Food cost 2026: the traditional numbers against the Masterestaurant numbers

The 30.4% that hid a negative cash position

A bookkeeping food cost of 30.4% can coexist with four straight months of negative cash, and that is exactly what happened to a 62-seat restaurant in Bogotá that closed 2025 with that figure on its report and nothing left at month end. The owner did what everyone does: renegotiated with two suppliers, tightened protein portions, swapped the loin cut for a cheaper one. The balance did not move a single point. The reason is arithmetic and has nothing to do with the kitchen: that 30.4% averaged dining-room tickets with app tickets, and the app tickets reached the bank with almost a third less money after the marketplace commission. While the numerator of the formula stayed the real recipe cost, the denominator was inflated with pesos that never landed in the account.

Commission is not overhead, it is a discount on your selling price

Every marketplace commission reduces the selling price of that specific dish rather than the general expense of the business, and filing it under the wrong line is what produces false food costs. That same dish shipped through an app charging 27% leaves 29,200 pesos in the till, and those 12,000 of recipe now weigh 41.1%. Eleven points of difference on the same dish, same recipe, same cook. Traditional accounting books the commission as a selling expense and does its job well, because its job is filing taxes. But an income statement was never designed to decide what goes on Rappi and what stays dining-room only, and using it that way is asking a hammer to turn a screw.

Why do two restaurants with the same food cost live opposite realities?

Because the average hides the spread, and the spread is the business. Apply the channel effect and the first one's effective food cost sits near 31%, while the second jumps into the 38% to 41% range.

One pays payroll and rent comfortably; the other is funding its own bankruptcy with volume. Their accounting reports say the same thing. The SINGLE figure is what lies, not the accounting. And the practical consequence is uncomfortable: if you track consolidated food cost and make menu decisions with that number, you are deciding blind precisely in the channel bringing you the most volume.

The category benchmark still rules inside the dining room

Inside the dining room the classic benchmark by dish category keeps every bit of its validity and no owner should discard it while chasing the channel effect. Red proteins move in the 32% to 38% band, poultry between 26% and 31%, pasta and rice between 18% and 24%, and prepared drinks rarely pass 22%. A healthy local restaurant averages between 28% and 32% on gross dining-room sales. Coffee deserves separate attention: arabica prices rose 70% during 2024 according to Bellwether Coffee, and the US tariff on Brazilian imports reached a combined 50% in 2025, so a café menu costed with 2023 recipes is lying to its owner today. The concrete decision is to recost every hot beverage quarterly, not annually.

Five years that ate the margin: +35% in inputs and +35% in payroll

According to the National Restaurant Association, the average US restaurant saw food costs climb 35% and labor costs another 35% over the last five years, and that double blow explains why old recipes no longer describe the operation. When an input rises by a third while the menu adjusts with a couple of timid increases, the gap opens on its own without anyone making a wrong call. Here frequency matters as much as formula: a food cost calculated once a year is a portrait of a restaurant that no longer exists. The sector carries enough weight to deserve that rigor, with bars and restaurants at 3.6% of Brazilian GDP according to ABRASEL and 15.3% of Mexican tourism GDP according to SECTUR and CANIRAC.

How to read these numbers in YOUR operation?

Three scenarios, three different decisions.

If you run a small venue under 40 seats with app sales below 25%, calculate food cost by channel once a month and cap the digital catalog to dishes whose recipe stays under 26%, because that margin survives a 27% commission without going into loss. In a mid-size venue of 40 to 90 seats with significant digital sales, measurement becomes fortnightly and it pays to raise app menu prices above the dining room, already standard practice and accepted by customers. For a group of three or more locations, the cut is weekly and by point of sale: dispersion BETWEEN branches usually beats dispersion between dishes, and that is the leak nobody watches. Diego F. Parra keeps insisting at Masterestaurant that a group's consolidated food cost is the least useful number on the whole dashboard.

Where these benchmarks come from and how far they reach?

The category ranges come from Masterestaurant's field work with Latin American and Iberian operations, and the macro figures cited here come from verifiable public sources:

National Restaurant Association for cost evolution, Bellwether Coffee for the arabica market, ABRASEL and SECTUR for sector weight. It is worth stating honestly where they break. Second, none of these percentages accounts for waste, which in kitchens without portion control adds 2 to 5 extra points. Third, a benchmark describes a population, never your restaurant: it tells you whether your number sits outside the range, not why.

The decision that changes next month's balance

Open last month's sales report, split tickets by channel and recalculate food cost twice: once against gross dining-room sales and once against net app sales after commission. If the difference between both figures exceeds six points, you have a digital menu that needs another price or fewer dishes. The Bogotá case was solved that way, without changing a single supplier: they pulled eleven low-margin dishes from the app catalog, raised the survivors 15%, and operating margin turned positive in the second month. No cook worked differently. What changed was the number used to decide. And that is the hard part to accept, because it means admitting the problem never sat in the kitchen but in the spreadsheet used to judge the kitchen.

Where the blended average breaks?

A marketplace commission is not an administrative expense, it is a direct cut to that dish's selling price.

Traditional bookkeeping is not wrong in its own lane, it files taxes properly, but it cannot decide what you publish on a delivery app. Averages hide dispersion. The second is bleeding and its report reads identical to the first. The SINGLE figure misleads, not the accounting itself. Frequency matters as much as the formula. Food cost you learn twelve days after closing is history rather than management: by the time the number lands, you already bought a full month at the wrong price. Counting twelve items every Monday takes forty minutes and hands back the variance while you can still fix it. Percentage is the compass, currency margin is the map.

Where the blended average breaks — in practice?

A dish at 24% food cost yielding 9.000 in contribution pays you less than one at 33% yielding 21.000, and in a kitchen capped by tickets per hour that gap decides the month.

Classic menu engineering has known this since the eighties, yet most digital menus still get built by habit. Channel changes the recipe, not just the price. A dish designed to travel twenty minutes inside a box tolerates a different protein, a different side and a different plating than the same dish served at a table, and whoever skips redesigning the spec sheet ends up paying commission to transport a product that also arrives worse. Prime cost is the ceiling almost nobody watches. Food cost plus labor above 65% of net sales means rent, utilities and maintenance no longer fit, and no purchasing win repairs that while the channel mix stays put.