Menu engineering by the numbers: the mistakes that drain cash and the method that fixes them

The steakhouse fighting over two points of beef cost while its menu sold the weak dish

A 96-seat steakhouse in Medellín sold 340 tenderloins a week at 9,800 pesos of contribution margin each, and 610 pastas that returned 21,400; the menu placed the tenderloin dead center on page one, with a photo, and banished the pasta to the bottom right corner, greyed out, with no photograph. For eight months the owner had been negotiating two points of food cost with his beef supplier while his own menu was asking every guest to buy the dish that paid him least. Cornell measured that disciplined menu engineering lifts profitability by roughly 10%, and Oracle NetSuite reports a sustained 10% to 15%; neither number comes from a supplier negotiation. They come from MOVING THE MIX, the one variable you control without anybody's permission. Your unit of measure decides the conversation, and food cost percentage decides it badly. Take a salad at 22% food cost that returns 8,400 pesos and moves 90 plates a month: 756,000 pesos of contribution.

Percentages versus pesos: why the salad always wins the wrong argument

Now the cut at 34% that returns 26,000 and moves 210: 5.46 million, more than seven times as much. The percentage crowns the salad; the cash register pays payroll with the cut, and in 2024 that payroll climbed 4% to 14.20 dollars an hour in the United States according to 7shifts, with equivalent pressure across Latin America. When labor rises and percentage margin becomes religion, the operator shaves portions off the dish generating the most pesos and sinks himself. Contribution margin per unit multiplied by units sold, in pesos, sorted high to low: that column is your real menu. Ninety days of POS sales, not one day less. That window absorbs paydays, holidays, a long weekend and the rebound from a one-off campaign without diluting a genuine trend; at thirty days you are reading the noise of a promotion, and at a full year you bury the effect of the menu you changed in March.

How many days of mix do you need before deciding anything?

Menu prices at large U.S.

chains rose 42% between 2020 and 2025, nearly double the 22% of general inflation according to One Haus, which means guest behavior toward your menu changed at the root over that stretch and your old data lies to you. Without a consistent window there is no menu engineering: there is opinion with a spreadsheet. Set the cut, freeze it, and always compare quarter against the same quarter last year, never against last month. A dish with a photograph sells up to 30% more, and the measured effect per dish with professional photography runs around 6.5%, per Cornell's menu design research collected by NeatMenu. Labeling a dish «Most popular» or «Chef's favorite» moves orders between 13% and 20%, also per NeatMenu in its 2026 menu psychology report. Look at the steakhouse pasta: 21,400 pesos of margin, 610 units a week with no photo and no label.

The photo and the label are engineering, not decoration

A 15% push is 91 extra plates a week, 1.95 million pesos of additional monthly contribution, without raising the price a single peso and without touching the recipe. What most owners file under graphic design is in fact the cheapest financial lever in the house. The expensive mistake is handing that decision to the designer instead of to whoever watches the register. A creator's post lifts reservations 30% the following week, according to Marketing LTB's 2025 influencer marketing statistics roundup, and there sits the most expensive paradox in this trade: if that Reel pushes your low-margin dish, you just bought traffic to sell losses. Grubhub logged +34% in matcha orders during 2025 in its Delivered Report, a wave dozens of menus chased without ever calculating the drink's contribution margin. Diego F. Parra keeps insisting at Masterestaurant on a sequence that sounds obvious and almost nobody respects: identify your two high-margin, high-rotation dishes first, then design the content around them, never the other way around.

Content pushing the wrong dish is virality that costs you money

The inside half —standard recipe, contribution, POS mix— and the outside half —Thursday's seven o'clock Reel— are the same job. Self-service kiosk tickets run 8% to 15% above the counter, with Yum reporting close to 10% per QSR Magazine 2024, and McDonald's has communicated increases of up to 30% in average ticket from its kiosk rollout. Upselling inside digital and QR ordering adds 20% to 30% to order value according to sector providers. Those numbers delight any owner until somebody asks which dish is actually rising. If the kiosk algorithm suggests fries and a soda —high contribution, granted— but displaces the main plate that returned 26,000 pesos, the ticket climbs while total contribution falls. The decision these three figures trigger together is one: audit your mix BY CHANNEL, because your delivery menu is already a different menu and odds are nobody designed it. Forty-two percent of price increases in five years against 22% of general inflation (One Haus) carries a direct consequence: your guest arrives with a calibrated mental calculator, and any further adjustment reads as gouging.

Price is not the lever the 2026 guest will forgive

McDonald's proved it from the other side when the $5 Meal Deal lifted visits 8% against the average Tuesday of the year, per Restaurant Dive in 2024 —perceived value, not high price. What happens if, instead of raising the whole menu 6%, you rebuild page one and push 12% of your dishes into the high-margin, high-rotation quadrant? You gain 4 to 7 points of contribution margin, the guest perceives no increase at all, and your average review does not budge. Raising prices is the easy way out and the most expensive one twelve months down the line. Ten percent (Cornell): what disciplined menu engineering adds to profitability, so block two hours every quarter, pull the 90-day mix from your POS and sort the menu by pesos of contribution, not by percentage. Thirty percent (Cornell via NeatMenu): what a photographed dish gains, so shoot the four highest-margin dishes professionally this week and strip the photo off anything paying less than the house median.

The 3 numbers you should tattoo on yourself

Forty-two percent (One Haus): the accumulated 2020-2025 price increase against 22% inflation, so take a price hike off your options list for this half and work on labels, position and content instead. Start today with the cheapest one of all: the «Chef's favorite» tag on the dish that leaves you the most pesos. The first difference is the unit of measure. The average operator compares food cost percentages across dishes, and the salad always wins that contest; the method measures contribution margin in currency multiplied by units sold, and there the dish the house can sustain wins. A salad at 22% food cost returning 8,400 pesos across 90 covers delivers 756,000 pesos a month; a cut at 34% returning 26,000 across 210 covers delivers 5.46 million. The percentage says one thing, the bank account says another, and the bank account pays payroll. The second is time.

The three differences that move cash

Almost nobody reads the sales mix through a consistent window, and without a window there is no menu engineering, only opinion. Ninety days is the cutoff I use because it absorbs paydays, holidays and one-off campaigns without flattening a real trend. Under sixty days, a Reel that worked for a week distorts the entire quadrant and you end up killing a dish that was only starting to move. The third, and the one almost nobody connects, is that video content IS a menu engineering tool, not a separate marketing activity. When you decide what to film this week, you are deciding next month's mix. Producing content for a beautiful low-margin dish means paying to sell worse. The social calendar and the quadrant matrix should be the same document, reviewed by the same person, on the same day of the month.