Plate costing: the before and after of training your floor

Why does real food cost run 5.7 points above the recipe card?

It runs high because the recipe card measures intent while the P&L measures what actually walked out the door, and the dining room lives between those two documents.

That 84-seat Bogotá restaurant costed 61 recipes in March 2026, portion weights to the tenth of a gram, waste verified against invoices: 28.4 % theoretical. By June the P&L read 34.1 %, nearly two points above the 32.0 % median the National Restaurant Association reports for full service in its Restaurant Operations Data Abstract 2025. Those 5.7 points on 92,000 USD in quarterly sales are 5,244 USD gone without a single theft. The kitchen over-plated because servers asked for generous portions to avoid complaints, and the floor comped 213 desserts without logging them. Train the floor and the same location closes that gap below one point. Cards alone close roughly 1.5 points of the gap; with a trained floor the gap drops from 5.7 to under 1 point, and that is where all the cash sits.

Recipe cards alone versus recipe cards with a trained floor: the variance

On side A the document exists, the chef signs it and nobody else reads it: variance holds around 4.2 points because control depends on a line cook remembering 61 portion weights mid-rush. On side B the server knows 180 grams means 180 grams, calls the ticket with the right modifier and logs the comp in the system; the chef stops plating by eye. Across 92,000 USD in quarterly sales, those 4.2 points are 3,864 USD that stay in the business. Side B wins, and not narrowly: paper does not serve the food. Two hundred thirteen comped desserts in a quarter, at a plate cost of 2.40 USD, add up to 511 USD of food that never touched a ticket. That sounds minor until you look at the other side: every unlogged comp breaks theoretical inventory, so dessert variance read as kitchen waste and the chef spent three weeks chasing a ghost in the walk-in.

The 213 comps in one quarter: the case with numbers

That is the real cost, the poisoned diagnosis. When the same restaurant made comp logging mandatory in the POS, those 213 fell to 74 the following quarter and inventory reconciled at 99.2 %. Nobody banned the gesture toward a guest; the house simply learned who was giving it away and why. Clean data is worth more than the food you save. The mistake owners repeat most is raising prices to cover a 34 % food cost when the real deviation lives in portioning and comps. Diego F. Parra flips the order in Masterestaurant audits: close variance below one point, and only then argue price using menu engineering data. The reasoning is arithmetic. Raise the menu 6 % with 7.3 points of open variance and you multiply the chaos over a bigger base while losing traffic in a market where food-away-from-home inflation already hit +4.1 % in 2024, per USDA Economic Research Service.

Close the variance first, touch the price second

Tightening execution costs you nothing at the door. Raising prices on top of a broken system costs you covers. More than a large one, because it starts from a worse place: full-service restaurants with sales under 2 million USD reported 33.7 % food cost in 2024 against 31.0 % for those at 2 million or above, according to the National Restaurant Association. Those 2.7 points are not purchasing power alone; much of the spread is execution discipline, which is precisely what floor training buys. A location doing 92,000 USD a quarter that moves from 34.1 % to 29.5 % recovers 4,232 USD per quarter, close to 16,900 USD a year, without touching a supplier or the menu. An owner with one to five locations does not compete on volume against chains. The fight is over variance, and that one is winnable. A four-minute preshift covering the three critical dishes of the day beats a 60-page manual living in a drawer.

Automated preshift against the meeting nobody holds

Without the tool, preshift happens whenever the manager has time, meaning twice a week in high season, and the server learns portion weights by asking a coworker. With the Training Kit, the brief hits the phone before the shift, the simulator asks about the star dish margin, and gamification leaves a visible ranking: at the Bogotá restaurant, preshift adherence went from 31 % to 88 % in eight weeks. Adherence is not the number that matters. Portion-related send-backs dropped from 47 to 12 in the quarter, and defensive plating ended right there. Dish costing is a document; portion control is a behavior, and confusing the two costs thousands of dollars a quarter. A restaurant can hold 61 flawless cards, waste verified, prices updated against the +3.8 % inflation USDA reported for 2025, and still lose 5,244 USD because nobody translated paper into the hand that plates. The tension resolves once you see what each side measures: the card sets the standard and training produces compliance.

The paradox: the finest costing sheet drops zero points by itself

Without the first, you cannot tell how far you drift; without the second, you know exactly what you lose and cannot stop it. Pick one and you own half a system. In a business running single-digit margins, half a tool is not half a fix: it is an expensive report. Running one location with food cost inside the healthy 28-35 % band the National Restaurant Association marks? Start with pure recipe costing: cost the full menu, verify waste against invoices, and expect to close about 1.5 points. With two to five locations, or a gap between theoretical and actual above three points, cards alone stop paying and the money sits with your floor team: automated preshift, portion simulators, mandatory comp logging. Here is the working rule I apply: measure the gap for 30 days before buying anything. Under one point, your problem is price, not execution. Above three, train the floor this month and leave the menu exactly where it is.

What separates the two scenarios?

Variance first, price second. The mistake that repeats most among owners is raising prices to cover a 34% food cost when the real drift sits in portion and comps;

raise prices with 7.3 points of open variance and all you achieve is selling fewer equally uncontrolled plates. Diego F. Parra flips the order in Masterestaurant audits: close variance under one point, then argue price with menu engineering data, because margin earned by fixing execution costs you zero guests. Plate costing is a document; portion control is a behavior. A restaurant can hold 61 flawless recipe cards and still lose 5,244 USD in a quarter, because paper does not serve food. That is the tension the industry resolves badly: money goes into costing software and not one cent goes into helping the person on the floor understand the number. We resolve it by putting the figure inside the shift — preshift, simulator, scoreboard — instead of inside a file only the owner opens.

What separates the two scenarios — in practice?

Trained suggestive selling moves food cost without touching a recipe.

Once a server can describe the high-margin dish — the one leaving 71% contribution instead of the house favorite's 54% — the sales mix shifts on its own and weighted food cost falls even though no recipe card changed. In this comparison that single effect explained 1.8 of the 4.9 points recovered. Cash flow gets fixed before EBITDA. A location running 4.8% EBITDA and 68.4% prime cost survives on weekend cash; when prime cost drops to 59.1%, 9.3 points of sales stop leaving every week and the break-even point slides down by nearly 8,000 USD a month. That is what turns a business that survives into one that can absorb a bad month. On menus: if your menu moved to QR, keep the PHYSICAL menu. QR gives you instant price updates, analytics on what guests browse and accessibility for delivery, but the physical menu is control of the experience — service pacing, menu narrative, the suggestive sell your server just rehearsed.

What separates the two scenarios — key points

At Masterestaurant the verdict is BOTH, each with its role; the restaurant that scrapped its physical menu lost 9% of average check in the first month because guests ordered what they already knew.