🧠 Customer Churn Analysis — Real-World SQL + Python Project

Welcome to a hands-on project that simulates real-world data challenges — from raw data ingestion to drawing meaningful insights that support business decision-making.

📍 GitHub Repository: customer-churn-analysis

---

📌 Project Objectives

✅ Ingest messy, real-life-like data ✅ Clean and preprocess using Python (Pandas) ✅ Load structured data into SQL Server ✅ Transform and correct business logic in SQL ✅ Analyze customer behavior, churn patterns, and retention ✅ Summarize insights for decision-makers

---

🧰 Tools & Technologies

• Python (Pandas, NumPy)
• SQL Server (T-SQL)
• SQLAlchemy for DB connections
• Jupyter Notebook for analysis and documentation
• Matplotlib / Seaborn for visualization (optional)
• GitHub for version control

---

📂 Project Structure

customer-churn-analysis/
│
├── data/                    # Raw & messy input data (CSV, tsv, json)
├── project-env/             # python virtual environment
├── SQL Codes/               # SQL DDL, DML, transformations & analysis
├── SQL_Output_Tables/       # Files to analyze/visualize in Python after SQL analysis
├── py_cleaning_eda.ipynb    # python notebook for initial dat cleaning and EDA
├── py_analysis.ipynb        # python notebook to visualize reports from SQL analysis
├── requirements.txt         # required python libraries for analysis
└── README.md

---

🔄 Workflow Overview

1. Data Ingestion & Cleaning (Python)

• Loaded messy datasets from CSV, tsv & json formats
• Fixed data types, nulls, duplicate entries
• Aligned columns for SQL import

2. Data Modeling & Loading (SQL Server)

• Created normalized tables: customers, subscription, product_usage, support_tickets
• Loaded cleaned data using SQLAlchemy

3. SQL-Based Transformation & Fixes

• Fixed logical issues like:

  • end_date earlier than start_date
  • Feature usage before signup

• Added validation logic

4. Data Analysis (SQL)

• Identified churned customers

• Conducted cohort analysis to track retention over months

• Measured the impact of:

  • Product feature usage
  • Support ticket resolution

• Created a churn flag and compared against engagement metrics

---

📊 Key Business Insights

• 📉 Feature usage in the last 30 days moderately correlates with reduced churn
• ⏱️ Unresolved tickets have weak/no significant impact on churn
• 🔁 April 2023 cohort had the highest long-term retention
• 📆 Many customers churn within the first 3–4 months post-signup

---

🚀 How to Run

1. Clone the repo:

   git clone https://github.com/HeatTransfer/customer-churn-analysis.git

2. Set up Python environment and install dependencies

3. Run notebooks to ingest and clean data

4. Execute SQL scripts under /sql_scripts in SQL Server

5. Explore analysis and observations

---

📌 Author

Shreyajyoti Dutta 🔗 LinkedIn Profile 📫 Open to opportunities in Data Analytics, Data Engineering, and BI

---

🏷️ Tags

SQL Python Data Engineering Churn Prediction Cohort Analysis ETL Business Insights Data Analytics