Marketing Campaign Performance & Optimization Case Study

This repository contains an end-to-end analysis of a digital marketing campaign dataset to identify conversion drivers and recommend optimal budget allocations.

Project Overview

In this case study, we analyze a digital marketing campaign dataset from Kaggle to uncover the key factors driving customer conversions for a mid-sized e-commerce brand. Through exploratory analysis, predictive modeling, and what-if budget simulations, we quantify channel performance, evaluate engagement metrics, and recommend the optimal marketing mix under a fixed budget.

Objectives

• Quantify Channel Performance: Compute CTR, Conversion Rate, Ad Spend and Cost per Acquisition (CPA) for each campaign channel.
• Segment Audience Cohorts: Create demographic (AgeGroup, IncomeBracket) and engagement (EmailEngagementRate, SocialEngagementScore) segments to identify high-value users.
• Build & Evaluate Models: Train and compare Logistic Regression and Random Forest classifiers to predict conversion likelihood and assess feature importance.
• Simulate Budget Scenarios: Develop a simulation function to estimate conversions for different spend allocations and recommend the mix that maximizes ROI under a $50 K budget.

Dataset

The dataset is available on Kaggle: https://www.kaggle.com/datasets/rabieelkharoua/predict-conversion-in-digital-marketing-dataset

Requirements

Install the required Python packages:

pip install -r requirements.txt

Notebook

Open and run marketing_campaign_analysis.ipynb to perform:

1. Data Loading & Initial Exploration
2. Data Cleaning & Preprocessing
3. Exploratory Data Analysis (EDA)
4. Predictive Modeling (Logistic Regression & Random Forest)
5. Budget Allocation Simulation & Recommendations

Screenshots

Channel‐Level Performance

IncomeBracket Cohort Analysis

Intereactive Tableau Dashboard

Click here to view the interactive dashboard on Tableau Public

Key Insights

• Social Media delivers the highest conversion rate (0.1066) and the lowest cost per acquisition (~$46.6K).
• PPC has the highest click-through rate (0.158) and, when allocated 40% of the budget, maximizes conversions (~38.6 per $50K).
• Behavioral Metrics: TimeOnSite and PagesPerVisit are the strongest predictors of conversion.
• Cohort Uniformity: AgeGroup and IncomeBracket show minimal variation in conversion and engagement metrics.
• Model Performance:
  • Scaled Logistic Regression: ROC-AUC ≈ 0.78, balanced precision/recall.
  • Random Forest: ROC-AUC ≈ 0.80, high recall for converters but low recall for non-converters.

Usage

Adjust budget scenarios in the "Budget Simulation" section of the notebook to test different spend allocations and expected conversion outcomes.

This project is part of a broader portfolio showcasing real-world data analysis. Visit My GitHub Portfolio to explore more case studies.