Food cost variance: where it actually hides

Invoices get the blame because they are the only piece of paper in the building. Prices did go up, that part is true and visible, and it is also not variance. A price increase you can see on an invoice is a cost change. Variance is the part you cannot see.

Theoretical versus actual is the entire concept. Your recipes and your sales mix say the food should have cost a certain amount. Your count says it cost something else. The distance between them is the number worth chasing, and you cannot calculate it without counting.

Where does the variance usually come from?

Portioning. The largest and least dramatic source. A protein spec that says six ounces and a cook who plates seven. Nobody is doing anything wrong on purpose. The scale is across the kitchen, the ticket printer is going, and seven ounces looks right. Multiply the overage by the number of that item you sell and it becomes the whole variance on its own.

Waste. Split between production waste, spoilage, and remakes. Production waste is trim and prep error. Spoilage is ordering and rotation. Remakes are the ones nobody logs, because a remade plate feels like service recovery rather than an expense.

Receiving. Short deliveries, substitutions at a higher price, credits never issued, and cases signed for by whoever happened to be near the back door. If the person receiving is not checking weight and count against the invoice, you are paying for whatever was written down.

Theft. Real and usually last on the list, not first. It is also the one operators jump to emotionally, which is a shame, because the accusation costs more than the food when the actual cause was a portion spec nobody enforced.

There is a fifth that is not really variance but shows up in the same number: menu mix. If your low margin item sells better than it used to, your food cost rises with nobody making a single mistake.

What is a point of food cost actually worth?

More than most operators think, because of what sits underneath it.

The National Restaurant Association describes a typical independent restaurant, before the pandemic, as spending about 33 cents of every sales dollar on food and about 33 cents on labor, with everything else taking roughly 29% and leaving a pre-tax margin near 5%.

That 5% is the whole reason variance matters. At a five cent margin, a dollar of profit requires about twenty dollars of sales behind it. Every dollar of food that leaves the building without a guest paying for it has to be replaced by twenty dollars of new revenue. Nothing in marketing is that expensive.

The input prices moved too. The same association estimates average wholesale food prices are up 35% since before the pandemic, and total expenses for an average restaurant jumped 36% between 2019 and 2026. A recipe costed in 2019 and never recosted is not a recipe, it is a memory.

Here is the sizing arithmetic on a restaurant at $1.2M in annual sales. One point of food cost is $12,000 a year. Two points is $24,000. That is the range most operators find on their first honest count, and it is larger than most independent marketing budgets.

Why does the number bounce so much month to month?

Because most food cost figures are built on an inventory count that either did not happen or happened badly.

Purchases minus ending inventory equals usage. If nobody counted the walk in, the software substituted an assumption. The result is arithmetic, not measurement, and it will swing several points based on when invoices landed relative to your period close.

Three habits fix the bouncing:

• Count on the same day and time every period, ideally after close and before the next delivery
• Count the same way every time, in the same order, with the same units, ideally by the same person
• Cut off invoices consistently. An invoice booked one day late moves your food cost and nothing physical changed

Once the number is stable, a two point move means something. Until then you are managing static.

How do you find which item is doing it?

Stop looking at total food cost and start looking at the two or three categories that carry most of your spend, usually protein and one other.

Take one high volume item and run it end to end. Pull the spec, weigh five plates as they leave the line during a real service, and compare. Do it quietly, not as an inspection.

Then work the arithmetic on what you found. Here is the version I have run more times than any other.

The spec says six ounces. The five plates you weighed averaged seven. Your protein costs $5.60 a pound, which is 35 cents an ounce. One ounce over per plate is 35 cents of food that nobody ordered.

You sell 180 of that item a week. That is $63 a week and $3,276 a year, from one ounce.

Now convert it to what it takes to replace. At a 5% margin, $3,276 of lost profit needs roughly $65,520 in additional sales to make back. One ounce, one item, one line cook who was never given a scale.

Then check the same item at receiving for two weeks. Weigh what arrives against what the invoice says.

That single item will usually explain more of your variance than a full inventory audit, and it takes a few hours spread across two weeks. Find the one item, fix the one item, then move to the next one. A kitchen wide crackdown produces two weeks of compliance and a return to baseline.

If you run more than one location, the fastest diagnostic you own is comparing them. Identical menu, identical vendor, different variance means the difference is behavior, not cost. That comparison is the first thing to run when two locations perform differently, and it will point at a person or a habit within a day.

Is your food cost percentage telling you the truth?

Sometimes not, and there are two specific ways it lies right now.

The denominator moved. Food cost as a share of sales falls when you raise menu prices, even if the kitchen wastes exactly what it wasted before. The Association's July 2026 outlook has industry sales growing 4.3% in nominal terms and only 0.8% after inflation, with much of that growth coming from higher menu prices rather than more traffic. If your food cost percentage improved in a year when your prices rose, you have not proven anything about the kitchen. Check the dollars per cover, not just the percentage.

The benchmark you are comparing against may not exist. People will quote you an industry standard food cost for your segment. Ask where it came from. The Association's Restaurant Operations Data Abstract does break out food cost by check average, sales volume, community size and location, but those cuts sit behind a paid publication, and the Association attaches its own caveat to the whole thing: the data is "not intended to represent standards or goals for individual restaurants" and is meant as a management tool for gauging performance.

Your theoretical cost is the only benchmark that is actually about your restaurant. It is built from your recipes, your specs, and your own sales mix. Nobody else's median can tell you whether the seven ounce plate happened.

Why do the fixes stop holding?

Because the fix is usually written down rather than built in.

You publish the spec. You retrain the line. Variance improves for three weeks. Then a new cook starts, the trainer is on vacation, service gets busy, and the spec quietly reverts. This is not a discipline failure, it is the ordinary way written procedures get ignored by about week two when nothing in the shift makes following them easier than not.

The controls that survive are physical. Portion scoops instead of a written ounce. A scale on the line instead of a spec on the wall. Pre portioned proteins for the item that keeps drifting. A waste container that has to be logged before it is emptied.

Notice what pre portioned product actually is. It is a decision to spend food dollars to buy back labor and consistency. The purchase price per pound goes up and the variance goes to near zero. Whether that trade is worth making is a prime cost question, not a food cost question, and it is one of the few places where paying more per case is obviously correct.

The other half is the first two weeks of any new hire. Portioning habits form immediately and are almost impossible to correct later, which makes how you onboard a new kitchen hire in their first two weeks a food cost decision and not only a training one.

Where does variance hunting stop paying?

Three places.

When the movement is price, not behavior. With wholesale food up 35% since 2019, a rising food cost percentage can be entirely explained by the market. Run theoretical against actual before you go looking for a cause inside the building. If theoretical rose by the same amount actual did, your kitchen is fine and your menu prices are the conversation.

When the movement is menu mix. Sell more of the low margin item and your blended food cost rises with nobody making an error. Check item level mix before you check the line.

When the count is not reliable enough to support the precision. A variance calculated from an assumed ending inventory will produce a confident number describing nothing. If you cannot count consistently yet, the honest move is to fix counting and postpone the analysis, not to run the analysis on bad inputs.

What to do this week

Count inventory properly one time, on a fixed day, after close. Even a partial count of your top spend items beats an estimate.

Calculate theoretical cost for your five highest volume menu items from the actual recipes, not from what the recipe used to be. Recipes drift, and half the time the spec on file is not what the kitchen makes.

Weigh five plates of your highest volume protein item during a real service. Write down what you find and do nothing else yet.

Run the ounce arithmetic above on whatever gap you found, using your own per pound cost and your own weekly item count. Then multiply the annual figure by twenty to see how much revenue it would take to replace. That second number is the one that gets a decision made.

Then pick the single largest gap and fix only that one, with a physical control rather than a memo. Recheck in two weeks.

Questions about your food cost?

Email me at eric@seod.com with your last three periods of food cost percentage and your top five selling items. I will tell you whether you have a variance problem or a menu mix problem, which are different diseases with the same symptom.

Sixteen years running multi unit restaurants and more than $54M in annual P&L, most of it spent finding out that the answer was a portion scoop.

Otherwise here is more on running the operation.