How to Cut Theft in Stores with Facial Recognition

cut theft

Security guards can’t catch every shoplifter, but modern facial recognition technology can.

Retail theft is a huge problem for stores. According to one survey, the cost of retail crime to US stores topped $50bn in 2016.

The most obvious cause of retail theft comes from shoplifters, who specifically target shops to take products they either use themselves or sell on. While organised shoplifting gangs do exist, most offenders aren’t professional thieves. They don’t fit a typical profile either, with both men and women equally as likely to shoplift, and 75 per cent of those are adults.

Preventing theft, reducing losses

Shockingly, shoplifters are typically only caught once out of every 48 times that they steal. It’s a statistic that demonstrates just how easy it is for shoplifters to get away with their crimes and how hard it is for security guards to do their jobs effectively.

The good news is that facial recognition software is helping shops and businesses fight back. As someone who starts shoplifting is likely to continue doing so, spotting a known shoplifter (and having security guard deal with them quickly) can prevent theft and reduce losses.


cut theft
Facial recognition technology can scan faces quickly, enabling stores to spot known criminals.

Thanks to the increased power of today’s facial recognition systems, cameras can scan faces and match them to images held in a database almost instantly. Using blacklists, any suspected or proven thief can be detected, with an alert sent to security staff. This allows stores to either remove the person (if they’re banned from the shop) or monitor them to see if they can be caught in the act.

The key to catching shoplifters this way is a reliable database. Photos provide a good starting point, perhaps shared by the police or other shops. But camera footage can also be scoured, with profiles constructed from anyone caught stealing on camera.

Training facial recognition systems

According to the Guardian newspaper, some US stores are giving shoplifters the opportunity to have their photograph taken rather than be arrested. This photo is then used to train the facial recognition system, helping to ban the offender and to prevent theft in the future.

The appeal of a facial recognition system is that it can be trained to spot and remember thousands of faces, far outperforming what human guards are capable of. Thanks to agile cloud technology, multiple branches of a store can use the same database, so a shoplifter caught in New York would have trouble entering a store in California. Likewise, disgruntled ex-staff can be added to a blacklist, so that security staff are warned should they return.

cut theft
Spot a thief on CCTV, then ban them from stores using powerful facial recognition ‘blacklist’ functionality.

As good as these systems are, they currently work by blacklisting known shoplifters. But what about those thieves who haven’t been caught yet? Or those who prefer ‘softer’ crimes, such as self-checkout fraud or getting a friend (who’s a member of staff) to skip over scanning the occasional item?

AI and object recognition are improving fast to combat theft in these areas too. Systems are already being developed to spot the cashier that doesn’t scan an item, or the self-checkout user that enters the wrong information to pay less for a product. At the high-end, emotion detection can highlight suspicious behaviour, even before a crime is committed.

An alternative to these types of casual theft is to prevent them altogether using AI. Inside the Amazon Go shop, for example, cameras are used to track what people pick up and put in their bag, forgoing checkouts all-together. Everything you take from a shelf is identified and billed to your Amazon account, without the need to scan anything.

Facial recognition and object detection technologies are going to be crucial in spotting criminals and preventing theft. Find out how the VIA Smart Access Control System is already helping stores to combat shoplifting by clicking here.

VIA Technologies, Inc.