Avoiding Scan Avoidance With Computer Vision

Retailers lose $14 billion annually in “scan avoidance” — that is, when items are bagged at the register without ever being scanned or paid for. That’s according to the National Retail Federation, and, scarily enough, it actually represents a decline since the height of theft at the register in 2015.

Scan avoidance can come in a variety of forms. It can be accidental or deliberate, committed by customers or by employees themselves. But one thing is for sure: For retailers, especially ones with mass merchandise, high volume and thin margins (like a grocery store), $14 billion hurts. A lot.

“Scan avoidance and ‘sweethearting’ have traditionally been the Achilles heel for retailers in terms of asset protection, because they leave no data trail,” said Malay Kundu, CEO and founder of StopLift.

Sweethearting is when someone inside the organization deliberately passes an item over the scanner without ringing it in, typically so a family member or friend can get the item for free.

Of course, sometimes it’s just negligence: For instance, the cashier didn’t feel like typing in the code for an item that wouldn’t scan properly, so she gave it away for free.

And shoppers can be even more negligent at self-checkouts, where, if the system doesn’t work right, they may feel justified in simply taking an item without paying. Other times, they just forgot the cat litter was on the lower rack of the shopping cart and walked out without paying for it.

Kundu said that, when an item is lost due to either deliberate or accidental theft, the retailer often has to sell 50 more of the same item to make up the difference because profit margins are so thin.

Identifying when scan avoidance is happening can be nearly impossible with traditional tools, said Kundu. Again, this type of theft leaves no data trail because the item is never scanned, so the store can’t tell which employee is doing it (in the case of sweethearting) and which customers are thieves versus which ones just made an honest mistake or got frustrated with a malfunctioning system.

That’s why Kundu believed a tool external to the system was needed to reveal when people were bypassing the scanner. And that’s what he founded StopLift to deliver.

The tech company provides artificial intelligence (AI)-powered computer vision technology that compares items that have been scanned to items that have been bagged to help identify items that have been missed in either a self-checkout or manned checkout scenario.

When the items in the bagging area don’t match those scanned into the system, Kundu said an employee is notified via a smart device worn around the wrist, which vibrates when an issue is detected and can even replay a video of the incident so the employee has contextual knowledge when he approaches the customer.

Kundu said the goal is never to accuse a customer of stealing at a self-checkout, but simply to let them know an item may not have scanned properly and to help them ring it through the right way.

He said even if the scan avoidance was deliberate, and the shopper knows it and the employee knows it, there’s no need to make an accusation of theft. The store gets paid for the item this time, and the thief now knows this is not a place where he can get away with bagging unscanned items.

Kundu said computer vision can also reveal when customers have made an honest mistake, or when they’ve set off an alert with an action that isn’t theft, such as placing a purse on the weight sensor in the bagging area or holding a bag open while a partner fills it with legitimately scanned items.

In systems that don’t leverage computer vision, said Kundu, those actions would freeze the transaction until the customer moved the purse or put the bag down. But with computer vision, the system knows to keep functioning because everything is just as it should be.

Kundu believes AI at the register will soon go the way of digital music. In the early days, services like Napster existed to provide digital copies of songs illegally, but people didn’t mind that it was piracy, because downloading music from the internet was convenient. Then Apple came along with a way to get music online without breaking the law, and customers were happy to pay for the convenient, legal service.

In the same way, Kundu said, the transition toward self-checkout lanes is inevitable in retail, so it’s critical to button up the system against abuse — and to cut back on unwanted interventions to make it more friendly and usable to the average, honest shopper.