The shrinkage that survives every closing count

Voids, refunds and shrinkage rarely look like theft at a cafe counter. They look like normal operations, slightly tilted: a sale voided after the customer left, a refund processed with nobody at the counter, a no sale drawer open a dozen extra times a shift. Each event is defensible on its own, which is exactly why you watch the pattern, not the event.

What should a void record capture?

Voids are a legitimate tool; the control is recording them completely, not preventing them. Every void should carry who performed it, who approved it, when, which items, the amount and a reason code.

The line that matters most is before versus after payment. A post payment void on a completed sale is the classic move for pocketing cash: take the money, cancel the record, and the drawer still balances. If any cashier can void a completed sale alone, that is not a control gap, it is an invitation.

Which actions need a manager PIN?

• Voiding a transaction after payment
• Refunds above a small threshold, say ₱200
• Manual discounts outside the programmed ones
• Opening the drawer without a sale

The point is not that managers never collude. It is that every sensitive action now carries two names, and casual opportunism dies when it needs a witness. Keep the threshold low enough to matter and high enough that the queue does not stall over a ₱35 cookie.

What patterns surface in a weekly review?

Ten minutes with the right report: void rates by cashier, where a team average of 1 percent makes one person's 4 percent a kind and early question; voids clustering near closing or on one recurring shift; no sale opens far above the norm; refunds without an original receipt; the same friendly discount appearing daily.

Watch refunds hardest. A fake ₱150 refund vanishes into a busy day, but refund frequency per cashier over a month does not lie, and a GCash reversal never lightens the drawer, so only the report will ever raise its hand.

Schedule the review like a shift. The check that happens "when there is time" happens twice and dies; the one on the calendar every Monday pays for itself the first time a pattern shows up early.

Cross check the stockroom

Transaction shrinkage has a twin: product leaving with no transaction at all. If sales say you used 12 kg of beans and the shelf says 14 kg left the bag, something moved without being rung up, and recipe level inventory is what makes that gap computable.

Neither report alone proves anything. Run them side by side weekly and they narrow the question to a week, a shift, sometimes a person.

Boring reports are the goal

After real controls go in, the exception reports come back dull, and that is the win. When staff know voids carry names and someone reads the report every Monday, the temptation math changes for everyone, and honest staff can no longer be falsely suspected either.

The goal was never to catch people. It is to run a counter where there is nothing to catch, and where everyone behind it knows that for certain.