Problem Statement : The E-commerce customer churn rate is up to 80% compared with traditional business customer management (Wu & Meng, 2016). From dataset customer churn is 17%.
Goals : The purpose of this project is to predict customer churn, loss opportunity and provide recommendations to the business team so the company can implement a persona customer retention strategy and can monitoring throught dashboard interactive.
Result : The results show that tenure and complaint has the greatest impact on a churn rate. With applying an actionable recommendation from insights, company can avoid loss oppurtunity up to $900,000 and revenue lift up to $ 150,000.
Tools: Python, JupyterLab, Git, PowerBI
Libraries: Pandas, Numpy, Feature-engine, Scikit-learn, Imbalanced-learn, kaplan-meier survival curve, lifelines.CoPHfilter, lifelines.predict_survival_function, K-means, gaussian, rfm-segmentation
Dataset: Ecommerce Customer Churn Analysis and Prediction [source]
Summary of the analysis
- This dataset has 5630 observations and 20 variables with 15 numerical variables, 5 categorical variables and 2 target variable.
- From the data visualization, it is obtained that the churn ratio has a correlation with tenure, complaints, cashback Amount, & preferred order cat.
- The results of predicting churn are strongly influenced by the level of Tenure, Complaint, Number of Addresses, and cashback Amount.
- The results of the Survival Analysis, the customer has the greatest survival chance in No Complain, Marital Status Married, Payment Mode Credit Card, Order Category Grocery.
- RFM Segmentation results show priority customer treatment in the Loyal, New, Promising, and Lost Potential segments.
- Total Expected Loss of $ 910,687
- Estimated Revenue Uplift
- Order category Grocery $42,448
- Payment Credit Card $ 91,785
- Payment Debit Card $ 78,543
What I have learned
- Framing the business problem.
- Create a machine learning model and extract insight that generates churn & retention from it to make an actionable recommendation for the business team.
- Create a survival analysis, predict customer who will churn in Future and extract insight that generates churn & retention from it to make an actionable recommendation for the business team.
- Create a customer segmentation with RFM Segmentation, KMeans and Gaussians that can generates strategy-strategy personal customer.
- Create a dashboard interactive can monitoring business metrics, operational, and sales.
- Make a business simulation from insights that calculate loss oppurtunity and revenue uplift.
File Dictionaries
- EDA_Churn_Analysis_Ecommerce_Customer_.ipynb: this notebook contains all of project details, such as business understanding, exploratory data analysis & insights from bivariate and multivariate analysis
- Supervised_Classification_Churn_Analysis_Ecommerce_Customer.ipynb: data preprocessing, Selection Models & Cross Validation, Handling Imbalance, Hyperparameter Tuning, Feature Importance with SHAP
- Model_Survival_Analysis_Churn_Analysis_Ecommerse_Customer.ipynb: Using Kaplan-Meier (KM) and COx Proportional Hazard (CPH) Model, Business Recommendation, Loss Oppurtunity and Revenue Lift
- Unsupervised_Churn_Analysis_Ecommerce_Customer.ipynb: Using RFM Segmentation, K-Means, and Gaussian, Business Recommendation and Customer Segmentation & Strategy
- ANN_E-Commerce.pbix: dashboard interactive to monitoring business metrics
- Churn Analysis Ecommerce Customer-Presentasi-ANNTeam_TSDN-2022.pdf: summary of the project.