Posted on February 20, 2017 by staff

How Amazon Go is taking tills out of supermarket shopping


Amazon Go is not exactly a gadget, but definitely a hot topic and one that I thought worthy of further exploration in my column.

Typically bricks and mortar businesses move to eCommerce, but there is an increasing trend of eCommerce players moving into the high street.

The latest and greatest is Amazon, first with a physical bookstore and now with Amazon Go, a more traditional convenience store with an untraditional twist.

In 1997 Amazon launched ‘1-click ordering’ and many people thought that was a crazy idea fraught with potential error and false transactions.

That frictionless experience is probably one of the reasons that so many people, myself included, buy so much through Amazon, even if it costs slightly more. They are effectively doing the same now for bricks and mortar shopping experience.

So how does it work? Well, nobody can say with certainty because, unsurprisingly, Amazon aren’t yet willing to share the secret. We can, however, deduce a few things from a patent that was filed in 2015 with a bit of educated guesswork.

It’s worth saying this isn’t a new concept. Tesco, among others, trialled shopping trolleys with RFID (radio frequency identification) readers some time ago but the accuracy wasn’t good enough and therefore it didn’t take off.

It is that accuracy that Amazon will have improved by using multiple types of technology.

That same key concept is still in play here. At the doors there are scanners, similar to the ones for shoplifting that will have an NFC (near-field communication) or RFID reader in them. There are two other – possibly more – technologies at play here to deal with the accuracy problem.

The first, and I think main one, is computer vision, basically using lots of cameras for analytics. When you scan your app to come in, these cameras take multiple photos of you to track you around the store, so your proximity to a shelf can easily be seen.

Action recognition with computer vision is a fairly well established science, so the action of you reaching to the shelf can be recognised. There will then also be sensors on the shelves, weight or possibly camera again that will detect an item being removed or returned. Marry the two up and you’ve got a good idea that the person has taken a specific item.

The accuracy can then further be improved by understanding shoppers’ specific buying habits. Did you know that the average person buys no more than 200 different types of products from the thousands available in a typical grocer? This means understanding you further improves the odds.

Keep in mind they’ll also have an understanding of your online shopping habits, so they know you pretty well.

Add this to the RFID/NFC reader and we have till-free, queue-free shopping or, more aptly, 1-click for the real world.

Unfortunately it’s only one store in the US right now, and I imagine it will be quite some time before we see Amazon Go on the streets of Blighty.

That said, I expect retailers all over the world will have an agenda item at the board meeting to discuss this. It will inevitably jumpstart many of these into action about how to take the next step in retail experience, so we may well see something similar over here sooner rather than later.

For me, this is an exciting development in retail and I’ll be trying to find a good reason to visit Seattle on my next trip over the Pond.

For others, it will have raised privacy prickles on your skin.

The truth is, though, that this simply wouldn’t work without a better understanding of you.

Gavin Wheeldon is CEO of Purple