Plate costing: the before and after of knowing what every menu line earns

The four inputs you need before opening the spreadsheet

Before costing a single dish, gather four things: the last 30 days of purchase invoices, the full menu list with pre-tax selling prices, recipes with genuinely weighed gram amounts, and the sales-by-dish report from the last quarter. The deliverable at this stage is one folder holding those four files, and the checkpoint telling you whether to move on has two parts: fewer than 35 menu lines, and gram weights present on at least 80% of recipes. When you buy the same input from three suppliers, drop the latest-invoice price and use a volume-weighted average, because a gap of USD 1.40 per kilo on a product moving 60 kilos a month is USD 84 that your spec sheet will quietly hide all year long. Each dish spec sheet carries four columns and nothing else: ingredient, served gram weight, supplier price per kilo, and trim plus cooking loss percentage.

One spec sheet per dish, with weighed grams and yield loss charged where it belongs

A tenderloin bought at USD 14/kg with 28% loss costs USD 19.44 per kilo SERVED, and that is the only figure entitled to enter the costing. When you finish, add the raw material cost of every sheet multiplied by units sold that month and compare it against what you actually spent on purchases: it should explain between 92% and 100%. Explaining 84% means 16% of purchases no dish is claiming, and that is where the leak lives, almost always in waste, in friends-and-family covers, or in the corner where nobody weighs anything. Food cost percentage comes from dividing raw material cost by the pre-tax selling price and multiplying by one hundred, while contribution margin is plain subtraction in cash, and that subtraction is what decides. A dish at 34% food cost leaving USD 16 clean per unit is worth more to your till than one at 22% leaving USD 5, because nobody pays rent with percentages.

Food cost and contribution margin: why cash outranks percentage

Sort the menu from highest to lowest margin in USD, never by percentage. The healthy band runs 28% to 32%, with 32% as the CEILING, and no dish above it stays on the menu without a written correction plan carrying a date. Bear in mind that food and beverage cost absorbs close to 33% of industry sales according to the National Restaurant Association 2026. This is where costing stops being accounting and becomes the script for your content calendar. Join the margin table to the sales-by-dish report and four quadrants appear, straight from the classic menu engineering of Kasavana and Smith: high rotation with high margin are the stars, high rotation with low margin are the traps to redesign, low rotation with high margin are the ones needing camera, and the rest goes. The deliverable is a named list of eight dishes, and the checkpoint is that those eight hold at least 45% of projected margin.

Crossing margin with popularity: from the costed table to the eight dishes your camera chases

It adds up once you remember that 20% of dishes usually carries 70% of sales. Diego F. Parra insists across Masterestaurant engagements that this list, not the chef's taste, approves the shooting brief. Every Reel, TikTok and carousel of the quarter gets built on one of the eight hero dishes, carrying a call to action for that specific plate —name and price on screen— rather than a generic invitation to visit the restaurant. The deliverable is a 12-week calendar with an assigned dish per piece and a unit sales goal per piece. The checkpoint stings, which is exactly why it works: at quarter close, contribution margin per ticket must rise at least 15% against the prior quarter WITH FLAT COVERS, because rising covers leave you unable to tell what moved what. The octopus case sums it up whole: 180,000 views, 340 plates sold, 46% food cost, worst month of the year.

Rewrite the 12-week calendar around those eight, then change the metric

Reach without margin is paid advertising for losses. Five failures concentrate almost all the damage, and none of them is arithmetic. First, costing on purchased weight instead of served weight, which turns a declared 27% into a real 38%. Second, loading payroll, rent and utilities inside the plate, when those costs get covered at the break-even of the whole business. Third, sorting by percentage rather than cash. Fourth, keeping the chef's favourite dish on the content list, which is precisely the decision the margin cross-check exists to block. And fifth, the most expensive of them all, costing once and filing the document away: with food-away-from-home inflation near 5% year over year according to the Bureau of Labor Statistics 2025, a menu untouched for a year sheds two or three margin points in silence. On the first of each month, refresh the price of your 15 core inputs, which usually explain 80% of spend, let the template recalculate, and review the alert tray for every input that climbed above 8%.

The twenty-minute monthly cycle and the checklist that closes everything

To know EVERYTHING landed, verify six figures: spec sheets covering 92% or more of purchase spend, none above 32% food cost, a gap between theoretical food cost and real inventory below three points, eight hero dishes holding 45% of projected margin, a 12-week calendar fully assigned, and margin per ticket growing 15% on flat covers. Meet all six and your menu is genuinely costed; fail the third one and the problem no longer sits in the spreadsheet but in the pass, on the scale, or at the back door. The first difference is about FILTER, not arithmetic. A restaurant without costing owns a content calendar; a costed one owns the criteria to fill it. Once you know the lasagna leaves USD 11.80 per unit and the ceviche USD 4.20, choosing Tuesday's shot stops being an aesthetic call. And if that answer stings because the ceviche photographs better, that is precisely the owner's job.

The four differences that actually move the till

Yield loss is the second. Theoretical food cost built from recipe ingredients ignores that a 5-kilo tenderloin yields 3.6 kilos of sellable portions, with 28% lost to fat, bone and cooking. Costing on purchased weight instead of served weight is the correction I have had to make most often on menus already in print, and it is what separates a declared 27% food cost from a real 38%. Third comes cadence. A menu costed in January is fiction by June if oil climbed 22% and nobody logged it. Plate costing is not a project with a delivery date, it is a twenty-minute monthly routine sitting on a template that already carries the formulas. The fourth difference is the uncomfortable one: what you do with the dish that does not pay. Almost everybody raises its price. The right play is usually different, and it is redesigning portion or garnish so the guest perceives the same value at a different cost.

The four differences that actually move the till — in practice

Price is the last lever, never the first, because a neighbourhood restaurant burns through its elasticity fast and you will not notice until average check drops.