Myth: media mix modeling is the old, imprecise method that MTA replaced

Myth: media mix modeling is the old, imprecise method that MTA replaced

The question: with granular user-level tracking, did media mix modeling (MMM — top-down regression of sales on aggregate spend) become an obsolete relic next to multi-touch attribution?

What actually happened: the trajectory reversed. As third-party cookies, ATT (Apple's App Tracking Transparency opt-in), and privacy regulation eroded user-level paths, the granular data MTA depends on became sparse and biased. MMM — which never needed individual identifiers — quietly returned to the center. Google open-sourced Meridian, Meta released Robyn, both MMM frameworks, precisely because aggregate modeling survives signal loss that breaks deterministic attribution.

The nuance: MMM and MTA were never measuring the same thing. MMM estimates aggregate, channel-level response curves including saturation and adstock (the decayed carryover effect of advertising over time) and can capture offline and brand effects MTA never saw. MTA gives granular path-level credit but is blind to anything untracked and assumes the conversion was driven by observed clicks.

The modern consensus from measurement literature is triangulation: MMM for strategic budget allocation, experiments to calibrate the model's causal claims, and attribution for in-flight tactical decisions — each checking the others.

Bottom line for practitioners: MMM is not the past. If your MTA is degrading under cookie loss, a calibrated MMM plus a few geo-experiments will likely give you a more honest picture of channel contribution than any user-level model can now deliver.