What This App Does
#Two parts, both inside Excel:
- Database: list tables, pull first or random rows, and run any SQL on a Neon Postgres database, directly from Excel over HTTPS with no server in between
- Analysis: EDA on any Excel table: summary statistics, count and response rate for every category, correlation matrix, charts
- Model: XGBoost response model with a stratified train / test split, Gini, deciles with lift, a score per row, ROC and gain charts
- Ships with a 10,000-row bank marketing sample, so the analysis works without a database
- The connection string is read from an xlwings Lite environment variable and is not saved in the workbook
Important Note
The model is a worked example, not a point-and-click product. For a new dataset, review the column choices and settings, or work through them with an AI coder.
How to Use
#- Download & open: get the workbook and open xlwings Lite (Home > Add-ins). The HOW_IT_WORKS sheet explains every step; the buttons are on the MASTER sheet.
- Try it first: click "4. Run EDA", then "5. Build Model" on the sample table DEF
- Connect your database (optional): copy the connection string from the Neon console, then in xlwings Lite add it under Settings > Environment Variables as NEON_DATABASE_URL, scope Local, and restart xlwings Lite
- Explore: "1. List Tables", "2. Get Table Data" and "3. Run SQL Query"; the results are Excel tables you can analyze in step 3
How It Works
#Database connection
- Python in xlwings Lite runs in the browser, which cannot open a direct database connection
- Neon offers an HTTPS SQL endpoint that browsers can call, so Excel queries it directly with your connection string
- For MySQL or another Postgres, put a small web API in front of the database (example code linked below)
Analysis and model
- All processing runs locally in Python inside Excel
- XGBoost and scikit-learn run in the browser environment
- Results are written as Excel tables and charts; the input table is not changed
