Amazon Go – “Just Walk Out” is a revolutionary concept of shopping.
After more than a year of beta testing, Amazon opened the first Amazon Go store in Seattle on 22nd January.
What is it about? Essentially, it’s the first “cashierless” shop, it means no cash in circulation. You identify yourself, you enter, you take what you want, and then… you leave. The slogan that stands out in the shop window is “No lines, no checkout. No, seriously, ” in addition to the gadget slogan, ” Just Walk Out”.
Yes, because one of the curiosities is that the entrance to the shop vaguely recalls that of the subway, with a series of turnstiles. Practically the opposite of traditional shops, where you enter freely, but checkout at the exit. Here, on the other hand, you identify yourself at the entrance, and then you freely go out with your shopping bag.
The items on sale in Amazon Go mainly concern gastronomy, drinks, and what they call “meal kits,” or all the pre-dosed ingredients for a dinner for two.
Any product taken is loaded onto the phone’s electronic wallet, for many of the products the shelf from which they are taken is used for authentication, while for others (see below) a special code on the packaging is used for recognition. In both cases the authentication process is carried out by a series of intelligent cameras placed on the ceiling.
If you change your mind, you can put the object back on the shelf, so the item will be canceled from the total bill.
There is no checkout process: you just quit shopping and walk out. After a few minutes the receipt arrives punctually on the phone.
Returns are not a problem, simply select the object you want to contest on the receipt, and this is automatically scaled, without any questions. You don’t need to return it, although you can physically return it to one of the store’s clerks.ù
Amazon has not been particularly lavish with details on the technology used until now, but inferences can be made on the basis of what is known, namely that at the base of everything there is a sensor fusion technology, combined with computer vision and deep learning algorithms.
We also know (as reported by Techcrunch) that facial recognition algorithms are not used.
Let’s try to be clear: sensor fusion means that there is a sort of “triangulation” process between the incoming data from the different sensors. One of these sensors are the ceiling cameras, which each shop (for the moment there is only one) is equipped with.
These cameras employ sophisticated Computer Vision algorithms, with which they are able to capture and track the product codes and the shape of the customers. The cameras are not only on the ceiling, but also on the shelves, to intercept the picking of the product, or its repositioning on the shelf. To identify the specific package, AI could use a combination of bar code, expiry date, and other data such as the change in weight on the shelf.
Identifying which product was taken is obviously not enough, it is also necessary to understand who took it. For this purpose, a Bluetooth beacon network is able to identify the closest smartphone to the shelf, and maintain communication with the Amazon app.
Free shoplifting for everyone?
In reality, the system seems rather difficult to “screw or cheat”, even covering the objects or attempting other tricks. It is not impossible, the system is not perfect, but the Amazon principle is that the vast majority of users are not thieves, and since designing a system around the few who could intentionally steal, is not ideal, everything has been designed around good faith.
This is also demonstrated by the fact that it is extremely simple to make returns: you simply indicate on the app the product that you are not satisfied with (or that perhaps you think is unduly loaded), and this will be automatically deducted from your account, without any questions. It is not even necessary to physically bring the product back to the store, altough it is possible to do so by contacting the clerks.
What about job positions?
In Amazon Go there are no checkouts, it’s true, but Amazon declares that so far the cash staff has not undergone downsizing, but has been “recycled” on other functions, such as cooks, clerks (always close in case of problems) and PR staff.
Hard to say if this trend will remain, but at the moment there are no cuts in staff in the Amazon stores.
What is the purpose?
Although the cashier-less selling principle is not new, and someone is already thinking of competing with Amazon Go through vending machines that work on similar principles (AI, computer vision, tracing via telephone…), Amazon’s purpose is likely to go further that of simple smart retail. As well as some speculation about the possibility of selling the system to other retailers, a likely scenario is actually using these stores as bridgeheads.
Initially, no membership was required or “suggested” for shopping, but on May 15th, Amazon began introducing a 10% discount to all Prime members, in addition to sell Echo in the Whole food chain stores, acquired only last year by the company.
But probably the business core is data: it is true that, according to Amazon, facial recognition is not used, but we are still talking about an impressive amount of “first-hand” data. In fact, there is data on how customers interact with products, which shelves make them more curious, which products possibly raise more doubts, which ones are preferred, which profile tends to “explore” the shop more and which one goes straight to the point.
These are all data points of great interest for an e-commerce such as Amazon, and we will see how they will be managed.
D. Kumar et al.: DETECTING TEM INTERACTION AND MOVEMENT – Amazon Technologies, Inc. (US Patent Application Publication, US 2015/0019391 A1), 15/1/2015