Theoretical vs. Actual Food Cost: Closing the Gap

The Most Important Gap in Restaurant Finance

Every restaurant has two food costs: what it should be, and what it actually is. The difference between them is one of the most revealing numbers in the entire operation.

Theoretical food cost (also called ideal food cost) is calculated from your recipes and sales mix — it is what your food cost would be if everything went perfectly: no waste, no theft, no portioning errors, no unrecorded usage. Actual food cost is what you really spent, calculated from purchases and inventory counts.

Why the Gap Exists

The variance between theoretical and actual food cost is caused by real operational problems. Understanding what drives it is the first step to closing it:

Portioning Inconsistency

If your recipe calls for 180g of chicken and your cooks routinely plate 210g, your actual cost is 17% higher than theoretical. At 200 covers per day, this 30g over-portion adds up to 6kg of extra chicken daily — a significant cost overrun that is invisible without tracking.

Unrecorded Waste

Food that goes in the bin without being logged does not appear in your recipe costs (because no dish was sold) but does appear in your actual purchases. Every unlogged waste event widens the gap.

Spoilage and Overproduction

Prepping more than you sell creates the same problem. Theoretical cost accounts for what sold — actual cost includes what was prepared and discarded.

Theft and Unrecorded Consumption

Staff meals, comped dishes without POS records, and theft all appear in actual cost but not theoretical. A gap that cannot be explained by waste or portioning deserves a closer look at these categories.

Recipe Inaccuracies

If your costed recipes do not reflect how dishes are actually made, your theoretical cost is wrong from the start. Recipes built on old prices or wrong portion weights generate a fictitious baseline.

What Is an Acceptable Variance?

The industry standard is a variance of 2–3% between theoretical and actual food cost. If your theoretical food cost is 30% and your actual is 33%, that 3-point variance represents significant hidden losses.

For a restaurant with $100,000 in monthly food revenue, a 3% variance is $3,000 per month — $36,000 per year — that cannot be accounted for. Top operators target below 1%.

How to Measure the Variance

1. Calculate theoretical food cost: multiply recipe cost by sales volume for every menu item sold this period
2. Calculate actual food cost: (beginning inventory + purchases) − ending inventory
3. Divide both by total food revenue to get percentages
4. Variance = Actual % − Theoretical %

Manual calculation is feasible if you have accurate recipe costs and a reliable POS export. Integrated systems like Karu automate this entirely — theoretical cost updates live as sales come in, and actual cost flows from purchase records and inventory counts.

Closing the Gap: A Systematic Approach

Step 1: Improve Recipe Accuracy

Audit your top 20 dishes. Verify that costed quantities match what is actually used in production. Update recipes for any ingredient where the costed weight differs from actual use by more than 5%.

Step 2: Implement Portioning Controls

Weigh proteins at plating, not at prep. A 180g portion that goes on the scale is consistently 180g. One that gets portioned by eye drifts 10–25g per plate. Assign scales to every station and make weighing non-negotiable for high-cost items.

Step 3: Log All Waste

Every item discarded for any reason should be logged: item, weight, reason. This creates a data record that can be compared against the theoretical-actual gap. If your gap is 3% and your logged waste accounts for 2%, you are looking for 1% elsewhere.

Step 4: Track Comps and Voids

Every comped dish and POS void should be recorded with a reason code. Comps should never close the gap — they should be recorded and budgeted separately.

Step 5: Review Weekly

A variance review done monthly catches problems after 4 weeks of accumulation. Weekly reviews catch them after one service week, when the cause is still recent and correctable.