How to design a menu that increases profits: the classic matrix, its limits and four honest alternatives

When classic menu engineering runs out of road?

The classic matrix of stars, plowhorses, puzzles and dogs runs out of road at the exact moment the guest stops reading and starts listening.

The tell comes from the floor, not from the card: a guest spends an average of 109 seconds reading the menu, according to NeatMenu (Menu Psychology 2026), and most of that minute and a half evaporates the second a server reaches the table and delivers the same recommendation as always. You rebuilt the card on a popularity-versus-margin grid, printed it, and the real mix kept being governed by the habits of eight people who never saw that grid. The alarm is concrete: if your best-selling dish does NOT match any of your four target items after three months with the new menu, the problem stopped living on paper. Food cost percentage is a purchasing indicator, and payroll gets paid in dollars, never in percentages.

Measure contribution margin in dollars, not in food cost percentage

A dish at 24% that leaves $2.20 of contribution loses against one at 31% that leaves $4.00, and yet the first one usually wins every committee argument because its number looks prettier on a slide. With median pre-tax profit in full service sitting at 2,8% of sales during 2024 (National Restaurant Association, Restaurant Operations Report 2024/25), the room for picking the wrong metric is literally three cents on every dollar sold. At Masterestaurant the first sheet we ask for is neither the menu nor the recipe book: it is the dish list sorted by contribution margin in dollars, with units sold beside it. The same uncomfortable finding shows up almost every time, because the flagship dish sits in the bottom half. Divide margin in dollars by the minutes a dish occupies your bottleneck and you end up with a metric nobody around you is using.

Option 1 — Contribution margin per station minute (for kitchens turning tables away)

A $4.00 contribution plate that eats 11 minutes of griddle yields 36 cents per minute; a $2.75 pasta that leaves in 4 minutes yields 69. At peak, when the griddle is the ceiling of the whole restaurant, pasta wins. Cost of the exercise: zero, since it comes from timing 20 dishes across two full services. Learning curve: medium, and the real risk is human rather than technical, because somebody has to hold the stopwatch without helping out the cook they happen to like. Who it fits: operations that fill up and turn tables away between 8 and 10 at night, where the constraint was never margin but station rotation. Leave the card exactly as it is and move the mix from the service itself, which is where it actually gets decided. The setup is easy to describe and hard to sustain: three target dishes per shift, a 15-second script per dish, simulator practice before service and a per-server points board that closes every week.

Option 2 — Floor training with a simulator and a points board

Cost: between 480 and 1.200 USD a year in platform, plus 12 daily minutes of pre-shift that come out of payroll you already pay. Set that against the 9.400 USD of a full graphic redesign and the arithmetic defends itself. Who it fits: table service with a stable crew and a floor manager who wants to measure. Contraindicated if your server turnover runs past 80% a year, because you will train people who leave before they pay back. Cutting categories is the cheapest option on this list and the one your kitchen will fight hardest. Aggregated menu design research places the healthy range between 7 and 15 items per category, and going past that pushes the guest into decision paralysis, which in practice means ordering whatever they had last time or whatever the server names first. If you carry 26 appetizers, you do not have 26 appetizers: you have 9 that carry the sales and 17 that inflate inventory, stretch line times and hide waste.

Option 3 — Prune the card down to the range the brain tolerates

Cost: reprinting, somewhere between 300 and 900 USD depending on format. Who it fits: menus that grew by accumulation over five years with nobody ever removing anything. Honest risk: every pruned dish has its loyal customer, and that customer complains loudly during week one. When most of your sales no longer happen in your dining room, redesigning the placemat means optimizing the wrong channel. Off-premises traffic reached 30% in full service during 2024, up from 19% in 2019, and 83% in limited service (National Restaurant Association, Off-Premises Report 2024). On a screen the four hot spots of a printed sheet simply do not exist, the first three scrolls do, and category order weighs more than any typeface. Add that 34% of online ordering customers spend $50 or more per order (Statista): that is where a well-built combo and automatic upsell lift the check without depending on anyone speaking.

Option 4 — Redesign the digital menu before the printed one

Who it fits: operations where owned delivery or aggregators clear 35% of sales. Cost: hours from whoever manages the catalog, usually under 40. Answer these four and the decision makes itself, no committee required. First: are you turning tables away at peak? If yes, start with contribution margin per station minute, because your constraint is time rather than money. Second: does your best seller sit in the bottom third of dollar contribution? Then your problem lives on the floor, and training with a 15-second script will pay back faster than any print shop. Third: does any category run past 15 items? Prune first, since none of the other tools work on a card nobody finishes reading. Fourth: does off-premises clear 35% of your sales? Then the menu that matters is the digital one. Diego F. Parra keeps hammering the sequence, and the sequence is not up for negotiation: measure first, then choose what moves it.

When NOT to change anything (and hold your hand steady)?

Sometimes standing still is the profitable call, and saying it out loud costs consulting clients.

Do not touch the menu if the current version has been out less than 90 days, because you still lack a clean read on the mix and you will credit your change for what is really seasonality. Leave it alone as well if you just changed chefs, if you are mid-renovation, or if your raw material cost moved more than four points last quarter: with the base shifting, any measurement you take lies to you. And if your pre-tax profit already sits above the sector median, 2,8% of sales per the National Restaurant Association, a redesign has far more to lose than to gain. Let two full purchasing cycles run, measure contribution per dish with real numbers, and come back in April. ALTERNATIVE 1 — Marginal contribution per station minute. Instead of splitting by popularity and margin, divide dollar margin by the minutes the dish occupies its bottleneck (plancha, oven, fryer).

The four alternatives, each with its who-it-is-for

Cost: zero, it comes from timing 20 dishes across two services. Curve: medium, someone has to hold the stopwatch without bias. Who it is for: kitchens that fill at peak and turn tables away, where the constraint is station speed rather than margin. ALTERNATIVE 2 — Floor training with simulator and gamification. Leave the card alone and move the mix from service: three target dishes per shift, a 15-second script each, simulator practice and a points board per server. Cost: 480 to 1,200 USD a year plus 12 daily preshift minutes. Curve: low, a shift lead sustains it without a consultant. Who it is for: table-service operations with high turnover, where new hires arrive monthly and the script has to be reproducible. ALTERNATIVE 3 — Reverse-engineered menu from the target check. Set the check you need to hit break-even, subtract drinks and dessert, then design the card backwards: how many dishes per price band and what contribution each band must deliver.

The four alternatives, each with its who-it-is-for — in practice

Cost: nothing in money, plenty in owner hours. Curve: high, it demands understanding break-even without confusing it with plate food cost. Who it is for: openings and relaunches, where no historical mix exists yet. ALTERNATIVE 4 — Short physical card plus extended QR with analytics. Cut the printed menu to the 18-24 references carrying the margin and move the rest (seasonal, extended drinks, allergens, delivery) to the QR menu, which also tells you what guests browse before ordering. Cost: 0 to 900 USD depending on platform. Curve: low. Who it is for: houses whose card swelled from years of adding without subtracting. Masterestaurant rule, no exceptions: the PHYSICAL card stays. It controls service pace, menu narrative and suggestive selling; QR complements with price updates and analytics. Never QR-only. And one I rule out: graphic redesign as the opening move. Typography, photography and dropping the currency symbol shift something once everything else is settled; as a first step it is expensive makeup over costing nobody reviewed.