Actual vs theoretical food cost compares what your recipes say you should have used with what the counts say you used. The gap is food cost variance, also called inventory variance, and it is the most useful number in the kitchen: over-portioning, waste, theft and bad receiving cost you money without appearing on an invoice, and the variance line is the one place they show up.
Actual vs theoretical food cost: the definitions
Theoretical usage is what the kitchen should have consumed, based on sales. If the POS says you sold 120 salmon plates and the recipe calls for 170 g of salmon each, theoretical salmon usage is 20.4 kg.
Actual usage is what left the building, from the counts:
Actual usage = opening stock + purchases − closing stock
If you opened the week with 8 kg of salmon, received 18 kg and closed with 3.2 kg, actual usage was 22.8 kg.
Theoretical food cost is the theoretical usage priced at usable cost; actual food cost is the actual usage priced the same way. The difference is product you paid for and did not sell.
The food cost variance formula, worked in kilos and dollars
Variance = actual usage − theoretical usage
Variance % = variance / theoretical usage
For the salmon: 22.8 − 20.4 = 2.4 kg. At $22.16 per usable kg, that is $53 of salmon in one week with no sale against it. As a percentage, 2.4 / 20.4 = 11.8%, which is high for a protein.
| Units (kg) | Value at $22.16 per usable kg | |
|---|---|---|
| Opening stock | 8.0 | $177 |
| Purchases | 18.0 | $399 |
| Closing stock | 3.2 | $71 |
| Actual usage | 22.8 | $505 |
| Theoretical usage (120 plates × 170 g) | 20.4 | $452 |
| Variance (11.8%) | 2.4 | $53 |
You need four inputs, and most restaurants are missing at least one:
- An opening and closing count for each item, in the same unit, for the same period
- Each purchase in the period, line by line, in that unit
- The POS item mix for the period
- Recipes mapped to each POS item, with yield-adjusted quantities
Miss number 2 and your actual usage is wrong; miss number 4 and your theoretical is wrong. Either way the variance is noise, so fix the inputs first. Input 1 fails for lack of time, which is why how to take restaurant inventory with a blind count targets 30 minutes. Input 4 fails on stale specs, where accurate recipe costing earns its keep.
Compute variance by item, multiply by usable cost per unit, then sort by money. The top ten lines hold 70 to 80% of the total variance value in most kitchens.
Realistic benchmarks: why vendors disagree, and the 2% target
Ask three software vendors what an acceptable food cost variance is and you get 2 to 3%, 3 to 5% and 10%, sometimes from the same blog. Our target comes from independents who count weekly and receive at the door: under 2% is reachable without a controller, and anything above 5% is systemic.
Total variance as a percentage of theoretical cost:
| Variance | What it means in practice |
|---|---|
| Under 1% | Tight; a QSR with few SKUs and strict specs |
| 1 to 2% | Good; achievable for full-service restaurants with weekly counts |
| 2 to 4% | Normal for a kitchen that has been counting for one to three months |
| 4 to 8% | Something upstream is wrong: inputs, portions or receiving |
| Over 8% | The calculation is broken more often than the kitchen; check mapping and units first |
By category, proteins and spirits should sit under 2%. Produce can run 3 to 5% because of trim and spoilage. Bar variance under 3% is good; above 6% points to pour size or unrecorded comps.
On a restaurant with $25,000 a month in food purchases, moving from 5% variance to 2% is $750 a month.
The five root causes
1. Portioning
The most common cause and the easiest to fix. A 170 g salmon portion cut by eye averages 185 to 195 g, a 10% variance on the line before anything else goes wrong. Fix: a scale at the station for proteins and cheese, portion scoops for rice and sauces, and a weekly spot check of ten plates during service.
2. Waste
Over-prep, spoilage from poor rotation, plate returns, dropped product. Fix: a waste log with item, quantity and reason; two weeks of logging tends to show two or three items behind most of it. Prep lists sized to the forecast cut over-prep.
3. Theft
Less common than owners assume and more common than staff admit. Spirits, premium proteins and anything that fits in a bag are the usual targets. Fix: daily counts on the A-list, two people on the count, receiving signed by someone other than the person who ordered. The pattern to look for is variance on specific high-value items spiking on specific shifts.
4. Receiving errors
Short deliveries, substitutions, damaged cases accepted, invoices for product that did not arrive. These inflate purchases and therefore actual usage, so variance appears with nothing wasted. A 2 kg short on a 20 kg case is 10% variance on that item for the week. Fix: check each delivery against the invoice at the door, weigh the proteins, and note shorts on the invoice before anyone enters it.
5. Recipe mapping
The quiet one. A POS item mapped to the wrong recipe, a recipe with no yield factor, a modifier (“extra cheese”) mapped to no ingredient, a special with no recipe behind it. Each of these makes theoretical usage wrong, so variance appears where there is none and disappears where there is. Fix: a monthly review of POS items with no recipe and of recipes with the biggest variance. Often the recipe is wrong and the kitchen is fine.
A diagnosis sequence
For an item with a large variance, check in this order, which is also the order of likelihood and of cost:
- Is the unit consistent across count, purchase and recipe? (kg on the count, cases on the invoice, grams in the recipe)
- Was each delivery entered? Check invoices received against purchases recorded.
- Is the recipe mapped to each POS item and modifier that uses this ingredient?
- Weigh ten portions. Compare to spec.
- Check the waste log for the item.
- Look at the variance by day and by shift.
Most large variances resolve at steps 1 to 3 with no fault in the kitchen, which is why the number looks scary in the first month and settles after.
Getting from 5% to under 2%
The sequence we see at operators who get there:
- Weeks 1 to 2: Fix units and mapping. Count the A-list (top 20 items by value) daily. Expect variance to look worse at first, because the inputs are now correct.
- Weeks 3 to 4: Scale on the line for proteins. Start the waste log. Receiving check at the door. Variance on proteins tends to halve here.
- Month 2: Weekly full count. Review the top ten variances by value every Monday. Re-cost recipes whose variance stays off; in most cases the spec is wrong.
- Month 3: Variance should be under 3% overall and under 2% on proteins. The remaining items are the ones to have conversations about.
The whole thing takes about 90 days and almost no capital.
Where the tooling fits
Variance is only as good as the four inputs, and the two that fail most often are purchases (invoices entered late or not at all) and counts (skipped because they take too long). We built bilz around those two: line-level invoice OCR at delivery keeps purchases current, phone counts in shelf order mean the counts happen, and theoretical usage comes from POS sales mapped to costed recipes (features). The output is a weekly variance report by value and by item, with no spreadsheet, at $149 per location per month. The method above works with any tooling.
Frequently asked questions
What is an acceptable inventory variance for a restaurant?
Under 2% of theoretical cost for a well-run independent with weekly counts; under 1% for a QSR with few SKUs. Between 2 and 4% is normal in the first three months. Above 5% means something systemic: inputs, portions or receiving.
How do you calculate inventory variance?
Actual usage is opening stock plus purchases minus closing stock. Theoretical usage is units sold on the POS multiplied by the recipe quantity per item. Variance is actual minus theoretical; divide by theoretical for the percentage, and multiply by usable cost per unit for the dollar figure.
What is theoretical food cost?
Theoretical food cost is what your food should have cost given what you sold, calculated from POS sales and costed recipes. If each portion matched spec and nothing was wasted, stolen or short-delivered, actual food cost would equal theoretical.
What causes food cost variance?
Five things, in rough order of frequency: over-portioning, waste and spoilage, theft, receiving errors such as short deliveries, and recipe mapping errors in the POS. The last creates variance on paper with no product leaving the building, so check how recipes and item lists get set up before you blame the line.
Can variance be negative?
Yes. Negative variance means you used less than the recipes say: under-portioning, a spec that is too generous, or a count or delivery entered wrong. Treat it as a mapping or spec problem first; a kitchen serving under spec is shortchanging guests.
