A couple of decades ago, I co-founded a mail order catalog business. We never grew past eight figures in sales, but we were successful in our niche in large part because of our efficient systems. Over a dozen years, we punched above our weight by using the latest and best technology we could find or create for order processing, customer support, and call center scheduling.
I exited that business just as the transition to e-commerce was beginning. Although not apparent at the time, my timing was fortuitous. I was leaving the direct-to-consumer space just as Amazon was starting to sell its first product category, books. Sooner or later, we surely would have been a casualty of Amazon’s drive to e-commerce dominance. Today, Amazon owns 38% of U.S. online sales; the next closest competitor, Walmart, has just 6.4%.
Amazon’s key strategy as it pushed to the top of the e-commerce world was an incredibly easy shopping experience. As far back as 1997, Jeff Bezos was talking about “frictionless shopping.” Two years later the company was awarded a patent for “One Click” ordering, a small advantage but one they vigorously defended. From frustration-free packaging to free, simple returns, Amazon has focused on eliminating even minor points of friction in their customer experience.
Amazon eventually decided to go beyond e-commerce and target in-person retail. In 2015, they opened their first physical Amazon Books store. That was a modest start, but two years later they acquired Whole Foods Market, a comparatively small but well-known national supermarket chain.
Presumably, acquiring Whole Foods was less about selling more organic groceries and more about learning the ins and outs of retail. The chain’s younger, more affluent customers would likely be more accepting of digital innovation, Amazon’s core strength.
They also opened Amazon Go, Amazon Style and Amazon Fresh stores to test retail concepts and to demonstrate and develop retail technology. Many of these have been closed, although some Fresh and Go stores remain open.
Perhaps Amazon thought it could leverage its technology and logistics prowess to conquer retail the way it did e-commerce. So far, that hasn’t happened.
Amazon has now shifted strategy. They have moved their retail technology operations to Amazon Web Services. At AWS, the emphasis will be not on Amazon stores but on third-party sales.
As a long-time observer of retail technology, I’ve seen many attempts to streamline the shopping experience. Most don’t accomplish that goal, and some even frustrate consumers. Who wants to self-scan a shopping cart full of groceries?
But, there’s reason for optimism. My recent visit to Amazon’s Just Walk Out (JWO) lab in Seattle provided an interesting behind-the-scenes look at their approach to friction-reducing shopping technologies. Taken in total, these innovations represent a significant step in retail evolution.
Amazon’s vision for the future of retail is not unlike its strategy for dominating e-commerce: remove every possible obstacle between the customer and their desired purchase. This isn’t just about convenience – it’s about fundamentally changing how we interact with physical stores.
The technologies I saw won’t revolutionize the shopping experience for most of us, at least in the near term, but they have the potential to eliminate major pain points in specific situations and environments.
The best-known of Amazon’s retail innovations is Just Walk Out (JWO) technology. While it seems magical to shoppers – walk in, grab what you need, and leave without checking out – the reality is a complex interplay of multiple technologies.
JWO combines overhead cameras, 3D shelf mapping, product appearance recognition, and sophisticated tracking software. The system doesn’t just see products; it understands shopper behavior, tracking individuals through the store while maintaining their privacy.
Visually determining exactly which product a customer is taking turns out to be a very complex problem. And, Jon Jenkins, vice president at JWO, notes, “close” isn’t good enough – 90% accuracy doesn’t sound bad, but it would mean a clearly unacceptable 10% error rate. Both for retailers and customers, the error rate must be near-zero if the technology is to be trusted.
Potential problems abound. The customer’s hand or body might prevent a camera from getting a good view. The customer might take two of an item, with the second one concealed behind the first. Items can end up in the wrong location – for example, a customer grabs a Coke but then takes a Pepsi from a different shelf, leaving the Coke in its place.
According to Jenkins, moving to diverse retail locations has created interesting challenges. For example, in humid climates, a shopper opening a glass refrigerator door can instantly fog the window, temporarily blinding cameras from that angle. It’s this kind of real-world complexity that Amazon must solve to achieve near-perfect checkout accuracy.
Accurately identifying items requires an algorithm to take into account all of the various data points (location, appearance, etc.) and make the determination. Extensive testing and machine learning have boosted the accuracy to acceptable levels, Jenkins says.
Surprisingly, one of the easier problems to solve was tracking individuals as they move about the store. The cameras track heads, hands, and feet as unique identifiers to keep people straight, even when they cross paths with each other or join groups and separate. Privacy is assured by not using personal indentification techniques like facial recognition.
Shopping groups are also identified. When a family enters a store, for example, they can move about separately but their selections are consolidated at payment time.
JWO has been implemented in stores as large as a 40,000 square foot Whole Foods. But, that’s not the target market – testing has shown that JWO is best suited to smaller stores with “visible and different” items.
For larger format stores, Amazon has developed the Dash Cart – a smart shopping cart that brings some of the benefits of JWO to a more traditional shopping experience. The Dash Cart helps another frequent shopper pain point, finding products in the store. The Dash Cart display helps locate items and guides the shopper to their location.
While JWO’s camera technology is impressive, it’s not the only way Amazon can eliminate checkout lines. For example, apparel and other soft goods can change shape (e.g., a sweatshirt can be neatly folded, dropped in a heap, or put on a hanger). They are also hard to distinguish visually – a “large” looks about the same as an “extra large” or a “medium.” Both of these characteristics present a challenge to Amazon’s JWO cameras.
For situations like these, Amazon has developed an RFID-based JWO system that’s both simpler and cheaper. Instead of a web of cameras, RFID-based JWO uses exit gates to detect tagged items. This technology is particularly well-suited for pop-up stores or temporary installations at events.
Although RFID checkout seems simple, the underlying technology is far more sophisticated than the simple anti-theft gates seen in mall stores. RFID checkouts must differentiate between an item the customer is leaving with, items in an adjacent checkout lane, items displayed close to the checkout, and even RFID tags discarded close to the exit after purchase. The algorithm must correctly count only the items being purchased by the current customer and reject all other spurious signals.
Since the customer doesn’t need a payment method to enter an RFID-equipped store, it’s suitable for situations where customers want to browse the merchandise before deciding to buy.
Perhaps most importantly for retailers considering these technologies, Amazon is committed to continually driving down costs. The JWO team has annual cost reduction targets, which, according to Jenkins, they consistently meet. Amazon develops its own cameras for both performance and cost reasons.
While the infrastructure for JWO definitely has cost associated with it, there are tangible benefits to offset those costs. Shoplifting is virtually eliminated, since removing an item from a shelf is logged as a purchase. Labor costs can be reduced or personnel redeployed to improve customer experience. Most importantly, in some situations throughput can be dramatically increased. Stadium concession stands, for example, often have long lines of customers waiting to order and check out. Pre-staging drinks and food items can combine with JWO to speed up operations.
The Seattle Seahawks have been early adopters of Just Walk Out technology. After converting a traditional concession stand to a JWO store, they saw a dramatic increase in sales. The number of purchases jumped by 85%, while per-game sales more than doubled.
Perhaps the easiest-to-implement retail technology Amazon offers is Amazon One. Once integrated with a retailer’s point-of-sale system, a wave of the customer’s palm supplies both credit card and loyalty information, no phone or wallet needed.
Amazon One is even being used for age verification at Coors Field in Denver, eliminating yet another friction point.
Amazon has created an impressive array of technology to make shopping easier – Just Walk Out, RFID, Dash Carts, Amazon One, and more. Most of the early adopters, though, aren’t mainstream, typical retailers. Rather, they are stores in airports, stadiums, hospitals, theme parks, etc.
As more stores adopt the technology, there’s little doubt it will get cheaper and easier to implement, even as accuracy and performance metrics improve. But will that be enough for a retailer who sees Amazon as their biggest competitor? Will they be willing to make Amazon a key and visible part of their customer experience?
AWS cloud services are indeed used by many retailers globally. Amazon displays dozens of case studies on their website, and surely there are far more brands that don’t publicize using AWS. These companies trust Amazon to keep their data secure and to not use it for the benefit of Amazon’s own retail businesses.
To gain acceptance of JWO and other retail technologies, Amazon will have to double down on reassuring customers that their data is secure and private. Additionally, they will likely have to white-label their products, i.e., avoid any labeling that mentions Amazon.
I don’t expect Walmart to adopt Amazon technology. Retailers that lack the technical prowess and resources of the biggest chains, though, may decide that when it comes to improving the shopping experience, Amazon isn’t the enemy.
Personally, I’d love to see more stores adopt Amazon One for payment and loyalty. I use it often at my closest Whole Foods, and now extracting a credit card from my wallet or even opening a payment app on my smartphone when paying at other stores seems effortful.
I’m excited to see how retail evolves with these and other technologies. I’m annoyed by checkout lines, and even more annoyed by self-checkout processes that transfer the work of efficient cashiers to far less capable consumers (like me).
Truly frictionless shopping may be an unattainable goal, but we’re heading in the right direction.