Welcome to mmm-eval¶

A comprehensive evaluation framework for Marketing Mix Modeling (MMM) frameworks.

Quick Start¶

Get started with mmm-eval in just a few steps:

1. Install mmm-eval¶

pip install git+https://github.com/mutinex/mmm-eval.git

2. Prepare your data¶

Your data should include: - Date column - Target variable (e.g., sales, conversions) - Media spend columns - Revenue column (for ROI calculations)

Example data structure:

date,sales,revenue,tv_spend,digital_spend
2023-01-01,1000,7000,5000,2000
2023-01-02,1200,8000,5500,2200

3. Create a configuration¶

For PyMC-Marketing:

from pymc_marketing.mmm import MMM, GeometricAdstock, LogisticSaturation
from mmm_eval.configs import PyMCConfig

model = MMM(
    date_column="date",
    channel_columns=["tv_spend", "digital_spend"],
    adstock=GeometricAdstock(l_max=4),
    saturation=LogisticSaturation()
)

config = PyMCConfig.from_model_object(
    model_object=model,
    revenue_column="revenue"
)

4. Run evaluation¶

mmm-eval --input-data-path data.csv --config-path config.json --output-path ./output --framework pymc-marketing

Documentation¶

Getting Started¶

• Installation - Get up and running with mmm-eval in minutes.
• Quick Start - Learn the basics with a hands-on example.
• Configuration - Configure your MMM frameworks.

User Guide¶

• CLI Reference - Learn how to use mmm-eval effectively.
• Data Requirements - Understand data format and requirements.
• Frameworks - Supported MMM frameworks.
• Tests - Available validation tests.
• Metrics - Understanding evaluation metrics.

Examples¶

• Basic Usage - Practical examples and use cases.

Development¶

• Contributing - How to contribute to mmm-eval.
• Setup - Development environment setup.
• Testing - Testing practices and procedures.

Features¶

Multi-Framework Support¶

• PyMC-Marketing: Bayesian MMM framework using PyMC
• Google Meridian: Google's MMM framework
• Extensible: Easy to add new frameworks

Comprehensive Testing¶

• Accuracy Tests: MAPE, RMSE, R-squared metrics
• Cross-Validation: Time series cross-validation
• Refresh Stability: Model stability over time
• Performance Tests: Computational efficiency metrics

Standardized Evaluation¶

• Consistent metrics across frameworks
• Reproducible results
• Industry-standard validation approaches

Quick Navigation¶

Getting Started¶

Get up and running with mmm-eval in minutes.

User Guide¶

Learn how to use mmm-eval effectively.

Examples¶

Practical examples and use cases.

Development¶

Contribute to mmm-eval development.