Hacker News Reader

Comment by amelius | Hacker News Reader

I'm wondering why, in this day and age, it is still difficult to search for a specific type of jeans, with a specific size, style and color, and availability. And filter it by shipping method/cost.

Same for other products. E.g. resistor 0805, 1/8W or better, etc. etc.

Why does every online shop have their own (often broken) filtering system? Why can't the information be gathered by an AI, and turned into a super filter?

View on HN

Google had "Froogle" specifically for shopping. Not sure what happened there, as it looks like they renamed it to "Google shopping" (https://www.google.com/shopping) but it just redirects to broken google search page.

https://en.wikipedia.org/wiki/Google_Shopping

I feel that

I was looking for braided BSP flexible stainless hoses in 3/8th of an inch with female to female ends that were longer than 1m

Boy I've had to scroll through pages of 1/2 inch hoses and female/male ends because fuzzy word search is completely unfit for that stuff

You can't even get Amazon to give you the right results based on filtering items by title. I was searching the other day for lithium rechargable AAA batteries and it gives you a mess of mostly AA batteries. Including ni-cad.

Interestingly eBay does a better job now of guiding sellers in filtering items into the proper categories, at least for the products I shop for.

Amazon is a dumpster fire.

Because nobody benefits from providing that information. Having choice isn't good for sellers. They want to sell you their product, not allow you to check whether their product+shipping is the cheapest.

1) AI isn't that smart and doesn't search the Internet in real time, nor can it distinguish between real and counterfeit products. There's a significant liability issue if BigCorp AI sends visitors to IP-infringing shop sites.

2) Search engines would rather show you irrelevant results than an empty page. This is true of Google, Amazon, Netflix, you name it.

3) The more custom a product is, the less likely it is to actually be in stock. I've seen plenty of cool products on Chinese electronics stores, many of which have quietly refunded my order after determining it wasn't in stock and wasn't coming back.

That's a pessimistic view. I think with even small effort a lot of improvement can be made here.

There will no doubt be some techbro startup that promises to solve this: after all, who wouldn't want to build a platform that guides you to these products and take a cut of every sale?

Whether or not such a platform actually meets your desired specifications is another matter. The money in this industry is in affiliate marketing and ads, not in helping people find stuff.

I can see where this is coming from but if your view was correct, then companies like Kagi would not exist.

When I wrote my first webstore I thought the product data would be reasonably static but I've gradually learned it is more like a first person shooter.

AI could still gather it but one would have to accept that all parts of a product may change which may create an undesirable game.

The most annoying thing, and reason why I don't "search" for clothes online, is that all the (web) store results are US based. What's the point of getting results in the US when I don't live there?

If you immediately find what you want you won't browse around and likely buy something with bigger profit margin or just more stuff.

It's not good for business to provide a good user experience, correct information and so on.

For some products, this exists. I use Geizhals/Skinflint, which is technically a price comparison site, for pretty much every purchasing decision. If they support that product category (they're mostly focused on tech).

But if they do support it, you get all the filters you could ever ask for. This is the online searching experience the 90s promised. e.g., https://skinflint.co.uk/?cat=hdssd or

Yes! But the point is that nowadays an AI should be able to collect all this information for all product categories, and provide a uniform UI. But indeed, this is the kind of filtering I'm thinking of. It would also be nice, if, on the product page you can easily search for alternatives. I.e., all the filtering options should be presented for that product, and you should be able to make modifications and then hit "find alternatives".

AI is a heuristic like any other. You don't want heuristics for this, you want actual data.

Manufacturers should just be required to provide a standardized data sheet for their product (as they already are for e.g. energy consumption data in EU), which could then be automatically mapped. Then you just need stores to list the barcode for each product, and you can map everything automatically.

Still not complete. Try searching for Elitegroup in the category Mainboards there, and see what you get.

It seems like they closed shop, or stopped distribution.

Which isn't the case at all.

That applies to several other brands in that category.

Extrapolating from that to other categories I know less of, I assume similar omissions in coverage.

Why that is escapes me.

Maybe there is a fee to be paid, and they couldn't be bothered?

You are incredibely right, I recently buy 2 jeans 514 33/34 ... and they are not the same size as thus bought one year before ... Seriously Levi's are you drunk?

Drunk on cost savings from moving production from factory to factory around the globe.

I can think of a couple of reasons: manufacturing tolerance, differently sourced / dyed fabrics (make sure to wash your clothes before first use).

There is also something called vanity sizing (https://en.wikipedia.org/wiki/Vanity_sizing) where sizes go up without the numbers going up, so that people still think they're the same size.

If you're buying clothes a year apart, they're likely using a different supplier using different grades of fabric, differently calibrated machines, so the clothes won't feel or fit the same.

Google Shopping probably comes closest with its Vision Match feature: https://www.google.com/shopping/page/visionmatch

I think we’re close to the solution. The problem for the 2000s was getting everything online, 2010s was getting APIs setup and I hope the 2020s will be when we build truly useful filter engines.

Probably for the same reason you can't ask in a Target whether the Wal-mart up the street has it in your size - the individual vendors are incentivized to avoid you going to another source, and hope you'll decide to go with something "good enough" rather than paying the cost of going to another place and comparing things.

That cost is much lower outside of brick and mortar travel times, but not 0, especially for people on crap internet connections or slow computers, and people tend to just want to get it done rather than shopping through N sites for the best option, so they pick "good enough".

That, plus the vendors don't want you to think they're just an Amazon frontend so they have their own UI, plus if they allow suppliers to populate the item fields, good luck getting the suppliers to provide consistent results...

> the individual vendors are incentivized to avoid you going to another source

Yes, but a search provider would not have that incentive. It can simply scrape the shop websites and parse it with AI, then turn it into filters.

Only if that AI does not hallucinate, otherwise it would be useless.

Sounds like you are talking about Amazon? Which is a front end for online shops.

Amazon's search is somehow even worse even though they already have a lot of the information in a structured format.

I suspect that's by design.

For the same reason some department stores have the up and down stairs/escalators on opposite sides.

https://ux.stackexchange.com/questions/49307/why-do-departme...

I'm talking about the web, not just Amazon. Amazon's filtering is limited too. I want a universal filter with many options. And not just shops that are captured in Amazon's sphere. And my search engine should have no incentives other than providing me with good search results.

Because most products are in their own ecosystem with no overarching standard way to share information. You have some standards but usually for specific niches, no product FHIR.

Just making sure every product has a unique ID would be a revolution. https://en.wikipedia.org/wiki/Global_Trade_Item_Number

Google’s Froogle was a bit like this. Obviously they had to cancel it.

People talk about the web being corrupted by ads, but it's important to recognize that it completely fails at informing people what products are available to solve the problems they have, at what prices, from which vendors, and how to transport them most efficiently. Instead it's like walking through an unregulated flea market full of pickpockets and everybody's asking you to sign documents.

Yeah, everyone is talking about free markets. But no one likes the idea of a transparent market.

I guess they talk about free-to-scam-you markets, not about free-to-move-and-see markets.

It's not the web that fails. It is the vendors who fail, which is their own individual fault. Any vendor who takes online sales seriously can have tremendous benefit.

Amazon has terrible filters for that type of thing.

Reply on news.ycombinator.com