Facebook News Feed video auto-play: test design, primary metric and guardrails
Company: Meta
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
Videos in the Facebook News Feed can either start playing on their own as they scroll into view or stay still until the person taps play. You are the product manager who has to decide whether feed videos should play automatically.
How would you reach that decision, and which metrics would tell you whether the choice succeeded?
```hint A started video is not a watched one
Automatic playback changes what counts as a view. Work out which signals show that someone chose to watch, and what playback might cost people who only wanted to scroll past.
```
### Clarifying Questions
- Is this a new launch, or a review of a default that already exists?
- Would videos play muted with sound on tap, or with sound? Does the decision also cover video ads, or only posts from friends, Pages and creators?
- Must there be one default for everyone, or can it differ by person, by connection type (Wi-Fi or mobile data) or by country?
- Which outcome matters most: time spent watching video, interaction between people, or video ad revenue?
### What a Strong Answer Covers
- The goal of the feature and everyone it affects: viewers, people who post non-video content, video creators and advertisers.
- A decision process built on a controlled experiment, with a sensible randomization unit and enough time for novelty to wear off.
- A primary metric that captures viewing people chose rather than views that playback creates mechanically, plus supporting metrics along the viewing funnel.
- Guardrails covering the rest of the feed, annoyance and negative feedback, mobile data and battery, and app performance.
- A decision rule set in advance, including partial outcomes such as automatic playback only for some users or contexts.
### Follow-up Questions
- Video watch time rises, but comments on friends' posts fall. Do you ship?
- How would you decide between muted playback and playback with sound?
- Suppose advertisers report much higher video ad view counts under automatic playback. How do you keep ad measurement honest?
Overview: A metrics and experimentation question about whether videos in the Facebook News Feed should play automatically. It tests framing the goal for viewers, creators and advertisers, designing a fair A/B test, picking a primary metric that captures intentional viewing, and setting guardrails and a decision rule.