[arijun]

A few thoughts:

One 5 star rating is worse than 1000 4.5 average ratings. There's a few ways to deal with this, Bayesian average is one.

I think for most aggregate ratings, thumbs up/down is all the useful signal you can get. If you're reading a review, it can be useful to see someone say "I gave this 3.5 stars instead of 4 because the last time I went there the fries were cold." However, on aggregate, that distinction becomes almost entirely lost given 1) peoples varying rating schemes (my 4 could be your 3 stars) and 2) often lovers/haters will just give 5/0 stars, drowning out any nuance. That's why Steam and Netflix switched to thumbs up/down.

The categories are wonky--under TV shows I found Netflix, a video game franchise, running shoes (!)... Maybe have user generated tags?

[arijun]

But generally: good luck! I've thought something like this was needed for a while. I've considered building similar things in the past, and always couldn't figure out how to get a critical mass of users. Here's hoping you succeed where I didn't even try.