Morrisons is using computer vision to make self-checkout less dependent on suspicion

Self-checkout has spent years creating an awkward bargain for retailers: lower labour requirements in exchange for more shrink risk, more interventions and a customer experience that can feel like being monitored by a machine that does not quite trust you.

Morrisons is trying to improve that bargain with computer vision.

The UK grocer is deploying Everseen’s Evercheck technology in an initial 200 stores. The system watches activity around self-checkout and can prompt a shopper to correct an item when it believes something has been missed.

That sounds simple, but the useful distinction is between prevention and accusation.

Traditional loss-prevention systems often intervene after a transaction looks suspicious. Computer vision allows the checkout to respond during the task, ideally catching accidental non-scans, product mismatches or handling errors before they become a loss event.

The technology is increasingly common because self-service now produces enough operational data — cameras, scanner events, weight information and transaction records — to compare what the system believes happened with what was visible at the checkout.

Accuracy is the entire product.

A false alert creates friction for a legitimate customer and sends a colleague to resolve a problem the technology invented. Too few alerts undermine the shrink case. The best deployment is therefore not the one that generates the most detections; it is the one that quietly reduces exceptions while making intervention more targeted.

Morrisons’ initial 200-store scope is large enough to reveal whether Evercheck works across different formats and customer flows. It is still not proof of an estate-wide business case.

The broader trend is clear, though. Retailers are not abandoning self-checkout. They are adding a perception layer around it.

The next generation of self-service will be judged less by how many tills are unattended and more by how well software can distinguish an ordinary shopping mistake from a genuine loss event without making honest customers feel like suspects.