Menu engineering in restaurants: myth vs reality with the numbers on the table

Contribution in cash beats cost percentage every time

A dish running at 22% cost that yields 6 USD of margin and sells 40 units a month brings 240 USD to the register, while another at 34% yielding 11 USD across 90 units brings 990 USD: the second one has the worse ratio and quadruples the contribution. That misreading is the costliest one on any menu, because payroll is not paid with ratios but with the sum of (price minus portion cost) times units sold. The arithmetic turns dramatic once you recall that median pre-tax profit in full service was 2,8% of sales in 2024 and 4,0% in limited service, per the National Restaurant Association Restaurant Operations Report 2024/25. On that cushion, shifting 300 USD of monthly margin between two dishes equals nearly eleven thousand dollars of extra sales. Decide in currency. The sales mix reveals which dishes hold up the business and which merely occupy refrigerator space, something a per-portion costing sheet never shows because it looks at each recipe in isolation.

What does the sales mix reveal that the costing sheet hides?

At the Bogotá steakhouse that opens this piece, billing 92.000 USD monthly with an average food cost of 31,4% —inside the 32% ceiling we set as the maximum—, 18% of the menu generated 61% of the contribution margin.

Eleven dishes failed to reach 1,2% of units sold and still demanded three exclusive suppliers, two mise en place stations and a quarter of the service refrigerator. They cost 4.100 USD a month in frozen inventory, waste and prep hours nobody ever billed. The accounting audit cleared them one by one; the cross with units sold condemned them in an afternoon. Cross both columns before touching a single price. Treat the menu as a portfolio where every dish plays a distinct financial role, because handling it as a list is what produces sixty-item menus carrying twenty orphans. There are anchor dishes that build the bulk of the margin, image dishes that justify the check and hold up price perception, and convenience dishes that settle the undecided guest without wrecking the kitchen.

A menu is a portfolio with roles, not a list of options

The Cornell finding (Food and Brand Lab, Wansink) is blunt: dishes with descriptive copy sell 27% more than the very same dishes without a description. Applied to an anchor moving 90 units monthly at 11 USD of margin, that rewrite is worth 267 USD a month with no change to recipe, supplier or price. Rewriting four descriptions costs one afternoon. Switching four suppliers costs a quarter. Start with the cheap lever. Translate the benchmarks to your own scale before applying them, because the same percentage means different things across three sizes of business. In a SMALL restaurant, say 25.000 USD a month, 61% of the margin usually concentrates in six or eight dishes and pulling five orphans frees between 900 and 1.400 USD of trapped capital: you feel it in next week's register. In a MEDIUM one like the 92.000 USD steakhouse, the figure was 4.100 USD monthly and the real impact arrived through freed mise en place, not through inventory.

How to read these numbers in YOUR operation?

In a GROUP of several locations, run the analysis per location and consolidate afterwards, never the other way round, because an orphan dish at the north store can be the anchor at the south store and the group average buries both truths.

Measure location by location. The profitability figures in this piece come from the National Restaurant Association Restaurant Operations Report 2024/25, which reports median pre-tax profit —2,8% in full service, 4,0% in limited service— across United States operators. The +27% description figure comes from Cornell's Food and Brand Lab. The limits are worth stating plainly: these are United States medians, built on rent, payroll and tax structures that do not map onto Latin America, and a median hides enormous dispersion between the top and bottom quartile. They work as a yardstick to know whether you sit far from or near the normal range, not as a target.

Where these benchmarks come from and how far they stretch?

Your actual food cost, your mix and your margin per dish come out of YOUR point of sale system and YOUR standard recipe. Nobody is going to audit that for you.

The dish yielding 11 USD in the dining room may yield 4 USD through delivery, and that gap stopped being marginal once off-premises consumption stopped being the exception. In full service it went from 19% of traffic in 2019 to 30% in 2024, and in limited service from 76% to 83%, according to the National Restaurant Association Off-Premises Report 2024. With nearly a third of orders walking out the door, keeping a single menu engineering matrix for both channels is bookkeeping from another decade. An anchor in the dining room can be a disaster in packaging, arrive cold and hand its margin to the platform commission. Build two matrices, one per channel, with the same cross of portion cost, units and contribution in currency.

Channels rewrote margin per dish and almost nobody recalculated it

You will see the ranking shift, and sometimes it shifts entirely. What would happen if you raised your fastest-moving dish 12% to fix the food cost? Take an anchor at 90 monthly units priced at 32 USD, with 21 USD of cost and 11 of margin: price climbs to 35,8 USD, unit margin reaches 14,8 and the ratio improves from 65,6% to 58,7%. But if elasticity costs you 25 units —ordinary for a high-rotation dish with a substitute on the same menu—, you are left with 65 units at 14,8 USD, meaning 962 USD against the previous 990. Better ratio, less cash, and an idler kitchen on top. The tension is real: percentage works for negotiating with suppliers and comparing recipes against each other, and it does not work for setting a selling price. Total contribution is the bridge. Raise the price of image dishes, which absorb elasticity, and protect the anchor.

The quarterly protocol Masterestaurant leaves installed

Install a four-step quarterly cycle and stop improvising the menu every time an input goes up. First, export units sold per dish for the quarter from the point of sale. Second, cross each line with its standard-recipe portion cost and work out the contribution in currency, keeping per-dish food cost never above 32%, which is the CEILING and not the target. Third, sort by total contribution from high to low and flag everything below 1,5% of units: that is where the orphans live. Fourth, decide dish by dish between rewriting the description, repositioning it on the menu, raising the price or pulling it. At Masterestaurant, Diego F. Parra runs this cross with the kitchen and the cash teams in the same room, because the chef knows the waste the P&L never shows. Block the first ninety-minute session this week. The myth measures in percentage, reality measures in currency.

The three differences that decide the outcome

Percentage is a useful ratio for comparing recipes and for negotiating with suppliers, but it does not pay payroll: payroll gets paid with the period's accumulated contribution margin, which is the sum of (price − portion cost) × units for every dish. When an owner chases the ratio and raises the price of the fastest-moving dish, he usually loses more units than he gains in unit margin, and closes the month with a better ratio and less cash. The myth treats the menu as a list, reality treats it as a portfolio with roles. A well-built portfolio has anchor dishes doing the volume of the margin, image dishes that justify the ticket and hold price perception, convenience dishes that solve the difficult table, and bridge dishes that push beverage or starter consumption. A dish can carry high food cost and still earn its place because it drags two side orders along; cutting it on its individual ratio dismantles the cross-sell nobody was measuring.

The three differences that decide the outcome — in practice

The myth reviews once a year, reality reads the mix every month. Input costs move with seasonality, exchange rates and supplier contracts, while guest behavior moves with weather, calendar and campaigns. A standard recipe recosted in January and left untouched until December may have spent six months selling below the point where it contributes margin, and nobody notices because the monthly average papers over the hole.