Strategy question
YouTube has 1 million users and you're supposed to build a video recommendation engine. How would you approach this task?
- Strategy
- 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
Strategic and technical product thinking for a recommendation system at an early growth stage, checking whether you can sequence a realistic build given limited data.
How to approach it
- Clarify the constraint: 1 million users is a meaningful but still limited dataset, so early recommendations will likely have significant cold-start and sparsity problems.
- Define the goal: increase watch time and session frequency by surfacing relevant content, while being honest about data limitations at this scale.
- Start simple: use collaborative filtering on available watch history where possible, combined with content-based filtering using video metadata like genre and creator for users with sparse history.
- Prioritize solving cold start for new users first, since with only 1 million users, a large share will have thin interaction histories.
- Propose an iterative rollout: launch a simple 'more like this' content-based model first, then layer in collaborative signals as interaction data accumulates.
- Define success as watch time lift over a non-personalized baseline like trending videos, validated through an A/B test before wider rollout.
What a strong answer includes
- Correctly identifies data sparsity as the central constraint at this scale, rather than jumping to a sophisticated model that needs far more data.
- Sequences a realistic build, content-based first then collaborative filtering layered in later, appropriate for the data maturity stage.
- Ties success to a lift over a simple baseline, trending videos, a defensible and testable comparison point.
Common mistakes
- Proposing a complex deep learning recommendation system without addressing whether 1 million users provides enough data to train it well.
- Ignoring the cold-start problem, which will affect a large share of users at this relatively early stage.
Likely follow-up questions
- How would you validate the model is actually better than just showing trending content?
- How would your approach change once the platform reaches 100 million users?
More strategy questions
- Google Keep is a free product to save, share notes etc. How would you make it a subscription product & monetize it?Google · Strategy · Hard
- How would you launch (roll out) Amazon Go?Amazon · Strategy · Hard
- With an unlimited network bandwidth what would you build?Google · Strategy · Hard
- Strategize and implement omni-channel initiatives to improve customer LTV for Walmart.Shopify · Strategy · Hard
- How would you decide between showing more ads on the Facebook Newsfeed vs showing a "People you may know" recommendation widget?Meta · Strategy · Hard
- Should Uber Eats be a different app from Uber Rides?Lyft · Strategy · Hard
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 9: Prove it paid off: outcomes, economics, and pricing
- Chapter 14: Get the job: the AI PM interview loop