Lead scoring in HubSpot is only useful if sales trusts it. A scoring model that marketing built without sales input, validated against criteria that marketing cares about rather than criteria that predict deals, will produce MQLs that sales ignores - and an ever-widening gap between marketing's pipeline contribution numbers and sales' experience of lead quality.
This guide covers the full lead scoring process in HubSpot: the difference between fit, engagement, and combined scores; how to build a model that reflects the ICP accurately; how to set thresholds that produce MQLs sales will act on; how to wire scores into routing and automation; and how to establish the sales-marketing agreement that makes the whole system credible and sustainable.
For how lead scores connect to deal creation and pipeline management, see our HubSpot deals guide 2026. For the broader automation framework that surrounds scoring, see our marketing automation strategy guide 2026.
What Lead Scoring Is in HubSpot
HubSpot's lead scoring system assigns numerical values to contacts based on properties and behaviours that indicate how likely they are to become customers. The total score (the sum of all applicable criteria) provides a single number that reflects a contact's current qualification status - their fit with the ICP, their level of engagement with marketing content, or a combination of both.
In 2024–2025, HubSpot restructured its scoring tools. The legacy 'HubSpot Score' property (a single combined score) was sunsetted and replaced with three separate scoring properties: Contact Fit Score, Contact Engagement Score, and Contact Score (a combined property). Company-level scoring was also introduced. These scores exist alongside HubSpot's AI-powered predictive scoring for Enterprise portals. (Source: HubSpot Knowledge Base - Lead scoring: knowledge.hubspot.com/reports/create-use-deal-lead-scores)
The three scoring properties and their purpose:
- Fit Score: Measures how well a contact matches the ICP - based on demographic and firmographic properties (job title, company size, industry, geography). This score is static until properties change.
- Engagement Score: Measures how actively a contact is interacting with marketing content - based on page visits, email interactions, form submissions, and content downloads. This score changes as behaviour accumulates or decays.
- Combined Score (Contact Score): A composite of fit and engagement, weighted to reflect the relative importance your team assigns to each dimension. A high-fit, low-engagement contact is a different routing decision from a low-fit, high-engagement one.
Fit vs Engagement vs Combined: When Each Model Applies
Fit-only scoring is most useful for outbound-led or ABM teams where the primary qualification question is 'does this contact match our ICP?' - not 'have they shown marketing engagement?' In an outbound motion, the rep initiates contact without waiting for engagement signals. Fit scoring identifies who is worth prioritising for outreach.
Engagement-only scoring is most useful for inbound-led teams where traffic volume is high and the primary question is 'which inbound contacts are showing commercial intent?' Engagement scoring surfaces contacts who are actively evaluating without relying on ICP data that may be missing or incomplete.
Combined scoring is the right model for most B2B SaaS and professional services teams. It rewards contacts who are both a good fit and showing meaningful intent - the highest-quality MQLs - while still surfacing high-intent contacts who need fit data enriched, and high-fit contacts who need nurture to produce engagement.
A practical rule of thumb: if your database has at least 60% ICP field completeness (company size, industry, job title populated), combined scoring will produce more reliable MQLs than engagement-only scoring. Below 60% completeness, engagement-only scoring is safer - an engagement-only score does not penalise contacts for missing data.
Building Your Scoring Model
Step 1: Validate the ICP in Data Terms
Before building any scoring criteria, translate the ICP from a qualitative description ('we sell to mid-market B2B SaaS companies') into specific, queryable data fields. The ICP must be expressible as HubSpot contact and company property values - not as attributes that require a discovery call to establish.
For each ICP dimension, identify:
- The HubSpot property that captures it: Job Title (contact), Number of Employees (company), Industry (company), Country/Region (company).
- The specific values that indicate fit: Job titles that match the buyer persona, employee count ranges that match the target company size, industry verticals that have converted historically.
- The values that indicate poor fit: Job titles outside the buyer persona, company sizes outside the target range, competitor domains - these drive negative scores.
Validate ICP criteria against historical closed-won data before setting point values. If companies with 500–2,000 employees closed at twice the rate of companies with 50–200 employees in the last 12 months, the scoring should reflect that pattern. Do not score based on who you want to sell to - score based on who has bought. For how website data feeds these signals, see our HubSpot Webflow integration guide.
Step 2: Define Engagement Signals and Their Weight
Engagement signals should be weighted by their correlation with commercial intent - not by how impressive they sound. A pricing page visit is a higher-intent signal than a blog post view; a demo request is a higher-intent signal than a webinar registration. The point values should reflect that hierarchy.
Fit Score Criteria
The following criteria table provides a starting-point framework for fit scoring. Adjust point values based on historical win-rate data for your specific product and market.
Engagement Score Criteria
Engagement criteria are dynamic - they should accumulate as contacts take actions and, where appropriate, decay when contacts go inactive. The table below lists the most commonly used engagement signals for B2B SaaS and professional services companies.
A note on email open rates as a scoring signal: Apple Mail Privacy Protection (MPP) pre-fetches email content on behalf of recipients, causing opens to register for contacts who may not have actively read the email. Using opens as a scoring signal inflates scores for Apple Mail users. Use clicks instead of opens as the primary email engagement signal in scoring models built or updated from 2022 onward. Weight email opens at 1 point or zero and rely on clicks for meaningful email engagement scoring.
Score Decay: Keeping Scores Current
A scoring model without decay produces contacts that score high based on activity from 18 months ago and never fall back down - even if they have not engaged since. Score decay is the mechanism that keeps the model current.
How to implement score decay in HubSpot:
- Time-based decay via workflow: Create a workflow that runs on a weekly or monthly schedule and decrements the Contact Score by a defined amount for contacts who meet an inactivity criterion (e.g., no email click in 60 days). The decrement value should be set so that a fully disengaged contact drops below the MQL threshold within 90–120 days.
- Negative scoring for inactivity signals: Add a negative score criterion for specific inactivity states: 'No email open in 90 days = -15 points', 'Unsubscribed = -30 points'. These are static negative criteria that apply permanently once the condition is met.
- Score reset on re-engagement: When a contact who has decayed below threshold re-engages (opens a link, visits a high-intent page), the positive score for that action should fire and begin rebuilding their score from the decayed baseline. This is the correct behaviour - no manual reset needed.
Thresholds and Routing: Turning Scores into Actions
The scoring model only produces value when scores are connected to actions. Without threshold-triggered routing, a contact's score change is a number in a database that nobody acts on. The threshold framework maps score ranges to specific next actions, owners, and SLAs.
The threshold values in this framework are starting points - the MQL threshold specifically must be calibrated to your sales team's capacity. An MQL threshold that is set too low produces a high volume of MQLs that sales cannot action within SLA, which trains sales to deprioritise the routing queue. An MQL threshold that is set too high produces too few MQLs and leaves genuine pipeline opportunities uncontacted. The right threshold is the one that produces a volume of MQLs the sales team can contact within 24 hours while maintaining a sales acceptance rate above 70%.
Wiring Thresholds into HubSpot Workflows
In HubSpot, score thresholds connect to actions via workflows. See our marketing automation workflow guide 2026 for the full MQL routing workflow template. The core logic:
- Enrollment trigger: Contact Score is greater than or equal to [MQL threshold] AND Lifecycle Stage is not 'Marketing Qualified Lead' or beyond.
- Action 1: Set Lifecycle Stage to Marketing Qualified Lead.
- Action 2: Rotate contact owner to the assigned sales rep (by territory, account size, or round-robin depending on team structure).
- Action 3: Create an internal task for the rep: 'New MQL - [Contact Name] at [Company]. Score: [Score value]. Last action: [Last Page Seen]. Contact within 24 hours.'
- Action 4 (high-intent only): For contacts who triggered a high-intent signal (pricing page visited, demo request), send an immediate Slack or email notification to the rep rather than creating a task - the SLA should be same-day, not 24-hour.
Company-Level Scoring
HubSpot also supports scoring at the company level - assigning scores to company records rather than individual contacts. Company scoring is most relevant for ABM (Account-Based Marketing) programmes where the target is a named account rather than a specific contact within it.
Company scores aggregate signals from:
- Contact-level fit and engagement scores for all associated contacts (multiple engaged contacts at one company indicates genuine buying intent)
- Company property matches to the ICP (revenue, employee count, industry, technology stack if enriched)
- Company-level engagement signals (if multiple contacts from the same company are engaging simultaneously, the company score reflects the collective interest)
Company scoring is most useful when sales teams prioritise accounts rather than contacts - where the question is 'which companies should we focus on this month?' rather than 'which individual contacts are ready to talk to sales?'. For teams running a contact-first motion, company scores are supplementary context rather than the primary routing signal.
HubSpot's AI Predictive Scoring
HubSpot's AI-powered scoring (available in Marketing Hub Professional and Enterprise, and accessible via the Contact Scoring section in the scoring tool) uses machine learning to predict the likelihood a contact will convert to a customer, based on the engagement and property patterns of contacts who converted historically. Unlike manual scoring, predictive scoring does not require you to define criteria - it identifies patterns automatically from historical data. (Source: HubSpot Knowledge Base - Predictive lead scoring: knowledge.hubspot.com/reports/create-use-deal-lead-scores)
When to use predictive scoring:
- Use predictive as a complement to manual scoring. Predictive scoring surfaces contacts that manual criteria may have missed - a contact who does not match any defined engagement criterion but whose property profile closely resembles past converters. Running both in parallel and cross-referencing discrepancies identifies gaps in the manual model.
- Require 90 days of conversion data before enabling. Predictive scoring needs a sufficient conversion history to train against. Enabling it in a new HubSpot portal with fewer than 50–100 closed-won contacts produces an unreliable model. Wait until the contact database has meaningful conversion signal before relying on predictive scores.
- Do not replace manual scoring with predictive scoring in the first year. Manual scoring is transparent - sales can see exactly why a contact scored 65 points. Predictive scoring is opaque - sales cannot easily verify why a contact has a high predicted likelihood. Sales trust requires explainability, at least initially. Maintain the manual model as the primary routing mechanism and use predictive as a secondary signal.
Aligning Lead Scoring with Sales: The Agreement Framework
A lead scoring model that marketing built without sales input is a model sales will not trust. The most common failure mode in lead scoring: marketing configures a model, sets an MQL threshold, begins routing MQLs to sales, and discovers six weeks later that sales is either ignoring the queue or disputing every MQL that arrives. The problem is not the scoring model - it is the absence of a shared agreement.
The scoring alignment session should happen before any scoring criteria are configured, and should produce written agreement on the items in the table below:
The alignment session should be facilitated by RevOps or a neutral party - not by marketing presenting a scoring model to sales. The session should start with closed-won deal analysis: which contacts became customers, what did they have in common, and what signals preceded their conversion? This data-first approach produces criteria that sales recognises from their own experience rather than criteria marketing invented.
Measuring Whether Lead Scoring Is Working
The measurement framework for lead scoring has three layers:
Layer 1 - Volume and velocity. Track MQL volume per week and MQL-to-SQL conversion rate (the percentage of MQLs that sales accepts as Sales Qualified Leads and creates deals for). An MQL-to-SQL rate below 50% suggests the threshold is too low or the scoring criteria are not aligned with what sales considers qualified. Above 80% may suggest the threshold is too high and genuine pipeline is being left below the MQL line.
Layer 2 - Revenue impact. Track pipeline influenced (total deal value in open pipeline that had an MQL-routed contact) and revenue attributed (closed-won deals where the closed contact was an MQL-routed lead). These numbers connect the scoring model to actual revenue, which is the metric that justifies the investment.
Layer 3 - Model accuracy. Compare the score of contacts who became customers versus contacts who were rejected by sales or did not convert. If contacts who converted had an average score of 55 and contacts who did not had an average score of 60, the model's discriminative power is weak. The converted contacts should score meaningfully higher on average than the non-converted ones - if they do not, the scoring criteria do not predict conversion.
Review the scoring model at 90 days post-launch, then quarterly. The 90-day review should focus on model accuracy: are the contacts scoring above the MQL threshold converting at a higher rate than contacts scoring below it? If not, identify which scoring criteria are adding noise rather than signal and remove or reduce their weight. See our marketing automation best practices 2026 for the governance framework that keeps this review process on a regular cadence.
Common Lead Scoring Mistakes
- Scoring without data validation. Setting point values based on intuition ('pricing page should be worth 20 points') rather than historical conversion correlation. Validate every scoring criterion against closed-won data before assigning point values.
- No negative scoring. A model without negative criteria inflates scores for misfit contacts - a CEO at a 5-person startup accumulates the same behavioural score as a VP Marketing at a 500-person target account. Negative criteria for non-ICP job titles, company sizes outside the target range, and competitor domains prevent score inflation for poor-fit contacts.
- Threshold set without considering sales capacity. Setting the MQL threshold to produce 200 MQLs per week when the sales team can contact 50 per week trains sales to ignore the queue. The threshold must be calibrated to what sales can realistically follow up with within the agreed SLA.
- No decay mechanism. Contacts who engaged heavily 18 months ago and have not engaged since remain at their historical high score, appearing as MQLs when any new minor engagement pushes them over threshold. Score decay prevents this accumulation of false-positive pipeline.
- No shared agreement with sales. Any scoring model without a written sales-marketing agreement on MQL definition, routing SLA, and rejection criteria will face ongoing credibility challenges from the sales team. Build the agreement before building the model - not after the first disputed MQL.
Work with Belt Creative
Belt Creative implements HubSpot for B2B teams including lead scoring setup, threshold calibration, MQL routing workflow design, sales alignment facilitation, and the 90-day measurement review. If you are building a new lead scoring model or fixing one that sales has stopped trusting, we can help.
See our work for examples of HubSpot implementations, or get in touch to discuss your lead scoring and marketing-to-sales alignment requirements.
A Note on Sources
HubSpot Knowledge Base - Lead scoring: knowledge.hubspot.com/reports/create-use-deal-lead-scores. HubSpot Knowledge Base - Contact scoring: knowledge.hubspot.com/contacts/set-up-lead-scoring. HubSpot Knowledge Base - AI predictive scoring: knowledge.hubspot.com. All HubSpot feature and configuration guidance verified against official documentation, mid-2026. Verify current feature availability at knowledge.hubspot.com before implementation.
Frequently Asked Questions
What Is the Difference Between Fit Score and Engagement Score in HubSpot?
Fit Score measures how well a contact matches your ICP - based on static demographic and firmographic properties like job title, company size, and industry. Engagement Score measures how actively a contact is interacting with your marketing content - based on dynamic signals like page visits, email clicks, form submissions, and webinar attendance. Combined Score is a composite of both, weighted by the relative importance you assign to each dimension. Most B2B teams should use Combined Score as the primary routing signal.
How Do I Get Sales to Trust Lead Scoring?
Build the model with sales input, not for sales. Run a closed-won deal analysis together: which contacts converted, what did they have in common, and what signals preceded their conversion? Use those patterns to define scoring criteria. Then document the MQL definition, routing SLA, and rejection criteria in a written agreement that both VP Sales and VP Marketing sign. Review the model against actual MQL-to-SQL conversion rates at 90 days and quarterly thereafter. Trust is built by demonstrating that scored MQLs convert at a higher rate than unscored contacts - not by telling sales to trust the model.
Should I Use Manual or AI Predictive Scoring?
Start with manual scoring and add predictive scoring after 90 days of conversion data has accumulated. Manual scoring is transparent - sales can understand why a contact scored 65 points. Predictive scoring is more accurate over time but less explainable, which makes it harder to build sales trust with. Run both in parallel once predictive scoring is available; use the discrepancies to identify gaps in the manual model.
How Do I Set the MQL Score Threshold?
Calibrate the threshold to produce a volume of MQLs that the sales team can contact within the agreed SLA. Run a test: lower the threshold and measure how many additional contacts would qualify as MQLs per week. Ask sales how many MQLs they can realistically contact within 24 hours. Set the threshold so that the resulting volume is manageable. Then validate: are contacts above the threshold converting to SQLs at a meaningfully higher rate than contacts below it? If not, adjust the criteria rather than the threshold.
Can I Score Companies as Well as Contacts?
Yes. HubSpot supports company-level scoring in addition to contact scoring. Company scores aggregate signals from associated contacts and company-level property matches to the ICP. Company scoring is most relevant for ABM programmes where the primary question is 'which accounts should we prioritise?' rather than 'which individual contacts are ready for sales outreach?'
How Often Should I Update My Lead Scoring Model?
Review at 90 days post-launch to assess whether scoring criteria are accurately predicting conversion. Then quarterly - comparing the average score of contacts who became customers versus contacts who did not convert. Update the model when: win rates by segment change significantly, the ICP evolves, a new product changes the buyer profile, or the sales team reports systematic disagreement with MQL quality. The model should be treated as a living document, not a one-time configuration.
What Properties Should Not Be Used in Lead Scoring?
Email opens (unreliable due to Apple Mail Privacy Protection pre-fetching), properties that are rarely populated (below 40% completeness across the database), job title free-text fields without standardisation (use dropdown or normalised values instead), and any property that reflects your internal assessment rather than a contact's own action or declared attribute. Also avoid scoring on properties that change frequently without deliberate action - for example, a contact's geographic region if your data has poor address quality.
Sources: knowledge.hubspot.com/reports/create-use-deal-lead-scores; knowledge.hubspot.com/contacts/set-up-lead-scoring. All configuration and feature information verified mid-2026.