Lead scoring

It is the main mechanism most teams have for handling more leads than sales can call, which makes it the thing standing between a spam wave and a wasted quarter of SDR time.

And it is built, almost universally, on fields the submitter filled in themselves. That is fine when the submitter is a person with no reason to lie. It is the entire problem when they are not.

A fabricated submission claiming VP of Engineering, 500–1000 employees, budget approved outscores a real founder who honestly answered 1–10 employees, just looking. The score does what it was built to do and gets it exactly backwards.

• Weight verified attributes above declared ones. An enriched company record from the email domain is evidence. A dropdown selection is a claim.
• Score the request, not just the requester. Which page, which campaign, what time, how long they spent, how they arrived — none of it is self-reported.
• Back-test against closed-won. Take last year’s leads, apply the current model, and check whether the high scores actually closed. Most models have never been through this.
• Recalibrate on a schedule. A score inherited from a template built for someone else’s business is a horoscope with arithmetic.

A lead score is a prediction. Predictions are worth exactly what their track record says they are worth, and almost nobody keeps the track record — the model is set up during implementation, tuned once by someone who has since left, and trusted indefinitely.

Grading it is not hard. It requires that each submission eventually carries what happened to it, which is the same requirement as everything else on this site and the same thing no form builder currently provides. See verdict and outcome webhook.