Diagnose and Respond to a 5% Decline in Facebook Like Rate
Company: Meta
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
You are the product manager who owns the Like feature on Facebook. Your dashboard shows that Like acceptance has dropped by 5%. Explain what you would do: how you would work out what caused the decline and how you would respond once you know.
```hint Trust the number last
Before you look for a change in user behavior, confirm the drop is real and pin down exactly which numerator and which denominator moved. Then split the metric so every hypothesis points to one piece of it.
```
### Clarifying Questions
- How is Like acceptance defined: Likes per post seen in the feed, the share of daily users who Like anything, or something else? Does it include the other reactions?
- Did the 5% fall happen suddenly or gradually, and against which baseline: last week, last month, or the same period a year ago?
- Is the drop global, or limited to particular countries, platforms or app versions?
- Have related actions such as comments, shares, sessions and feed scrolling moved too?
### What a Strong Answer Covers
- A precise metric definition and a judgment of whether 5% is outside normal fluctuation for this metric.
- Ruling out data and instrumentation causes (logging, pipeline delays, definition changes, removal of fake accounts) before behavioral ones.
- Decomposing the metric into its drivers and segmenting by geography, platform, app version, user type, content type and surface, with attention to mix shifts.
- Internal causes (releases, experiments, ranking changes, bugs) weighed against external ones (seasonality, events, competitors, news and public sentiment), including substitution toward other reactions.
- A way to confirm the leading hypothesis, the action each finding would trigger, and how findings are communicated.
### Follow-up Questions
- Likes fell 5% but total reactions are flat. Is that a problem, and what would you do?
- The drop began the day a new feed ranking model reached all users. How would you decide whether to roll it back?
- Every internal check comes back clean and the drop is confined to one country. What would you investigate next?
Overview: A metric-diagnosis case in which the product manager who owns the Facebook Like feature sees Like acceptance fall by 5%. It tests defining the metric, ruling out data problems, decomposing and segmenting the drop, separating internal from external causes, and choosing a response.