• There are currently no items in your cart.

$0.00
View Cart

Cashierless checkout tech is having a rough year: The more commonplace self-checkout scanning stations are under fire amid concerns over shrink and consumer complaints, and on the other end of the spectrum, there also appears to be a pullback on fully autonomous solutions like Amazon’s Just Walk Out (JWO) stores. It’s enough to make one wonder if cashierless tech overall is simply a failed experiment.

The short answer is no, although the experiment is still very much in progress — and there have been some big failures. Most notable is Amazon’s recent pullback of its own tech, with the removal of JWO from its Amazon Fresh supermarkets and its plans to focus on smart carts instead at new locations.

And yet, Just Walk Out and similar technologies are racking up major wins in distinct, primarily smaller, retail environments like sports and entertainment venues, travel hubs and places where staffing for extended hours would be cost-prohibitive, like hospitals and college campuses.

“It is an enduring truth that people don’t want to stand in checkout lines,” said Jon Jenkins, Vice President of Just Walk Out Technology at Amazon Web Services in an interview with Retail TouchPoints. “When you build a product around that, it allows you to focus on long-term solutions that solve real problems for customers. This is where I think Just Walk Out is fundamentally different than self-checkout: we’re not making [checkout] the customer’s problem, we’re making the problem go away altogether.”

But Jenkins is the first to admit that JWO-style tech won’t be a fit for every retail situation. So let’s take a look at where fully autonomous retail solutions are working, and where they aren’t.

The Challenges of Fully Autonomous Retail

1. It’s Intimidating

The first, and perhaps biggest, problem with fully autonomous stores is that they’re new and therefore make shoppers a little uncomfortable. That’s certainly not reason enough to write the tech off completely — after all, it took self-checkout 30 years to reach mainstream adoption — but it has definitely slowed the tech’s ability to scale.

“Not everyone is savvy enough or intuitive enough to use the tech,” said Sandeep Unni, Senior Director and Retail Analyst in Gartner’s Research and Advisory practice in an interview with Retail TouchPoints. “I’ve seen this firsthand on store visits. Pretty much every single time there’s someone at the front of the store walking people through the process, setting up the app or getting people registered. There’s definitely a learning curve.”

“When a store first launches, it’s not uncommon for the store associates to be somewhere near the gates, helping people understand how things work and how to use this type of store,” acknowledged Jenkins. “But customers very quickly pick up on how it works and even start teaching each other. It’s a very different experience, but once you’ve done it, it’s very appealing to do again.”

Because it is so different, even once a consumer figures it out, it can feel a bit strange: “In a Just Walk Out store it’s totally valid to walk in, grab a candy bar and stick it straight in your pocket,” added Jenkins. “In a traditional store, that would be theft, but we know you took the candy bar off the shelf, and we just charge you for it.”

An interesting facet of the ongoing retail shrink conversation, which also brings us to…

2. The Creep Factor

Most consumers are aware that these stores are operated via a host of high-tech tracking tools, including cameras and AI, and in today’s more privacy-conscious world that idea can be off-putting. Amazon and other companies with similar technologies, like Google and Verizon, all claim this data is completely anonymized, but reports like one earlier this year — that the majority of JWO sales involve a human review of camera feeds — throw doubt on those reassurances.

Still, the companies are adamant that personal data (beyond payment method) is not used nor stored to operate their autonomous stores. Amazon vehemently insisted that the aforementioned reports were “erroneous,” adding that “associates don’t watch live video of shoppers to generate receipts — that’s taken care of automatically by the computer vision algorithms.”

“Privacy has always been an extremely important component of what we do here at Amazon; no one wants knowledge of their purchases to be leaked out to the world,” said Jenkins. “We’ve applied those same longstanding principles around privacy to Just Walk Out. We go to great lengths to make sure that the video is used by the machines and then deleted very quickly.”

3. The Cost

In the case of existing stores, shifting to a fully autonomous checkout model is, as one might guess, quite expensive — which is why the most successful of these types of stores are new builds or have smaller footprints.

That’s likely to change, though: “At AWS we have a relentless focus on reducing costs over time,” said Jenkins. “As we reduce the cost of the service over time, more and more stores will be able to make the economic case work out. Just because a store might not work today as a Just Walk Out store, doesn’t mean it won’t next year or two years from now as we drive down the cost of the service.”

The technology itself is also advancing with new solutions like RFID coming to bear and further expanding the use cases for retailers in different categories.

Where Autonomous Retail is Working

Despite these challenges, autonomous retail experiences are making inroads around the world. Amazon now counts more than 140 third-party locations of its JWO tech worldwide, in addition to its own stores, and it’s not the only player in the game. In recent months:

“In smaller-format stores — arenas and stadiums, travel retail — it’s a very different value proposition, because it’s all about how quickly you can get customers in and out, and the assortment is much smaller,” said Unni. “It’s about throughput, convenience and speed. That value prop is more compelling [in these types of retail locations] than in a traditional grocery format.”

According to Jenkins, the “increase throughput scenario” is one of the four most common value-driving scenarios for JWO tech. “That’s your stadium use case,” he said. “Places where there used to be a five-minute line to grab a hot dog and a Coke, and we make that line go away entirely. For example, at Lumen Field [home of the Seattle Seahawks and Sounders] they do twice as many transactions per game in that store as they did before.”

Other top retail use cases for autonomous stores include:

  • Extended operating hours to, for example, keep an airport or college campus convenience store open throughout the night; and
  • Labor optimization: “Say you have a store on a High Street in London that gets a ton of traffic for a two-hour period, between 11 a.m. and 1 p.m.,” posited Jenkins. “You have to staff people for longer than an 11-to-1 shift; people won’t come in for just two hours. A JWO store can absorb that increase in traffic with technology, without having to add people.”

The final, perhaps least intuitive use case, particularly since it’s a main reason for retailer’s pullback on self-checkout, is shrink reduction, according to Jenkins. “We didn’t intentionally build JWO for that,” he said. “What we’re focused on is making sure the customer’s receipt is accurate, and part of that means we have to have really good technology that combines input from cameras, shelf sensors and RFID tags, and uses machine learning techniques and sensor fusion to blend all that information together to determine what actually happened. The side effect is we’re really good at figuring out what has left the shelf and making sure that we charge that to a person’s credit card.”

While he certainly has reasons to cheerlead for the technology, Jenkins truly believes this is just the beginning for autonomous retail: “Originally people thought it’s cameras on the ceiling and sensors in the shelves. And that is what it is, but with the recent launch of our RFID stores we’re showing that there are other sensing technologies that can be brought to bear,” he said. “When you apply machine learning across all these sensor technologies, you can support new and interesting merchandise. I don’t think we’re anywhere near done in terms of the technologies that we will bring to these types of stores. We’re constantly evaluating what we add in terms of a technical platform to enable faster, more convenient, more seamless merchandising experiences, and I’m excited over the coming years to see what we will create to do things that people didn’t think were possible.”