Given that I have, more than once, received a huge box from Amazon holding an extremely small item, I don't think they have unless they're really bad at packaging stock management and had to stuff things into whatever they had available?
(To be fair, it has been infrequent but then I'm just one person and I don't believe I'm a statistical anomaly which suggests that at Amazon scale, it's happening thousands of times a day.)
I assume that's exactly what happened.
You don't have to be "really bad at packaging stock management", you just have to have a supplier that is late delivering certain box sizes on a certain day. Late fulfillment is far more common than late ordering when it comes to automated systems like these.
(And the more days' worth of box stock you try to maintain in case box delivery is late, the more it costs you in warehousing. So occasionally getting a box too big is probably going to be economically optimal, when the box suppliers aren't completely perfect in supplying on time.)
Amazon employees are just scrambling to meet inhumane and unrealistic metrics set and enforced by assholes and algorithms that have zero concern for human rights, health, or dignity.
I saw a recent poll that many Americans list Amazon as their favorite retailer, and it blows my mind. Do people really need same day delivery on consumer goods, for the not-low cost of the aforementioned problems?
I think the main reason people use/prefer them is familiarity and availability. There are lots of things that are hard to find in local shops or even other stores online. That's in no small part because of what amazon as done to brick and mortar retail and their online competitors, but I do feel for people who want something but have no idea where they'd even start looking for it except for amazon. The situation is likely to only get worse as search engines grow increasingly less useful.
Even when you want to quit them it can be hard to escape amazon. It's work to try to determine the trustworthiness of random storefronts you've never heard of before entrusting them with your credit card info, and I've even heard stories of people who ordered from other sites, paying more than they would have on amazon, only to get their item in an amazon box.
No, but what I do need is low prices, reliable deliveries, easy returns, and a lot of customer reviews with photos and videos so I can get a sense of what the product actually is, and what kinds of things people like about it, and what kinds of problems they're having with it.
Amazon is literally the only delivery service in my area that doesn't just leave packages in front of my building to be stolen, but actually follows my delivery instructions for a code that lets them in -- and even has a place to put that in in the first place.
Amazon is the only store where if a product doesn't work or wasn't accurately described in the listing, I can easily drop it off at Whole Foods or UPS without having to box it up and print a label, and I don't have to worry about checking four weeks later whether I actually got refunded on my credit card because they always refund within a couple of hours. I also don't have to worry about submitting a "return request" and then waiting for an "RMA number" and then paying for return shipping.
And Amazon is the only place where popular products across all categories have thousands of reviews, rather than 10 or 5 or just 1, which lets me avoid a lot more returns than I would have to make otherwise.
Amazon sure isn't perfect, but the reality is that it's both cheaper and better than anything else for a large proportion of the things I need to buy.
> Amazon is the only place where X
It will continue to be that way, until every other option closes, and then Amazon reverts to the norms you see elsewhere, because they have no competition remaining and no reason to provide good service.
Time in a weekday tends to be at a steep premium for me personally; between Amazon and my local grocery store that offers online ordering and in-person pickup, these offerings are a lifesaver.
If I were cording this for amazon, I would start with a known algorithm, something like the skyline texture atlas packing algorithm. Texture atlases are old school ways to send graphics to GPUs and they had to be certain sizes even if what you wanted to draw was much small. Like if you want to draw single letters, for efficiency packing a whole font and related graphics onto a single bitmap and sending it one go was usually more efficient. This has to do with batches once being very expensive and the only way to send stuff over AGP/PCIE to the GPU, I here it is better now but I don't really know.
The skyline algorithm starts by lining up all the biggest textures along one edge, then it works down the size until one edge was full. Then it fills gaps with the largest remaining textures. This sometimes fails because "largest" is fuzzy, both longest single edge and total pixel can lead to quirks so game specific heuristics often need to be used (like creating a point system for "largest" where each item gets 1 point per pixel and 2 points per pixel if one dimension is over X or something). It was called the skyline packer because if you ran it with just colored boxes what it drew often looked like a low res city skyline, if you rotated it or packed along the bottom edge (but you always pack from topleft out because of how RAM and storage are efficiently iterated and prediction hardware, and other details)
Here is a stackexchange post where they discuss "sorting things in scanline order" by which I think they mean height and then they pack the top edge left to right. If they packed bottom first or if you flip/rotate it you might see the skyline effect, if you squint real hard:
https://gamedev.stackexchange.com/questions/2829/texture-pac...
But once that is done for the bottom plane of the box other stuff can be jammed in on top. Also not doing it in... Ruby(?), which is what I think candy Japan used, is likely to help. I like Ruby just fine, but I did a pixel comparisons routine for 1080p screenshots once in it and it took like a minute for one comparison of two images. I wrote a c-extension and that dropped to 1 second. I would guess that modern multithreaded/SIMD C++ or Rust could knock out most Amazon packing sized packing problems exhaustively really fast. Still not a great solution, but 10s per package is likely faster than the human packers can work and I would feel comfortable making such promises to a customer (then hopefully deliver ms length solutions).
And then for bigger problem spaces we can talk Monte Carlo... some other time.