Lead scoring in B2B: a one-page model, the seven factors, and when it is worth it
Tips
The short answer
Lead scoring is giving leads points to decide whom you call first. In B2B it only works in this order: first points for the fit of the company (sector, size, region, can it pay), then points for the trigger and the behaviour (a vacancy, a new director, a visit to your pricing page, a demo request). A model with those two axes and three thresholds fits on one page and beats any more elaborate system, because your team understands it and keeps it up. Lead scoring is worth it as soon as you have more leads than you can call in a day. Below that, qualify by hand, which is covered in lead qualification.
What lead scoring is, and what it is not
Lead scoring puts a number on every lead, built from characteristics of the company and behaviour of the person, so that a list sorts itself. It is not qualification: qualifying is a conversation in which you answer four questions (does the company fit, are you talking to the right person, is there a trigger, can they pay). Scoring only decides who gets that conversation first. And it is not a prediction. A lead with 85 points can cancel tomorrow; a lead with 30 points can become the biggest customer of the year. The score allocates time, not outcomes.
The difference with "AI lead scoring" is smaller than the term suggests. A model learns the points from your closed deals instead of you choosing them, and adjusts them as more data arrives. That is useful from a few hundred leads a month. Below that, a model you write yourself is more transparent and just as good.
A model that fits on one page
Two axes, each up to 50 points, and three thresholds.
Axis | What you score | Source |
|---|---|---|
Fit (0 to 50) | Sector, size, region, legal form, credit score | Registers and your CRM |
Trigger and behaviour (0 to 50) | Vacancy, director change, move, growth, website visit, demo request, reply to contact | Signals, website, email |
Thresholds: 70 or more is A, call today. 40 to 69 is B, keep warm and wait for a trigger. Below 40 is C, park it. Subtract points for what makes a lead unsuitable: a visit to your careers page (looking for a job), a student address, a competitor, a company outside your market. And let points decay: a website visit from three months ago no longer counts. Without decay every old lead becomes an A by itself.
The seven factors worth scoring
Fit with your ideal customer profile. Sector, size, region and legal form. This is the gate: if the company does not fit, the score stays low whatever the person does. How to derive that profile from your best customers is in building an ideal customer profile.
Ability to pay. A credit score or the latest annual accounts. A company that cannot pay is not a lead, however warm. In Belgium and the Netherlands this is free to check; how is in how to run a company credit check.
A trigger from the registers. A vacancy in the department that uses your product, a new director, a move, more staff in the accounts. This is the factor most models miss and the one that predicts most in B2B: something is changing at the company. Which triggers exist and where to find them is in finding new customers per trigger.
Behaviour on your website. A pricing page weighs more than a blog article, a demo request most of all. Repetition counts: three visits to the same solution page in a week is a pattern, one visit is chance. Which companies visit you see through visitor identification; the person stays anonymous there.
Response to contact. A reply to an email, a returned call, an accepted invitation. The strongest behavioural factor, and the only one that says the person themselves wants to talk.
Recency. Everything above with a date, and with decay. A trigger from this week is gold; the same trigger from six months ago is history.
Relationship. An existing customer, a former customer, a referral. Give points for it, but score the relationship separately from fit: a friendly former customer in the wrong sector is still the wrong sector.
When lead scoring makes sense, and when it does not
Yes when your team gets more leads each week than it can call, when marketing and sales argue about "good leads", or when leads come from several sources (website, signals, bought lists) that you cannot compare on one measure.
No when you have fewer than fifty leads a month: you qualify them by hand in less time than the model costs. No when you have no definition of a good customer yet: the model then scores your assumptions. And no when nobody maintains it; a model untouched for a year sorts for last year's market.
How to set it up in a week
Monday: the fit. Take your twenty best customers and write down what they share: sector, size, region, legal form. Those are your fit criteria and their points.
Tuesday: the triggers. For every deal closed in the last year, look at what preceded the first conversation. A vacancy? A website visit? A referral? Give the recurring triggers the most points.
Wednesday: the test. Run the model over last quarter's leads. Do the closed deals score high and the lost leads low? If not, the error is in the points, not in the leads.
Thursday: thresholds and decay. Set A, B and C so that A matches exactly what sales can call in a week. Set the decay.
Friday: automate. Put the fit criteria and behaviour points into your CRM. In Bizzy, fit (register data, credit score) and triggers (vacancies, website visits, job changes) already sit on the same company record, and the lead-generation agent learns from which leads you accept and reject, so the sorting moves with your feedback.
Where it goes wrong
Scoring behaviour without fit. The most active visitor on your site is often a student, a competitor or a job applicant.
Points that never decay. After a year every lead is an A and the model sorts nothing.
Nobody calibrates. Compare the scores with closed deals every quarter and adjust the points.
The score replaces the conversation. An A lead is a lead you call today, not a lead that is already qualified. Qualifying stays human work.
Scoring only earns its place as the hand-back rule inside a nurturing flow. Our lead nurturing guide shows where that rule sits, and lead management the process around it.
Scoring ranks leads; a conversation qualifies them. The BANT method covers the second step.
Frequently asked questions
What is lead scoring? Assigning points to leads based on company characteristics and behaviour, so your team knows whom to call first. In B2B: first the fit of the company, then the trigger and the behaviour of the person.
What is a lead scoring model? The list of criteria with their points, the thresholds that separate A, B and C leads, and the rules for deductions and decay. A good model fits on one page and is checked against closed deals every quarter.
How does AI lead scoring work? A model learns the points from your closed and lost deals instead of you choosing them, and adjusts them as new data arrives. It pays off from a few hundred leads a month; below that a hand-written model is more transparent and just as good.
What is the difference between lead scoring and lead qualification? Scoring sorts the list with a number; qualification is the conversation in which you establish that the company fits, you are talking to the right person, there is a trigger and they can pay. Scoring sets the order, qualification the outcome.
Which tools do lead scoring? Most CRMs (HubSpot, Salesforce, Pipedrive) score on behaviour and fields you fill yourself. For the fit and trigger side you need company data and signals; Bizzy supplies those for Belgium and the Netherlands on the company record, with an agent that learns from your feedback.
Read next: lead qualification, what is a lead, B2B leads in the Netherlands and the B2B sales funnel.