8.4
Marketing Mix Modeling
A regression on aggregate historical data that estimates each marketing input's contribution to sales.
No single originator; econometric sales-response modelling from the 1960s onward
What it does
Regresses aggregate historical data to attribute sales contribution across channels. Back in favour now that user-level tracking has degraded.
When it breaks
Correlational. Uncalibrated by experiment it moves budget systematically to the wrong channel — and whoever builds the model can largely choose the answer.
Case
P&G cut over $200 million of digital advertising spend in 2017 after media modelling and transparency audits, and increased reach by 10%. Marc Pritchard reported the removed spend was largely waste, reinvested into higher-reach media.
Adweek — P&G's $200m digital cut ↗Diagram — not yet drawn
The same sales line decomposed twice under two defensible model specifications, producing two different channel attributions — the researcher-degrees-of-freedom problem.
In the wild
Unvetted · not part of the tier assessment
What has been written about this tool in the last twelve months. Machine-retrieved and unchecked — everything above this line was checked.
Loading…