Amazon Go-es After IRL Retail... Again

The biggest news in the IoT this last week was Amazon's announcement of their updated Brick-and-Mortar shopping experience: Amazon Go.  The major selling point is simple: check-in, grab what you want and leave.  There is no check out.  The system simply knows what you have picked up, bills your Amazon account and that's it.  It does all of this in real time while you are in the store.  There is no TSA-style sensor pad at the exit point to walk through, no scanning of bar codes or RFID tags.  It just knows.

Tech

In many ways, the experience is reminiscent of an old IBM ad from a decade ago (and lots of people are making the comparison).  However, while the experience appears to be similar, how the two companies are going about it are very different.

The old IBM ad is based entirely on RFID chips on the product.  These are scanned on leaving the store and a receipt is generated (to be delivered by a more robust Paul Blart type).  To make this work, the chips need to be embedded in the packaging for each product at manufacturing which will increase costs.  Admittedly, passive RFID chips can cost as little as 7 cents each, but over hundreds and thousands and millions of items, that adds up quickly.  Keep in mind that most packaging companies are looking to shave every penny off of their product.  Just look what they are willing to go through to reduce materials in an aluminum can.  Still, the IBM thing is just a concept, albeit one that no one has implemented in a retail environment.  Not even Amazon.

Instead, Amazon is using a combination of 'computer vision, sensor fusion, and deep learning.' Whatever that all means, it does not mean chips on product.  They call it 'Just Walk Out' technology and reading into what little they say about it, it appears to be using the location of the shopper's phone (which is running their app while in the store), shelf triggers and cameras.  The last are no doubt heat mapping the store to know where everyone is.

The end result is an IRL implementation of Amazon's '1-Click' ordering.  It is on-line shopping in meat space.  All of the mechanics are mapped across: you check in, you search for what you want, you put it in your cart/bag and you leave.  Done.

Implications

For the consumer, there is a whole lot of win involved here.  Checking in is a lot less time consuming than checking out.  There are no worries about price checks or mis-read barcodes (at least not in the picture that Amazon is painting).  And prices should be lower because there is less overhead as there is no need for checkers or even self-checking stations.

For Amazon, it also appears to be a win.  In essence, they are opening one of their warehouses to end consumers and then outsourcing the 'picking' job to those same consumers.  And those consumers are paying them (in the form of product margin) for the priveledge.  Looked at in that light, Amazon is Tom Sawyer, the products are the picket fence and we are all the rubes paying to paint it.

That's the first of two things to be careful of with this system: fewer basic service jobs.  I'm all for automation and removing jobs that are better done by machines.  I've even written about it before.  However, it is going to remove a bunch of 'unskilled' jobs from the market (though if you pay attention to grocery checkers, it is far from 'unskilled').  That means that these people will have less money to buy things at places like Amazon and may lead to a vicious cycle of cost/job-cutting to meet lower prices because people can't afford things at the old price because they no longer have a job.

The second issue is privacy.  This is an old one with Amazon that is now being pushed into the real world.  They already know what you buy on-line (and because so many people use Amazon Web Services, they know more about you than merely your Amazon shopping habits... or could if they wanted to).  Now they are going to know not only what you buy in the real world, but also what you looked at but did not buy.  What you put back.  What you glanced at.  Whether the 'Special Buy' bin at the front of the store worked on you or not.

From this, they will then be able to make more educated guesses at what will entice you to spend more.  Amazon is very good about this on-line.  You can bet that they will just as good in the real world.

Gaming the System

My final set of thoughts on this (and then I'll stop typing because what the hell this is freakin' long!) have to do with retail security.  The dreaded Loss Prevention that every retailer worth their company shirt has in one form or another.  And this is because one of the basic rules of retail is "If it isn't nailed down, it will get stolen".  Shrinkage is very real and has been estimated as high as $45 Million in 2015.

For Amazon Go, they have a couple of things going for them.  First, the whole store is apparently blanketed by cameras.  Secondly, each item, when picked up, triggers an inventory 'event'.  So, in theory, if an item is picked up and there is no phone nearby to register the event, then no doubt alarms go off.

On the other hand, those check-in barriers look awfully easy to hop.  A savvy thief could jump them and then closely shadow a legitimate shopper, picking things up as they walk by them.  There may also be ways to spoof the sensors and the rest of the 'fusion'.  These will be explored and limits will be pushed and Amazon will learn.  However, they are going to need some human guidance in all of this.  Imagine a child being dragged to the store by a parent.  This child grabs an item without the parent noticing (because that has never happened before).  What happens then?  Is that theft?  How is delicately can the 'deep learning' handle all of this?

Speaking as a lifelong retail professional, I can't wait to find out.