End-to-end ML pipeline for understanding what drives customer conversion. Predict customer conversion from product features and review sentiment using Snowflake ML and Cortex. Includes explainability, NLP diagnostics, and a Streamlit dashboard for business insight.
Disclamer: This project was forked from Snowflake-Labs, and modified: https://github.com/Snowflake-Labs/sfguide-build-ml-models-for-customer-conversions.git
PROJECT STRUCTURE ├── README.md ├── Customer Conversions.ipynb # Full Snowflake ML pipeline (rebuilt from Snowflake's lab) ├── streamlit_CCapp.py # The interactive dashboard (coded with Copilot) ├── EnableCrossRegion.sql # Configuration for cross-region Snowflake access ├── Setup ACCOUNTADMIN Role.sql # Role setup ├── SetupGuide.txt # Step-by-step process ├── .gitignore # Ignore checkpoints, cache, etc. ├── assets/ # Optional: screenshots or visuals └── Dashboard_ML.jpg └── Dashboard_NLP.jpg └── Environment Setup.jpg └── data/ # Synthetic data └── synthetic_review_data_sourcetable.csv └── synthetic_review_data_text.txt
This lab demonstrates how to build an end-to-end machine learning (ML) pipeline in Snowflake to predict customer conversion based on product features, review sentiment, and NLP signals — and visualize the results in an interactive Streamlit dashboard.
🚀 Features
- Train and deploy an XGBoost model using Snowflake ML
- Enrich reviews with sentiment using Snowflake Cortex
- Segment users by conversion likelihood and emotional tone
- Diagnose model residuals and feature impact
- Explore insights in a Streamlit dashboard
📊 Interactive Dashboard
https://app.snowflake.com/.../#/streamlit-apps/HOL_DB.HOL_SCHEMA.NZC_RN57VD3SQK25