Live demo · synthetic model

Lead Scoring Simulator

A live logistic propensity model — the pattern behind a Salesforce-integrated scorer that lifted conversions +15% for a premium automotive brand. Move the CRM signals on the left; the score, tier, and feature contributions react instantly.

Logistic regressionExplainabilityCRM scoring

Lead signals

Model output

Why this score? (feature contributions)

How it works

Score = σ(w·x + b)

Each signal is normalized, multiplied by a trained weight, summed with a bias, and squashed through a sigmoid into a 0–100 propensity score. The contribution bars show each feature's w·x term — the same explainability view sales teams used in the production version to know why a lead was hot, not just that it was.

In production this ran end-to-end: scikit-learn training on CRM history, scores written back into Salesforce for real-time call-list prioritization, with Power BI tracking score-band conversion to keep the model honest.