You enter a shop and browse the shelves. Something catches your eye. A new brand of cereal, perhaps. You pick up some milk to go with it. Some orange juice. Maybe some bread to make toast. With this bundle of products in your arms, you walk straight out of the shop door without saying a word. No alarms; no security guards; no confused frowns from other shoppers. You left without paying, but you didn’t steal a thing: you will be charged for it later.
At the new Amazon Go grocery stores, this bizarre scenario will be a reality.
The world’s largest online retailer made its first move onto the high street with a Seattle bookstore in 2015, sending older, established bookstores spiralling in condemnation. The company’s latest retail venture will likely shake things up even more.
Amazon Go will offer all the convenience of a grocery store with none of the checkouts. Walk in, pick up your food, walk out. With this, Amazon could change the way we all shop once again, but a project as ambitious as this raises one important question: is it safe?
The technology behind checkout-free payment
To examine the safety of checkout-free stores we have to take a look at how they work. Amazon themselves say their stores will utilise “the same types of technologies used in self-driving cars: computer vision, sensor fusion, and deep learning.” This statement does very little to shed light on the mysteries behind these stores. On the one hand, Amazon is simplifying things by likening checkout-free stores to another ingenious technology. And on the other hand, they use complicated-seeming terms like “computer vision”, “sensor fusion” and “deep learning”.
We can look at each of these technologies individually to get a better picture of what they mean. Computer vision involves transforming images captured by cameras into data. A CCTV camera could see an orange circle and understand it as an orange (the fruit) for example. Sensor fusion is simply combining (or “fusing”) data from multiple sensors to gain a more precise picture of where things are. And deep learning is an advanced form of artificial intelligence (AI) that sees computer “brains” growing more intelligent, based on experience.
With these terms demystified, it becomes clearer to see how these technologies could be combined to create a checkout-free store, but we still need more clarity. Luckily, an academic paper was published five years ago that discussed the security concerns of a then-hypothetical unattended store.
As the paper explains, any unattended store would need to make use of RFID technology. Hailed by retail tech experts like OCS Retail Support as “revolutionising retail”, RFID sensors are embedded in tags which are attached to products in many retail stores already. These tags emit data about what kind of item a product is and, crucially, where it is.
These tiny emitters likely form a large part of Amazon’s “sensor fusion” plan. As the paper’s authors note, though, it would be difficult to make a checkout-free store work unless it was “integrated with an E-commerce application”. From this conclusion, they did not seem to see the store on the horizon. But for a retail giant like Amazon, with an app on many people’s phones already, it is suddenly a lot more achievable.
Their combination of technologies and RFID tags could lead to a store that knows (and learns) where every item and shopper is at any given time, and how many items to bill them for when they go home. But what if it doesn’t work? What if people find a way around it? Say they don’t download the app, or they sign up with a fake account or even find a way to block the cameras and sensors? Problems like these should worry those who wish to sell their wares in a checkout-free store, and of course Amazon themselves.
How can checkout-free stores be secure?
If implemented intelligently, there may be a way for Amazon and potential followers to use the smart technology of unattended stores to combat shoplifting and other instore crime. The cameras, for example, could witness people untagging items, sneaking items into others’ baskets and stealing them back when they get outside or placing items into RFID-proof bags, but this will not be easy.
Even if the cameras did spot a crime, what would the store then do about it? It could set off an alarm, but many people ignore alarms in stores. The best approach could be to hire security guards, but that might seem counter-productive if the aim of checkout-less stores is to rely less on a human workforce.
Perhaps some form of access control could make these stores more secure. Security firm Bridger Security argue that access control systems are the best way for a business to “exercise control over entry points”. A system like this would allow for an automated system to check customers before they enter and exit the shop to determine whether any foul play had taken place. But taken to its logical extreme, this would end up like an airport security point—perhaps the only place more notorious for queues than theme parks and, yes, grocery stores.
With a combination of tight traditional security and high tech deep learning, the engineers and planners Amazon no doubt has the money to hire will likely come up with an ingenious solution to these potential problems. We will see when the first Amazon Go store opens its doors. But for any smaller business thinking of switching to a checkout-less model, it might be best to wait and see how Amazon Go goes.