Plate costing mistakes versus the method that actually holds your margin

The 30% average food cost hiding two dishes at 44%

A steakhouse in Guadalajara billed 1.9 million pesos a month and closed every quarter with the checking account at zero, while its average food cost read 30% because nobody had gone from the aggregate down to the dish. Taking apart a 46-item menu surfaced two skirt-steak plates with real costs of 41% and 44%, plus a fish of the day that climbed to 39% once fillet waste was counted: those three moved 34% of tickets. The arithmetic behind the average was not wrong, it was BLIND. It never tells you which dish eats the cash, and the server, who actually decides the sales mix, always pushes whatever sells easiest. With 90% of full-service operators raising prices in 2024 and 60% pulling dishes off the menu (National Restaurant Association 2024), the menu gets trimmed blindfolded whenever costing lives inside a single division. It means calculating the cost of a portion from what the cut actually yields after trim loss, not from the price on the supplier invoice.

What standard recipe costing really is?

That is the whole difference. You buy loin at 320 pesos a kilo, but if yield after cleaning fat and sinew runs 68%, your usable kilo costs 470 pesos, 47% above what the previous manager's spreadsheet claims.

Multiply that by a portion the cook plates at 240 grams when the spec says 200, and the dish you thought sat at 31% lives at 41%. A standard recipe locks three numbers nobody argues with afterward: gram weight, yield and waste, all measured in your kitchen, with your supplier and your knife. Recost every 90 days, or sooner when an input moves more than 8%. Coffee is the obvious warning: Brazil supplies roughly 38% of the world's beans (Bellwether Coffee), so one frost there rewrites your entire beverage menu. Standard recipe costing on a spreadsheet stops working past roughly 40 menu items, or the moment you run more than one location.

When the original approach falls short?

Inventory drift is the giveaway: when the month-end physical count differs from theoretical by more than 3% two months running, the sheet no longer describes what happens in the kitchen.

The other symptom is a calendar one, and it is the worst of all: once your last recost is more than 120 days old, you are pricing with costs from another quarter. With 99% of operators reporting higher labor spending in 2024 (TouchBistro 2024) and restaurant price inflation that peaked at 8.8% in March 2023 (National Restaurant Association), a static file expires before you open it. There is also a human ceiling. Nobody updates 46 recipe cards by hand every quarter, and sustaining that for a full year is the exception, never the rule. Grouping the menu into six or eight families —grill, pasta, cold starters, desserts— and costing one representative recipe per family, adjusting the rest by protein difference, takes about 6 hours to set up and any manager grasps it in an afternoon.

Option 1 — Costing by dish family, a bridge for one quarter

Who it fits: menus of 20 to 35 items built on heavily repeated ingredients, and teams that have never costed anything and need a quick win that builds the habit. Switching effort is close to zero, since it touches neither software nor inventory. The downside bites hard: it leaves out the odd dishes, which tend to be exactly the ones bleeding. At that Guadalajara steakhouse, neither skirt steak at 41% and 44% would have surfaced this way, because the grill representative was a clean 29% plate. I would use it as a bridge for one quarter, with an expiry date written on the calendar, never as a permanent system. Toast, Lightspeed and Fudo ship recipes with automatic inventory depletion for somewhere between 60 and 220 dollars a month depending on modules and terminal count, and their real virtue is that cost updates itself the moment a purchase invoice changes. Who it fits: operations with two or more locations, a formal storeroom and someone entering receipts daily.

Option 2 — The costing module inside your point of sale

That last point is the trap. The module does not cost money, it costs DISCIPLINE: with nobody logging waste and inter-store transfers, the system produces a pretty theoretical figure that never matches the physical count, and you end up trusting it less than your old sheet. Real setup runs 30 to 50 hours of loading recipe cards plus a two-month curve before the team respects it. Since 75% of traffic now happens off premise (Circana), automatic depletion by channel is what justifies the spend. Menu engineering sorts every dish by popularity and contribution margin, and it pays off only once costing is already measured: without a real cost per portion, the matrix just organizes your mistakes. Cost: nothing beyond time, roughly 4 hours per quarterly review. Who it fits: owners with living recipe cards who now want to decide what stays on the menu. Measure cost first, set price second, redesign the menu only then; the reverse order is what I see fail most often.

Option 3 — Menu engineering on top of costing, not instead of it

And here I got it wrong for years: I would raise the price of the expensive dish, the guest noticed, the mix drifted on its own toward the cheap badly-costed plate, and margin ended up flat or worse. Add waste control —foodservice generated 17.9% of all U.S. food surplus in 2024, and full-service restaurants contribute over 43% of that surplus, per ReFED 2024— and the matrix finally decides on evidence. Inside the Masterestaurant framework, 32% food cost per dish is the MAXIMUM you tolerate, not the goal, and no plate on the menu should be born there. Diego F. Parra hammers one point that changes the conversation with your accountant: payroll, rent and utilities do NOT load onto the plate, they belong to break-even, because spreading them per unit produces prices nobody buys while hiding the real problem, which is volume or mix. What would happen if you raised the price of your two 44% dishes by 8% to bring them to 40%?

Diego F. Parra: why 32% is a ceiling, never a target

If those plates move a third of your tickets, elasticity takes part of the volume, the mix migrates to the cheap badly-costed dish, and you end up behind where you started. The way out runs backward: cut the portion back to spec, renegotiate the cut, swap the expensive garnish for a seasonal one, and touch the price last. Cash improves without the guest noticing. Running one location with fewer than 20 menu items, a spreadsheet updated within the last 90 days and inventory drift under 3%? Stay put: migrating to a point-of-sale module will cost you 60 to 220 dollars monthly and some 40 hours of setup to solve a problem you do not have. The uncomfortable honesty is that the most expensive method never wins; the one your team sustains all year does. Waiting also makes sense during high season, or within 60 days of an opening, because changing systems at the peak guarantees half-loaded recipe cards and a useless theoretical number.

When NOT to switch methods?

And when the real problem is volume —Mexico, where restaurants account for 12.2% of the country's economic units (INEGI–CANIRAC 2024), has markets like that— no costing method fixes it for you.

Open your menu this week, take the five best sellers and weigh them raw on the scale. That number tells you whether you need a new method or just a smaller spoon. ALTERNATIVE 1 — Costing by dish family. You group the menu into six or eight families (grill, pasta, cold starters, desserts), cost one representative recipe per family and adjust the rest by protein difference. Cost: roughly 6 hours to build. Learning curve: low, any manager gets it in an afternoon. Who it fits: menus of 20 to 35 dishes with heavily repeated ingredients. Verdict: a decent bridge for one quarter, but it leaves out the odd dishes, which tend to be exactly the ones bleeding.

Four honest alternatives to standard recipe costing

Use it only when the team has never costed anything and needs a quick win that builds the habit. ALTERNATIVE 2 — The costing module inside your POS. Toast, Lightspeed and Fudo all ship recipes with automatic inventory depletion. Cost: 60 to 220 USD monthly depending on locations and modules. Curve: medium, two to three weeks to load the recipe cards properly. Who it fits: two-location operations upward with disciplined inventory. Verdict: excellent to SUSTAIN costing, terrible to originate it, because software inherits whatever garbage you feed it; enter invented yields and the system will hand you beautiful, false reports for years. ALTERNATIVE 3 — Menu engineering costing (popularity × margin matrix). Instead of chasing the cost of all 45 dishes, you measure contribution margin in currency and units sold, sort into stars, plowhorses, puzzles and dogs, and work only the quadrant that moves cash. Cost: 4 analysis hours a month using data your POS already holds.

Four honest alternatives to standard recipe costing — in practice

Curve: medium-high, since it demands accepting that margin is defended in currency, never in percentage. Who it fits: owners with basic costing already in place. Verdict: the best profitability lever per hour invested, though it does NOT replace costing; it needs reliable costs to compute that margin. ALTERNATIVE 4 — An outside consultant who costs the whole menu. A third party builds the recipe cards, measures yields and hands you a closed manual. Cost: 800 to 3,500 USD per menu depending on city and size. Curve: zero for you, and that is precisely the risk. Who it fits: openings, full menu changes, or owners who need an auditable document for a partner or a bank. Verdict: it compresses months of work and usually pays back within a quarter, yet if the internal team never learns to recost, eighteen months later you own an obsolete manual and a second invoice.

Four honest alternatives to standard recipe costing — key points

THE LIMIT OF THE ORIGINAL OPTION. The full standard recipe, which Masterestaurant recommends as the base, carries a real price: 18 to 25 hours of build time, a scale, and a cook willing to weigh waste three times per cut. In a coffee bar running six SKUs that effort is disproportionate and a well-built spreadsheet serves you fine. The method earns its place once the menu passes 20 dishes, once protein by weight enters the kitchen, or once a second sales channel with a different commission shows up. ON CAPEX AND OPEX, WHERE MOST OWNERS GET LOST. The hood, the oven and the buildout are CapEx: recovered through depreciation and break-even, never inside the cost of a plate. Oil, gas and film wrap are variable OpEx and they do belong, though almost nobody loads them. Pushing rent into the plate is the most common capital leak in a badly costed menu: it inflates the price, kills competitiveness, and leaves the owner convinced his signature dish is unprofitable when the unprofitable thing is his lease.