Your food cost percentage is probably wrong, and the count is why

31.4 percent. It was 29.8 last month and 30.1 the month before, and now it is 31.4, and the chef says nothing has changed and the bookkeeper says the figure is the figure. Somewhere in the next fortnight this gets blamed on beef.

It might well be beef. More often it is that the number is not measuring what everyone in the room believes it is measuring.

The formula, quickly, so we can get past it

Food cost percentage is opening stock plus purchases, minus closing stock, divided by food sales. Both sides excluding VAT, which is the first place a South African kitchen loses a point and a half without noticing: an invoice total includes VAT and a menu price includes VAT, and mixing an inclusive cost with an exclusive sales figure gives you a number that is wrong in a direction you will not question, because it looks bad enough to be believable.

Four ways the figure lies

The closing count

Anything you did not count is treated as eaten. That is what the formula does: it has no way to tell the difference between a case of chicken that went out on plates and a case of chicken sitting behind a stack of boxes in the walk-in where nobody looked. Miss R14 000 of stock on the shelf and your food cost goes up by a point and a half, and the reason is that somebody was in a hurry at half past eleven at night.

The same works in reverse, which is worse. A generous count makes this month look excellent and next month look terrible, because that phantom stock becomes opening stock and has to come out somewhere. A kitchen with a good month followed by a bad one, over and over, usually does not have a purchasing problem. It has a counting problem with a two-month cycle.

Purchases that are not purchases

A credit note filed as an invoice adds cost twice. An invoice dated the 31st for a delivery that arrived on the 2nd puts the cost in one period and the stock in another. A statement paid as though it were an invoice double-counts everything on it. None of these is exotic. All three happen in most restaurants in most months, and none of them announces itself.

The denominator

If drinks revenue is in your sales figure but drinks stock is not in your cost figure, your food cost percentage is flattering and meaningless. Staff meals, comps and platform commission all do their own damage. Delivery is the one that catches people now: a R220 order that nets R180 after commission still costs you the same food, and if you are booking the gross you are overstating your margin on every order that comes through an app.

Stock that moved without being sold

Wastage, transfers to a sister site, and anything you manufacture. Twelve litres of stock reduced from bones is no longer bones. If the system cannot express that, the bones look stolen and the stock looks free, and both errors land in the same percentage.

The formula is not the hard part. The formula has never been the hard part.

The number worth chasing is a gap, not a percentage

Your recipes already say what a plate should cost. Multiply that by what the till says you sold, and you have theoretical cost: what the month should have cost if every portion was on spec and nothing left the building unrecorded.

Actual cost is what the count says happened. The distance between the two is the interesting figure, because unlike the headline percentage it has no innocent explanation. It is over-portioning, waste that was never logged, breakage, theft, or a recipe that stopped matching the kitchen some time last winter.

A restaurant running at 31 percent against a theoretical 30 has a small, boring problem. A restaurant running at 31 against a theoretical 26 has a five point hole, and no amount of renegotiating with suppliers will close it, because the money is not going out through the back door of purchasing. Same headline number. Completely different week ahead.

On benchmarks

You will find plenty of blogs that will tell you the right food cost percentage for your restaurant. Nearly all of them are American, most are quoting each other, and none of them know whether you run a steakhouse, a coffee shop or a wine bar that sells three snacks.

A butchery-led menu and a pasta-led menu are not playing the same game and never will be. What is worth comparing is your own trend, your own theoretical gap, and what you are paying for an ingredient against what the market is paying for it in the same period. That last one is a real benchmark, because it holds the thing being measured constant.

Making the count worth trusting

• Count at the same point in the cycle every time, and preferably before a delivery rather than after one.
• Count in the unit you buy in. A kilogram counted as a case is a rounding error with a multiplier attached.
• Split the sheet by area so the walk-in, the dry store and the bar can be done at the same time by different people.
• Do not count during service. Nobody has ever counted accurately while the pass is calling.
• Log wastage as it happens. Reconstructing a month of waste on the last day of it is fiction, however honest the person doing it.
• Close the period and stop editing it. A count somebody can amend next Tuesday is not a control.

Load shedding deserves its own line here. A fridge that lost temperature overnight is a real cost, it is not a purchasing failure, and if it goes in as unexplained variance instead of as wastage then somebody spends a week interrogating a chef about a generator.

Where software actually helps

Not with the arithmetic. Any spreadsheet can divide. Where it earns its place is in the inputs: reading invoices so the purchases figure is complete and correctly dated, keeping recipes costed as prices move so the theoretical number is current, holding wastage and transfers and manufacturing as first-class events, and refusing to show a closing valuation that does not follow from the opening one.

That last one is the useful discipline. Most systems will happily give you a tidy percentage regardless of whether the period reconciles, because a tidy percentage is what the buyer asked to see. It is more use to be told that R11 400 is unaccounted for and here are the eight ingredients it is hiding in.

One point on R980 000 of monthly food sales is R9 800. You do not need a heroic improvement to justify counting properly. You need to stop being wrong by a point and a half in a direction you cannot see.