How to Catch a Dishonest Employee Without Confrontation

Suspecting an employee of theft puts most shop owners in an uncomfortable position: say something too early, based on a feeling, and you risk accusing an innocent person, wrecking morale across your whole team, and possibly damaging your standing as an employer if the accusation turns out to be wrong. Say nothing and wait, and the loss continues. The way out of that bind isn't to pick a side quickly — it's to slow down and let evidence do the work before any conversation happens.

Why you shouldn't accuse anyone without evidence

In Nigeria, as anywhere, a wrongful accusation of theft is not a small thing. It can damage a person's reputation in a way that follows them to their next job, it opens you up to a legitimate grievance if the employee disputes it, and — even if you never say a word to anyone else — it changes how the rest of your staff see you as an employer. Teams talk. If word gets around that the owner accused someone based on a gut feeling and was wrong, you lose trust with everyone, not just the person you accused. The standard to clear before any conversation should be a real pattern, not a suspicion.

What actually counts as evidence, not coincidence

A single bad shift, one short delivery, or one till discrepancy is not evidence of theft — it's a data point, and on its own it's far more likely to be a genuine mistake, a supplier error, or a miscount. Evidence is a pattern that repeats in a way random chance doesn't explain. Concretely, that means:

• The same person's shifts show discrepancies significantly more often than their colleagues', across at least two to three weeks — not one bad day
• The pattern is concentrated on specific high-value or easily concealed items, not spread randomly across the shelf
• Discrepancies line up with days that person worked alone or had unsupervised access, and largely disappear on days they didn't
• There's a documented record — counts, timestamps, photos — not just a memory of "it felt like stock was low that week"

Notice what's missing from that list: lifestyle, personality, or how someone "seems." Those things might eventually be consistent with a conclusion you've already reached with real evidence, but they should never be the basis for reaching it. People have side income, family help, and habits that have nothing to do with your shop, and treating those as evidence is exactly how honest employees get wrongly accused.

How to build the trail quietly

The goal before any confrontation is to have a record that speaks for itself — one that doesn't depend on your memory, doesn't depend on anyone's willingness to admit anything, and can't be argued away as "that's not what happened."

1. Run counts at the start and end of every shift, not just once a day. This is what lets you attach a discrepancy to a specific person instead of a vague 24-hour window where three people had access.
2. Keep a written or photographic record of every count, dated and tied to whoever was on shift. A number in your head is not evidence six weeks later; a dated record is.
3. Don't change anything about how you supervise that specific person during this period. Altering their schedule or watching them differently can tip them off, and it can also taint the pattern you're trying to observe honestly.
4. Let the pattern run its natural course for several weeks before drawing a conclusion. Resist the urge to act on week one's numbers.
5. When you do have a pattern, review it against other possible explanations first — a new supplier, a change in product mix, a genuinely careless but honest mistake — before treating it as theft.

This is where photo-based counts change the nature of the record entirely. Shelfie has attendants photograph the shelf at the start and end of every shift on their own phone, and its AI counts what's visible and flags discrepancies the same day — automatically tied to whoever was clocked in, with clock-ins verified by face so the record can't be logged by someone who wasn't actually there. Instead of a manager's memory of "I think stock was low that week," you end up with a dated photo and count for every shift, which is a very different thing to bring into a conversation than a suspicion.

Turning the evidence into a conversation, not an accusation

Once you have a genuine pattern, the conversation itself should still start from a place of asking, not telling. Show the records, explain what you're seeing, and give the person room to respond before you draw a conclusion. Sometimes what looks like theft is a process problem — nobody trained them on how to log damaged stock, for instance — and the record will make that clear just as easily as it makes theft clear. The value of building the evidence trail properly isn't just that it protects you legally and reputationally. It's that it turns a conversation that would otherwise be "he-said-she-said" into one grounded in facts both sides can look at together.