The labor cost formula is not the hard part:
Labor cost ÷ revenue × 100 = labor cost percentage.
That is all of it. You could teach it in ten seconds, and every operator who tracks the number already knows it. Yet two shops with identical books routinely produce figures four points apart, which means the formula is not where the difficulty lives. The inputs are.
Here are the three places it goes wrong, in the order they cost you.
One: the top half is wages
Covered at length in what is labor cost, and worth repeating because it is far and away the most common error. Wages are not labor cost. Fully loaded — employer payroll taxes, workers' comp, benefits, paid time off — usually runs 15–25% above the timesheet total.
The reason this one is expensive rather than merely wrong: every published benchmark is fully loaded. Comparing your wage-only 26% against a fully loaded 30% band tells you that you are comfortably fine, when the honest comparison is 32% against 30% and you have a problem that started three months ago.
Two: the bottom half is the wrong sales
Use the same revenue figure your profit and loss uses — net of comps, discounts, voids and refunds.
This sounds pedantic until you run a promotion. Count discounted covers at menu price and your labor percentage improves without a single hour changing, because you have inflated the denominator. The month looks like an efficiency win. It was a discount.
The same trap catches gift cards, which should count as revenue when redeemed rather than when sold, and delivery platforms, where the honest figure is what actually reaches your account rather than the menu total the customer saw.
None of these are large individually. Together they are comfortably enough to move the number a couple of points, in the flattering direction, every time.
Three: the two halves cover different weeks
This is the one nobody warns you about. Payroll runs on its own cycle — biweekly for most people — and sales run on yours. Divide a fortnight of labor by a month of sales and you get a number that is not wrong so much as meaningless.
It gets worse at the edges. A pay period that happens to contain three weekends against a sales period containing two will produce a spike you will spend an afternoon investigating, and the answer will be the calendar.
Fix it by picking a period and forcing both halves into it, accruing the partial pay period rather than letting the payroll dates decide your reporting dates. It is slightly more work once and removes a whole category of phantom problem forever.
The one judgement call
Salaried managers. Do they belong in labor cost?
There is no universally right answer — some operators hold management out and track it separately as overhead, others load everything in. Both are defensible. What is not defensible is switching, or doing it differently in two different reports, because the entire value of this metric is the trend and a definition change breaks the trend without announcing itself.
Write down which you chose. Put it next to the number.
Then stop calculating it by hand
Once the definitions are settled the arithmetic is mechanical, and mechanical arithmetic done monthly by a person is arithmetic that eventually gets skipped. The restaurant labor cost calculator loads the payroll taxes for you and puts the result against the 25–35% band — and breaks out how many of those hours went on answering the phone, which is the only part of the total you can cut without cutting a person.