Customer Segmentation using RFM Analysis + K-Means Clustering

Overview

This project segments customers of a UK-based online retailer into distinct behavioral groups using RFM (Recency, Frequency, Monetary) analysis combined with K-Means clustering. The goal is to identify which customers drive the most revenue, which are at risk of churning, and what marketing actions are appropriate for each group.

Key Finding

Champions (30.6% of customers) generate 81.6% of total revenue — making retention of this segment the single highest-priority business action.

Dataset

• Source: UCI Online Retail Dataset
• Also available on: Kaggle
• Period: December 2010 – December 2011
• Raw size: 541,909 transactions, 8 columns
• After cleaning: 386,019 rows, 4,325 unique customers

Download the dataset from the links above and place Online Retail.xlsx in the root directory before running the notebook.

Project Steps

Step	Description
1	Data Cleaning — remove cancellations, nulls, price outliers, duplicates
2	RFM Table — build Recency, Frequency, Monetary per customer
3	Clustering — log-transform, scale, elbow method, silhouette score, K-Means
4	Segment Profiling — label clusters in plain English
5	Visualizations — scatter plot and bar chart of segments
6	Business Framing — revenue analysis and actionable recommendations

Results

Segment Profiles

Segment	Customers	% of Customers	Avg Recency	Avg Frequency	Avg Monetary
Champions	1,325	30.6%	29.8 days	9.74 orders	£5,272.90
Casual Customers	2,015	46.6%	54.6 days	2.03 orders	£591.31
Lapsed Customers	985	22.8%	254.5 days	1.38 orders	£393.76

Revenue Contribution

Segment	Total Revenue	% of Revenue
Champions	£6,986,589	81.6%
Casual Customers	£1,191,495	13.9%
Lapsed Customers	£387,852	4.5%

Business Recommendations

• Champions: VIP retention program — loyalty rewards, early product access, personalized outreach. Losing this segment would be catastrophic for revenue.
• Casual Customers: Increase purchase frequency via targeted promotions at the 45-60 day inactivity mark and product recommendations. Largest growth opportunity.
• Lapsed Customers: Single win-back email campaign. If no response in 30 days, deprioritize — marketing spend is better directed at Casual Customers.

Tech Stack

• Python 3
• Pandas — data manipulation
• NumPy — numerical operations
• Scikit-learn — K-Means clustering, StandardScaler, silhouette score
• Matplotlib — visualizations

How to Run

1. Clone this repository

   git clone https://github.com/mishapatel2537/customer-segmentation-rfm.git
   

2. Install dependencies

   pip install -r requirements.txt
   

3. Download the dataset from UCI or Kaggle and place Online Retail.xlsx in the root folder
4. Open and run Customer_Segmentation.ipynb top to bottom

Project Structure

customer-segmentation-rfm/
│
├── Customer_Segmentation.ipynb   # Main analysis notebook
├── README.md                     # Project documentation
├── requirements.txt              # Python dependencies
└── .gitignore                    # Excludes dataset file

Acknowledgements

Dataset provided by Dr. Daqing Chen, London South Bank University, via the UCI Machine Learning Repository.