Work out whether a new feature changed behaviour or just moved it around

You are an analyst checking whether a shipped feature created new activity or cannibalised existing activity. Below is usage data before and after launch, plus what the feature does. Produce:

1. WHAT MOVED - table: Metric | Before | After | Change | Where the change most plausibly came from.
2. SUBSTITUTION CHECK - for each metric that fell, whether the drop is consistent with users switching to the new feature (same users, sametime  window) or with something unrelated.
3. NET EFFECT - your single best estimate of new activity created, with the arithmetic shown.
4. WHO IT ACTUALLY HELPED - segments where net effect is clearly positive, and where it is flat or negative.
5. WHAT WOULD SETTLE IT - the one query, cohort cut, or holdout you would run next, and what result would change your conclusion.

Rules: no preamble. Do not attribute a change to the feature without saying what else could explain it. Do not invent numbers - if a needed metric is missing, list it under MISSING. State the comparison window you used.

USAGE DATA BEFORE AND AFTER LAUNCH: {{paste}}
WHAT THE FEATURE DOES AND WHICH EXISTING WORKFLOW IT COMPETES WITH: {{paste}}