[heisenzombie]

The product database behind Digikey is much larger but the metadata of lower quality than McMaster. It means that filtering can be very frustrating if the parameter you’re filtering on has entries for, say,

    1 in
    1.0”
    25.4mm
    1”
    1lpi

Another example: If you’re buying something like a cable, they have a parameter for “end 1” and “end 2” with each product making an arbitrary choice about which end is which.

I guess my point is that the taxonomy and cleanliness of the product database is just as important as the front-end UI

[RobertKerans]

Oh definitely. In terms of reference model what I'm looking at is how easy it is use that exposed taxonomy, how fast it is, etc. I can only make inferences about the taxonomies from that (though the quality of the search gives extremely strong indications as to how the data is structured).

The example you give is really good though. It is an indication how well structured/clean the metadata is, and it's something that should be dealt with, even if it's not at the metadata level -- ie some kind of sanitisation/formatting layer just behind the UI layer which parses sizes/weights/etc to a single common unit that can be switched at the customer's preference (even this breaks down though due to sizing subtlteies in groups of similar things). I've worked for a large electrical/electronic components wholesaler (CEF), and was always an issue -- every supplier has their own metadata conventions, and it didn't matter how careful we were, the sheer amount of products flooded our ability to ensure complete consistency.

A different market, but same issue: the ASOS storefronts I like very much, but as they've absorbed different brands and suppliers, the search filtering has visible suffered over the last few years. Sizing (common problem for all retailers) is the most obvious issue