The short answer
Qualify B2B leads on four signals: fit (do they match your ICP on size, sector, and geography), intent (what did they actually do, not what they opened), authority (can this person start a buying process), and timing (is there a dated event forcing a decision). Score fit and intent separately, route only leads that clear both thresholds to sales, and nurture the rest. Reviewing the thresholds every quarter against closed-won data is what keeps the model honest.
Why most scoring models fail
The typical model awards points for email opens, page views, and form fills. All three measure curiosity. None of them measure whether a company can sign a contract. That is why sales stops trusting the score within a quarter and goes back to working the list by gut feel.
A model earns trust when the leads it flags close at a visibly higher rate than the ones it does not. That is the only success criterion that matters.
The four signals worth scoring
Score fit and intent on separate axes. Collapsing them into one number hides which half is broken.
- Fit — company size, sector, geography, and tech stack against your actual closed-won profile, not your aspirational one.
- Intent — pricing page visits, demo requests, repeat sessions from multiple people at the same company, competitor comparison reads.
- Authority — job title plus whether they referenced a budget, a team, or an existing vendor.
- Timing — a contract renewal, funding round, new hire, or stated deadline. Without timing, high-fit leads sit for months.
A simple scoring model you can run this week
Give fit a score out of 50 and intent a score out of 50. Route anything above 35 on both straight to sales with a same-day SLA. High fit and low intent goes to nurture. High intent and low fit goes to a self-serve path or a polite no.
Keep the model to eight or fewer inputs. Every extra input adds maintenance cost and reduces the chance anyone keeps it current.
How to settle the lead-quality argument for good
Once a quarter, pull every closed-won deal from the last twelve months and check what score it had at creation. If your winners were scoring low, the model is wrong — not the sales team.
Publish that review to both teams. The argument stops being an opinion contest the moment it becomes a table of numbers.
Frequently asked
Should we use MQL and SQL labels?
Only if both teams agree on the definition in writing and the handoff has a time limit. Otherwise the labels just move the argument rather than resolving it.
How many inputs should a scoring model have?
Between five and eight. Beyond that, nobody maintains it and the score drifts away from reality within a couple of quarters.
Can AI do the qualification?
It can read behaviour, enrich firmographics, and rank leads far faster than a person — but a human still sets the thresholds and owns the quarterly review against closed-won data.
Free GTM Diagnosis
Want this run on your pipeline?
Send two details. We come back with the two or three gaps costing you the most revenue.
No spam. No sales sequence. One reply from a human.
Prefer to talk? Book a 30-minute call