Key takeaways
- Good lead scoring combines fit (is this the right company and person?) with engagement (are they showing buying intent?).
- Build the model with sales, using won and lost deals, not marketing intuition alone.
- Scores should decay over time so old activity does not keep leads artificially hot.
- Review the model quarterly against conversion data and adjust.
Lead scoring promises that sales will spend time on the leads most likely to buy. In practice, many scoring models are ignored because reps do not trust them. The difference between a trusted model and an ignored one is usually how it was built.
Two kinds of signals
| Fit (who they are) | Engagement (what they do) |
|---|---|
| Industry, company size, region | Pricing or demo page visits |
| Job title and seniority | Content downloads and webinar attendance |
| Technology they use | Email clicks and replies |
| Budget or revenue band | Return visits within a short period |
A lead with perfect fit and no engagement is a good target for marketing. A highly engaged lead with poor fit may be a student or a competitor. Prioritize leads that score well on both.
Build the model from real data
- Pull closed-won and closed-lost opportunities from the past year.
- Compare the attributes and behaviors of won deals against lost deals and unqualified leads.
- Weight the signals that genuinely differ, and give negative points to disqualifiers such as personal email domains or student titles.
- Agree a threshold for a marketing-qualified lead with sales, and a service level for follow-up.
Add decay
Engagement from six months ago means far less than engagement this week. Reduce engagement points over time, or score only activity within a recent window, so the score reflects current intent.
Running it in Salesforce
- Simple models can be built with formula fields or record-triggered flows that add points for fit attributes.
- Marketing Cloud Account Engagement (formerly Pardot) provides scoring and grading based on prospect activity and fit.
- Einstein-based scoring, where licensed, uses your historical conversion data to predict which leads are likely to convert.
Running it in HubSpot
HubSpot includes lead scoring tools in its paid tiers, letting you add or subtract points based on properties and behaviors and use the score in workflows, lists and lead routing. If HubSpot and Salesforce are connected, sync the score to Salesforce so sales sees it on the lead.
Make it visible and actionable
- Show the score and the reasons behind it on the lead record.
- Route high-scoring leads automatically and alert the owner.
- Build a dashboard comparing conversion rates by score band, and revisit the model every quarter.
An example scoring model
| Signal | Points |
|---|---|
| Company size in target range | +15 |
| Industry in target list | +10 |
| Decision-maker title (director and above) | +15 |
| Visited pricing page | +10 |
| Requested a demo | +30 |
| Clicked emails three times in 30 days | +5 |
| Personal email domain | −10 |
| Student, job seeker or competitor | −50 |
This is illustrative, not a recommendation; your weights should come from your own conversion data. Notice that a demo request outweighs many small engagement signals, and that disqualifiers carry strong negative weight so they cannot be offset by enthusiasm.
Common mistakes
- Scoring email opens heavily, even though privacy features in some email clients make open tracking unreliable.
- Building the model without sales input, so reps do not trust it.
- Never revisiting weights after products, markets or campaigns change.
- Using one model for very different products or segments.
- Routing leads by score without a follow-up service level, so hot leads still wait.
Lead scoring and lead grading
Some teams separate the two ideas: a grade (A to F) for fit and a score for engagement. An A-grade lead with a low score is a nurture target; a high-score lead with a poor grade may need qualification before sales time is spent. Account Engagement in Salesforce supports this split natively, and the same logic can be built in HubSpot with properties and workflows.
Frequently asked questions
What is a good lead score threshold?
There is no universal number. Set it where conversion rates rise noticeably, based on your own data, and adjust as you learn.
Should we use predictive lead scoring?
Predictive scoring helps when you have enough historical conversions. With little data, start with a simple rules-based model.
Why does sales ignore our lead scores?
Usually because the model was not built with them, scores are not explained, or high-scoring leads have not converted. Involve sales and show results.
