A customer browses the spirits aisle, picks up a bottle of whisky, reads the label. Puts it back. Picks up another. This one doesn’t go back on the shelf. It goes inside another item, like a container. By the time they reach the self-checkout, scanning bread, milk, just the container without incident, the highest-value item in their basket has already disappeared.

For most retailers, this is invisible. The checkout cameras see a normal transaction. The till balances. The loss only surfaces weeks later, buried in an inventory count.

Netto Marken-Discount (Netto) took an innovation-led step to make shrink transparent and measurable, without slowing the checkout experience.

The German discounter, part of the EDEKA Group, has just won the reta award 2026 in the Artificial Intelligence category for its deployment of AI-powered computer vision from Trigo, recognised as Top Supplier Retail 2026. The key is in using innovative digital applications to increase transparency and efficiency in the daily working routines within their stores.

A familiar problem, compounded

The challenge of German grocery retail will resonate across the UK grocery industry. Labor shortages have accelerated self-checkout adoption, and Trigo’s internal findings suggest self-checkout concentrates a large share of avoidable loss, from operational mis-scans to deliberate scan avoidance. Add economic pressures driving both increased theft attempts and rising operating costs, and margins get squeezed from every direction.

The traditional response (more cameras, more security staff, blanket measures) ties up resources without solving the fundamental problem: retailers don’t have enough transparency into where losses occur.

“Hope isn’t a strategy,” says Trigo’s Dominic Brynolf. “At some point you have to pick something important to the business, build a use case, measure the ROI.”

Three layers of defense

Trigo’s system transforms existing CCTV infrastructure into an intelligent monitoring network—no new hardware, no store renovations. But the real shift is in what it watches for, working across three distinct layers.

The first validates the scan, catching operational errors, intentional fake scans, or mis-scans. For example, if a customer scans incorrectly and the barcode doesn’t register, a prompt nudges them to retry. The second verifies the basket by identifying items present at checkout but never scanned, products left in the trolley or sitting on the counter.

The third layer is the one that changes everything: identifying items concealed long before checkout. The bottle in the container. The product that slipped into a pocket three aisles ago. Losses that checkout-focused security was never designed to see.

“Retailers have been trying to solve shrinkage at the point of sale for years,” says Brynolf. “But the data is telling us the problem starts much earlier.”

What the data revealed

Shrink isn’t a self-checkout problem. Self-checkout is just where it becomes visible.

In Trigo’s broader analysis of commonly stolen items tracked from shelf to exit, 80 per cent were concealed before the shopper ever reached checkout, meaning they never appear in plain view at the scanning area.

Once concealment becomes detectable as a pattern (where it happens, when it spikes, and which categories it targets), prevention stops being a blanket response and becomes a targeted operating discipline.

Subscribe to Grocery Gazette for free

Sign up here to get the latest grocery and food news each morning

From guesswork to evidence

When shrink becomes measurable, it stops being a debate and becomes an operating discipline. Instead of competing theories across departments, leaders get a shared view of what’s happening, where it’s happening, and how it changes over time.

And instead of asking teams to watch everything, the system highlights the few moments that need attention, so people can act with context, not suspicion.

What comes next

For retailers watching shrink erode margins, computer vision offers a way to augment existing systems, surface previously invisible intelligence, and enable targeted intervention instead of costly blanket measures. The technology exists. The ROI is measurable. What’s required now is the clarity to start and the discipline to scale what works.