Netflix assigns attributes to each film, such as feel-good, steamy, understated, chilling. Then it regularly asks us to rank movies we've seen and matches attributes of our favorites to create personalized recommendations--for example, 3.5 stars, its best guess for Carol. 3.5 stars usually means I'll like the movie.
Because of the huge book numbers, delivering a personal recommendation may be too difficult or impractical for Amazon, and it's certainly not possible for reviewers. But without personalization, star rankings for books don't mean a lot.
Then there's the other thing: some reviews and likes of reviews may not be legitimate. I heard recently of an author with review blogs under several different names who trashes his rivals and praises the books he writes under a number of pseudonyms. This is a bad thing for his rivals as well as for readers! Maybe it's an urban legend--I have no facts. But there's a scam for everything, so why not? Would a real reader finish a book he could give only one star?
Recipe and product reviews can be helpful--I've bought products (luggage, for example) based on reviews. Occasionally I review a book and give stars, if the website requires it. Five stars to me means one of my all-time favorite books. People give five stars too often--a five star review seems as suspect as a one-star.
For me, the best book recommendation still comes word-of-mouth from someone I know or a reviewer I follow who has preferences similar to my own.
What's your take?