MMM 3 published · 5 pillars

Marketing Measurement Operating System

MMM is not regression. It is a way to make better decisions when data is incomplete.

Core thesis

Constraints + Causality + Decisions

This module studies how marketing teams reason from messy aggregate data to incremental evidence, uncertainty-aware planning, and organizational action.

Operating model

Three questions every MMM must answer.

Constraints

What does the world prevent us from seeing cleanly?

Incomplete data, collinearity, lag, seasonality, taxonomy drift, and market shocks shape what any model can learn.

Causality

What did marketing actually change?

Incrementality, identification, experiments, calibration, priors, and triangulation decide whether fit becomes evidence.

Decisions

How should the organization act under uncertainty?

Budget allocation, scenario planning, posterior risk, and operating cadence turn measurement into a decision system.

Shared resource

One MMM Dataset for the whole learning path.

Synthetic source of truth

MMM Dataset

Synthetic DTC subscription dataset with known ground truth for adstock, saturation, contribution, and budget scenarios.

Dataset guide Download CSV View parameters

MMM editorial architecture

Five pillars for a measurement lab.

01

Core ideas

Foundations

Define MMM as a way to reason about constraints, causality, and decisions rather than a regression tutorial.

Published

Marketing Mix Modeling: Measuring Marketing Beyond Last-Click

Roadmap

Why MMM Is Not Just RegressionIdentification > FitWhy Attribution Breaks at ScaleBayesian Thinking for Marketing MeasurementThe Problem of Hidden Demand
02

Business assumptions

Modeling

Show how memory, diminishing attention, hierarchy, geography, and experimental priors enter the model.

Published

Adstock as Memory: How MMM Remembers Advertising

Roadmap

Building a Minimal Bayesian MMM in PyMCSaturation as Diminishing AttentionHierarchical MMMGeo-level ModelingPriors from Experiments
03

Evidence quality

Validation

Connect MMM to experiments, placebo checks, synthetic controls, calibration, and triangulation.

Planned

Why MMM Needs ExperimentsGeo Holdout DesignSynthetic ControlPlacebo TestsCalibrationTriangulation
04

Operating decisions

Decision Systems

Move from estimates to planning: risk-aware allocation, scenario simulation, and executive decision rituals.

Planned

Budget Allocation under UncertaintyRisk-aware PlanningScenario SimulationWhy Executives Misuse MMMMedia Planning with Posterior Distributions
05

Measurement sociology

Real World Frictions

Treat messy organizations, channel taxonomies, agency incentives, and team burnout as part of the measurement system.

Planned

Dirty TaxonomiesBroken Channel DefinitionsAgency Incentive ProblemsDCR LimitationsWhy Measurement Teams Burn OutOrganizational Misalignment

Suggested path

Start with identification, then move toward decisions.

  1. Why MMM Is Not Just Regression
  2. Identification > Fit
  3. Adstock as Memory
  4. Why MMM Needs Experiments
  5. Budget Allocation under Uncertainty

Published work

Current MMM notes.

Reference

MMM Dataset

A shared synthetic DTC subscription dataset with known ground truth for MMM lessons on adstock, saturation, validation, and budget decisions.

 6 min Foundations

Marketing Mix Modeling: Measuring Marketing Beyond Last-Click

How adstock and saturation turn aggregate spend and sales data into actionable budget decisions — with a 26-week simulated example you can reproduce.

 10 min Modeling

Adstock as Memory: How MMM Remembers Advertising

A practical MMM lesson on why this week's outcome can still contain last week's media, using the shared MMM Dataset as source-of-truth evidence.

 9 min