Metrics question

Twitter - DAU dropped from 70% to 60%. Why?

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

Root-cause investigation for a DAU-MAU stickiness ratio drop, requiring clarification of the metric and structured diagnosis.

Clarify the metric: this looks like a stickiness ratio dropping from 70 to 60 percent, meaning DAU fell, MAU grew, or both, and confirm this interpretation.
Check which side moved: if MAU grew while DAU stayed flat, the drop reflects an influx of less-engaged users, not existing users disengaging.
If DAU itself fell, segment by user cohort and platform to localize where daily engagement dropped.
Check for a recent product change: a feed algorithm update, a controversial policy change, or a technical issue affecting daily usability.
Check for external factors: a major news cycle previously driving daily visits that has since quieted, since this product's usage is tied to real-time events.
Propose next steps based on the diagnosed driver, such as investigating retention among a recent influx of new users.

Correctly interprets the stat as a stickiness ratio and considers both possible directions of change before concluding disengagement.
Names the product's specific behavioral pattern, event-driven daily spikes, as a plausible external explanation many candidates would miss.
Segments by user cohort to distinguish an engagement regression from a dilution effect from new user growth.
Proposes distinct next steps depending on which underlying cause is confirmed.

Assuming DAU necessarily fell without checking whether MAU simply grew instead.
Not considering the product's real-time, news-driven usage pattern as a possible external cause.
Jumping to a product change as the cause without segmenting or checking timing correlation first.

How would you determine whether this is a dilution effect from new user growth?
What would you do if this coincided with a major news event ending?
How would you weigh a genuine engagement problem against a metric definition issue?