Investigate MAU Drop and Test Coupons
Company: Lyft
Role: Data Scientist
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Technical Screen
You are interviewing for a Data Scientist role at a ride-hailing company.
The company defines **MAU (Monthly Active Users)** as the number of unique riders who complete at least one ride in a calendar month. MAU has dropped by **7% month over month**.
Answer the following:
1. **Investigating the MAU drop**
- How would you diagnose the cause of the 7% decline?
- What metric decomposition, user segmentation, funnel analysis, and operational checks would you use?
2. **External vs. internal causes**
- How would you determine whether the drop was caused by **external competition** rather than internal factors such as pricing, product bugs, rider experience, or marketplace quality?
- What data, quasi-experimental methods, or market comparisons would you use?
3. **Coupon intervention for retention**
- Suppose the team concludes that the problem is mainly **retention among existing users**.
- The proposed intervention is a **coupon that expires in 7 days**.
- How would you estimate the **cost** and **business value** of this coupon program?
- How would you decide which users should be **eligible** for the coupon?
4. **Experiment design**
- How would you design an experiment to measure the causal impact of the coupon?
- What should be the primary success metric, what secondary metrics and guardrails would you track, and how would you handle the trade-off between incremental rides, retention, revenue, and coupon cost?
5. **Launch decision**
- Suppose the experiment shows a **statistically significant** improvement.
- Ride volume increases by about **10%**, but **total revenue increases by only a few percentage points** because of the discount.
- Would you launch the coupon program? Why or why not?
- What are the limitations of making a launch decision from those topline results alone?
6. **Statistical vs. practical significance**
- If the result is statistically significant but not practically significant, would you ship the feature? Explain your reasoning.
Quick Answer: This question evaluates a data scientist's competency in product analytics, causal inference, experimentation design, metric decomposition, user segmentation, retention modeling, and business-impact estimation.