How To Reframe A Business Question As A Causal Question

Translating between the business and Data Science team is critical to project success. I will explain how to do that for complex problems and even more complex research based solutions.

The business has a customer churn problem and they come to me to resolve it. Customers make an initial purchase but only a small percentage return to make a second purchase. The business asks me, “How do we get more people to make a second purchase?”

I talk a lot about reframing a business question as a Data Science question. In this post, I want to go a little deeper and reframe a business question as a causal question.

Any business question needs to be broken down to create a causal question. Why?

The example question hides many smaller questions inside. Without breaking down the initial question, I can add an assumption to my approach, a bias to my model, or a flaw to my experiment. Those hide in poorly assessed problems.