Exploratory Data Analysis of BNPL User Behavior & Repayment Risk

📌 Project Overview

This project performs end-to-end Exploratory Data Analysis (EDA) on a large-scale Buy Now Pay Later (BNPL) dataset to identify borrower characteristics and loan attributes associated with higher repayment risk.

The analysis focuses on understanding repayment behavior, risk signals, and segment-level patterns relevant to credit underwriting and risk policy decisions.

---

🎯 Objectives

• Analyze distributions of key financial and behavioral variables
• Examine relationships between borrower attributes and loan outcomes
• Identify high-risk segments based on purpose, housing status, and pricing
• Create a clean, analysis-ready dataset with documented assumptions
• Derive actionable insights for BNPL risk strategy

---

📂 Dataset Details

• Source: Kaggle – BNPL Dataset (v1)

• Records: ~1,048,000

• Features: 29 (16 numerical, 13 categorical)

• Target Variable: loan_status
  (Current, Fully Paid, Charged Off, Default, Late, etc.)

  Dataset is sourced from Kaggle and is subject to its original license and usage terms.

📂 Dataset Access

Due to GitHub file size limits, the dataset is not included in this repository.

🔗 Dataset Source:
Kaggle – BNPL Dataset (v1)
https://www.kaggle.com/datasets/bdoey1/bnpl-data-v1

How to use:

1. Download the dataset from Kaggle
2. Place the CSV file inside the data/ folder
3. Run the notebook normally

---

Key Features

• Financial: loan_amount, interest_rate, monthly_payment
• Credit Profile: annual_income, total_dti, credit_limit
• Behavior: delinq_2yrs, num_accts_120_pd, mort_acc
• Segmentation: loan_purpose, home_ownership, application_type

---

🧹 Data Cleaning & Preparation

• Parsed text-based fields (loan term, employment length)
• Median imputation for missing numerical values
• Dropped high-missing columns (e.g., emp_title)
• IQR-based outlier capping for financial stability
• Validated data types and removed inconsistencies

---

📊 Analysis Performed

✔ Univariate Analysis

• Distribution of loan status (class imbalance observed)
• Numerical distributions before & after outlier treatment
• Categorical breakdowns (purpose, housing, application type)

✔ Bivariate Analysis

• Numerical vs numerical (income vs loan amount, payment vs term)
• Categorical vs numerical (loan amount & interest rate by loan status)
• Categorical vs categorical (purpose & housing vs loan outcomes)

✔ Multivariate Analysis

• Parallel coordinate plots
• Interaction between income, loan size, interest rate, and outcomes
• Segment-level risk concentration analysis

✔ Pivot Table Insights

• Charge-off rates by home ownership
• Risk concentration across loan purposes
• Segment-level volume vs risk trade-offs

---

🔍 Key Insights

• Higher interest rates cluster with adverse loan outcomes
• Debt consolidation & credit card loans show elevated risk by volume
• Renters exhibit higher charge-off rates than mortgage holders
• Income alone is insufficient—multivariate interactions matter
• Outlier treatment improved distribution quality without biasing medians

---

📊 Key Visual Insights

Loan Amount Distribution - View visualization

Loan Amount vs Loan Status - View visualization

Loan Purpose vs Repayment Risk - View visualization

Visuals are generated from the full dataset and represent portfolio-level risk patterns. Additional exploratory plots are available in the /visuals directory.

---

🛠 Tools & Technologies

• Python
• Pandas, NumPy
• Matplotlib, Seaborn
• Jupyter Notebook

---

👤 Author

Teja Kesarapu
Data Analyst | Python | SQL | Power BI | FinTech Analytics