With a spike in shoplifting incidents, retailers are turning to smart AI solutions by Neto Baltic and Agmis to prevent retail losses.
Walmart, Target, Foot Locker, Dollar General and other major retailer investor calls this quarter had one thing in common – the shrinkage, caused by increase in retail theft incidents – is getting too big to ignore. According to the ECR Retail Loss survey, the losses due to theft have increased by 33% compared to pre-pandemic levels. The British Retail Consortium states that the retail theft rate has grown by 28 percent.
In the Baltic states retailers are also reporting a threefold increase in retail theft related losses.
“These are the numbers that nobody wants to advertise publicly. Yet in our talks with clients this year, we can identify the spike in retail theft as a major challenge faced by retailers across the board” – says the CEO of Neto Baltic, a retail technology and loss prevention company, Rokas Budvilaitis.
How AI can reduce losses in self-checkouts
When shoplifters are increasingly going after such everyday products as cheese, meat, alcoholic drinks or cleaning products – they are also targeting self-checkout desks.
“Back in 2017 we started to talk about the loss prevention risks involving self-checkouts. Despite being equipped with AI-powered control scales, self-checkouts are not well proofed against different theft scenarios” – says R. Budvilaitis.
In 2020 Neto Baltic and technology company Agmis founded a new startup ScanWatch, which develops Computer Vision solutions for retail loss prevention.
One of ScanWatch products – Crime Predictor – makes use of security cameras in the checkout area and self-checkouts. By using AI and Computer Vision, Crime Predictor identifies scanned items at the checkout and checks them against the product images in its database.
Crime Predictor mitigates the risks when a cheaper bulk product (e.g. cucumbers) is selected in the picklist instead of a more expensive one (e.g. avocados). The product also prevents barcode switching, no scan or no pay fraud scenarios. When AI detects a discrepancy, Crime Predictor automatically notifies store security or a store employee.
“When 2% of all retailer self-checkout transactions may include intentional fraud or unintentional misscanning, we can safely prevent 90% of such incidents” – says CEO of ScanWatch Simas Jokubauskas.
AI boosts security in autonomous stores
ScanWatch products can be installed in self-checkout and manned checkout counters. In the latter example the product helps to prevent employee errors and employee fraud.
New product use cases are also trialed in autonomous stores.
“In today’s store we identify the checkout as a security barrier. EAS antennas and gates are installed at the checkout area. Yet with the advent of Scan&Go or autonomous shopping concepts, retailers need to rethink their article surveillance strategy” – tells R. Budvilaitis.
According to R. Budvilaitis, by utilizing in-store security cameras, shelf and checkout area monitoring, AI can operate as a holistic, ever present security layer all across the store.
“Where are testing a product use case, where every shopper is identified by a unique ID. We can trace what items the shopper has taken from the shelf and put into this shopping basket. This data can be cross-referenced with what items were paid for at the checkout” – says R. Budvilaitis.
The AI platform can be integrated with security gates or sound alarms. If all items were paid for, the gates open automatically. If not – an alarm is sounded and a security officer is notified.
Furthermore such use cases are fully compliant with GDPR requirements – no biometric data is used for shopper identification, the data is not stored as well.
“Retail security is a compromise between convenience and loss prevention. AI can already replace physical EAS tags, reducing the investments of the store owner and ensuring a smooth checkout experience for the customer“, says R. Budvilaitis.
AI for better self-checkout experience
Loss prevention is one of ScanWatch AI product usage scenarios. The product also helps to simplify self-checkout usage experience.
By automatically identifying bulk, unpackaged items – such as fruit, vegetables, bakery goods – the shopper no longer needs to look for those items in the checkout picklist menu.
“The shopper no longer needs to explore 3 levels of the picklist menu or – worse – use a text item search. AI can automatically identify products in a matter of milliseconds“ – says S. Jokubauskas.
More than half of grocery shoppers have at least one unpackaged product in their shopping basket. Manual picklist menu increases the shopper checkout time. It also prevents less technology savvy customers from using the self-checkout option.
Combining technology and retail know-how
ScanWatch is already employed by leading retailers in the Baltics, Poland, Germany and Switzerland. The product is making inroads into the North American market. ScanWatch development is based in Lithuania by leading experts in AI and retail.
“In 2018 we launched a dedicated startup focused on AI and machine vision product development. Retail was one of our key interest areas from the get go. In five years we have collected invaluable knowledge, which helps our products to operate seamlessly in real-world retail environments“, says CEO of technology company Agmis Saulius Kaukenas.
ScanWatch products were developed by joint effort from Agmis, Partner Tech Europe and Neto Baltic. Neto Baltic has more than 20 years of experience in the loss prevention sector. The deep understanding of the retail domain, combined with advanced technologies, are the main advantages of ScanWatch products.
“For the product to make a real-world impact, it must be simple to integrate with self-checkouts from different hardware vendors, the product database needs to be scalable and easy to update. The product needs to react in real-time and identify suspicious incidents in less then a second, minimizing the rate of false positives and accounting for different shopper behaviors“, concludes R. Budvilaitis.
More about ScanWatch here.