Behavioral signal scoring ranks accounts and people by what they do, firmographic scoring ranks them by what they are, and close rates show whether either works. Neither score proves itself: sort resolved leads into fit-by-behavior cells, compute each cell's closed-won rate with its sample size, and trust a difference between cells only when the intervals do not overlap.
Behavioral Signal Scoring, Firmographic Scoring and Close Rates: The Short Answer
- Firmographic scoring describes the account; behavioral scoring describes activity. Firmographic inputs are company attributes such as industry, employee range, revenue, region and technology in use. Behavioral inputs are actions such as a pricing-page visit or a demo request.
- The pages that rank give point values and thresholds, and no closed-won rates by score. Where they give outcome figures, the figures are conversions to MQL or opportunity, not closed deals.
- Each input needs a passport. The Input Passport records an input's source, the date it was observed, how it was verified and the confidence it deserves, so a stale company field does not look like fresh evidence.
- Each score needs a ledger. The Score Cell Ledger counts resolved leads in each fit-by-behavior cell and reports the close rate with an interval.
- A score becomes useful when a cell has an action. The ledger ends in a rule for who works the lead, what they check first and what they send back.
How the Pages That Rank Handle Scoring
The results for this phrase are glossaries and guides from marketing-technology and data vendors. They define the two scoring types well. They differ in how they set weights and in what they offer as proof.
| Source | What it answers | Numbers and claims it gives | What it leaves open |
|---|---|---|---|
| CDP.com, AI Lead Scoring (September 19, 2026) | Machine-learning scoring trained on historical conversions, compared with rule-based scoring | Credits Forrester with a 20% sales-productivity increase and a 30% acquisition-cost reduction; says a model needs about 1,000 or more conversions | No link to the Forrester source; no firmographic input table |
| The Pedowitz Group, What is behavioral scoring | Weights tied to pipeline correlation, recency decay, multi-signal patterns, validation by score band | Process guidance; no figures | No example of a score-band table |
| Saber, Behavioral Lead Scoring | Point tables, decay (75% of points for days 31 to 60, 50% for days 61 to 90) and example programmes | Credits Forrester with 20 to 30 percent higher lead-to-opportunity conversion for behavioral plus firmographic scoring | No link to the Forrester source; examples are unnamed companies |
| Plura AI, Automated Lead Scoring Examples (May 22, 2026) | Six worked point models; says a common starting allocation is 40 points for demographic and firmographic fit and 60 for behavior | Point tables and thresholds | The 40/60 split has no stated source or outcome |
| ZoomInfo, Lead Scoring: A Step-by-Step Guide | Fit, engagement and intent attributes; a starting MQL threshold of 50 to 70 on a 0 to 100 scale; test for 60 to 90 days | A customer result of an 84% MQL increase and a 26% jump in opportunity rates after connecting scoring to ZoomInfo data | Both figures measure MQLs and opportunities, not closed-won; the customer's baseline is not given |
| HubSpot Knowledge Base, lead scoring tool (performance reports last updated August 14, 2026) | Engagement, fit and combined scores; score and group limits; three performance reports | Score distribution and averages; combined scores grouped into six labels by fit and engagement | The performance reports list scored records, distribution and scores over time; closed-won by band has to come from your own outcome data |
The gap is the same on every page. They tell a team to validate against conversions and none shows what a validation table looks like or how many records it needs. The two Forrester figures appear on two pages without a link, so this guide does not repeat them as findings. The Score Cell Ledger below is the missing table.
Definitions: What Each Score Measures
- Firmographic score. Points for company attributes: industry, employee range, revenue, geography and technology in use. HubSpot's fit score also uses person properties such as job title, which other sources call demographic. This guide keeps the two apart: a firmographic field describes a company, and a demographic field describes a person.
- Behavioral score. Points for actions by the account or its people: site visits, content downloads, demo requests, replies. HubSpot calls this an engagement score and lists visiting a website, subscribing to a newsletter, clicking a call to action and opening an email.
- Signal. A dated event at the account, such as a hire or a funding round. A signal can feed either score depending on whether you treat it as a fit attribute or as an action. The buying signals vs intent data guide separates events that were observed from behavior that a vendor reports or infers.
- Close rate. Closed-won divided by a stated denominator. Two denominators are in use and they give different numbers: leads that reached a resolved outcome (lead-to-won) or opportunities created (opportunity-to-won). A close rate without its denominator cannot be compared.
The Input Passport
The Input Passport is a record kept with every scoring input. It has four fields: source, observed date, verification and confidence. Without it, a score mixes a field verified last week with one copied from a form two years ago, and the number looks the same.
| Input | Type | Typical source | Freshness rule (recommendation) | Verification | Confidence in this guide's view |
|---|---|---|---|---|---|
| Employee range | Reported | Data vendor or company page | Re-check when the record is older than the period you set | Compare two sources | Partial: ranges are coarse |
| Industry | Reported | Data vendor or registry code | Re-check on a change of company name or ownership | Compare with the company's own description | Partial |
| Revenue | Reported or inferred | Filing for public companies; vendor estimate otherwise | Filing date for a filing; vendor date otherwise | Filing is a primary source; an estimate is not | Strong for a filing; weak for an estimate |
| Technology in use | Inferred | Technographic vendor detection | Detection date | Check a second source or the company's own pages | Partial |
| Job title | Reported | Contact record | Date last verified | Check against the company's page or a current posting | Partial |
| Pricing-page visit | Observed (first party) | Your analytics | Visit date | Your own log | Strong that a visit happened; it does not show who decides |
| Demo request | Observed (first party) | Your form | Submission date | Your own log | Strong |
| Email open | Reported | Email platform | Event date | Platform record | Weak: it records a message event, not a decision |
| Third-party topic surge | Inferred | Intent feed | Feed period | Combine with an observed event | Weak |
| Funding, hire or posted role | Observed | Public record | Record date | Open the source | Strong that the event happened; weak on what it means for a purchase |
"Observed", "reported" and "inferred" are used as in the buying signals vs intent data guide. The confidence column is this guide's judgment and not a measured accuracy rate. The best B2B contact database comparison covers how databases differ in the freshness of person records, and the predictive lead scoring guide covers the model side.
From Score to Sales Action
A score is stored as a row until a rule says who acts on it. Splitting fit and behavior into a two-by-two grid gives each cell an action. HubSpot's combined scores use the same idea, grouping records into six labels by level of fit against level of engagement.
| Cell | What it suggests | First action | Owner |
|---|---|---|---|
| High fit, high behavior | An account that matches and is active | Seller reviews the account today and accepts or returns it with a reason | Seller |
| High fit, low behavior | A match with no current activity | Research the buying group and watch for an observed event | Marketing and seller |
| Low fit, high behavior | Activity from an account that does not match | Check the fit inputs first, since a stale company field can produce this cell; then nurture | Marketing |
| Low fit, low behavior | Neither | Suppress from outbound | Marketing |
The third row is where the Input Passport matters most. If the firmographic fields behind a low fit score are old, the lead is a data question before it is a nurture question. The weighting signals in lead scoring guide covers how to weight inputs by source confidence, and the ICP scoring that sales trusts guide covers building the fit side. Percepture's guide on using intent data to identify sales qualified leads covers the step from a signal to a qualified lead.
The Score Cell Ledger
The Score Cell Ledger answers the question the ranking pages leave open: do the cells close at different rates? It is built from your own outcomes.
- Freeze the score at creation. Store the fit score and behavior score as they were when the lead reached sales. A score that updates after the outcome cannot be tested against that outcome.
- Choose resolved leads only. A lead is resolved when it is won, lost or aged out after a window you set in advance. Open leads are left out, since counting them as losses understates every cell.
- Cut each score at a threshold. Pick a high and a low threshold before looking at outcomes. Choosing the cut after seeing results fits the ledger to the past.
- Count by cell. Record resolved leads and closed-won deals per cell, and compute the close rate as closed-won divided by resolved leads.
- Add an interval. Report the 95 percent Wilson interval beside each rate. It shows the range of close rates the data are consistent with, and it widens when the cell has few records.
- Compare intervals. Treat two cells as different only when their intervals do not overlap.
An Invented Worked Example
Tamsin Workflow Software is invented for this guide, and its numbers are invented. The ledger covers 800 resolved leads from one year.
| Cell | Resolved leads | Closed-won | Close rate | 95% Wilson interval |
|---|---|---|---|---|
| High fit, high behavior | 120 | 30 | 25.0% | 18.1% to 33.4% |
| High fit, low behavior | 200 | 18 | 9.0% | 5.8% to 13.8% |
| Low fit, high behavior | 150 | 9 | 6.0% | 3.2% to 11.0% |
| Low fit, low behavior | 330 | 6 | 1.8% | 0.8% to 3.9% |
| All resolved leads | 800 | 63 | 7.9% | 6.2% to 9.9% |
Two readings follow. The combined high cell's interval (18.1% to 33.4%) sits above every other cell's, so the data support treating it as a different group. The two middle cells overlap (5.8% to 13.8% against 3.2% to 11.0%), so the ledger does not show that fit outweighs behavior or the reverse. A team that read 9.0% against 6.0% as proof that fit matters more would be reading a difference the sample cannot support. The next step for Tamsin is to collect more resolved leads in those two cells, and to check whether the low-fit, high-behavior cell has stale company data.
Why Sample Size Changes the Reading
The same close rate carries different weight at different sizes. Each row below is a 25 percent rate.
| Resolved leads in the cell | Closed-won | 95% Wilson interval |
|---|---|---|
| 8 | 2 | 7.1% to 59.1% |
| 20 | 5 | 11.2% to 46.9% |
| 48 | 12 | 14.9% to 38.8% |
| 120 | 30 | 18.1% to 33.4% |
A cell with 20 leads cannot separate a 12 percent true rate from a 45 percent one. This is why vendor guidance on the minimum conversions for a model matters: CDP.com says a machine-learning model needs about 1,000 or more conversions, and House of MarTech says a team with fewer than 1,000 historical conversions should start with a rule-based model. Those thresholds are the publishers' guidance. The interval is a method any team can apply to the cells it has.
Where These Numbers Will Not Answer the Question
- Intervals measure sampling noise, not bias. If sellers work high-score leads harder, the high cell's close rate reflects effort as well as score. A control group that works leads without seeing the score separates the two.
- The method assumes one stable process. A pricing change or a new ideal customer profile mid-year mixes two populations in one cell.
- Resolved leads exclude the open pipeline. A long sales cycle leaves recent cells thin; use a cohort window longer than the cycle.
- Vendor outcome figures use other denominators. An MQL or opportunity lift is not a closed-won rate, and none of the cited pages states a baseline.
- Fit data ages. A firmographic score built on old company fields scores the company as it was.
Fact, Inference and Our View
| Statement | Type | Basis |
|---|---|---|
| HubSpot's combined scores group records into six labels by fit and engagement, and its performance reports show distribution, averages and scores over time | Fact (HubSpot documentation) | HubSpot Knowledge Base |
| ZoomInfo reports an 84% MQL increase and a 26% jump in opportunity rates for one customer | Fact (vendor claim) | ZoomInfo lead scoring guide |
| CDP.com and Saber each attribute performance figures to Forrester without a source link | Fact (about the pages) | The cited pages |
| None of the ranking pages in the table shows closed-won rates by score band | Fact (about the pages) | The cited pages |
| A cell with few leads gives a wide interval | Fact (statistical method) | Wilson interval arithmetic |
| A stale firmographic field can put an active account in the low-fit cell | Inference | Follows from how fit is computed |
| The Input Passport and the Score Cell Ledger make a score testable | Our view | This guide's framework |
| Confidence ratings in the passport table | Our view | This guide's judgment |
Frequently Asked Questions
What is the difference between behavioral scoring and firmographic scoring?
Behavioral scoring awards points for actions such as a pricing-page visit, a demo request or a reply, so it describes timing. Firmographic scoring awards points for company attributes such as industry, employee range, revenue, region and technology in use, so it describes fit. Many tools combine the two, and each can be tested separately against closed deals.
Should behavioral or firmographic scoring carry more weight?
There is no split that holds for every business. Plura AI describes a common starting allocation of 40 points for demographic and firmographic fit and 60 for behavior, and states no source for it. Set your own weights by filling the Score Cell Ledger and checking whether the cells separate, then revisit it when the process or the ideal customer profile changes.
How many closed deals do you need before trusting a score?
For a machine-learning model, CDP.com says about 1,000 or more conversions, and House of MarTech advises a rule-based model below that. For a ledger of cells the question is the width of the interval: a cell of 120 resolved leads at a 25 percent close rate has an interval of about 18 to 33 percent, and a cell of 20 has one of about 11 to 47 percent.
How do you measure close rates by score?
Store the fit and behavior scores when a lead reaches sales, keep only resolved leads, split them into cells using thresholds set in advance, and divide closed-won deals by resolved leads in each cell. State the denominator and report an interval with each rate. Compare cells only when their intervals do not overlap.
Sources
- CDP.com, AI Lead Scoring: How It Works, Models and CDP Integration (September 19, 2026); The Pedowitz Group, What is behavioral scoring; Saber, Behavioral Lead Scoring: Definition, Examples and Use Cases; Plura AI, Automated Lead Scoring Examples: 6 Worked Models (May 22, 2026); House of MarTech, Lead Qualification Models: Behavioral and Intent Scoring (April 20, 2026).
- ZoomInfo, Lead Scoring: A Step-by-Step Guide for B2B Teams.
- HubSpot Knowledge Base, Overview of the lead scoring tool and View lead score history and performance (last updated August 14, 2026).
- Wilson score interval, the standard binomial confidence interval used for the close-rate ranges above (95 percent, z = 1.96).
- Percepture, How to Use Intent Data to Identify Sales Qualified Leads.
- Lead Seeker, how Lead Seeker works, Lead Compass, best B2B contact database, buying signals vs intent data, weighting signals in lead scoring, predictive lead scoring and ICP scoring that sales trusts. Vendor pages and documentation are an October 2026 snapshot.
About the Author
Bob Generale is President of Percepture. He works across SEO, AI search, digital PR, sales intelligence and AI-powered revenue systems, with a focus on connecting visibility, buyer intent and sales action.
Next Steps
Pull the last 12 months of resolved leads, freeze the scores each had when it reached sales, and fill one Score Cell Ledger with intervals. Add the four Input Passport fields to the five inputs that carry the most points. Read how Lead Seeker works to see how a signal, a verified person and a source link sit on one record, and use Lead Compass to choose the accounts to test. Then start free with five verified contacts to compare a dated, source-linked record with a lead already in your score.