How AI is helping reduce Shrinkage

Vinay Vaswani
Oct 13
5 min read

Updated: Nov 3

Retail has always been a game of thin margins. And one of the line items that constantly eats away at retailers' margins is shrinkage. However, it attracts far less attention than most other aspects including those that come into more prominence such as tariffs etc.

At about 1.5% of sales, shrinkage translates into €30-50 billion in losses across European Retailers - every euro straight off the bottom line.

What is shrinkage exactly?

Shrinkage broadly comprises three components:

Theft, which accounts for roughly 55-60 % of shrinkage losses, both external and internal. Shoplifting and organised crime drive much of this, but around 29 % comes from employees.
Spoilage and waste is the second component, which contributes about 20-25 %. In food retail, this is especially prominent, where supermarkets throw away 2.5 - 4 % of potential revenue as surplus goods since they go bad.
Administrative errors make up the balance, surprisingly large at around 20 %. Mis‑counts, pricing mistakes and other “paper shrink” quietly eat away at stock.

Retailers have tried to solve this problem for decades

Shrinkage is not a new problem for Retailers. And they’ve tried to solve each of the three key components with solutions throughout the industry. They’ve deployed RFID tags, CCTV, staff training and inventory systems for years. Loss prevention has its own department in many chains.

Yet shrink remains stubbornly high - partly because theft and error adapt as quickly as the defences. Cost‑of‑living pressures have driven a surge in shoplifting across parts of Europe, and organised retail crime rings treat self‑checkout terminals as low‑risk, high‑reward opportunities. Something more dynamic is needed.

Enter AI, the new nemesis of shrinkage

Artificial intelligence doesn’t replace security guards or store associates; it augments them. Machine vision and pattern‑recognition models watch every transaction and every aisle, spotting anomalies that humans overlook and sending alerts before losses occur. Early deployments show promise.

There are various AI use cases doing the rounds, trying to battle the shrinkage problem.

AI‑powered smart checkout: French grocer Intermarché turned to AI after noticing mis‑scans and barcode swaps at self‑checkout terminals. Cameras and weight sensors verify that the scanned item matches what goes into the bag. The results have been dramatic: mis‑scans have fallen from 3% to under 1%, and inventory discrepancies at self‑checkout have dropped by up to 75%. Notably, the system reduces the need for human intervention, speeding up queues and improving customer experience.

AI‑enhanced CCTV analytics: Masses of video footage often go unwatched. In February 2025, Tesco opened a 24/7 security hub that uses pattern recognition to analyse thousands of hours of CCTV from across its network. The system flags repeat offenders, unusual behaviour around high‑value products and suspicious interactions. Considering UK retailers lose about £2 billion a year to customer theft, the opportunity is enormous. Rather than relying on individual store teams to spot issues, Tesco centralises intelligence and coordinates responses in real time. The result is faster intervention and, importantly, a safer environment for staff and shoppers.

However, the question of privacy always plays in this part, especially given Europe and its regulators' view towards it.

AI-driven instant alerts: Laurel Ace Hardware, a modest US retailer, installed AI‑equipped cameras that detect concealment gestures and send instant alerts to staff. Losses dropped by about 50 %, proving that even independent stores can leverage cutting‑edge tech.

At the other end of the spectrum, Amsterdam’s Schiphol Airport deployed an AI‑powered theft detection system that recovered more than €163,000 in retail sales between April and September 2024. Despite an existing suite of anti‑theft tools, the airport retailer found that AI could identify suspicious behaviour in busy travel environments and notify staff before items left the premises. As the system learns from each detection, false alerts fall and accuracy improves.

Inventory optimisation algorithms: Shrinkage isn’t just about theft. Ordering too much or too little stock also destroys value. Swiss cooperative Migros implemented AI‑driven forecasting models that predict demand for every product, at each store, for each day. The adaptive system factors in promotions, seasonality and local events. After adoption, inventory days fell by 11 %, product availability rose 1.7 % and lost sales decreasedi. By aligning supply with demand, Migros reduced waste and avoided stockouts - a double win for margins and customer satisfaction.

How does the business case look on AI in shrinkage?

Several forces make AI‑driven loss prevention more than a nice-to-have. First, shrinkage is rising. Organised crime rings are more sophisticated, and economic pressures drive opportunistic theft. Second, technology costs are falling. Cloud computing and computer vision are accessible even to mid‑sized chains. Third, the stakes have never been higher: with margins compressed by energy prices and wage inflation, a 0.5‑point reduction in shrink can mean the difference between profit and loss.

Early pilots have shown that AI investments can pay for themselves quickly. For example, a Forrester study of one AI loss‑prevention vendor found a 374% return on investment and payback in six months. The same report noted a 15% improvement in staff productivity thanks to fewer manual interventions and clearer priorities. Early adopters like Intermarché, Tesco and Schiphol aren’t just testing technology; they’re proving that AI can recover millions of euros in revenue and free up employees to focus on service.

What should retailers do to reduce losses?

Like tariffs or supply chain disruptions, shrinkage is a strategic risk. Treat it with the same rigour:

Map your shrinkage hot spots. Use existing data to pinpoint where and when losses occur. High‑risk categories and high‑traffic times are ideal starting points for AI pilots.
Start with quick wins. Self‑checkout terminals, alcohol aisles and cosmetics counters tend to see high loss rates. Target these with smart checkout or real‑time alert systems.
Integrate, don’t rip‑and‑replace. Leading AI platforms layer onto existing CCTV and POS infrastructure. That means you can test without massive upfront capital expenditure.
Prepare your people. AI is a tool for the front line, not a replacement. Train staff to respond to alerts with empathy - sometimes a gentle reminder is all that’s needed. Involving employees also builds trust and surfaces edge cases the AI might miss.
Measure and iterate. Track shrink before and after implementation, as well as intervention times and customer satisfaction. AI models improve with feedback; retailers who share data with vendors will see faster gains.

Looking ahead

Shrinkage is unlikely to disappear. But the narrative is shifting from “unavoidable cost” to “controllable risk.” Europe’s retailers have already proven that AI can reclaim millions of euros by catching theft, reducing waste and preventing errors. As technology evolves, expect AI to move beyond detection into deterrence - for example, prompting price adjustments in real time when shelf‑life dwindles or predicting fraudulent returns before they occur.

The underlying message is simple: shrinkage is a margin killer, but AI offers a lifeline. If your strategy hasn’t yet addressed this silent P&L drain, now is the time. With regulators tightening data rules and consumers demanding seamless experiences, the retailers who succeed will be those that weave AI into their operations responsibly and transparently. The question is no longer whether AI can reduce shrink. It’s how quickly you can scale it before lost margin becomes lost opportunity.

If you would like to share your opinion on this topic or clarify any questions, please feel free to write to vinayvaswani@strivo.nl

Written by Vinay Vaswani for Strivo B.V (edited and source checked by AI)

Sources: