📣 Campaign Performance Prediction

End-to-end Machine Learning & Analytics project for DOOH Advertising

Predicting ROI · Classifying Campaign Success · Visualizing Business Intelligence

---

📌 Project Overview

This project was built during my internship at Lemma Technologies, a DOOH (Digital Out-of-Home) advertising company. The goal was to analyze advertising campaign data, predict ROI using regression, classify campaign success using Random Forest, and surface insights through an interactive Power BI dashboard.

The project covers the full data science pipeline — from raw data exploration all the way to a business-ready dashboard used by the team.

---

📊 Dashboard Preview

Executive Business Overview

6 KPIs tracked: ₹7.42M Revenue · ₹5.33M Ad Spend · 42.24 Avg ROI · 27M Impressions · 8M Clicks · 30.82% CTR — filterable by ScreenType, Industry, Time Slot, and Weather Condition.

Campaign Intelligence & Optimization

CTR vs ROI scatter by industry · Revenue treemap by ScreenType · Decomposition tree for weather/time analysis · Top revenue-generating clients (Ola, Netflix, OnePlus, Zomato).

---

🗂️ Repository Structure

Campaign-Performance-Prediction/
│
├── _EDA Lemma.ipynb                        # Exploratory Data Analysis
├── Campaign Performance Prediction.ipynb   # Classification model (Random Forest)
├── Regression Model to predict ROI.ipynb   # Regression model (Linear + RF)
├── Insights.ipynb                          # Business insights & findings
├── Execute Business Overview.pbix          # Power BI dashboard file
|── dashboard_overview.png
|── campaign_intelligence.png
└── README.md

---

🔬 Project Workflow

1. 📁 Exploratory Data Analysis — _EDA Lemma.ipynb

• Distribution analysis of revenue, impressions, clicks, CTR, and ROI
• Industry-wise and screen-type-wise breakdowns
• Correlation heatmaps and outlier detection
• Time slot and weather condition impact on performance

2. 🤖 Campaign Performance Classification — Campaign Performance Prediction.ipynb

• Binary classification: High vs Low performing campaigns
• Feature engineering on CTR, ROI, and Ad Spend
• Random Forest Classifier — outperformed Logistic Regression on non-linear patterns
• Evaluation: Accuracy, Precision, Recall, F1-Score, ROC-AUC
• Feature importance analysis to identify top campaign drivers

3. 📈 ROI Regression Model — Regression Model to predict ROI.ipynb

• Predicting campaign ROI from input features
• Compared Linear Regression vs Random Forest Regressor
• Metrics: R², MAE, RMSE
• Residual analysis and prediction vs actual plots

4. 💡 Business Insights — Insights.ipynb

• Industry-level ROI benchmarks (Entertainment & Food top performers at ROI ~44–45)
• Evening time slots drive the highest revenue (₹2.7M vs ₹0.7M at night)
• Revenue nearly evenly distributed across all 5 screen types (~20% each)
• Top clients: Ola, Netflix, OnePlus, Zomato, Samsung, Flipkart

5. 📊 Power BI Dashboard — Execute Business Overview.pbix

• 2-page interactive dashboard
• Cross-page slicers: ScreenType, Industry, Time Slot, Weather Condition
• Treemap, donut chart, scatter plot, bar charts, decomposition tree
• Designed for executive-level decision making

---

🛠️ Tech Stack

Category	Tools
Language	Python 3.x
Data Manipulation	Pandas, NumPy
Visualization	Matplotlib, Seaborn
Machine Learning	Scikit-learn (Random Forest, Linear Regression, Logistic Regression)
BI Dashboard	Microsoft Power BI
Notebook Environment	Jupyter Notebook

---

📐 Key Results

Model	Task	Metric	Result
Random Forest Classifier	Campaign Success Prediction	ROC-AUC	Strong separation
Linear Regression	ROI Prediction	R²	Baseline established
Random Forest Regressor	ROI Prediction	R² / RMSE	Improved over linear

Exact metric values are documented inside the respective notebooks.

---

💼 Context

This project was developed as part of my role as a Technical Intern at Lemma Technologies (May 2026 – Present). Lemma is a DOOH advertising platform, and this analysis directly supported campaign optimization decisions by the business team.

---

🚀 How to Run

# 1. Clone the repository
git clone https://github.com/prathmesh0work/Campaign-Performance-Prediction.git
cd Campaign-Performance-Prediction

# 2. Install dependencies
pip install pandas numpy matplotlib seaborn scikit-learn jupyter

# 3. Run notebooks in this order
#    _EDA Lemma.ipynb  →  Campaign Performance Prediction.ipynb
#    →  Regression Model to predict ROI.ipynb  →  Insights.ipynb

jupyter notebook

For the Power BI dashboard, open Execute Business Overview.pbix in Microsoft Power BI Desktop.

---

👤 Author

Prathmesh Ingole Data Analyst Intern @ Lemma Technologies · Pursuing M.Sc. Data Science

---

Built with real internship data · Designed for business impact · Owned end to end.