The recommender system has just been released yesterday. I am sure it has a significant cold-start problem! Wondering how many of its users can actually receive recommendations at all.

I want to know the precision of the recommender system at the point (length of recommendation list) where precision equals recall.

I am not in favor of computing this metric on a global level, but since it is often done in literature let's compute the inverse of the overall item popularity in my recommendation lists.

Wow, this is a fancy new music recommender! I need to know each and every viking metal song it can provide.

I need relevant recommendations be placed on top of the recommendation list.

I want to know how users experience the usability of the system.

Look at this fancy new movie recommender! What is the entropy over genres of the recommendation list?

I am worried about the producer fairness of the new system. Therefore, I am really curious about the fraction of items that never appear in any recommendation list.