Metrics question

How would you evaluate the success of Reactions in Facebook?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Tests evaluating a lightweight engagement feature for its real impact on expression and content signal, not just usage count.

State Reactions' goal: give richer emotional signal than Like alone, improving both user expression and feed ranking quality.
Reject 'total reactions given' alone as the top metric since it doesn't show whether the richer signal is actually used.
Propose the key metric: share of reactions that are non-Like (Love, Haha, Wow, Sad, Angry), since that measures whether users use the nuance offered.
Add a downstream metric: whether feed ranking models that use Reaction type see improved content relevance, measured by session engagement.
Add a guardrail: Angry/Sad reaction rate on News content, since spikes can indicate harmful or polarizing content spreading.

Goes beyond raw usage to measure whether the feature achieves its actual purpose, richer emotional signal.
Connects Reactions to a downstream system (ranking) rather than treating it as a standalone engagement feature.
Uses an illustrative number, e.g. assume 70% of reactions are still plain Likes, to argue there's room to grow non-Like usage.
Flags a real content-integrity risk (reaction-based signal on news content) that a naive metrics answer would miss.

Treating 'total reactions' as success without checking whether the richer reaction types are actually used.
Ignoring the content-integrity angle where negative reactions correlate with sensitive or harmful content.

Chapters of the AI PM course that teach what this question tests.