The Clickbait Trap

Senior ML Engineer interview at Meta, and the interviewer asks:

“You built a recommendation engine for an e-commerce site. Precision, recall, and CTR are all hitting all-time highs, but the actual user sign-up rate hasn’t budged. What fundamental assumption in your proxy metrics just failed in production, and how do you course-correct?”

Don’t say: “The model is overfitting to historical data. We need to introduce exploration-exploitation strategies like UCB or epsilon-greedy to diversify the recommendations, or adjust the learning rate and wait for better data.”