Pymc Marketing Mmm Clv

This is PyMC's full-featured Bayesian marketing analytics suite, and it's legitimately production-ready. You get proper MMM with adstock transforms (geometric, delayed, Weibull), saturation curves, and time-varying effects. The budget optimizer uses your fitted model to reallocate spend across channels, and you can calibrate against actual lift tests. CLV models cover the standard BTYD approaches (BetaGeo, Pareto/NBD, Gamma-Gamma). It supports NumPyro, BlackJax, and Nutpie samplers if PyMC's default NUTS is too slow. The Docker setup and extensive plotting methods suggest this was built by people who've actually shipped marketing models. If you're doing attribution or LTV work and want proper uncertainty quantification instead of point estimates, this beats rolling your own Stan code.

Install to Claude Code

npx -y skills add aradotso/marketing-skills --skill pymc-marketing-mmm-clv --agent claude-code