MODEL RUNNING IN YOUR BROWSER

Explore customer churn.

Choose an example, change its features and see how the model score changes.

This is an educational demo. Use fictional inputs, not real personal data. Calculations run on your device; form values are never sent to a server. The score does not guarantee that a customer will leave.

The interactive form needs JavaScript and the model file. The explanation below remains available without JavaScript.

What runs in this demo?

The trained MLP v2 uses the saved order of 30 features and scaling statistics. Its architecture is 30 → 64 → 32 → 1, with ReLU hidden layers and a sigmoid output. Categories, numeric ranges and USD charges reflect the Telco sample dataset.

Earlier screenshots may use another model version. Changing a feature shows the model’s response; it does not establish real-world causation.