Bayesian Marketing Mix Modeling

Overview

This project applies Bayesian Marketing Mix Modeling (MMM) to analyze the impact of advertising, pricing, and competitor activities on sales performance in the fiber-optic telecommunications industry. The goal is to provide a robust, probabilistic approach to optimize marketing budget allocation.

Solution Approach

We developed and tested three different Bayesian models to address the challenges of traditional MMM:

• Model 1: Utilizes Fourier basis functions to capture seasonality in sales trends.
• Model 2: Implements a Gaussian Process (GP) to model time-varying advertising effectiveness.
• Model 3: Combines Fourier basis and Gaussian Process for a more comprehensive representation of seasonal and dynamic effects.

Each model incorporates non-linear transformations to account for diminishing returns on advertising spend and carry-over effects over time.

Repository Contents

• Models: Python scripts implementing the three Bayesian MMM models.
• Report: Full project report detailing the methodology, data preprocessing, modeling steps, and results.
• Final Presentation: Presentation slides summarizing the key findings and conclusions of the project.

How to Use

1. Review the Report folder for a detailed explanation of the methodology.
2. Explore the Models folder to see the Bayesian MMM implementations.
3. Use the Final Presentation folder for a concise project summary.

Authors

• Montagna Rosita
• Peracchione Zeno
• Pertusi Federica
• Riboni Alessandro
• Tomasi Francesco
• Vozza Daniele