Restaurant business model: the numbers that decide whether yours holds

Prime cost is the first number that validates a model, and the ceiling is 65%

Add food and beverage cost to fully loaded payroll: if that sum exceeds 65% of net sales, the business model no longer closes, however much you optimize everything else. In the Bogotá case that opens this document, the operation billed 47,000 USD a month with a declared food cost of 29% and payroll at 41%, a prime cost of 70%, and with the National Restaurant Association reporting a median sector operating margin between 3% and 5% before taxes in its State of the Industry 2025, those five excess points swallowed the entire year before rent was even paid. The decision that follows from this number is simple and hard: above 68% you freeze every hire and redesign the menu that same week, not next quarter. Thirty-two percent per dish is the MAXIMUM you can tolerate, never the goal, and that distinction decides whether a model survives a year of ingredient inflation.

Why is 32% food cost a ceiling and not a target?

The usual confusion is reading 32% as a menu average when it must be read dish by dish, because a menu with three stars at 38% and fifteen fillers at 24% averages nicely while bleeding on whatever sells most.

The National Restaurant Association projects real growth of just 1.3% for the U.S. sector in 2026 once inflation is stripped out, which means nobody will offset bad costing with new volume. Operating rule I apply: any dish above 32% enters portion or supplier redesign within fourteen days, and if it does not come down, it leaves the menu. Translate your break-even into a calendar date: if the operation does not cover its fixed structure before day 20, the model depends on no month ever bringing a surprise. The restaurant in our case crossed it on day 29, with a single night left to generate profit, and that fragility explains better than any average the figure the U.S.

Break-even is measured in days of the month, not pesos of the year

Bureau of Labor Statistics has sustained for more than a decade: around 60% of food service establishments close before their fifth year. They do not close for lack of customers; they close because their break-even day falls so late that one badly placed holiday or a rainy week takes the month. Calculate yours today by dividing fixed costs by daily contribution margin and mark it on the kitchen calendar. A business model where one channel contributes more than 30% of sales is not diversified, it is rented. In the case we analyzed, two delivery apps concentrated 61% of revenue at 27% commission, meaning that for every 100 USD billed through that route only 73 came in to cover a prime cost calculated on 100. Sensor Tower measured the fragmentation of that Latin American market in 2025: iFood leads with 40% of active users and reaches 89% in Brazil, while in Mexico DiDi Food hits 38% and Rappi 36%.

No single channel should exceed 30% of your revenue

That concentration is precisely the lever an aggregator uses to raise its commission without negotiating. Set the cap at 30% per channel and move the rest to dining room, direct pickup and corporate catering. The same benchmarks apply differently depending on size, and anyone who copies them without adjusting makes bad decisions with correct data. In a small venue of 12 to 30 covers, payroll rarely drops below 30% because the owner cannot dilute fixed positions, so food cost has to be worked toward 28% to respect the 65% prime cost ceiling. In a mid-sized venue of 60 to 120 covers, efficient payroll lives between 28% and 33%, and there the margin is won through menu engineering, not layoffs. In a group of three or more units, a central structure cost of 4% to 7% of consolidated sales appears that prime cost never captures and must be subtracted before celebrating: a group at 63% prime cost with 6% overhead operates like one at 69%.

Where these benchmarks come from and what they do NOT prove?

Methodological honesty before you make a capital decision with these figures. The market data cited here comes from public, verifiable sources —National Restaurant Association 2025 and 2026, the U.S.

Bureau of Labor Statistics, Sensor Tower 2025, CANIRAC with its projected growth near 6% for Mexico in 2025 and Abrasel with real growth of 0.92% in Brazil— and each measures its own universe with its own methodology, so they are not directly comparable across countries. The operating thresholds of 65% prime cost, 32% food cost and a day-20 break-even are management criteria that Diego F. Parra applies at Masterestaurant, not statistical averages from a sample. They let you decide fast; they do not replace the accounting close of your own operation. The structural difference between a model that gets corrected and one that gets discovered broken lies not in how many indicators you watch, but in when you watch them.

Measuring on Monday or waiting for the accountant: the gap is 33,840 USD

Closing Sunday's cut on Monday morning gives four chances to react each month; waiting for the accountant's income statement gives one, and it arrives forty days after a decision has already been made. There is a real tension here and I resolve it rather than dodge it: for a twelve-table business, measuring prime cost weekly sounds like bureaucracy, and partly it is. But the cost of not measuring is not the paperwork you save, it is the decision taken late. A venue billing 47,000 USD monthly at 71% instead of 65% loses 2,820 USD a month, 33,840 USD a year, more than its starting working capital. Picture that same Bogotá operation with the four figures reviewed before signing: rent adjusted so break-even fell on day 18, a delivery contract limited to one aggregator capped at 25% of sales, a 22-dish menu instead of 41 and payroll sized at 32%.

What happens if the model is validated before the lease is signed?

With that design, prime cost lands at 61% and the same 47,000 USD in sales leave roughly 4,200 USD of monthly operating profit, inside the 3% to 5% range the National Restaurant Association reports as the sector median, with room to negotiate.

The business changed neither concept nor chef; it changed structure. This week calculate your prime cost from Sunday's cut and write down the exact date of your break-even: with those two numbers on the table, you already know whether you have a model or a hunch. The structural difference is not how many indicators you watch, it is WHEN. Closing prime cost every week, with Sunday's cut settled by Monday morning, gives four chances to correct each month; measuring it when the accountant delivers gives a single reaction window that arrives forty days late. A restaurant billing 47,000 USD a month at 71% prime cost instead of 65% loses 2,820 USD every month, which compounds to 33,840 USD a year, more than the working capital it opened with.

The differences that move cash

There is a genuine tension here, and I will resolve it rather than dodge it: measuring everything weekly sounds like bureaucracy for a 12-table operation, and in some sense it is. But the cost of NOT measuring is not the paperwork you save, it is the decision you take too late. The mild version of that mistake —downgrading ingredient quality to fix a food cost that was fine, when the real leak sat in bar payroll— costs customers, and customers are the one asset you cannot repurchase. The traditional method loads payroll, rent and utilities onto plate cost because it feels prudent. It is not. A plate costing 4 USD in ingredients and selling at 14 runs a 28.6% food cost and contributes 10 USD of margin; add a prorated slice of rent and that same plate suddenly 'costs' 7 USD, so the owner concludes it must go up in price or come off the menu.

The differences that move cash — in practice

Profitable plates get cut, sales drop, and rent still costs the same spread across fewer dishes. That loop has closed more restaurants than any supplier crisis. In channel economics the arithmetic is merciless and hardly anyone runs it. A 20 USD order through a platform charging 27% leaves 14.60 USD; at a 30% food cost on menu price —6 USD of ingredients— real gross margin falls to 8.60 USD against 14 for the identical plate sold in the dining room. The channel is not bad. The channel is different, and a business model treating it as equivalent revenue rests on a false figure. On printed menus versus QR my position is firm, and I know it runs against fashion: you keep BOTH. QR handles price updates, delivery, accessibility and analytics on what guests browse without ordering. The printed menu controls service pace, sustains menu narrative and enables the server's suggestive sell, which is the cheapest average-ticket lever in existence.

The differences that move cash — key points

Dropping the printed menu to save on printing trades a two-figure monthly cost for a ticket decline nobody ever traces back to its actual cause.