Profitable menu: criteria to build it — traditional method vs Masterestaurant method: service and team

What does each method actually measure when you build the menu?

The traditional method measures one variable, plate cost, while the Masterestaurant method measures two, contribution margin and how many times the floor puts that plate on a table.

That arithmetic difference decides the month. Take the 62-item menu at the chef-driven restaurant in Medellín that opens this piece: 71% of sales concentrated in nine dishes, and four of those nine carried the LOWEST contribution margin in the whole document. Every plate costed with judgment, food cost under the 32% ceiling, and the month still closing on fumes. By traditional criteria that menu was well built. By Masterestaurant criteria it was badly built from day one, because nobody asked who would defend each dish on the floor. The second one wins, and it wins wide when median pretax profit in full service runs at 2,8% of sales, per the National Restaurant Association Restaurant Operations Report 2024/25: on that margin, four badly pushed dishes eat the year.

Suggestions per table versus food cost points: which one moves more cash

Going from 1,3 to 1,7 effective suggestions per table weighs more on monthly margin than shaving two food cost points in the kitchen, and almost nobody runs that comparison. Do the math on a 40-item menu with a 90.000-peso average check and 1.800 tables a month. Two fewer food cost points on the food portion of the check give back roughly 1,6% of revenue. Four extra tenths of effective suggestion, each landing on a high-margin dish, add a new line of incremental sales that carried no fixed cost with it. The traditional method fights on the first terrain because that is what the spreadsheet knows how to measure; the Masterestaurant method fights on the second because that is where money arrives without negotiating with suppliers. Verdict: Masterestaurant wins, with one non-negotiable condition — without documented floor training, those 1,7 suggestions are a wish, not a number.

The 32% food cost ceiling: a maximum in one method, a target in the other

Masterestaurant sets 32% food cost per plate as a tolerable MAXIMUM, never as a goal, and the traditional method reads it backwards, as a target you climb toward from below. That reversal destroys turnover. If a dish costs 24% and sells eighty units a week, the traditional manual suggests raising price until it approaches the theoretical 32%; the usual outcome is punished elastic demand and eighty units turned into fifty-four. I got this wrong for years, pushing price on anchor dishes because the percentage gave me permission. The other error in that same family is loading payroll, rent and utilities onto the plate: they do not belong there, they belong to the venue's break-even, and dropping them into unit costing inflates prices with no accounting basis. Verdict: Masterestaurant wins on direction — 32% is a roof you come down from, not a floor you climb to.

Popularity: a fact about the guest, or a result of training?

Classic menu engineering from Kasavana and Smith (1982) sorts dishes into four quadrants crossing popularity with margin, and it assumes popularity arrives given by the guest.

It does not arrive given. A good part of it is manufactured by the floor, which is why Diego F. Parra keeps insisting in Masterestaurant audits that a menu gets built while looking at who will defend it in the dining room. Picture what happens if you rotate 40% of your servers tomorrow without transferring the suggestion script: your matrix quadrants rearrange themselves within two weeks, your stars slide down into plowhorses, and you conclude, reading the report, that guest taste shifted. It did not shift. What changed is who speaks at the table. The paradox resolves like this: the matrix remains the right tool, but you read it as a snapshot of current training, not as stable diner preference.

The Medellín mini-case, with the closing numbers

Nine dishes carried 71% of sales on that 62-item menu, and four of them sat in the low band of contribution margin across the entire document. We reordered without touching recipes or prices: 19 dead items came out, the menu dropped to 43, and a two-line script got written for each of the six dishes with above-average margin. One single instruction went to the floor, suggest that block before anything else. By week six, effective suggestions measured per ticket had moved from 1,2 to 1,6, the menu's weighted contribution margin climbed close to six points, and the month stopped closing on fumes. None of that came out of the kitchen. It came from deciding which dish earns a slot by what it leaves AND by what the team can place, which is exactly where the two methods split.

How many items a menu carries before it gets in the way?

Between 7 and 15 items per category is the range menu design research points to for avoiding decision paralysis, and the average guest spends 109 seconds reading the menu, according to NeatMenu (Menu Psychology 2026).

With those two facts on the table, a 62-item menu is a document nobody finishes: the diner scans, gives up, and orders the usual or whatever the server names last. The traditional method trims by cost, cutting what is expensive to produce and keeping what is cheap even when it leaves little behind. The Masterestaurant method trims by weighted margin and by suggestibility, and that filter tends to produce shorter and more profitable menus at once. Verdict: Masterestaurant wins, though I concede that in high-volume rooms with small kitchens production cost rules, because there the constraint is the griddle, not the spreadsheet.

Off-premises: the channel that breaks the traditional comparison

Off-premises traffic reached 30% in full service during 2024, against 19% in 2019, and 83% in limited service, per the National Restaurant Association Off-Premises Report 2024. That share travels without a server, and there the traditional method regains ground: in delivery nobody suggests anything, so plate margin rules alone. A well-built menu today therefore splits into two lists with different criteria. The dining room list gets sorted by margin crossed with suggestibility. The off-premises list gets sorted by margin, transport resilience and dispatch time, and it gets pruned without mercy. Mixing the two is the error that repeats most when a restaurant opens a digital channel: it loads the same 43 dishes into the app, packaging ruins four of them, and the guest rates the whole operation by its worst performers.

What to pick for your profile, no hedging?

If your menu runs past 30 items and your dining room works with three or more servers per shift, run the full Masterestaurant method and stop debating it:

the deciding variable there is suggestion, and without a floor script every high-margin dish sits orphaned. If you operate a bar, a dark kitchen or a counter format where nobody recommends anything, traditional per-plate costing is enough for you, and then some, provided you respect 32% as a ceiling and keep payroll, rent and utilities out of unit costing. The middle case, a 20-to-30-item restaurant with two servers, goes hybrid: classic margin matrix plus a suggestion script for the three highest-contribution dishes. Start this week with the cheap part — measure effective suggestions per ticket for seven days and set that number against your menu's weighted margin.

Where the two methods really split?

The traditional method asks what the dish costs. The Masterestaurant method asks what the dish earns AND how many times the floor will place it on a table.

Those are two different questions, and the second one moves more cash: on a 40-item menu, going from 1.3 to 1.7 effective suggestions per table weighs more on monthly margin than shaving two points of food cost in the kitchen. The 32% food cost cap per dish, which Masterestaurant sets as a MAXIMUM and not as a recommendation, gets used backwards in the traditional method: as a target to climb toward from below, which pushes prices up on elastic-demand dishes and sinks turnover. Payroll, rent and utilities never load onto the plate, they belong to the break-even of the venue, and confusing the two inflates prices for no reason. Standard recipes are where the traditional method quietly breaks. Without locked gram weights, Tuesday's portion costing is not Saturday's, and a profitable menu built on costs that swing 8% or 12% depending on who is on the line is not profitable, it is an estimate with good typography.

Where the two methods really split — in practice?

Restaurant menu design in the traditional method aims for a beautiful page; in the Masterestaurant method it aims for the eye to land where the margin sits.

On single-page menus the gaze enters through the upper right third, and that real estate does not belong to the owner's sentimental dishes. On printed menu versus QR, I was wrong for years recommending the printing savings: QR-only strips the restaurant of control over service pace, menu narrative and suggestive selling. The PRINTED menu is experience and hospitality; the QR menu is an operational complement for delivery, accessibility, price updates and analytics. Both, each in its own role. Dishes that hurt profitability are almost never the ones the owner suspects. They tend to be the mid-margin, high-popularity ones, because they occupy the server's mouth and the line's time without earning what the neighboring dish would, and they only surface once margin gets weighted by units sold.