What problem does it solve?
Self checkout moves the scanning from a cashier to the shopper, and items go unscanned, by mistake or on purpose. The Grocer reports that in June 2026 the largest study of self checkout loss to date by ECR Retail Loss, covering millions of transactions across 39 retailers, found losses 33% higher in stores with self checkouts than in stores without.
The usual defenses have a cost of their own. An attendant at every lane gives back the staff savings and a receipt check at the exit slows every shopper. An accusatory alert on a shopper who made an honest mistake loses goodwill. Retailers want to catch the item without treating the customer as a thief.
- The Grocer reports that a 2026 study by ECR Retail Loss, based on millions of transactions across 39 retailers, found losses 33% higher in stores with self checkouts than in stores without.Morrisons adds self-checkout AI tech to detect mis-scans (2026)
- The Grocer reports that the same ECR Retail Loss study found customers self correct in more than 80% of cases once an on screen prompt asks them to check or rescan items.Morrisons adds self-checkout AI tech to detect mis-scans (2026)
How does it work?
- Watch the lane. A camera above the self checkout, sometimes with a second camera facing the shopper, records the items placed on the scanner and in the bagging area. Computer vision identifies the items the shopper handles.
- Compare with the scan. The system compares what it sees with the point of sale data to find items that were not scanned, scanned as something else, or keyed in wrongly on the touchscreen.
- Nudge the shopper. On a mismatch the screen shows a message, in Tesco's trial a replay of the video, and asks the shopper to rescan. Most shoppers fix the error themselves.
- Escalate when the nudge fails. After a second failed scan it alerts a store associate on a mobile device to check the lane.
- Verify at the exit. At Sam's Club, cameras under an exit archway identify the items in each cart, a model matches the cart to a receipt and the result goes to an exit associate in real time, replacing the manual receipt check.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Kiosk and branch, Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Automation rate | Too few to pool | at least 75% | 1 | 1 organization |
Value drivers: Risk and loss reduction, Lower cost to serve, Customer experience.
Indicative value
A grocery chain with 200 stores and self checkouts
USD 112,500 to USD 10 million
Annual self checkout loss recovered per year
How this is calculated
Formula: stores * selfCheckoutSales * lossRate * lossRecovered. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Stores with self checkouts stores, stores | 150 | 250 | Illustrative size, similar to the 200 stores in Morrisons' initial rollout referenced on this page. |
| Sales through self checkouts per store per year selfCheckoutSales, USD per store per year | 3,000,000 | 8,000,000 | Editorial assumption, replace with your own self checkout sales per store. |
| Value lost at self checkout as a share of self checkout sales lossRate, fraction of sales | 0.005 | 0.02 | Editorial assumption, replace with your own measured loss. The ECR Retail Loss study reported by The Grocer found losses 33% higher in stores with self checkouts, which does not give an absolute rate. |
| Share of that loss recovered by prompts and checks lossRecovered, fraction of loss | 0.05 | 0.25 | Editorial assumption. Kroger reports that customers resolve over 75% of scanning errors themselves and The Grocer reports more than 80% self correction after a prompt in the ECR Retail Loss study, but neither is a share of total loss, which includes theft the system does not see. |
What it leaves out: Counts only recovered product value. It leaves out the cost of cameras, edge servers and licenses, the staff time saved on lane interventions and receipt checks, the effect of shopper annoyance on sales, and the share of loss that is deliberate theft the prompts do not stop.
Who already uses it?
4 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Sam's Club
United States · Retail and ecommerce · 2024
Sam's Club announced at CES 2024 a computer vision system at the club exit that replaces manual receipt checks. Cameras capture images of each cart as it passes through an exit archway, models identify the items, an algorithm matches the cart to an outstanding receipt, and the system sends the comparison of cart and receipt to an exit associate in real time. Sam's Club says the technology was live at more than 120 clubs by 30 April 2024 and that members leave the club 23% faster. The exit associate still verifies carts, and the system flags new items for labeling so the models keep learning.
No outcome disclosed.
Morrisons
United Kingdom · Retail and ecommerce · 2026
In August 2026 Morrisons announced a long term partnership with Everseen to roll out its Evercheck computer vision platform at self checkouts in 200 stores, after a successful trial. An overhead camera identifies when a shopper fails to scan an item or enters a product wrongly on the touchscreen, plays back a video of the error on screen and asks the shopper to try again, and a second failed scan alerts store staff. Morrisons and Everseen describe the aim as a fast checkout that frees colleagues for service. Neither discloses trial results.
No outcome disclosed.
Tesco
United Kingdom · Retail and ecommerce · 2025
In 2025 Tesco trialed cameras at self service checkouts that look at the shopper and down on the checkout from above and play back video of an unscanned item with the message that the last item was not scanned properly. Tesco said the system made checkout quicker and easier but did not say how many stores had it. The sources do not name who built the system. Shoppers mocked the replay as a video assistant referee and some raised privacy concerns.
No outcome disclosed.
Kroger
United States · Retail and ecommerce · 2020
Kroger signed with the Irish company Everseen in 2020 to put computer vision cameras on its self checkouts, with the aim of catching items that pass without being scanned. According to a Lenovo case study, which supplied the edge servers, the system analyzes checkout video together with point of sale data, flags a missed scan on screen so the customer can correct it, and alerts a store associate on a mobile device if the customer cannot. After a pilot, Kroger had deployed it at 1,700 grocery stores with plans for all locations. Kroger's senior manager of asset protection said that customers resolve scanning errors themselves more than 75% of the time.
- Automation rate: at least 75%, share of scanning errors, at the time of the case study
"In fact, over 75% of the time, customers are able to resolve scanning errors themselves, with no intervention from our associates, which makes their job a little easier."
Claimed by: organization
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Real time scan events from the point of sale or self checkout system, matched to lane and time
- Product images and assortment data, including produce and items without barcodes
- A baseline of unknown loss and manual intervention counts per store before launch
Systems to integrate
- Self checkout and point of sale software, for scan events and screen prompts
- Camera hardware and edge servers at the lane
- Staff mobile devices, for escalation alerts to associates
- Store loss prevention or incident systems, for review and reporting
Complexity: Medium
The build needs cameras at each lane and edge computing to process the video, as Kroger did with Lenovo edge servers, plus a live feed of point of sale scan data to compare against. Item recognition has to be trained on the retailer's own assortment before go live, and the store's operations and legal teams need to agree how staff are alerted and what is retained.
- 1
Trial in a few stores with a baseline
Measure unknown loss, interventions per lane and shopper satisfaction before launch, and compare trial stores with matched stores. Morrisons ran a trial before committing to 200 stores.
- 2
Prompt first, escalate second
Show a prompt and let the shopper correct the item before involving an associate. Kroger reports that customers resolve over 75% of scanning errors themselves.
- 3
Tune the wording and the replay
Test the screen message with real shoppers. Tesco's video replay drew mockery and privacy concerns, so decide whether a replay, a still image or a plain message fits your customers.
- 4
Decide what staff do with an alert
Write the rule for who responds, in how long and what they say, so an alert leads to a helpful check and not an accusation.
- 5
Extend to the exit or other lanes after the first results
Add exit verification or staffed lanes once the self checkout results are measured, so each step is judged on its own numbers. For scale, Sam's Club announced its exit technology at CES 2024 and reported it live at more than 120 clubs by April 2024.
Guardrails
- The system compares items with scans and does not identify or profile shoppers, with no facial recognition
- An alert is a request to check, never an accusation, and staff are trained in the wording
- Video is kept only as long as needed for the check and any dispute, with a stated retention period
- Signage tells shoppers that cameras are used at the checkout and why
KPIs to instrument
- Share of detected scan errors the shopper corrects without an associate
- Associate interventions per lane per hour, before and after
- Unknown loss as a share of self checkout sales, trial against matched stores
- Shopper satisfaction and abandoned transactions at self checkouts with the system
Human in the loop
The system prompts the shopper directly on the lane screen, and after that a store associate makes every decision that involves the shopper: a person checks the lane and talks to the customer. At Sam's Club an exit associate receives the cart and receipt comparison and verifies it. Loss prevention staff review flagged patterns, and no automated action blocks or penalizes a shopper.
Common failure modes
- Alerts that feel like accusations
- A flagged honest mistake damages goodwill, and reports of shoppers mocking Tesco's replay show the risk. Test wording, keep alerts private to the lane and track complaints.
- Misidentified items
- Produce in bags, similar packages and new products confuse recognition and create false alerts. Retrain on new assortment, measure false alert rate by category and fix the worst first.
- Savings measured on a self selected trial
- Trial stores differ from the rest. Use matched control stores and measure over a full season before projecting a chain wide result.
- Privacy concerns about cameras
- Cameras pointed at shoppers raise data protection questions. Check items and not people, keep video short term and explain it on signage and in the privacy notice.
What are the risks and rules?
EU AI Act
Depends on design
A system that compares the items a shopper handles with the items scanned does not appear in Annex III and carries minimal risk, though the retailer owes AI literacy under Article 4. The tier changes with the design: for a retailer, facial recognition or other biometric identification of shoppers, or inferring their emotions, falls under Annex III point 1 and is high risk, with Article 50(3) transparency duties for emotion recognition. Inferring emotions of store staff would fall under the Article 5(1)(f) prohibition. A design that identifies or profiles people needs a separate legal assessment.
Guidance
- Guidelines 3/2019 on processing of personal data through video devices (European Data Protection Board, Europe). Covers the lawful basis, transparency, retention and data subject rights for cameras in shops, which apply to checkout video.
- Regulation (EU) 2024/1689 (EU AI Act), Articles 4 and 5 and Annex III (European Union, Europe). Annex III point 1 covers biometric identification and emotion recognition and Article 5 prohibits emotion inference at work, which a checkout system based on items and scans avoids by design.
Controls to put in place
- A data protection impact assessment for camera use at checkouts, covering retention, access and signage
- Written rules for how staff respond to an alert, with training and complaint tracking
- Regular accuracy checks of item recognition by category, including produce and promotions
- Matched control stores for measuring loss and shopper satisfaction
When it went wrong elsewhere
- OECD.AI incident entry: Tesco AI 'VAR' cameras at self checkouts. Shoppers mocked the replay as a video assistant referee, some raised privacy concerns and a few refused to use the checkouts, as summarized by OECD.AI from press reports.
Frequently asked questions
- Does this replace the self checkout attendant?
Not in the evidence on this page. Morrisons says its system gently prompts the shopper and alerts staff after a second failed scan, and Sam's Club still has an exit associate verify the cart. The aim is fewer routine interventions, so staff handle the cases that need a person.
- What results have retailers reported?
Kroger's senior manager of asset protection said customers resolve over 75% of scanning errors themselves, in a Lenovo case study. Sam's Club says members leave the club 23% faster with its exit technology. Morrisons and Tesco have not disclosed trial results. None of the figures is a measured reduction in shrink, and all come from the retailers or their vendors.
- Is it legal to point cameras at shoppers at the checkout?
In the EU and UK, cameras that record shoppers are subject to data protection law, which requires a lawful basis, transparency, limited retention and a data protection impact assessment where the risk is high. The systems on this page are described as comparing items with scans and not as identifying shoppers, which is the design to keep, but the cameras, the signage and the retention period still need a legal review.
How to cite this page
Blits.ai AI Use Case Library, "AI loss prevention at self checkout and store exit", last verified 14 October 2026, https://www.blits.ai/ai-use-cases/self-checkout-and-exit-loss-prevention. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 14 October 2026: First published