Plate costing: the 35% food cost myth and what your bank account actually pays

The dish that lost $1.10 in margin without anyone touching the price

The best-selling dish at a Medellín steakhouse delivered 11,400 pesos of contribution margin on the January 2026 menu and 6,900 on the July one, with the exact same selling price printed on the card. Three things shifted underneath: the loin went up 18%, trim loss climbed from 22% to 27% when a new supplier came in, and the recipe costing still lived in a 2024 spreadsheet nobody had opened since the previous chef walked out. Cash doesn't drain from lack of volume, it drains through this silent erosion that no sales report will ever show you. The backdrop pushes the same way: the producer price index for all food in the U.S. sits 35% above its February 2020 level (USDA ERS / BLS 2026), while Colombian restaurants raised prices 9.8% from February 2025 (ACODRÉS 2025) just to hold 98,000 jobs in place. Charge the plate with raw material at its real measured yield loss, and nothing else.

What belongs in the plate cost, and what the sum of margins pays for?

Spreading payroll, rent and utilities into unit cost inflates that cost by 18 to 24 points and produces selling prices the market simply will not pay.

The logic test takes one sentence: sell one more plate tonight and your rent doesn't move a single peso, your fixed payroll doesn't either; the only thing that moves is what you spent on protein, side and sauce. Costs that depend on the month rather than on the unit sold belong in the break-even calculation, where they get covered by the SUM of the contribution margins of everything that crossed the pass. Diego F. Parra has taught it this way at Masterestaurant for twenty years, and the rule hasn't budged, because the arithmetic of variable versus period costs doesn't change with software fashion. The lower a dish's food cost percentage falls, the less money that dish usually contributes to the register, and that trap catches almost every owner.

The food cost paradox: cutting the percentage can drain your register

Look at real menu numbers: a salad at 19% food cost leaves 8,000 pesos per sale; a steak cut at 34% leaves 26,000 pesos. Chase the percentage and you tilt the menu toward salad, celebrate a 26% average, then discover in March that break-even moved three million pesos further away. The way out demands tracking both at once —percentage and pesos— and deciding ALWAYS with the second, keeping 32% as a hard safety ceiling rather than a target to chase. The percentage protects you from a pricing mistake; the pesos pay the rent. The tiers I see across Latin America today fall into three bands, and you should read them by what they deliver, not by the sticker. From 0 to 400 USD sits the artisanal level: your own recipe-costing template, a gram scale, two weeks of weighing trim loss, and the discipline to update by hand; it covers menus up to 40 dishes with someone dedicating four hours a week.

What each investment tier includes for real costing (as of September 2026)?

From 900 to 2,600 USD a year you get costing software wired to inventory, with purchase-price history and per-ingredient variance alerts.

And from 3,000 to 9,000 USD a year comes the layer with automatic invoice reading, which refreshes 340 ingredients and recalculates 62 recipe costings in fourteen seconds. These ranges are from September 2026 and they expire fast: the U.S. services producer price index closed 2025 at +3.2% (U.S. BLS, PPI 2025 M12). Five levers move your unit cost, and knowing the weight of each one matters before you touch a single price. Real yield loss against the textbook 15% explains 4 to 9 points of food cost on protein-driven menus, and it's the biggest lever almost nobody measures. Switching suppliers moves another 3 to 7 points, because cleaning yield changes even when the price per kilo drops. Commodity swings can be brutal: arabica jumped 70% through 2024 (Bellwether Coffee), which wrecked the margin of any coffee menu costed back in 2023.

Five factors that move your plate cost, with their weight

Unaudited portion size quietly takes 2 to 5 points through kitchen drift. And update cadence —monthly versus daily— is the difference between correcting in time and finding out from a P&L two months later. Weighing the cleaning loss on your own cuts for two weeks changes your costing more than any license purchase will. The procedure fits in one paragraph: weigh the gross that arrives, weigh the clean product left, weigh bone and trim separately, note the supplier and the date, and do that with the eight ingredients that make up 70% of your purchasing. The gap that shows up between the textbook 15% and your measured loss ran, across the protein menus I've reviewed, from 4 to 9 points of real food cost. A recipe costing that doesn't weigh its own loss isn't costing at all, it's an estimate wearing the format of a table and the look of rigor.

Weigh your own trim loss: fourteen days with a scale beat any software

Two weeks cost you a 60 USD scale and forty minutes a day from one cook; the return lands in the first month of a recosted menu. Negotiate on yield and on calendar, never on list price, because price per kilo lies whenever yield loss shifts. With weighing data in hand you can sit down and say something concrete: your loin yields me 73% clean and the other one yields 78%, so your kilo at 42,000 pesos costs me more than the competitor's at 44,000. No serious supplier can argue with that. Then ask for three measurable things: fixed pricing per quarter on your four highest-turnover ingredients, written notice of any increase fifteen days ahead, and an automatic credit note whenever yield falls below what you agreed. The macro data gives you footing to demand it: U.S. final-demand PPI closed 2025 at +3.0% after +3.5% in 2024 (U.S.

How to negotiate with your supplier once you hold the number?

BLS), meaning input inflation is decelerating and a general rise has stopped being a universal excuse. Artificial intelligence changed the dirty work of plate costing, never its arithmetic.

A model that reads the supplier invoice, refreshes the unit cost of 340 ingredients and recalculates 62 recipe costings in fourteen seconds turns a six-hour monthly chore into a daily routine, and that cadence gap is real: a costing updated yesterday and one from March are not the same instrument even with identical columns. But when the costing is badly built —payroll inside it, theoretical loss, yields copied from a book instead of measured in your kitchen— AI just hands you the wrong answer faster and with better charts. Automate price capture, variance alerts and recalculation. Do not automate the definition of what belongs in the plate, nor the pricing decision; you sign that one. Start tomorrow: weigh a single ingredient and compare it against your costing.

Four differences that move the result

ALLOCATION. Pushing payroll and rent into the dish inflates unit cost by 18 to 24 points and produces menu prices the market refuses to pay. Those costs do not depend on selling one more plate; they depend on the month. They belong at the break-even point, covered by the sum of every contribution margin you booked. WEIGHED YIELD. The gap between a textbook 15% and two weeks of weighing your own trimming ran 4 to 9 points of real food cost on protein menus. Plate costing that never weighs its own loss is not costing: it is an estimate wearing a table for a costume. CADENCE. A cost sheet refreshed yesterday and one refreshed in March are not the same instrument, even with identical columns. AI adds no new intelligence here; it adds CADENCE, which is exactly what the spreadsheet lacked. DECISION UNIT. Decide by percentage and you end up with a menu of profitable salads and a tight till; decide by contribution margin currency and you build a menu that pays the rent. Percentage patrols, currency decides.