Acting on a Polarized 1-5 Satisfaction Survey for a Video Conferencing App
Company: Google
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
Set the clock to a time before the COVID-19 pandemic. You are the product manager for a video conferencing app, and your team has just closed a customer satisfaction survey. It drew 400,000 responses on a 1-to-5 scale, and the answers clump at two points: 40% of respondents chose 2 and another 40% chose 4. How would you put these results to use?
```hint Shape before score
Before planning any fix, ask what the shape of this distribution tells you that one summary number would hide, and what you could attach to each response to explain why people landed where they did.
```
### Clarifying Questions
- What exactly did the survey ask about (the product overall, a recent call, or one feature), and is 5 the most satisfied score?
- How are the remaining 20% of responses spread across scores 1, 3 and 5?
- Who was invited and how: a random sample of all users or a prompt after a call, and what share of those invited responded?
- Can each response be linked to the respondent's account, plan, device and call-quality data, and did the survey collect written comments?
### What a Strong Answer Covers
- Recognizing that the scores are polarized and what an overall average would hide
- Testing whether the split is real or an artifact of survey wording, sampling and response bias
- A plan to explain the split by segmenting respondents and joining usage and call-quality data, followed by qualitative research
- A reasoned choice of which group to act on first, tied to retention or revenue and to how businesses used video meetings before the pandemic
- How progress will be tracked: the score distribution by segment over time and its link to behavior such as churn
### Follow-up Questions
- Your segment cuts show no difference between the people who gave a 2 and those who gave a 4. What do you look at next?
- Sales wants to advertise the average score. How do you respond?
- How would you redesign next quarter's survey so its results are easier to act on?
- Suppose the same split appeared instead in public app-store ratings for a messaging and video-calling app just after its launch: 40% gave 1 star, 40% gave 5 stars, the other 20% gave 3 stars, and the average is exactly 3. What changes in how you investigate and what you do?
- Now take the store-rating version as a stand-alone question about a named product: you are the product manager for Google Hangouts, Google's messaging and video-calling app, and its Google Play Store listing shows an average of 3 stars across 100,000 reviews, with 40% of reviewers giving 5 stars and 40% giving 1 star. What would you do?
Overview: A product analytics question about a pre-pandemic video conferencing app whose satisfaction survey of 400,000 users clustered at two scores, with 40% choosing 2 and 40% choosing 4 on a 1-5 scale. It tests reading a polarized distribution, checking for survey bias, segmenting respondents, and turning findings into prioritized action.