Metrics question

Time spent on Youtube has gone down by 20% day over day. What would you do?

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

Whether the candidate can run a structured root cause investigation before proposing a fix for a sudden engagement drop.

Confirm the drop is real: check for a logging or measurement bug first, since a clean 20 percent day-over-day cliff is unusual for a mature product.
Segment by platform, region, and content type (Shorts versus long-form) to see if the drop is broad or concentrated in one area.
Check for a recent change: an app release, a recommendation algorithm update, or a UI change to the watch page or autoplay behavior.
Check external factors: a competing major content event, an outage, or a seasonal effect explaining a temporary dip.
Once isolated, propose a fix and define a recovery metric, such as watch time per session returning to baseline within a set number of days.

Checks for a tracking bug first, since a sudden clean 20 percent drop often signals instrumentation issues rather than real behavior change.
Segments by content type specifically, since Shorts and long-form watch time can move in opposite directions and mask each other in aggregate.
Names a specific likely cause, a recommendation algorithm change, as the first hypothesis given the timing and scale of the drop.

Proposing a fix before ruling out a data or tracking issue.
Looking only at aggregate watch time without segmenting by content type or platform.

How would you distinguish a genuine content quality issue from an algorithm bug?
What would you do if the drop is real but limited to one content category?