Retail fraudsters are always coming up with new ways to sneak unpaid merchandise out of stores, making it hard to proactively mitigate losses. However, if you’re waiting until a cycle count reveals missing merchandise to react, it’s going to be impossible to prevent any loss. By then it’s too late.
Therefore, we must be looking at new ways to get ahead of these fraudsters. How can we see them coming, or at least identify who they are so we know when they’re in the store?
One thought: We could figure out what they DO buy so it becomes easier to spot their fraudulent activities.
Consider the so-called “banana scheme” at self-checkout, whereby a customer weighs something expensive at the price of bananas. Many of us have reports set up to flag transactions involving, say, excessive quantities of bananas or bananas scanned separately throughout an order. The principle here is that, although the fraudster is technically bypassing the point of sale with the stolen merchandise, they’re not avoiding the point of sale. In fact, they’re using it as a mechanism to cover their tracks. What they don’t know is that every time they try to make the fraudulent activity look like a legitimate transaction, they are creating data that we can analyze to identify the suspicious behavior.
The same general principle applies in other cases as well. Here are two prime examples of how we can analyze the data that we do have—their purchasing behavior—to make up for the data that we don’t: their other theft tactics.
Items Available Near the Register
Here’s a question: Have you ever gone to the grocery store to buy something? Assuming you have, here’s another question: have you been to the grocery store consecutive times, each time buying only candy and lip balm?
Probably not. Even if you are a creature of habit and you make small convenience purchases at the grocery store because it’s on your daily route, it’s likely that you’ve made at least one major grocery run in between those buys, breaking up the pattern. That’s the logic behind a set of algorithms we have deployed for many retailers using the Zebra Prescriptive Analytics solution. Whenever a customer or cashier shows a pattern of transactions involving items that are available in the queue line or at the cash wrap, the algorithms send an alert to the relevant employee (typically a loss prevention investigator), along with course-corrective actions. They may be told to review transactions, check CCTV, or complete other specific investigative tasks.
Remember, the cashier could be complicit, too. As you can imagine, the implication is that the customer or cashier is ringing these easy-to-grab items to generate a receipt, which they hope will make them look less suspicious when walking out with other stolen items. Recently, a pattern just like this helped one of my customers uncover a shopper swiping a couple packs of gum at self-checkout before walking out with cartloads of diapers, baby formula, energy drinks, and other products.
Low-Dollar Items and Transactions
This one is tried and true. Retail fraudsters love to scan the cheapest items they can get their hands on, whether to get a receipt to appear less suspicious, or to cover up the barcode of a more expensive item. There’s good news, though. Using statistical benchmarks, it’s easy to configure an advanced analytics solution like prescriptive analytics to identify when customers or cashiers are manipulating transactions to lower item prices or tender totals.
For example, one of my grocery customers recently used its prescriptive analytics solution to resolve an employee theft case. The solution alerted a loss prevention investigator to several transactions, each totaling just 15 cents, sending along CCTV footage related to the incident for on-the-spot review. In just a few minutes, the investigator could see that a produce clerk had created a 15 cent price sticker and was using it repeatedly to buy expensive meat items. The items he was stealing increased in value with each occurrence, meaning the fast alert was critical to minimizing losses.
To increase the number of such positive outcomes, many of my customers will combine these low-dollar triggers with other elements, such as:
- Transactions occurring at self-checkout
- Transactions that take an exceptionally short time to execute
- Combinations of the same cashier and customer loyalty account
- Low-dollar transactions that started with several items that were later voided
By properly leveraging your data in a flexible solution like Zebra Prescriptive Analytics, you can use a fraudster’s efforts to cover their tracks to track them down more quickly and stop them from continuing to cause loss.
For more information on the use of AI-powered analytics to stop fraud, ping me anytime at scott.pethuyne@zebra.com or visit www.zebra.com/zpa.
