How I Used RFM and Regression to Uncover Business Insights from Customer Data

Introduction

In today’s competitive business world, data isn’t just numbers — it’s insight, strategy, and opportunity. In this project, I explored real-world transactional data to analyze customer behavior, sales performance, and future value using end-to-end data analytics.

This assignment involved RFM-based customer segmentation, cohort-based retention analysis, product/location insights, and predictive modeling using regression techniques — all executed using Excel and Python.

Project Overview

• Objective: Understand customer segments and predict purchasing behavior to support marketing and strategic decisions.
• Dataset: Realistic customer transaction dataset
• Tasks:

1. RFM Segmentation
2. Cohort & Retention Analysis
3. Product & Location Analysis
4. Predictive Modeling using Regression

• Tools Used:
• Excel (Data cleaning)
• Python (Pandas, NumPy, Matplotlib, Scikit-learn)
• Jupyter Notebook

Data Cleaning & Preparation

• Cleaned missing/null values and removed inconsistent records using Excel
• Ensured uniform formats for date, product, and customer fields
• Imported the .csv into Jupyter Notebook
• Feature Engineering included:
• Customer Age
• Signup Tenure
• Average Purchase Value
• Total Purchases
• Discount Used

Task 1: RFM Analysis & Segmentation

🧮 What is RFM?

• Recency: Days since last purchase
• Frequency: Total purchase count
• Monetary: Total spending

Using these metrics, I segmented customers into:

• High Value
• At Risk
• Lost
• New Customers
• Potential Loyalists

Business Strategy:

• High Value: Loyalty programs, upselling
• At Risk: Re-engagement offers
• Lost: Exit surveys or cost-effective promotions

Task 2: Cohort & Retention Analysis

Grouped customers into monthly cohorts based on first purchase date and tracked retention over time.

📌 Insight: Most cohorts dropped after Month 2. Retention rates declined steeply over 3–4 months.

🧠 Actionable Strategy:

• Build onboarding campaigns for new users
• Add post-purchase follow-ups or loyalty incentives

ask 3: Product, Category & Location Insights

🛍️ Category-Level Findings:

• Home and electronics performed well
• Books and grocery underperformed

🌍 Location-Level Findings:

• High performing: Mumbai, Bangalore
• Low performing: Jaipur, Bhopal

📌 Business Strategy:

• Focus marketing and inventory on high-performing locations and categories

ask 4: Predictive Modeling — Regression

Built a regression model to predict Final Purchase Amount using features like:

• Age, Tenure
• Avg Purchase Value
• Total Purchases
• Discount Usage

Model Used:

• Compared Linear Regression and Random Forest Regressor
• Chose Random Forest due to higher accuracy

🧪 Evaluation Metrics:

• R² Score: 0.985
• RMSE: 20.58
• MAE: 9.8

Key Business Insights

• Most revenue comes from a small group of loyal, frequent buyers
• Retention drops steeply after 2–3 months — early loyalty efforts are essential
• Certain cities and categories outperform others — optimize resource allocation
• Predictive models can help personalize marketing and increase conversion

Let’s Connect!

If you enjoyed this project breakdown or found it useful, feel free to:

• 💬 Share your thoughts in the comments
• 🌐 Connect with me on (3) shubhi verma | LinkedIn

Tags:

#DataAnalysis #CustomerSegmentation #PowerBI #Python #RFM #Regression #BusinessIntelligence #MachineLearning