Whenever you're ready, walk me through how you'd approach this.
You are evaluating a recommendation system for a content platform. The model ranks items for the home feed, and the team wants to know whether it is actually better than the current approach.
How would you evaluate the performance of a recommendation system?