Shrink, whether it be through intentional acts, such as shoplifting, or simply as a result of human error, has historically been one of the biggest pain points for retailers. In fact, according to the most recent National Retail Security Survey conducted by researchers at the University of Florida in partnership with the National Retail Federation (NRF), retail shrink totaled $50.6 billion in 2018, representing 1.38% of sales for the year.
Aside from the use of good, old-fashioned manpower as a theft deterrent, one of the ways that retailers have attempted to curb shrink in their stores is by leveraging various security technologies but even the use of some traditional solutions, such as video surveillance, have fallen out of favor with some brick-and-mortar shops more recently. The NRF study found that only 22% of respondents reported using IP analytics in 2018, a decrease of 9.5 percentage points from the prior year’s survey and only 57% said they were using remote IP CCTV monitoring, a decrease of nearly 20 percentage points from the previous year.
Despite this, however, it should be noted that most retailers still see security technology as the answer to their shrink problems and that several them, particularly large big-box stores, are increasingly turning to solutions powered by artificial intelligence (AI) and machine learning tech. One company that has made significant inroads with such a product in the retail industry is Everseen. Based in Ireland, the company was started about 10 years ago with the goal of curbing shrink – shoplifting, sweethearting, etc. – in manned checkout lanes by detecting irregular activity patterns with the help of AI algorithms.
However, according to Huw Lloyd, Everseen’s Chief Strategy Officer, the company’s growth really began to accelerate when retailers began adopting self-checkout lanes in grater numbers. The “seenicself” loss prevention solution for self-checkout lanes began trialing in 2016 and the product is now live in about 3,000 stores around the world, including some of the biggest names in retail in both the U.S. and Europe like Walmart.
“Retailers have relatively little data on where their shrink actually happens. If you talk to a store manager, they will tell you what, in theory, turned up at the back of the store – they know that that’s not accurate as they are constantly at war with the supply chain – and then they know what went through the point of sale, but actually where the loss happened… they don’t have much data on that,” Lloyd explains. “What we do is actually shine a light and give accurate figures around the checkouts and the amount of loss that is happening around the checkout. Universally, I would say up until about six to 12 months ago, people when we ran a trial were always shocked at how high the size of the shrink was through checkout.”
In grocery store retail, for example, Lloyd says that shrink has typically been around 1.8% to 2% of revenue but that that figure is growing rapidly with the increased adoption of self-checkouts. Today, Lloyd says that 30% to 40% of grocery retail shrink occurs at the checkout – both manned and self-checkouts – representing 0.5% to 0.8% of revenue.
How It Works
While the loss prevention industry has taken a number of different approaches to reducing shrink through the years, Lloyd says Everseen’s philosophy has been to focus on catching incidents as they occur at the point of sale with as minimal amount of false positives as possible while simultaneously offering the best customer experience via a computer vision AI system.
“We take a camera feed over a checkout and then wrap that around an AI platform. We also feed the point of sale (data) through that platform and look for irregularities in that process,” he explains. “Initially we started just with what we call the classic non-scan and that was a product passed over the scanner bed but nothing (registered). That caught a lot of the irregular activity and I think it is important to note that some of this is accidental and some of it is deliberate.”
Once a suspicious transaction occurs, the system plays a GIF for employees monitoring the self-checkout lanes who can then talk to customers about a possible mis-scan of an item and subsequently ask them if they want to rescan it. The GIF of the mis-scan can also play on the screens of the self-checkout themselves, which serves as an increased deterrent. “The customer, although they’re not accused of by the self-checkout assistant of stealing, still has to stand there and watch a video of what they’ve just done and for most people that sort of shames them into good behavior,” Lloyd adds.
Lloyd says the company has now moved well beyond catching irregular transactions to looking for additional avenues of shrink in checkout lanes.
“There are some more advanced use cases and we’ve now moved on to extending the area we look at; for example, people in both manned lanes and self-checkouts that don’t even go near the checkout (with merchandise) – they either leave it in the cart or they drop products straight from the cart or basket to the bagging area,” he adds. “The most advanced and perhaps most nefarious act that happens is when people do what we call a product switch or ticket switch. A product switch is taking a small item and placing it over the barcode of a more expensive item – the classic example being a pack of Kool-Aid placed over the barcode of a nice piece of filet steak. The ticket switch involves taking a stick-on barcode like you would find on cheap deodorants or candles and placing it on a more expensive product like a bottle of whiskey.”
A Real-Time Solution
While there are other companies that offer monitoring of retail transactions, Lloyd says one key differentiator for Everseen is that they can do this is real-time, which is essential for a store with self-checkout options.
“Particularly at the self-checkout you need a real-time solution and when we say real-time, we’re talking about making a decision in less than two seconds,” he says. “They are some new entrants in the market that have some AI credentials and can certainly in labs or test stores deliver very good results, maybe as good as ours, but what I think differentiates us – from being in 3,000 stores and multiple retailers – is our knowledge about the many different realities of being in a live retail environment. Retail is a messy environment. Yes, every retailer has a beautiful, new refitted store normally near headquarters where they show investors around but there is also the store in a rough neighborhood that hasn’t been refitted for nearly 20 years and it drops eight new self-checkouts to replace manned lanes. In that store you’re dealing with a very different situation, so we made our solution to be scalable to a number of different environments.”
In the average store where Everseen’s solution has been deployed in self-checkout lanes, Lloyd says they typically intervene in about 2% to 4% of transactions. If the platform is running in what Lloyd refers to as “silent mode,” essentially running in the background without active monitoring, he says they have found that in those transactions they would traditionally intervene is has about 1.6 to 2 mis-scans in that transaction on average. Once the system goes live, Lloyd says they have discovered that if someone is caught once in a transaction that rarely will there be another mis-scan.
“The savings vary with regards to the size of the store but in percentage terms, we estimate that we save – if the total loss at checkout is in the 0.5% to 0.8% range of revenue – 0.4% to 0.5% of revenue,” Lloyd adds. “We’re not saving everything, nor we don’t claim to but we’re continuing to push that up as we look at more use cases. If you’re a retailer working on 4% to 5% margin and the Everseen solution can offer a 10% improvement to you profit margin… you can see that is a very significant number and something that people get pretty excited about.”
Joel Griffin is the Editor-in-Chief of SecurityInfoWatch.com and a veteran security journalist. You can reach him at [email protected].
