Control group


Fortran must die
In theory we would like to know those who got our service results compared to those who did not. And we would like to randomly assign people to services. The latter is not legal and we do not have information on those who are similar to our customers, but got no services from us. So while that limits our analysis we are stuck.

So given that, how invalid is it to analyze the impact of getting a service by looking at our customers who get the service and those that did not controlling for factors like gender that impact result. All we want to know is, did a specific service or behavior or ours have an impact. Of course analyzing the relative impact of getting service 1 compared to service 2 would be awesome, but after ten years of looking for ways to do that (different issue)I don't think there is an agreed way to do that :)