Food cost

How to find food cost variance per outlet, and why your POS report can't show it

A practical guide to theoretical vs actual food cost for restaurant chains: the formulas, the data you need, and how to find which ingredient and outlet is losing margin.

Paneer tikka · 1 platePaneer180 gCurd40 gSpice mix8 gOnion60 gFood cost31.4%of menu priceVariance +0.8%

Food cost % hides more than it shows

Most restaurants track food cost as one percentage: purchases divided by sales for the month. It's a useful headline, but it can't tell you why an outlet is at 36% when its sister outlet is at 31%. Was it a price increase, over-portioning, wastage, theft, or simply a different sales mix?

Your POS knows what you sold. Your purchase records know what you bought. Neither, on its own, knows what you should have used. That's the gap variance analysis fills.

Two numbers you need

Variance compares two measures of consumption for each ingredient over the same period.

Theoretical consumption = items sold × recipe quantity

If you sold 400 plates of paneer tikka and the recipe uses 180 g of paneer, theoretical paneer consumption is 72 kg.

Actual consumption = opening stock + purchases + transfers in − transfers out − closing stock

If the outlet opened with 20 kg, received 90 kg, sent nothing out and closed with 30 kg, actual consumption was 80 kg.

Unexplained variance = actual − theoretical − recorded wastage

With 3 kg of logged wastage, the unexplained gap is 80 − 72 − 3 = 5 kg. Valued at your purchase price, that's the money to go looking for.

What you need in place

  • Recipes for your top sellers. You don't need every item on day one. The items that make up most of your sales usually account for most ingredient consumption.
  • POS items mapped to recipes. Each menu item, including variants and add-ons, needs to point to a recipe.
  • Receipts recorded at the outlet. Purchases must be booked to the outlet that received them, on the day they arrived.
  • Regular counts. Weekly counts of high-value ingredients are more useful than one exhausting monthly count of everything.
  • Wastage logged with reasons. Otherwise genuine wastage and real losses look the same.

Reading the results

Start with the ingredients where variance is highest in value, not percentage. A 20% variance on coriander matters less than a 6% variance on paneer or chicken.

Then look for patterns. Variance concentrated in one outlet points to local practice: portioning, receiving or security. Variance across all outlets for one ingredient often points to the recipe itself being wrong, or a yield that's changed with a new vendor.

A negative variance (using less than the recipe says) isn't good news by default. It usually means portions are short, recipes are overstated, or sales are being recorded without the ingredient being used.

Making it routine

Variance is most useful weekly, while the week is still fresh in the manager's memory. Assign each outlet a target band, review the top five ingredients by value, and record what was found. Over a few months, the conversation shifts from "food cost is high" to "paneer portions at Outlet 3 are running 15 g over".

DeepRestaurantAI calculates theoretical consumption from your POS sales and recipes, actual consumption from receipts, transfers and counts, and shows the unexplained variance per outlet and ingredient. You can also try the basic calculation for a single dish with our food cost calculator.

See DeepRestaurantAI with your own outlets

We'll connect a sample of your POS data and walk you through sales, purchasing, food cost and DeepObserve.ai camera AI for your format.