Metrics question

Diagnose a 15% drop in Instagram engagement.

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 structured metrics diagnosis: narrowing a broad engagement drop to a root cause using segmentation before proposing any fix.

Clarify what 'engagement' means here: DAU, time spent, posts shared, or likes and comments, since the diagnosis path differs by metric.
Check for measurement issues first: a logging change, a new experiment, or a metric definition change that could cause a false drop.
Segment by platform, geography, and user cohort (new vs returning) to see if the drop is broad-based or concentrated.
Segment by content type and feature (Stories, Reels, feed) to see whether the drop is specific to one surface.
Check for external factors: a competitor launch, a platform policy change (such as an iOS privacy update), or seasonality.
Once isolated, form a hypothesis and propose a targeted next step, such as an A/B test or a rollback if a recent change is implicated.

Rules out a measurement or logging artifact before assuming a real behavioral drop, which is a common false alarm in metrics questions.
Segments systematically (platform, geography, cohort, surface) rather than guessing at a single cause immediately.
Considers an external cause, such as iOS App Tracking Transparency's effect on tracked engagement, showing awareness of real industry events.
Only proposes a fix after isolating where the drop concentrates, keeping diagnosis and solution clearly separate steps.

Jumping to a fix or explanation before checking whether the drop is even real (a logging or measurement issue).
Treating a 15 percent drop as uniform without segmenting to find where it actually concentrates.

What would you check first if the drop were concentrated in one country only?
How would you distinguish a real user behavior shift from a measurement artifact?