Stock loss you can't explain: the fortnight that finds it

The count says £600 has gone and you've already checked the usual suspects. The buttons look right, the lines were cleaned on schedule, nobody remembers a leaking keg. This is the point where a lot of licensees do the worst possible thing: they start watching staff. Put that away. Most "unexplained" loss turns out to be a recording problem or a counting problem, and there's a boring, methodical way to find out which. It takes about two weeks and it starts with proving the loss actually exists.

I've written before about telling pouring, waste and measuring error apart from the numbers. That guide is for reading one result. This one is for what to do when you've read it, checked the obvious causes, and the gap is still staring back at you.

Day one: prove the loss exists

A stocktake is a measurement, and measurements have error in them. Before you investigate a loss, make sure you're not investigating a counting mistake.

Go back through the count itself. Part bottles are the first place to look: if spirits were estimated by eye rather than weighed or done in honest tenths, forty bottles of guesswork can move a result by a couple of GP points on their own. Check the keg yields the count assumed. An 88-pint keg does not sell 88 pints, and if your system thinks it does, every keg looks like loss. Check the VAT basis, because a count valued one way compared against sales valued the other manufactures a deficit out of thin air. And check the period boundaries: a delivery that arrived after the last count but got booked before it, or the other way round, moves real money into the wrong week.

Then check the paperwork feeding the count. Was every delivery checked off at the door, or just signed? A short delivery you signed for looks identical to theft by count day: stock booked in that never physically existed. Credits and returns that never got posted do the same thing. If any of this turns something up, adjust and re-run the numbers before going further. More often than not, this step alone explains the gap.

Day two onwards: shrink the haystack

Say the count survives checking. You now know money is leaving, but a whole-site variance number tells you almost nothing about where. The fix is to stop measuring everything monthly and start measuring a few things daily.

Pick the five products with the biggest cash variance from the count. Not percentages, pounds. Then line check just those five every day: count them, note the figure, compare movement against till sales since yesterday. Five lines takes five minutes with the kettle on. Within a week you'll know which products are actually walking and, roughly, which days.

The reason this works is that unexplained loss is almost never spread evenly. Vianet's dispense data across 11,000 pub sites found around 2% of draught drinks poured never reach the till, and 61% of pubs miss their target pouring yield. Those aren't site-wide averages inside one pub, they cluster on particular lines and particular sessions. Daily counts find the cluster.

While you count: list every route stock leaves without a sale

Run through the exits that don't ring a till, because each one is a legitimate explanation hiding in plain sight:

• Line cleaning. A ten-tap pub bins around 20 pints a week just cleaning lines. If nobody logs it, it reads as loss.
• Prepaid pints. Every "one in the pump" is money in the till this week and a pint leaving the cellar next week. Uncounted, they make one count look brilliant and the next look criminal.
• Voids and replaced drinks. A pint poured wrong, binned and replaced is one sale and two pints. If the replacement doesn't get rung at zero, the second pint is invisible. The voids and refunds guide covers reading that report.
• Comps and staff drinks. Whatever your policy is, if it isn't rung through a button, it's "loss".
• Wastage nobody wrote down. Split casks, out-of-date bottles, the dropped tray. The wastage log exists precisely so this stuff has somewhere to go other than your variance line.

The pattern across all five is the same: the stock really did leave, the business already paid for it, and the only thing missing was a record. Fixing the recording fixes the "loss" without a single difficult conversation.

One deadline you're on: the CCTV clock

If it does turn out you need to look at footage, know that most licence conditions requiring CCTV specify around 31 days of retention, and systems overwrite on that cycle. Westminster's model condition, used widely as a template, says 31 days. A monthly stocktake plus two weeks of investigating means the early part of your loss period may already be gone. That's an argument for starting the daily counts immediately, not for reviewing footage speculatively. Footage is for confirming something specific once the counts have narrowed it, not for fishing.

Day fourteen: the second count, and the three outcomes

Two weeks in, count the narrowed area again properly. You'll land in one of three places.

The gap is explained. The line checks and the exit list found it: unlogged cleaning waste, a prepaid pot, a miscoded button, a short delivery. This is the most common outcome. Fix the recording habit, add the till button, log the waste, and the "loss" disappears from the next count. Done.

The gap stopped. Nothing found, but the daily counts show the products behaving now. Sometimes measurement itself changes behaviour, sometimes the cause was a one-off you'll never identify. Keep the five-line daily check running for another month and move on. Not every mystery gets solved, and chasing a stopped loss costs more than it recovers.

The gap continues, in one place. The counts show a specific product still moving without sales, on specific shifts, with the recording routes ruled out. Only now is this a genuine shrinkage question, and even now the move is process, not accusation. Tighten the till discipline on that line. Make the daily count on it visible and routine, done together at handover, so counting is just what happens here rather than something aimed at anyone. In most cases visible measurement ends the problem on its own.

What not to do, and one bit of law worth knowing

Don't announce an investigation, don't stage a dramatic till check on one person's shift, and don't touch anyone's pay. On that last one: the Employment Rights Act 1996 specifically limits what can be deducted from a retail worker's wages for cash shortages or stock deficiencies, at most 10% of gross wages in any pay packet, and only where the contract provides for it and written notice rules are met. Docking pay for a stock gap you haven't even located isn't just corrosive, it's very likely unlawful. I'm a licensee, not a solicitor, so take employment questions to someone qualified.

The deeper reason not to go straight to suspicion is simpler: you'll usually be wrong. Run an investigation on gut feeling and the likely ending is an innocent person treated badly while the actual cause, a miscoded button or an unlogged waste bin, sits there still costing money. Systems first. People last, if at all.

Where software fits

Everything above can be done on paper. What software changes is the speed of the loop: daily line checks that take minutes because the maths is done for you, a wastage log staff will actually use because it's on the phone behind the bar, and counts you can compare period against period without rebuilding a spreadsheet. StockTap does spot checks against till figures, a wastage ledger, and weighed counts for the spirits that started this whole investigation. The free trial gives you two proper counts to test it on, which happens to be exactly the shape of the fortnight above. The 30-day GP plan extends this investigation into a full month of fixes if you want the longer version.

Common questions

How much variance is normal before I should investigate?

Rule of thumb: investigate anything beyond about 1% of wet sales at retail, and investigate a trend in the same direction over consecutive counts whatever its size. The acceptable variance guide goes deeper on where to set your own line.

Is unexplained stock loss usually theft?

No. In my experience and in the dispense-industry data, the bulk of it is unrecorded legitimate use: cleaning waste, replaced drinks, prepaid pints, comps, short deliveries and counting error. Theft exists, but it's the last explanation standing, not the first.

Should I bring in an independent stocktaker?

If the loss is big, persistent and you've done the fortnight above without an answer, yes, a fresh pair of professional eyes on your counting method is worth the fee. The audit vs stocktake guide explains what an audit checks that your own count doesn't.

How far back can I realistically investigate?

Paperwork, months. Footage, about 31 days on most systems. Memory, a week if you're lucky. That asymmetry is the argument for counting more often: a monthly count gives a loss thirty days' head start, a weekly one gives it seven. How often to count covers the trade-off.