[pradn]

This is why listings on Amazon are a million words long. They do an adequate job pulling important attributes into the metadata so you can filter on them, for some categories. For example, the size and type of TVs. But it's not really possible to do that for all product categories. Moreover, users can't easily see the attributes in the search page, so it's best to just put it all in the title. For things like motherboards, they get super dense with jargon.

[ThePadawan]

I worked at Switzerland's biggest online retailer whose USP was that it actually did strongly typed attributes really really well.

To date, it's still incredibly searchable. Want to find a 5.5" phone running Android 10+ and at least a 12MP back-facing camera for below $350? You can filter for it.

It was an absolute nightmare on the (normalized, backend) technical level, but a real pleasure to consume on the denormalized, frontend level.

[leokennis]

https://tweakers.net/pricewatch/ is also a great example of this. No idea how they do it.

[ThePadawan]

As someone who worked in the space - there are standards and commercial solutions which all do lots of good work (some are now unfortunately having to break into the AI space, literally because manufacturers are too cheap to use standards).

But they all come at some cost - my employer decided to cheap out slash invest in a highly custom in-house solution instead that relied on lots and lots and lots of data entry.