Frictionless checkout is rapidly becoming the future of retail, especially with Amazon.com's push to expand its checkout-free “Amazon Go” stores, which leverage digital imaging to redefine the checkout experience. We can expect to see increasingly fewer cashiers, registers and lines as more retailers turn to frictionless checkout to improve efficiency, convenience and satisfaction.
While retailers are still assessing the various types of risk associated with frictionless checkout technologies, they can look to the most common form today for inspiration: self-checkout. While self-checkout is a critical component of the modern retail operation, retailers have had to find new solutions to protect their assets from risk in this area. Prescriptive analytics — the artificial intelligence- and machine learning-powered methodology that sends the right information, to the right person, at the right time — is one proven solution. Here are some ways I’ve seen retailers leverage prescriptive analytics to mitigate risk around self-checkout:
Five-Finger Family Discount
Often, retailers will leverage self-checkout attendants to mitigate risk. However, this strategy can backfire. Even measures designed to mitigate risk can become a source of risk themselves. It’s important to augment humans with prescriptive analytics to maximize visibility.
This was seen in the case of a general retailer that adopted a prescriptive analytics solution. The retailer had long relied on self-checkout attendants, but decided to test prescriptive analytics too to see whether the attendants were actually effective.
Less than 24 hours after going live, the solution alerted an asset protection (AP) manager to a store in her district with a higher amount of self-checkout markdowns than similar stores. Interestingly, the analysis also found that nearly all of the suspicious markdowns had occurred when a recently hired self-checkout attendant was on duty. The solution sent the AP manager a prescriptive action that included CCTV footage from the times the markdowns were scanned, plus directives to interview the employee.
The CCTV footage showed several self-checkout customers removing “75 percent off” markdown tags from Halloween candy and reattaching them to expensive energy drinks, teeth-whitening strips, and even premium-brand electronics. They did this right in front of the suspect attendant, who simply stood at his station and watched. Using the prescriptive analytics solution, the AP manager further identified through loyalty card data that most of the suspicious customers shared the employee’s last name. They were his family members, and this was clearly evidence of organized retail crime (ORC) activity.
The AP manager interviewed the employee about the results of her investigation. Although he initially denied involvement, he confessed when confronted with the CCTV footage showing him watching the fraud take place. The employee was terminated and the retailer pressed charges against him. Ultimately, the retailer recovered $10,000 in restitution.
Where’s the Beef?
A national grocer adopted a prescriptive analytics solution to ensure better margins and inventory accuracy. Soon after deployment, the solution’s machine learning and AI technology flagged an inventory anomaly. A specific store’s meat department had begun the week with just 250 pounds of chicken parts on hand; by Wednesday of that same week, records showed it had sold 505 pounds, with no new deliveries. The numbers just didn’t add up. The solution also identified in parallel that beef was moving slowly based on the store’s typical ship-to-sales ratio. Interestingly, another nearby store showed similar behaviors. A prescriptive action directed the store operations managers to check pricing-sticker accuracy at their respective stores, and for AP managers to interview the meat employees on duty over the past several days.
Several meat workers turned out to be colluding with a local caterer in an ORC ring. The caterer would come into the stores several times per week and order very large quantities of expensive beef cuts, like rib roasts or tenderloins. The colluding employees would attach price tags for chicken parts to the beef, allowing the caterer to purchase them at a fraction of their actual price. To avoid suspicion at the register when the beef rang up as cheaper chicken, the caterer used the self-checkout line each time. The caterer would later give the meat employees a kickback for their help.
The retailer pressed charges against all four involved employees and the caterer, ultimately recovering $90,000 in losses. It also updated its self-checkout procedures to mitigate future risk.
Going Bananas
Another large grocer adopted a prescriptive analytics tool to more closely monitor its product movements and behaviors. The system identified multiple stores showing monthly sales of more bananas than they had purchased. The solution performed root-cause analysis and traced most of these excess purchases to the self-checkout line, a common area of risk for grocers. The retailer’s supply chain manager received a prescriptive action informing her of the anomaly and directing her to investigate CCTV footage and self-checkout practices.
The manager found that self-checkout customers were entering the PLU code for bananas (4011) to ring up more expensive items like organic fruits, meat, olive oil, and detergent. The prescriptive analytics solution calculated that the grocer was losing around $8,500 per week. The losses stemmed from both the pricing fraud and the lack of replenishment for the products rung up as bananas, which caused widespread inventory and allocation inaccuracies. In turn, these inaccuracies negatively impacted the customer experience through poor on-shelf availability and increased waste for the retailer.
The prescriptive analytics solution recommended a fix for the problem, and sent it to the grocer's IT team. Per the prescriptive actions, self-checkout registers were reconfigured to loudly announce "Bananas!" whenever a customer entered the fruit's PLU code. This made it easier for self-checkout attendants to identify fraudulent uses of the code, and also served as a deterrent for future fraud. The fix worked, and the grocer saw an overall margin increase of 1.2 percent within one week of identifying the problem. Based on this success, the prescriptive analytics solution later sent another prescriptive action to the IT team, directing them to apply the high-volume announcement to all of its highest-risk products at self-checkout.
Retailers too often write off frictionless checkout risk as a “cost of doing business.” The reality is, they can in fact mitigate that risk with the right tools, such as prescriptive analytics. With the increased visibility and efficiency provided by prescriptive analytics, retailers can identify areas of highest risk and deploy the right countermeasures to mitigate it (e.g., launching investigations, requiring loyalty membership for certain activities, etc.). Doing so will leave them better positioned to improve their sales and margins and ensure future success with frictionless checkout.
As general manager of Zebra Analytics, Guy Yehiav is responsible for setting the organic and nonorganic growth, leadership strategy, and customer success for the Zebra Analytics business unit.
Related story: Getting the Most Out of Your Gen Z Workforce
Guy Yehiav is the president of SmartSense by Digi. He is a recognized thought leader in retail, CPG, supply chain, and complex manufacturing with a proven track record of success in M&A, B2B enterprise software solutions, SaaS metrics, and AI and IoT solutions. Guy most recently served as the GM and VP of Zebra Analytics. He supported the overall AI, machine learning, and analytics strategy by driving M&A, and the development of B2B enterprise solutions.