Theoretical food cost is what your food should have cost: every dish sold, multiplied by its recipe cost, at current ingredient prices. Actual food cost is what it did cost: opening inventory plus purchases minus closing inventory, over the same period's sales. The difference is your variance — and the industry bands are blunt: under 2 points is well-managed, 2–3 is acceptable, 3–5 is common but improvable, and anything over 5 points signals a systemic problem (meez, Supy).
Variance is the honest number because it doesn't care about your intentions. A 4-point gap on $1M of food sales is $40,000 leaving through some combination of waste, portioning drift, unrecorded comps, theft, and recipes costed at last spring's prices. Your profit margin doesn't distinguish between those causes — it just shrinks.
How to calculate both numbers
The formulas are trivial; the inputs are the work. Theoretical requires every menu item mapped to a costed recipe at today's supplier prices — a recipe costed in March is fiction by July. Actual requires clean inventory counts at both ends of the period. Get either wrong and the variance number turns into noise.
Check a dish's cost in 30 seconds
Ingredient cost, ex-tax price, and your target ratio — free tool, no sign-up.
The prerequisite nobody skips successfully
You can't measure variance on items you never costed. At Chat Noir, my club in Geneva, 645 of the 649 items we sell are mapped to a costed recipe — drinks and plates. Not because we're obsessive, but because the four uncosted items are exactly where a variance calculation goes blind. Whatever tool you use, coverage comes first: an 80%-costed menu gives you a variance number with a 20% hole in it.
Reading the gap: what each point usually means
- Stale recipe costs — supplier prices moved, the fiche didn't; you're measuring against fiction
- Portioning drift — the 30g garnish that became 45g on busy nights
- Waste that never hits a log — trim, over-prep, the fryer batch that died in the pass
- Comps and staff meals rung as zero instead of tracked
- Receiving gaps — invoiced quantities that never fully arrived
Work the list top-down: stale costs first, because they're the cheapest to fix and they contaminate the measurement itself. Only when the theoretical side is current does the remaining gap point at operations. That's also why supplier price drift deserves its own routine.
The cadence that makes it useful
Monthly variance is an autopsy; weekly variance is a diagnosis. The venues that hold under 2 points do the same boring thing: count the high-value items weekly, keep recipe costs synced to invoices, and put the variance number in front of the chef and bar manager — the people whose decisions move it — not just in the owner's spreadsheet.
