5 ms·
"Search and categories are complete garbage." I'm consulting in the pet space, if you search for 'Mousr' which is the name of a very cool pet toy brand that th
by sonnyblarney 8y ago
"Search and categories are complete garbage."
I'm consulting in the pet space, if you search for 'Mousr' which is the name of a very cool pet toy brand that they carry - you get PC mouses (ie 'Mouse'). Not the brand 'Mousr'.
So basically their 'spell correct' 'Mousr->Mouse' utterly and completely overrides the notion that the users wanted to in fact input 'Mousr' to get that product.
Instead, in order to find the product, you have to type the company name 'Petronics'.
FYI - if you've worked in NLP, you might recognize this problem: the edit distance from Mousr->Mouse is short, and the 'e' and 'r' keys are beside one another, and of course the word 'Mouse' is very common while the word 'Mousr' is very uncommon, leading the error correction algorithm to 'over correct'.
This is a problem in NLP where very uncommon things that are lexically adjacent to common things get suppressed.
Since 'Mousr' however, is 100% in the Amazon catalougue, you'd think they'd include it!
But no.
Consider the resulting 'lowest common denominator' approach: statistically speaking, it might be slightly more economically efficient to actually assume the error and show PC mice, besides, 'everyone' needs a PC mouse anyhow!
But the interesting, long-tail stuff gets suppressed.
This is to me the statistical epitome of not just Amazon, but of our new, globalized culture - the big brands have such power they just bulldoze through everything else - and the 'short run' economics are in their favour! They can point to 'the efficiency of the search' etc. to validate to everyone else in the equation (ie city councillors approving permits, banks giving loans, buyers - heck, even consumers) to validate their case on competitive, culturally secular grounds.
It takes quite some magic, a loyal following, a magic meme of attention, a super faithful investor, or deep rooted cultural behaviours (like Italians who could care less about the new 'Starbucks' because they already have good coffee).
Most big CEO founders have made some kind of 'big claims' or almost arbitrary statements about 'how it ought to be' and they stuck to their guns on those, I wish Bezos et. al. would think just one tiny bit about the 2cnd order effects of their businesses.
- manigandham 8y agoIt is completely reasonable to optimize for the vast majority typing 'mouse' and potentially misspelling it. Also "mousr petronics" shows up as the 1st choice in the autocomplete when typing "mousr". This is probably one of those products that can and should have its own website and sell from there or link directly to Amazon instead.
- sonnyblarney 8y agoI don't think it's reasonable to show one thing when the user typed another. There are dozens of product results, there's no reason why one of the proper results wouldn't show up as well. "Results for Mousr ..." "Results for Mouse ..." etc
- whoisjuan 8y agoThe 'r' is next to the 'e' in a qwerty keyboard. Quite honestly it's more likely for someone to make a typo and enter "mousr" than your case which is someone searching for a brand name that happens to be "mousr". It seems like the right optimization IMHO. I presume the p90 case for this would be people looking for an actual "mouse."
- cloakandswagger 8y agoThere is a link on the search result page for "Search instead for X" if you really want to search for "Mousr". There's also nothing wrong with correcting what appears to be a typo'd word for a user (Google and every other search engine I can think of does this). Seems more like poor naming on Mousr's part.
- sonnyblarney 8y agoTotally disagree, and it seems nobody understood my comment. I pointed out in that I'm fully aware where the probabilities are, and the fact the 'keys are close to one another' - that's my whole point, it's a statistical suppression. The 'short view' is to just show 'mice' - but there are problems with that: 1) 'over correction' is far worse error that 'insufficient correction' 2) There are tons of 'Mousr' products out there, i.e. products that overlap closely to more common things. The likelihood of some random error being a product is low - and since 'Mousr' is an actual product, it makes sense to show it. In this case, the rational product choice would be to show 'Mousr' first, and to not assume over-correction. To take a more extreme example i.e. going with the 'highest probability case' ... it might make sense by that logic to show only the most commonly purchases products purchase, almost whatever is typed.
- blackoil 8y agoIt is 150 USD toy for cats. i.e. 99.9% people typing mousr are searching for mouse(PC).