📌 Interview Question
“Instagram noticed a 5% drop in DAU (Daily Active Users) last week. As a data scientist, how would you investigate the cause?”
This type of question is common in interviews at product-focused tech companies like Meta, TikTok, and Pinterest. It tests your ability to analyze ambiguous business problems, form hypotheses, and communicate a data-driven investigation strategy.
✅ Step-by-Step Answer
1. Clarify the Objective
Start by clarifying what “DAU dropped 5%” really means.
“Can I confirm the time window, platform (mobile vs. web), and whether this is seasonally adjusted? Is the drop based on global data or a specific region?”
Questions to ask:
- Is the drop real or a data anomaly (e.g., instrumentation issue)?
- Is it across the board or localized to a segment (e.g., Android users, new users)?
2. Define Key Metrics to Monitor
- DAU by segment: country, platform, app version, acquisition source
- Login attempts or session start counts
- Feature-specific usage (e.g., Stories, Reels)
- Crash rate, error logs, and API failures
“I'd look at supporting metrics like daily session count, time spent per user, and app load success rates to triangulate the issue.”
3. Form and Prioritize Hypotheses
Use a MECE (Mutually Exclusive, Collectively Exhaustive) approach to brainstorm hypotheses:
A. Product-related Hypotheses
- A recent feature rollout introduced friction or bugs
- UI changes increased bounce or confusion
- App store reviews or social media sentiment turned negative
B. Technical Hypotheses
- Backend issues or app crashes after update
- Login server failures or increased 401/500 errors
- App not opening due to deployment problems
C. External Factors
- Natural disasters, political events, or holidays affecting usage
- Competitive product launch
- App removed or downranked in stores in some countries
D. Data Quality Issues
- DAU tracking event not firing
- Metrics pipeline delayed or broken
“I’d prioritize hypotheses by likelihood and user impact. For example, if there was a version release right before the drop, I’d first analyze that cohort.”
4. Analyze by Segments
Break DAU into key segments:
- Platform: iOS vs. Android vs. Web
- Region: Global vs. specific markets
- User type: New vs. returning, power users vs. casual users
- Acquisition channel: Organic, paid, referrals
“If the drop is isolated to Android users in Southeast Asia, I’d dig into app version releases, store rankings, and crash logs in that region.”
5. Investigate Technical and Feature Logs
Ask engineering:
- Any recent deployments or rollbacks?
- Any increased crash rates or downtime alerts?
Ask product:
- Any A/B tests or new feature gates?
- Any recent changes to onboarding, ads, or homepage layout?
“I’d correlate the timing of metric changes with any version deployments, feature rollouts, or experiment activations.”
6. Recommend Next Steps
Once a root cause is suspected or identified, propose data-informed actions:
- Fix a bug in the latest version and release a patch
- Revert a confusing UI change or feature flag
- Refine onboarding if new users are bouncing
- Launch a re-engagement campaign if necessary
Also recommend:
- Improving anomaly detection (e.g., time-series monitoring)
- Building dashboards to break down DAU by risk segments
- Postmortem documentation if a confirmed failure occurred
💡 Sample Answer Summary (1-Min Version)
“First, I’d confirm whether the DAU drop is a real behavioral change or a data issue. I’d segment DAU by geography, platform, and user type to isolate where the drop occurred. If it’s localized, I’d investigate recent code deployments, feature releases, and app version changes. I’d also work with engineering to check for crashes or downtime, and with product to review A/B tests. Based on the findings, I’d recommend specific fixes or recovery actions and suggest better monitoring tools going forward.”
🧠 What Interviewers Look For
| Skill | What They Want |
|---|---|
| Structure | Clear investigative process with prioritization |
| Product Intuition | Realistic hypotheses grounded in user experience |
| Analytical Thinking | Segmentation, pattern recognition |
| Communication | Explaining your reasoning to technical and non-technical audiences |
| Action Orientation | Data-backed decisions and next steps |
✅ Takeaways
Business case interviews simulate what you’ll actually do on the job. Success means going beyond just listing metrics—you need to diagnose problems, reason through ambiguity, and connect metrics to user behavior and business outcomes.