Why your GP% jumps around week to week, and when a bad number is real

A pub's GP% is supposed to wobble. Two points of movement week to week usually means nothing was stolen and nothing is broken, it means you counted on a different day, a keg landed at the wrong moment, or the rugby crowd drank different beer than the Friday regulars. The skill is telling the wobble from the slide. Here are the seven things that move a weekly GP% with nobody doing anything wrong, and the one pattern that means the number is real.

I watch licensees chase a one-week GP dip like it is a crime scene. I have done it myself, marched into the cellar with a printout, ready to find the leak, and found instead that the leak was my own count timing. So before you re-brief the team or ring the stocktaker, run the number past this list.

First, remember what the number is

GP% over a week is a ratio of two estimates taken through a narrow window. Sales for the period, minus the cost of what the period consumed, over sales. Both ends of that sum depend on exactly when you counted, what was in transit, and how well you measured the part-full containers. Narrow window, small denominator, noisy number. On a £6,000 wet week, which is £5,000 ex VAT, one GP point is just £50 of cost. Fifty quid. Keep that in your head as we go, because several perfectly innocent things move £50 of apparent cost without breaking sweat.

If you are not sure your GP is even being calculated on the right VAT basis, stop here and read how to read a GP report first, because an inc-VAT calculator will disagree with an ex-VAT one by six points all day long, and that has nothing to do with your cellar.

1. The delivery that straddles the count

This is the biggest one and it is pure timing. Say a £160 keg arrives Thursday, the invoice goes in the week's purchases, and your count on Sunday somehow misses it, still on the dray trolley, behind the empties, wherever. The sum now thinks that keg was consumed. On our £5,000 ex-VAT week that is 3.2 GP points gone, from one keg and one piece of paper. Reverse it, count the keg but miss the invoice, and your GP looks 3.2 points better than it is, and next week pays it back.

The fix is boring discipline: the count and the paperwork must describe the same moment. Same cutoff, every week, deliveries either side of it belong to either side of it. I wrote more about delivery-day chaos in beer deliveries and your stocktake.

2. Unequal weeks

A count done Sunday night this week and Tuesday morning next week is not comparing two weeks, it is comparing six days with nine. Bank holidays, a staff no-show on count day, doing it after the quiz instead of before, all of it stretches or squeezes the window. The percentage does not care about your reasons. Count on the same day at the same time, and if you want the method nailed down, here is how often and when.

3. Quiet weeks read worse, and that is just arithmetic

Some losses are fixed pints per week, not a percentage of sales. Line cleaning is the obvious one: the beer in the lines goes down the drain on cleaning day whether you sold 1,000 pints that week or 2,000. Twenty pints of cleaning waste is 1% of a 2,000-pint week but 2% of a 1,000-pint week. Same routine, same waste, and the quiet week's GP reads a full point worse. Your GP% is partly a weather report.

So when the number dips in a dead January week, check the volume before you check the staff. The line cleaning guide puts real costs on the cleaning cycle, and draught wastage: what is normal covers the rest of the fixed losses.

4. Mix shift: your customers changed the number, not your cellar

Every product on your bar carries its own GP. Premium lager at £5.94 a pint and cask at £4.91, the Morning Advertiser's 2026 averages, do not make you the same margin, and spirits, post-mix and wine sit somewhere else again. Your weekly GP% is the blend of whatever people happened to drink.

A big cask weekend, a cocktail crowd in for a birthday, the football lot on session lager instead of the Friday gin circle, any of it moves the blended number a point or more with zero loss anywhere. If your report breaks GP down by category, and it should, check whether the categories held steady before deciding the pub has a problem. My GP benchmark guide shows the category spread you should expect.

5. Part containers, estimated instead of measured

Eyeballing gets the full and the empty right. It is everything in between where the guessing lives. Call a 70cl bottle half full when it is actually at 40% and you have invented 7cl of spirit, about £2 of stock at cost on a £20 bottle. Do that across 40 optics and back-bar bottles, kind guesses one week, harsh the next, and the estimate noise alone swings the count the best part of £100 either way, getting on for two points on our example week, without a drop actually moving.

Kegs and casks are worse because they are bigger and opaque. This is exactly why I weigh instead of guessing: spirits on a scale, and kegs by weight with proper calibration. Measured part-containers are the difference between a GP that wobbles half a point and one that lurches two.

6. The till and the stock telling different stories

Prepaid pints, the one-in-the-pump culture, wrong till buttons, staff ringing the house double as two singles, refunds and voids done generously. None of this changes what is physically in the cellar, but all of it changes the sales figure your GP divides by. It is common enough to have an industry number: Vianet's dispense monitoring found 2% of drinks poured never reach the till at all. If the variance report and the GP move together in a way the bar cannot explain, read why the till does not match the stocktake and the prepaid pints guide, because the answer is usually on the till side.

7. A price moved and your book did not

Suppliers reprice constantly, and the annual list rise is only the loud one. If a cost price went up on the invoice but your system still holds the old one, your GP is being calculated against fiction. It reads fine until the day someone updates the file, then appears to fall off a cliff in a single week. Check cost-price dates before panicking, and track the rises line by line so the book moves when the invoice does.

So when is a bad number real?

Direction, not wobble. One week two points down, with a delivery on count day or a festival weekend in the mix: noise, note it, move on. The same report four weeks running, each one a little lower, on steady volume with clean count timing: that is a slide, and slides have causes. A four-week rolling average smooths the timing noise and shows the trend your weekly print hides.

And when you do investigate, investigate in units, not percentages. A GP% cannot tell you which tap is leaking; a line-level variance can. Find the products that are short, in pints and bottles, and follow those. What variance is acceptable gives you the thresholds, and if the slide survives all of that scrutiny, the 30-day GP recovery plan is the systematic way back.

The point is not that theft never happens. It does, and a real slide deserves a real look. The point is that the weekly wobble is mostly physics and paperwork, and every hour spent interrogating noise is an hour not spent on the four-week trend, which is where the money actually is.