Every Attribution Model Is Wrong in a Different Way

Attribution debates run for years in most companies because they are framed as a search for the correct answer. There is no correct answer. Every model takes a set of observed touchpoints and applies a rule for splitting credit. The rule is an assumption, and the assumption is usually wrong in a predictable direction.

What each model assumes

The observability problem

Data-driven models sound like they escape the assumption trap, but they only learn from the journeys they can record. Cross-device, in-app, offline, and consent-declined activity is invisible, so the model confidently allocates credit within a partial picture. Worse, each platform sees mostly its own touchpoints, which is why the sum of platform-reported conversions routinely exceeds the orders you actually shipped.

If your platforms together claim more conversions than your finance system recorded, attribution is not measuring — it is competing.

Pick one and hold it still

The practical answer is to standardise on one model for reporting, document why, and stop revisiting it quarterly. Consistency lets you see trends. Switching models resets every baseline and creates an argument about whether performance changed or the ruler did.

Calibrate against reality

Run periodic holdout or geo experiments on your largest channels and compare the incremental result to what the attribution model reported. The ratio between them becomes a correction factor you can apply, and it turns a philosophical dispute into a number that gets updated twice a year.

What to tell the board

Report blended acquisition cost and total revenue alongside channel figures. Blended numbers cannot be gamed by attribution choices, and they are what actually determines whether the marketing programme works.