This is pretty sad to read. Before Covid, the Amazon Go store experience was phenomenal. All the convenience of a 7-Eleven but with the pricing of a normal grocery store. The food options were really good and the BlueApron style meal-kits were amazing. The Alexa integration was also nice for being able to just verbally ask what's the next step on a recipe while you're busy stirring or chopping things.
When it rolled out to Amazon Fresh stores, it was a breadth of fresh air. The painful clunkiness of self-checkout was gone. The slow and pointless exercise of unloading and reloading your cart was gone. You could just bring your reusable shopping bags, throw stuff in, and walk home. By far the most hassle-free shopping experience to be had.
Scan as you shop is a big step backwards and feels like you've got the annoying self-checkout experience looming over you the entire time you're there.
The selection and operating hours both took a hit during covid and never recovered.
Lots of services would be phenomenal when you can offload much of the cost to run them to cheap offshore labor.
Remember how great Uber and Doordash were when much of the cost of operating was offset by underpaid workers and VC funds? Now that they don't have unlimited piles of money and cities/states are making them pay more fair wages, the cost-benefit of those services has diminished. I was ok paying $5 for $20 of food to be delivered, but now it's more like $10 - $15 in fees/tip, plus fees hidden in menu prices making that $20 food cost $27.50.
From my point of view, Uber and DD were always going to head in that direction because by any objective measure, they are worse than the solution that already existed. The introduction of a middleman created new costs for drivers, customers, and restaurants alike, yet it's actually worse at delivering food than the conventional "drivers work for the restaurant" model. Sure, you get one website where you can see all the delivery options in the area, but the cost is making all of those options shittier. I could rant about how stupid that industry is for hours, but I digress.
What Amazon is doing is different because they're actually doing something that adds value, and they're solving a problem that actually exists. Maybe I'm naive for believing any of this matters.
To be honest, I never thought much about how the whole system worked. I've only been in an Amazon Fresh twice. It had the "just walk out" thing but I don't have a Prime account so I never used it. I just figured it was UHF RFID. Seems like an obvious solution, but I'm not the engineer responsible for figuring out what to do if you put two exits next to each other.
"Underpaid" and "fair" is a modern political myth, except in the instances when there is a concerted effort to undermine the semi-natural labor market. Pay will never be enough, as a point of fact. It will always be "underpaid" and "unfair" in discussion. As this is an immortal pressure and negotiating tactic. It is incentivized, and therefore we will get more of it. That works both ways, but at some point there has to be a reckoning that accepts that it is better to have employed than unemployed unskilled workers. In the context of a massive and growing unskilled population base. Yet the same people who demand that everyone earn a living wage, sometimes punctuating their points with riots, tend to be the same crowd who wants to import unskilled workers by the millions.
There is no effectively borderless world in which most are both employed and not "underpaid". It won't happen both ways. In fact, the pay trend is about to tip massively against fairness. In a manner that even will make economic rationalists uncomfortable. These "fair pay" efforts are a distraction before that storm.
Problem is, when you get an economy acclimated to false prices set by adversarial “business” “strategy”, the only logical result is depression when it’s time to actually create money for investors
Fair warning to all who come after - the children comment for a decent distance below are a battle based on correct/ incorrect/ technical/ personal definitions of skilled labor.
wrt DoorDash (and Instacart etc), along with this squeeze I’ve observed the quality of service decrease severely. Could just be that my sample of a now wider employee pool is crappy, or workers are trying to “scale” to make a living wage with many concurrent orders, but I believe it’s part of the race to the bottom. I love the convenience, but the experience is worsening. How can these vendors continue to profit when lower quality leads to increased refunds?
> Though it seemed completely automated, Just Walk Out relied on more than 1,000 people in India watching and labeling videos to ensure accurate checkouts. The cashiers were simply moved off-site, and they watched you as you shopped.
Wow - I did not know this. This makes it all a whole lot less impressive and interesting that it was just people off shore watching you.
Apparently this news came out in May 2023. I also was bamboozled into thinking that sophisticated computer vision algos -- that worked -- were doing this.
I'm now picturing the remote workers constantly switching between cameras, studying: Did he put down the can of kidney beans or the can of corn there? Wait, the man picked up the bananas, but then he handed them to the woman in the white shirt. Let's charge them to her account. Wait, she handed them back at 14:42 in aisle 16. Going to switch them back to the man.
All day long, day in day out, for years. I am the last to criticize 'low wage jobs' paternalistically, because I know they may be much better than what the workers might otherwise be doing: perhaps just toiling in the fields for 16 hours a day (or worse: something like 'melting down discarded PCBs to recover trace metals'). But still, I do not think I would want to do this job nor that it was worth it. I get that they were supposedly trying to train an algorithm to do it. I'm glad that they aren't keeping at it any longer though now that it's proven so unworkable.
> Though it seemed completely automated, Just Walk Out relied on more than 1,000 people in India watching and labeling videos to ensure accurate checkouts. The cashiers were simply moved off-site, and they watched you as you shopped.
Reminds me of the "delivery robots" that weren't really automated after all, they were remotely navigated by cheaper workers on playstation controllers in Brazil and the Philippines driving them on the streets through cameras.
It's just a tech-illiterate journalist who can't seem to understand the difference between "annotators watching and labeling videos to validate the model" vs "people watching the videos live to remotely decide the cost of every user's purchase".
Or maybe they do know the difference, but wanted to bait audience.
Sounds like fake-it-till-you-make-it stuff right? This is exactly how I would bootstrap a startup that wanted to do this. Kind of impressed with their scrappiness to be honest.
What I think is funny is that circa 2008 I had a manager who used to work at Amazon who told me that "a surprising amount of Amazon artificial intelligence is artificial artificial intelligence, low paid workers".
I heard this was behind mechanical turk. Sounds like the playbook remained the same.
Regular grocery stores have really gone downhill as well. Whenever I shop, there’s at most 1 cashier doing checkout and usually 0 (only self-checkout being open). I consider myself pretty proficient about knowing what sets off the machine but still set it off 60% of the time (about some weight imbalance etc) that requires an attendant to come fix manually. I’ve gotten to dread the grocery store trips as they require so much overhead time. I really wish the “just walk out” could’ve been popularized and caught on at more stores.
I live in Western Europe and the self checkouts don’t have scales, except for weighing produce, and even for buying alcohol, they trust you to say you’re over 16. It’s always fast because you wait in one line for 4-8 self checkouts and take the first free one.
On the other hand, using the cashier is often frustrating. You have to self-bag anyway and the cashier won’t start scanning your items until the person in front of you has bagged all of their stuff, had a chat about the weather and counted their change. The bagging area is always way too small, making all of this take a while. A few times I’ve even waited while one cashier counts their float, leaves, and a new one comes and counts their float.
There is one grocery store near me that always has at least two staff members ready to work the checkout and no self checkout at all. There are virtually never lines more than two persons deep. The only time I ever go anywhere else is when I absolutely have to. Vote with your wallet.
Yeah, the "if anything goes wrong you now stand around with your thumb up your ass while one employee makes their way from broken kiosk to broken kiosk to manually resolve the problems" model has soured me on self-checkout. I find I have a very low "dealing with this bullshit" limit, to the point that if I have trouble I'm likely to just say fuck it and walk out of the store without completing my purchase.
What never made sense to me with their go stores was why a store that only needed 1-2 people max to operate had such bad hours. Hearing now that getting the bill is a mainly manual process i guess their hours had to line up with their data entry team in india so people could get their recipes quickly. insane to think about
I'm glad it's gone for good, if the process really works like how it's described in the article. Thousands of poor souls doing terrible pointless menial work just so that a few entitled customers can avoid the clunky self checkout (eww- the horror!). The entitlement of the west has no bounds really.
It wasn’t pointless, they were training an AI that never fully worked. It’s ultimately the same kind of thing as people monitoring self driving cars, a boring task that may be pointless or possibly remove a lot of drudgery longer term.
According to The Information, 700 out of 1,000 Just Walk Out sales required human reviewers as of 2022. This widely missed Amazon’s internal goals of reaching less than 50 reviews per 1,000 sales. Amazon called this characterization inaccurate, and disputes how many purchases require reviews.
> Thousands of poor souls doing terrible pointless menial work just so that a few entitled customers can avoid the clunky self checkout
Wanting to make your life more convenient and pleasant isn't "entitled".
99% of jobs are things people would rather not be doing (otherwise you wouldn't be getting paid for it). The point is that we can allocate this work in a way that minimizes the amount of time everyone has to spend doing undesirable work.
Are you mad that I sometimes pay "poor souls" to do the "menial work" of cooking me food so I can avoid doing it myself?
> The entitlement of the west has no bounds really.
A very bizarre response to "darn, this was so convenient" - I wonder if this is a troll.
The point of the store was to provide the appearance of conventional retail while being exclusive to "members". This would theoretically cut down on shoplifting and boost profits at the expense of shutting out marginalized people.
My first thought is that this was an attempt at satire, but now I realize it's a privileged Western person who has never traveled to a poor country and thinks that if these workers weren't labeling data they'd be self-actualized DJs or crypto investors.
"Amazon Go stores will still use Just Walk Out technology, and the company will continue to license it to other retailers. Smaller stores in the UK also will keep using the system."
> The painful clunkiness of self-checkout was gone.
It could be worse. Imagine a smart gate that refuses to open for wheelchair users, or claims the child in your arms is an unpaid item - something that is getting rolled out in many Australian supermarkets.
Shopping there always felt supremely weird to me. Scanning in, getting stuff, and walking out, but I always felt incomplete without a receipt being right there in my email, wondering if I'd done something wrong and had just inadvertently shoplifted.
I shop more often at Walmart, which has recently increased the number of manned checkout lanes and restricted their self-checkout to 15 items or less.
The local Walmart expanded self-checkout and it works surprisingly well (the bag scale seems to be very loosely calibrated or off). I wonder if they're doing things differently depending on how much product walks out the door.
> Scan as you shop is a big step backwards and feels like you've got the annoying self-checkout experience looming over you the entire time you're there.
How do the Amazon "Dash Carts" actually work though? If it were a traditional bar code scanner, I'd agree with you. But reading up on the tech makes it sound like you just place the item in your cart and that's it. If it worked like that I don't see at all what the problem would be.
But this was an obvious sham and a fraud that couldn't scale. I don't think its demise is sad at all.
What's sad is putting cashiers out of work, and even worse is replacing them with outrageously piss-poor alternatives. Self-checkout is a clinic on incompetent system design, and has been for DECADES now. It's mind-bogglingly bad, all to take jobs away from people. Fuck that.
Why is it a fraud? Did they ever tell you it was entirely computer-driven? I thought their value proposition was you could just ... walk out. This feels pretty inescapably the future. Why have people do bad jobs like cashier work when computers could do them instead? This is why UBI is going to be critical. We're moving to a world where we'll have more people than work, and that's not just ok - it's fantastic.
They're not wrong, they're just early. Which you could argue is the same thing on a micro scale, but not on a macro.
there's no way they were turning a profit. they definitely missed stuff all the time even with a ton of sensors. and sensors aren't the only cost. annotation is by far the most costly operational expense. new product? needs several annotated photos and recalibrated weight sensors. merchant decides to put Christmas branding on the same UPC? now all your vision models are poisoned for that product. it needs to be re-annotated for the month and a half it exists and the models need to be swapped out once inventory changes over again. as long as merchants are redesigning products (always) your datasets will be in a constant state of decay. even if your vision sensors are stationary and know the modular design up front, you still need to be able to somewhat generalize in case things get misplaced (big problem for weight sensors) or the camera gets bumped.
between dataset management, technology costs, research costs, rote operational costs, etc this is a very expensive problem to solve. and large models with a ton of parameters are little help; they may lower annotation costs a bit but will increase the cost of compute.
once i really dug into this problem i saw Amazon Go's Just Walk Out for what it really was: a marketing stunt
the biggest cost is not annotators at the scale you're imagining. it is labor costs.
Amazon bet that the federal govt would raise labor costs to $20/hr and all their competitors (besides themselves with this tech) would get wiped out. They even publicly campaigned and lobbied. That didn't come to fruition as the election promises turned to fluff, and the populists simply chose to empower unions instead.
i mean, labor cost (as in in-store labor) is the target for this cost optimization. unfortunately for the time being labor cost is not as significant as the other costs associated with annotation and dataset curation. technology costs are not really significant if this can be pulled off at scale.
in-store employees know where things are supposed to be and why, if at all, items are "misplaced" according to the modular design
I think this failing, and tesla failing to actually ship "self driving" is clear : machine learning definitely has complexity limits, that we are a long way off from perfecting or even getting beyond some reasonable threshold.
> The company’s senior vice president of grocery stores says they’re moving away from Just Walk Out, which relied on cameras and sensors to track what people were leaving the store with. [emphasis mine]
> Though it seemed completely automated, Just Walk Out relied on more than 1,000 people in India watching and labeling videos to ensure accurate checkouts.
> According to The Information, 700 out of 1,000 Just Walk Out sales required human reviewers as of 2022. This widely missed Amazon’s internal goals of reaching less than 50 reviews per 1,000 sales. Amazon called this characterization inaccurate, and disputes how many purchases require reviews.
> “The primary role of our Machine Learning data associates is to annotate video images, which is necessary for continuously improving the underlying machine learning model powering,” said an Amazon spokesperson to Gizmodo. However, the spokesperson acknowledged these associates validate “a small minority” of shopping visits when AI can’t determine a purchase.
The article is kind of all over the place, but it sounds like there were lots of sensors and also lots of human intervention.
How well does simulated data work in this space? My first stab at doing this scalably would be as follows: given a new product, physically obtain a single instance of the product (or ideally a 3d model, but seems like a big ask from manufacturers at this stage), capture images of it from every conceivable angle and a variety of lighting conditions (seems like you could automate this data capture pretty well with a robotic arm to rotate the object and some kind of lighting rig), get an instance mask for each image (using either human annotator or a 3d reconstruction method or a FG-BG segmentation model), paste those instances on random background images (e.g. from any large image dataset), add distractor objects and other augmentations, and finally train a model on the resulting dataset. Helps that many grocery items are relatively rigid (boxes, bottles, etc). I guess this would only work for e.g. boxes and bottles, which always look the same, you'd need a lot more variety for things like fruit and veg that are non rigid and have a lot of variety in their appearance, and we'd need to take into account changing packaging as well.
as mentioned in another comment, "scale" is not just horizontal, it's vertical as well. with millions of products (UPCs) across different visual tolerances it's hard to generalize. your annotation method is indeed more efficient than a multistep "go take a bunch of pictures and upload them to our severs for annotators" but is still costly in terms of stakeholder buy-in, R&D, hardware costs, and indeed labor. if you can scope your verticals such that you only have, say, 1000 products the problem become feasible, but once you start to scale to an actual grocery store or bodega with ever-shifting visual data requirements the problem doesn't scale well. add in the detail that every store moves merchandise at different rates or has localized merchandise then the problem becomes even more complex.
the simulated data also becomes an issue of cost. we have to produce a realistic (at least according to the model) digital twin that doesn't interfere too much with real data, and measuring that difference is important when you're measuring the difference between Lay's and Lay's Low Sodium.
i'm not saying it's unsolvable. it's just a difficult problem
I want to know why these places don't simply dramatically drop the accuracy requirement.
Rather than giving itemized receipts, give just a total dollar value. Then just make sure that most customers are charged within 10% of the correct amount.
Only if the customer requests and itemized receipt, then go watch the video and generate them one. But after a while most customers won't request it, and that means you can just guess at a dollar amount and as long as you're close-ish (which should be easy based on weight and past purchase history), that's fine.
The problem is taxes. Different items can be taxed differently. Itemized receipts are a must from the get-go unless you want to explain to the government that you underpaid them because you charge within a 10% error-margin.
Come on, it isn’t anything to do with being a “marketing stunt.” Often products like this are expected to lose money at first, but they hope with enough R&D and scale that they can make it successful eventually.
For example, you are pointing out that annotating is costly, but that’s an expense that scales independently of the number of stores. So with enough scale it wouldn’t be as big a deal. Or if they figured out some R&D that could improve it too.
right, that's how it starts. but the improvements in methodology simply aren't there as the ML sector has been laser focused on generality in modeling (GenAI as it's affectionately known). "at scale" doesn't just mean more stores; it means more products and thus more annotation. how many UPCs do you figure there are in a given Target or Whole Foods? i assure you it's in the millions.
one advantages of the Amazon Go initiative is a smaller scope of products.
a big problem in the space is that products that look very similar will be clustered in the same section. large models are very good at generalizing, so they may be more attuned to "this is a Christmas thing", but they won't know that it should be classified as the same UPC as the thing that was in that spot yesterday without you specifically telling it to. how would it know it's not a misplaced product or a random piece of trash? (you won't believe the things you find on store shelves once you start looking) you can definitely speed up your annotation time with something like SAM[1], but it will never know without training or context that it's the same product but Christmas (ie it resolves to the same UPC).
Uniqlo (in Japan and at least in SF) has a cool checkout method.
You drop your clothes into a big bin (I’ve always done it one piece at a time, didn’t think to do all at once) and it adds up all your items. I’ve used this maybe 8 times, 100% accurate so far.
Not sure how it works but I guessed RFID and quick Google appears to confirm:
Self checkout systems based on RFID are very convenient, quick and quite accurate compared to self checkout on grocery stores'l. The main problem on groceries is that the tags are expensive compared to the product (tags prices are in the order of tens of cents depending on manufacturer and size)
Disclaimer: I work on a RFID reader manufacturing company
If we had a better Blue Bin system it might work. Remove tags at home and place in compartment of Blue Bin, and at the recycling centre they get sorted out and sent back to the retailers.
But if we really could scale RFID tag production down to 1 cent each, then we’d likely just throw them out. Not that I’m a fan of throwing away silicon…
The tech in store is powered by RADAR (https://goradar.com) - next gen RFID sensor that can work with lower cost antennas. Still unlikely to ever get to unit economics that work for groceries, but really exciting what this team has created.
As far as I can tell, every Uniqlo has this. And you’re right, this is by far the best checkout experience. It’s what Amazon wishes they’d built.
I can’t imagine there’s that much untapped profit in the grocery business that Amazon could turn a profit with such an expensive and unreliable mess like Just Walk Out and I find it’s so typical Amazon to find out that the man behind the curtain is actually a bunch of offshore workers in India.
Decathlon does it but it's a little harder to implement in a grocery store.
You need to basically have custom packaging so the tags are not easily swappable and the RFID tag is not free too, it could really add up for small price items.
The maybe e-waste problem?
I'm not really familiar with how much of a problem this really is but on a grocery store size it could add up
Uniqlo checkout experience is really how self-checkout should be look like, but unfortunately relies on controlling the manufacturing of all products to implement efficiently.
The closest thing I've seen in a convenience store, was a fully automated 7-Eleven that I ran across in Taipei last year. It seems to rely on smart shelves[1], rather than just product video recognition. This could work for most types of groceries.
This exists in most UNIQLOs it's the same mechanism through which the alarm goes off when someone attempts to steal things with the magnetic RFID attached.
It's borderline ancient tech. So why is it not applied everywhere? Because stores would rather have you line up and buy stuff from the "impulse aisle" and tip or donate at their PoS (higher rates with an employee glaring at your selection)
RFID tags are expensive compared to printing barcodes. The volume of products moved in a grocery store is far larger than that of a clothing store like Uniqlo.
I've had a few times when I bought lots of small items and it missed a few of them. Once, they were recognized when I moved stuff in the bin, another time I made sure I didn't move anything before paying and it worked in my favor.
> Uniqlo (in Japan and at least in SF) has a cool checkout method.
Saw this in the Honolulu location and was similarly impressed. It's got to be pretty close to a theoretical minimum in check out speed, while requiring no labor and no accounts.
Some thrift stores let you buy clothes by the pound here in the US. There are always the same people picking out stuff I swear. Probably picking stuff to resell.
"Ginger Market" on the SJSU campus has attempted to use a similar "Just Walk Out" approach. It has not worked well. I've been double charged, charged for items I did not take, not charged for items I've taken, etc. The refund process was also a pain. It was so bad that they had to stop using it last spring, although they claim they are going to give it another go.
Why even bother? Self checkout or a cashier work so much better and I have a hard time believing they are less expensive. The store is plastered with cameras. Seriously, there must be 100 cameras in the place. That's a lot of video to process, which has got to be costly, whether it is a machine or people who are reviewing it.
There was a startup, Standard Cognition, that offered the same experience, but I checked their website (https://standard.ai/) and it seems they have given up on it too.
Edit:
Looking over their marketing videos now and taking a more optimistic approach, is "just walk out" technology all that useful? It seems they pivoted to a product where is much clearer what the value add is (Predictive analytics, loss prevention, context-aware marketing). I imagine "just walk out" technology was likely pretty expensive to implement, but wouldn't have saved much more that self checkout.
Maybe the lesson here isn't that "it didn't work", and more so "it wasn't economically efficient"
Well from the article it's clear that this camera+ai based detection of purchases never worked for Amazon. They had to rely on Indian contractors watching people remotely. It never technically worked, and even if it did, then yeah I agree that it wouldn't make economic sense. Cameras with that level of fidelity and with 100% coverage, tracking N customers at once, are probably a huge capital expense. And all for slightly faster shopping.
I tried the store in Seattle in 2021 and it was a shitty experience. Overpriced, bad stock, and since few are going to actually trust Amazon to get it right, you still find yourself with the Amazon app open the whole time
Wanted to comment on this part -- Prime itself was nothing more than faster shipping, though yes the difference between 4-6 weeks standard vs 2-days was massive (not just slightly faster). But the point is, Amazon excels at identifying friction points that others have just accepted as industry norms, but which, if unblocked, could actually meaningfully shift consumer behaviors.
Go might be a failed experiment, but "slightly faster shopping" is probably an unfair trivialization of what the experiment hypothesis was really about. A core thesis of Amazon in general is basically, to fanatically remove any unnecessary extra steps/friction/bureaucracy/etc., between a consumer and their act of purchasing.
For another example, think back to why Amazon cared so much about the 1-click patent -- legal validity aside, the idea that you can have 1-click checkout, was pretty revolutionarily customer-obsessed, compared to the average online shopping experience of the early 2000s.
And in fact, the natural progression from 1-click, is of course going down to "0 clicks", which is what things like, subscribe-and-save, memberships/subscriptions, or Alexa/Echo styling clothes for you, are meant to do -- they are meant to shift the consumer's mental model of shopping away from emphasizing the build-up from browsing into the climax moment of then clicking to purchase, to instead make the actual purchase decision more of a hidden-in-the-background/automatic thing, instead of a foreground conscious choice.
I would pay 5% more for my groceries if I never had to wait in the checkout line. I’m a kind of customer that has tried grocery delivery, and didn’t reject it because of the cost, but instead for other reasons like I want to pick out my own food, I don’t want it left sitting when they deliver it, I can get it faster myself etc. I would definitely pay money to make grocery shopping faster and more convenient. I suspect there are a lot like me.
I recently finished a book called "The Secret Life of Groceries" by Benjamin Lorr. Amazing book. I learned that grocery and supermarket stores of the past, seemed to have worked very hard to give the best customer experience possible. Today it feels that customer experience has been replaced by cost cutting.
Self checkout has been rife with problems, internally and socially, but supermarkets keep pushing them. I can only conclude it is cost cutting.
I hate it, especially for large trips to the store.
The scale freaks out too often after I scan an item and put it on the bed, because the weight is slightly off for whatever reason. Then I have to stop scanning and wait for an employee to come over and scan their badge to authorize the sale.
I've gone back to just waiting in line at the standard checkout unless I have < 5 items in my cart.
At least where I live it's a nightmare. Instead of having someone work as a cashier and weight my veggies I now have to weight my veggies myself under the supervision of a worker who's only job is to supervise a machine that is buggy enough to need human intervention for every other client.
It only serves the shop owner because he can have a single drone operating a few self checkout machines instead of a single line, but is that a net positive for society ? doubt, going to a supermarket is alienating enough, we didn't have to make it worse.
And of course they had to add a gate so now you can't even leave the shop without showing the receipt, when you don't find what you want to buy you have to call and wait for an adult to allow you to leave the premise like you're a fucking toddler getting out of the playground area.
> Today it feels that customer experience has been replaced by cost cutting
This seems to be true of all businesses, and I really hate it. Flying sucks, trying to reach a human for any reason sucks, lines are huge everywhere. Life just seems a lot more stressful purely to put a few extra bucks in shareholders' pockets.
One of the various jobs I had during the 90s to put myself through college was stocking store shelves in a large grocery store. During my time there, I placed every item on the shelf by hand, rotated forward, so that the shelves were easy to front and look respectable by the time the morning came. After a few months, it was fairly easy to eyeball an item in the case, lift it while rotating, and place it on the shelf facing forward. It also became reflex to hear something fall and catch it without looking. We also mopped the entire store by hand, twice, before getting down to the business of fronting the shelves one final time.
As time went on, they hired a cleaning crew to clean the floors with a machine. They'd miss things. After my time, they started having their employees cut boxes and just plop entire boxes onto store shelves, ripping the fronts off of them. I'm sure it saves a little bit of time, but you end up with a store that is a complete mess full of discarded cardboard. All of that adds up to a terrible customer experience.
There is a possibly unexpected advantage to stocking the shelves with boxed items: if the customer wants to purchase items in bulk, they can retrieve the boxes just as easily as they were put there. As a customer, I do that a lot for unperishable goods.
> Today it feels that customer experience has been replaced by cost cutting.
I feel like it's the case for literally every single industry, everything is getting worse for the end user point of view. Public services, jobs, hospitals, cars, new houses, online services, mobile apps...
I think globalisation allowed things to get better and cheaper up to a point and now that we can't delocalize to cheaper places we have to cut corners somewhere else and the entire stack is suffering from it
When the savings are not passed on to the customer (and they usually are not, as in this case) it's more accurate and descriptive to call it "profit maximizing."
I think that the scan-as-you-shop style systems are definitely the next realistic step for grocery stores. I really like using them, as they allow me to put stuff directly into my bags as I go.
I think a further improvement on a system like that would be to use the cheap RFID tech (UHF EPC) so that when you walk through the anti-theft barriers, the system knows exactly what you got, and then there's no need to scan anything.
> I think that the scan-as-you-shop style systems are definitely the next realistic step for grocery stores.
Scan-as-you-shop has been around for at least a decade at this point I think, at least in some parts of the world. ICA (supermarket in Sweden) been doing it for as long as I can remember, and I came across this image of the scanner being used at the store on Mediawiki: https://commons.wikimedia.org/wiki/File:Self_checkout_scanne... (Image uploaded 2011)
Surely this exists elsewhere too, or been judged to only work in certain contexts (like a high-trust environment like Sweden) and won't be the next realistic step.
Jup it has been available for a very long time in the Netherlands as well.
When I moved to the UK, what surprised me is that Tesco self check out is so much more cumbersome. It weighs your stuff, stops working when something isn't exactly the right weight. Super annoying to use if you're used to the system used in Sweden and The Netherlands. But that was in stores in the middle of central London, it may be very different in more suburban "big" stores which is where it was first rolled out in NL as well.
Yes. In Switzerland Migros you have this as well, at least in the bigger ones. But they also added scanning with the phone. So you don't need a special scanner, just your phone.
I see a couple people using it at the shops near me in the UK (but it's not that popular here either). It's pretty universally available in the bigger shops, but rare in the little metro shops. I really like just slinging my bag in the trolley and running around the store with a scanner. It's dramatically easier than unpacking a trolley for self-checkout or the old school checkouts. It's annoying when there are security tagged items though, because it defeats the purpose of the self scanning.
> We've had that in France for years, and noone uses that as far as I can tell.
Which part of France are you in? Where I am, Paris suburbs, I see it used all the time in Carrefour.
There is a learning curve, which I think inhibits wider usage. I used it a couple of times, and every time I forgot to scan stuff, or forgot whether I scanned stuff, and had to go over the last scans to check.
Yes, same in the UK. Most stores have scan as you shop, but most people do self checkout. I get the impression scan as you shop is seen as being a bit prissy.
We have it in Germany for years too (in densely populated areas) and my subjective impression is that next to none was using it for a long time but that it's picking up slowly.
For one thing I think these things take years for broad acceptance and for another the current scanners with their bright and large displays are just what was needed to make it attractive for the young and elderly.
Similar in my area (in the US). We've had scan-and-go at the local grocery/everything store for many years, but I really don't see that many people actually using it.
> I think a further improvement on a system like that would be to use the cheap RFID tech (UHF EPC) so that when you walk through the anti-theft barriers, the system knows exactly what you got, and then there's no need to scan anything.
Isn't that basically what Amazon had claimed they were doing here, except apparently maybe they weren't?
Albert Heijn does this in the Netherlands, tied to your loyalty card. You can even import your shopping list and show it on the little hand scanners. At the end, you scan your loyalty card and it shows up on the self checkout.
I think it’s great, I get to go around the shop with a little laser gun.
An even nicer thing about the Albert Heijn self checkout is that I can use my own phone from start to finish. Connect to store wifi, scan items using their app (using the phone camera), pay with my phone (contactless payments) on the self checkout and use my phone to scan an exit barcode at the turnstile. My visits to the store usually last only a few minutes and I don't mind popping in multiple times a week.
Yes it's great. Put everything straight in your bag, no packing at the end.
The only thing I don't like is that they have added the option to the machines to scan all your things there instead of using the portable scanner. Which in my area has resulted in queues at the self checkout... Which used to be very quick when it was only used for the portable scanners.
Scanning with your phone camera sucks its very slow and scanning as you pass the anti theft barriers normalizes running the barriers which will make catching thieves basically impossible. It also runs contra to shoppers expectations. Raises the question of charging the wrong person when multiple phones are nearby as when multiple people are shopping together or just adjacent to one another or when carts are just too close to the barriers. It also assumes that reading is instant and faultless when its not and fails to support anyone who wanders in from the street without an account set up expecting normal payment options. This makes it suitable for a membership club which expects to set up payment as part of setting up membership and useless for the 99% of stores that don't work like that.
Why not wait a few years until even a can of beans has an rfid tag, put a rfid tag in 4 corners of the cart. You should be able to compute which tags are within the bounds of the cart and allow you to pay.
The hardware to read the tags is still too expensive to put in the carts and not liable to be available as part of everyone's phones that soon so the easiest thing to do is retain the existing self checkouts and just skip the part where you scan anything. Roll up to the front. Tap your prominently displayed cart number or numbers tap your card done.
Reducing checkout to 6 seconds from multiple minutes will obviate the new loathsome situation where you actually wait in line to self checkout and will be close enough to normal procedures that it should be easy to transition to.
The scan as you shop systems that I am referring to are where you are given a little barcode scanner with a display, and you go around the shop scanning your items as you pack them. At the end of the shopping trip, you still go to the self checkout machine, but you just transfer your scan list from the portable barcode scanner to the till. At no point are you using a phone camera to scan items. Usually the trolleys have a little mount point for the scanner so you don't even need to touch anything.
I agree that having to scan out via the anti theft barriers is idiotic.
It's like a gift-registry, but you take all the stuff with you when you're done instead of letting other people buy it for you later.
We've got Barcode guns at self-checkout here, and I've begun to arrange my items barcode-up in the cart, so I can just get everything rapidly. I often don't bring my bags with me, because I'm going to have to load everything into the car anyway.
That's become a thing at Home Depot, and I do exactly the same. I wish all the self-checkouts would use guns. Once you get used to just putting things in the cart code-up, it makes checkout pretty painless.
Heck, I even do the 'leave the bags in the car' trick too. I'm going to be using the cart anyway, so why do the extra step at the checkout instead of just loading it directly into the bags in the car.
Glad to know I'm not the only person who thinks this is a good way to do it.
>I often don't bring my bags with me, because I'm going to have to load everything into the car anyway.
Thats an interesting optimization. Don't bag your stuff in the store just walk out with a cart of loose groceries. Especially for someone who simply MUST bring all the groceries inside with one trip, I can just load up a crate.
Interesting, it works the other way around here. You get a portable barcode gun that you carry around the store or put in a sort of cup holder on the cart.
So I always bring a bag or a foldable crate, throw everything in while shopping and then put that in the car.
peel off the RFID tags from the expensive vine ripened tomatoes and replace them with tags from the really cheap ones (people already do this with the plastic stickers)
In the US (mid-Atlantic states) the supermarkets have tried this for years with very little success. Giant (Ahold) in particular persists with this, perhaps because they have a unionized workforce and would like nothing more than to slash their employee count.
Stop and Shop (in New England) had hand scanners you could do this with 15 years ago. I have been shocked that the technology didn't spread. It was so easy to scan and bag then just plug the scanner in at checkout and pay.
Stop and Shop now has self-checkout where you have to scan and move your items one at a time into your bag, because it's doing continuous weight-based reconciliation.
So if you have six of the same yoghurt, you have to scan them and then place them in your bag one at a time. And if you have a 24 pack of sodas, you need to haul it out of your cart onto the scales as you're checking out. And if anything goes wrong (item didn't weight what the system expected; you moved the item too fast onto the scales etc) then you get a "please wait while someone helps you", which involves an employee having to come and clear the error.
I refuse to use those (although I will probably need to use it, eventually), because they do this annoying "BING!" when I pass whatever item they are trying to push.
2+3=5. In general, the weight idea is impossible. For any two items, you can find the LCM in units of the precision of the scale, and then you won't know if they got LCM/a of one or LCM/b of the other.
I think there's another technical issue with using RFID. If you have a basket of a few dozen different RFID chips, there's going to be a lot of collisions in the data they transmit, further reducing reliability.
At Meijer stores I've shopped at, there are fancy scales in the produce section.
They work like this: Scan the barcode on the produce if it has one (or pick it from a picture-list, or search by name, or just key in the 4-digit PLU if you're cool like that), put it on the scale, and it spits out a barcode label that identifies the product and the weight.
At self-checkout, one just scans the generated barcode label and puts the produce in the bagging area like any other item.
Unlike the self-checkout kiosks themselves: There's never any wait to use these machines, so it's an easy process that doesn't involve making other people wait.
This process would work the same, I think, for applications where portable scanners are in-use.
The Wegmans I shopped at in Rochester had scales that would produce a special barcode you could scan into the app that would record both the item and the weight.
See also the RFID tags Uniqlo and Decathlon use for self-checkout. You just place your shopping bag in a bin and it reads all the items within in a second.
My guess is that right now this isn't viable for lower margin food products. Uniqlo and decathlon probably don't sell that much for less than £10. It's going to be a while before you can reasonably do that for a pint of milk.
When I worked there back in 2005, Best Buy had an RFID test setup in the basement of their HQ that could tell what you had in your cart without scanning.
Walmart was also working on similar technology, but from what I heard they couldn’t convince suppliers to include RFID tags on all of their products.
The problem with your second suggestion is that it gives the customer no recourse for improperly tagged/scanned items. Interaction with a human or checkout system before leaving the store is necessary, given how ripe for abuse a totally automated checkout system would be.
Agreed - I think there's also the issue of people needing to be pre-registered with such a system, and not being able to accept cash. I would expect that in the future, systems like that will be an "express lane" for the particularly prepared, but not necessarily the default.
Nothing, the same as current self-checkout systems. You can already easily just put stuff in your pocket or ring expensive items up as incorrect, cheaper things. I suspect shops have done a risk analysis and decided that they'd prefer to have more shoplifting and fewer staff.
I use this regularly at the local stop & shop -- it's been in place for easily 10 years.
You scan your loyalty card (in my case a picture of it on my phone) and it releases a wireless barcode scanner with an android touchscreen phone-like thing in it.
Walk around the store, scan things and put them directly in my bags.
My favorite part is checking out without scanning anything and getting bemused looks (and occasional challenges) from other customers -- sometimes even from store staff.
Definitely saves time in my opinion over any other option. I only need to touch things twice vs 3 or 4 times and the wireless barcode scanner is much more reliable than the cacophony of things yelling at me with the self checkout. Weights don't always match, vision sensors think i'm doing the wrong thing.
I remember the one time I went to a just walk out store, I bought a chocolate and got the bill the next day which felt weird but now makes complete sense if they couldn’t get the Computer Vision to work but just relief on humans to identify this stuff.
I’d contend that even if a much more capable company like OpenAI tried this, they’d get much closer, maybe good enough for a product but this a really ill defined problem where even the best results may not be worth it.
Putting cameras on the roof doesn’t let you see everything a person grabs, there will always be weird occluded angles. On top of that the cameras need to be super high resolution and zoomed in to detect some of the smaller stuff in shelves. This is just for the trustworthy customers, now consider the thieves, they’re gonna have a field day with this, it’s not going to be easy to prevent a determined thief from not fooling the systems. A self checkout cart honestly makes much more sense, still prone to thieves but it at least lets you skip the line much more reliably.
Readit News
When it rolled out to Amazon Fresh stores, it was a breadth of fresh air. The painful clunkiness of self-checkout was gone. The slow and pointless exercise of unloading and reloading your cart was gone. You could just bring your reusable shopping bags, throw stuff in, and walk home. By far the most hassle-free shopping experience to be had.
Scan as you shop is a big step backwards and feels like you've got the annoying self-checkout experience looming over you the entire time you're there.
The selection and operating hours both took a hit during covid and never recovered.
Remember how great Uber and Doordash were when much of the cost of operating was offset by underpaid workers and VC funds? Now that they don't have unlimited piles of money and cities/states are making them pay more fair wages, the cost-benefit of those services has diminished. I was ok paying $5 for $20 of food to be delivered, but now it's more like $10 - $15 in fees/tip, plus fees hidden in menu prices making that $20 food cost $27.50.
What Amazon is doing is different because they're actually doing something that adds value, and they're solving a problem that actually exists. Maybe I'm naive for believing any of this matters.
To be honest, I never thought much about how the whole system worked. I've only been in an Amazon Fresh twice. It had the "just walk out" thing but I don't have a Prime account so I never used it. I just figured it was UHF RFID. Seems like an obvious solution, but I'm not the engineer responsible for figuring out what to do if you put two exits next to each other.
Dominos had this down pat twenty years ago, how come everything with an app is so much more expensive?
There is no effectively borderless world in which most are both employed and not "underpaid". It won't happen both ways. In fact, the pay trend is about to tip massively against fairness. In a manner that even will make economic rationalists uncomfortable. These "fair pay" efforts are a distraction before that storm.
Problem is, when you get an economy acclimated to false prices set by adversarial “business” “strategy”, the only logical result is depression when it’s time to actually create money for investors
Cheap wages? Both of your examples paid decent wages, especially in the example of Uber.
It was just an unprofitable business model subsidized by VC money or Amazon corporate profits.
That sounds like a good thing?
Wow - I did not know this. This makes it all a whole lot less impressive and interesting that it was just people off shore watching you.
I'm now picturing the remote workers constantly switching between cameras, studying: Did he put down the can of kidney beans or the can of corn there? Wait, the man picked up the bananas, but then he handed them to the woman in the white shirt. Let's charge them to her account. Wait, she handed them back at 14:42 in aisle 16. Going to switch them back to the man.
All day long, day in day out, for years. I am the last to criticize 'low wage jobs' paternalistically, because I know they may be much better than what the workers might otherwise be doing: perhaps just toiling in the fields for 16 hours a day (or worse: something like 'melting down discarded PCBs to recover trace metals'). But still, I do not think I would want to do this job nor that it was worth it. I get that they were supposedly trying to train an algorithm to do it. I'm glad that they aren't keeping at it any longer though now that it's proven so unworkable.
This a real-life, genuine Flintstones-esque cartoon gag.
Reminds me of the "delivery robots" that weren't really automated after all, they were remotely navigated by cheaper workers on playstation controllers in Brazil and the Philippines driving them on the streets through cameras.
https://www.thestar.com/news/gta/they-are-cute-pink-robots-w...
Or maybe they do know the difference, but wanted to bait audience.
I heard this was behind mechanical turk. Sounds like the playbook remained the same.
Deleted Comment
On the other hand, using the cashier is often frustrating. You have to self-bag anyway and the cashier won’t start scanning your items until the person in front of you has bagged all of their stuff, had a chat about the weather and counted their change. The bagging area is always way too small, making all of this take a while. A few times I’ve even waited while one cashier counts their float, leaves, and a new one comes and counts their float.
Before covid, the hours were much better too.
According to The Information, 700 out of 1,000 Just Walk Out sales required human reviewers as of 2022. This widely missed Amazon’s internal goals of reaching less than 50 reviews per 1,000 sales. Amazon called this characterization inaccurate, and disputes how many purchases require reviews.
Wanting to make your life more convenient and pleasant isn't "entitled".
99% of jobs are things people would rather not be doing (otherwise you wouldn't be getting paid for it). The point is that we can allocate this work in a way that minimizes the amount of time everyone has to spend doing undesirable work.
Are you mad that I sometimes pay "poor souls" to do the "menial work" of cooking me food so I can avoid doing it myself?
> The entitlement of the west has no bounds really.
A very bizarre response to "darn, this was so convenient" - I wonder if this is a troll.
https://www.bloomberg.com/news/articles/2024-04-02/amazon-pu...
From Bloomberg:
"Amazon Go stores will still use Just Walk Out technology, and the company will continue to license it to other retailers. Smaller stores in the UK also will keep using the system."
It could be worse. Imagine a smart gate that refuses to open for wheelchair users, or claims the child in your arms is an unpaid item - something that is getting rolled out in many Australian supermarkets.
I shop more often at Walmart, which has recently increased the number of manned checkout lanes and restricted their self-checkout to 15 items or less.
How do the Amazon "Dash Carts" actually work though? If it were a traditional bar code scanner, I'd agree with you. But reading up on the tech makes it sound like you just place the item in your cart and that's it. If it worked like that I don't see at all what the problem would be.
What's sad is putting cashiers out of work, and even worse is replacing them with outrageously piss-poor alternatives. Self-checkout is a clinic on incompetent system design, and has been for DECADES now. It's mind-bogglingly bad, all to take jobs away from people. Fuck that.
They're not wrong, they're just early. Which you could argue is the same thing on a micro scale, but not on a macro.
i've worked on a similar product before.
there's no way they were turning a profit. they definitely missed stuff all the time even with a ton of sensors. and sensors aren't the only cost. annotation is by far the most costly operational expense. new product? needs several annotated photos and recalibrated weight sensors. merchant decides to put Christmas branding on the same UPC? now all your vision models are poisoned for that product. it needs to be re-annotated for the month and a half it exists and the models need to be swapped out once inventory changes over again. as long as merchants are redesigning products (always) your datasets will be in a constant state of decay. even if your vision sensors are stationary and know the modular design up front, you still need to be able to somewhat generalize in case things get misplaced (big problem for weight sensors) or the camera gets bumped.
between dataset management, technology costs, research costs, rote operational costs, etc this is a very expensive problem to solve. and large models with a ton of parameters are little help; they may lower annotation costs a bit but will increase the cost of compute.
once i really dug into this problem i saw Amazon Go's Just Walk Out for what it really was: a marketing stunt
Amazon bet that the federal govt would raise labor costs to $20/hr and all their competitors (besides themselves with this tech) would get wiped out. They even publicly campaigned and lobbied. That didn't come to fruition as the election promises turned to fluff, and the populists simply chose to empower unions instead.
in-store employees know where things are supposed to be and why, if at all, items are "misplaced" according to the modular design
> Though it seemed completely automated, Just Walk Out relied on more than 1,000 people in India watching and labeling videos to ensure accurate checkouts.
> According to The Information, 700 out of 1,000 Just Walk Out sales required human reviewers as of 2022. This widely missed Amazon’s internal goals of reaching less than 50 reviews per 1,000 sales. Amazon called this characterization inaccurate, and disputes how many purchases require reviews.
> “The primary role of our Machine Learning data associates is to annotate video images, which is necessary for continuously improving the underlying machine learning model powering,” said an Amazon spokesperson to Gizmodo. However, the spokesperson acknowledged these associates validate “a small minority” of shopping visits when AI can’t determine a purchase.
The article is kind of all over the place, but it sounds like there were lots of sensors and also lots of human intervention.
the simulated data also becomes an issue of cost. we have to produce a realistic (at least according to the model) digital twin that doesn't interfere too much with real data, and measuring that difference is important when you're measuring the difference between Lay's and Lay's Low Sodium.
i'm not saying it's unsolvable. it's just a difficult problem
Rather than giving itemized receipts, give just a total dollar value. Then just make sure that most customers are charged within 10% of the correct amount.
Only if the customer requests and itemized receipt, then go watch the video and generate them one. But after a while most customers won't request it, and that means you can just guess at a dollar amount and as long as you're close-ish (which should be easy based on weight and past purchase history), that's fine.
For example, you are pointing out that annotating is costly, but that’s an expense that scales independently of the number of stores. So with enough scale it wouldn’t be as big a deal. Or if they figured out some R&D that could improve it too.
one advantages of the Amazon Go initiative is a smaller scope of products.
[1]: https://github.com/facebookresearch/segment-anything
You drop your clothes into a big bin (I’ve always done it one piece at a time, didn’t think to do all at once) and it adds up all your items. I’ve used this maybe 8 times, 100% accurate so far.
Not sure how it works but I guessed RFID and quick Google appears to confirm:
https://www.wsj.com/articles/uniqlos-parent-company-bets-big...
Disclaimer: I work on a RFID reader manufacturing company
https://kioskindustry.org/walmart-self-checkout/
https://news.ycombinator.com/item?id=38715111
Unlikely to work for groceries given the costs of RFID tags, I imagine.
But if we really could scale RFID tag production down to 1 cent each, then we’d likely just throw them out. Not that I’m a fan of throwing away silicon…
I can’t imagine there’s that much untapped profit in the grocery business that Amazon could turn a profit with such an expensive and unreliable mess like Just Walk Out and I find it’s so typical Amazon to find out that the man behind the curtain is actually a bunch of offshore workers in India.
You need to basically have custom packaging so the tags are not easily swappable and the RFID tag is not free too, it could really add up for small price items.
The maybe e-waste problem? I'm not really familiar with how much of a problem this really is but on a grocery store size it could add up
The closest thing I've seen in a convenience store, was a fully automated 7-Eleven that I ran across in Taipei last year. It seems to rely on smart shelves[1], rather than just product video recognition. This could work for most types of groceries.
[1] https://www.retail-insight-network.com/news/7-eleven-itri-st...
It's borderline ancient tech. So why is it not applied everywhere? Because stores would rather have you line up and buy stuff from the "impulse aisle" and tip or donate at their PoS (higher rates with an employee glaring at your selection)
RFID tags are expensive compared to printing barcodes. The volume of products moved in a grocery store is far larger than that of a clothing store like Uniqlo.
Saw this in the Honolulu location and was similarly impressed. It's got to be pretty close to a theoretical minimum in check out speed, while requiring no labor and no accounts.
Why even bother? Self checkout or a cashier work so much better and I have a hard time believing they are less expensive. The store is plastered with cameras. Seriously, there must be 100 cameras in the place. That's a lot of video to process, which has got to be costly, whether it is a machine or people who are reviewing it.
Edit:
Looking over their marketing videos now and taking a more optimistic approach, is "just walk out" technology all that useful? It seems they pivoted to a product where is much clearer what the value add is (Predictive analytics, loss prevention, context-aware marketing). I imagine "just walk out" technology was likely pretty expensive to implement, but wouldn't have saved much more that self checkout. Maybe the lesson here isn't that "it didn't work", and more so "it wasn't economically efficient"
Deleted Comment
I tried the store in Seattle in 2021 and it was a shitty experience. Overpriced, bad stock, and since few are going to actually trust Amazon to get it right, you still find yourself with the Amazon app open the whole time
Wanted to comment on this part -- Prime itself was nothing more than faster shipping, though yes the difference between 4-6 weeks standard vs 2-days was massive (not just slightly faster). But the point is, Amazon excels at identifying friction points that others have just accepted as industry norms, but which, if unblocked, could actually meaningfully shift consumer behaviors.
Go might be a failed experiment, but "slightly faster shopping" is probably an unfair trivialization of what the experiment hypothesis was really about. A core thesis of Amazon in general is basically, to fanatically remove any unnecessary extra steps/friction/bureaucracy/etc., between a consumer and their act of purchasing.
For another example, think back to why Amazon cared so much about the 1-click patent -- legal validity aside, the idea that you can have 1-click checkout, was pretty revolutionarily customer-obsessed, compared to the average online shopping experience of the early 2000s.
And in fact, the natural progression from 1-click, is of course going down to "0 clicks", which is what things like, subscribe-and-save, memberships/subscriptions, or Alexa/Echo styling clothes for you, are meant to do -- they are meant to shift the consumer's mental model of shopping away from emphasizing the build-up from browsing into the climax moment of then clicking to purchase, to instead make the actual purchase decision more of a hidden-in-the-background/automatic thing, instead of a foreground conscious choice.
It's just early tech: They were working out the tech, then they'd figure out their market, stock, pricing, etc.
Self checkout has been rife with problems, internally and socially, but supermarkets keep pushing them. I can only conclude it is cost cutting.
The scale freaks out too often after I scan an item and put it on the bed, because the weight is slightly off for whatever reason. Then I have to stop scanning and wait for an employee to come over and scan their badge to authorize the sale.
I've gone back to just waiting in line at the standard checkout unless I have < 5 items in my cart.
It only serves the shop owner because he can have a single drone operating a few self checkout machines instead of a single line, but is that a net positive for society ? doubt, going to a supermarket is alienating enough, we didn't have to make it worse.
And of course they had to add a gate so now you can't even leave the shop without showing the receipt, when you don't find what you want to buy you have to call and wait for an adult to allow you to leave the premise like you're a fucking toddler getting out of the playground area.
This seems to be true of all businesses, and I really hate it. Flying sucks, trying to reach a human for any reason sucks, lines are huge everywhere. Life just seems a lot more stressful purely to put a few extra bucks in shareholders' pockets.
As time went on, they hired a cleaning crew to clean the floors with a machine. They'd miss things. After my time, they started having their employees cut boxes and just plop entire boxes onto store shelves, ripping the fronts off of them. I'm sure it saves a little bit of time, but you end up with a store that is a complete mess full of discarded cardboard. All of that adds up to a terrible customer experience.
One thing that really bothers me is that older stock is not brought to the front and they just push it back with the new stuff.
It's doubly frustrating because sometimes products change so you'll find what you're looking for after digging through three layers of assorted items.
I don't really blame the apathy directly, but there are doubtless solutions over the horizon.
I feel like it's the case for literally every single industry, everything is getting worse for the end user point of view. Public services, jobs, hospitals, cars, new houses, online services, mobile apps...
I think globalisation allowed things to get better and cheaper up to a point and now that we can't delocalize to cheaper places we have to cut corners somewhere else and the entire stack is suffering from it
When the savings are not passed on to the customer (and they usually are not, as in this case) it's more accurate and descriptive to call it "profit maximizing."
I think a further improvement on a system like that would be to use the cheap RFID tech (UHF EPC) so that when you walk through the anti-theft barriers, the system knows exactly what you got, and then there's no need to scan anything.
Scan-as-you-shop has been around for at least a decade at this point I think, at least in some parts of the world. ICA (supermarket in Sweden) been doing it for as long as I can remember, and I came across this image of the scanner being used at the store on Mediawiki: https://commons.wikimedia.org/wiki/File:Self_checkout_scanne... (Image uploaded 2011)
Surely this exists elsewhere too, or been judged to only work in certain contexts (like a high-trust environment like Sweden) and won't be the next realistic step.
When I moved to the UK, what surprised me is that Tesco self check out is so much more cumbersome. It weighs your stuff, stops working when something isn't exactly the right weight. Super annoying to use if you're used to the system used in Sweden and The Netherlands. But that was in stores in the middle of central London, it may be very different in more suburban "big" stores which is where it was first rolled out in NL as well.
I stopped using it after they required using their loyalty cards/apps to use it.
You maybe don't know that Sweden is consistently near the top of shoplifting statistics in Europe.
Looks like they had it in 1996, 28 years ago. https://www.supermarketnews.com/archive/safeway-uk-expands-s...
I just scan everything at the end. Not really any different than going to self checkout.
I also use checkout as a time to organize my groceries for unloading at home.
The main usage I see is mainly self checkout.
Please note that I'm biased towards dense areas. We have on-the-go scanning in super stores as well, I feel it's not being used but I could be wrong
Which part of France are you in? Where I am, Paris suburbs, I see it used all the time in Carrefour.
There is a learning curve, which I think inhibits wider usage. I used it a couple of times, and every time I forgot to scan stuff, or forgot whether I scanned stuff, and had to go over the last scans to check.
For one thing I think these things take years for broad acceptance and for another the current scanners with their bright and large displays are just what was needed to make it attractive for the young and elderly.
Why not? what makes it unpopular?
Isn't that basically what Amazon had claimed they were doing here, except apparently maybe they weren't?
I think it’s great, I get to go around the shop with a little laser gun.
The only thing I don't like is that they have added the option to the machines to scan all your things there instead of using the portable scanner. Which in my area has resulted in queues at the self checkout... Which used to be very quick when it was only used for the portable scanners.
Definitely makes big shops easier.
Why not wait a few years until even a can of beans has an rfid tag, put a rfid tag in 4 corners of the cart. You should be able to compute which tags are within the bounds of the cart and allow you to pay.
The hardware to read the tags is still too expensive to put in the carts and not liable to be available as part of everyone's phones that soon so the easiest thing to do is retain the existing self checkouts and just skip the part where you scan anything. Roll up to the front. Tap your prominently displayed cart number or numbers tap your card done.
Reducing checkout to 6 seconds from multiple minutes will obviate the new loathsome situation where you actually wait in line to self checkout and will be close enough to normal procedures that it should be easy to transition to.
I agree that having to scan out via the anti theft barriers is idiotic.
We've got Barcode guns at self-checkout here, and I've begun to arrange my items barcode-up in the cart, so I can just get everything rapidly. I often don't bring my bags with me, because I'm going to have to load everything into the car anyway.
Heck, I even do the 'leave the bags in the car' trick too. I'm going to be using the cart anyway, so why do the extra step at the checkout instead of just loading it directly into the bags in the car.
Glad to know I'm not the only person who thinks this is a good way to do it.
Thats an interesting optimization. Don't bag your stuff in the store just walk out with a cart of loose groceries. Especially for someone who simply MUST bring all the groceries inside with one trip, I can just load up a crate.
So I always bring a bag or a foldable crate, throw everything in while shopping and then put that in the car.
So if you have six of the same yoghurt, you have to scan them and then place them in your bag one at a time. And if you have a 24 pack of sodas, you need to haul it out of your cart onto the scales as you're checking out. And if anything goes wrong (item didn't weight what the system expected; you moved the item too fast onto the scales etc) then you get a "please wait while someone helps you", which involves an employee having to come and clear the error.
Unfortunately I doubt you'll ever see this happen.
Sure, it's supposedly possible to buy an RFID chip for 1p - but you can buy a can of beans for 23p so that 1p is probably their entire profit margin.
I've also worked with a variety of RFID readers; none of them can provide reliable reading, even in favourable conditions.
My city use it for public transit tickets
They work like this: Scan the barcode on the produce if it has one (or pick it from a picture-list, or search by name, or just key in the 4-digit PLU if you're cool like that), put it on the scale, and it spits out a barcode label that identifies the product and the weight.
At self-checkout, one just scans the generated barcode label and puts the produce in the bagging area like any other item.
Unlike the self-checkout kiosks themselves: There's never any wait to use these machines, so it's an easy process that doesn't involve making other people wait.
This process would work the same, I think, for applications where portable scanners are in-use.
Walmart was also working on similar technology, but from what I heard they couldn’t convince suppliers to include RFID tags on all of their products.
You scan your loyalty card (in my case a picture of it on my phone) and it releases a wireless barcode scanner with an android touchscreen phone-like thing in it.
Walk around the store, scan things and put them directly in my bags.
My favorite part is checking out without scanning anything and getting bemused looks (and occasional challenges) from other customers -- sometimes even from store staff.
Definitely saves time in my opinion over any other option. I only need to touch things twice vs 3 or 4 times and the wireless barcode scanner is much more reliable than the cacophony of things yelling at me with the self checkout. Weights don't always match, vision sensors think i'm doing the wrong thing.
I’d contend that even if a much more capable company like OpenAI tried this, they’d get much closer, maybe good enough for a product but this a really ill defined problem where even the best results may not be worth it.
Putting cameras on the roof doesn’t let you see everything a person grabs, there will always be weird occluded angles. On top of that the cameras need to be super high resolution and zoomed in to detect some of the smaller stuff in shelves. This is just for the trustworthy customers, now consider the thieves, they’re gonna have a field day with this, it’s not going to be easy to prevent a determined thief from not fooling the systems. A self checkout cart honestly makes much more sense, still prone to thieves but it at least lets you skip the line much more reliably.