- Score two dimensions separately: fit (who they are) and engagement (what they do).
- A score is only useful if it changes what happens next — wire thresholds to actions.
- Recalibrate quarterly against actual closed-won data.
Every sales team eventually drowns in leads that all look equally urgent. Lead scoring exists to answer one question fast: who deserves the next hour of a rep's attention? Done well, it quietly compounds — hot leads get called while they're hot. Done badly, it's a number nobody trusts, decorating a CRM field nobody reads.
Two dimensions, kept separate
The classic mistake is mashing everything into one number. Keep two: fit — how closely they match your ICP (right industry, size, role) — and engagement — what they've actually done (opened, replied, visited pricing, attended a webinar). A perfectly-fitting company that's never engaged needs nurture; a poor-fit lead clicking everything needs a polite pass. Only high-fit + high-engagement deserves a same-day call. One blended number hides exactly this distinction.
Weight behaviors by intent, ruthlessly
- Pricing page visit, demo request, reply to outreach: heavy points — these are hand-raises
- Case study reads, repeat site visits, webinar attendance: moderate
- Email opens, single blog visits: near zero — curiosity, not intent
- Negative scoring: unsubscribes, careers-page visits, student email domains, competitor domains — subtract hard
Negative scoring is the underused half. A model with no way to say 'this lead got worse' inflates forever, and inflated scores are why reps stop believing.
Wire scores to actions, not dashboards
A score that doesn't trigger anything is trivia. Set thresholds that do work: above 80, create a task and alert the owner within the hour; 50–80, drop into an active nurture sequence; below 50, quarterly touch at most. This is where scoring earns its keep — in BixJet, score thresholds can trigger sequences and tasks automatically, so the hot lead gets touched while the signal is fresh rather than at Friday's pipeline review.
Calibrate against reality
Quarterly, pull the last 90 days of closed-won and closed-lost and ask: did our scores predict this? If half your wins scored under 40, your weights are fiction — fix them with what the data says actually predicted revenue. And keep the model explainable: when a rep asks why this lead is an 85, there should be a two-line answer. Reps follow scores they understand; they audit ones they don't, and then ignore them.