Causal effect

Find out whether changing this factor actually changes the outcome, not just whether they appear together.

Your online shop's 15-percent-off coupon feels like it's driving sales, since coupon orders roll in daily. But here's the uncomfortable question: how many of those buyers would have purchased anyway at full price? If most coupon users were already committed, the coupon isn't causing sales, it's discounting revenue you already had.

Causal effect is the evidence job of finding out whether changing this factor actually changes the outcome, not just whether the two appear together. Coupon use and purchases co-occurring proves nothing about what the coupon adds. The cue is crediting or blaming a factor when you've never seen the outcome without it under similar conditions. The operation: create the comparison that isolates the factor. Turn the coupon off for two weeks and compare total sales and profit against two matched weeks with it on, or show the coupon to a random half of visitors. Decide in advance what difference would justify keeping it. What changes is that you learn the coupon's real contribution, which is sometimes negative once margin is counted. Don't run causal experiments when the intervention could do irreversible harm to something you can't restore, like testing whether your best client "really needs" the attention you give them.