Lead Scoring Without a Data Science Team
By CodexierPublished 5 min read
Lead scoring sounds like something that needs data scientists and predictive models. For most small and mid-size Swedish companies it does not. A transparent points model, built from what you already know about good customers and checked against deals you have actually won, gets you most of the value. It tells sales who to call first and marketing which leads need more nurturing. This guide shows how to build one in your CRM.
Fit and behaviour signals
Fit is about who the lead is: industry, company size, role, location. Behaviour is about what they have done: visited the pricing page, booked a demo, opened emails, downloaded a guide. A perfect fit who shows no interest is not ready. A very engaged lead who is a student writing a thesis is not a customer. You need both dimensions.
| Fit signal | Behaviour signal |
|---|---|
| Industry you serve well | Visited the pricing page |
| Company size in your sweet spot | Requested a quote or booked a call |
| Decision-making role | Returned to the site several times in a week |
| Located in your service area | Opened and clicked recent emails |
| Uses a system you integrate with | Downloaded a buying guide or checklist |
A simple points model
Give each signal points based on how strongly you believe it predicts a deal. Keep numbers coarse, such as 5, 10 and 20, because false precision gives a false sense of accuracy. Add negative points for signals that disqualify: a competitor's email domain, a student address, a country you do not serve, or an unsubscribe. Let behaviour points decay over time, so a visit three months ago counts less than one yesterday.
- Hand-raiser actions such as booking a call or requesting a quote get the most points; many teams route them straight to sales regardless of score.
- Pricing and service page visits score higher than blog visits.
- Fit signals come from form fields or data enrichment; ask only what you will actually use.
- HubSpot, Pipedrive and most CRMs let you build this with properties and workflows, without code.
Checking against won deals
A model built on opinions needs to meet reality. Export your last twenty to fifty closed deals, both won and lost, and score them retroactively with your model. If won deals do not clearly score higher than lost ones, the points are wrong. Look at which signals the won deals had in common and adjust. This takes an afternoon in a spreadsheet and is the step most teams skip.
Won deals score high
The model separates good leads from bad. Keep it and set the threshold.
No clear difference
Your signals do not predict. Revisit which fit criteria really matter.
Lost deals score high
You are rewarding the wrong behaviour, often content consumption by people who never buy.
Hand-off thresholds to sales
Scoring only helps if it changes what people do. Agree between marketing and sales what score makes a lead sales-ready, what sales commits to do with it and how fast, and what happens to leads below the threshold. Write it down. Automate the hand-off: a task or notification to the right salesperson when the threshold is crossed. Our guide to getting website leads into the CRM covers the plumbing.
- Above threshold: assigned to a salesperson, contacted within an agreed time.
- Near threshold: nurtured with relevant email content.
- Below threshold or poor fit: stays in marketing, no sales time spent.
- Rejected by sales: returned with a reason, so the model can learn.
Reviewing the model quarterly
Markets and offers change, and so do the signals that predict a deal. Once a quarter, re-run the won-and-lost check, look at leads sales rejected and why, and adjust points. Remove signals nobody can explain. Keep the model simple enough that a salesperson can understand why a lead scored as it did; a black box nobody trusts gets ignored.
We set up lead scoring as part of a CRM implementation. When you do not need it: if you get a handful of leads a week, sales can simply look at each one, and a scoring model adds admin without value. It becomes useful when volume makes prioritising hard. A CRM health check is often the better first step. To talk through your lead volume, book a free call.
Frequently asked questions
Do we need predictive lead scoring?
Not at the start. Predictive models need a lot of historical deal data to be reliable. A transparent points model works well for most small and mid-size companies and is easier to trust and adjust.
Can we use lead scoring in HubSpot's free CRM?
Advanced scoring features sit in HubSpot's paid tiers. On the free tier you can approximate it with properties and manual rules, or use a simpler CRM feature set. Check the current plan contents before choosing.
Is lead scoring allowed under GDPR?
Generally yes for B2B, based on legitimate interest, provided you inform people in your privacy policy and do not make decisions with legal effects purely automatically. Keep tracking consent rules in mind for website behaviour.
How many signals should the model have?
Start with five to ten. More signals make the model harder to understand without making it more accurate. Add signals only when the won-deal check shows a gap.
Too many leads to call them all?
Tell us your CRM and lead volume. In fifteen minutes we will sketch a scoring model you can check against your own deals.
Book a free 15-minute call