InterviewCompany
Product analytics, experimentation, ranking, and high-scale system questions.
Define primary, guardrail, and long-term metrics before launch.
A good answer separates engagement from quality, includes retention and negative feedback, and explains how to avoid optimizing for shallow clicks.
Randomize users, define metrics, choose sample size, and monitor guardrails.
Mention network effects, novelty effects, experiment duration, statistical power, and interpreting practical versus statistical significance.
Generate candidates, score them, apply policy filters, and diversify results.
A feed system uses retrieval, ranking features, model scoring, freshness, creator diversity, integrity checks, and online feedback loops.
Group events by week and count distinct users who meet the active definition.
Clarify the active event, timezone, deduplication, bot filtering, and whether users can be counted once per week.