Define Success for a New Group Feature Without Hiding Cannibalization

Context

You are the lead data scientist supporting a major social platform's Groups product. The product team has built a new feature called "Sub-Spaces" (topic-based, lightweight sub-channels within large existing groups) designed to help active members have more focused discussions without creating entirely separate groups.

Early testing shows strong feature adoption: members in the pilot groups are actively posting and reacting in these Sub-Spaces. However, leadership is concerned that this engagement is merely cannibalizing standard main-feed group posts, reducing participation from more casual members who do not enter Sub-Spaces, and masking potential network spillovers.

Your Task

Design an end-to-end analytics and experimentation framework to determine whether Sub-Spaces should roll out globally.

Specifically, structure your response around the following:

1. Metric Hierarchy & Framework: Define your primary North Star metric, supporting primary metrics, and counter/guardrail metrics. Distinguish between feature adoption (shallow engagement) and durable platform value (retention, ecosystem vitality).
2. Cannibalization Measurement: Define explicit formulas or operational metrics to detect whether Sub-Spaces activity represents net-new engagement or zero-sum shifting of attention away from the main group feed, individual direct messaging, or the home feed.
3. Experimentation & Spillover Strategy: Since groups are highly interconnected networks, standard user-level randomized controlled trials (RCTs) will suffer from SUTVA violations and social spillover. Specify your randomization unit, experiment design, and how you will account for interference.
4. Launch Criteria & Trade-off Decisions: Outline clear quantitative criteria for a launch decision. Explain how you would advise product leadership if Sub-Space engagement is up +25%, but casual member retention drops -1.5% and safety report volume increases.