Amazon has hundreds of millions of items, search at that scale is not as easy as it is for a few thousand books. E-commerce search is not the same as google search. Relevance by words and associations like google creates is much different than how people search when shopping/exploring.

If you have a better way to do it, Amazon is hiring and I am sure they would love to hear your ideas.

I am not so sure they really care about changing their search.

They only recently started to change their recommendation stack, after years of using outdated technologies. I think it is somewhat similar with search.

It's not so much that "it's complicated at their scale" as "It works somewhat okay, which translates into X hundreds of millions of $, so don't touch it".

I know they could (and they most probably do) run A/B tests, but the thing is, how do you run MVT tests on such a large infrastructure, with such a varied catalog, etc...it all gets very complicated at their scale, just from an org perspective.

Many "power users" complain about the state of Amazon search, but I doubt the average joe really cares about that.

A more recent (and with a much smaller catalog) player, zalando (rocket internet) has a modular org and a modular approach to their stacks and they, apparently, change or update their tools much more often than Amazon.

I am pretty sure zalando being more "agile" come from their much smaller size (both in inventory, turnover and org).

Large organisations have a harder time moving than smaller ones, after all...

Thank you!