This question evaluates a data scientist's competency in experimental design, metric selection, and causal inference within the analytics and experimentation domain.

The Info Stream currently prioritizes content from a user's friends. The product team plans to introduce recommended ("unconnected") content from accounts the user does not follow for the first time. The goal is to understand whether friends' content is inherently more "social" (e.g., comments, replies, reshares) than unconnected content, and to measure whether launching unconnected content improves overall engagement without harming core health metrics.
Assumptions (for clarity):
Hints: Consider A/B tests (treatment vs. control), primary engagement metrics (CTR, reactions), guardrails (retention, time spent), and segmentation by relationship.
Login required