Attributed is not incremental

Attribution answers a question about proximity: which of our touchpoints was near the purchase. It is frequently mistaken for an answer about causation, which is a different question: would this purchase have happened without us.

The gap between them is largest exactly where spend is largest. Branded search is the clearest case. Someone who already decided to buy from you types your name, clicks the ad above the organic result, and buys. Attribution credits the ad with the entire sale. The incremental contribution is close to zero, because they would have clicked the link below it.

Retargeting has the same shape. It follows people who already demonstrated intent, so it attributes beautifully and frequently contributes little.

The only reliable correction is a holdout: withhold the activity from a random, meaningful slice and compare outcomes. It feels wasteful and it is the cheapest thing in this course, because the alternative is spending the same money indefinitely without knowing.

Run it long enough to cover your actual sales cycle, and pre-register what you expect before you look. A holdout analysed after the fact, by people who know which group is which, will produce whatever the analyst already believed.