Lead Score
A numerical value assigned to a lead that represents how likely they are to convert, based on a combination of behavioral and profile signals.
A lead score is a numerical value assigned to a lead that represents how likely they are to convert, based on a combination of behavioral signals (what they’ve done) and profile signals (who they are). The score is the input to most B2B sales prioritization decisions: which leads sales calls first, which leads enter outbound sequences, which leads are flagged for AE attention.
In modern B2B stacks, lead scores live as a property on the contact record in CRM (HubSpot, Salesforce, Pipedrive). The score updates automatically as new events fire — a page view adds points, a demo request adds more, time without engagement subtracts via decay.
How lead scoring works
Most lead scoring systems combine four kinds of signals:
- Engagement — recent behavior: page views, content downloads, email opens, meeting attendance, demo requests
- Profile fit — who the person is: title, seniority, function, decision authority
- Account fit — what company they’re at: industry, size, tech stack, geography
- Deal context — current sales motion: open opportunities, last touch, pipeline stage
Each signal has a weight. The total score is the sum (or weighted average) across all signals. A lead crossing a threshold (often 50, 75, or 100 points) becomes an MQL and gets handed to sales.
Some systems also apply decay: a lead that did a high-value action a year ago shouldn’t score the same as a lead that did it last week. Without decay, scores accumulate forever and your “hot” list silently fills with cold leads.
What makes a lead score trustworthy
The single biggest predictor of whether sales will use a lead score is whether they can defend it in a forecast meeting. “This lead scored 87” is useless if the rep can’t answer “why?” Three things make a score trustworthy:
- Per-event audit trail. Every score change is logged with the event, the points it added, and a human-readable reason.
- Authored rules, not auto-AI. RevOps writes the rules in plain logic. The rep can read them, sales leadership can debate them, and anyone can change them when the model drifts.
- Calibration against closed-won data. The model is checked quarterly against actual conversion rates by score band. If the top decile doesn’t convert at higher rates than the bottom decile, the model is broken.
Common lead score failure modes
- Black box. The score is a number with no explanation. Sales ignores it.
- No decay. Stale signals stay hot forever. Pricing-page views from 6 months ago still count.
- Weights set once and forgotten. Calibration drifts. A score that worked for last year’s ICP doesn’t work for this year’s.
- One-dimensional. A single number can’t represent “high engagement, low fit” vs “low engagement, high fit.” A multi-dimensional model treats those as different plays.
Lead score vs lead grade
Some systems combine a numerical lead score (engagement) with a letter grade (A/B/C/D for fit). The grade is a coarse summary of profile match; the score is a fine-grained measure of activity. Salesforce’s legacy Pardot grading model is the canonical example. HubSpot’s Modern Lead Scoring (Aug 2025) treats them as separate properties: a numerical Fit score and a numerical Engagement score, optionally combined into a single value on Enterprise.
Related at kenbun
kenbun ships rules-based lead scoring with per-event audit trails and per-event half-life decay. Every score change is explainable; every rule is authored by your team; every event ages at its own rate. See the 4-dimension framework for the full model.