Lead scoring

How lead scoring works

Lead scoring gives every lead a number so a team can sort a queue instead of guessing. Salesforce Trailhead describes it as a way to measure a lead's interest so time goes to the leads most likely to convert 2.

Most models combine two kinds of evidence 1.

• Fit asks whether the lead looks like a good customer. HubSpot builds fit scores from property values such as job title, company size or annual revenue 1.
• Engagement asks whether the lead is showing interest. HubSpot builds engagement scores from actions such as visiting the website, subscribing to a newsletter, clicking a CTA or opening a marketing email 1.

Fit usually draws on firmographics and the role of the person, which is why it ties back to the ideal customer profile. Engagement draws on marketing activity and, in some stacks, third-party intent data.

Points come from rules. In a Trailhead example, an automation rule adds 50 points when a prospect requests a demo 2. HubSpot lets teams set a maximum total score and optional maximums per group of criteria 1. That lets awareness events such as page visits earn limited points while conversion events such as sales form submissions earn more 1.

Scores also need to fall. Interest from half a year ago says little about today. HubSpot's score decay reduces an event's points based on how long ago it happened, on intervals of 1, 3, 6 or 12 months 1. Trailhead calls adjusting for score decay as important as earning points 2.

A threshold turns the score into action. When a lead crosses it, the lead becomes a marketing qualified lead or enters a lead routing rule that assigns an owner. The threshold should come from conversion data rather than a round number someone liked.

Rule-based vs predictive lead scoring

Rule-based scoring is transparent. Someone decides what each action and attribute is worth, and anyone can read the rules. A Trailhead example adds 10 points for a pricing page view and subtracts 10 points for a jobs page view 2. The weights stay guesses until someone checks them against closed deals.

Predictive scoring learns the weights. Einstein Lead Scoring analyzes historical field data on the Lead object, finds patterns in past conversions and scores current leads by how well they match 3. Each score comes with insights about which fields affect it most 3. HubSpot offers AI-built engagement or fit scores for contacts on its Marketing Hub Enterprise tier 1.

Predictive models need history. A team with few closed deals gives a model little to learn from, so rules usually come first. Both approaches share a weak spot. A fit score built on missing fields rates the lead as a poor fit when the real answer is unknown, and good accounts with empty industry or size fields drift down the queue.

Filling company fields before scoring makes the fit half honest. Some fit rules cannot live in a field at all, such as whether a company sells into regulated healthcare. A model can read enriched company facts and return a structured verdict for those rules 4. The RichAPI ICP playbook requires a third answer, unknown, and sends it to a person rather than treating it as a no 4. That keeps a gap in the data from turning into a low score.

Example of a lead scoring model

A B2B software company runs a simple two-part model with caps on each part.

Fit (max 50)
  Company in target industry               +15
  Headcount 51 to 500                      +15
  Title matches a buyer persona            +15
  Region the team sells to                 +5
Engagement (max 50, decays 50 percent a month)
  Demo request                             +25
  Pricing page view                        +10
  Webinar attended                         +10
  Marketing email click                    +5
  Jobs page view                           -10
Route to sales at 60 or more

Fit and engagement each carry a cap, the same pattern HubSpot supports with group limits 1. The engagement points decay, so last quarter's demo request no longer pushes a lead over the line on its own. HubSpot's own decay example halves a 10 point event to 5 points after a month at 50% decay 1.

Fit points depend on company fields, and inbound forms rarely collect them. The team enriches each new lead's company first. On RichAPI, enrich_company takes a LinkedIn company page URL and returns industry, headcount, specialties and description at 1 credit per call 4. One fit rule, whether the company sells into regulated healthcare, goes to ai_enrich at 2 credits per call with a yes, no or unknown schema 4. A yes adds the points. An unknown sends the lead to a person instead of scoring it as a poor fit 4. Both calls bill on any 2xx response, including an unknown verdict, so the team skips the second call when the enrichment payload is empty 4.

Every quarter the team compares score bands with closed deals. If leads below the threshold convert as well as leads above it, the threshold moves down.

Lead scoring vs lead qualification

Lead scoring ranks leads with a number. Lead qualification is the decision about whether a lead is ready for sales time. A score is one input to that decision, alongside conversations and frameworks that check budget, authority, need and timing.