Measurement
What the data can and cannot tell you: attribution, incrementality, causation and the counterfactual you never observe. The distinction between measurement and action also matters in workforce optimization software, where reported gains need a clear baseline. For the broader principle of defined and comparable measurement, see NIST SI units guidance.
6 articles
Measurement
What Attribution Actually Measures, and What It Does Not
An attribution report tells you which touchpoints appeared before a conversion. It does not tell you which caused it, and no model weight changes that.
Measurement
Incrementality: The Only Question Worth Asking
What would have happened without the spend. How to actually measure it, what each method costs, and why the answer is usually smaller than the report.
Measurement
Correlation, Causation and the Marketing Report
Everyone knows the phrase and the reports are built on the confusion anyway. The four specific ways it appears in marketing data, and what each requires.
Measurement
Measuring the Channels That Have No Click
Television, audio, outdoor and video get no credit in click-based reporting, so budget drifts away from them. What actually measures them, and what does not.
Measurement
Marketing Mix Modelling: What It Is Good For
It survives every privacy change and it is correlational. What it can support, and the questions that separate a good model from a plausible one.
Measurement
Geo Experiments and Holdout Tests
The practical way to measure incrementality for channels you cannot randomise per person. How to design one that works, and the three ways they fail.