Plate Costing: What the Data Says Against What the Sector Keeps Repeating

How much does food cost really weigh on a restaurant's cash register?

Median food cost sits at 32.0% of sales in full service and 32.4% in limited service, according to the National Restaurant Association with 2024 close, and that figure is a CEILING, not a target.

An operator who costs dish by dish almost always finds that the median hides brutal dispersion: six items above 45% that turn every single day, propped up by two appetizers at 18% nobody orders. Average arithmetic is reassuring and that is exactly why it deceives. When we at Masterestaurant open a location's sales matrix and sort items by descending turnover, that aggregate 32% breaks into two different businesses living under one roof, and one of them is financing the other with margin the owner thought he had. That is the first decision this number triggers: stop reading the menu as a block. MSMEs account for roughly 60% of formal employment in Latin America and contribute only around 25% of regional GDP, a gap ECLAC has documented for years and one that in food service has a very concrete translator: the absence of elementary management instruments.

The missing productivity: why Latin America's food MSME employs a lot and produces little

Add labor informality of 47.6% per the ILO's Labour Overview 2025 and the picture closes by itself. A restaurant that does not know the unit cost of its flagship dish cannot project formal payroll, cannot negotiate terms with a supplier and cannot back a credit application with anything other than its word. SATE Institute produced this analysis precisely because the micro-operation moves macro indicators. Four chained effects start with a recipe card nobody filled out, and the first one is called productivity. Two external costs distort any costing done the old way. First: card commissions average 2.35% per transaction according to the Texas Restaurant Association in 2025, and at country scale swipe fees add up to nearly $187 billion a year in the United States, a National Restaurant Association figure. The second bites far harder: DoorDash charges between 15% and 30% per order, with a 30% standard marketplace rate, and Uber Eats moves in the same range, while Grubhub runs between 15% and 25%, all measured by Rezku in its 2026 fee review.

The commissions that eat your margin before you ever see it

Run the math on a dish costed at 32%. If that same dish goes out through a platform at 30% commission, effective cost climbs to 62% before touching payroll or rent. The decision this block triggers is not negotiating the commission; it is costing a separate menu for the digital channel. Food waste costs the U.S. restaurant industry close to $162 billion a year, per The Restaurant HQ's 2025 count, and in Latin America food loss equivalent to 11.6% of production turns shrinkage into a solvency matter rather than a conscience one. Let us be blunt here: shrinkage is not a separate cost booked at the end, it is part of the plate cost you already paid for and never sold. If your recipe card records 180 grams of tenderloin and the real portion leaves the pass at 210 because the cook cuts by eye, your true food cost runs 17% above the calculated one and no spreadsheet will ever tell you.

Waste is a costing figure, not an environmental problem

Input traceability, moreover, is the condition without which no loss-reduction program under SDG target 12.3 can even begin. The full-service segment in the United States is roughly 18% smaller than in 2019, according to Technomic with 2024 data, and that contraction was not evenly distributed: it swept away operators who entered the pandemic without unit costing and came out adjusting prices by instinct. Put it as a counterfactual. Picture two identical locations in 2019, same menu and same traffic. The first raises prices a flat 12% in 2021 because input inflation hit; the second raises 22% on the eight highest-shrinkage items and freezes the rest. Three years later the first lost its low-ticket customers without recovering margin on the dishes that were bleeding, and the second kept traffic with a profitable mix. Diego F. Parra, restaurant consultant at Masterestaurant with twenty years of field work across 43 countries, puts it without decoration: most restaurants that close did not close for lack of sales.

Opening costs what it costs, and costing decides whether you get it back

Opening a small takeout restaurant in the United States demands between $75,000 and $150,000, while median opening cost sits at $450 per square foot with a range of $100 to $800 according to Square in 2024, and the 2025 bottom quartile closed at $175,500 —$59 per square foot— per Rezku. Those figures matter for a reason almost nobody states out loud: initial investment comes back through MARGIN per dish multiplied by turnover, not through total sales. A location billing well with a real food cost of 41% takes twice as long to return capital as one billing half that at 28%. The operator who only watches the daily cash count cannot tell those two scenarios apart until the flow dries up, and by then the conversation with the bank has already started badly. When costing lives in the owner's head, the business is illegible to any outsider, and that illegibility carries a market price.

Costing so that a third party can read it

A commercial bank with an MSME portfolio, an IDB Lab program or an impact fund cannot rate risk on a hunch; they need a data series. Masterestaurant S.A.S., exclusive technology partner of the SATE model, turns recipe cards, shrinkage, yields and food cost variance into auditable records, and that is where the jump happens: the establishment stops being an applicant with no history and becomes a subject with structured information, which is exactly what enables alternative scoring under SDG 9. There is a real tension here, and it deserves acknowledgment: filling recipe cards consumes hours from a team already stretched thin. Sequence resolves it —first the twenty items that concentrate 80% of sales, then the rest— not the promise of digitizing everything in one month. 32.0%. That is the full-service median food cost per the National Restaurant Association for 2024, and your action is concrete: measure food cost for each item separately this week and flag in red every one above 35%, no matter how much you love the dish.

The 3 numbers you should tattoo on yourself

30%. That is the standard marketplace commission from DoorDash and Uber Eats per Rezku 2026, and the action is to build your own digital menu with portions and prices calculated to absorb that bite, instead of publishing the same dining-room menu. 47.6%. That is regional labor informality reported by the ILO in its Labour Overview 2025, and the action is to use the margin recovered from costing to formalize ONE position, the person who controls portions and receives deliveries. Start with the twenty items that give you 80% of sales. The rest can wait. The core difference is not the formula, which fits on one line, but the unit of analysis: the myth costs the MENU as a block while reality costs the DISH as an independent economic unit, with its own yield, trim loss and rotation.

The differences that change the decision

A menu averaging 31% food cost can hide six dishes running at 45% that sell every single day, subsidised by two dishes at 18% that almost nobody orders, and that reassuring arithmetic is exactly what leads an operator to tell the accountant the business is healthy while the till empties out. A second difference cuts deeper, and it concerns who gets to read the numbers. When costing lives inside the owner's head, the business is illegible to any third party: a commercial bank with an MSME portfolio, an IDB Lab programme, a development agency weighing equipment co-financing. When costing lives in versioned technical sheets, with supplier, date and yield, that same business produces evidence. Diego F. Parra puts it bluntly in his work with Masterestaurant: a restaurant without technical sheets is not an informal restaurant, it is an INVISIBLE one, and nobody finances what they cannot see. The third difference is temporal, and hardly anyone argues about it.

The differences that change the decision — in practice

Costing is an act; recosting is a system. An operator who sets food cost in January and reviews it in December is not costing badly: they are costing once and assuming eleven times. With protein and oil prices moving several points inside a single quarter, that assumption costs real, measurable margin, which shows up as an unexplained cash shortfall sometime in August. One more difference separates the consultant from the accountant. The accountant wants to know what it cost; the operator needs to know which decision that cost triggers. A dish at 38% does not get cut automatically: if it drives Tuesday traffic and pulls two beverages per table, its contribution margin may beat a 24% dish that only sells on weekends. Menu engineering exists precisely to settle that tension, and plate costing is its mandatory input.