Measuring customer satisfaction with a mix of survey and behavior signals
Company: Amazon
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
A product manager is asked how they would find out whether the people who use their product are actually satisfied with it, and which specific techniques they would use to measure that satisfaction. Treat this as a measurement-design problem: say what you would collect, how you would collect it, and how you would read the results.
```hint Two kinds of evidence
What people say and what people do can disagree. For each source you propose, think about what it would fail to show you.
```
### Clarifying Questions
- What kind of product is it (a subscription service, a marketplace, a one-time purchase), and how often would a satisfied customer naturally use it: daily, weekly, or a few times a year?
- Whose satisfaction matters here: paying customers, free users, or both, and is the buyer a different person from the end user?
- Is the goal a one-time health check, an ongoing dashboard, or an early-warning system for customers who are about to leave?
- Which data already exists today: surveys, support logs, product telemetry, reviews?
### What a Strong Answer Covers
- A working definition of satisfaction, and a clear split between stated signals (surveys, interviews, reviews) and behavioral signals (retention, frequency, repeat purchase, support contact).
- Specific techniques with their mechanics, such as how a recommendation-likelihood score is computed and when a short post-interaction rating is better.
- Awareness of each signal's blind spots, for example response bias or metrics that can be read two opposite ways.
- A plan to triangulate signals, segment the results, and turn them into a decision rather than a report.
- Leading versus lagging indicators, and who owns acting on a drop.
### Follow-up Questions
- Session length went up after a redesign. How would you decide whether customers are happier or just lost?
- Your survey response rate is 4 percent and skews toward people with strong opinions. What would you change?
- Satisfaction scores are flat but cancellations are rising. What would you investigate first?
Overview: A product manager is asked how to tell whether the people who use a product are actually satisfied and which techniques measure it. The question tests the ability to combine survey metrics such as NPS and CSAT with behavioral signals like retention and support contacts, and to avoid misleading proxies.
