Attribution credit is presented as incremental lift
A fictional team presents a dashboard showing purchases attributed to a campaign and claims every one is a purchase that would not have occurred without advertising. The supplied dashboard applies an attribution method but contains no randomized holdout, suitable counterfactual comparison or analysis of customers who would have bought anyway. A colleague suggests keeping the useful reporting result while narrowing the claim and considering an appropriate incrementality study. This written exercise creates no experiment and changes no ad delivery. It asks the learner to distinguish assignment of credit from a causal estimate, identify the missing evidence and explain why even a planned holdout needs attention to assignment, contamination, measurement and uncertainty.
1. What does attribution alone not establish?
Read the worked answer
It does not establish the number of purchases caused by the campaign compared with a credible no-ad counterfactual.
2. What additional approach can help?
Read the worked answer
A suitable controlled holdout or other defensible causal design can estimate lift, with attention to actual eligibility, assignment, contamination, sample size, measurement and uncertainty.