Amazon’s ‘Just Walk Out’ Shopping Is Out, Moves To Dash Carts At Its Grocery Stores

After a few years of Amazon promoting a grocery shopping experience without checkout lines and frustrating self-checkout experiences, it is now ditching its Just Walk Out technology. Conceptualized as a store where you can walk in, grab the items you need and walk out with said items automatically charged to your registered payment method, it never really caught much traction. More recently it was revealed that the technology wasn’t even as automated as portrayed, with human workers handling much of the tedium behind the scenes. This despite claims made by Amazon that it was all powered by deep machine learning and generative AI.

An Amazon Dash Cart's user interface, with scanner and display. (Credit: Amazon)
An Amazon Dash Cart’s user interface, with scanner and display. (Credit: Amazon)

Instead of plastering the ceilings of stores full with cameras, it seems that Amazon instead wishes to focus on smart shopping carts that can keep track of what has been put inside them. These so-called Dash Carts are equipped with cameras and other sensors to scan barcodes on items, as well as weigh unlabeled items (like fruit), making them into somewhat of a merging of scales at the vegetable and fruit section of stores today, and the scanning tools offered at some grocery stores to help with self-checkout.

As the main problem with the Just Walk Out technology was that it required constant (700 out of 1,000 sales in 2022) human interaction, it will be interesting to see whether the return to a more traditional self-service and self-checkout model (albeit with special Dash Lanes) may speed things along. Even so, as Gizmodo notes, Amazon will still keep the Just Walk Out technology running across locations in the UK and elsewhere. Either this means the tech isn’t fully dead yet, or we will see a revival at some point in time.

60 thoughts on “Amazon’s ‘Just Walk Out’ Shopping Is Out, Moves To Dash Carts At Its Grocery Stores

    1. People will totally forget the concept of hubris every five years before nature once again smashes their teeth with it and afterwards they will act like they’ve discovered the ancient wisdom of Atlantis for the very first time. It is not funny anymore, it’s shocking to see it happen so many times in my short life.

    2. Indeed, and I’m very gald they’re shuttering this one. I had the misfortune of being forced into one to return a package once, I can only describe it as a dystopian camera infested Orwellian hell hole. Just to try and get to the returns counter it wanted me to install an app to sign over data and card details, I refused the kind offer and just tailgated another customer through the barrier.

    3. It’s not that anyone is dreaming about this except for companies who are hoping to save on the labor costs of check-out personel.

      Amazon isn’t doing this because of some dream of the future, but because of a dream about their bottom line.

    1. The real problem is why just walk out in a store that will still charge you money versus one that will let you do it for free? The police won’t do anything. Their lunch was eaten by the consequences of their own bourgeois morality.

    1. If course, that’s if we assume shoplifting is a technical problem to be solved.

      What if it’s an economical one instead? Do you think the Bay Area’s housing policies, high cost of living, etc contribute to the plague of shoplifting? If so, is pursuing a technological solution the right way to handle it?

        1. Considering at least half of the articles on HaD are about circumventing technical solutions, I suspect you may be vastly underestimating the simplicity, speed, oversight, and costs associated with any such solution.

          Perhaps that’s why Amazon abandoned it. ;)

          1. > No I was commenting on the complexity of the social changes needed

            So was I. ;) I suspect those will still wind up being less complex than the seemingly-simple technological solution, given how technology is often able to be circumvented.

          2. Big garage door, order your groceries on an app, garage door opens with your items inside, closes behind you.

            Customer shoplifting has now been solved.

            But customers don’t spend as much when they can’t casually peruse the aisles. So shoplifting is an accepted loss by retailers.

            Anyway, if Amazon really cared about the customer experience, they’d just hire an abundance of well -trained cashiers to check people out, but that cost money.

      1. “if we assume shoplifting is a technical problem to be solved.”

        Maybe some sort of very fast/high power converted RC model truck running openCV/AI route planning, that takes a trolley full of goods, and guns it for the exit door whilst avoiding guards and other shoppers? Then you go collect it 3 miles down the road… :-)

      2. Even if we assume that the bulk of shoplifters aren’t just opportunistic scumbags (which is a big assumption, IMO), it doesn’t really matter to the company that is being robbed. They can’t fix the socio-economic problems. Therefore any solution that mitigates their losses, is potentially valuable.

        1. A single mom-and-pop store won’t be able to do much by itself, but the massive retail corporations certainly can make a dent in the problem.

          Assuming one of the drivers of shoplifting is economic (in reality, this is a complex issue, of which only one driver is likely to be economic, but let’s focus on this for simplicity’s sake!), corporations, which have been driving wages lower, while the cost of living continues going up, can… simply pay people more for the work they do.

          Imagine that those who work for retail stores, who typically live in the area they work, being able to afford basic necessities on a retail paycheck. Do you think that might reduce shoplifting?

          1. You say the basic motivations aren’t economic – but you keep coming back to people shoplifting because they can’t afford to pay.

            Which is economic.

            And not true. Retail workers can and do afford to pay for what they consume.

          2. I think you misread what I said.

            I said there are multiple factors, only one of which is economic. While it’s debatable if HaD is the appropriate forum to discuss socioeconomic issues in, it’s almost certainly not the right place to delve into an entire dissertation on everything that drives a social ill like shoplifting.

            So I focused on the economic driver, given that the original commenter discussed the shoplifting epidemic in the SF Bay area, which is a horrendously expensive area, and not one retail employees are likely to be able to afford to live in.

    2. Why on earth do you think that? Do you think that theft occurs simply due to inefficiencies in processing bank accounts? A junkie with a knife walks into this place without an account or a phone, does what he will and leaves and the police show up four hours later if ever, what does this solve? Or if they set up their own security, he simply waits for other customers to leave its boundaries and steal from them instead.

    3. Unless there’s a security entrance where you are unable to enter, even by force, that won’t solve it. California is known for shoplifting and high crime. There are tons of options to limit that. End the 951 rule to start with. I’m from the Netherlands. California is the last place I’d visit. There are so many states I want to see (especially Montana and Idaho), but California has been scrapped off that list years ago. There are more and more road trips that you can book (flight, hotels and car included) that will have you pick up the car at LAX or SFO, with the next stop being in Arizona or even Nevada, so you don’t have to stay in California. My parents love those roadtrips (they have done dozens of them) and they even get warnings from the travel agency nowadays to leave the state as soon as they get the rental car. That wasn’t a thing 10 years ago.

    4. The best way to massively reduce shoplifting is to make companies like Amazon pay living wages and taxes to fund social care so people don’t have to choose between luxuries like healthcare, food, heat, housing etc.

    1. Most people do. And that’s why I prefer self-service, queues tend to be shorter, or even empty.
      But back on article, it is over-engineered solution to a simple problem. For example, Decathlon uses RFID tags instead of bar code, you just put your stuff in a bin, and everything is scanned. No cameras, no AI…

        1. I’ve tried it in several countries now and I agree for most of them. But here in Switzerland it’s actually great: Discounts are always applied automatically (even single items, they just stick a new barcode over the old one), weighting vegetables is done ahead of time anyway (there are scales in the produce section that print barcodes), there is no scale that weights your bag so you can scan several things after each other quickly (i.e. they trust the customers more), there is no barrier at the exit (so it doesn’t even print a receipt if you don’t want one) and the clerks have remotes, so most of the time they can quickly flag you through remotely. Even the UI is clean and user friendly.

          1. Here they just stick red labels on soon-to-be expired goods for 30% discount, and that isn’t recognized by the scanners.

            People also cheat with the scales and press “carrots” for apples and oranges, since that’s cheaper. A lot of people simply walk through the gates, so there’s a scanner that reads your receipt before it lets you through… etc. The cashier has to physically check that you aren’t presenting a fake ID to buy alcohol…

            It all works if everyone is being totally honest, which they aren’t.

          2. All that distrust started when they cut down on policing and removed the conversion of multiple minor fineable offenses to jail time. Shoplifting and gasoline theft at pumping stations went up the roof.

            But in the official statistics it looks nice because the recidivism rate went down: habitual criminals don’t return to prison because they were never put there in the first place.

    2. If you think that in today’s world things have to be fast (you are always lagging behing) a machine interface would be preferrable when going shopping, but I get the point that humans sometimes will solve your problem faster than machines, sure

    3. If the “Just walk out” thing worked as advertised, it would have been the best of both worlds. You wouldn’t have to interact with either human or machine.

  1. In the UK Aldi also does this just walk out tech.

    Frustratingly it has fairly short opening hours and after a walk around it was clear it could only work with humans monitoring the cameras a significant amount of the time.

    Annoyingly the receipt system doesn’t work with Amex so you simply have to trust you will eventually be charged correctly.

    I don’t mind it I just don’t see much upside I suppose no queuing is cool but it’s fairly empty whenever I go anyways.

    1. Tesco have a handful of stores where they do it too. I’ve visited the Holborn one, which is GetGo only, and Welwyn Garden City (Tesco HQ), which is a hybrid store. That’s a weird feeling – just strolling past the checkouts with a handful of stuff trying not to look like you’re stealing it.

      I’m pleased to see the Aldi page about their Greenwich Shop&Go store specifically addresses one question I’d wondered about ever since Amazon first announced it, but never seen answered before – the tall person problem:

      “Helping other shoppers

      Items you pick up will be added to your bill. When passing other
      shoppers’ items, place the item on an accessible shelf for them to reach”

  2. ACM just did an interview with Medioni, one of the main people behind Just Walk Out:
    https://www.acm.org/articles/people-of-acm/2024/gerard-medioni

    Everyone always says “just use RFID”, but there are operational issues there. Retailers would want manufacturers to put the RFID tags inside packaging, to make it more difficult to swap a tag from a cheap item to an expensive one, for example. Identifying items with high assurance is tricky. As it is, I expect stores see a certain amount of UPC-replacement fraud (shoppers could print UPC codes for cheaper items of the same weight on stickers and substitute them before getting to the self-checkout), though I haven’t looked into it.

    I’ve never been a fan of these schemes myself.

  3. Experiments on reducing product interaction (shelf to cart, item through scanner, item into bag, bag into cart, etc.) has been tried for years. One trial was to have an RFID tag on each item (from the manufacturer, imbedded in the packaging and replacing barcodes) then the entire cart would be rolled through an RFID scanner which would tally the items and either direct debit your account or allow you to swipe your card. However, cost to implement and inefficiencies held this back even if it could reduce labor and deter crime. Another attempt was with scanners on the cart. The customer scanned the item, put it in the cart and before leaving paid on the terminal attached to the cart. It was in effect a rolling self checkout. One modification to this was the cart scanner printed a barcoded receipt which was scanned at a self payment terminal before leaving. However, technology (and costs of the hardware) was not up to it at the time of the trial. Which leaves us with the best (and worst) of technology reduced personal store interaction…the self checkout. Not my favorite (/Luddite mode off) even though I used to work in retail technology.

  4. This kills the Amazon Fresh store for me. The prices aren’t great, and the selection is abysmal. Trying to get anything cut at the deli is amateur hour. The only thing going for it was just-walk-out, that I could fill a bag and not need to unpack-scan-repack it to leave: the promise of technology to free us from cashiers and checkout counters.

    What’s the point of going there now? Enjoying a mediocre store experience? If I’m going to be forced to play with smart carts and still go through a checkout, I’ll go to a real grocery store with decent selection and prices, and I’ll let their clerk to the scanning and bagging for me.

  5. Looks like a whole lot of future e-waste to avoid just hiring a cashier. That said, I’m pretty sure “just walk out” technology works at most stores already (unless they hire security).

    1. > That said, I’m pretty sure “just walk out” technology works at most stores already (unless they hire security).

      LOL, only if the stores don’t mind not receiving payment. ;)

  6. Funny how in a few short years, we went from “AI is going to let tech finally replace drudgery” to “AI is going to create art so that humans are freed up for more drudgery”.

  7. Yes it’s AI. Assistant Indians…

    It’s like when we found out those self driving food delivery robots were actually being driven from low paid workers in Columbia.

      1. I’m not OP, but it looks like Kiwi was doing this around 2019: https://www.dailycal.org/archives/kiwi-hires-colombian-students-to-supervise-kiwibots/article_6065ef5a-0c8f-53d4-8d36-83206668fd72.html

        That was also briefly mentioned in a LA Times story about Coco, which uses drivers who are more local to the service: https://www.latimes.com/business/story/2022-03-17/california-autonomous-sidewalk-food-delivery-robots-coco-starship-kiwibot

  8. Putting scales in the trolleys would be useful as verification, but will likely not be certifiable for trade weights; ie still need to weigh apples on another station, unless the apples are sold “each” rather than by the kg.

  9. Amazon used a known strategy, the Turk, to gather data about a certain process so as to train an AI to be able to do that task, then they decided to scrap the project. Conclusion, the training results failed to meet expectations.

  10. We have an Amazon-unaffiliated chain in Poland doing this sort of thing in some locations and it appears to work. There are caveats, such as that these being tiny stores with mostly ready-to-eat food, and that such store legally qualifying as a vending machine sidesteps Sunday shopping ban. But I’ve talked with some programmers behind the system and that 70% failure rate sounds insane, their early prototypes had something like 20% instead.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.