Life is easy if you have an experiment. Be it a completely randomized AB trial, a near experimental design, or an instrumental variables scenario: explicit randomization allows you to model the effect of moving the treatment variable independently from the other covariates. This is the key to counterfactual modeling: since you will be acting to apply treatment, you need to know how the response changes with independent movement in this variable.
Unfortunately, life doesn’t always give you experiments. In business, it is typical that you will be asked to make a decision about future actions based upon historical data without the benefit ...