From Stock Variance to Investigation: Closing the Loop

Most retail teams are good at finding variances and poor at doing anything with them. A count reveals a gap, the gap gets a note, and the note gets forgotten. The distance between “we are short” and “we know why, and it has stopped” is where loss actually lives.

A variance is a question, not a conclusion

When a count comes back short, the honest answer at that moment is: we do not yet know what happened. A negative variance could be theft, a receiving error, a mis-scan, an unrecorded transfer, damaged stock never written off, or a supplier who short-delivered. Treating the variance as automatic proof of theft is as wrong as treating it as automatic proof of an admin slip. The variance is the start of an enquiry, and the quality of your loss prevention depends almost entirely on how disciplined that enquiry is.

The loop that most retailers leave open

Picture the ideal path from count to closure:

1. A count produces a variance above a threshold that matters.
2. The variance is triaged — is this material, is it recurring, is it concentrated on one product, till, shift, or branch?
3. The credible innocent explanations are checked and ruled in or out: transfers, receiving, returns, known damages.
4. What remains unexplained becomes a case with an owner, a record, and a next action.
5. The case is worked, resolved, and — crucially — its lesson feeds back into the process that let it happen.

Most retailers do step one well and steps two through five badly. The count exists, but the triage is informal, the innocent-explanation check is skipped, and nothing ever becomes a proper case. So the same variance reappears next quarter, and the business slowly accepts it as “normal shrinkage” when it was a solvable, repeating leak.

Triage before you investigate

You cannot investigate every rand of variance, and you should not try. Triage is what makes the loop affordable. A useful triage asks three questions: is it big enough to matter, is it repeating, and is it concentrated? A small, one-off, evenly-spread variance is usually process noise and can be logged and monitored. A variance that is large, or repeating, or tightly concentrated on a single point deserves attention now, because concentration is the fingerprint of a specific cause rather than general slippage.

This is why frequent, localised stock counts are worth more than rare, total ones. A count that can pin a gap to one category in one branch in one week hands you a lead. A count that only tells you the whole estate is 1.8% short at year-end hands you nothing but anxiety.

Rule out the innocent explanations first

Before anyone is under suspicion, the mundane causes must be eliminated on the record. Were there transfers in or out that were not captured? Was the delivery actually received in full? Were refunds processed with product returned to stock? Were damages binned without a write-off? Doing this first protects staff from unfair suspicion and protects the eventual investigation from falling apart. If you skip it and jump straight to accusing a person, you both damage trust and hand any real wrongdoer a defence.

When it becomes a case, treat it like one

What survives triage and the innocent-explanation check is your genuine unexplained loss, and it deserves a real record rather than a mental note. A case needs an owner, a timeline, the evidence gathered, the actions taken, and an outcome. Structured investigation and case management matters here for an unglamorous reason: consistency. A case that is properly documented can be reviewed, escalated, handed over, and — if it ever leads to disciplinary or legal action — actually stand up. A case that lives in someone’s memory cannot.

It also matters for POPIA. Once you are recording enquiries that touch on named individuals, you are processing personal information, and that processing should be lawful, purposeful, and secured. A structured case system makes that far easier to demonstrate than a folder of loose notes and screenshots.

Feed the lesson back

The loop is not closed when the case is resolved; it is closed when the process changes. If the enquiry found a receiving gap, tighten receiving. If it found refund abuse, review the refund control. The whole value of connecting counts to cases is that each closed case should make the next variance less likely, not just punish the last one.

If your counts currently end in a spreadsheet rather than an outcome, the cheapest improvement is a written triage rule and a simple case record with a named owner. If it would help to see how other retailers structure that hand-off from count to case, we are glad to walk through it.