You launched a new growth feature and adoption rose afterward, but rollout timing and user selection may make the raw lift misleading. You need to separate true incremental impact from underlying trends and selection effects.
How would you apply causal inference to understand whether a new feature actually drove growth?