Food cost is usually the second-largest line on a restaurant’s P&L, and it is the one that drifts fastest. Before you change anything, know where you stand: most full-service restaurants run a food cost of roughly 28-35% of food sales, and many operators use about one-third of the menu price as a working rule of thumb. Those are general industry ranges, not targets handed down from on high — your own break-even math decides what you can actually afford. If you are not sure how the percentage is built, read food cost percentage explained first, then come back here to lower it.
The single most useful diagnostic is the gap between theoretical food cost (what your recipes say you should have spent) and actual food cost (what a physical inventory count says you did spend). That gap is your variance, and it is where almost all the recoverable money hides. A steady 3-4 point variance is a signal to look hard at portioning, receiving and waste. Run your plate math through the food cost calculator so you have a theoretical baseline to compare against.
Over-portioning is a slow leak that never shows up as a dramatic event — it just quietly adds points month after month. It is also the fastest thing to fix because it costs almost nothing.
Money is won or lost before food ever reaches the line. Two failures dominate: paying too much, and accepting less than you paid for.
| Lever | What it looks like in practice |
|---|---|
| Buy to par | Order to a set par level, not to a nervous guess. Overbuying perishables is just future spoilage. |
| Compare and negotiate | Price a few high-volume items across vendors quarterly, then negotiate. Loyalty is fine; blind loyalty is expensive. |
| Check every invoice | Watch for price creep line by line. A supplier’s quiet 4% increase on ten items is a real margin hit. |
| Weigh receiving | Verify case weights and counts at the door. You cannot claim credit for short or spoiled deliveries you already signed for. |
| Consolidate orders | Fewer, fuller orders can unlock volume pricing and cut delivery minimums. |
Waste is the difference between the tidy recipe card and the messy walk-in. It comes from spoilage, over-prep, trim, spillage, comps and theft — and it is the biggest single reason actual cost beats theoretical.
A simple two-bin waste log — one for spoilage, one for prep and cooking errors — sitting on the prep table for two weeks will usually pay for a month of improvements just by making the problem visible.
Rotate stock first-in-first-out, date everything on receipt, and repurpose usable trim into stocks, staff meals and specials rather than the bin. Waste reduction is a big enough topic on its own that it earns a dedicated playbook — see how to reduce food waste in a restaurant for the full routine.
Not every dish deserves its spot. Menu engineering sorts items by margin and popularity so you can promote the winners, rework the losers, and quietly retire the dishes that lose money on every ticket. Two moves do most of the work:
Cooking more from scratch is another lever: made-in-house items usually cost less than their ready-made equivalents, though they trade food cost for labor, so weigh both sides before you commit.
The operators who hold food cost down do not do a heroic cleanup once a quarter; they run a short, boring weekly loop. Count key inventory weekly rather than monthly — a weekly number catches a problem while there is still time to fix the month. Keep beverage cost on its own line, because liquor, beer and wine run very different percentages and blending them hides which side is leaking.
Finally, remember that food cost is only half of your prime cost. Chasing food cost down to 24% by starving portions can cost you repeat guests and does nothing if labor is out of control. Track the two together, and sanity-check any big change against your break-even analysis so a cost cut does not accidentally hurt volume. For deeper questions on specific ingredients, formats or violations, the answers library covers the common cases.
This is educational guidance, not financial advice — run your own numbers before acting on any benchmark.