Metrics question
How would you prevent bad ads (e.g., those violating privacy) from appearing on Meta?
- Meta
- 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
Tests trust and safety metrics design for a specific, high-risk ad category: can you name detection and measurement approaches for privacy-violating ads specifically, not generic ad-quality metrics.
How to approach it
- Clarify what privacy-violating ads look like: ads using sensitive inferred categories like health conditions for targeting without consent, or ads that feel surveillance-like to users.
- Design detection at the targeting layer: flag campaigns targeting sensitive inferred categories, using automated classifiers on targeting parameters before an ad goes live.
- Design detection at the creative layer: use ML classifiers on ad copy to catch ads referencing sensitive attributes directly, similar to other policy-violating content detection.
- Use user signals as a second layer: track 'feels intrusive or creepy' negative feedback specifically, since users report this even when automation misses it.
- Add human review escalation: flagged ads above a threshold go to human reviewers, given the nuance in distinguishing relevant targeting from a violation.
- Define success as prevalence of privacy-violating impressions, time-to-detection, and reduction in privacy-specific complaints.
What a strong answer includes
- Separates detection at the targeting-parameter layer from the creative-content layer, since violations can occur in either.
- Names a specific negative-feedback category, 'feels intrusive or creepy', as a distinct signal not captured by generic ad-quality feedback.
- Proposes pre-flight blocking at the targeting stage as the strongest prevention layer, not just post-publish detection.
- Uses prevalence and time-to-removal as metrics, the standard way trust and safety teams measure this.
Common mistakes
- Treating this like a generic ad-quality problem (CTR, relevance) without addressing the privacy-specific targeting and creative dimensions.
- No pre-flight prevention mechanism, relying only on post-publish detection which lets violating ads reach users before removal.
- No metric specific to privacy complaints, conflating this with general ad-quality or CTR metrics.
Likely follow-up questions
- How would you distinguish legitimately relevant targeting from privacy-violating targeting in a borderline case?
- How would you handle an advertiser who repeatedly tries to circumvent the sensitive-category restrictions?
- How would you measure the false-positive rate on ads blocked for privacy violations?
More metrics questions
- How would you measure the success of Facebook Likes?Meta · Metrics · Medium
- How would you measure improvements made to Facebook messenger?Meta · Metrics · Easy
- Imagine you are the PM in charge of Reactions on Facebook - the new way to interact with posts by using “love”, “haha”, “wow”, “sad”, and “angry” reactions. What would success look like in terms of number of non-like reactions per post at launch and how do you come up with this? Would this number differ by reaction? Why or why not?Meta · Metrics · Hard
- 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 are the PM of Instagram app. The MAU has been constant but DAU has declined. What will you do?Meta · Metrics · Medium
More questions from Meta
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