Profitable menu: criteria to build it — what agencies charge, what it truly costs and the mistake you pay for monthly

What is the first criterion for building a profitable menu?

The first criterion is your SALES MIX over the last 90 days, not price and not design. Before moving a single item on the page, export unit sales per dish from your POS for that quarter and sort them highest to lowest:

at the 92-seat steakhouse we reviewed in November 2025, 11 dishes out of 74 carried 68% of revenue, and the remaining 63 lived off the cook's inventory rather than the guest's appetite. With food costs up 28,7% since 2020 according to the Bureau of Labor Statistics, carrying 63 dead references means paying waste, cooler space and prep time for variety nobody asked for. Average operating margin in full service runs 3% to 5% per the National Restaurant Association, and a decorative menu does not fit inside that number. As of September 2026, menu redesign work sits in three clear ranges and it pays to know what each one buys.

What each investment range buys when you redesign a menu?

Under 800 USD you are buying layout: someone drops your dishes into a template, picks a typeface and hands over a PDF; costing is on you, or on nobody.

Between 1.500 and 3.500 USD the menu engineering work begins — standardized recipe cards, portion costing for every dish, sales-mix analysis and a dish-by-dish call to raise, rework or retire. Above 6.000 USD, which is what that steakhouse paid for its brochure, you should be getting studio photography, premium printed stock and support across two or three measurement cycles. The owner's mistake was never the price tag: he bought the third range without ever having bought the second. Costing without a written, weighed, signed recipe card is not data — it is an estimate with decimals that buys you false confidence. When the card says 180 grams of protein and the cook on shift plates 220, that dish is not running at 29% food cost but at 35,4%, and you find out during month-end inventory, after selling 400 units of the mistake.

Portion costing is an opinion until a standard recipe exists

That was exactly the steakhouse's jump between quarters: 33,1% to 35,4% consolidated, with nobody having touched a single price. At Masterestaurant, Diego F. Parra hammers an order that sounds obvious and almost nobody honors: weigh first, cost second, price third, and only then lay out the page. Invert that sequence and every aesthetic choice becomes a financial hypothesis nobody will be able to defend with numbers. Four factors explain most of the price variation inside one menu, and the owner's intuition is not among them. Input cost carries the obvious weight: with food and labor 35% above 2019 levels according to the National Restaurant Association, an imported-protein dish tolerates less margin than a vegetable-based one. Demand elasticity matters more than owners assume and shifts tolerable pricing by 6 to 14 points. Placement and labeling are priced in too: tagging a dish as chef's favorite lifts orders by 13% to 20% according to NeatMenu's 2026 menu psychology analysis.

The factors that actually move a dish's price

And reputation moves the whole register — each additional review star is worth 5% to 9% of revenue, per Michael Luca at Harvard Business School. Four levers, four different decisions. Raising the whole menu 8% is what owners without data reach for, and it hurts at both ends. On an elastic dish — the pasta, the burger, whatever the guest compares against the place across the street — that 8% costs you units, and losing units on your highest-turnover item means losing the volume that pays fixed payroll. On the inelastic dish — the signature cut, the dessert nobody else makes — you gave margin away: it would have absorbed 14% without a single complaint, because it has no substitute in the guest's mind. The arithmetic is uncomfortable. If 68% of your revenue lives in 11 dishes, a differentiated increase on those 11 moves more EBITDA than a blanket increase across 74, and it annoys fewer people.

Why an across-the-board price increase destroys margin?

A menu is not raised; it is tuned, reference by reference. Negotiate the project as two separate contracts and pay the second one against a measured result.

The first covers data: recipe cards, portion costing and the 90-day mix analysis, delivered as a spreadsheet that stays in your hands rather than on the agency's server. The second covers design and printing, and gets signed only once the first one has said which dishes survive; that way you do not lay out 74 references to end up printing 40. Ask for final payment on the data contract to release at 60 days against an agreed indicator: consolidated food cost down at least 1,5 points, or average contribution margin up. And if you already paid for the pretty brochure, do not bin it. That steakhouse salvaged its 6.800 USD by reprinting only the inner pages with 38 costed dishes, keeping cover, stock and photography.

What happens if you leave the menu alone this quarter?

Suppose you leave the menu untouched and wait for the next cycle. Inputs keep climbing at the rate the Bureau of Labor Statistics records, so your 35,4% food cost drifts to 36 or 37 without you doing anything;

against an operating margin of 3% to 5%, those two points swallow half your profit. Then you react late, hike prices all at once in January, and the guest who was quietly absorbing gradual adjustments feels a jump and starts comparing. That is where you lose traffic exactly when you need volume most. Here is the paradox of the trade: the menu you touch least is the one that costs you most, because price goes stale on its own while cost moves every month. Review the mix every 90 days even if you change nothing. Measuring is cheap; correcting late is not. A dish leaves the menu when it fails two of three tests: turnover, contribution margin and shared inputs.

The retirement criterion: when a dish stops earning its slot

Selling under 5 units a week, contributing less margin than the median of its category, and demanding an ingredient no other dish uses — there is no defending that combination, because the item is costing you cooler space, waste and a purchase line. For years I kept dishes out of affection for the chef, and that affection gets paid for in idle inventory. The legitimate exception exists and deserves respect: a low-margin dish that brings in the party of four, or one that solves a serious dietary restriction — remember that eight food groups account for 90% of food allergies according to the FDA — earns its slot even when the number argues otherwise. Outside those two cases, pull it this week. Design works on the menu; data works on the sales mix. A 74-dish menu where 11 items carry 68% of revenue does not have a layout problem, it has dead inventory dressed up as variety, and no typeface fixes that.

The differences that change the month's result

Across-the-board price hikes are the tool of an operator without data. Raising everything 8% looks fair and does the opposite: it punishes the elastic dish, which sheds units, and gives away margin on the inelastic one, which would have absorbed 14% unnoticed. Cost per portion without a standard recipe is an opinion with decimals. If the cook on shift plates 220 grams where the spec says 180, your real food cost is not 29%: it is 35.4%, and you find out at month-end inventory, three weeks late and with no idea which dish ate it. The dishes that hurt profitability are rarely the suspects. At the Bogotá steakhouse the problem was never the 42 USD imported cut, which cleared 24.80 USD; it was a 14 USD salad with avocado, goat cheese and walnuts that cleared 3.90 USD and sold 61 times a week, blocking the table from a dish worth three times more.

The differences that change the month's result — in practice

QR does not replace the printed menu: it completes it. Print governs service pacing and backs up the server's suggestive selling; QR handles delivery, accessibility, same-day price changes and analytics on what gets viewed but never ordered. Drop either one and you lose a whole function.