Profitable menu: criteria to build it — building by taste vs the Masterestaurant method

Which one wins: the menu built on the chef's taste or the menu built on contribution margin?

The menu built on contribution margin in dollars wins, and the gap is not a matter of opinion. For example, take two dishes from the same list:

the mushroom risotto sells at a high percentage margin but leaves little money per unit, while the charcoal-grilled steak has a lower percentage and leaves much more per unit. A taste-driven menu puts the risotto top right and banishes the steak to the bottom corner, because the percentage looks prettier; a criteria-driven menu multiplies dollars by units and finds that the steak, with its steady monthly volume, contributes several times what the risotto does. That monthly gap is easily one line cook's salary. Percentages pay vanity, dollars pay payroll, and the owner who ranks the list by percentage is subsidizing his own ego every single week.

Sixty-eight references against twenty-eight: what happens to waste in each scenario

The short menu wins on waste for a physical reason, not an aesthetic one: every reference removed frees walk-in space, cuts a long-tail supplier and shrinks the inventory exposed to spoilage. In operations where the cut is applied with discipline, food waste drops to a fraction of what it was as a share of food cost. Run the math on your own restaurant: on a month of food cost, cutting waste from a high level to a low one returns a sum every month, without selling one extra plate or raising a single price. A 68-item menu also forces you to buy small formats at list price. That is the invisible invoice of generosity misunderstood.

The 90-day sales mix outranks intuition, and here is the numerical proof

Before touching a typeface you have to measure the real mix of the last 90 days, because the dining-room instinct and the POS ticket rarely agree. The menu engineering benchmark is explicit: a large share of the orders in each category should concentrate in the star dishes, while the rest of the menu supports variety without competing with them. On long menus that figure usually runs higher, with demand scattered across twenty references nobody remembers. The criterion Diego F. Parra applies through the Masterestaurant method flips the usual order: mix first, dollar margin second, and only then restaurant menu design, boxes and visual hierarchy. Whoever starts with design is dressing up a cost structure he has not measured, and the makeup lasts until the first month-end close.

Food cost per recipe: 32 % is a ceiling, not a target, and the difference costs cash

Treating 32 % food cost as a goal instead of a hard ceiling is the costliest costing mistake I see repeated. The criteria-driven menu calculates recipe by recipe, with weighed gram counts and purchase prices updated against the USDA food price index, and accepts an anchor dish touching 32 % only when its dollar margin justifies it and its mix sustains it. The taste-driven menu averages: it declares that everything sits near target while hiding a few dishes with a much higher food cost that carry a large share of volume. At the same sales, those three mispriced dishes drain a meaningful sum each month compared with properly costed ones. And beware the creative accounting of loading payroll or rent onto the plate: that belongs in break-even, never in the recipe.

Physical architecture and descriptions: where the criteria menu pulls away from decorative design

Menu architecture steers the eye and copy closes the sale, but it only works when the dishes underneath actually leave dollars. The hard data came from Cornell's Food & Brand Lab, with Brian Wansink: dishes carrying suggestive labels and descriptions sell up to 27 % more than the same dishes with no description. Now set the two scenarios side by side. Apply that lift in sales from descriptive labels (+27 %, according to Cornell University Food and Brand Lab) to the risotto and the gain is small; apply it to the steak, which leaves more per unit, and you gain much more for exactly the same writing effort. Decorative design pushes whatever looks pretty, criteria-driven architecture pushes whatever pays. Same resource, more than double the return, and that decision gets made before you order the printing.

The case: a grill house trimming a long menu down to a short one, with monthly sales as the starting point.

An urban grill house with a sprawling menu and healthy monthly sales arrived with the classic diagnosis: an average check frozen for eighteen months and a declared food cost that, measured recipe by recipe, turned out to be several points higher. We cut to 26 dishes using one single criterion, contribution dollars per unit multiplied by units sold over 90 days. Most of the references fell away; together they carried a small fraction of sales and consumed four suppliers. Four months later the average check had climbed, waste had dropped noticeably and real food cost landed under the ceiling. In cash, that meant a clearly higher monthly contribution margin. No new dishes, no across-the-board price increase, just three signature plates absorbing the freed capacity.

The real tension: fewer references against a menu with character

There is an honest tension here, and settling it with a lukewarm middle ground would be lying to you: the kitchen with the most soul wants more references, and the owner who cuts blindly ends up with a sad menu nobody remembers or recommends. The way out is not the average, it is asymmetry. You cut from the bottom with data, removing anything with a marginal share of mix that fails to contribute a meaningful monthly amount, and reinvest that freed capacity into two or three signature dishes priced above their category. What would happen if you cut without reinvesting? Waste would fall for a year, identity would fade by the second, and you would end up competing on price against the place next door, the one war an independent restaurant cannot win.

What to choose based on your business profile?

If you run a single location under 40,000 USD in monthly sales and you have not pulled the 90-day mix from your POS, start there and leave the menu alone for now:

without that data, any cut is a bet. If you already bill between 40,000 and 120,000 USD and your menu passes 45 references, the Masterestaurant method applies in full and the return shows up within the first quarter, with waste as the earliest visible signal. Multi-unit operators reverse the order: standardize recipes and gram counts first, because a badly costed short menu replicated across four sites multiplies the error by four. And if you sell delivery, remember that more than 40 % of adults order delivery or takeout 3 to 5 times a month, according to UpMenu (2024), so the digital menu needs its own cut. Pull the last 90 days of mix this week and rank it by contribution dollars.

The four differences that move the cash

Contribution margin is measured in DOLLARS, not percent. A dish with a high percentage margin that leaves little cash per unit is a worse business than one with a lower percentage that leaves much more, and the menu built by taste makes that call backwards every week. Sort your list by contribution dollars times units sold and the real engine of your restaurant appears — rarely the dish the chef is proudest of. The cut is operational, not cosmetic. Every reference you remove frees walk-in space, kills a long-tail supplier and lowers waste, and that waste is money you already paid. Where this gets applied with discipline, food waste falls to a fraction of what it was, which in a busy restaurant means real money every month that stops going into the bin.

The four differences that move the cash — in practice

Menu architecture is worth as much as the price. Menu Engineering, the classic work by Kasavana and Smith at Michigan State University, proved forty years ago that position and visual emphasis shift the sales mix without touching a single price; what changed by 2026 is that the POS confirms it in two weeks instead of leaving it to assumption. A dish moved to the top-right vertex of a block sells noticeably more units. The QR replaces nothing: it covers what the printed menu cannot. I got this wrong for years, recommending full digitization to save printing, and the hidden cost surfaced in suggestive selling, which collapsed once the server lost the object to point at. The printed menu governs service rhythm and menu narrative; the QR handles delivery, allergens, price changes and analytics. Both, each in its role.