Metrics question

There is an 8% drop in Google Search. 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

Tests root-cause diagnosis under a metrics-drop scenario, moving from data segmentation to hypothesis to action.

Confirm the metric and scope: is the 8% drop in query volume, click-through, or revenue, and is it global or a specific segment.
Rule out measurement issues first: check for a logging bug, tracking pixel change, or a recent release before assuming a real user drop.
Segment the drop by platform, geography, and query category to isolate whether it's broad or localized to one cause.
Check external factors: a competitor launch, a seasonal pattern, or a regulatory change in a specific market.
Once isolated, form a hypothesis (e.g. a recent ranking algorithm change hurt a query category) and validate with a quick rollback test or holdout comparison.

Follows a clear funnel: rule out instrumentation bugs before assuming a genuine behavioral drop, which is the standard first move on this type of question.
Uses segmentation (platform, geography, query type) to narrow the search space instead of guessing at a single cause.
Considers both internal causes (a bad deploy) and external causes (competitor, seasonality), showing balanced hypothesis generation.
Proposes a concrete validation step, like a rollback or holdout test, rather than stopping at a hypothesis.

Jumping to a root cause without first checking for a data or tracking bug.
Treating the drop as uniform without segmenting to see where it's actually concentrated.

What would you do in the first hour versus the first week of noticing this drop?
How would you tell a genuine user behavior shift apart from a measurement artifact?