"How much does marketing actually earn us?" has two main paths to an answer — and each looks at the problem from a different side. Attribution bottom-up, marketing mix modeling top-down. Here's when to use which.
Attribution (bottom-up)
Attribution follows the specific customer journey and splits credit between touchpoints. It's detailed and fast, but depends on tracking individuals — which weakens in the privacy era and after the end of third-party cookies. The models are covered in conversion attribution.
Marketing mix modeling (top-down)
Marketing mix modeling (MMM) statistically estimates channel impact from aggregated data — comparing spend and results over time, without tracking individuals. It's more privacy-resilient, but less detailed and more demanding on data and time.
When to use which
- Attribution — for day-to-day campaign optimization and fast decisions.
- MMM — for strategic budget decisions across channels and the long term.
- Incrementality — experiments (holdout, geo) that verify real impact. See incrementality.
Why they complement each other
In the privacy era no method gives the whole truth alone. The trend is triangulation — combining attribution, MMM and incrementality so their weaknesses cancel out. The common denominator is data quality: the more accurate the measurement (server-side, clean first-party data), the better the input for all three.
Summary
Attribution is detailed and fast but depends on tracking individuals; MMM is more privacy-resilient but coarser. In the privacy era they complement each other best together with incrementality — and all rest on quality data that reliable server-side measurement provides. More in the complete guide.