How lead scoring works
Score fit and behaviour separately, then read them as a grid rather than a single total. A perfect-fit lead who has done nothing needs outbound; a highly engaged poor-fit lead needs disqualifying, not a call.
| Input | Example signals | Weight guidance |
|---|---|---|
| Fit | Industry, employee count, revenue band, region, job title | Fixed attributes; set once from closed-won analysis |
| Behaviour | Pricing page views, demo request, email replies, repeat visits | Decays over time; recent activity should outweigh old |
| Negative | Free-mail domain, competitor, student, unsupported region | Subtract aggressively to keep the top of the list clean |
How to build your first scoring model
You do not need machine learning to start. Build the first version from your own closed-won data.
- Export the last 100 to 200 closed-won and closed-lost deals.
- Find the attributes that appear far more often in won deals than lost ones.
- Assign points to those attributes; total them to roughly 100 for a perfect lead.
- Set the qualification threshold where roughly the top 20 to 30 percent of leads land.
- Review the threshold monthly against actual conversion and adjust.
When AI lead scoring helps
Rule-based scoring reflects what you already believe. A model trained on your historical outcomes finds patterns you did not encode — combinations of source, timing and engagement that correlate with closing. AI scoring is worth adding once you have a few hundred closed deals; below that, a hand-built model is more reliable and far easier to explain to the team.
Alegria RevOps scores leads from synced CRM data and shows the reasoning behind each score, so reps can see why a lead is ranked highly rather than being asked to trust a number.
Prioritisation and follow-up speed
Scoring is worthless if the top-scored lead waits three hours for contact. Pair scoring with routing rules that assign an owner instantly and escalate anything above the threshold that has not been touched within your response target.