Plate costing: the myth that costs four margin points

How wrong is plate costing built on purchase weight?

Between 12% and 30% off on proteins and produce, and that spread isn't a napkin estimate: it comes from comparing the kilo your supplier invoices against the kilo that actually crosses the pass.

A tenderloin arriving at 1,000 grams leaves at 620 once you strip the chain, bone it out and lose weight in the sear, so a recipe card written on those 1,000 grams lies by 38% in the denominator. The two-unit steakhouse that opened this analysis billed 1.4 million dollars a year and believed it ran 28% food cost; inventory came back at 34.4%. Six and a half points. Eighty-four thousand dollars nobody was stealing, sitting in no line of the management P&L, evaporating in the gap between what was bought and what was served. Median full-service food cost closed 2024 at 32.0% of sales (National Restaurant Association), meaning that steakhouse ran two points below par without knowing it.

The right denominator is portions served, not portions promised

Divide preparation cost by the portions you actually rang up, not by the ones your recipe card promises. A 12-liter stockpot does not yield 40 servings of 300 milliliters: it yields 37 or 38, because 3% to 7% disappears into taste tests, returned plates, comps for the annoyed guest and that last ladle that never quite fills. It reads like penny accounting until you multiply it: in a 120-cover-a-day operation that single adjustment moves a full point of food cost, and a point on a million in sales is ten thousand dollars. Limited-service came in at 32.4% of sales during 2024 according to the food cost ratios the National Restaurant Association publishes, barely four tenths above full service, which dismantles the lazy assumption that quick formats buy better. They buy the same. They portion differently. Weigh the output of twenty plates during any ordinary shift and brace yourself: deviations against the recipe card usually run 15% to 25%, and they grow on Friday at nine, when the hand turns generous and nobody watches the scale.

With no standard recipe on paper, every cook invents the gram weight

No spreadsheet catches that from an office, because the file says whatever you typed the day you typed it. A standard recipe isn't paperwork: it's the only document that turns a kitchen opinion into an auditable number. And it has a side effect in the trash. Waste costs the U.S. restaurant industry roughly 162 billion dollars a year according to The Restaurant HQ, a figure that aggregates spoiled product, overproduction and precisely that extra ladle nobody records. Document gram weight, a plating photo and expected trim loss per recipe. Three fields. Nothing more. That's the conceptual mistake with the biggest price tag, and the tidy owners commit it most, the ones who want every dish to "carry its share". It doesn't. Fixed payroll, rent, utilities and depreciation on the new walk-in live in the break-even calculation, not in the recipe card, because none of them move when you sell one more tenderloin.

Payroll and rent never belong on the plate

Loading them onto the plate inflates unit cost, pushes prices out of market and kills anchor dishes that were working fine. There is one legitimate exception that does belong to the variable cost of a sale: channel commission. Uber Eats and DoorDash charge 15% to 30% of the ticket, Grubhub 15% to 25% (Rezku fee analysis, 2026), and cards take an average 2.35% per transaction according to the Texas Restaurant Association. A dish running 30% food cost in the dining room hits 60% through delivery. Your bank won't take percentage points, and that's where half of a well-meaning menu sinks. Compare two dishes: a 14-dollar salad at 24% cost leaves 10.64 dollars per cover; a 38-dollar steak at 38% cost leaves 23.56. The salad wins the percentage contest and loses the till by thirteen dollars on every table. At 80 covers a day across 300 operating days, that gap is 316,800 dollars of gross contribution.

Contribution margin gets banked in dollars, not percentages

This is why serious menu engineering ranks by absolute contribution times popularity, and only then looks at the ratio. The contraction of the full-service segment, which Technomic measured at roughly 18% fewer locations than in 2019, punished exactly those operators who defended a pretty percentage while their average ticket drained. Rank your menu by dollars. We'll argue about points afterward. Three scenarios, three different cadences. Small venue, under 60 covers a day and revenue below 400,000 dollars a year: cost by hand the six dishes driving 70% of sales, review them every 90 days and live comfortably with a two-point variance band across the rest of the menu. Mid-size operation, 120 to 200 covers and sales between 800,000 and 1.5 million: weekly costing of the top ten, output weighing one shift a month, biweekly inventory on your ten most expensive SKUs.

How to read these numbers in YOUR operation?

Group of three units or more: the discipline changes nature, because you're no longer chasing pennies but consistency between kitchens; there the recipe card gets centralized, each unit reports variance against the same standard and purchasing is negotiated on consolidated volume.

That last tier is where Masterestaurant works with most of its clients. The food cost ratios cited above —32.0% in full service and 32.4% in limited service, 2024 medians— come from the operator analysis published by the U.S. industry association. Channel commissions were compiled by Rezku in its 2026 fee comparison, the card figure was reported by the Texas association in November 2025, and the waste number is aggregated by The Restaurant HQ. Now the limits, which matter as much as the figures: these are United States medians, not your city's; a median hides that half of all operators sit above it; and no benchmark carries your menu mix, your supplier or your yield rate in the walk-in.

Where these benchmarks come from and what they won't tell you?

They're useful for knowing whether you're playing inside the trade's range. They're useless for setting your price. That one comes out of your kitchen.

Suppose that letter lands on a Monday. The reflex is to raise the menu price and hope the guest doesn't notice. Follow the thread: you push 18%, the dish goes from 38 to 44.84 dollars, crosses the psychological forty-dollar line, demand drops maybe 12%, and with fewer covers the rent burden per ticket rises, so break-even moves further away right when you thought you'd protected it. The real alternative is redesigning the cut: dropping the portion from 280 to 240 grams with a more generous side, or moving that dish to the menu's secondary visual position and pushing another with similar contribution and stable protein. Diego F. Parra puts it this way in Masterestaurant audits: price is the last lever you touch, because it's the only one the guest can see.

What would you do if your supplier raised tenderloin 18%?

Review your yield before you review your tariff. The biggest gap sits in the denominator, not the formula. Dividing by theoretical portions assumes a kitchen cutting with laboratory precision;

dividing by portions actually served absorbs the 3% to 7% lost to tests, service errors and comps. That single adjustment moves a full food cost point in operations above 120 covers a day. Documentation is the second break. Where no standard recipe exists on paper, every cook invents a gram weight and variance explodes: weighing 20 plated dishes across one shift usually exposes 15% to 25% deviations against the spec, and no spreadsheet catches that from an office. Then comes the conceptual error with the highest price tag: charging the plate for what belongs to structure. Fixed payroll, rent, utilities and depreciation stay out of plate costing —they live in the break-even— yet once mixed in, the operator raises prices to patch an occupancy problem, kills demand and deepens the very hole he meant to close.