A rating is subjective in nature as a single person gives it to “rate” the quality of something. As if you say “I liked it” or “Well it was average” to “I didn’t like it at all”. A rating is not very differentiated as such, and also very dependent on what you like category-wise as well as how high your expectations are. Highly subjective thing, such a rating.
But a rating is also a message to others, a recommendation (or the opposite), expecting that they feel similar to you about certain categories and the expectations you have. For someone, who has the same expectations in the diverse categories, you most likely can give very good recommendations (or advice on what he/she won’t like too).
But if you understand how people feel about certain categories of consumables, what they rate about the consumable and what they don’t rate, and what expectations they have in these rated subproperties, it should be possible to calculate a distance between the peoples expectations about categories in a consumable space, and to tell them reliably things they’re missing out on still.
I deem this doable technically, but as I don’t have the time / money to build this platform (that I have sketched out already), I just want to tell you programmers out there that you should really build a platform, that can show people what they really would love. I’d use it.