📊 Strategic Data-Driven Media Campaign Evaluation Framework A full-scale, end-to-end data science project simulating real-world media campaign analysis using predictive modeling, scenario simulations, and agent-based modeling—culminating in a dynamic Tableau dashboard for stakeholder insights and strategic decision-making.
📜 Table of Contents 🎯 Project Objective
📁 Dataset Overview
🧠 Methodology
📈 Machine Learning Models
🧪 Scenario Simulation
👥 Agent-Based Modeling (ABM)
📊 Tableau Dashboard
🛠️ Tech Stack
📂 Folder Structure
🚀 How to Use This Repo
📩 Contact
🎯 Project Objective This project demonstrates how data science can drive media planning strategy by evaluating marketing campaigns and customer behavior through:
✅ Predictive modeling ✅ Scenario-based simulation ✅ Agent-based behavioral modeling ✅ Interactive visualization for business decision-making
📁 Dataset Overview The dataset contains information about marketing campaign interactions for a financial services company, including:
Demographics (e.g., income, age, marital status)
Product purchases (e.g., wine, meat, gold products)
Channel interactions (web, catalog, store)
Campaign responses (Yes/No)
Data Size: 2,240 records × 29 columns Source: Simulated / Mock Marketing Data
🧠 Methodology Data Cleaning & Preprocessing
Missing values handled (e.g., income)
Feature engineering: total spending, engagement rate, customer tenure
Encoding categorical variables
BigQuery Storage
Data stored and queried using Google BigQuery
Cleaned datasets exported to Excel for Power BI/Tableau
📈 Machine Learning Models 1️⃣ Regression (XGBoost) Goal: Predict total customer spend
Metrics: MAE = 3.57, R² = 1.00
Top Features: Total purchases, web visits, meat product spending
2️⃣ Classification (Random Forest) Goal: Predict whether a customer will respond to a campaign
Accuracy: 87%
Key Insight: High-engagement customers are more likely to respond
🧪 Scenario Simulation What-If Analysis:
Increase digital ad spend by 20%
Reduce spend on underperforming channels
Focus only on high-value customer segments
Scenario Avg Predicted Spend A - Increased Digital Spend $575.67 B - Cut Low Channels $576.01 C - Target High-Spenders $1132.81 📌 Outcome: Focusing on high-spenders has 2× the ROI.
👥 Agent-Based Modeling (ABM) Simulated long-term purchasing behavior using MESA framework:
Each customer = Agent with income, engagement & purchase probability
Agents react to campaign exposure over 12 months
Spending distribution shows most customers spend between $2,000–$3,500
📈 Insight: Personalized marketing boosts average spend and loyalty.
📊 Tableau Dashboard ✅ Live Dashboard Features: Customer spending distribution
Income vs. spending segmentation
KPI cards: Avg spend, customer count
Filter by income, campaign response, and engagement
Scenario simulation summary
📌 Link to Dashboard: [https://public.tableau.com/app/profile/brian.balbuena/vizzes]
🛠️ Tech Stack Category Tools Used Programming Python (Pandas, Scikit-learn, SHAP, XGBoost), SQL Data Storage Google BigQuery Modeling XGBoost, RandomForest, MESA (ABM) Visualization Tableau Project Mgmt Git, GitHub, Excel