Skip to content
  • overview

Overview

AMMM is a Python library for Bayesian Marketing Mix Modelling (MMM), designed for practical decision support with explicit diagnostics.

What AMMM V2 Emphasises

Section titled “What AMMM V2 Emphasises”
  • Stage-gated workflow: each run writes structured outputs from 00_run_metadata/ to 80_interpretation/.
  • Diagnostics-first execution: convergence, calibration, Pareto-k, and structural checks are produced in 50_diagnostics/.
  • Configurable gate policy: diagnostics_gating: strict | warn | off.
  • Bayesian uncertainty propagation: decomposition, response curves, and optimisation outputs are grounded in posterior inference.
  • Pragmatic runtime tooling: CLI runner, cache management, and reproducible output contracts.

Typical Workflow

Section titled “Typical Workflow”
  1. Prepare data/config and run the V2 driver (python runme.py or MMMBaseDriverV2 API).
  2. Validate diagnostics in 50_diagnostics/.
  3. Use decomposition (40_decomposition/) and response curves (60_response_curves/).
  4. Run optimisation/scenarios (70_optimisation/).
  5. Generate reporting artefacts (80_interpretation/ when enabled).
Section titled “Recommended Next Pages”