RetailPulse AI: B2B Churn Prediction Engine

A Compound AI System for predicting B2B Churn using AWS SageMaker Canvas.

🚀 Project Overview

RetailPulse is an autonomous retention architecture designed to detect "Silent Attrition" in offline B2B retail. By leveraging AWS SageMaker Canvas, this project operationalizes a predictive model to identify churn risk 60 days before revenue impact.

📂 Repository Contents

• RetailPulse_Research_Paper.pdf: Full architectural breakdown and methodology.
• AWS_Clean_Upload.csv: The synthesized training dataset (N=1,000 retailers).
• Screenshots/: Evidence of model accuracy and single-prediction simulations.

🛠️ Tech Stack

• Platform: AWS SageMaker Canvas (No-Code ML)
• Model Type: Binary Classification (XGBoost)
• Target Variable: Churned_YesNo
• Performance: >95% Predictive Accuracy

⚡ Key Findings

1. Recency is Critical: The DaysSinceLastOrder variable accounted for the majority of model impact.
2. The 45-Day Threshold: Retailers inactive for >45 days show a non-linear spike in churn probability.

---

Created by Sumanta Pani - Product Manager Portfolio