Aisle two, jacket pocket
She's watched him for ninety seconds. Two items into a bag that isn't a store bag, a third moved from the shelf to a jacket pocket, and a route toward the doors that avoids the tills.
The associate is twenty-two. There's no loss prevention officer on today. The store policy — which she read during onboarding and has not looked at since — says something about not approaching, but she can't remember the specifics, and he's forty feet from the exit.
Everything she feels says do something. The stock is the store's, she's watching it walk out, and standing still feels like complicity.
Almost everything she might do is forbidden, and the reasons are better than she knows.
The conversation you may not be allowed to have
This is the only scenario across these series where the correct action is frequently to not act at all — and where the instinct to act is strongest.
Three things make it unlike anything else on a shop floor.
You may be wrong, and you cannot know. Bags get reused. People pocket things absently and pay for them. A person who looks like they're concealing is sometimes a person with their hands full. Human suspicion in this setting is unreliable — not because staff are careless, but because the behaviours that look like theft overlap heavily with ordinary shopping. It's also unreliable in patterned ways: misidentification doesn't fall evenly across customers, and a store that leaves the judgment to individual instinct will produce a distribution it would not defend if it were written down. That is a liability question as much as an ethical one, and it is the strongest argument for having a procedure that doesn't depend on being right.
Most policies forbid most instincts. Detention, pursuit beyond the door, physical contact, blocking an exit, accusing — the majority of retailers prohibit all of these for associates, and the ones that permit any of it usually restrict it to trained loss prevention staff acting under specific conditions. Associates rarely know the detail, which means the policy exists and doesn't operate.
And the exposure is asymmetric in the same direction as everything else in retail. Recovering a £40 item is worth £40. A wrongful accusation is worth a complaint, a claim, a viral clip, and occasionally litigation — and an injury to a staff member is worth more than any of it.
The move that resolves it
There is one approach that works regardless of whether the suspicion is correct, and it's the reason loss prevention associate training should lead with service rather than security.
"Hi — can I help you find anything? I can take those to the till for you if it's easier."
Consider what that does. If he's stealing, he has been seen, acknowledged and offered a way out that costs him nothing — and presence is the most effective deterrent available to an untrained associate. If he isn't stealing, he has been offered help. The approach requires no judgment about guilt, carries no accusation, and is defensible in every direction.
Compare it to anything that presumes theft. "Can you empty your pockets?" is an accusation, is almost certainly outside policy, and is catastrophic if wrong. "I saw you put that in your bag" commits the store to a version of events before anyone has checked.
Approach as service. It's not a euphemism — it's the only version that survives being wrong.
The rest of it
Know your policy specifically, not generally. Not "we don't confront" but: can you approach, can you ask, can you follow to the door, can you follow past it, who can you call, and what do you do if they run. Associates who half-remember a rule improvise under pressure.
Never touch, never block, never follow outside. These three cause nearly all the injuries and nearly all the claims. The door is the line, and it exists to protect the associate more than the customer.
Let it go when it goes. If he leaves with the goods, he leaves. The merchandise is not worth a chase through a car park, and no retailer's policy actually wants one — even when store culture quietly suggests otherwise.
Record behaviour, not people. Afterwards, write what was done: items, times, actions, route. Not what someone looked like. A behaviour-based record is more useful to the people whose job this is, and a description-led one is where profiling problems become documented ones.
And escalate to the people trained for it. Loss prevention, the duty manager, the police where policy directs. The associate's job is to observe, deter by presence, and report.
What the store rewards is what the store gets
This is the operational finding underneath the scenario, and it sits with management rather than with staff.
If a manager praises an associate for chasing someone down and recovering a jacket, every associate on that shift has learned what the store actually values — whatever the policy says. Recovery stories get told. Near-misses don't. Over a few months that culture will produce someone standing in a car park holding a stranger's sleeve.
The correction is equally simple and almost never made: praise the disengagement. Name it in the shift huddle. He got out with two items, nobody touched him, everyone's fine, that's exactly right. Until a store does that out loud, the policy is competing with a story about someone who got the jacket back.
Four ways it goes wrong
The confronter accuses, on suspicion, in public.
The pursuer follows past the doors. Outside policy almost everywhere, and where the serious injuries happen.
The blocker stands between a person and an exit. Reads as detention regardless of intent, and escalates instantly.
The celebrated hero, praised for a recovery, who teaches the rest of the team what the store really wants.
Why current training leaves the gap
Theft training is detection training. Concealment behaviours, high-risk categories, sweethearting, how to watch. That content is real and it stops at the moment a person is standing in front of you.
The policy is read once at onboarding. In an environment with high turnover, that means a large share of the floor cannot state the rule they're expected to follow under pressure.
The service approach is rarely framed as the security approach. Presence and engagement are the most effective untrained deterrents available, and they're taught as customer service rather than as loss prevention — so the two never connect.
And peer role play cannot approach it. Colleagues will not play a shoplifter credibly, the scenario is uncomfortable to stage, and the version that gets rehearsed is a compliant person who admits it. The real difficulty is acting well under uncertainty about whether anything is happening at all.
What retail theft policy training can rehearse
A simulation can present the ambiguous case repeatedly — sometimes theft, sometimes not, with the associate unable to tell — which is the only honest way to practise a procedure designed to work without certainty. Foretell AI supplies the counterparty configuration, transcripts and rubric-based scoring; the theft policy, escalation routes and legal thresholds stay with the retailer.
Four to build:
- The ambiguous shopper, whose behaviour reads as concealment and who is paying. The essential persona, because the whole procedure exists for him.
- The one who is stealing and responds to a service approach by putting the item back. The most common real outcome and the one associates don’t expect.
- The one who runs, testing whether the associate holds the line at the door.
- The angry innocent, who reacts furiously to being approached at all — rehearsing recovery when the approach itself has caused offence.
Design caution. This scenario involves accusation, potential violence and real discrimination risk. Build it against the retailer's actual policy and local law, make debriefs mandatory, keep every persona behavioural rather than descriptive, and state plainly that it rehearses communication and confers no security or detention competence.
Designing the module
Pass one — the service approach. Score whether the opening presumed guilt in any form and whether help was genuinely offered.
Pass two — the innocent customer. Score whether the associate could withdraw without implying suspicion, and whether they made any accusation they'd have to retract.
Pass three — the exit. Score whether the associate held the door line, avoided contact and blocking, and escalated correctly.
Rubric on observable behavior: Did the opening work equally well if the customer were innocent? Was any accusation made? Was contact, blocking or pursuit attempted? Was the door line held? Was the report written in behavioural rather than descriptive terms?
That first criterion is the single best test in the post, and it's assessable from a transcript alone: read the opening line and ask whether it would have been fine said to someone doing nothing wrong.
The operator case
The recovery is never worth the exposure. Stock value against injury, claims, reputational damage and staff attrition — the arithmetic is not close, and it's rarely laid out for the people making the decision in the aisle.
Policy knowledge decays with turnover. A rule read once at onboarding, in a workforce that turns over substantially each year, is not an operating control.
Discrimination risk is concentrated exactly here. Where approach decisions rest on individual suspicion, the resulting pattern becomes the retailer's problem — and a service-first procedure that doesn't require a guilt judgment is the most practical mitigation available.
And culture beats policy. What gets praised in the huddle determines what staff risk. That's a management behaviour, it's free to change, and almost nobody audits it.
For retail programmes, this is a strong judgment-under-uncertainty exercise and an unusually clear case where the legally, ethically and commercially correct answers converge on the same behaviour.
Frequently asked questions
What should a retail associate do if they suspect someone is shoplifting? Approach as service — offer help, offer to take items to the till — which works whether or not the suspicion is correct. Don't accuse, touch, block or follow outside, and report to loss prevention or a manager.
Can retail staff stop or detain a suspected shoplifter? At most retailers, no. Detention, pursuit and physical contact are typically restricted to trained loss prevention personnel under defined conditions, and associates need the specifics of their own policy rather than a general impression.
What if they leave with the goods? Let them go. The merchandise is worth less than an injury or a wrongful-accusation claim, and no policy genuinely wants a chase — even where store culture implies otherwise.
How should the incident be recorded? By behaviour: items, times, actions, route. Not by physical description of the person, which is where profiling problems become documented ones.
The short version
She can't know. That's not a failure of attention — it's the permanent condition of the job, and every approach that depends on being right is a bet the store shouldn't be asking a twenty-two-year-old to place.
There's one opening that works either way, and it's the same sentence she'd use for any customer in aisle two. If he's stealing, being seen and offered help is the most effective thing she can do without training. If he isn't, she's helped a customer. Nothing else on the menu is safe in both directions.
Foretell AI lets retailers build conversational simulations — including suspected-theft approaches, ambiguous-behaviour judgment, and frontline safety scenarios like the one above — with configurable counterparties, transcripts, recordings, and rubric-based evaluation. If your theft policy is read once at onboarding and improvised thereafter, we're happy to walk through how other operators have structured it.