Till voids and refunds: what they actually do to your stock variance

A void is not one thing. The timing decides everything. A void before the drink is poured cancels a bit of bookkeeping and no stock moves. A void after the pour deletes the sales record while the pint is already sitting in front of someone, or down the drain, and that pint will turn up on your next stocktake as missing. Refunds, replacements and no-sales each do something different to the till-vs-stock match, and most pubs never separate them. This is what each button actually does, and the ten-minute weekly habit that stops "voids" being the polite word for variance you can't explain.

I run a wet-led pub in Washington, Tyne and Wear. I have sat through five independent liquor audits in eight months, and the conversation that follows a bad one always circles the same drain: the stock says one thing, the till says another, and nobody can remember why. A decent chunk of that gap, in my experience, is not theft and not short measures. It is voids and replacements that were done for perfectly good reasons and recorded nowhere.

The five buttons and what they do to stock

Every EPOS has its own names for these, but the mechanics are the same everywhere.

What happened	What the till shows	What stock did	Variance effect
Void before pour (wrong item rung, customer changed mind)	Transaction cancelled	Nothing moved	None
Void after pour (drink made, sale cancelled)	No sale recorded	Stock gone	Stock short vs sales
Refund, drink replaced (flat pint swapped for a fresh one)	One sale, or one sale plus a refund	Two drinks gone	One drink's worth short unless logged
Refund, money back, drink binned	Sale reversed	Stock gone	Stock short vs sales
No-sale (drawer opened, no transaction)	Drawer event only	Nothing moved	None directly, but it hides cash handling around the other four

Look down the variance column. Three of the five leave stock short against sales, and none of the three is dishonest. A barrel-end pint that gets pulled and binned, a lager swapped because the customer wanted the other lager, a round rung twice and one copy voided after both were poured. All normal Friday-night traffic. All invisible at the stocktake unless somebody wrote them down.

Why this reads as theft when it isn't

An auditor, or your own stocktake, works from one sum: opening stock plus deliveries minus sales should equal what is on the shelf. Post-pour voids and unlogged replacements break that sum in exactly the same direction as pilfering does. Stock down, sales not up. So a pub with sloppy void habits produces a variance report that looks like a theft problem, and the pubs that then start watching staff harder tend to find nothing, because there was nothing to find. The buttons were the problem.

The industry-wide numbers say the leak is real. Vianet's dispense data across roughly 11,000 metered sites found 2% of drinks poured never reach the till at all. Some of that is line cleaning and training pours. Some of it is exactly this, drinks that existed, got poured, and left no record.

The sum on one wrong pint a night

Say a pint gets replaced most nights. Flat, wrong beer, dropped at the hatch, doesn't matter. The original was rung and paid, the replacement was poured with no transaction. Till looks perfect. Stock is one pint short. At the average £5.34 pint that is about £32 a week, call it £1,665 a year, and every penny of it lands in the "unexplained" column of your stocktake. My arithmetic, illustrative volumes, but run it on your own numbers and it rarely comes out smaller. A ten-second wastage entry at the moment it happens moves the whole lot from "unexplained" to "explained", which is the difference between a stocktake you act on and one you argue about.

Your till already keeps the report, you're just not reading it

Nearly every modern EPOS tracks voids, refunds, discounts and no-sales as standard, by day and by user. Epos Now, for one, lists refunds, voids, discounts and no-sale counts in its stock reporting toolkit, and the others have equivalents. The report exists whether you look at it or not.

One warning about how you use it. These reports break the numbers down by staff member, and the temptation is to read them as a suspect list. Don't. A high void count usually marks your busiest server, your newest starter, or the person honest enough to void mistakes instead of burying them. Read the report by reason and by time, not by name. If Tuesday's voids doubled, something happened on Tuesday. That is a process question, and process questions get answered. Accusations just teach people to stop using the void button, which makes your numbers worse, not better.

What to actually change

Four habits cover it. First, one wastage button that asks for a reason, and a house rule that a poured drink which doesn't get drunk gets logged, same minute, no exceptions, no blame. Second, a written refund line: does the liquid go back on sale (bottles, unopened cans) or in the bin, so the stock answer isn't improvised at 11pm. Third, pull the void and refund totals once a week, the same evening you do your line checks, and skim for the week that looks different. Fourth, count often enough that the gap stays small. A £30 weekly variance is one conversation. The same leak discovered at a quarterly count is £400, six theories and a bad atmosphere.

If your variance is still ugly after the voids are clean, the gap is coming from somewhere else, and the usual suspects are miscoded till buttons, prepaid pints crossing count periods, or the pouring and measuring issues in the stock loss diagnosis guide. Voids are just the cheapest one to fix, because the fix is a habit, not a purchase.

Where software earns its keep here

Weekly counting is what makes void discipline visible. StockTap's spot checks compare a line's movement against till sales between full counts, so a post-pour void problem shows up inside the week it started, on the line it started on, instead of surfacing months later as one big unexplained number. The wastage log sits in the same place you count, which is the only reason staff actually use it.