North-star and guardrail metrics for Vibes, Meta's AI-generated short video feed
Company: Meta
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
How would you measure the success of Vibes? Assume Vibes is Meta's feed of short AI-generated videos in the Meta AI app, where people can watch clips, create new ones from a prompt, remix someone else's clip, and share the results to Instagram and Facebook.
```hint Watching is not the same job as making
Viewing an AI clip and creating or remixing one are different activities with different costs; decide which of them Vibes most needs to prove before you pick a single headline number.
```
### Clarifying Questions
- Is the main aim of Vibes to give the Meta AI app a regular habit, to get more people creating video, or to supply new short videos to Instagram and Facebook?
- Is Vibes being judged as an early product that still has to prove people will create, or as a mature feed judged on viewing?
- Should views of Vibes clips that happen on Instagram or Facebook count toward Vibes' success?
- Is the compute cost of generating videos in scope, or only user outcomes?
### What a Strong Answer Covers
- The purpose of Vibes for Meta, stated and justified before any metric is chosen.
- The roles in the product (viewers, people who create from a prompt, remixers, sharers) and the loop that connects them.
- One north-star metric with a precise definition that reflects value rather than raw views or raw generations.
- Supporting metrics for viewing, creation and distribution, plus guardrails specific to AI-generated video: harmful or misleading clips, misuse of real people's likeness, cannibalization of other Meta feeds, and generation cost.
- How the metrics would be read given novelty effects, and how their weighting changes as the product matures.
### Follow-up Questions
- Watch time on Vibes doubles, but the share of viewers who ever create or remix falls. Is that good news?
- Most views of Vibes clips turn out to happen on Instagram Reels rather than in the Meta AI app. How do you credit that, and does it change your north star?
- How would you tell whether people are tiring of AI-generated video in general rather than of the particular clips the feed shows them?
Overview: A product manager metrics question about how to judge whether Vibes, Meta's feed of short AI-generated videos that people can watch, create and remix, is succeeding. It tests setting a goal, defining a north-star metric, choosing funnel metrics for viewing and creation, and adding guardrails suited to generated video.