PyMC-Marketing
Bayesian MMM library (Python)Open source, license $0. PyMC Labs sells consulting separately.
PyMC-Marketing is an open-source marketing analytics library maintained by PyMC Labs, the company formed around the PyMC probabilistic-programming ecosystem. Its MMM module gives you a full Bayesian model with adstock and saturation transforms, plus a customer-lifetime-value module in the same package.
Two things distinguish it. First, lift-test calibration is a documented first-class workflow: you feed geo-test results in and the model treats them as evidence, pulling channel estimates toward experimental truth. Second, it is a library, not a framework; every prior, transform, and likelihood is yours to change, which is either the point or the problem depending on your team.
Development is active and public on GitHub, and PyMC Labs offers paid consulting for teams that want the library plus expert hours.
At a glance
- Model
- Bayesian regression on PyMC; geometric and Weibull adstock, logistic and Hill-type saturation.
- Calibration
- Lift-test calibration is a documented workflow, not a feature request.
- Stack
- Python end to end. Runs where PyMC runs.
- Governance
- PyMC Labs maintains it; no ad platform owns it, which is worth something in this category.
Who it is right for
- Python-native data teams that want full control of model specification.
- Brands already running geo tests that want the model calibrated to them.
- Teams that value a maintainer with no ads business.
Verdict: The control-maximalist choice. Right when your data scientist has Bayesian depth; wrong as a first model for a team without one.
Source ledger
Every numeric claim on this page resolves to one of these primary sources.
- [1]PyMC-Marketing documentationhttps://www.pymc-marketing.io/retrieved 2026-08-02
- [2]PyMC-Marketing repository (GitHub)https://github.com/pymc-labs/pymc-marketingretrieved 2026-08-02