On Amazon Go: It’s Not About Retail!

supply chain dloboda/Adobe Stock

In my first post on Amazon Go, I talked about some operation details of the Amazon Go store model, and it’s certainly tempting to think about Amazon Go as a retail disrupter. To some extent it is, especially as it relates to the sensor fusion technology, which was lampooned in a recent spoof video. It's a fair bet that many grocers are working hard on their strategies for mobile shopping apps, click-and-carry models, small footprint formats, use of RFID and vision systems and so on. For example, it will be particularly interesting to see how companies like Trader Joe’s (in the U.S.) or Tesco (in the U.K.) respond.

Yet this story isn’t really about retail, but rather about the broader supply chain. Amazon finally topped the leaderboard on the Gartner supply chain rankings last year — and not just because of its role in retail, but because of its broader supply chain and digital capabilities. These capabilities have been built methodically and also incrementally. The capabilities then help “unlock” value down the road. What is being built is an operating model, not a store. In fact, you can see this in Amazon’s patent filing for the sensor-fusion-enabled operating model applied here.

If you go through the filing (which is fascinating in terms of its combination of specificity and generality through the use of massive compound sentences describing thousands of potential scenarios), you’ll notice that this operating model doesn’t mention shoppers in retail stores. It talks about users, not shoppers. It talks about order picking, not consumer shopping. And it talks about “material handling locations” — not stores! Per the filing, the patent “describes for tracking and identifying the removal of items from inventory locations and the transition of items from a materials handling facility.” And the scope of these facilities includes “warehouses, distribution centers, cross-docking … rental facilities, libraries, retail stores, wholesales stores, museums” and more.

In essence, a “user” (a human for now, including the consumer shopper) becomes an object to which to associate chosen items, sort of like a virtual shopping cart that follows the user. And according to the filing, “transitioning” the items from stock to the user “may include charging a fee to the user for the purchase, rental, lease, etc. of the items, consumption of the items … updating the user profile [and] transferring ownership.”

As you read through the patent and its process/data flows, you get a sense of the gory complexity involved in keeping the information system synched with the reality of the material or user flow, as well as all the anomalous conditions and thresholds that will require intervention from the user and/or facility staff. If the system gets easily confused from some real-life scenarios (some specified in the patent filing itself, such as grabbing multiple potato chip bags), then the security alerts are going to be very noisy.

I think that there’s no way the system will be scaling up as rapidly as is being reported in the popular business press. However, the machine learning will get smarter rapidly, especially if Amazon Alexa is listening to you at home when you say you are in the mood for a cookie. Just kidding here, of course. But nevertheless, the patent says that user information may include “images of the user, height of the user, weight of the user, username and password, user biometrics, purchase history, payment instrument information, purchase limits, and the like.”

Yes, this might seem like the blueprint for a Minority Report-like world, but it’s really a way to establish IP rights to a supply chain operating model that is as pertinent to a tool crib or vending machine as it is to a retail shelf. The system could enable creating a virtual vending machine of sorts that transcends a physical box. So you could perhaps see its applicability in an MRO environment where the retail shelf in essence becomes a tool crib. The user recognition system could also just be a less “visible” biometric system to help replace or augment current badging systems (including the use of NFC/RFID tags and your smartphone).

Supply chain professionals who read the patent might get other insights from it. I personally thought about the supply network design implications. For example, if the store essentially becomes a warehouse and shoppers are order pickers, how should the inventory be optimally configured? Would it be different from the usual supermarket layout? For the click-and-collect model, the consumer car is really a container coming to pick up its load. Therefore, should that load be only picked from that location (e.g., a large format Amazon Go store with an adjoining warehouse), or should it be cross-docked via fulfillment coming from a nearby Amazon Fresh warehouse? Could Amazon Pantry items be merged in transit with Amazon Fresh orders to create a unified model for the consumer, as opposed to the current disjointed model? Could Amazon Fresh trucks pick up a click-and-collect order and deliver to the home if the consumer wants to change the method?

These scenarios illustrate not just the strategic importance of supply chain to enable new consumer facing value, but also how smaller initiatives that build supply chain capabilities can pay off in bigger ways down the road. In subsequent posts, I’ll highlight many areas where procurement and supply chain organizations can take some lessons from Amazon (and others) about building a more flexible operating model for physical goods and services (including internal procurement/supply service) — much of it digitally enabled — that will enable new sources of value. Stay tuned for more!

Discuss this:

Your email address will not be published. Required fields are marked *