A playbook for reconciling MMM and MTA when they disagree

A playbook for reconciling MMM and MTA when they disagree

The question: your media mix model (top-down, aggregate) and your multi-touch attribution (bottom-up, user-level) give different channel verdicts. Which do you believe, and how do you reconcile them?

The reconciliation steps:

— Map what each can and can't see. MTA tracks trackable digital touches but is blind to TV, OOH, and brand effects. MMM sees all spend but at coarse weekly granularity. Disagreement often just reflects different fields of view.
— Identify channels both cover (paid search, paid social). These are where disagreement is genuinely informative, not just scope difference.
— Hold an incrementality test as the tiebreaker. Where MMM and MTA conflict, the experiment is the arbiter — neither model is causal on its own.
— Use MMM for top-level budget splits and MTA for in-channel optimization. They operate at different altitudes; forcing one number flattens that.

The nuance: the modern frontier (Google's Meridian, triangulation frameworks) doesn't pick a winner — it calibrates MMM with experiments and cross-checks MTA against MMM. Each is a biased estimator; the consensus across three independent methods is more trustworthy than any one.

Bottom line for practitioners: don't force MMM and MTA into agreement — triangulate. Use experiments to anchor both, assign each model the altitude it's good at, and treat persistent disagreement as a question for a test, not a tie to break by gut.