Which Menu Items to Cut to Gain Profitability: the numbers before and after

Which dish should you cut from the menu first?

The first one out is the dish that fails both tests at once: it sells below 70 % of its fair share of the mix and it returns a dollar contribution margin under the median of its own category.

On a 24-item menu, each dish's fair share is 4.17 % of units sold, which puts the suspicion line at 2.9 %; any dish under that line that also earns fewer dollars per unit than its family median goes on the cut list without argument. Watch the order of those two tests, because that is where almost everyone gets it wrong: dollars first, percentage second. A dish running 38 % food cost that keeps 14 USD and moves 300 units a month brings in 4,200 USD of margin; another at 22 % that keeps 6 USD and moves 40 brings in 240 USD. The first one pays payroll. Cutting by food cost percentage is the most expensive mistake made on menus, and it gets made every single month.

Food cost percentage as your only test destroys margin

The reason is arithmetic: percentage measures purchasing efficiency, not contribution to the till, and you do not pay rent with percentages, you pay it with dollars. Cornell measured that disciplined menu engineering lifts profitability by roughly 10 % on average, and Oracle NetSuite puts the sustained range between 10 % and 15 %; neither one gets there by deleting high-food-cost dishes. Run the counterfactual: cut your five dishes above 35 % food cost tomorrow, and if those five carried 28 % of your units, the menu looks cleaner on the spreadsheet while the till loses thousands of dollars that month. Sort your menu by dollar contribution margin multiplied by units sold rather than by percentage, and the candidate list flips from top to bottom. Across the menus Masterestaurant reordered during 2026, somewhere between 18 % and 24 % of the items accounted for under 3 % of total sales, and that fifth of the menu exists for nobody.

How much menu exists for nobody: the long tail of the mix?

Translate the number into operations:

68 dishes in a 45-table dining room means 68 recipe cards to update every time a supplier moves a price, 68 supply chains and 68 possible waste lines, with a third of your inventory either turning every fortnight or dying in the freezer. Diego F. Parra puts it in one dry sentence in every engagement: a menu is not designed so the guest can choose, it is designed so the guest DECIDES fast. That context weighs more now than before, because the National Restaurant Association reported in its State of the Restaurant Industry 2026 that 62 % of operators name food costs their leading operational challenge. Here is the exception that will save you from some dumb decisions: a dish with a high dollar margin and weak sales is almost never a bad dish, it is an invisible one. Move it and label it before you delete it, because the placement levers are measurable.

Before you cut, test whether the dish is simply misplaced

NeatMenu documented in its Menu Psychology 2026 that tagging an item as «most popular» or «chef's favorite» raises orders by 13 % to 20 %, and Cornell's menu design research reports as much as 30 % more sales when the dish carries a photo, with an effect near 6.5 % per professionally photographed item. Give it thirty days with new placement, a tag and a photo; if it still cannot clear 70 % of its fair share after that full cycle, visibility was never the problem and it goes. Dishes with low dollar margin and low sales do not earn that test: those leave on Monday. The chef wants room to express, the till wants turnover, and that tension is real, so do not dodge it with talk of balance. The way out is not the chef backing down, it is splitting the menu into two pools with different rules: 20 to 24 permanent items defended on dollar margin, plus 3 to 5 rotating slots where the chef rules and owes nobody a mix report during the run.

The tension between the chef and the till is settled with a quota

Permanent items get audited every 90 days against both tests; rotating items get judged when the season closes, and any that beat their category margin median earn a permanent seat by pushing out the weakest occupant. On a 24-item menu, five rotating slots are 21 % of the live card and cost five recipe cards a quarter, not sixty. The till sets the bias; the talent decides inside the quota. The thresholds do not change, the size of the cut does, so place your case in one of three scenarios. Small room, up to 40 seats and fewer than 30 items: fair share per dish sits near 3.3 %, the cut line lands at 2.3 %, and you remove 3 to 5 dishes per cycle, never more, because a three-person kitchen absorbs mise en place changes badly. Mid-size restaurant, 60 to 120 seats with 40 to 60 items: this is where the 18 % to 24 % long tail lives, and cutting 10 to 14 dishes on the first pass is normal.

How to read these numbers in YOUR operation?

Multi-unit group: do not cut on a consolidated average, because an office-district location and a neighborhood one never share a mix;

measure each point of sale on its own and keep any dish that clears its fair share in even one location, even when the consolidated view condemns it. Be honest about what these numbers can and cannot do for you. The sector figures come from verifiable public sources: the National Restaurant Association for costs and operating pressure, through its State of the Restaurant Industry 2026 and its reading of +35 % in food and +35 % in labor since 2019; Cornell and Oracle NetSuite for menu engineering returns; NeatMenu for the menu psychology levers. Three limits deserve stating out loud: most of that evidence is American and its margins do not transfer to a market with a different labor cost structure; menu engineering studies measure an average effect, not your menu; and label and photo levers get measured in short windows, with decay once everything carries a tag.

Where these benchmarks come from and where they break?

The 70 % fair share rule and the category median are internal decision rules of the Masterestaurant method, calibrated on real mix data rather than an academic finding.

A menu gets cut on a calendar, not on impulse, and the cycle that holds margin is 90 days. Pull 90 days of mix by units and by dollar margin, work out fair share by dividing 100 by your item count, flag everything sitting under 70 % of that share, then cross that list against each category's margin median; whatever lands in both columns leaves, and whatever lands in only one moves into the thirty-day test with a tag and a photo. Without that calendar a menu grows by accumulation: in comes the new supplier's dish, in comes the partner's pet project, in comes the special nobody ever pulled, and two years later you carry 68 items nobody decided to carry.

The cutting calendar, and what happens without one

Open your mix sheet this week and calculate fair share for every dish before you argue with anyone about what stays. The most repeated mistake: cutting by food cost percentage. A dish at 38 % that leaves USD 14 per unit and sells 300 a month contributes USD 4,200 of margin; a dish at 22 % leaving USD 6 that sells 40 contributes USD 240. The first one pays payroll. The second one buys someone breakfast. Second confusion: believing a long menu protects you against the guest who finds nothing. Sales-mix data says otherwise. Across menus Masterestaurant reordered during 2026, between 18 and 24 % of references concentrated under 3 % of total sales, which means a fifth of the menu existed for nobody. There is a real tension and it deserves resolution rather than avoidance: the chef wants expression and the register wants turns. The answer is not that the chef gives in; it is that expression lives in 4-6 signature dishes, with calculated margin and price headroom, while the rest of the menu works as profit structure.

Where the decision actually happens?

Diego F. Parra frames it backwards from the usual approach: first decide what margin the business needs per guest, then design the menu that produces it.

On digital menus versus printed ones, the Masterestaurant position has not moved: BOTH, each with its role. The PHYSICAL menu controls the experience, service pacing, menu narrative and the server's suggestive selling; the QR is a complement for delivery, accessibility, price updates and analytics on what gets viewed without being ordered. Killing the printed menu to save on printing is among the most expensive decisions I have watched an owner make with a savings face on. What happens if you pull 20 dishes at once without reading the mix? Most likely this: you take down two items that anchored a specific segment (the group's vegetarian, the kids' plate), that group stops coming, the traffic drop reads as "the cut was a mistake", all 20 dishes come back, and the business lands exactly where it started minus six weeks and a demoralized team. Cut in two rounds, with eight weeks of measurement between them.