Which dishes to cut from your menu for profitability: myth, reality and four honest alternatives

The two numbers that decide, and the one that decides nothing

Cut a dish only when it fails on both dimensions at once: contribution margin in dollars below your menu's median AND units sold in the bottom quartile, measured across 90 days of real sales, never across one stray month. That intersection —menu engineering calls them the dogs— rarely exceeds 12 % of a long menu, and cutting them almost never hurts. Food cost percentage, the first number everyone stares at, decides nothing on its own: a ceviche at 34 % that leaves USD 5.20 per plate and sells 180 units a month hands you USD 936, while a salad at 22 % that leaves USD 1.60 and sells 14 hands you USD 22.40. The full-service median ran to 32.0 % of sales in 2024, according to the National Restaurant Association's Restaurant Operations Report, and that figure is a traffic light on your purchasing process, not a verdict on the dish.

When the long menu stops working as a diagnosis?

The signal that gut-feel pruning has run out appears when the owner shows up with candidates picked by eye and, once the standard recipes are costed, half of them turn out to sit among the top margin generators.

It happened in Bogotá with a 74-dish menu in a 46-seat room, declared food cost of 31 % and cash that never closed: of twelve candidates for the guillotine, seven ranked in the venue's top fifteen by margin. Cutting them would have shaved roughly 9 % off the quarter's cash. One cheap test exposes the limit: if you cannot say, without opening the POS, how many dollars your seventh best-selling dish leaves, you do not have a diagnosis, you have an opinion about what tires you to cook. And I got this wrong for years, recommending short menus out of doctrine before evidence. Before deleting a dish, raise its price and measure elasticity for six weeks; it is the cheapest of the four routes and the one that destroys the fewest assets.

Option 1: reprice before you reach for the guillotine

It suits the high-rotation, thin-margin plate that the kitchen fires 200 times a month for crumbs: an adjustment of 8 % to 12 % on the menu price usually holds volume, because the guest already ordering it orders for taste, not for cents. Profile: chef-owner with a POS that splits units by dish and the ability to reprint the menu or swap the QR. Switching cost: near zero in money, two weeks of nerves. Downside: if that plate is your price hook on delivery, the increase moves your ranking inside the app and you lose on the other side. Keep per-dish food cost at 32 % as a MAXIMUM, remembering that this ceiling is a limit, not a target. The second exit touches cost without touching price, and the industry already picked it: switching suppliers was the number-one strategy against rising costs for 40 % of operators in 2024, according to TouchBistro.

Option 2: redesign the recipe and renegotiate the input

It works when two or three inputs concentrate the plate's spend —the protein cut, the cheese, the oil— and a renegotiation of 6 % to 10 % on those lines restores margin with the guest noticing nothing. Profile: a venue with volume enough to demand scale, or one inside a buying group. Switching cost: four to eight weeks of testing, plus the very real risk that the substitute ruins the dish. Downside: a botched substitution gets paid in reviews, and a review that says it no longer tastes the same costs more than the dollars saved. Run blind tastings with regulars before you sign with anyone. Placement sells. The high-margin, low-rotation plate —menu engineering's puzzle— seldom needs to die: it needs to leave the footer. Move it into the first third of its category, give it a photograph, name it after the ingredient rather than the adjective, and drill the server on a twelve-word pitch.

Option 3: move the dish on the page instead of killing it

The same principle Taco Bell measures in its digital channel, where self-service generates 20 % more spend than a human cashier according to Yum! Brands, operates on your printed menu: choice architecture moves money. Profile: a venue with healthy margin and stable traffic that needs a higher check without a higher price. Switching cost: a menu redesign and two service briefings. Downside: if the plate takes fourteen minutes to leave the pass on a full Friday, pushing it wrecks your kitchen timings and punishes the next table. Cut when the dish fails on both dimensions and you have already exhausted price, recipe and placement; the removal then frees up purchasing, cold storage, mise en place and minutes on the line. Book what goes dark, though: every retired dish is a content asset that dies —the photographs, the Reels built around it, the reviews that name it, the branded searches of the restaurant-plus-dish kind— and winning that traction back costs months of posting.

Option 4: cut it, and count the hidden cost nobody books

Diego F. Parra keeps pressing an uncomfortable point with chefs: the menu is not a kitchen document, it is a cash asset and a marketing asset at once, and at Masterestaurant we score a cut by what it frees and by what it switches off. Working rule: never retire more than 15 % of the menu in one pass, and if the dish shows up in your last ninety days of content, keep it three more months and migrate the audience to its replacement. Payroll, rent and utilities do not load onto the plate: they belong to the venue's break-even. When someone spreads USD 9,500 of monthly payroll across 74 dishes to see which one fails to pay, that person manufactures a false culprit and ends up cutting the plate that carried Thursday's cash. Plate costing covers inputs, shrinkage and portioning waste; nothing else.

The allocation error that manufactures false culprits

Test the opposite and watch it collapse: load fixed costs per dish and the highest-rotation plate absorbs the largest share of overhead, so it surfaces as the least profitable and the model advises you to amputate your own engine. That is the paradox that sinks menus: the more a dish sells, the worse it looks under proportional allocation. The National Restaurant Association places optimal food cost between 28 % and 35 % depending on concept, and that range only makes sense read over inputs, never over a plate carrying the rent. Standing still is the right call in three concrete situations, and it pays to recognise them before you touch the menu. First: your POS has been properly configured for fewer than ninety days and your standard recipes remain uncosted, which makes any pruning a bet. Second: the venue has just changed chef, location or prices within the last sixty days, and the noise from that transition contaminates any reading of rotation.

When NOT to change anything, said plainly?

Third: aggregate margin is already healthy —positive cash, food cost inside the sector range, which the National Restaurant Association measured at 33.7 % for full-service operators under USD 2M in sales in 2024— and your real problem is traffic, not menu.

Cash flow is the leading cause of financial stress and closure among small businesses, according to Inc., and cutting dishes on bad data speeds that closure rather than braking it. Spend eight hours instead: sit with the POS, export ninety days, and cost your fifteen best sellers before deleting a single one. A pruning that works is decided with two numbers per dish—contribution margin in currency and units sold across 90 days. A pruning that sinks the till is decided by the chef's taste or by the server who hates plating the risotto. A dish at 31 % food cost selling 200 units leaves more cash than one at 19 % selling 11; the percentage is a traffic light, the currency is the destination.

What separates a pruning that lifts the till from one that sinks it?

Hold food cost per dish at 32 % as a MAXIMUM, and treat that ceiling as a limit rather than a target. Payroll, rent and utilities never load onto the plate:

they belong to the break-even of the house. Whoever spreads payroll across 74 dishes to see which one loses money invents a false culprit and ends up cutting the dish that carried Thursday's till. Diego F. Parra presses an uncomfortable point: the menu is a marketing instrument before it is a kitchen document, and Masterestaurant works it with the rigor of a sales funnel—what the guest sees first, what they remember, what they order by name on Instagram. The printed menu and the QR menu do different jobs and neither replaces the other: paper controls service rhythm, menu narrative and suggestive selling; QR handles delivery, accessibility, price updates and analytics on what gets viewed without being ordered.

What separates a pruning that lifts the till from one that sinks it — in practice?

Drop the printed menu to save on printing and you lose the single instrument that moves average check most. The datum that kills the myth:

a shorter menu does not sell more by itself. It sells more when what remains is properly costed, properly photographed and properly suggested in the dining room—three separate jobs, only one of which happens in the kitchen.