I think what recsys really should be trying to target is finding the group most similar to your tastes, and then picking based on the outliers of that group. The goal being to find something within your tastes, but far enough from your normal to be interesting
Recommending based on star personalities preferences, is I think, a way of approaching this idea: why would you want to know what the average item is? You want the best one.
I'll be sourcing recommendations from the silent majority of experts as well (https://prog21.dadgum.com/143.html).
Reference: https://news.ycombinator.com/item?id=17784551.
OT: I checked your profile but couldn't find an email to contact you. It'd be cool if I can shoot an email once in a while to brainstorm.
Then you can also do SEO analytics on site for domain authority too, that tells you roughly how many people use it vs other software. Or check how old a site is, how old a video is etc