Self checkout shrink: reading the patterns instead of blaming the machines
Three behaviours, one number
Start by splitting the till data before arguing about causes. Weight mismatches and scale interventions cluster differently from void and supervisor override patterns, and both differ from the clean, fast, no intervention transactions that still lose stock because an item never crossed the scanner at all. Each leaves a different residue in the logs, which is the only reason the split is possible.
Honest error is wide and shallow: many customers, tiny amounts, random items. Deliberate skip scanning is narrow and deep: few baskets, repeated categories, the same high value lines missing again and again. Any response that does not distinguish them will spend its budget on the wide shallow problem because it is the visible one.
Every time an attendant clears a weight warning, a restricted item, or an unscanned item alert, the till writes it down. Read as a whole, that log is a map of where your process fights your customers: the products whose packaging defeats the scale, the produce codes nobody can find, the bagging area that rejects a light item. A store that fixes those three frictions cuts its honest error without accusing anyone of anything.
What the intervention log really says
The same log flags the deliberate pattern when you slice it by time and till rather than by day. Repeated interventions on one machine in one hour, baskets with many items and suspiciously few scans, and attendants being waved past are signals worth a human look. The log is evidence for a review, not a verdict, and treating it as a verdict is how stores end up challenging pensioners over a carrier bag.
One attendant watching eight tills is not supervision, it is presence. The stores that hold shrink down staff the bank so that interventions happen at the till where the problem is, quickly, with a script that assumes error first. Speed matters more than sternness: a fast, cheerful rescan teaches the honest customer the process and tells the deliberate one that this bank is watched.
Attendants are the control, not the furniture
Rotate attendants across the bank and log who cleared what. Collusion at self checkout is rare but real, and it hides in a fixed pairing of one attendant and one regular customer. Rotation breaks the pairing without accusing anyone, which is why it belongs in the standard rota rather than in an investigation.
A bank level shrink target invites gaming: tighten everything, slow every transaction, watch the honest customers leave. Category level counts are harder and more useful. Count the lines that walk: small high value goods, spirits where they are sold at self scan, meat, cosmetics. If those counts fall while overall throughput holds, the controls are working. If total shrink falls because total sales fell, nothing has been fixed.
Review the counts on the same cycle as the stocktake, with the till logs beside them. The conversation that matters is specific: this category, these hours, this intervention pattern. Anything vaguer produces another poster by the tills, and posters have never scanned a tin.
Measure by category, not by bank
Keep reading: How do bots abuse price match guarantees to drain retail margins?, How do credential lists from old breaches keep powering retail account takeovers?.
Frequently asked questions
Is most self checkout loss deliberate theft?
No single split fits every store, but honest error is usually the larger count of incidents and deliberate loss the larger value per incident. The till logs let a store measure its own split instead of importing someone elses headline figure.
Is most self checkout loss deliberate theft?
They catch items placed in the bagging area unscanned, and they miss items that never reach it. Weight checks are one signal. Intervention patterns, category counts, and attendant coverage carry the rest of the control.
Is most self checkout loss deliberate theft?
Cameras record, they do not intervene, and reviewing footage after the fact is the most expensive way to learn what the till logs already show. Cameras support a specific investigation. They are a poor substitute for staffing the bank properly.
Is most self checkout loss deliberate theft?
Assume error, offer the rescan, and keep it light. The large majority of approaches end there because the large majority are errors. A pattern that repeats after friendly interventions is a different conversation, held away from the till bank with records in hand.