Correlation Is Not Cause
Which one is causing which — and is it either?
The idea
Two things moving together may share a cause, run in the opposite direction, or coincide.
In the real world
Customers who use a feature retain better, possibly because committed customers explore more.
Going deeper
Two things moving together admits at least four explanations: A causes B, B causes A, something else causes both, or coincidence. Only the first supports acting on it, and it is the one people assume.
Reverse causation and common causes are the ones that catch teams out. Customers using a feature retaining better is easily explained by committed customers exploring more, or by larger accounts doing both. Pushing everyone into the feature tests neither, and will often move the metric without moving retention — which then looks like the intervention working.
Where it stops applying
Demanding proof of causation before acting is impractical for most business decisions. The realistic standard is knowing which explanation you are assuming and what would distinguish it.
Why it matters
It is the difference between a finding you can act on and one that will waste a quarter.
Try this today
For one metric relationship you believe in, name a third factor that could drive both.
Test yourself
Customers using the reporting feature renew at twice the rate. The team plans to push everyone into reporting. Name two readings other than reporting causing retention.
Show the answer
Reverse causation, where customers already committed enough to renew are the ones who explore extra features; and a common cause, such as larger or better-onboarded accounts doing both. Forcing the feature tests none of these and may move the metric without moving retention.
Learn this in the feed Answering from memory, then again days later, is what makes it stick.