Few retail criminals can resist the allure of the self-checkout (SCO) lanes, where a lack of cashiers presents a golden opportunity for fraud. One of the most popular types of SCO fraud is price switching.
The most severe cases of price switching involve the perpetrator entering the produce lookup (PLU) code for an extremely cheap produce item, like bananas ($0.49 per pound) or potatoes ($0.99 per pound), placing a much more expensive product on the scale and paying the low amount “scanned” instead of the latter product’s SKU. If this behavior goes unnoticed (and it often does), the customer is free to walk out with expensive item(s) purchased at a significantly reduced price. Even if the SCO attendant notices the activity, proving it to be fraud versus a simple, honest mistake can be very difficult. Even scale monitoring would be unable to diagnose the issue since the weights are aligned. It is the product that is wrong.
Machine learning provides a solution by monitoring transactional behaviors. By setting a machine learning-powered analytics solution like prescriptive analytics to monitor unusually low-value SCO orders, you can quickly identify any suspicious activity and take appropriate action. The right solution will immediately recognize any telltale data behaviors that suggest fraud and alert you to take appropriate action before losses increase.
As an example, one of our grocery customers deployed its analytics solution to flag SCO transactions with multiple low-value, sold-by-the-pound produce items. Most customers buying bananas, for instance, will place all the bananas in their carts on the scale and scan them in as a single “line item.” A fraudster, on the other hand, may be scanning multiple products as bananas, and thus would have the item on multiple lines in his transaction. For good measure, the solution was also configured to flag customers who appeared to be buying large quantities of bananas frequently. Think about it – do you typically buy bananas every day or only a couple times per week as needed?
The solution quickly detected a customer who was making three weekly SCO transactions, each containing four separate purchases of bananas. Surveillance footage provided by the solution showed the customer wait for the SCO attendant to get distracted, then enter PLU code 4011 for bananas ($0.49 per pound), while placing items like beef tenderloin (which alone was $28.99 per pound), organic fruit, energy drinks and olive oil on the scale instead. Thanks to the alert, the customer was flagged and audited the next time he came to the store. He was caught trying to ring up racks of lamb ($30 per pound) as bananas.
Checking his old transactions over the past three months, the retailer’s LP found the customer had stolen a total of $2,300 worth of product. Unable to dispute the evidence the analytics solution provided, he later paid full restitution. $
(Editor’s Note: Guy Yehiav is General Manager of Zebra Analytics and former CEO of prescriptive-analytics provider Profitect Inc. Reach out to his team at fran@zebra.com for more information on the use of prescriptive analytics in retail.)
