The parable of the supermarket self-checkout

Imagine an invention that is worse at what it does than humans, threatens jobs and increases the potential for crime. You might think it would go nowhere fast. In fact, the supermarket self-checkout machine is a parable of technology adoption—how something can spread despite imperfections—and also one of management: how real-world trade-offs affect rates of change.

Start with its spread. Before self-checkout came self-service. The Piggly-Wiggly grocery chain, which opened in Memphis, Tennessee in 1916, was the first in America to allow customers to pick items from the shelves for themselves rather than asking a clerk to fetch them. It took a while for American shoppers to be able to scan and pay for goods themselves. Kroger, another retailer, introduced self-checkout machines in 1986. It was not until this century that they became widespread.

All new technologies must get over the problem of forcing people to change their behaviour. In Britain, initial encounters with self-scan machines involved a disembodied voice shouting “unexpected item in the baggage area” at you no matter what you did.

Even now, they bring their own very specific irritations—age-verification checks to buy alcohol even though you remember using faxes; a ten-step process to confirm that the yellow, waxy citrus fruit you put in your basket is a lemon. A cashier at an assisted lane can handle these thorny problems without having to think; skilled ones can also scan items much faster than customers.

Why then are self-service machines a fixture in most retailers? The big reason is economic: the ratio of staff to customers is much lower for self-checkout machines than it is on assisted lanes. The lure of cost savings in a highly competitive industry gave supermarkets a strong incentive to stick with them even as customers griped at doing something new. Persisting with them meant that shoppers became more used to self-checkout, and gave the technology a chance to improve.

It will, however, be a long time before assisted lanes disappear entirely. When queues build up and stores need to speed up transactions, human cashiers are better. And customers’ preferences vary greatly. Older folk still tend to like cashiers; younger ones are less keen. Chatty types might want a natter; introverts would prefer an acid bath. Levino Perrucci of NCR-Voyix, an American company which develops self-checkout technology, says that users fall into several categories. “Tech-enabled shoppers” will seek out the whizziest checkout, for example; “necessity shoppers” will pick the option that enables them to get in and out of the store as quickly as possible. Most supermarkets must cater to them all.

Self-checkout machines are themselves designed to negotiate several trade-offs. Employers want lower costs. But they also need to provide a decent customer experience and control “shrink”, the name that retailers give to lost stock when shoppers—inadvertently or maliciously—fail to pay for things.

These goals do not line up neatly. Employing fewer cashiers cuts costs, for example, but also opens the door to more shrink. So lots of self-checkout machines have weight-sensitive surfaces where you have to place each scanned item before proceeding to the next one; some retailers install exit gates that only open for people who have paid. Measures like these deter shoplifters, but at the cost of slowing down customers.

These same trade-offs explain why checkout nirvana is so difficult to achieve. In theory, Amazon Go, a sensor-packed store where customers can just walk out without going anywhere near a checkout, offers a perfect customer experience and a way to reduce shrink to nothing. In practice, industry observers say that it is extremely expensive to track every item in a shop digitally (and consumers are still required to learn strange new behaviour). RFID tags do work beautifully for higher-margin products like clothes, but the sums don’t add up in the cut-throat world of cheap groceries.

Real-world constraints mean that technologies often take small steps rather than great leaps. Self-checkout technology is no different. In the offices of NCR-Voyix in Atlanta, Mr Perrucci’s team is working on machines with overhead cameras which use AI to identify items without the need for scanning. These machines can still get confused if items are piled on top of each other. But they fit into existing consumer behaviour. They should also be better at identifying that bloody lemon. ■