Four lines decide the month. Everything else is commentary.

Every ratio below has net sales in the denominator, so an error here silently distorts all four. Net sales means gross receipts less sales tax collected, less comps and voids, less discounts actually given. Sales tax is money held for the state and was never revenue; leaving it in flatters the denominator and makes every cost line look better than it is.

Two further deductions get missed. Delivery-platform commission is not a marketing expense in any useful sense — it is a reduction in what you were paid for the food, so many operators track platform sales net of commission on a separate line rather than mixing them with dine-in at menu price. Gift certificate sales are a liability until redeemed, not sales on the day of purchase.

Fix the denominator once, define it in writing, and never change it mid-year. A definition change halfway through creates a step in every trend line that will be mistaken for an operational event.

Food cost is not the sum of what you bought. It is what you consumed, and the difference between the two is inventory. The calculation only works if the count is real.

Three things break this. Invoices dated in the period for goods that arrived after the count — the purchase is in, the stock is not, and cost looks high. Transfers between kitchen and bar recorded on neither side, which moves the error from one ratio to the other without changing the total. And a count taken at a different time of day than last time: a Friday-evening count and a Monday-morning count are measuring different restaurants.

Merging beverage into food hides the two most informative signals in the whole statement. Beverage cost percentage is normally far lower than food and far more stable, which means a movement in it is a strong signal rather than ordinary variance.

Run at least the split between liquor, beer, and wine. Wine cost percentage drifts when the list turns over; draught beer cost drifts with line loss and keg yield; spirit cost drifts with pour discipline. These are three different problems with three different fixes, and one blended number tells you which of them you have — none.

The bar count has its own trap: partial bottles. Weighing is more accurate than eyeballing tenths, but whichever method is chosen has to stay chosen. Switching from eyeball to scale mid-year produces an apparent cost swing that no operational change caused.

Wages means hours worked at the rate paid, in the period the work happened. The two habitual distortions are pay-period boundaries that do not line up with the accounting period, and salaried managers loaded entirely into one week.

Accrue across the boundary rather than assigning a whole pay run to whichever period it settled in. A restaurant whose four-week accounting period contains three pay runs one time and two the next has a labour ratio that swings on payroll administration rather than on anything a manager did.

Track hours as well as dollars. Dollars alone cannot separate a rate increase from an hours increase, and those need opposite responses: one is a pricing question, the other a scheduling question.

Burden is everything paid because someone is employed, over and above their wage: employer payroll taxes, unemployment insurance, workers' compensation premium, and whatever benefits are provided. It is real money leaving the account, and a prime cost calculated without it is not prime cost.

Because the first three are driven by wages, burden usually tracks the wage line closely — which is exactly why it gets omitted and why omitting it understates the true labour position by a consistent margin every single period.

Prime cost is a ratio of two moving quantities, so it can move without either underlying cost changing. Reading a movement correctly means asking which side moved first.

1. Mix, not price. A week when guests order more of the low-margin item and less of the high-margin one raises food cost percentage while every recipe cost, every purchase price, and every menu price is unchanged. Check item mix before checking invoices.
2. Yield, not theft. A supplier switch that looks like a saving per case can lose more in trim than it saved in price. Compare cost per portion served, not cost per unit purchased.
3. Volume, not efficiency. Labour percentage falls in a busy period because the fixed part of the schedule spreads over more sales. That is arithmetic, not improvement, and it reverses.
4. Timing, not consumption. A delivery received before the count and invoiced after it moves cost between periods. Two periods read together will show the swing and the reversal.
5. Waste, which is the residual. Only after the four above have been excluded is a movement genuinely about spoilage, over-portioning, or loss — and by then you know where to look.

The number is a ratio of costs to sales. Anything absent from both terms is invisible to it, and treating it as a summary of the business produces bad decisions with confident arithmetic behind them.

It cannot tell you whether the restaurant is profitable, because rent, utilities, insurance, marketing, repairs, and debt service sit below it. A 60% prime cost under a rent that is too high still loses money every month.

It cannot tell you whether a menu item earns its place, because percentages and dollars disagree. An item with a high cost percentage and a high absolute margin can be worth far more than a cheap item with a flattering ratio; the register counts dollars, not percentages.

It cannot tell you whether cutting worked. Cutting hours lowers the ratio immediately and may lower sales next quarter, by which time the two changes look unrelated in the data.

And it cannot be compared to somebody else's. Published averages blend service models, geographies, lease structures, and definitions that differ from yours — particularly around whether burden and platform commission are included at all. Your own trailing periods are the only honest comparison you have.