Metrics question
How would YouTube go about detecting if a video is watched in a group or not?
- Metrics
- Hard
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
Metrics and product thinking for inferring an unobserved behavior (group viewing) from indirect signals, a common ambiguous data problem.
How to approach it
- Clarify the goal: why does YouTube care if a video is watched in a group, likely for better engagement measurement or targeted content and ad relevance.
- State the challenge: YouTube cannot directly observe how many people are in a room, so this must be inferred from indirect signals.
- Propose signals: device type (smart TV or living room device usage correlates with group viewing more than mobile), time of day (evening and weekend viewing on TV), and engagement patterns like pause and rewind frequency, which may differ for group co-viewing.
- Propose a secondary signal: multiple accounts logging engagement (likes, comments) shortly after a single viewing session on a shared device.
- Build a probabilistic model combining these signals into a group-viewing likelihood score rather than a binary yes or no answer.
- Define success as the model's ability to predict group viewing sessions validated against a small opt-in survey sample.
What a strong answer includes
- Frames this correctly as an inference problem using indirect signals, not something directly observable, and says so explicitly.
- Proposes a specific, plausible signal, smart TV plus evening or weekend timing, grounded in real viewing behavior patterns.
- Recommends validating the inferred model against a real ground truth, an opt-in user survey, rather than trusting the model blindly.
Common mistakes
- Assuming group viewing can be measured directly without acknowledging it must be inferred.
- Proposing a model with no way to validate its accuracy against real ground truth.
Likely follow-up questions
- How would you validate your group-viewing model's accuracy without invasive tracking?
- How would this signal change YouTube's content recommendations or ad targeting?
More metrics questions
- How do you define success for Yelp reviews?Google · Metrics · Medium
- Utilization went down by 45% on app XYZ in Italy for the month of August. Give a reason why and draft a plan to fix it.Spotify · Metrics · Medium
- You launched a new signup flow to encourage new users to add more profile information. A/B test results indicate that the % of people that added more information increased by 8%. However, 7 day retention decreased by 2%. What do you do?Google · Metrics · Hard
- Define the metrics for YouTube search.Google · Metrics · Medium
- You walk in to your office and find that Google cloud subscription has gone down by 20%. What do you do as a product manager?Google · Metrics · Medium
- What would be the top six metrics for WhatsApp? Now pick one from the six. Tell me three things you would do to improve that metric.Google · Metrics · Medium
More questions from Google
Learn the skill behind it
Chapters of the AI PM course that teach what this question tests.
- Chapter 9: Prove it paid off: outcomes, economics, and pricing
- Chapter 2: Data fluency: SQL, logs, and reading the truth yourself
- Chapter 14: Get the job: the AI PM interview loop