Lead scoring assigns points to leads so you know who deserves attention first. It uses two kinds of signals: who someone is (profile) and what someone does (behaviour). The goal is not a neat number but a better allocation of limited follow-up time.
This page explains how to build a workable model, how to define MQL and SQL, and why many models deliver nothing in practice. For wider context, see lead generation and CRM.
Two kinds of signals
Explicit (profile): job title, industry, company size, country, whether the company is on your target list. This tells you whether someone fits.
Implicit (behaviour): which pages were viewed, a pricing or services page more than once, a demo request, email clicks, repeat visits. This tells you whether someone is interested.
You need both. Someone who fits perfectly but does nothing is not a near-term opportunity; someone very active who doesn't fit only costs time.
Building a simple model
- Define your ideal customer profile, based on your existing best customers rather than wishful thinking.
- Pick five to eight signals, not thirty. More signals create false precision.
- Assign points from closed deals. Look back: which attributes and behaviours recurred in deals you won?
- Subtract points for disqualifying signals: wrong country, competitor, job applicant, free email domain on a business offer, careers page as the only visit.
- Set thresholds where a lead moves to sales, and agree what happens then.
- Let scores decay. Behaviour from three months ago says little; without decay everyone eventually scores high.
| Signal | Type | Example value |
|---|---|---|
| Role matches decision maker | Explicit | +15 |
| Company size in target range | Explicit | +10 |
| Pricing or services page viewed twice | Implicit | +15 |
| Contact form submitted | Implicit | +30 |
| Only visited careers page | Implicit | -20 |
| Outside your market | Explicit | -30 |
These values are an example; your model should come from your own data.
MQL, SQL and the hand-off
An MQL is a lead ready for contact by marketing's criteria. An SQL is a lead sales has accepted. The hand-off is where most models die: without an agreement on how fast contact happens, who does it and what happens on rejection, scoring stays a dashboard.
So agree explicitly: which score hands over, with which context, how quickly, and on what grounds sales may send it back. That feedback is the most valuable input for improving the model.
Where it goes wrong
- Invented points. Without looking back at closed deals you measure your assumptions.
- Too many signals. Complexity makes the model opaque and unreliable.
- No decay. Old behaviour keeps counting, so everything looks warm.
- A score without an owner. If nobody acts on the list, nothing changes.
- Counting only forms. Also weigh signals like repeat visits or company identification through Leadinfo.
- No quality feedback. Without an outcome per lead you can never validate the score — the same applies to ad platforms, see smart bidding.
Frequently asked questions about lead scoring
Can I start without tooling?
Yes. Three categories (call now, nurture, ignore) in your CRM already help.
How much data do you need?
Enough won and lost deals to see patterns. At low volume, manual judgement beats a model.
Does lead scoring work in B2C?
Yes, but mostly on behaviour and value: purchase history, CLV and recency.
How is it different from lead qualification?
Scoring is ranking at a distance; qualification is the conversation where you establish whether there is a real opportunity.
Further reading
See online leads, Leadinfo and CRM, or spar via /en/book-call.
Questions, or just want to spar?
We're happy to think along — call, email or drop by in the heart of Eindhoven.