Lead scoring models, compared.

A lead scoring model is the framework you use to turn lead behavior and attributes into a single number representing buying intent. The three main types are rule-based (manual point rules), predictive (statistical regression on past conversions), and AI/machine-learning models that continuously retrain on new data.

Rule-based is easy to start with but goes stale fast and reflects opinions, not data. Predictive models are more accurate but usually require a data team and a CRM. AI models like Catch before they bounce's combine the accuracy of predictive scoring with continuous learning, and run on raw website behavior — no CRM required.

Traditionally: list the signals that correlate with conversion, assign points to each, sum them, and threshold. Modern approach: feed historical behavior + conversion outcomes to a model, let it learn the weights, and retrain regularly. Catch before they bounce handles the modern approach automatically — you don't build the model, the model builds itself.

Demographic/firmographic (industry, company size, role), behavioral (pages viewed, dwell time, return visits, form actions), source (UTM, referrer), and friction (rage clicks, form abandonment, exit intent). Catch before they bounce captures all behavioral and friction signals out of the box.

Static rule-based models should be reviewed quarterly — buyer behavior drifts. Predictive models are typically retrained monthly. Catch before they bounce retrains continuously as new conversions come in, so the model never goes stale.