An evaluation framework for Marketing Mix Models (MMM).
mmm-eval provides a standardized approach for evaluation and comparison of different MMM frameworks using a full suite of validation tests including holdout accuracy, in-sample accuracy, cross-validated holdout accuracy, refresh stability, perturbation, and placebo metrics.
- Multi-framework support: Evaluate PyMC-Marketing and Google Meridian models
- Standardized metrics: Consistent evaluation across different frameworks
- Easy to use: Simple CLI interface and Python API
- Extensible: Add new frameworks and tests easily
You can install mmm-eval directly from the repository:
pip install git+https://github.com/mutinex/mmm-eval.gitmmm-eval currently supports Python 3.11 and 3.12 only.
If you’re building locally with Poetry, make sure you have Poetry 2.x installed.
# Run with default settings
mmm-eval --input-data-path data.csv --framework pymc-marketing
# Run with custom configuration
mmm-eval --input-data-path data.csv --framework pymc-marketing --config-path config.json --output-path results/
# Run specific tests
mmm-eval --input-data-path data.csv --framework pymc-marketing --test-names holdout_accuracy in_sample_accuracy cross_validationThe official hosted documentation can be found here.
We welcome contributions! Please see our Contributing Guide for details.
mmm-eval was created at Mutinex by Joseph Kang, Sam McCormick, Benjamin Dobrin, and Phil Clark.
If you use mmm-eval in your research or work, please cite it — citation metadata is available in CITATION.cff.
This project is licensed under the Apache 2.0 License - see the LICENSE file for details.