Your question is Causal Inference Without Clean Experiments. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
You are evaluating a product change on a feed or notification surface, but rollout was not randomized and adoption was correlated with user behavior. You still need to estimate whether the change caused a shift in outcomes rather than just reflecting selection or timing effects.
How would you think about causal inference in a product setting when you can’t run a clean experiment?