You can get better results with a 5minute tsvector in Postgres, At least it will show the words you typed in in the results!
And yes I understand it may be a different "scale" of products, but do people not test these things?
You can get better results with a 5minute tsvector in Postgres, At least it will show the words you typed in in the results!
And yes I understand it may be a different "scale" of products, but do people not test these things?
If you want to get academic with it, you could say something about revealed preferences vs stated preferences (they asked for Coke but they really wanted Pepsi), but I really think they’re just throwing as many suggestions at the user as possible in the hopes they stay engaged on the site and eventually click buy on something.
The only search tools that actually give me great value are history shell with fzf on the terminal, and grep (and similar like ag or ripgrep) in combination with a fuzzy search frontend (similar to piping grep to fzf).
I don't know where I'm going with this. I guess searching just sucks now.
Good website search is hard. Funnily in some respects it's harder than web search, since there are no popularity signals a la pagerank. (There could be such signals, but you have to work hard to create them.)