Redefining the DNA of the shopping experience—or trying to

People walk by the Amazon Go brick-and-mortar grocery store without lines or checkout counters, in Seattle Washington, U.S. December 5, 2016. REUTERS/Jason Redmond – RTSUU23

By Alan O’Herlihy, Everseen

It’s easy to point to Amazon’s lack of experience in brick-and-mortar retail as an explanation for the difficulties it’s facing as it tries to launch Amazon Go. Retail is not for the faint of heart, so entering the space and trying to transform it seemingly overnight has inevitably resulted in some skepticism—and even a little bit of schadenfreude when the company missed its self-imposed, and highly publicized, launch date.

But the same can be said of many of the industries that Amazon has reinvented. Amazon had no business getting into books, video content, clothing, warehousing and delivery, etc. And yet it continues to redefine all of these, much to the dismay of early skeptics (not to mention, the businesses it has upended).

Common among Amazon’s largest successes has been its ability to understand the nature of a real world transaction, translate it into the online world, and establish what the experience there should be. Where it will continue to struggle in the real world is doing the reverse.

Online it was able to create the rules from scratch. Offline it has to work within the confines of established customer expectations, two hundred years of “this is how we’ve always done it,” and its own notable lack of information about how this all plays out day to day.

Oh yeah, and the technology. What they’re attempting to build is not only extremely sophisticated, it’s dependent on data they don’t have. To their credit, as of their late March deadline they’d completed quite a bit of the heavy lifting: cameras that can identify and track faces, sensors that know when a product has been removed from the shelf, and algorithms that tie this data together to tell the system who to attribute behaviors (and charges) to.

The postponement of Amazon Go’s official launch in late March wasn’t due to a lack of highly impressive technology; it goes a bit more granular (and deeper) than that. Their current issues are experiential in nature—details that only come from intimately understanding the DNA of customers’ behaviors and the shopping experience itself.

Having spent the past 7 years working with 5 of the top 10 global retailers to perfect my company’s mature non-scan detection AI technology, the type of customer behaviors you witness on the ground during a typical day-in-the-life would surprise you at the least, and frighten you at worst. Eliminating checkout will do away with some of these behaviors, but not the psyche of the shopper responsible for them. That’s the part that Amazon’s missing. And ultimately, it’s precisely why they’ve underestimated the challenges and made the missteps they have along the way.

They didn’t (and probably still don’t) have the use case data they need to properly analyze and understand what building this technology will actually take. Algorithms need to be trained using data, so their lack of it not only affects their understanding, but also the technology itself. This data and understanding come from examining and analyzing hundreds of thousands to millions of customer transactions, and then tracking back the flawed transactions to either the customer, the cashier or a gap in the supply chain. Understanding where flawed transactions originate is critical to Amazon’s ability to account for them when developing a structure for the new checkout process.

Not only do they not have an in-depth enough understanding of the flow of customers and products in the physical space, they also don’t seem to understand the mentality of a retailer. Before breaking and rewriting the rules, they need a better understanding of why retailers make some of the decisions they do. And what customer behaviors inspire those decisions.

From a pure tech standpoint, Amazon didn’t fully realize the computer power processing challenges they would face, as is being well documented now. The power requirements for running a convenience store with checkout-free technology are completely different than those for running the same size convenience store with a standard checkout.

Training in one store also doesn’t give Amazon the data they need to scale their technology across different customers types and to multiple locations. As far as anyone can tell, Amazon has taken its technology from the lab to its single Seattle store location, and conducted their testing in those two locations. That’s not nearly an adequate test case. As a result, their algorithms are left starving for data and desperately out of tune.

When my company was perfecting our technology, we trained our algorithms in over 100 test stores throughout Europe before our commercial launch. We gave retailers the technology in return for letting us fine-tune our algorithms. Amazon simply doesn’t have this luxury. Not only do they not have the physical stores; no retailer in their right mind would trust them with their customer data as part of the tests they’d need to run to get their technology working.

To solve this, their two obvious options are to either buy a grocery chain or do what IBM, Salesforce, Oracle, Google, Facebook and others are doing, and just purchase start-up AI companies until they find the technology they need (which still may not fill the customer data gap).

Now, to be fair to Amazon, missing a tech deadline is par for the course—but that’s why you don’t talk about it so far in advance. They jumped the gun on their announcement and now they don’t have the luxury of working out the kinks behind the scenes.

When Amazon mastered the online retail game, every physical retailer out there had to fight for dear life to compete in the digital space. Amazon will figure this out eventually, but now that the roles are reversed, there are a lot of companies out there popping popcorn as Amazon feels the heat.

Alan is the CEO and visionary behind Everseen, a company that uses artificial intelligence to convert video into a sensor, with applications across retail, healthcare and security.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s