How do you measure the long-term impact of a product decision made two years ago?

Model answerMeasuring retrospective impact is hard because you rarely have a clean control group. Best case: the decision was rolled out in phases, giving you a natural experiment. Worst case: it launched everywhere at once. In that case, I'd use pre/post trend analysis with external benchmarks as control — industry averages, competitor proxies. I'd be transparent about confidence levels: this is directional evidence, not causal proof.