Chapter 11: Food Costing: Recipe Costing, Plate Cost, Food Cost Percentage, and Why 28–32% Keeps You Alive

"The card says eight fifty-two. The walk-in says otherwise. Believe the walk-in." — constructed; a chef to a new sous, on the last day of the month

Overview

Here is the question that starts this chapter, and it is the most ordinary question in the business. A cook hands you a plate. Half a chicken off the hearth, roasted roots, a spoon of salsa verde. It is delicious. What do you charge for it?

Most operators answer by looking at what the restaurant down the street charges, adding a dollar, and moving on. That is not a stupid method — the market is real information, and a price nobody will pay is worth nothing regardless of what your spreadsheet says. But it is incomplete, because it tells you what you may charge and says nothing about whether you can afford to make the thing. Between those two facts is the entire margin of a restaurant.

In Chapter 1 we said that if you take one number away from this book, take prime cost. Half of prime cost is food and beverage. This chapter is where that half becomes a skill. By the end you will be able to build a cost card that foots to the cent, convert a supplier's price into what the ingredient actually costs on the plate, price from a target and know when the target is lying to you, and — the part almost nobody does — compute what your food cost really was.

That last point is the chapter's central corrective. Chapter 1 made an accusation: that most operators who say "my food cost is thirty percent" cannot produce the calculation, because they divided their invoices by their sales and never counted the walk-in. That is easy to say and it sounds like scolding, so this chapter does the harder thing. It builds a two-period example in which the invoice shortcut reports a falling food cost while the true food cost rises by three points — and then shows the arithmetic that would have caught it in eight days.

We do all of it on one plate: the Hearth Chicken, the signature dish of the restaurant whose business plan runs through this book. Its cost card is the chapter's spine. We build it line by line, take it apart, break it deliberately, and re-cost it when the poultry contract moves. By the end, an \$8.52 plate cost will not be a number you were told. It will be one you can reproduce, defend, and audit.

In this chapter, you will learn to:

• Compute food cost percentage the only way it is ever true — from a count — and state precisely how wrong the invoice shortcut is in any given period.
• Build a recipe cost card from batch recipes and portion specs, including a waste allowance, and check that every line resolves.
• Distinguish as-purchased from edible-portion cost, run a butcher or trim test, and compute a yield percentage you can defend.
• Price a plate from a target food cost — and name the five circumstances in which to ignore the answer.
• Specify portion controls for a dish and quantify what portion drift costs per year.
• Compute ideal food cost, compare it with actual, and decompose the variance into causes you can investigate in a defined order.
• Re-cost a plate when a price moves, and choose among the four honest responses.

Learning Paths

🏗️ Opening — this is your chapter. Build every card in §11.2 by hand; the costed menu is a required section of your plan and a lender will read it. §11.5 sets your prices; §11.1 tells you how you will know whether they worked. 📋 Managing — weight §11.1, §11.6, and §11.7. You may not set prices, but you own the variance, and the variance is what your prime cost is made of. If you read one section, read §11.7. 🍸 Beverage — the method transfers exactly: substitute a bottle for a bird, a jigger for a scale, pour cost for food cost. Chapter 15 does the beverage version, but learn the logic here. 🚚 Small Format — costing matters more in a small format. Fewer items, no room to hide a mistake, and a menu narrow enough that one mis-costed dish moves your whole number. §11.3 and §11.4 are your highest-value sections.

11.1 What food cost percentage is, how it's computed, and how it lies

Food cost percentage is food cost of goods sold divided by food sales, for a defined period. It is the most quoted number in the restaurant industry, it appears within ninety seconds of two operators meeting, and it is wrong more often than it is right.

$$\text{Food cost \%} = \frac{\text{food cost of goods sold}}{\text{food sales}}$$

Simple enough. The problem is the numerator. "Food cost of goods sold" does not mean what you bought. It means what you used — different numbers, because a restaurant carries inventory and inventory moves.

The formula that is actually true

Food cost of goods sold for a period is:

$$\text{Food COGS} = \text{beginning inventory} + \text{purchases} - \text{ending inventory}$$

And therefore:

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

Read it in words, because the words are what make it stick. What you started with, plus what you bought, minus what is still on the shelf, is what you used. Everything else is guessing.

You cannot compute that without counting — no software workaround, no clever proxy. Somebody walks into the walk-in on the last night of the period with a clipboard and a scale. Chapter 13 owns that procedure: the count sheet, the sequence, the discipline, the frequency, and how to organize a stockroom so the count takes forty minutes instead of three hours. This chapter owns the arithmetic and what it does to your judgment.

⚠️ Where the Money Leaks

"I know my food cost." — the payoff.

In §1.3 we said that almost nobody who says this can produce the calculation, and promised the proof later. Here it is.

What most operators mean by "my food cost is 30%" is "I divided my food invoices by my food sales." That is not food cost. It is a purchasing ratio, and it answers a fine cash-flow question and a useless cost-control one.

The difference between the two methods is exactly computable, and it is one line of algebra worth more than most of what gets written about restaurant finance:

$$\text{shortcut \%} - \text{true \%} = \frac{\text{ending inventory} - \text{beginning inventory}}{\text{food sales}}$$

The shortcut is wrong by the change in your inventory, divided by sales. Build inventory and the shortcut overstates your food cost. Draw inventory down and it flatters you. Neither has anything to do with what happened in the kitchen.

Now notice the sting. Over a long enough window, inventory change is small and the two methods converge — which is precisely why an operator can be wrong for eleven months and then have the annual figure "come out about right" when the accountant finally trues it up at year end. The shortcut is not reliably wrong. It is reliably wrong in the short run, which is the only run in which you can still do something.

The two-period example that ought to end the argument

Chapter 1's Exercise 1.27 set exactly this task. Here is the worked version.

The restaurant below is a constructed teaching example: a neighborhood American restaurant of roughly Bellwether's size, in its third year. It counts inventory once a quarter because the accountant asks for it, and computes "food cost" monthly by dividing invoices by sales. The owner is proud of the number. Two consecutive four-week periods:

🧾 Read the Numbers

```text FIGURE 11.1 — "The month food cost improved" [constructed teaching example] THE ARTIFACT Two consecutive four-week periods of food purchases, food sales, and counted inventory, for a third-year neighborhood American restaurant. THE CONTEXT Period 1 ends the week before a long holiday weekend; the chef takes a forward buy on a poultry and beef promotion and fills the walk-in. Period 2 runs that inventory back down. Nobody on staff considers either fact relevant to food cost, because nobody is computing food cost from a count.

                                               PERIOD 1        PERIOD 2
                 Beginning inventory            $19,500         $23,500
                 Purchases (food invoices)      $28,940         $25,660
                 Ending inventory               $23,500         $21,000
                 ────────────────────────────────────────────────────────
                 Food used  (BI + P − EI)       $24,940         $28,160
                 Food sales                     $86,000         $88,000
                 ────────────────────────────────────────────────────────
                 TRUE food cost                   29.0%           32.0%
                 Invoices ÷ sales ("food cost")   33.7%           29.2%

WHAT IT SHOWS Two stories, one restaurant. The shortcut says food cost fell 4.5 points and the operator is doing a great job. The count says food cost rose 3.0 points and something is badly wrong. Both arithmetic operations are correct. Only one of them is about food. The error is exactly the inventory swing: Period 1 built $4,000 of inventory ($4,000 ÷ $86,000 = 4.7 points of overstatement); Period 2 drew down $2,500 ($2,500 ÷ $88,000 = 2.8 points of understatement). WHAT IT DOESN'T It does not say what caused the true 3-point rise. Usage went up relative to sales; that could be portioning, waste, spoilage of the forward buy, theft, uncosted specials, comps, or a menu mix that shifted toward expensive items. A usage number identifies the wound. §11.7's variance identifies the weapon. It also does not tell you whether the forward buy was a good idea — that is a purchasing and cash question, and Chapters 13 and 33 own it. THE DECISION Count weekly, starting this week, and compute food cost from the count. Then compute ideal food cost from the cost cards and look at the gap. Do not change a single menu price until you know which of the six causes it is. THE LESSON A ratio computed from the wrong numerator is not a small error. It can point in the opposite direction from the truth, and it does so precisely when you most need it to be right — in a single period, close to the event. ```

Sit with that for a moment. The operator did not merely fail to see a problem. The operator received a congratulatory signal at the exact moment the business deteriorated. That is worse than no information: no information leaves you uneasy, wrong information leaves you confident.

Now the money. Three points on Period 2's \$88,000 of food sales is \$2,640 in four weeks. Run it for a year at Bellwether's planned food sales of \$1,116,000 and three points is **\$33,480** — the same order as the \$38,700 that four and a half points cost the restaurant in Figure 1.3, the one that closed in March. Same mechanism, same ending.

And here is what makes the shortcut dangerous rather than merely sloppy. Take both periods together:

Over eight weeks the two methods differ by 0.9 points instead of 4.5. The shortcut's error shrank, because the inventory change over the longer window was small (\$21,000 − \$19,500 = \$1,500, and \$1,500 ÷ \$174,000 = 0.9 points — exactly the identity from the callout above). The longer you wait, the more right the wrong method looks, and the less you can do about anything.

The four ways food cost percentage lies even when you compute it correctly

Counting fixes the numerator. It does not fix the ratio, and a costing chapter that stopped here would be selling a formula without its limits. Food cost percentage misleads in four structural ways.

One: it is a ratio, and the denominator moves. Raise a price and food cost percentage falls though nothing in your kitchen changed. The Hearth Chicken: \$8.52 against \$29.00 is 29.4%. Raise the price 5% to \$30.45 and it becomes 28.0% — a 1.4-point "improvement" produced entirely by charging more. Menu inflation flatters food cost across the whole industry every year, and operators quietly credit themselves with discipline they did not exercise.

Two: it is blind to mix. Sell more cheap items and the percentage improves while your dollars fall. This is important enough to work:

🧮 Run the Numbers

The week the food cost improved and the restaurant made less money.

Bellwether's five highest-volume dinner entrées, at planned weekly volumes:

Item	Units	Plate cost	Ideal cost	Menu price	Sales
Hearth Chicken	96	\$8.52 | \$817.92	\$29.00 | \$2,784.00
The pork chop	54	\$10.09 | \$544.86	\$31.00 | \$1,674.00
Squash and grains	62	\$3.95 | \$244.90	\$24.00 | \$1,488.00
Ember trout	45	\$10.40 | \$468.00	\$28.00 | \$1,260.00
The burger	73	\$5.10 | \$372.30	\$21.00 | \$1,533.00
Total	330		\$2,447.98** | | **\$8,739.00

Food cost on this extract: \$2,447.98 ÷ \$8,739.00 = 28.0%. Gross contribution: \$8,739.00 − \$2,447.98 = \$6,291.02.

Now move twenty covers from the Hearth Chicken to the burger. Nothing else changes — same guests, same night, same kitchen.

• Chicken: 76 units → cost \$647.52, sales \$2,204.00
• Burger: 93 units → cost \$474.30, sales \$1,953.00
• New totals: cost \$2,379.58**, sales **\$8,579.00

Food cost is now \$2,379.58 ÷ \$8,579.00 = 27.7% — a 0.3-point improvement. Gross contribution is \$8,579.00 − \$2,379.58 = \$6,199.42**, which is **\$91.60 less. Annualized, \$4,763.20.

The number got better and the restaurant got poorer. You do not deposit percentages. Chapter 12 builds an entire discipline on this observation; here it is simply a reason to distrust the ratio you are about to spend a chapter learning to compute.

Three: it is blind to labor. A cost card prices ingredients. It knows nothing about the four hours somebody spent butchering, brining, and turning birds. Buy pre-portioned chicken at a higher price per pound and food cost rises while labor falls — whether that was a good trade is a prime cost question, which is why Chapter 1 insisted on prime cost first. §11.2 returns to this with a specific dollar figure, and it is uncomfortable.

Four: it is usually mis-measured, which we have now spent half a section on.

None of this makes food cost percentage useless. It makes it a diagnostic ratio with known failure modes, which is what most useful numbers are. Use it to compare a period against the same period last year, to compare actual against ideal (§11.7), and to sanity-check a price. Do not use it to rank menu items, judge a chef, or tell you whether you had a good month.

The benchmark, honestly stated

Industry guidance generally puts full-service food cost in the high twenties to low thirties — the chapter title says 28–32% and that is a fair statement of the convention. Fine dining commonly runs higher, because expensive proteins carry more trim and more waste; concepts built on vegetables, grains, and pasta run lower; bar-driven rooms blend down hard because pour cost is so much better.

Treat those as orientation, not as a target for your restaurant. The number that matters is the one your own plan requires, and the plan running through this book requires 30% food cost against a blended cost of goods of 27.8% and a 60.0% prime cost. That 30% is not borrowed from a trade magazine. It is a line the plan has to earn, and this chapter is where it earns it.

11.2 The recipe cost card: building one, line by line

A recipe cost card lists every component of one plated portion, the quantity of each, the cost of that quantity, and the resulting total. Its output is the plate cost: what it costs you in product to put that dish in front of a guest, once. It is the only honest basis for a price and the input to everything else in this chapter.

A cost card requires two things first: a standardized recipe (Chapter 10) saying what is in the dish and in what quantity, and a portion spec saying how much goes on the plate. If either is missing you are not costing, you are estimating. The reason Chapter 10 comes before this one is that you cannot cost a dish that is not the same dish twice.

The frozen card

Here is Bellwether's signature dish, and the number this entire chapter orbits.

Menu price \$29.00**. Food cost **29.4%** (\$8.52 ÷ \$29.00 = 0.2938). Contribution margin \$20.48** (\$29.00 − \$8.52).

(All prices in this book are illustrative. Ingredient markets move constantly, and a cost card is a snapshot that begins decaying the moment you print it — §11.8 is about that decay. Do not treat \$3.20/lb as a current market fact; treat it as what this restaurant's invoice said on the day the card was built.)

Now build it. Every line below derives from a batch recipe or a portion spec, and every line resolves.

Line 1 — the protein: \$5.60

The spec is an air-chilled whole bird, 3.5 lb, at \$3.20 a pound: \$11.20 a bird, half a bird on the plate, so the card carries \$5.60. Two things about this line matter more than they look.

First, \$5.60 is an allocation of a purchase, not a measured cost of what lands on the plate. It assumes the bird weighs exactly 3.5 lb and that the halves are equal. Neither is quite true, and §11.3 and §11.6 are about how much that costs you.

Second, the brine. Bellwether brines its birds overnight in salt, sugar, bay, and pepper — roughly nine cents a bird, four and a half cents a plate. There is no brine line on this card. Either the brine is hiding inside "oil, salt, and misc," or the card is four cents light. On 4,992 plates a year that is \$225, and the point is not the \$225. The point is that a cost card is a claim about reality and you should be able to say where every input on the prep list lives. When you find a line that isn't there, you add it, fold it explicitly into another line and write that down, or accept the error knowingly. What you must not do is not notice.

Line 2 — the roasted roots: \$0.95

A batch item, and batch items are where cost cards go wrong. Cost the batch, weigh the yield, divide.

Peeled and trimmed, those twelve pounds yield 9.0 lb — a 75.0% yield. Roasted, the 9.0 lb of raw vegetables cook down to 7.6 lb, or 121.6 ounces, of finished product. The batch still cost \$23.10, so:

$$\text{cost per finished ounce} = \frac{\$23.10}{121.6 \text{ oz}} = \$0.19$$

The portion spec is five ounces. Five ounces at nineteen cents is \$0.95; one batch yields about twenty-four portions.

Notice what happened. The invoice said carrots cost \$1.35 a pound. By the time they reach the plate, the blended root mixture costs **\$3.04 a pound** — \$0.19 × 16 — because a quarter of the weight went in the compost and another sixth evaporated in the oven. That gap is §11.3's subject, and it is the single most common reason a cost card is wrong.

Line 3 — the salsa verde: \$1.05

Another batch item, worth seeing in full because it looks cheap and isn't.

Measured yield: 40 fluid ounces. Cost per fluid ounce: \$28.00 ÷ 40 = **\$0.70. Portion spec: 1.5 fl oz. Card line: 1.5 × \$0.70 = **\$1.05.

Two disciplines hide in that arithmetic.

Measure the yield; do not assume it. If the batch actually returns 36 fl oz — what stays in the blender, what evaporates — the cost per ounce is \$0.78 and the card line is \$1.17. Twelve cents high on every plate, \$599 a year, because nobody weighed the container.

Cost the recipe in the units the recipe is written in. This batch is written in bunches, so we cost bunches, and the discarded stems are already inside the \$1.95. Had the recipe said "two cups of picked parsley leaves," we would have needed a yield test to convert bunches into picked leaves, and the cost per plate would be higher, correctly. The rule: the yield conversion lives at whichever step the recipe crosses from purchased units into prepared units, and it lives there exactly once. Cost it twice and you inflate the card; skip it and you understate it.

Lines 4, 5, and 6 — the small ones

These three lines total seventy-five cents — nine percent of the components — and they are the lines operators leave off entirely. A card with no line for fat, salt, and garnish is not slightly imperfect. It is systematically nine percent light on every dish in the building.

The waste allowance

Components sum to \$8.35. The card then adds a waste allowance of 2%: a percentage added to the costed components to cover what the recipe cannot see — trim you didn't measure, spillage, a sauce that broke, a plate that went back, the last inch of a batch that dried out on the line. Two percent of \$8.35 is \$0.17, giving a plate cost of \$8.52.

Be honest about this line, because it is the softest number on the card. Two percent is at the low end of common practice, which runs roughly 2% to 5% depending on how much of the dish is made in house and how expensive the protein is. Consider what it has to cover here: if one Hearth Chicken in every sixty-seven is dropped, overcooked, or sent back, the remake alone costs 1.5% — three-quarters of the entire allowance — before a single carrot has been over-trimmed. The 2% is defensible only if the kitchen runs a genuinely clean waste log. It is a watch item, not a settled fact, and Chapter 13's waste log is what settles it.

🧾 Read the Numbers

(This figure is written forward. Right now, in the planning file, every number in it is a quoted price. What follows is the card as it will read in week six, when the quotes have been replaced by invoices and somebody has checked. That re-verification is not optional — see §11.8.)

text FIGURE 11.2 — "The chicken that pays the rent" [the Bellwether plan] THE ARTIFACT Recipe cost card, Hearth Chicken, priced at $29.00 on the dinner menu. THE CONTEXT Bellwether, week 6 of operation. Half a 3.5 lb air-chilled bird at $3.20/lb, roasted roots, salsa verde, butter and aromatics, oil and seasoning, garnish. WHAT IT SHOWS Components total $8.35; a 2% waste and spillage allowance brings the plate to $8.52. At $29.00 that is a 29.4% food cost and a contribution margin of $20.48 — the single most profitable dollar-contributor on the menu. WHAT IT DOESN'T It does not include labor to butcher and brine, the gas and hardwood to run the hearth, or the birds that come in over-weight and blow the portion. A cost card prices ingredients, not the dish. Butchery and brining run about $0.79 a portion in fully loaded prep labor; hardwood allocates at roughly $0.63 a hearth plate. Add those two and the plate consumes $9.94 — 34.3% of its price, not 29.4%. The card is not wrong. It is answering a narrower question than the one an owner actually has. THE DECISION Hold the price at $29. Move the item to the top-right of the menu panel and train the servers to name it. Weigh the first three birds of every delivery. Re-cost when the poultry contract renews. THE LESSON You bank contribution margin, not food cost percentage. A 29.4% item that sells 200 times a week beats a 22% item that sells nine — and the card that produced both numbers is a purchasing document, not a profitability document.

That WHAT IT DOESN'T field is the most important paragraph in this chapter. Here is its arithmetic.

🧮 Run the Numbers

What the card leaves out, in dollars.

Butchery and brining. A prep cook breaks down, halves, trims, and brines twelve birds in an hour — twenty-four portions. At \$19.00 an hour fully loaded (wage plus payroll taxes plus benefits): \$19.00 ÷ 24 = **\$0.79 per portion**.

Hearth fuel. The hearth burns roughly \$180 a week of hardwood across about 285 hearth-cooked plates — sixty percent of dinner entrées. \$180 ÷ 285 = **\$0.63 per hearth plate**.

The shadow card, which is not a cost card and must never be filed as one:

Plate cost (the card)	\$8.52
+ butchery and brining labor	\$0.79
+ hearth fuel	\$0.63
What the plate actually consumes	\$9.94
As a percentage of \$29.00	34.3%

Thirty-four point three, on the dish the plan calls a 29.4% item.

Now: do not "fix" the cost card. Food cost percentage is defined to exclude labor and utilities, because those live in the other half of prime cost and in other operating expense. A card that absorbed them would be uncomparable to every benchmark in the industry and to every other card in your own binder.

Instead do what this book has argued since Chapter 1: stop managing to food cost and manage to prime cost. The \$0.79 of butchery is a real choice with a real alternative — pre-portioned half birds cost more per pound and less per hour, and the only way to know which is better is to watch the total. That is §1.3 with the decimal points filled in.

👨‍🍳 On the Line

How a cost card actually gets built, which is not how it gets taught.

The textbook version — open the recipe, look up prices, multiply, add — produces a card wrong by ten to twenty percent, because the standardized recipe is aspirational and the prep list is what actually happens. The version that works takes one prep shift and a scale.

1. Stand at the station while somebody makes the batch. Not the chef demonstrating — the prep cook who makes it every Tuesday, at normal speed. You will find one ingredient that is not on the recipe and one that is on it and never used.
2. Weigh the finished batch in a tared container. Everybody skips this step and it is worth more than all the others combined.
3. Weigh a plated portion. Have the line cook plate three the way they always do, before you say why. Do not correct anybody yet — correcting it destroys your measurement.
4. Price it from invoices, not a catalog. Use the delivered price, including any fuel or small-order fee attached to it.
5. Foot the card. If the components do not add to the total you have a typo, not a rounding issue. Fix it before it propagates into a price.

The failure mode: cards built in an office in January and never opened again. A cost card is perishable inventory. Treat it like the product it describes.

The plate-cost stack

FIGURE 11.3 — The Hearth Chicken plate, by the cent           [the Bellwether plan]

  chicken (½ bird)     █████████████████████████████████████████████   $5.60
  salsa verde          ████████                                        $1.05
  roasted roots        ████████                                        $0.95
  butter & aromatics   ███                                             $0.42
  oil, salt, misc      █                                               $0.18
  garnish              █                                               $0.15
  ────────────────────────────────────────────────────────────────────────────
  components                                                           $8.35
  + 2% waste allowance                                                 $0.17
  ============================================================================
  PLATE COST                                                           $8.52

  Menu price $29.00  →  food cost 29.4%  →  contribution margin $20.48
  Bars to scale, 8 characters per dollar.

Two thirds of this plate is the bird, and that generalizes: on almost every plate in a full-service restaurant one component is most of the cost, and it is almost always the protein. So if you have one afternoon to improve your food cost, spend it on the protein line of your five highest-volume dishes. A nickel saved on garnish across the whole menu is worth less than a dime saved on the chicken. Operators routinely get this backwards — a three-week campaign about lemon wedges and never once a weighed bird — because small economies feel virtuous and large ones feel like they need somebody's permission.

⚠️ Where the Money Leaks

The bread basket, and the point and a half nobody costs.

Look again at the card. It prices everything the guest ordered. It prices nothing the guest received for free.

The Q factor is the per-cover cost of everything served with a meal but not separately priced: bread and butter, condiments, the oil and vinegar on the table, the cream with the coffee, the amuse, the lemon in the water. Add it to the plate cost of every entrée, or carry it as a separate line in your food-cost build. Do not do neither, which is what most restaurants do.

Item	Build	Cost per cover
Bread and cultured butter	1.4 oz bread @ \$0.19/oz = \$0.266; 0.4 oz butter @ \$0.29/oz = \$0.116	\$0.38
Condiments, table oil, lemon, salt		\$0.06
Coffee and tea accompaniments		\$0.04
Q factor		\$0.48

Forty-eight cents. On the Hearth Chicken that turns an \$8.52 plate into \$9.00 and a 29.4% item into a 31.0% item.

Across the restaurant it is worse, because it applies to every cover whether or not they bought an entrée. At Bellwether's projected 36,140 annual covers, \$0.48 is **\$17,347 a year — 1.55 points of food cost** against \$1,116,000 of food sales. A point and a half, made entirely of bread nobody thought of as food cost.

What the disciplined operator does: compute the Q factor once a season, carry it as an explicit line, and — this is the part that pays — decide about it. A smaller, better piece of bread served on request rather than automatically is a legitimate, non-cheap answer. Pretending the bread is free is not.

11.3 AP vs. EP: yield, trim, and the cost of the parts you throw away

Every ingredient has two prices and operators routinely use the wrong one.

As-purchased (AP) cost is what the invoice says: the price of the product as it arrived, including peel, bone, stem, fat, water, and outer leaves. It is what your distributor quotes and what your accountant sees. Edible-portion (EP) cost is what the usable product costs after everything you cannot serve has been removed. It is what belongs on a cost card.

The bridge between them is the yield percentage: the proportion of as-purchased weight that survives into usable product.

$$\text{Yield \%} = \frac{\text{edible-portion weight}}{\text{as-purchased weight}} \qquad\qquad \text{EP cost} = \frac{\text{AP cost}}{\text{Yield \%}}$$

That second formula goes the direction people find counterintuitive. Yield is less than one, so dividing by it makes cost go up. A 75% yield does not mean your ingredient costs 75% of what you paid; it means it costs 1 ÷ 0.75 = 133% of what you paid.

🧮 Run the Numbers

A trim test on a case of carrots.

You buy a 25 lb case of carrots for \$33.75 — \$1.35 a pound as purchased. The prep cook peels and tops the case in the normal way. Weighed on the way out: 19.25 lb of usable carrot. The peel, tops, and tips — 5.75 lb — go in the compost.

$$\text{Yield \%} = \frac{19.25}{25.00} = 77.0\%$$ $$\text{EP cost} = \frac{\$1.35}{0.77} = \$1.75 \text{ per pound}$$

Check it the other way: \$33.75 ÷ 19.25 lb = \$1.75. The two routes agree, which is how you know you did it right — always compute a yield test both ways.

Forty cents a pound, which the invoice never mentions. On a four-ounce portion that is the difference between \$0.34 and \$0.44 — a dime a plate, on one component of one dish.

Scale it. A restaurant costing its entire vegetable program at AP, averaging a 78% yield, carries every vegetable line about 28% below its true cost. Where vegetables are roughly 15% of food cost, that is about a full point of food cost sitting invisibly inside a card binder that looks meticulous.

Where the AP–EP gap actually is

Yields vary enormously, and the pattern tells you which lines on your cards deserve a test.

Two things fall out of that table.

The test list is short. You do not need a yield test on ketchup. On a typical independent menu, ten to fifteen ingredients account for nearly all of the AP–EP gap, and they are the proteins and the produce you break down yourself. Test those; revisit the rest annually.

Portion-cut products have a yield of 100%, and that is precisely why they cost more. Buying a portioned breast or a cut steak buys somebody else's yield problem solved, their labor to solve it, and their margin on doing so. That is a legitimate trade — the same trade from §1.3 — but you cannot evaluate it until you know your own yield. An operator who does not know their yield cannot tell whether a convenience product is expensive.

Cooking loss is not yield loss, and you must not double-count it

Trim yield is what survives butchery and prep. Cooking yield is what survives the fire. They are different losses at different steps, and which belongs on your card depends entirely on the unit your portion spec is written in. A spec reading "5 oz of roasted roots" carries both losses, as §11.2 did: 12 lb AP → 9.0 lb trimmed → 7.6 lb roasted. A spec reading "6 oz raw beef, cooked to order" carries trim yield only, because the cooking loss happens after the portion is weighed.

Get this wrong one way and you overstate costs and price yourself out of the market; the other way and you understate them and wonder why the walk-in never balances. The discipline: write on every card, next to every quantity, whether the weight is as-purchased, trimmed-raw, or as-served.

11.4 Running a yield test (and why your supplier's number isn't yours)

A yield test is a controlled measurement in which you weigh a product as purchased, break it down the way your kitchen actually breaks it down, weigh every output stream separately, and compute the resulting yield percentage and edible-portion cost. Applied to meat and fish it is traditionally called a butcher test, and it has one extra move: assigning value to the by-products.

A butcher test, worked

Bellwether's pork chop comes off a whole bone-in loin. The spec is a 14 oz bone-in chop.

FIGURE 11.4 — From the invoice to the plate: 18.4 lb of pork loin  [constructed teaching example]

  AS PURCHASED       18.4 lb @ $6.40/lb ............................. $117.76
    │
    ├─ fat and sinew, discarded ........  1.9 lb  (10.3%)   value  $0.00
    ├─ bone, to stock ..................  2.3 lb  (12.5%)   credit  $2.07
    ├─ trim, to sausage ................  1.6 lb  ( 8.7%)   credit  $7.60
    │
    └─ SALEABLE CHOPS ..................  12.6 lb (68.5%)
                                          ─────────────────────────────────
       cost carried by the chops:   $117.76 − $9.67 = $108.09
       EP cost per pound:           $108.09 ÷ 12.6 lb = $8.58
       14 oz portion (0.875 lb):    0.875 × $8.58 = $7.51

  Streams reconcile: 1.9 + 2.3 + 1.6 + 12.6 = 18.4 lb.
  Percentages reconcile: 10.3 + 12.5 + 8.7 + 68.5 = 100.0%.

  The invoice said $6.40 a pound. The chop costs $8.58 a pound. The difference is
  not waste — it is the arithmetic of buying an animal instead of a portion.

Walk the figure. The loin arrives at 18.4 lb and \$117.76, and four streams come out of it that must add back to 18.4 lb — if they don't, you missed a stream, usually the one that went straight into the stockpot without being weighed.

The two credits are the move most operators skip. The 1.6 lb of trim becomes sausage, which means you are not buying 1.6 lb of ground pork; credit it at what that ground pork costs, \$4.75/lb, for \$7.60. The 2.3 lb of bone becomes stock; credit it at what you would pay for pork bones, \$0.90/lb, for **\$2.07**. Total credits \$9.67, off the cost carried by the chops.

The difference is not trivial:

The AP figure understates the true protein cost by \$1.91 a plate — a quarter of it. Costing without the credits overstates it by 67 cents, which sounds like the safe direction to be wrong until it makes you price the item out of the market or delete it from the menu.

The credit rule: credit a by-product only at the price of the thing it genuinely displaces, and only if you genuinely use it. Bones in the trash are worth zero. Trim that becomes a staff meal is worth zero on the card — a real benefit and a real cost, but not a purchase you avoided on this menu. Crediting a by-product you don't use is not conservatism; it is inventing revenue, and it reappears in §11.7 as an unexplained variance you will spend a month chasing.

The pork chop card

Now build the plate:

At a \$31.00 menu price: food cost **32.5%**, contribution margin **\$20.91. This item has a worse food cost percentage than the Hearth Chicken and contributes more dollars per plate. Chapter 12 turns that into a system; here it is one more reason the percentage should not be trusted alone.

Note also the roasted roots appearing on a second card — cross-utilization (Chapter 10) showing up in the arithmetic: one batch, one prep task, one line on the order guide, two dishes. Cross-utilization is a food-cost technique, not only a menu-design one, because it reduces the number of items you can be caught holding when a dish doesn't sell.

👨‍🍳 On the Line

How to actually run a yield test on a Tuesday.

Nobody has a spare afternoon. Here is the version that fits inside a real prep shift.

Do it on a normal delivery, in a normal week, with the person who normally does the job. The commonest way to ruin a yield test is to have the chef run it carefully on a quiet Sunday. That measures the best case, which is not the case you are buying. Your yield is a property of your kitchen, not of the product.

Three tared containers and a scale: primary cut, usable secondary, unusable. Everything goes in one of the three, including the bits that fall on the board.

Reconcile before you leave the station. Outputs must equal input within about 2%; more than that and something left unweighed.

Run it three times, on three deliveries. One test is an anecdote:

Test	AP weight	Saleable chops	Yield
1 (Tuesday)	18.4 lb	12.6 lb	68.5%
2 (Friday)	19.1 lb	12.9 lb	67.5%
3 (Tuesday)	17.8 lb	12.4 lb	69.7%
Combined	55.3 lb	37.9 lb	68.5%

Use total-out ÷ total-in, not the average of the three percentages. They differ — the simple mean here is 68.6% — and the weighted one is correct.

The spread is the finding, not the average. A 67.5%-to-69.7% range is 2.2 points of yield, which at \$6.40/lb AP is \$9.48 versus \$9.18 a pound of chop — 26 cents on a 14 oz portion. That is your real risk, and it is why the answer to "what should we cost it at?" is the conservative end.

Re-run when anything changes: new supplier, new spec, new prep cook, new season. A yield test has a shelf life measured in months.

Why your supplier's yield number isn't yours

Distributors publish yield figures and sales representatives quote them from memory. They are not lies, and they are not yours, for four reasons.

Their conditions. Their number came from a processing facility, from someone who breaks down four hundred loins a day with a knife they sharpen hourly, on a lot selected for the demonstration. Yours comes from a prep cook who does eight loins a week between other tasks. Their spec. They trimmed to a commodity standard; you trim for presentation, which means you cut away more — every kitchen with a plating aesthetic buys that aesthetic in yield points. Their by-products. The identical loin has a different effective cost in a restaurant that grinds its trim into a burger and one that bins it, and the supplier cannot know which you are. And the product itself varies by season, lot, grower, animal, and how long it sat. A yield is a distribution, not a constant.

Use a supplier's figure as a starting hypothesis. If your test comes back ten points off, that is worth a conversation — your spec may be wrong, your knife skills may need work, or the product you received may not be the product on the spec sheet. But cost from your own number. The yield you have is the yield you pay for.

🔍 Check Your Understanding

1. A case of romaine costs \$26.40 for 24 lb. After removing outer leaves and cores you have 16.8 lb. What is the yield percentage and the EP cost per pound?
2. Why does dividing by a yield percentage make a cost go up rather than down?
3. A supplier's spec sheet claims a 74% yield on a whole fish. Your three tests average 66%. Name three explanations that do not involve anyone being dishonest.

(1: 16.8 ÷ 24 = 70.0% yield; \$26.40 ÷ 24 = \$1.10 AP; \$1.10 ÷ 0.70 = **\$1.57/lb EP — confirm with \$26.40 ÷ 16.8 = \$1.57. 2: Because you paid for the whole thing but can only sell part of it, so the cost of the whole must be recovered across a smaller usable weight. 3: Their test was run by a specialist on selected product; your spec trims harder for presentation; you are not crediting frames and collars that they counted as yield; and the fish you received may simply be a different size grade — larger fish generally yield better.)

11.5 Pricing from a target: plate cost ÷ target food cost, and when to ignore the answer

Target-cost pricing sets a menu price by dividing plate cost by the food cost percentage you want to run:

$$\text{Menu price} = \frac{\text{plate cost}}{\text{target food cost \%}}$$

For the Hearth Chicken, at the plan's 30% target:

$$\frac{\$8.52}{0.30} = \$28.40$$

Bellwether prices it at **\$29.00**, which gives 29.4% and a contribution margin of \$20.48.

That sixty-cent difference between the formula's answer and the actual price is the subject of this section, and it is worth \$0.60 × 4,992 plates a year = \$2,995 of pure contribution — money that exists only because somebody declined to obey a formula.

Run it in reverse; it is more useful that way

Inverted, the same formula answers a better question. Given a price the market will bear, what plate cost can you afford?

$$\text{Allowable plate cost} = \text{menu price} \times \text{target food cost \%}$$

At \$29.00 and a 30% target, \$8.70. The Hearth Chicken at \$8.52 has eighteen cents of headroom — which tells you how much room there is to add a component, absorb a price increase, or improve a garnish before the item stops meeting the target.

This is how experienced operators actually use the number. You do not usually get to pick your price; the market, the neighborhood, and your price ladder pick most of it (Chapter 10 owns that argument). What you pick is what goes on the plate for the money. Target-cost pricing is a smoke detector, not a thermostat. It is very good at telling you an item is wildly wrong and bad at telling you what to charge.

🧮 Run the Numbers

Five items, one target, and why obeying it would cost \$34,916 a year.

Bellwether's five costed entrées, each priced strictly to the plan's 30% target versus what the menu actually charges:

Item	Plate cost	Price at 30%	Actual price	Actual FC%	CM
Hearth Chicken	\$8.52 | \$28.40	\$29.00 | 29.4% | \$20.48
The pork chop	\$10.09 | \$33.63	\$31.00 | 32.5% | \$20.91
Squash and grains	\$3.95 | \$13.17	\$24.00 | 16.5% | \$20.05
Ember trout	\$10.40 | \$34.67	\$28.00 | 37.1% | \$17.60
The burger	\$5.10 | \$17.00	\$21.00 | 24.3% | \$15.90

Squash and grains. The formula says charge \$13.17; the menu says \$24.00. The formula's price earns \$13.17 − \$3.95 = \$9.22** of contribution; the actual price earns **\$20.05. Obeying the formula on this one item gives away \$10.83 a plate** — at 62 plates a week, **\$34,916 a year. A composed plate of roasted autumn squash, grains, ember tomato, and ricotta salata is worth \$24 in this market, and that has nothing to do with grains being cheap. Note what this produces: a vegetarian dish contributing \$20.05, within forty-three cents of the signature Hearth Chicken's \$20.48. The formula prices ingredients. The guest prices dinner.

The ember trout. The opposite case, and the more painful one. The formula says \$34.67. The menu says \$28.00** — because a mid-priced fish in a neighborhood restaurant has a ceiling, and \$35 would not sell. Obeying the formula here does not gain \$6.67 a plate; it loses the item. So the trout runs a 37.1% food cost, seven points above target, and it is still worth carrying: 45 plates a week at \$17.60 of contribution is **\$792 a week, \$41,184 a year that would not exist if the dish were priced to the ratio and stopped selling. What matters is that somebody priced the decision instead of drifting into it.

And the blend. Weighted across the volumes in §11.1, these five items produce a 28.0% food cost while individual items range from 16.5% to 37.1%. You do not need every item at 30%. You need the blend at 30%, which is a different and much easier problem.

The five times to ignore the answer

1. When the item is cheap to make. The formula caps your price at plate cost × 3.33, which has nothing to do with what the dish is worth to a guest. Pasta, grains, eggs, beans, vegetables, and most desserts fall here. Charge what the dish is worth and bank the dollars.

2. When the item is expensive to make and the market has a ceiling. The formula prices you out. Accept a worse percentage for better dollars — high-cost, high-price items are usually your best contributors, and the ember trout is the example — 37.1% food cost, and worth every point of it.

3. When labor, not ingredients, is the real cost. The card excludes labor (§11.2), so target-cost pricing systematically under-prices dishes that take three hours of skilled work and over-prices dishes assembled to order. A twelve-hour braise and a composed salad can share a plate cost and must not share a price.

4. When the price is a positioning statement. The top and bottom prices on a menu tell the guest what kind of restaurant this is before they read a dish. Chapter 10 owns menu psychology and the price ladder; respect it here.

5. When it is the signature dish. A signature item's price is load-bearing for the whole room's perception. Bellwether's chicken is \$29 and stays \$29; that number is doing brand work. When its cost moves — and §11.8 moves it — the price is the last lever you reach for.

🤝 Hospitality

The price you can defend at the table.

There is a test worth applying to every price on your menu, and it has nothing to do with arithmetic. Can a server say this price out loud without flinching?

Servers know. One who believes an item is a rip-off will steer guests away from it without ever deciding to, through a half-second hesitation and a slightly warmer description of something else. You will see it in the mix and never know why. One who believes an item is a genuinely good thing to buy will sell it all night — and the Hearth Chicken at \$29, with real hardwood, a good bird, and enough on the plate, is exactly that kind of item.

This is why the honest response to a cost increase is rarely to quietly shrink the portion. A guest who has eaten the dish four times will notice the fifth one is smaller, and will not conclude that input costs have risen. They will conclude that you are getting away with something — a judgment that is expensive to reverse and impossible to see on a P&L.

Raise the price and say why, or hold the price and absorb it, or change the dish and present it as a change. All three are defensible. The secret shrink is the only one that costs you the second visit, which Chapter 23 will show is the visit that actually pays.

11.6 Portion control: scoops, scales, ladles, and the four-ounce that became six

Portion control is the set of tools, specs, and habits that make the portion on the plate equal the portion on the card. Without it, a cost card is a work of fiction — a careful, internally consistent document describing a plate that is not being served.

Nobody decides to over-portion. It happens through five ordinary mechanisms, and the countermeasure differs for each. The card was never communicated — the spec is in a binder in the office and the cook has never seen it. The cook was trained by another cook, who was trained by another cook, and the portion has drifted upward two percent a generation for three years. The plate is bigger than the food — a wider rim makes a correct portion look mean, and every cook in the world fixes that by adding food. The rush — on a Saturday at 7:40 nobody weighs anything, and whatever the hand knows is what the plate gets. And generosity is a virtue, and cooks have it: the one nobody says out loud. Over-portioning is very often an act of care, and treating it as a discipline problem rather than a specification problem will cost you a good cook.

The tools

The disher numbering system is simple and people still get it wrong: the number is scoops per quart.

Grab a #12 when the card says #16 and you have plated 33% more product — every plate, invisibly, with no possibility of noticing by eye. That is portion drift in one sentence, and it is why the disher number belongs on the card and the disher lives in a labeled slot rather than a drawer.

⚠️ Where the Money Leaks

The drift card: how a 29.4% item becomes a 32.8% item without anybody doing anything wrong.

Three ordinary things happen at Bellwether over about six weeks. None is a scandal, none is visible to the naked eye, and nobody is dishonest, lazy, or careless.

• The poultry order is placed against a weight range, and the birds that arrive average 3.7 lb rather than 3.5. Nobody weighs them, because they look like chickens.
• The roasted roots get plated at about 6.5 oz instead of 5, because the new plate is wider and five ounces looks stingy on it.
• The salsa verde is free-poured with a spoon rather than a 1.5 oz ladle and averages 2.0 fl oz, because a generous spoon of salsa verde is a nice thing to give somebody.

Line	On the card	On the line	Difference
Chicken, ½ bird	\$5.60 (3.50 lb) | \$5.92 (3.70 lb)	+\$0.32
Roasted roots	\$0.95 (5 oz) | \$1.24 (6.5 oz)	+\$0.29
Salsa verde	\$1.05 (1.5 fl oz) | \$1.40 (2.0 fl oz)	+\$0.35
Butter and aromatics	\$0.42 | \$0.42	—
Oil, salt, misc	\$0.18 | \$0.18	—
Garnish	\$0.15 | \$0.15	—
Components	\$8.35** | **\$9.31	+\$0.96
+ 2% waste allowance	\$0.17 | \$0.19	+\$0.02
Plate cost	\$8.52** | **\$9.50	+\$0.98
Food cost at \$29.00	29.4%	32.8%	+3.4 pts
Contribution margin	\$20.48 | \$19.50	−\$0.98

Three point four points on the signature dish. Ninety-eight cents a plate; on 4,992 plates a year, \$4,892 — one item, three invisible drifts, six weeks.

This is exactly the cost drift Chapter 1 named as the second of the four killers, with the decimals filled in. And notice: no report in the building shows it. The invoice shortcut won't. The monthly P&L won't, for seven weeks. Two things catch it — a scale at the station, and the ideal-versus-actual variance in §11.7.

👨‍🍳 On the Line

The three plates, and the twenty seconds.

Two habits catch almost all portion drift, and both fit inside a real service.

The first three plates. At the start of every service the chef or sous weighs the first three plates of the two or three highest-volume items — out loud, in front of the station, as a normal part of setting up. Ninety seconds. It re-calibrates the hand for the whole night and makes the spec a shared fact rather than a private accusation.

Weigh the protein, every time. A scale on the hearth and a rule that every bird gets weighed before halving adds about twenty seconds a plate. At 96 plates a week that is thirty-two minutes; at \$19 an hour fully loaded, about **\$10 a week — \$527 a year — against a drift worth thousands on that item alone. It is a prime cost** trade: spend labor to buy food cost, and the total falls.

What does not work: memos, a laminated card on the wall, telling people to be careful. If the correct portion requires judgment you will get judgment, and judgment varies by cook, by hour, and by how the night is going.

The honest limit. Portion control fails when it fights the concept. If your restaurant sells generosity, policing to the gram hollows out the thing guests come for. The answer is not a secret shrink; it is to cost the generous portion honestly and price it, which is what §11.5 was for.

11.7 Ideal vs. actual: the variance that tells you what's really happening

This is the section that turns costing from paperwork into management.

You now have two numbers describing the same period. Ideal food cost (also called theoretical food cost) is what the period should have cost: every item sold, multiplied by its plate cost. Actual food cost is what it did cost: beginning inventory plus purchases minus ending inventory. The gap is the ideal-versus-actual variance, one of the two or three most valuable numbers an independent restaurant can produce.

$$\text{Ideal food cost \$} = \sum_{\text{items}} (\text{units sold} \times \text{plate cost})$$

$$\text{Variance} = \text{actual usage} - \text{ideal usage}$$

Read what the variance is, because the definition is the insight. Ideal cost is what the menu, the cards, and the sales mix say you consumed. Actual cost is what the walk-in says you consumed. Everything in between is waste, over-portioning, spoilage, uncosted items, comps, receiving errors, theft, and counting mistakes — in some proportion you now have to determine. The variance does not tell you which. It tells you how much is unexplained.

Computing the ideal

You need three things: a costed menu, a unit count by item from the point-of-sale system, and the discipline to include everything — specials, off-menu items, staff meals if you count them in food cost, and comps.

Bellwether's five-item extract from §11.1 gives ideal cost \$2,447.98 against \$8,739.00 of sales: 28.0%. A real calculation runs every item and every unit sold in the period; the method is identical, and it is what inventory software is actually for.

Two cautions matter more than the arithmetic. An ideal computed from stale cards is worthless — if your cards were built in January and the chicken contract moved in March, your variance will show a problem in the kitchen that is really a problem in the office. And ideal moves when mix moves: if your ideal percentage changes between periods, the first question is not what happened in the kitchen, it is whether the mix shifted. This is why you compare actual against ideal, not actual against last month.

The variance, read properly

Return to the constructed restaurant from §11.1. Its cost cards produce an ideal food cost of 28.9% in both periods — the mix did not move.

Period 1 is a healthy restaurant. Period 2 is a restaurant with a problem, and — this is the value of the method — the ideal held steady, so the entire three-point deterioration is usage. Not pricing. Not mix. Not the menu. Something is leaving the building without being sold.

Meanwhile, the invoice shortcut was telling this operator that Period 2's food cost had improved by 4.5 points.

🧾 Read the Numbers

```text FIGURE 11.5 — "The variance report that found it" [constructed teaching example] THE ARTIFACT Period 2 ideal-versus-actual food cost report, four weeks, produced from cost cards, POS unit counts, and a physical inventory count. THE CONTEXT Third-year neighborhood restaurant. Period 1's variance was 0.1 points and nobody looked closely. Period 2 followed a large forward buy on poultry and beef, a new prep cook in week two, and four weeks of unpriced daily specials.

                 Food sales                                      $88,000
                 Ideal usage    (from cost cards × units sold)   $25,432    28.9%
                 Actual usage   (BI $23,500 + P $25,660 − EI $21,000)
                                                                $28,160    32.0%
                 ──────────────────────────────────────────────────────────────
                 VARIANCE                                        $2,728     3.1 pts

                 Investigated over eleven days:
                   Protein over-portioning (birds over-weight; line portions)  $980
                   Waste log: over-production and spoilage of the forward buy  $640
                   Daily specials sold below their uncosted plate cost         $520
                   Comps and remakes, never costed against usage               $310
                   Receiving: two short deliveries and one uncaught price move $180
                   Unexplained after recount                                    $98
                   ──────────────────────────────────────────────────────────────
                   TOTAL EXPLAINED                                          $2,728

WHAT IT SHOWS Three points of food cost, decomposed into five nameable causes and a $98 residual. Not one of the five is theft. The largest single item is portioning, which is a training and tooling problem; the second is spoilage of inventory the restaurant bought on promotion and could not use in time. WHAT IT DOESN'T It does not prove the decomposition is right — these are the operator's best attributions after eleven days of looking, and the $98 residual is honest rather than reassuring. It cannot distinguish spoilage from unrecorded waste, or over-portioning from theft, without controls that Chapter 34 supplies. And it says nothing about whether the forward buy was profitable on a cash basis. THE DECISION Scales on two stations and the first-three-plates check, starting tomorrow. Every special gets a cost card before it is sold, no exceptions. Comps costed weekly against plate cost. Stop buying promotional volume the walk-in cannot turn in ten days. Recount weekly for six weeks, then reassess. THE LESSON The variance does not tell you what happened. It tells you how much happened, which is what buys you the right to go looking — and it is the only report in the building that would have contradicted the invoice shortcut in time. ```

What is a normal variance?

Industry practice generally treats an ideal-to-actual gap under about one point as normal operating friction in full service, one to two points as worth investigating, and more than two as a real problem requiring action. Those are rules of thumb, and they tighten in high-volume, low-item-count operations and loosen where there is a lot of scratch production and daily specials.

Set the threshold in advance — say 1.0 point or \$500, whichever is smaller — and commit to investigating anything above it. Setting it before you see the number is what keeps you from rationalizing a bad period.

The order to investigate in

When a variance appears, investigate in this order — deliberately the reverse of instinct.

1. Count and math errors. Somebody priced the sheet wrong, counted a case as an each, or missed a shelf. The commonest cause of a large surprise variance and the cheapest to check; recount two or three high-value categories.
2. Receiving. Short deliveries, substitutions accepted without a credit memo, price increases nobody flagged. Reconcile a week of invoices against the order guide (Chapter 13).
3. Uncosted items. Specials, off-menu requests, dishes that hit the menu before the card was built. The commonest genuine cause in restaurants that change the menu often.
4. Portioning. Weigh plates, weigh batches, compare against the cards.
5. Waste and spoilage. Read the waste log; if there is no waste log, that is your finding.
6. Comps, voids, and staff meals. They consume product and produce no sales. Cost them.
7. Theft — last.

Everybody's instinct is to start at seven, and it is almost always wrong. Theft is real, it happens, and Chapter 34 takes it seriously — but it is the least likely single explanation for a first large variance, the slowest and most expensive to investigate, and the most destructive to a kitchen's morale if you lead with it. Work the list in order. If you have cleared one through six honestly and the variance persists, you have a controls problem, and Chapter 34 is where you go.

The limits of the variance

Say it plainly, because a section this useful invites over-trust. A variance is only as good as your cards — garbage cards produce a confident, precise, meaningless number. It cannot separate causes: over-portioning and theft look identical in the arithmetic, and so do spoilage and unrecorded waste. It is noisy at small scale, since one miscounted case of protein can produce a point of apparent variance, which is why the trend across several periods is worth more than any single period and why you recount before you act. And it does not measure cash — a restaurant can hold a beautiful variance and still not make payroll, which is Chapter 33's subject entirely.

🔍 Check Your Understanding

1. A period shows food sales of \$92,000, ideal usage of \$26,680, and actual usage of \$28,520. Compute ideal %, actual %, and the variance in dollars and points. How would you characterize it?
2. Your ideal food cost percentage rose from 28.4% to 29.6% while your variance stayed at 0.5 points. What happened, and is it a problem?
3. Why is "theft" the last item on the investigation list rather than the first?

(1: Ideal 26,680 ÷ 92,000 = 29.0%; actual 28,520 ÷ 92,000 = 31.0%; variance \$1,840 = 2.0 points — at the threshold where this stops being friction and becomes a problem to act on. 2: The mix moved toward higher-cost items, or prices changed, or cards were updated after a cost increase — the kitchen is executing fine. Whether it is a problem depends on contribution margin, not percentage, and that is Chapter 12's question. 3: It is the least likely single cause of a first large variance, the slowest and costliest to investigate, and leading with it damages the team you need in order to fix the six more probable causes.)

11.8 Re-costing: contracts, seasonality, and what to do when the protein price moves

A cost card is a snapshot, and snapshots decay. Produce moves with the season and the weather; proteins move with commodity cycles, disease, feed costs, fuel, and trade policy; dairy, oils, and packaging all move, and packaging is inside a lot of your cards whether you have a line for it or not.

So the question is not whether to re-cost, but on what cadence, and what to do about the answer.

The re-costing cadence

Re-costing everything constantly is a way to spend a manager's life producing no decisions. Use three tiers.

The trigger threshold is the important one. It converts re-costing from a chore into an exception process, so the work happens when it matters rather than on a calendar.

When the poultry contract moves

🧮 Run the Numbers

Chicken goes from \$3.20 to \$3.68 a pound.

A fifteen percent increase — the kind of move that follows an avian-influenza event or a broad feed-cost rise. Everything else on the card holds.

New protein line: 3.5 lb × \$3.68 = \$12.88 a bird; half a bird = **\$6.44**, up \$0.84.

Line	Before	After
Chicken, ½ bird	\$5.60 | \$6.44
Roasted roots	\$0.95 | \$0.95
Salsa verde	\$1.05 | \$1.05
Butter and aromatics	\$0.42 | \$0.42
Oil, salt, misc	\$0.18 | \$0.18
Garnish	\$0.15 | \$0.15
Components	\$8.35** | **\$9.19
+ 2% waste allowance	\$0.17 | \$0.18
Plate cost	\$8.52** | **\$9.37
Food cost at \$29.00	29.4%	32.3%
Contribution margin	\$20.48 | **\$19.63**

Eighty-five cents a plate. At 4,992 plates a year: \$4,243 off the bottom line, from one ingredient, on one dish.

Now the four responses, priced.

(a) Absorb it. \$4,243 a year on this item. The dish still contributes \$19.63 a plate, which is excellent. If the increase looks seasonal this is frequently right, and it costs you nothing with guests.

(b) Reprice to restore the dollars. \$9.37 + \$20.48 = \$29.85, so price at **\$30.00. Food cost 31.2%, contribution margin \$20.63** — slightly better than before.

(c) Reprice to restore the percentage. \$9.37 ÷ 0.294 = \$31.87, so price at \$32.00. Contribution margin becomes \$22.63**, \$2.15 a plate more than before the cost increase. Chasing the percentage overshoots:** it takes a \$3 price increase to undo an 85-cent cost increase, and it does so by making the guest pay for the ratio's arithmetic rather than for the chicken. This is the commonest pricing error in the industry and it is why (b) exists.

(d) Re-spec. Buy a 3.25 lb bird: 3.25 × \$3.68 = \$11.96, half = \$5.98, components \$8.73, plate cost \$8.90** — food cost 30.7%, contribution margin \$20.10. Nearly all the damage repaired without touching the price. And the guest gets a smaller chicken**, which on a signature dish is a real risk (§11.5). Defensible if you say so; corrosive if you don't.

The fifth option, not on the card: contract forward. A fixed price for six or twelve months converts a variable into a constant; you give up the upside if the market falls, plus usually a volume commitment you must take. Chapter 13 owns that decision — but the cost card is what tells you how much the certainty is worth.

The answer for Bellwether's plan, for the record: hold the price at \$29.00, absorb a short-term move, and re-cost at the contract renewal. The signature price is doing brand work and the item still contributes nearly twenty dollars a plate. But the plan should say so explicitly, and say what would change the answer — a sustained price above roughly \$4.00 a pound puts the plate near \$10.00 and the item at 34.3%, and at that point the conversation is a real one. (3.5 × \$4.00 = \$14.00; half = \$7.00; components \$9.75; +2% = \$0.20; plate cost **\$9.95; ÷ \$29.00 = 34.3%.)

Seasonality, and the trap of costing in June

Produce cards costed in high season are lies in February. Build your cards in the summer when the tomatoes are \$1.20 a pound and never revisit them, and your winter food cost will run above ideal every year while your variance insists the kitchen has a problem it does not have.

Two honest fixes. Cost seasonal items at a blended annual price, and accept running below ideal in season and above it out of season — simple, and it works for items you carry year-round. Or change the menu with the season and cost each menu at the prices you will actually pay while it runs. That is better, and more work, and it is one of the underrated financial arguments for a seasonal menu: not that the food is better, though it is, but that a menu you rewrite four times a year is a menu you re-cost four times a year.

⚖️ Code and Compliance

Re-speccing has legal edges, and they are sharper than they look.

When you change an ingredient to manage cost, three obligations follow you.

Truth in menu. If the menu says "air-chilled chicken," "local," "wild-caught," "grass-fed," "house-made," or names a specific breed, region, or producer, that claim must remain true after you re-spec. Federal and state consumer-protection law, health codes, and in some places specific menu-labeling rules all reach this, and enforcement is real. The cheapest control is a rule: any spec change routes past whoever owns the menu copy, before it is ordered.

Allergens. Substituting an ingredient can introduce a major allergen into a dish that did not have one — a different oil, a thickener, a stock base, a sauce with soy or wheat in it. A guest who has ordered that dish safely for a year has no reason to re-ask. Every spec change must be checked against the allergen matrix and communicated at pre-shift. Chapter 25 covers allergen management properly; this is the corner where the change actually originates.

Menu-labeling and nutrition disclosure, where it applies. Requirements differ sharply by jurisdiction and by the size and type of operation, and independents are frequently outside them — but "frequently" is not "always," and a spec change can move you.

All of this varies by state, county, and city, and it changes. Verify locally, and for anything consequential use an attorney. What does not vary is the operating principle: a cost decision that changes what a guest is eating is not only a cost decision.

🍽️ The Business Plan

Checkpoint 11 of 40 — the costed menu, and the 30% line is earned.

Until now the plan's 30% food-cost target has been an assertion. Chapter 4 wrote it into the pro forma because a full-service restaurant of this type ought to be able to run 30% — a respectable reason and not evidence. This chapter converts it into arithmetic somebody can check.

What this checkpoint adds: the costed menu, and the bridge from the cards to the line.

The cost cards

Every card in the plan's appendix is built the way §11.2 built the Hearth Chicken: from batch recipes, measured yields, and portion specs, footing to the cent, with the assumptions written on the card. That last detail is what a lender is reading for. Anyone can print a number. A card that says this yield came from three tests on three deliveries and this price came from the invoice of March 14th is a different kind of document.

The bridge from the cards to the plan's line

The weighting behind that 28.1%, stated so it can be argued with:

What this checkpoint settles. The 30% is now reachable rather than asserted. The cards are real, the Hearth Chicken foots to \$8.52 against a \$29.00 price, and the theoretical blend at 28.1% leaves genuine room underneath the target.

What it does not settle, and this is the honest part. Theoretical cost including the Q factor is 29.7%, and a well-run kitchen still produces half a point to a point of ideal-to-actual variance. That puts a realistic year-one landing zone at 30.2% to 30.7%, and the plan's line is 30.0%.

The plan's line does not move. What moves is that we now know what has to be true for it to hold:

1. Variance held below 0.3 points — tighter than most operations achieve, and requiring weekly counts from opening week, not from month three.
2. Or the Q factor managed. Reducing bread service from \$0.48 to \$0.36 a cover — a smaller, better piece served on request — recovers 0.4 points on its own.
3. Or the difference accepted and disclosed. Half a point of food cost is \$5,580 a year on \$1,116,000 of food sales: 0.36% of total sales, moving prime cost from 60.0% to 60.4% — still inside the full-service benchmark, but the cushion is visibly thinner.

That third line is the one to sit with: a fraction of a point, on one cost line, moving the number that predicts survival. Nothing here is a crisis. Everything here is exactly the kind of small, invisible, compounding thing that closed the restaurant in Figure 1.3.

What the plan commits to as a result — this is a new section, and it is worth more than the cards:

• Physical count of the walk-in, the freezer, and dry storage weekly, from opening week.
• Food cost computed from the count, never from invoices.
• Ideal-versus-actual computed weekly against the cards, with a 1.0-point action threshold.
• Every special costed before it goes on the board.
• Full menu re-cost twice a year; trigger re-cost on any component moving 10% or more.
• Scales on the hearth and garde manger stations from day one, and the first-three-plates check at every service.

Open questions carried forward:

1. Who actually counts the walk-in, on which night, and how long does it take? (Chapter 13)
2. Is the 2% waste allowance enough, and what will the waste log say? (Chapter 13)
3. Which items should the prices actually be, once the menu is treated as a portfolio rather than a list? (Chapter 12)
4. Does the kitchen have the labor to hold portion discipline at 95 covers on a Saturday? (Chapters 14 and 19)
5. What happens to every one of these cards when the poultry contract renews? (§11.8, and Chapter 13)

Conclusion

Food costing is the part of this business that most rewards being boring about it.

Everything in this chapter is arithmetic a competent fourteen-year-old could do — no derivation, no model, nothing you could not check on the back of a guest check. What there is instead is a set of disciplines almost nobody maintains, and the gap between the people who maintain them and the people who don't is measured in points of prime cost, which is to say in whether the restaurant is open in three years.

Four things to carry out.

Food cost is what you used, not what you bought. The invoice shortcut is wrong by the change in your inventory divided by your sales, and in a single period it can point in the opposite direction from the truth — which is exactly what it did in Figure 11.1, reporting a 4.5-point improvement in the period the restaurant lost three points. Software cannot fix this. Somebody has to count.

A cost card is a purchasing document, not a profitability document. The Hearth Chicken costs \$8.52 and consumes \$9.94 once you count the labor to butcher and brine it and the hardwood to cook it. The card is not wrong; it answers a narrower question than the one an owner has. The wider question is prime cost, and it has been the wider question since Chapter 1.

The price is not the output of the formula. Target-cost pricing is a smoke detector. Obeying it on a cheap-to-make dish gives away real money — \$34,916 a year on one plate of squash and grains — and obeying it on an expensive one prices you out of your own market.

The variance is where the truth lives. Ideal versus actual is the only report in the building that would have contradicted that operator in time. It does not tell you what happened; it tells you how much happened, and it buys you the right to go looking — in the right order, with theft last.

Chapter 12 takes the same cards and asks a better question. Not what does this item cost? but which items make me money, how often, and what should I do about the rest? Menu engineering turns the plate costs and contribution margins you just built, plus the unit counts you already used to compute your ideal, into a portfolio with four quadrants and four different actions. The costing was the unglamorous half. The next chapter is where it starts paying.

Key Terms

Food cost percentage — food cost of goods sold divided by food sales, where cost of goods sold means beginning inventory plus purchases minus ending inventory. The most quoted number in the industry and the most frequently mis-measured. (Ch. 11)

Plate cost — the total product cost of one plated portion: costed recipe components plus a waste allowance. The only honest basis for a menu price. (Ch. 11)

Recipe cost card — the document listing every component of one portion with its quantity and cost, totalling to the plate cost. Requires a standardized recipe and a portion spec first. (Ch. 11)

As-purchased (AP) cost — the price of an ingredient as delivered, including peel, bone, stem, fat, and trim. What the invoice says. (Ch. 11)

Edible-portion (EP) cost — the cost of the usable product after trim, butchery, and prep loss; AP cost divided by yield percentage. What belongs on a cost card. (Ch. 11)

Yield percentage — the proportion of as-purchased weight surviving as usable product. Because it is less than one, dividing by it raises cost: a 75% yield means the ingredient costs 133% of the invoice price. (Ch. 11)

Yield test — a controlled measurement in which a product is weighed as purchased, broken down the way your kitchen actually breaks it down, and every output stream weighed separately. Called a butcher test for meat and fish, where by-products are credited at the price of what they displace. (Ch. 11)

Waste allowance — a percentage added to costed components to cover trim, spillage, remakes, and small unmeasured losses; commonly 2–5%. The softest line on a cost card. (Ch. 11)

Portion control — the tools, specs, and habits that make the portion on the plate equal the portion on the card: scales, numbered dishers, sized ladles, pre-portioning, the first-three-plates check. A physical-systems problem, not an exhortation. (Ch. 11)

Target-cost pricing — setting a menu price by dividing plate cost by a target food cost percentage. A screen for items that are wildly mispriced, not a rule for every price. (Ch. 11)

Ideal vs. actual food cost — the comparison of theoretical usage (units sold × plate cost) with counted usage (beginning inventory + purchases − ending inventory). The gap is waste, over-portioning, spoilage, uncosted items, comps, receiving errors, and theft, in unknown proportion. (Ch. 11)

Q factor — the per-cover cost of items served with a meal but not separately priced: bread and butter, condiments, table oil, coffee accompaniments. Frequently worth a point or more of food cost and almost never costed. (Ch. 11)

Spaced Review

1. A standardized recipe (Chapter 10) must exist before a cost card can be built. Explain in two sentences why — and what specifically goes wrong when a kitchen costs a dish it has not standardized.
2. Chapter 10 argued for cross-utilization: one delivery becoming six dishes. Name two distinct ways cross-utilization shows up in this chapter's arithmetic, and one cost it imposes that the arithmetic does not show.
3. From Chapter 1: state the correct formula for food cost percentage from memory, then state what the invoice shortcut's error is exactly equal to. A restaurant with \$21,000 of beginning inventory, \$24,000 of ending inventory, and \$110,000 of food sales in the period — how many points is its shortcut off by, and in which direction?
4. From Chapter 4: the plan's assumptions register lists a 30% food-cost assumption. After this chapter, is that assumption more or less credible, and what single piece of evidence would you now attach to it?
5. The recurring question: the chef proposes putting a portion scale on the hearth station and requiring every bird to be weighed before it is halved — about twenty seconds a plate. Does this decision move prime cost, and in which direction? Show the arithmetic on both halves.