4 ms·
Deciding what star level to accept should not be the rating-reader's problem. The reviewers need to have their own ratings statistically analyzed, so that thei
by logfromblammo 8y ago
Deciding what star level to accept should not be the rating-reader's problem.
The reviewers need to have their own ratings statistically analyzed, so that their median rating is adjusted to 3 stars out of five for the purposes of averaging across multiple reviewers.
If you only rate things as 5 stars, you have no dynamic range. I want your reviews to all be 3 stars in the average I see.
Another optimization could be to change the aggregation function such that reviewers that rated well the same things I rate well, and who rate poorly the same things I rate poorly, and those reviewers I explicitly mark as having helpful reviews, would be weighted more heavily in the number I see. That would effectively filter out bots that rate their brand as 5 stars and competitors as 1 star, because that pattern is so dissimilar from normal consumer behavior. The bots see cheap white-label crap as 5-star because the recommendations of their fellow ratings-gaming spam-bots dominate their own aggregate rating. If you rate a microwave oven as 1-star because it shipped in 3 days rather than 2, I don't really want your reviews polluting my averages. If you rate a cheap-ass USB cable as 5-stars because it "works good i had no problems", I don't want your reviews polluting my averages. If you rate a USB charger as 1-star because it "Violates USB 3.0 spec and will probably burn your house down; I am an electrical engineer and here's how I figured that out..." then you're danged right I want that reviewer's reviews to have more weight.
I don't see how a system like that would be any more complex than Google's ranking algorithms for search. If they can discover the most useful sites and weed out SEO gaming attempts by analysis of links and traffic, so too could Amazon put an actually useful rating value on the products it ships, and so too could Uber show you what drivers and riders are really like.