Marketing Mix Modeling

Marketing Mix Modeling Overview

Marketing Mix Modeling (MMM) is a statistical analysis technique that quantifies the impact of various marketing inputs on sales and other business outcomes. It uses historical data to understand how different marketing channels and external factors contribute to performance.

What is Marketing Mix Modeling?

MMM is an econometric approach that uses regression analysis to determine the relationship between marketing investments and business outcomes. Unlike digital attribution, MMM can measure both online and offline channels, making it essential for holistic marketing measurement.

Core Components of MMM
  • Dependent Variable: The outcome you're measuring (sales, revenue, conversions)
  • Marketing Variables: Spend, impressions, or GRPs by channel
  • Control Variables: Seasonality, pricing, promotions, economic factors
  • Adstock/Carryover: The lingering effect of advertising over time
  • Saturation: Diminishing returns as spend increases
Key Benefits of MMM
Privacy-Safe

Uses aggregated data, not individual user tracking. Unaffected by cookie deprecation or privacy regulations.

Cross-Channel Measurement

Measures all channels including TV, radio, print, outdoor, and digital in a unified framework.

Budget Optimisation

Provides data-driven recommendations for optimal budget allocation across channels.

External Factor Control

Accounts for seasonality, competition, economic conditions, and other external influences.

Long-Term Effects

Captures brand-building effects and advertising carryover that attribution misses.

Scenario Planning

Enables "what-if" analysis to predict outcomes of different budget scenarios.

MMM Data Requirements
Data TypeExamplesMinimum History
Outcome DataSales, revenue, conversions, leads2-3 years weekly
Marketing SpendTV, digital, print, radio, OOH by week2-3 years weekly
Media MetricsGRPs, impressions, reach, clicks2-3 years weekly
Pricing DataAverage selling price, promotions, discounts2-3 years weekly
DistributionStore count, online availability, inventory2-3 years weekly
External FactorsWeather, holidays, economic indicators, competitor activity2-3 years weekly
Key MMM Outputs
Contribution Analysis

Breakdown of what percentage of sales each marketing channel and factor contributed. Shows the relative importance of each driver.

ROI by Channel

Return on investment for each marketing channel, enabling comparison of efficiency across the mix.

Response Curves

Visualize how sales respond to different spending levels, showing saturation points and optimal investment ranges.

Optimal Budget Allocation

Recommendations for how to redistribute budget across channels to maximise overall ROI.

Scenario Simulations

Predict outcomes for different budget scenarios, helping with planning and forecasting.

Open-Source MMM Tools
ToolDeveloperKey Features
RobynMeta (Facebook)Automated hyperparameter tuning, budget optimiser, R-based
LightweightMMMGoogleBayesian approach, JAX-based, geo-level modeling
MeridianGoogleNext-gen MMM, reach/frequency support, Python-based
PyMC-MarketingPyMC LabsFlexible Bayesian modeling, CLV integration
MMM Limitations
!Data Requirements: Needs 2-3 years of historical data to build reliable models
!Granularity: Typically weekly or monthly, cannot provide real-time optimisation
!New Channels: Difficult to measure channels without sufficient historical data
!Correlation vs Causation: Statistical relationships don't always imply causation
!Resource Intensive: Requires statistical expertise and significant data preparation