For the best viewing experience, please open this page on a desktop or laptop.

Machine Learning Project

E-Commerce Customer Churn:

An E-Commerce Churn Prediction Engine

Leveraging machine learning to identify at-risk customers in an imbalanced dataset (16.8% churn rate), with strongest business value from recall-focused targeting and scenario-based retention planning.

0

%

Churner Recall

Decision Tree Classifier

0

%

Current Churn Rate

947 of 5,630 customers

0

K

Revenue at Risk

Immediate intervention needed

0

K

Annual Retention Potential

With targeted strategies

Executive Summary

This project addresses a critical business challenge: 16.8% customer churn resulting in significant revenue loss for an e-commerce platform serving 5,630 customers.

Using machine learning techniques, I developed a predictive model where the priority metric is 52% churner recall (on a 16.8% churn base rate), enabling proactive intervention while accounting for class imbalance.

Key Discoveries:

  • Early Tenure Crisis: 0-3 month customers show ~50% churn rate
  • Complaint Impact: Customers with complaints have 31.7% churn (2x higher)
  • Cashback Paradox: Mid-satisfaction (3-4 score) customers with low cashback usage at highest risk
  • Projected Impact (Illustrative): Scenario modeling suggests targeted strategies could reduce churn to below 12%, depending on intervention success rates.

Business Impact

Customers Saved ~270
Immediate Revenue Illustrative
Annual Potential Illustrative
Churn Reduction 29%

Model Performance

Churner Recall
52%
Overall Accuracy
89.3%
AUC Score
0.88

Project Journey

Navigate through each phase of the analysis

Project Resources

Download presentation materials and access code repositories

Code Repository

Full analysis code, model training, and evaluation scripts

View on GitHub

Jupyter Notebook

Interactive analysis notebook with detailed methodology

Download Notebook

Tableau Dashboard

Interactive visualizations on Tableau Public

Open Tableau