Design a lead scoring model that predicts customers
A lead score is a number that says how likely a person is to buy, so sales can call the likeliest first. A good model is built from what your customers actually had in common before they bought, decays with silence, and is judged by one test: do high scores become customers more often than low ones? This article gives the standard model shape, starting points for the points, the setup in each tool, and how MarqIQ tells you whether your score sorts buyers from the rest. For your own score’s verdict, open the Lead scoring analysis.
Two scores, added at the end
Keep fit and engagement as separate scores, then combine them. Fit rarely changes; engagement changes weekly. A single blended number hides which one moved, and you cannot tune what you cannot see.
| Score | Built from | Typical range | Changes |
|---|---|---|---|
| Fit | Industry, company size, region, job title, technology in use | 0 to 50 | When the profile is enriched or corrected |
| Engagement | Forms, page views, email clicks, webinar attendance, replies | 0 to 50, decaying | Every action; minus points for silence |
| Total | Fit + engagement | 0 to 100 | The MQL threshold sits on this one |
Starting points for the points
- Fit: +15 for a target industry, +15 for company size in range, +10 for a buying title, +10 for a target region. Zero, not minus, for unknown — a blank field is not a bad fit, it is a missing fact.
- Engagement, strong: +20 demo or contact request, +15 pricing page visit, +10 webinar attended. Medium: +5 content download, +5 email click, +3 repeat visit. Weak: +1 email open (opens are inflated by mail apps; keep them near zero).
- Negative: −5 personal mailbox, −10 job title contains student or intern, −50 competitor domain, −25 unsubscribed.
- Decay: −10 after 30 days with no engagement, −20 after 60, back to zero at 90. A hot score from last quarter is not a hot lead.
- Threshold: pick the score above which sales accepts most leads, not a round number. Start at 60 and move it after the first month of accept and reject data.
The one test that matters
Take everyone who became a customer in the last year and everyone who did not. If the customers’ scores were not clearly higher before they bought, the model is decoration. This needs enough customers to judge — MarqIQ asks for 30 — and it is why the score must be stamped before the deal, not after.
Set it up in your tool
Pull the traits of your last 50 customers before you write a point value.
Industry, size, title, the forms they filled, the pages they read. Points come from what buyers had in common, not from what marketing wishes mattered.
Create the two score fields and the total.
HubSpot: two Score properties plus a calculated total, or the built-in HubSpot Score for engagement. Marketo: Behavior Score, Demographic Score and Lead Score fed by scoring smart campaigns. Salesforce: number fields updated by flows, or Einstein Lead Scoring. Pipedrive: a custom number field kept by an automation.
Write the decay rule.
A scheduled workflow that subtracts points from engagement after 30, 60 and 90 quiet days.
Tie the threshold to the MQL rule.
Reaching the total threshold is one of the ways a person becomes an MQL — see Define an MQL.
Review monthly against acceptance and wins.
Raise points that appear in accepted and won leads; cut points that appear in rejected ones.
How MarqIQ checks it
The Lead scoring analysis reads every field with a score in its name, judges each on its values, and tells you whether customers scored higher than non-customers — or that there are too few customers to say yet, and which score fields exist but were never switched on. Ask “is our lead scoring working” or “how many leads scored above 50” for your own numbers.
Next steps
Related articles
Was this article helpful?