Optimizing Home Delivery in Small Grocery Stores
Manideep racharla
Optimizing Home Delivery in Small Grocery Stores
A Data-Driven Strategy for Customer Segmentation and Marketing
✨ Overview
This project investigates how small grocery stores can leverage data analytics to optimize customer engagement, retention, and marketing strategies. Using RFM analysis, K-Means clustering, and machine learning models for churn prediction, this study provides actionable insights for improving customer retention and targeted marketing.
🛠 Objectives
RFM Analysis: Quantify customer engagement using Recency, Frequency, and Monetary metrics.
Customer Segmentation: Segment customers into actionable groups (e.g., loyal, at-risk) using K-Means clustering.
Churn Prediction: Predict disengaged customers using machine learning models like Logistic Regression, Random Forest, and Gradient Boosting.
Insights and Recommendations: Develop targeted marketing strategies for each customer segment.
📊 Methods and Techniques
RFM Analysis: Customers are analyzed based on:
Recency: Days since the last purchase.
Frequency: Total number of transactions.
Monetary: Total spending value.
K-Means Clustering:
Used to segment customers into 4 clusters (Loyal, At-Risk, Moderate, Occasional).
Optimal clusters determined using the Elbow Method.
Silhouette Score: 0.601.
Churn Prediction:
Models used: Logistic Regression, Random Forest, Gradient Boosting.
Achieved 100% accuracy with Recency as the most critical feature (based on feature importance).
🧰 Technologies
Languages: Python
Libraries:
Data Manipulation:
pandas, numpyVisualization:
matplotlib, seabornMachine Learning:
scikit-learnTools: Jupyter Notebook
📈 Key Results
RFM Analysis:
Segmented customers based on their engagement and spending patterns.
Identified clusters:
Loyal Customers: High frequency and monetary value, low recency.
At-Risk Customers: High recency, low frequency and monetary value.
Clustering:
Four customer groups identified using K-Means clustering, validated with a silhouette score of 0.601.
Churn Prediction:
Models achieved 100% accuracy in predicting churn.
Recency was the most important feature, contributing over 85% of the model's predictive power.
🌟 Insights and Recommendations
Loyal Customers (Cluster 0):
Offer loyalty rewards and exclusive deals to retain these customers.
At-Risk Customers (Cluster 3):
Focus on re-engagement campaigns like personalized discounts or reminder emails.
Occasional Buyers (Cluster 2):
Upsell with time-sensitive offers to encourage more frequent purchases.
🔮 Future Work
Real-Time Analytics: Implement a real-time pipeline for dynamic customer segmentation.
Incorporate Additional Data: Add demographic and behavioral data for deeper insights.
Scalability Testing: Apply methods to larger datasets or multi-store environments.
Run individual scripts:
RFM Analysis:
python Code/rfm_analysis.pyClustering:
python Code/kmeans_clustering.pyChurn Prediction:
python Code/churn_prediction.py👨🎓 About the Author
This project was developed as part of my coursework at Saint Louis University. Special thanks to my instructor, Dr. Ravindranath Arunasalam, for his guidance throughout this study.
Feel free to reach out via LinkedIn or email me at manideepguptha139@gmail.com.
Like this project
Posted May 5, 2025
This project explores how small grocery stores can optimize customer engagement using RFM analysis, K-Means clustering, and churn prediction.
