Strategy question

A new product video review feature, which increases sales, can be rolled out now with 2 hour video processing turnaround or 6 months later with instant processing using Machine Learning. Would you roll it out now or later?

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

Tradeoff reasoning between speed to market and product quality, grounded in the business impact of each option.

Clarify the business context, how much sales lift the feature is expected to drive and how time-sensitive the opportunity is, e.g. a seasonal shopping peak.
Quantify the cost of waiting, six months of foregone sales lift from the feature not existing yet.
Quantify the cost of the slower version, whether a 2 hour turnaround meaningfully hurts the shopping experience or conversion versus instant processing.
Consider a hybrid option, launch the 2 hour version now to capture value while the ML-based instant version is built in parallel.
Recommend based on the tradeoff, if 2 hour turnaround does not meaningfully hurt conversion, launch now and upgrade later.
Define success metrics, video review submission rate and downstream conversion lift, measured at launch and again after the faster version ships.

Quantifies both sides of the tradeoff, foregone sales from waiting versus a degraded experience from launching early, rather than a gut call.
Proposes a hybrid path, ship now and upgrade later, avoiding a false binary between the two options presented.
Identifies what data to check first, whether 2 hour turnaround actually hurts conversion meaningfully, before committing to a recommendation.
Ties the final decision to a measurable checkpoint, comparing conversion before and after the ML upgrade ships.

Treating this as a pure binary choice without considering a phased or hybrid rollout.
Not quantifying the cost of waiting six months in terms of lost sales.
Assuming a 2 hour delay is either clearly fine or clearly bad without checking user tolerance data.

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