you will actually use:

⚠️ Where the Money Leaks

The one-ounce over-pour on a glass of wine.

A bottle costs \$14.80 and pours at 5 oz for \$14.00 a glass.

text AT A 5 OZ POUR: 5 glasses x $14.00 = $70.00 revenue. $14.80 / $70.00 = 21.1% pour cost. AT A 6 OZ POUR: 4 glasses x $14.00 = $56.00 revenue. $14.80 / $56.00 = 26.4% pour cost.

One ounce of generosity costs 5.3 points of pour cost on that bottle. No one steals anything. No invoice changes. A bartender is being nice, and the bar's entire margin target is gone.

The fix is not a lecture, it is a tool: lined glassware, a jigger, or a metered pour, plus a weekly test-pour check. Discipline that depends on someone remembering is not discipline.

The conversion that costs people money. Fluid ounces are not ounces. A cup of water weighs about 8.3 oz; a cup of flour weighs about 4.25 oz. Cost by weight whenever the item is purchased by weight, and never convert volume to weight for a dry ingredient without a density you tested in your own kitchen with your own scoop.

A.11.4 Invoice reconciliation

The three-way match: the purchase order (what you ordered) against the delivery slip (what physically arrived, verified at the door before you sign) against the invoice (what you were billed).

🧾 Read the Numbers

```text FIGURE A.1 — "Thursday produce" [constructed teaching example] THE ARTIFACT One produce invoice, Thursday delivery, checked against the order guide and the delivery slip the receiving cook signed at 7:15 a.m. THE CONTEXT A 68-seat full-service restaurant, mid-week, ordinary delivery. The receiving cook was on the line by 7:30 and signed without a full count.

Item                Ordered  Received  Invoiced  Unit price   Extended
Carrots, 25 lb          2        2         2       $27.50      $55.00
Spinach, 4 lb case      3        3         3       $18.40      $55.20
Lemons, 165 ct          1        1         1       $46.00      $46.00
Parsley, 12 ct          4        3         4       $ 9.80      $39.20   SHORT 1
Shallots, 10 lb         2        2         2       $32.60      $65.20   was $28.90
Fuel surcharge          -        -         -            -      $ 6.50   new line
────────────────────────────────────────────────────────────────────────
INVOICE TOTAL                                                  $267.10
CORRECT TOTAL (short case credited)                            $257.30
CREDIT DUE                                                     $  9.80
PRICE VARIANCE TO INVESTIGATE  ($3.70 x 2 shallots)            $  7.40

WHAT IT SHOWS Three separate issues on one ordinary invoice: a short case billed in full ($9.80), a 12.8% unannounced price increase on shallots ($7.40 this week), and a fuel surcharge line that was not on the last invoice ($6.50). Total exposure on one delivery: $23.70. WHAT IT DOESN'T It cannot tell you whether the parsley was short at the truck or short in the walk-in an hour later, because nobody counted at the door. It cannot tell you whether the shallot price is a market move or a quiet reprice of a contracted item. And it says nothing about quality — a case that arrives at the right count and the wrong condition costs more than a short case. THE DECISION Call for the credit today, in writing, before the statement closes. Ask the rep, by name, for the shallot price history and whether the surcharge is permanent. Retrain receiving: nobody signs anything they did not count. THE LESSON $23.70 is not worth a phone call. $23.70 a week for a year is $1,232, and it is only the portion you happened to catch. Invoice discipline is not about any single invoice; it is about being the customer whose invoices get checked. ```

A.11.5 Price variance tracking

PRICE VARIANCE  =  ( new unit price  -  old unit price )  x  quantity used in the period

Worked [the Bellwether plan] — a hypothetical 7.5% increase on chicken, from \$3.20/lb to \$3.44/lb. This is a worked illustration; **the frozen Hearth Chicken cost card remains \$8.35 of components and a \$8.52 plate cost at \$3.20/lb.**

At 468 Hearth Chickens a period, Bellwether uses 234 birds — 819 lb.

Now trace it to the plate, against the frozen baseline:

And the two routes reconcile, which is the check: \$0.43 of plate-cost increase × 6,084 plates a year = \$2,616, of which \$2,555 is the raw price variance computed above and \$61 is the 2% waste allowance riding on top of the increase.

The decision that follows is not arithmetic. You can hold \$29.00 and absorb \$2,616 of margin, raise to \$30 or \$31, re-spec to a different bird, or change the plate. The formula tells you the size of the problem. It cannot tell you what the guest will pay, what the item's role on the menu is, or whether a Star with pricing power should be the item you move. That is a judgment, informed by A.5 and made by a person.

A.12 Benchmark bands

Read this paragraph before you read the table. Everything below is a directional industry range, not an audited figure. These are the shapes that show up repeatedly in industry guidance, operator experience, and the published references in the Bibliography; they are not the output of a survey this book conducted, and no decimal in them should be quoted to a lender. Ranges are given deliberately — a single point would be a lie with a decimal place. A coffee shop and a steakhouse do not share a cost structure, and neither do two American full-service restaurants in different wage regimes.

A.12.1 Full-service independent — the main bands

A.12.2 Beverage sub-bands

A bar-driven concept blends to 20–28% not because anyone is buying better but because the mix sits where the margins are. Read every beverage number against the program that produced it.

A.12.3 Labor sub-bands

A.12.4 By service style

Reproduced for reference; the same ranges appear in Chapter 1.

A.12.5 The two caveats that matter most

Occupancy percentage is a function of revenue as much as of rent. This is the standing warning on this whole table, and it is the one operators most reliably get wrong.

THE SAME 6% OCCUPANCY, TWO DIFFERENT BUSINESSES       [constructed teaching example]

  RESTAURANT ONE                        RESTAURANT TWO
    Rent + NNN ....... $95,200            Rent + NNN ....... $54,000
    Sales ......... $1,550,000            Sales ............ $900,000
    Occupancy ........... 6.1%            Occupancy ........... 6.0%
    On 2,800 sq ft ... $34/sq ft          On 2,800 sq ft ... $19/sq ft

Same ratio. Completely different business. Restaurant One is carrying an expensive lease and outrunning it with volume. Restaurant Two has a cheap lease and modest volume. Only one of those two positions is contractual. The cheap rent survives a bad year; the high volume does not. A restaurant can hit 6% occupancy by having cheap rent or by being busy, and those are not the same business — they have different downside, different lender profiles, and different answers to the question "what happens if sales fall twenty percent."

And watch what happens to the ratio when nothing about the lease changes at all:

The lease did not move. The denominator did. Always read occupancy in dollars per square foot as well as in percent, and understand that the "rent should be 6–10% of sales" rule of thumb means the lease you can afford is determined by a revenue number you have not yet earned.

Never move a band to accommodate your example. Bellwether's 27.8% blended COGS sits just below the 28–33% band, and that is left standing on purpose. It is a true and useful fact — a 28% beverage mix at a 22% pour cost pulls the blend under the range a food-weighted restaurant would hit — and the moment you widen a benchmark by a point so your own number lands inside it, the benchmark has stopped being evidence and become decoration. Report the number. Report the band. Explain the gap. That is the whole discipline.

Bellwether's projected operating profit sits far above the band. The plan shows 16.8% against a directional range of 3–10%. That is worth saying out loud rather than burying: this plan projects a result well above what an average independent full-service restaurant achieves.

That is what a plan is — an argument that this restaurant will outperform, made before the restaurant exists. The number to defend is not the 16.8%. It is the four assumptions underneath it: 1.4 turns on 68 seats, a \$46 dinner check, a 60.0% prime cost held from the first month, and \$1,550,000 of revenue in a first year. Every one of those is testable. None of them is yet true. Any serious reader of the plan — a partner, an accountant, a landlord, a lender — will go straight there, and you should get there first.

What benchmark bands cannot tell you. They cannot tell you what your restaurant should run. A band is the shape of a population; your restaurant is one member of it, with a specific concept, a specific wage regime, a specific lease, and a specific menu. Use these ranges to notice when you are outside the normal envelope and to ask why. Never use them as targets. Your targets come from your own cost cards, your own staffing guide, and your own lease — which is the work of Chapters 11 through 31, and the reason this appendix is at the back of the book rather than the front.

A.13 The quick card

The ten formulas an operator actually uses every week. Print this page. Tape it inside the office door.

The weekly close, in order

THE MONDAY MORNING SEQUENCE

  1.  Count food inventory.        Same person, same sheet, same route through the building.
  2.  Count beverage inventory.    Bar, cellar, storeroom, walk-in.
  3.  Pull purchases.              Every invoice RECEIVED in the week. Include the cash runs.
  4.  Compute usage.               Beginning + purchases - ending, food and beverage separately.
  5.  Pull sales.                  Food sales and beverage sales, separately, from the POS.
  6.  Compute food cost % and pour cost %.       Each against ITS OWN denominator.
  7.  Pull payroll.                Gross wages for the week, plus the burden.
  8.  Compute prime cost %.        (COGS + labor) / total sales.
  9.  Compare to last week, the same week last year, and the plan.
  10. Write down ONE number you will move this week, and who is moving it.

  Under two hours for a trained manager. Do it on Monday, every Monday, forever.

The five numbers to know without looking

All Bellwether figures in this appendix are illustrative and constructed; benchmark ranges are directional industry guidance, not audited data. Nothing here is legal, tax, or accounting advice — wage law, tax rates, health codes, and licensing vary by state, county, and city, and change regularly. Verify locally, and for anything consequential use an attorney and an accountant.