Metrics question

What goals would you set for the Reels recommendation engine?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Ability to set goals for a recommendation engine that balance short-term engagement against long-term content ecosystem health.

Identify the two sides the engine must serve: viewers wanting relevant, enjoyable content, and creators needing fair discovery to keep producing Reels.
Propose a primary viewer-side goal: session-level watch time or completion rate per Reel served, reflecting genuine relevance, not just autoplay continuation.
Propose a creator-side goal: the share of impressions going to mid-tier and new creators, not just already-viral content, to keep the content ecosystem healthy.
Add a quality guardrail: rate of user-reported low-quality or repetitive content, since a pure watch-time optimization can degrade content diversity over time.
Explain the tension explicitly: optimizing purely for watch time tends to concentrate attention on a small set of viral creators, so the creator-fairness goal exists specifically to counter that.

Names goals for both sides of the marketplace, viewers and creators, not just a single engagement number.
Explicitly states the tension between pure engagement optimization and content ecosystem health, a sophisticated point on recommendation systems.
Adds a concrete quality guardrail rather than only chasing watch time.

Naming only viewer-side engagement metrics, ignoring the creator-fairness and content-diversity side of a recommendation engine's health.
Not acknowledging the trade-off between engagement and diversity, a well-known failure mode of recommendation systems.

How would you detect if the algorithm is over-favoring a small set of viral creators?
How would you balance new-creator discovery against showing viewers content they're most likely to enjoy?