DeepCausalMMM

What it does

Analysis

It includes response curves for saturation analysis, constrained budget optimization, and DMA-level contribution calculations.checked 4 Oct 2026

Causal structure

It learns relationships between marketing channels with DAGs, using an upper-triangular mask by default and offering opt-in NOTEARS learning.checked 4 Oct 2026

Deployment limit

The installation page says a Docker image will be available soon.checked 4 Oct 2026

Geographic analysis

The package supports multi-region modeling and automatic seasonal decomposition per region.checked 4 Oct 2026

GPU support

The package automatically detects and uses CUDA when available; the documentation recommends a GPU for large models.checked 4 Oct 2026

Input data

The package expects NumPy arrays shaped by region, week, and channel for media and controls, plus a region-by-week target array.checked 4 Oct 2026

Installation

It can be installed from PyPI with pip or from the project’s GitHub repository.checked 4 Oct 2026

Intended users

The documentation identifies marketing mix modeling, attribution analysis, budget optimization, causal discovery, and multi-touch attribution as use cases.checked 4 Oct 2026

License

The project’s GitHub README states that it is released under the MIT License.checked 4 Oct 2026

Product

DeepCausalMMM is a Python package for marketing mix modeling that combines deep learning and causal inference to estimate marketing channel impacts on business KPIs.checked 4 Oct 2026

Requirements

The documented runtime requirements include Python 3.9+, PyTorch 2.0+, and NumPy 1.21 or later but below 2.0.checked 4 Oct 2026

Support

The project README directs users to GitHub issues for bug reports and feature requests.checked 4 Oct 2026

Temporal modeling

Its GRU-based temporal model is designed to capture time-varying effects.checked 4 Oct 2026

Visualizations

The documentation describes 14+ interactive visualizations, including performance metrics, channel analysis, economic contributions, and DAG networks.checked 4 Oct 2026