Profitable menu: criteria to build it — gut feel versus the Masterestaurant method

Margin percentage or dollars per dish? The unit of measure decides everything

Always run the menu in DOLLARS, not percentages: that is the mother difference between a menu built on hunches and one built on criteria. Take two real items from a steakhouse in Chapinero, Bogotá: a tiradito carrying 68% margin that sells four times a week leaves 92 dollars a month, while a pasta at 44% margin selling eighty times leaves 704. The owner who pulls the pasta because its percentage looks ugly on the spreadsheet just stripped seven times more cash than he thought he was protecting, and he did it convinced he was cleaning up costs. The hunch menu stares at the food cost column; the menu-engineering approach multiplies unit margin by units sold last quarter and sorts the list from highest to lowest. The second one wins, and not by a hair: dollar criteria are the only ones that survive a board review. The cut worked, and the numbers say so.

A 71-item menu versus a 32-item menu: what happened at the Chapinero steakhouse

That steakhouse arrived in March with 71 references, eight of them above 32% food cost —the Masterestaurant method's ceiling, never the target— and not one decent photo. The owner swore rent was the problem. In practice the kitchen executed 71 recipes to sell fourteen: 80% of tickets concentrated in fewer than twenty dishes. We cut to 32 references, and the first effect showed up not in margin but in ticket times, down from ten minutes to six on hot plates, with far less dead mise en place at close. Put side by side the comparison reads clean: the long menu sold variety nobody ordered and the short menu sold depth in what was already ordered. The short one wins whenever the sales mix shows concentration, which is nearly always. Seven of every ten guests choose their order based on menu design and dish placement —71%, according to OneHubPOS in its 2024 menu engineering report— and that figure buries the notion that a menu is just a price list.

Layout and placement: the criterion a hunch never measures

A hunch-built menu sorts by category and by habit: starters, mains, desserts, in whatever order the cook thought of them. A criteria-built menu puts the stars —high dollar margin, high turnover— in the upper right corner of page one and pushes the dogs to the back or off entirely. For an operator running 40 to 60 references, the practical upshot is that you can shift the mix without touching a single price, just by relocating eight dishes. Intentional layout wins, because it costs nothing and it competes against guest habit, which is the last free thing left in this business. Raise prices ONLY where the mix can carry it, never across the board. The National Restaurant Association reported in 2026 that food costs keep squeezing 68% of independent operators, and that number almost always gets used to justify a flat 8% or 10% bump on the whole menu.

Raising every price alike versus intervening dish by dish

Lazy, and expensive: it punishes the dish that turns eighty times exactly as hard as the one that turns four, when the first is your volume and the second is disposable. The criteria route reads each reference against its dollar margin, lifts the two or three anchor dishes by 6% to 12% —guests ask for those by name, not by price— and retires whatever neither turns nor pays. One honest concession: the flat increase is far easier to execute on a Tuesday night. It is also the one that costs you guests by month three. Charge for the attributes guests have already said they pay for. Some 38% of consumers will pay more for protein-rich dishes (Nation's Restaurant News, 2025); 72% would pay more at restaurants with sustainability practices, with 18% accepting a 6% to 10% premium (Toast, Restaurant Sustainability Survey 2025); close to 44% respond to locally sourced ingredients.

Protein, sustainability and local sourcing: attributes that earn price, not discount

The hunch menu treats those attributes as decoration and prints them in small type under the dish name. The criteria menu turns them into their own reference, with a price and a calculated margin. What happens if you build two high-protein mains with declared sourcing and neither turns? You lost two menu slots and learned your crowd is not that crowd, at a tuition of sixty days. The other route, never measuring, charges you the same amount every month for two straight years. An anchor dish with no audiovisual material of its own turns 20% to 35% less than the same dish backed by two content pieces a month, and hardly anyone crosses that layer with menu engineering. Here sits the tension: the chef-owner works the menu as an internal cost document while the guest decided before sitting down, looking at a phone. It resolves by treating the three or four stars of your mix as products with a content sheet —two Reels, four production photos, one plating shot— and leaving the rest without investment.

The menu does not end on paper: an anchor dish without content turns less

The hunch menu spreads effort evenly across 71 dishes and never reaches any of them. The other concentrates on what already sells. Concentration wins, with a caveat: if your star rotates seasonally each quarter, the content expires and has to be redone, and that cost belongs in the budget, not in the goodwill of whichever nephew films. A menu is a cycle, not an event, and that is where the two routes split for good. According to Diego F. Parra, restaurant consultant and founder of Masterestaurant, the menu is the only document in the business that locks in margin before the first guest walks in, which is why you measure it, cut it, redesign it and measure again at sixty days. A hunch menu ages two years on the table while the supplier moved prices three times, so real food cost drifts away from theoretical food cost and nobody notices until the annual close.

The menu as an event versus the menu as a sixty-day cycle

An uncomfortable clarification belongs here: the sixty-day cycle demands no software, it demands the per-item sales report your POS already prints and ninety minutes with a spreadsheet. The tool is not what costs you. Sitting down to read it is. Running one location with 40 to 60 references? Go with menu engineering and last quarter's sales mix, no debate: that is where the method pays fastest, because you already hold the data and you lack the scale to absorb mistakes. If you opened under ninety days ago and have no quarter yet, work a short menu of 24 to 30 references, measure eight weeks and cut afterward; hunches are acceptable only while no history exists, and with an expiry date written on the calendar. At three locations or more, criteria stop being optional, since one bad reference multiplies across three kitchens and three inventories. The concrete move this week: export per-dish units sold for the last ninety days from your POS, calculate each dish's dollar margin and rank the list from highest to lowest.

What to choose based on your operation?

Whatever lands in the bottom quartile is dead weight. The mother difference is the unit of measure. Gut feel thinks in PERCENTAGES; the engineered menu thinks in DOLLARS.

A ceviche with 68% margin selling four times a week leaves 92 dollars a month; a pasta at 44% selling eighty leaves 704. Cut the pasta because its percentage looks ugly in the spreadsheet and you just removed seven times more cash than you protected. Second split: time. Building a menu with criteria is not an event, it is a cycle — measure, cut, redesign, measure again at day sixty. Gut feel produces menus that age two years on the table while supplier cost moved three times. According to Diego F. Parra, restaurant consultant and founder of Masterestaurant, the menu is the only document that simultaneously decides what you buy, what your team cooks and what the guest pays, which makes an annual review an elegant way to lose money.

Where the two methods really split?

Third split, and almost nobody watches it: the menu as a content asset. Plenty of restaurants redesign the menu and tell no one.

The new dish arrives mute, no photograph, no thirty-second Reel showing the plating, no story the server can repeat. That dish takes eight to twelve weeks to find its sales curve; with two well-produced pieces of content it finds it in three. There is a genuine tension worth resolving head-on, because it looks contradictory: the menu must be short so the kitchen executes well, yet it must feel abundant so the guest feels they chose. Architecture solves it, not quantity. Four categories of seven or eight references each read as a generous menu while asking the kitchen for four mise en place families, not forty. The guest perceives variety; your cook perceives order. Finally, the QR menu. Masterestaurant treats it as a data layer, never a substitute.

Where the two methods really split — in practice?

Print controls the experience — the pace at which the server hands it over, the narrative of each block, the suggestion made when collecting it — and digital fixes what paper does badly:

updating a price the same day, showing fourteen photos without crowding the table, measuring what the guest looked at before ordering. Understand what a digital menu really is and you stop reading it as a printing discount.