Product design question
How would you improve the recommendations module 'For You' on TikTok?
- TikTok
- Product design
- Easy
Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.
Start a mock interview on this question · Mock interview from a job description
What this question tests
Tests product improvement on a core algorithmic feed: can you propose a specific, testable change to a recommendation system rather than a vague 'make it better'.
How to approach it
- Clarify what 'improve' means: could be relevance, diversity, creator fairness, or session length, since For You already performs very well on raw engagement and the interesting angle is a specific gap.
- Identify a plausible current weakness: over-optimization for short-term watch time can lead to filter bubbles or repetitive content, hurting long-term satisfaction even as short-term metrics look fine.
- Propose a concrete improvement: introduce a controlled content-diversity injection (occasionally surfacing adjacent-interest content) and measure its effect on long-term retention versus short-term watch time.
- Propose a creator-fairness angle: ensure smaller creators with genuinely engaging content get some exploration traffic instead of the feed purely reinforcing already-popular creators.
- Design the test: an A/B test comparing the diversity-injected feed against the control on session length, D30 retention, and content report rate, not just immediate watch time.
- Define success as improved long-term retention (D30) without meaningfully hurting short-term watch time, since that would indicate healthier long-run engagement.
What a strong answer includes
- Picks a specific, well-known tension in recommendation systems (short-term engagement vs. long-term satisfaction and diversity) rather than a generic 'personalize more' answer.
- Proposes a concrete, testable mechanism (diversity injection) with a clear metric trade-off to watch (watch time vs. D30 retention).
- Adds the creator-fairness angle, which is a real, discussed issue in feed design and shows broader platform thinking.
- Frames the evaluation as an A/B test with named guardrail metrics, not just an assertion that the change would help.
Common mistakes
- Vague answers like 'use better AI' or 'personalize more' with no specific mechanism.
- Ignoring the known trade-off between short-term engagement optimization and long-term user satisfaction.
- No testing plan or metric to judge whether the proposed change actually worked.
Likely follow-up questions
- How would you detect a filter bubble forming for a user?
- How would you balance creator fairness against pure engagement optimization?
- What would make you roll back the diversity injection experiment?
More product design questions
- How would you improve TikTok?TikTok · Product design · Easy
- Redesign a washer and dryer to increase sales and profit.TikTok · Product design · Hard
- How can you design TikTok, but for older people?TikTok · Product design · Medium
- How can TikTok improve communication between sellers (merchants) and customers?TikTok · Product design · Medium
- What user pain points does TikTok solve?TikTok · Product design · Easy
- How would you leverage the TikTok app to acquire customers for their shopping platform?TikTok · Product design · Hard
More questions from TikTok
Learn the skill behind it
Chapters of the AI PM course that teach what this question tests.
- Chapter 4: Discovery and strategy for AI products
- Chapter 7: AI UX and human oversight: design for a system that is wrong sometimes
- Chapter 14: Get the job: the AI PM interview loop