Plate profitability: the traditional method, its real limits and the alternatives that actually move cash

Row 19 of the spreadsheet that explained the whole menu

A dish running at 27% food cost can be the worst business on your menu, and that chef-owner's mushroom risotto proved it in row 19 of his own spreadsheet: best percentage among 43 dishes, two burners tied up for eleven minutes at the nine o'clock peak, and three pastas fit in that same window that together brought in triple the money. He was selling more than ever and keeping nothing, and he thought he had a purchasing problem. What he had was the WRONG ARITHMETIC. The percentage told him what the ingredient cost relative to the price; it never told him how much money hit the register, or how fast. Those two questions pay the payroll at month's end, and neither one shows up in a food ratio. Percentage food cost falls short the moment your kitchen develops a bottleneck, and the giveaway is simple: look at your prime cost.

When percentage food cost falls short on you?

If food plus labor eats 60% to 65% of sales in full service, which is the range the National Restaurant Association reports, then half of your relevant variable cost lives outside the very ratio you are using to rank the menu.

There is a second signal, less comfortable: when your two or three best-percentage dishes happen to be the longest to prepare. At that point the ratio and the cash register point in opposite directions. Let's be fair to the tool before proposing another one, because the percentage isn't broken: it catches purchasing and portioning drift with a precision no other reading gives you. Using it is not the error. Deciding which dish stays on the menu with it is. Kasavana and Smith published a matrix in 1982 that crosses popularity against contribution margin in dollars and sorts every dish into four quadrants: star, plowhorse, puzzle or dog.

Alternative 1: classic menu engineering, the 1982 matrix that still wins

It costs nothing if your POS exports the sales mix, which any POS does today. Two or three days to understand the quadrants, one week to make the first call without your hand shaking. Who is it for? Any menu of 15 to 60 items with at least 90 days of clean history, and especially the owner who has never ranked dishes by absolute dollars. The calibration benchmark helps: according to the National Restaurant Association, between 35% and 45% of orders per category should land on star dishes. Its honest limit is heavy, though: the matrix treats every kitchen minute as equal, when the Friday nine o'clock grill is worth ten times the Tuesday lunch grill. Divide contribution margin in dollars by the minutes a dish occupies your bottleneck station and you get the only figure that ranks a menu once the kitchen is full. Back to the risotto: 27% food cost, eleven minutes on two burners, against a pasta at 36% that leaves in four minutes on a single burner.

Alternative 2: margin per station minute, where the constraint rules

If the risotto yields 14 dollars and the pasta 11, the risotto returns 1.27 per burner-minute and the pasta 2.75 — more than double — and that gap multiplies across every hour of the peak. Who is it for? Kitchens that saturate at least three services a week. The switching cost is timing dishes with a stopwatch across two shifts, roughly four hours of work, plus pinning down which station is the real constraint. If your kitchen never fills, this reading is useless: with no constraint, the minute is worth nothing. Spreading payroll across dishes is the third route and the one I've most often seen end up in an abandoned spreadsheet. The logic appeals: if labor weighs as much as the ingredient, assign each dish the man-minutes it consumes and get a true total cost.

Alternative 3: per-dish costing with labor allocated, and why it's usually overkill

The problem is that kitchen payroll is a FIXED cost per shift, not a variable per dish: you pay the grill cook the same whether he sells 40 covers or 120, so any allocation depends on the volume you assume and shifts the moment it rains on a Tuesday. Diego F. Parra runs it the other way in the Masterestaurant method: payroll, rent and utilities are never loaded onto the dish, they go to the break-even point, and the dish answers only for its raw material and the station minute it consumes. It earns its keep in one place: negotiating a fixed-price banquet menu. Suppose tomorrow you raise the price of your best food cost dish by 12%, convinced it's the crown jewel. The percentage improves, of course, because the denominator grew. But that dish was the eleven-minute one, so its demand drops — say 20%, conservative for a double-digit increase — and bottleneck occupancy drops with it; so far it sounds fine.

What would happen if tomorrow you raised the price of your best-percentage dish?

The trouble shows up in the next link: the guests abandoning that dish don't leave the restaurant, they migrate to the cheap starter, which yields fewer dollars per ticket.

You improved a ratio, freed up burners nobody is using to sell something pricier, and lowered your average margin per table. That three-step chain is exactly what the percentage cannot show you, and it's why price gets decided after you rank the menu by money, never before. Three cases make staying with percentage food cost the right call, and saying so costs me less than watching an owner build a system he won't maintain. First: short menus, under twelve items, where you already know by heart what pays and what doesn't; there the matrix will confirm your hunch and charge you a week for it. Second: operations with no real bottleneck, kitchens that never saturate, because margin per minute needs a constraint to mean anything at all.

When NOT to switch methods and stay where you are?

Third, and the most common: when your POS doesn't give a trustworthy sales mix, because building menu engineering on dirty data produces worse decisions than not deciding.

Fix the sales register first, over 90 days. Meanwhile, use the percentage for what it genuinely does well, which is hunting purchasing drift week after week. ALTERNATIVE 1 — Classic menu engineering (Kasavana-Smith matrix, 1982). It crosses popularity against contribution margin in dollars and sorts every item into star, plowhorse, puzzle or dog. Cost: zero, if your POS exports a sales mix. Learning curve: two or three days to read the quadrants, a week to act on them without flinching. Who it is for: any menu of 15 to 60 references with at least 90 days of sales history. Its honest limit is that it treats all kitchen minutes as equal, when the grill on a Friday at nine is worth ten times the same grill on a Tuesday at noon.

The three real alternatives, with cost and learning curve

ALTERNATIVE 2 — Contribution margin per service hour (theory of constraints applied to a kitchen). Divide margin in dollars by the minutes the dish occupies the saturated station, then re-rank the menu by that figure. Cost: nothing in money, ten hours of stopwatch during real service. Curve: one week to measure, another to believe it. Who it is for: kitchens with an identifiable bottleneck —flat top, oven, fryer, pass— and a rising average check. It is the only alternative that answers the question that matters at peak: what do I push when no more tickets fit? ALTERNATIVE 3 — The Masterestaurant method. It takes the matrix from alternative 1, inserts the time divisor from alternative 2 and adds two layers nobody measures: estimated elasticity per dish (how many units drop against a 5% price increase) and contribution to average check through pairings. What comes out is not a list of good and bad dishes, it is a menu running order and a suggestive-selling script.

The three real alternatives, with cost and learning curve — in practice

Cost: 0 to 40 USD monthly depending on spreadsheet or POS module. Curve: two weeks. Who it is for: 25 dishes or more, or two or more locations. WHAT NONE OF THEM FIXES. No method here repairs a menu without identity. If your four appetizers compete with each other and the guest cannot tell what he came to eat, you may rank the numbers with surgical precision and you will still sell a little of everything. First you decide what the restaurant is; then you calculate. That order is not negotiable and it took me years to accept, because for a long stretch I believed a good spreadsheet could rescue a badly conceived menu. It cannot. THE ROLE OF THE PRINTED MENU. All this plate profitability work gets collected at the moment the guest decides, and that moment happens on paper. At Masterestaurant we ALWAYS recommend keeping the printed menu alongside the QR menu: print controls service pace, house narrative and suggestive selling, which is where margin per minute turns into money.

The three real alternatives, with cost and learning curve — key points

QR is a complement —delivery, accessibility, same-day price changes, click analytics—, never a replacement. A QR-only restaurant gives up 100% of menu design as a margin lever.