E-Commerce Transaction Data Analysis

A comprehensive data analysis project exploring transactional patterns, customer behaviour, and product performance using real-world e-commerce data. This project demonstrates advanced data cleaning, feature engineering, and exploratory data analysis (EDA) to extract actionable business insights.

Dataset Overview

• Filename: purchase_data.csv
• Rows: 541,909 transactions
• Columns: 8 (InvoiceNo, InvoiceDate, Quantity, UnitPrice, StockCode, Description, CustomerID, Country)
• Timeframe: Approximately 2 years
• Geography: 38 countries

Key Features

• Invoice-level granularity
• Customer-specific insights
• Product descriptions and unit prices
• Date and time stamps for time-series analysis

Data Cleaning Highlights

• Missing CustomerID replaced with "Anonymous"
• Description cleaned and title-cased
• Negative or zero Quantity and UnitPrice values removed
• InvoiceDate standardized using European format
• Removed cancelled transactions (InvoiceNo starting with "C")

🛠️ Feature Engineering

• Revenue: Quantity * UnitPrice
• Temporal Columns: Month, Quarter, Season
• Product Categories: Generated via clustering of Description

Exploratory Data Analysis (EDA)

1. Monthly Fluctuations

• Detected revenue spikes in July and October 2011
• Stable transaction count, but revenue increased due to high-value purchases

2. Top Product Categories

• "Gifts & Decorations" leads revenue across all months
• "Bags & Accessories" and "Household Essentials" show consistent growth

3. Seasonality

• September peaks tied to seasonal demand for gifts
• Stable product lines identified for year-round promotions

4. Customer Behaviour

• High-value outliers and loyal customer clusters identified
• Cross-category spending drives higher customer value

5. Anonymous Customers

• Represent significant revenue (e.g., £640,000+ in "Gifts & Decorations")

Limitations

• Lack of demographic data (age, income, etc.)
• Outliers affect distribution, even after removal
• Country-level analysis only—no granular geography
• Product grouping relies on unsupervised text clustering

Recommendations

• Add demographic fields to enable deeper segmentation
• Use supervised models (e.g. Random Forest, XGBoost) for predictive insights
• Integrate external data (holiday calendars, marketing spend)
• Build interactive dashboards for real-time decision-making

Tools & Libraries

• Pandas, NumPy – data wrangling
• Matplotlib, Seaborn – visualisation
• Scikit-learn – clustering and potential model integration
• Jupyter Notebook – analysis environment

License

This project is released for academic and non-commercial use. Attribution is appreciated.

References

• Pandas Documentation
• Matplotlib Documentation
• Seaborn Documentation
• NumPy Documentation
• Scikit-learn Documentation
• Python Documentation

---

Author: Fatimo Adenike Adeniya

Exploratory Data Analyst | E-Commerce Insights | Python Enthusiast