An amazing piece of pie. I have a question though about the implementation of the recommender. How or what are the ways to roll out the recommendation engine.

1- Do we generate the recommendations live when the user open the application?

2- Do we generate a list of recommendations for each user in our database and later feed the recommendations when the user open the application?

If we have millions of users recommending but only 30% are active then we have wasted resources perhaps, on the other side how we can make sure our recommendations live can maintain short time.

I would appreciate if you started conversation in this aspect because it completes the loop on building a recommendation engine.

Thank you very much.