A score earns its place when the accounts it ranks highest are the ones that buy.
HubSpot makes a lead score easy to build. Pick some properties, assign points, set a threshold, and every contact in the CRM carries a number by the end of the afternoon.
The harder question comes a quarter later. When a rep works the top of the list, are those the deals that close? A score can be calculated exactly as designed and still rank the wrong people first.
This guide covers how HubSpot lead scoring works today, how to build a model from the deals you already won, and how to test that it keeps predicting them. The method is the Magnetiz HubSpot Lead Scoring Playbook, with a worksheet for each step.
What is lead scoring in HubSpot?
Lead scoring in HubSpot is a tool that adds and subtracts points on contacts, companies and deals, so the CRM can rank them by fit and interest. It lives under Marketing, then Lead Scoring.
It is available on Marketing Hub and Sales Hub at the Professional and Enterprise tiers, according to HubSpot's overview of the tool (updated September 2, 2026).
HubSpot offers three kinds of score.
- Fit scores rate how closely a contact or company matches who you sell to, using properties like company size, industry and job title.
- Engagement scores rate how much a contact or company has interacted with you, using events like page views, form submissions and email clicks.
- Combined scores hold both, and they are the only kind available on deals. HubSpot notes that deal scores are combined scores by default.
Which objects you can score depends on the hub.
HubSpot lists contacts under Marketing Hub, companies under Marketing Hub or Sales Hub, and deals under Sales Hub. You can create more than one score per object, so a team can run a new model beside the old one before switching.
Watch for a score that was built once and never revisited. The points reflect what the team believed about buyers on the day it was set up.
What is the difference between a fit score and an engagement score?
A fit score says whether this company could buy. An engagement score says whether someone there is paying attention right now. They answer two different questions, and keeping them in separate fields is what makes each one testable.
HubSpot's combined score shows both at once with a letter and a number. The letter is fit, from A for high fit to C for low fit. The number is engagement.
HubSpot's own example is a low-fit but highly engaged contact, who lands at C1. Its nine grades run from A1 to C3, and engagement and fit scores on their own use High, Medium and Low.
The grid is useful because the two failures look different.
An A3 is the right company that has gone quiet, a case for outreach. A C1 is someone very active at a company that cannot buy, often a student, a job seeker or a competitor reading the pricing page.
Watch for a single blended number with no way to see which half drove it. When fit and engagement are summed into one total, a C1 and an A3 can carry the same score and route the same way.
Why does lead scoring reward page clicks instead of buying intent?
Engagement points pile up fastest on the people who click the most, and the people who click the most are often researching, comparing or job hunting. When engagement carries more points than fit, the top of the list fills with the busiest readers.
It happens through ordinary setup choices. Ten points for a pricing page view and five for an email click feel reasonable on their own.
Across a month, an analyst at a company outside your market can collect more points than a VP at your best-fit account who visited once and asked for a demo.
HubSpot gives you two controls for this.
Score limits cap the points a group of criteria can add (the default maximum per group is 100 points).
Score decay lowers engagement points over time on a 1, 3, 6 or 12 month interval, so a burst of clicks from last spring stops counting as interest today.
Watch for a top-scored list that sales keeps skipping. When reps work around the score, the weights are worth checking first.
How do you build a lead scoring model from closed-won deals?
Start from the deals you actually closed, and write the fit rules they share, before assigning any engagement points. Set them beside the profile in the strategy deck, and let the backtest below pick.
Here is the order Magnetiz uses.
- Pull the closed-won and closed-lost deals from the last 12 months. Read each company's size, industry, region and the title of the person who signed.
- Write the fit rules as checks a stranger could apply. "Employee band 100 to 500, checked on the LinkedIn company page" is a rule. "Mid-market and growing" is an opinion.
- Add the disqualifiers as large negative points. Schools, government, agencies and competitors get cut on sight.
- Add engagement criteria last, and only the events that preceded real deals. A demo request that came before most of your wins earns its points. A blog view that shows up equally on wins and losses has not earned any.
- Decide what each threshold triggers. An A1 might go to a rep the same day and an A3 into a sequence. A threshold with no named action tends to outlive the definition it was set for.
Keep a fourth outcome beside fit, marginal and no fit.
An account with missing data, or two sources that disagree, is unknown. It goes to a queue with an owner, so a guess about a thin record never routes as if it were a fact.
What does a lead scoring model look like?
A working B2B model fits on one page, with fit rules on top, engagement rules below, and the action each grade triggers. The example below is illustrative. The company and the points are made up to show the shape.
The example company sells revenue software to B2B firms of 100 to 500 people in North America.
| Group | Criterion | Points |
|---|---|---|
| Fit | Employee count between 100 and 500 | +25 |
| Fit | Industry is B2B software or business services | +20 |
| Fit | Contact title is VP or Director of Sales, Marketing or Revenue Operations | +20 |
| Fit | Headquarters in the US or Canada | +10 |
| Fit | Email domain is a school, government body or known competitor | -100 |
| Engagement | Demo request form submitted | +40 |
| Engagement | Pricing page viewed (counted once, decays after 3 months) | +15 |
| Engagement | Webinar attended | +10 each |
| Engagement | Marketing email clicked (capped at 10 points) | +2 each |
| Grade | What happens |
|---|---|
| A1 | A rep reaches out within one business day |
| A2 | A rep reaches out within three business days |
| A3 | Added to a nurture sequence, reviewed monthly |
| B1 | Rep reviews the account before any outreach |
| C1 to C3 | No sales outreach, the reason recorded |
The weights in a real model come from step 4 above. If the demo request preceded 80% of your closed-won deals and the webinar preceded 20%, the points should show that gap.
See the full method in the playbook →
How do you test whether your lead scoring works?
Run two tests, one against the past and one against the accounts the score will meet next. HubSpot's own testing tools confirm a score is calculated as designed. Whether that design predicts revenue is a separate check, run against outcomes.
HubSpot lets you test an individual record to see what its score would be, and preview the distribution of scores across your records (in large accounts the preview uses a sample of about 20% of eligible contacts).
Its build guide also suggests a custom report "to track how scores affect deal creation or close rates over time." Those are the right starting points. The two tests below go further.
The backtest scores last year's deals. Score every closed-won and closed-lost deal from the last 12 months and compare where they land. In the illustrative example, if 31 of 40 wins grade A1 to A3 and 22 of 60 losses do too, the fit rules separate winners from losers. If wins and losses land in A at the same rate, the score is ranking something other than buyers.
The golden set checks the accounts you have not closed yet. Pick 20 to 30 accounts from the sources your routing draws on, including clear misses and hard edge cases. A person labels each one against the written fit rules before seeing the score. Then compare. Agreement tells you the score still applies your definition. Run it again after any change to the rules, the properties or the data provider.
Then read outcomes by grade every month. Credit each meeting, opportunity, win and loss to the grade the account held before the month began.
A1 should convert better than A3, and A3 better than C. When two grades convert at the same rate, the line between them is costing effort and deciding nothing.
The labelling guide and the agreement check for a golden set are in The ICP You Can Prove.
Watch for a score that only ever gets checked by the person who built it. A second person labelling the golden set catches what the builder reads past.
Should you use HubSpot's AI lead scoring?
HubSpot's AI lead scoring can draft a contact score from your own data, and it is worth trying as a starting point you then test. It requires a Marketing Hub Enterprise subscription and covers contact fit and engagement scores.
HubSpot says your contacts are evaluated to train the model, which then recommends criteria and points based on your account's contact data.
It needs at least 50 contacts, with 25 converted and 25 not converted, and evaluation can take up to an hour. Separately, engagement and combined scores can suggest AI rules based on high-impact events.
The recommendations come from whatever your CRM already records as converted. If that definition is loose, or the property data is thin, the AI will learn those gaps faithfully.
The same backtest and golden set apply. An AI-drafted score is a hypothesis about your buyers, and the closed deals are where it gets graded.
Watch for a conversion definition that counts form fills as conversions. A model trained on that will learn to find form fillers.
When should you score accounts instead of leads?
Score accounts when several people from one company take part in the purchase. One engaged contact at a poor-fit company and three quiet contacts at an ideal one tell very different stories, and a contact-level score cannot see the difference.
HubSpot company scores can roll contact activity up to the company with Sum, Average, Min or Max.
Sum rewards companies where many people engage. Max rewards the single most engaged person. For fit, score the company directly from company properties, since fit belongs to the account.
A practical split is to keep fit at the company level and engagement at both levels. Route on the company's grade, and use the contact scores to decide who at that company a rep should call first.
Watch for duplicate company records. Engagement split across two records for the same company makes a strong account look like two weak ones.
The pattern across a score that works
Every step above serves one question, whether the accounts ranked first are the ones that buy. Fit rules written from real wins answer it on the first day.
The backtest, the golden set and the monthly readout keep answering it as the market, the product and the data shift.
The playbook holds a worksheet for each step, free by email. The fit rules, the engagement evidence, the HubSpot build sheet, the grade actions, the backtest, the monthly readout, and ten sample records with an answer key.
Get the playbook and the kit →
A score is a claim about who buys, and the closed deals are where it gets checked.
FAQ
What is lead scoring in HubSpot?
Lead scoring in HubSpot is a tool that assigns points to contacts, companies and deals based on fit and engagement, so teams can rank and route them. It is found under Marketing, then Lead Scoring. HubSpot lead scoring supports fit, engagement and combined scores, and combined scores show a grade from A1 to C3. It is available on Marketing Hub and Sales Hub Professional and Enterprise.
Which HubSpot plans include lead scoring?
HubSpot lists the lead scoring tool on Marketing Hub Professional and Enterprise and Sales Hub Professional and Enterprise. Contact scores need Marketing Hub, deal scores need Sales Hub, and company scores work with either. Building contact lead scores with AI requires Marketing Hub Enterprise. Check HubSpot's knowledge base for changes, since these pages are updated often.
What is a good lead scoring model for B2B?
A good B2B lead scoring model starts with fit rules taken from the deals you closed, then adds engagement points only for events that came before real deals. It includes large negative points for disqualifiers like schools and competitors. Each grade triggers a named action, such as a same-day rep call for A1. A good lead scoring model is one where the top grades convert measurably better than the bottom ones.
What is the difference between a fit score and an engagement score?
A fit score rates whether a company or contact matches who you sell to, using properties like size, industry and title. An engagement score rates recent interaction, using events like page views and form submissions. In HubSpot's combined score, the letter shows fit and the number shows engagement. Keeping fit and engagement separate lets you test each one.
Can HubSpot build lead scores with AI?
Yes. HubSpot can build contact fit and engagement scores with AI on Marketing Hub Enterprise. It evaluates your contacts and recommends criteria and points, and it needs at least 50 contacts, 25 converted and 25 not converted. Treat an AI lead score as a draft and test it against closed deals before routing on it.
How do you know if lead scoring is working?
Lead scoring is working when the accounts it grades highest convert to meetings, opportunities and wins at a higher rate than the ones it grades lower. Test it with a backtest on the last 12 months of closed-won and closed-lost deals, and with a golden set of 20 to 30 accounts labelled blind by a person. Then compare outcomes by grade every month.
What is account scoring?
Account scoring rates a whole company instead of one contact, which suits B2B purchases that involve several people. In HubSpot, company scores can roll contact engagement up with Sum, Average, Min or Max, while fit is scored from company properties. Account scoring helps a team route on the company's grade and use contact scores to pick who to call first.