Digital Attribution — The Great Measurement Lie
The productive tension
Attribution as precise scienceandas the biggest measurement illusion in marketing
last-click is wrong, MTA is broken, MMM is imprecise, and the truth lives in triangulation
The synthesis
Digital attribution is simultaneously the most granular measurement system marketing has ever had and the most misleading one, because granularity is not the same as accuracy. Every attribution model is wrong in some known direction; the question is not "which model is right?" but "which set of methods, used together, produces a decision-making picture of the business that is good enough to act on?" The mature answer is triangulation: MMM for long-run budget allocation, incrementality tests for causal claims, digital attribution for day-to-day in-channel optimisation, and cheap proxies like share of search for brand health between formal studies. No single lens is the truth.
Learning objectives
- →Compare the seven main attribution models (last-click, first-click, linear, time-decay, position-based, data-driven, MTA) and identify their biases
- →Explain why last-click attribution systematically under-credits brand and upper-funnel activity
- →Describe why multi-touch attribution became structurally broken after iOS 14.5 ATT (April 2021) and cookie deprecation
- →Apply a triangulation approach using MMM (Robyn, Lightweight MMM), incrementality testing, and share of search
- →Critique a vendor dashboard ROI claim using the language of causation, correlation, and counterfactuals
Members only
This lecture is part of a paid plan
The first lecture of every module is free — no account needed. The rest unlocks with a subscription. One price, all 120 lectures, both languages.