Strategy question

Product Manager comes to me (Data Scientist) and tells me that the "share" button feature is not improving engagement. Let's take it out. How would you, as a Data Scientist, evaluate that feature?

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 pushing back on a premature product decision with a rigorous, causal evaluation method rather than accepting the PM's framing.

Push back gently on the premise: ask what data showed the share button isn't improving engagement, since removal is a real cost if that read is wrong.
Define the metric correctly: engagement caused by sharing, like new sessions or content views from shared links, not just how often the button is clicked.
Design a proper test: an A/B experiment removing the button for a holdout group, comparing downstream engagement against a control that keeps it.
Check for confounds: sharing may drive engagement for content creators or new-user acquisition even if it doesn't move the sharer's own engagement.
Recommend holding the removal until the causal test completes, since correlational engagement data alone can't prove the button has no value.

Challenges the premise professionally instead of just complying, the core skill this question is testing.
Distinguishes correlation, engagement looks flat, from causation, a proper holdout experiment, as the standard for this decision.
Flags a real confound, sharing's effect on new-user acquisition, that a simple engagement metric would miss entirely.

Agreeing to remove the feature without first proposing a rigorous causal test.
Looking only at the sharer's own engagement while ignoring downstream effects on new users reached through shares.

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