The Till That's Always Short

The drawer is short again. Forty, sixty, sometimes a hundred and twenty.

Two reactions are common and both are wrong.

"It's theft." Sometimes. Usually not, and acting on it without evidence costs you a good member of staff and poisons the room.

"It's just noise." Also sometimes. But a variance that is always in the same direction is not noise — noise is symmetrical.

The direction is the whole diagnosis.

The first thing to look at

Not the size of the variance. The sign.

Pull the last sixty cash-ups and count how many were over and how many were short.

Roughly even, in both directions? That is genuine noise — miskeys, rounding, the odd wrong change given in a rush. Annoying, not a problem to investigate.

Almost always short? Something systematic is happening. It might be theft; it very often is not.

Almost always over? Worth just as much attention. Consistently over usually means transactions are not being rung — which is the same hole, seen from the other side.

What is usually happening when it is always short

In order of how often it turns out to be the answer:

One — the float is wrong. Nobody counted it in properly at the start, so the shortfall was baked in before service. This is the single most common cause and the easiest to eliminate. If the opening count is not signed off by two people, you cannot rule it out and everything below is unmeasurable.

Two — comps and staff drinks are not being rung. A drink given away in good faith, never recorded, is indistinguishable from a drink sold and the money missing. Most rooms have more of these than management believes.

Three — the drawer is shared. If four people use one till through the night, the variance belongs to nobody. A shared drawer converts an accountability problem into an accounting problem, permanently.

Four — change is being made from the drawer for other purposes. Taxi money, a supplier paid in cash, someone's float topped up. Usually recorded on a scrap of paper that does not survive the night.

Five — theft. Real, and it does happen. It is fifth on the list, not first, and you cannot distinguish it from the four above until those four are eliminated.

Why the order matters

If you investigate theft first, you will find it — whether or not it is there. A room where four people share a drawer and comps go unrecorded will produce a pattern that looks exactly like skimming, and the person who happens to be on the most shifts will look guilty.

Eliminating one to four costs a week and no accusations. If the variance survives all four, you have something real and you have it with evidence.

What to change, in order

Two-person opening count, signed. Removes cause one entirely. Takes ninety seconds.

A way to ring a comp as a comp. Removes cause two, and has the side benefit of finally telling you what comps cost.

One person per drawer per shift, or individual logins. Removes cause three. This is the biggest operational change on the list and the one that usually meets resistance — but without it, nothing downstream is attributable.

Any cash leaving the drawer gets a slip in the drawer. Removes cause four. The slip is counted as cash at cash-up.

The measurement that tells you where you are

Variance per shift, plotted, with the operator named.

A pattern that follows a person survives a change of night. A pattern that follows a night survives a change of person. That distinction is the entire investigation, and it is invisible if the drawer is shared.

⚠️ It also takes time. A month of clean data beats a hunch formed over a bad weekend, and the discipline of waiting is what stops you being wrong about someone.

The conversation, when it comes to that

If it does turn out to be a person, the evidence should be able to stand on its own before anyone is spoken to.

⚠️ What you may do next — investigating, suspending, deducting from wages, involving police — is governed by employment law and differs by jurisdiction. Deducting a till shortfall from someone's pay is unlawful in many places even when the shortfall is real. That is a question for your counsel before you act, not after.

Three to hold

Ratio of over to short, last sixty cash-ups. Tells you whether you have noise or a system.

Variance per named operator. Only exists if drawers are not shared.

Recorded comps as a share of sales. The number that stops "always short" being a mystery.

Where the record has to sit

If the float count lives on a clipboard, comps live in someone's memory, and the till report lives in a different system from the rota, none of the three numbers can be produced. That is why the drawer stays a mystery for years.

tasteck records shifts against the people who worked them and keeps comps as recorded transactions, so whose shifts does the variance follow is a question with an answer.

A drawer that is always short is telling you something. Usually it is not what people assume.