Managing Self-Checkout: The Potential Role of Video Analytics

The growth in the use of various types of self-checkout systems (SCO), especially, but not exclusively in grocery, is one of the defining transformations of 21^st century retailing. From a rather niche and generally unloved (certainly by some shoppers) form of checkout in the early 2000s, SCO is increasingly defining the design and development of what is often described as the ‘frictionless’ retail store customer experience.

While ‘fixed’ SCO machines continue to be the dominant technology, where the consumer scans and pays for their items at a static checkout station/robot, many other variants are now being deployed. This includes: ‘scan and go,’ where the retailer provides the consumer with a scan device; ‘mobile scan and go,’ where the shopper uses their own device; ‘smart’ shopping carts, which can detect items automatically placed within them; and autonomous stores, such as Amazon Go, which remove all forms of traditional checkout, enabling registered users to pick up items and simply walk out of the store.

Although this transformation is being embraced by many parts of the retail sector, it does not always come without an increase in risk—offering consumers less ‘friction’ (usually through removing various forms of established controls) can expose retailers to elevated levels of retail loss. This has certainly been the case when it comes to SCO; various large-scale studies undertaken on behalf of ECR Retail Loss have consistently identified the scale and extent of additional losses generated by these systems. Indeed, one study estimated that stores with 50 percent of their transactions being processed through fixed SCO could see their overall store losses increase by 75 percent.

This research has also shown that SCO losses are mainly caused by three problems: customers not always scanning the barcodes of all the products they wish to purchase (accidentally and on purpose); misrepresenting a cheaper product for another more expensive one (the carrots for grapes scam); and walk-aways, where a consumer scans all their items but leaves before initiating a successful payment.

Responding to SCO Losses

As retailers have become more aware of the scale and extent of these problems, they have begun to explore a range of ways in which they might be better controlled and prevented. These have coalesced around four key themes: design (the SCO technology itself, the space within which they are utilized, and the products being purchased); guardianship (the use of people to monitor SCO, including training, incentivization, and staff levels); store processes (such as when to open/close SCO stations); and technologies, in particular, video-based systems.

Using Video Analytics

Many retailers are now exploring the use of video analytics to address risks relating particularly to fixed SCO environments, looking to automate processes that can detect, direct, identify, and sanction.

• Detect: Automatically and accurately identify when a user has not scanned a particular item.
• Direct: Interact with users that have made an error to encourage them to correct a problem without recourse to a member of staff (such as rescan at item or correctly identify a product).
• Identify: Through product recognition, automatically block miscreant mis-presentation of one product for another e.g., the system can identify grapes and will not allow them to be labelled as carrots.
• Sanction: Only ‘allow’ shoppers that are associated with a form of payment to exit a SCO area.

Recent ECR research has shown the growing retailer appetite for these types of video-based interventions in their SCO environments. When a global sample of retailers were asked about their plans and current trials, 58 percent said product identification, 51 percent checking customers had paid before leaving, and 38 percent identifying non scanning.

In many respects the fixed SCO space offers many advantages to the use of video analytics compared with other parts of a retail store. The space is often well compartmentalized, known objects are typically systematically moved across a well-defined and identifiable space, and the flow of customers can be controlled through judicious design. This is important in terms of both designing and installing video analytics, but also improving their accuracy and reliability. It is often easy to install a video analytic, but much harder to take account of the vagaries of the space within which it must operate, which can significantly undermine its operating capability.

More SCO, More Analytics

It is highly likely that the use of SCO systems will increase further in the future, accounting for a growing percentage of transactions and visible in a wider array of retail settings. In addition, retailers will become more attuned to the way in which they generate losses and consequently, seek to invest in a range of control interventions. Evidence to date suggests that an effective strategy should carefully consider aspects of guardianship, design, store process, and not least, technologies. As the capability of a range of video analytics develops over the next few years, it seems clear that their role in helping retail businesses manage their SCO environments will undoubtedly grow.

Video Watch is a monthly column written by Professor Adrian Beck sharing insights on the proactive use and impact of video technologies in retail. It reflects the latest research and monthly discussions of the Video Working Group of ECR Retail Loss, the leading global think tank on retail loss. The research commissioned by ECR Retail Loss is supported by independent research grants provided by Genetec and other leaders in retail loss prevention.