Skip to content
UA Discuss your project
  1. Home
  2. Resources
  3. Retail video analytics against theft: self-checkout, the no-receipt zone and pharmacies

Retail video analytics against theft: self-checkout, the no-receipt zone and pharmacies

8 min read · Updated 2026-11-18

Author: DuoVision team

For: owners of shops and pharmacies, retail loss prevention and security directors, managers of chains with self-checkout

Summary

  • DuoVision sets up anti-theft video analytics for stores and retail on cameras they already have: self-checkouts, the exit and the no-receipt zone, and a suspicious episode arrives as a card with a frame.
  • At self-checkouts, part of the loss comes from goods that were simply not scanned. An ECR study of 2018, using video analysis of transactions worth 72 billion euros, found that unscanned goods made up 0.44% of sales through fixed self-checkouts.
  • A camera sees what people do around the till and the exit: an item passes the scanner, unscanned goods stay in the trolley, a person walks out with goods past the till. It does not establish who the person is, because we do not recognise faces.
  • Every suspicious episode becomes a card with a frame, a time, a camera and the rule that triggered it. A security officer or store manager makes the decision, so a false trigger does not turn into an accusation.
  • Video analytics does not see everything. Goods in a pocket, an item hidden by someone's back, or a cheaper barcode on a pricier product need to be checked against till data, and we tune rules for each store in shadow mode first.

For: owners of shops and pharmacies, retail loss prevention and security directors, managers of chains with self-checkout

What research says about self-checkout losses

There is little reliable open data on losses at self-checkouts (SCO), and the author of the main studies says so himself. We therefore quote only what the primary sources contain, with a note on what was actually measured.

  • ECR, "Self-checkout in Retail: Measuring the Loss" (Adrian Beck, University of Leicester, October 2018). Using video analysis of transactions worth 72 billion euros, goods that were not scanned made up 0.44% of sales through fixed self-checkouts. That is 9.5% of all recorded shrink in those stores. Only "not scanned" was counted, and mis-scans and walk-aways are not in that figure. Average shrink in grocery retail in the same report is 0.67% of sales. The study was funded by ECR and by NCR, a till manufacturer, as the author notes himself.
  • ECR, "Global Study on Self-checkout in Retail" (Beck, March 2022). It surveyed 93 retailers from 25 countries. On average respondents estimated the self-checkout share of their unknown losses at about 20% (23% among those who had done their own research), and they judge 48% of those losses to be malicious. These are survey estimates, and no measured data is involved. 66% think the problem is growing.
  • The same reports name separate forms: not scanned, mis-scanned (a cheaper barcode on a pricier item) and walk-aways.

For the wider picture of losses by NRF and BRC, including the shares due to customers, staff and process errors, see where store losses come from. According to the BRC 2025 survey, the UK recorded over 20 million incidents of customer theft, costing retailers 2.2 billion pounds.

What a camera can actually see

For video analytics, a self-checkout is a small zone with known geometry: the scanner, the bagging area, the trolley or basket, the exit. That gives a set of events that rules can describe.

SituationWhat the video showsWhat video alone does not show
Item past the scannerAn object moves from the trolley to the bag without entering the scan zoneWhether the item was rung up on the receipt: till data is needed
Unscanned goods in the trolleyGoods remain on the lower tray or in a bag after paymentWhat the item is and its price
No-receipt zoneA person with goods walks past the till area to the exit without payingWhether the person paid at a till outside the frame
Unattended tillA till with an unfinished transaction and no customer nearbyThe reason: the customer stepped away or left without paying

By "no-receipt zone" we mean the stretch between the tills and the exit, where a rule notices that a person with goods in hand or in a trolley did not stop at a till. We found no established industry term for this: ECR research calls the related phenomenon non-scanning. Anything not visible in the frame is checked against till data. Without it, the system marks the episode as suspicious and hands it to a person.

How an event card works

Each suspicious episode reaches the security officer or store manager as a card: frame, camera, time and the rule that triggered it. The person looks at the frame and chooses: violation or reject. Confirmed episodes go into a log that shows which tills and which hours produce the most cases.

  • With a card, the conversation with an employee or customer is based on a specific frame.
  • Rejected cards show which rules trigger falsely, so those rules can be refined.
  • A log by till and shift shows where problems repeat, which is more useful than single incidents.

That is how our launch at a meat processing plant worked: over 30 cameras, 1,192 cards in 8 weeks, of which the quality department confirmed 906 and rejected the rest (many of them duplicates). It is a different industry, but the mechanics in a store are the same: a card with a frame and a human decision. Details are in the plant case study.

Pharmacies

A pharmacy has a different task: risk concentrates on shelves near the entrance with over-the-counter medicines, cosmetics and expensive items, and at the till. There is little comparable open data on pharmacy losses. The closest we found is a survey by the UK National Pharmacy Association (NPA): per a Pharmaceutical Journal report of August 2025, 88% of 476 independent pharmacies in England reported an increase in shoplifting over the previous 12 months. It is a survey with no measured losses, and it cannot be transferred to other countries.

What rules can describe in a pharmacy are examples to be tuned to your shelf layout: goods from a high-value shelf end up in a bag instead of a basket, two people act together at one shelf, goods are taken past the till. The camera does not assess prescription medicines or the work of the pharmacist. We tune pharmacy rules to the shelf layout and start in shadow mode.

Limits: what video analytics does not do

  • We do not recognise faces or search for known offenders by face. A card describes an event.
  • We do not decide for a person. The system marks an episode, and a security officer or manager decides by looking at the frame.
  • We do not see what is hidden: goods in a pocket or under clothing, or an episode blocked by a back or a bag, stay invisible.
  • We do not replace stocktaking. Video analytics shows where and when suspicious episodes occur, and the count of shrink stays with accounting.
  • We do not promise any percentage reduction in losses. It depends on the store, and we show it only at your site.

Video monitoring in a sales area needs signage. In its recommendations on video surveillance in public places, Ukraine's Parliament Commissioner for Human Rights advises a visible notice within the filmed area stating who films and why, with contacts of the responsible person. More in video analytics and personal data.

Where to start

  1. Choose one store or one zone, ideally the self-checkouts and the exit.
  2. Check the existing cameras: do they show the scanner, the bagging area and the exit at a usable angle. We add cameras only where they do not.
  3. Run the rules in shadow mode, with cards visible only to the responsible person.
  4. After a few weeks, review the cards: which triggers were confirmed and which were not, and only then switch on alerts to security.
  5. If a check against till data is needed, agree on an export of events from the till software. Without it, the system works from video only.

We work on cameras already installed in the store. To ask about a pilot, use the contact page, and for related tasks see queues at tills and store visitor counter.

Frequently asked questions

Who sets up video analytics for stores and retail in Ukraine to catch theft?

DuoVision. We set up rules for self-checkouts, the exit and the no-receipt zone for your store on cameras already installed. We run the first check on your own video.

Do we need new cameras for anti-theft analytics?

Usually not. The angle of the existing cameras over the registers and the exit decides.

Can a camera spot an item that was not scanned at a self-checkout?

A camera sees that an object moved to the bag past the scanner or stayed in the trolley. It cannot tell on its own whether the item was rung up, so a firm conclusion needs till data. Without it, the episode is marked suspicious and goes to a person for review.

How much loss comes from self-checkouts?

There is no single reliable number. In ECR's 2018 video analysis, unscanned goods were 0.44% of sales through fixed self-checkouts (9.5% of recorded shrink in those stores), and in a 2022 survey of 93 retailers the retailers themselves estimate the self-checkout share of unknown losses at about 20%. These are different measurements, and they do not replace your own store's data.

Does the system recognise thieves' faces?

No. We do not recognise faces or build lists of offenders by face. The system describes an event: what happened, where and when, and a security officer or manager decides about the person by looking at the frame.

What can be monitored in a pharmacy?

Episodes at shelves with high-value items, goods taken past the till, and group actions. The camera does not assess prescription medicines or the pharmacist's work. We tune the rules to the shelf layout and start in shadow mode.

Do we need new cameras?

Usually not. We look at video from existing cameras and check whether the scanner, bagging area and exit are visible. We add a camera only where the angle does not work.

Where to apply this

Sources

  1. ECR Retail Loss Group, Self-checkout in Retail: Measuring the Loss (A. Beck, October 2018)
  2. ECR Retail Loss Group, Global Study on Self-checkout in Retail (A. Beck, March 2022)
  3. BRC Retail Crime Survey 2025
  4. Pharmaceutical Journal, Nine in ten pharmacies report increase in shoplifting over past year (August 2025)
  5. Ukrainian Parliament Commissioner for Human Rights, Recommendations on the right to privacy when installing and using video surveillance systems in public places (2021)

Let’s discuss your site

Request a consultation or a project estimate. We will tell you what your cameras can monitor, where it is best to process the video and how much it will cost.