Strategy question

Swap 'People you may know' with 'Products you may buy'. As a PM, how will you make a decision to swap an existing feature with a new feature without developing the new 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

Ability to validate a risky feature-replacement decision cheaply before committing engineering resources to build it.

Clarify the ask: decide whether to make this swap without first building the new feature, meaning validation must come from other signals.
Identify what's really being tested: whether users would engage as much with product recommendations as with people recommendations in that slot.
Propose proxy validation: analyze existing engagement with product recommendations elsewhere in the app to estimate likely interest.
Run a lightweight test: a fake-door test showing mocked cards to a small user sample, measuring click-through without building full backend logic.
Weigh the risk: the existing feature is long-standing and well-performing, so removing it without validation risks real engagement loss.
Recommend a small live A/B test with a minimal viable version before any full swap, rather than deciding from proxy data alone.

Directly addresses the without-building-it constraint with a concrete method, a fake-door test, rather than ignoring the limitation.
Uses existing product-recommendation engagement elsewhere in the app as a real, available proxy signal.
Names the real risk explicitly, losing a proven feature's engagement for an unproven one.
Recommends a minimal live test as the final validation step, balancing speed and rigor.

Ignoring the without-building-it constraint and describing a full build-and-test plan anyway.
Not weighing the risk of removing an already successful existing feature.
Proposing a decision based on opinion rather than any validation method.

How would you interpret a fake-door test result showing lower engagement than expected?
What would you do if the existing feature still outperforms in the test?
How would this decision differ across different user segments?