Attribution is collapsing. Marketing mix modeling is quietly back.
Why in-house MMM rebuilds are the logical response to attribution loss, and why the agencies that owned the model for thirty years are not getting the call.
Marketing mix modeling is coming back because the alternative stopped working, and the rebuilds we hear described are in-house: a data scientist and a Bayesian library, not the agency that historically owned the model. This is our reading of the shift, not a survey finding.
The reason is uncomfortable for the holding companies. Once a brand has its own first-party sales data, an MMM is closer to a six-week engineering project than a six-month consulting engagement. The hard part used to be the data plumbing. The plumbing is now standard.
What we're seeing in the rebuilds: weekly cadence (not quarterly), media + non-media variables in the same model (price, distribution, weather), and a UI that lets the CFO ask 'what if' without filing a ticket.
The implication for marketers is the same one we keep writing: the role is shifting from 'I commissioned the model' to 'I operate the model.' That role is closer to the CFO than to the agency.
Last-click died the moment iOS 14 shipped. We've been pretending otherwise for four years. The pretending is over.
- 01MMM is returning as an in-house build, not an agency engagement.
- 02Plumbing is standard now; the model is a six-week project.
- 03Weekly cadence beats quarterly. Non-media variables matter.
- 04The operator-of-the-model role sits next to finance, not creative.
- Published
- May 28, 2026
What changed. What it means. What to do next. Two briefings each week.
Two briefings a week. Unsubscribe anytime.
- Issue 13 · Intelligence · Jun 18, 2026Your AI search dashboard is lying to you (and the fix takes one spreadsheet column)
- Issue 11 · Intelligence · Jun 11, 2026The Generative Squeeze: why CMOs are quietly cutting agency retainers in Q3
- Issue 06 · Intelligence · May 24, 2026Five tools quietly replacing your $400k MarTech stack, and what they cost