User Retention Cohort Analysis

Tools: Python · Pandas · Matplotlib · Seaborn
Dataset: UCI Online Retail Dataset (541,909 transactions)
Domain: Product Analytics / E-commerce

Business Question

After a customer makes their first purchase, do they come back?
This project analyzes one full year of real e-commerce transaction data to measure customer retention across monthly cohorts.

Dataset Source: UCI Machine Learning Repository — Online Retail Dataset Period: December 2010 – December 2011 Size: 541,909 transactions, 4,338 unique customers (after cleaning)

Methodology

1. Data Cleaning — removed missing CustomerIDs, invalid quantities/prices
2. Feature Engineering — created Revenue, InvoiceMonth, CohortMonth columns
3. Cohort Analysis — tracked each monthly cohort across 13 months
4. Visualization — heatmap and bar chart of retention rates

Key Findings

1. Average Month-1 Retention: 20.62% — 4 in 5 customers never return
2. Strongest Cohort: Dec 2010 (36.6%) — likely driven by Christmas season
3. Weakest Cohort: Nov 2011 (11.1%) — sharp acquisition quality drop
4. Retention stabilizes at 20–25% from Month 2 onwards

Recommendation Focus re-engagement efforts within the first 30 days of acquisition — that's where 80% of customers are lost.

How to Run

1. Clone this repository
2. Place Online Retail.xlsx in the project folder or add appropriate path
3. Run Analysis Online Retail.ipynb in Jupyter or VS Code