Diagnosing a DAU drop

A classic prompt: "DAU fell 10% week over week. What do you do?" Work it like an incident plus a product diagnosis. Many candidates skip verification and jump to a redesign. Interviewers notice.

The same playbook is how you should talk about a real funnel drop at work. Engineers should prep one metrics diagnosis story from real work (an outage or funnel drop) using this language — authenticity beats invented consumer examples.

The big idea

1. Verify — Is the dashboard correct? Timezone, bot filter, logging change, instrumentation bugs (experimentation pitfalls mindset).
2. Scope — Platform (iOS/Android/web)? Geo? New vs returning? App version? Acquisition channel?
3. Decompose DAU — New users × retention of older cohorts. Is it acquisition collapse or retention collapse? See retention.
4. Funnel — Where in activation did conversion fall? Push notification deliverability? Login outage?
5. External — Competitor launch, seasonality, store ranking, OS permission change, holiday.
6. Hypotheses ranked — 3–5 causes with how you'd confirm each (query, survey, status page).
7. Actions — Mitigate (fix outage, revert bad release) vs investigate (week-long analysis). Say what you'd do today.

A fire alarm is the picture: first check it is not a burnt toast (instrumentation), then which floor (scope), then whether new guests never arrived or old guests left (acquisition vs retention). The analogy breaks when several causes stack — still rank them, still say what you do today.

Walking through it

DAU −10%. iOS-only, version 12.4, US. Returning users flat; new-user D1 retention cratered. Crashlytics shows a crash on the new onboarding permission screen. Action today: rollback 12.4 or kill-switch the screen; confirm DAU recovers; then write the postmortem. A candidate who jumps to "redesign the whole home feed" fails the diagnosis.

Practice saying the verification step out loud first. Then scope. Then split new vs returning before you invent a growth strategy.

Common mistakes

• Skipping verify. Logging changes look like 10% DAU drops.
• Redesigning before scoping. iOS-only + one version is a release, not a strategy problem.
• Listing every KPI. One north star, the decomposition (new vs returning) and the action for today.

Check yourself

1. DAU −10%, iOS-only, version 12.4, new-user D1 cratered, returning flat. First action today?

Rollback or kill-switch 12.4, confirm DAU recovers, then postmortem. Not a home-feed redesign.

2. Why split new vs returning before listing hypotheses?

Acquisition collapse and retention collapse are different problems with different owners and different same-day actions.

Key takeaways

• Verify data, then scope, then split acquisition vs retention, then hypotheses and a same-day action.
• A version-scoped crash is a rollback, not a strategy offsite.
• Bring one real diagnosis story from work; the language transfers.

Further learning

Lenny's definitive guide to product sense interviewsA step-by-step playbook to help you ace your PM interviewlennysnewsletter.com

Next: judgment and drills in product sense practice.