You are asked to judge whether a model or analysis is actually moving the business metric the team cares about. The model looks good on offline checks, but leadership wants proof that the observed lift is real and not just timing, selection bias, or a coincidental trend.
How would you investigate whether a model or analysis is actually driving the intended business outcome?