Check if a correlation you found is actually causal, or just correlated
Two lines moving together on a chart is the start of the question, not the answer.
- Use it for
- Anyone about to act on a pattern they spotted in data.
Finding two things that move together is easy; knowing whether one actually causes the other is a different, harder question that a lot of data analysis skips past. A correlation is consistent with the causal story you want to tell, with the reverse causal direction, with both being driven by a third factor, and with pure coincidence — and acting on the wrong one wastes effort or makes things worse.
This prompt forces the alternative explanations onto the table before treating a correlation as a reason to act.
When not to use this
If you're just exploring data for patterns worth investigating further, this level of rigor is premature — flag it as interesting and move on. This is for the moment right before you act on a pattern as if it were causal.
Did this work?
You can name at least one plausible alternative explanation for the pattern you found, and know what evidence would rule it out — not just a confident causal story.
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.