Your question is Science Problem Decomposition. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How did you break this down into a science problem? What alternative methods existed, and what was the justification for not using them?
Asked in the Round 2: ML Depth stage. Deep dive into the candidate's professional project. Reported follow-up: What were the offline and online metrics used and their tradeoffs?