Skip to content

Vasuk12/Marketing-Mix-Modelling

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Marketing-Mix-Modelling

Bayesian Marketing Mix Modeling (MMM)

This project estimates the Return on Investment (ROI) of various marketing channels using Bayesian Linear Regression with PyMC. It reflects key skills outlined in data science roles, particularly at organizations like Bank of Ireland, including statistical modeling, causal inference, and marketing analytics.


Objective

To quantify the impact of media spend across different channels (TV, Digital, SEM, etc.) on total GMV (Gross Merchandise Value) and identify the most efficient channels using a probabilistic approach.


Tools & Libraries

  • Python 3.x
  • PyMC – Bayesian inference
  • ArviZ – Posterior analysis and visualizations
  • Pandas, NumPy – Data manipulation
  • Matplotlib, Seaborn – Visualization
  • scikit-learn – Standardization

Dataset Overview

The dataset includes:

  • Monthly GMV and units sold per category
  • Marketing channel investments (e.g., TV, Digital, SEM)
  • Special events and Net Promoter Score (NPS)

Methodology

  1. Feature Selection: Marketing spend variables + NPS
  2. Standardization: To prepare data for PyMC modeling
  3. Bayesian Regression: Using prior/posterior distributions to model GMV
  4. ROI Estimation: Based on contribution per €1 spent
  5. Posterior Analysis: Interpretation of coefficients & credible intervals

Key Results

ROI Ranking (Top Channels)

  • ContentMarketing: Highest ROI
  • Sponsorship: Strong positive return
  • SEM & Digital: Moderate ROI
  • Affiliates / OnlineMarketing: Lower ROI

NPS Insight

  • Posterior mean: -1.2
  • 94% HDI: [-3.1, 0.7]
  • Suggests an inconclusive but potentially negative effect of NPS on GMV

Business Recommendations

  • Scale up ContentMarketing and Sponsorship efforts
  • Audit & optimize underperforming channels like Affiliates
  • Investigate NPS further — inconclusive link may suggest confounding factors or data limitations
  • Consider time-lagged modeling for long-term media effects

Next Steps

  • Add lag variables for time-series causal effects
  • Integrate A/B testing or simulated experiments
  • Extend with hierarchical or dynamic Bayesian models

Skills Demonstrated

  • Bayesian inference & causal analysis
  • ROI interpretation with uncertainty
  • PyMC modeling pipeline
  • Data storytelling for marketing analytics

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published