One way people try to tease out those differences is through methods like "Difference in Differences"[1] which try to estimate causal relationships from observational data. These methods are used in marketing analysis as well.

In my experience, it is easy to break one of the many assumptions inherent in methods like Diff-in-diff. Because of this I have little trust in these techniques in general.

[1] https://en.m.wikipedia.org/wiki/Difference_in_differences