Topic Cluster
Attribution
Attribution is the question of what caused what. In practice, no single system answers it cleanly: MMM, multi-touch attribution, brand tracking and incrementality tests each measure a different slice of reality. This hub explains what each method sees, where they disagree, and how to reconcile them into a decision that holds up commercially.
The three attribution surfaces
MMM answers 'what did each channel contribute at the aggregate level?' Multi-touch attribution answers 'which touchpoints preceded conversion?' Incrementality testing answers 'what would have happened anyway?' They rarely agree — because they aren't measuring the same thing.
When the data disagrees
If MMM, attribution and brand tracking all agreed, the decision would already be made. The disagreement is the finding. Reading the gap between systems — rather than picking whichever number is most flattering — is what separates measurement from reporting.
Incrementality as the tiebreaker
Smart bidding and last-click attribution both over-credit demand capture. Incrementality tests — geo holdouts, PSA tests, matched-market designs — are the closest thing to a causal check, and often the only way to see how much of the reported ROAS is real.
Essays in this cluster
MMM vs Attribution: When the Data Disagrees
If all your data agreed, the decision would already be made. How to navigate conflicting signals across MMM, attribution and brand tracking.
Read essayAI Bidding & Empty Conversions: The Digital Cobra
AI bidding systems are very good at collecting credit for conversions. The question nobody asks is whether those conversions would have happened anyway.
Read essayThe Cobra Effect: When Marketing KPIs Backfire
When the metric functions but the outcome fails. How efficiency-led marketing systems optimise the wrong thing with increasing precision.
Read essayMarketing Dashboard Pitfalls: Green Hides Decline
When every number looks healthy and growth is still declining. How dashboards can create the illusion of performance while hiding what actually matters.
Read essayMMM Limitations: Why a Good Model Can Mislead
Your MMM may be technically correct and still lead you in the wrong direction. Why model accuracy and model usefulness are not the same thing.
Read essayCPP Efficiency Trap: Low Cost Per Point & Growth
Low CPP looks like efficiency. It can quietly reshape your audience strategy without anyone deciding to change it. The CPP efficiency paradox explained.
Read essayMarketing Measurement Strategy & KPI Design
How you measure is a strategic choice, not just a technical one. The metrics you choose determine what your system optimises for over time.
Read essayReference material
Deep-linked entries from the Reference Library for terms used across these essays.