In brief: An attribution model is a rule by which analytics assigns a conversion to one or more marketing contacts on the customer journey.
How I use an Attribution model in practice
I choose the model according to the decision it should support. For campaign operations, I need rapid feedback. For budgets, I am interested in incremental benefit and channel overlap. Last click is understandable, but favours the end of the journey. First click highlights discovery. A data-driven model may describe recorded journeys better, but it cannot restore contacts that measurement never saw. I therefore compare multiple views and, for larger decisions, use an experiment or control group.
What to watch out for
Attribution is not the same as causality. A channel may often be on the path to purchase without causing the purchase. Changes in consent, devices, cookies and offline contacts also hide part of the journey. A precise decimal number therefore need not mean precise truth.
Questions for decision-making
- Which budget decision should the model change?
- Which contacts and conversions does measurement not see?
- Are we rewarding a channel for customers who would have purchased anyway?
- Can we verify an important hypothesis through an experiment?