How to Build a Lead Scoring Model in HubSpot

By Amit Gupta · Senior Director, Marketing Operations · Published 14 June 2026

Step 1: Define your MQL

Before you touch a score property, write down what "ready for sales" means in plain language and get sales to agree to it. A lead scoring model only works if the score it produces maps to a shared definition of a marketing-qualified lead. Skip this and you'll build a precise model of the wrong thing.

Anchor the definition in real customers

Look at your last 20–50 closed-won deals in HubSpot and ask what those people had in common before sales engaged: their industry, company size, job title, and the actions they took on your site. That profile, not a gut feeling, is what your scoring rules should reward.

Step 2: Pick fit and behavior criteria

Good lead scoring blends two dimensions: fit (is this the right kind of person and company?) and behavior (are they showing buying intent?). A lead that scores high on only one is not yet an MQL. A perfect-fit contact who never engages is a cold prospect, and a frequent visitor who can't buy is noise.

Keep the list short to start, roughly 8–15 criteria. It's far easier to add a rule later than to untangle why a bloated model is firing on everything.

Step 3: Assign points

Assign points to each criterion using the HubSpot Score property, choosing manual scoring for transparency or predictive scoring for scale. In manual scoring you add positive attributes (with point values) and negative attributes that subtract. HubSpot recalculates each contact's score automatically as their properties and activity change.

Manual scoring

In Settings → Properties → HubSpot Score, build "positive" rule sets (for example, +10 for a director-level title, +15 for a pricing-page view) and "negative" rule sets (−10 for a free-mail domain, −20 for unsubscribed). Weight behavior that sits closest to a purchase highest. A demo request should outscore a single blog visit by a wide margin.

Predictive scoring

On Enterprise tiers, HubSpot's predictive lead score uses machine learning across your historical conversions to rank likelihood to close. It needs a meaningful volume of closed-won and closed-lost data to be reliable, and it's a black box: useful as a second opinion, but harder to explain to a skeptical sales team than explicit point rules. Many teams run manual scoring for routing and watch predictive as a sanity check.

Step 4: Set the MQL threshold

The threshold is the score at which a lead becomes an MQL and earns a sales hand-off. Set it where historically good leads cross over, not at a round number you picked by feel. Score a handful of known great customers and a handful of bad-fit contacts, and place the line so it cleanly separates the two groups.

There is no universal "right" number, so resist copying a benchmark from another company. Start with the score your ideal customer profile would earn, then tune up if sales drowns in low-quality MQLs or down if good leads are getting stuck below the line.

Step 5: Route qualified leads to sales

Automate the hand-off with a workflow so a crossing of the threshold immediately changes the lifecycle stage and notifies the right owner. Manual triage at this stage is where speed-to-lead dies.

1. Create a workflow enrolled on HubSpot Score is greater than or equal to [your threshold].
2. Set the contact's Lifecycle stage to Marketing Qualified Lead.
3. Assign or rotate the contact to an owner using HubSpot's rotation action or your routing rules.
4. Send an internal notification or create a task so the rep sees it in minutes, not days.
5. Suppress re-enrollment edge cases so a lead isn't handed off twice for the same crossing.

Keep score-based MQL status separate from a rep's manual Lead status field so marketing's signal and sales' working state don't overwrite each other.

Step 6: Test, iterate, and add decay

Treat the first version as a hypothesis: review which scored leads actually became opportunities, adjust the weights that misfired, and add score decay so engagement that's gone cold stops counting. A model that's never revisited slowly drifts out of sync with how buyers really behave.

Validate against outcomes

After 30–90 days, compare MQLs that converted to opportunities against those sales rejected. If high scorers aren't converting, your behavior weights are too generous; if good buyers sit below the line, your fit criteria are too strict. Adjust a few rules at a time so you can tell what moved the result.

Add score decay

Behavior points should expire. A pricing-page view from six months ago is not the same intent as one from yesterday. Add a workflow that enrolls contacts with no activity in the last 30–90 days and subtracts a portion of their behavior score, re-enrolling on a schedule. Decay keeps dormant leads from sitting permanently above your MQL threshold and polluting the sales queue. Fit points generally shouldn't decay, since a director is still a director next quarter.

Frequently asked questions

Does HubSpot lead scoring require an Enterprise plan?

Manual scoring with the HubSpot Score property is available on Marketing Hub Professional. Predictive scoring and multiple custom score properties are Enterprise features. On any tier you can also approximate scoring with workflows that add or remove points from a custom number property.

What is the difference between manual and predictive lead scoring in HubSpot?

Manual scoring lets you set explicit point rules for fit and behavior criteria, so it is transparent and easy to debug. Predictive scoring uses HubSpot's machine learning across your closed-won history to rank likelihood to convert, which needs enough data and is harder to explain to sales.

What is a good MQL threshold score in HubSpot?

There is no universal number; the threshold is whatever score separates leads that historically convert from those that don't. A common starting point is the score a known good customer would earn, then tune it so sales accepts most of what crosses it without being overwhelmed.

How do I add score decay in HubSpot?

Use a workflow to subtract points when behavior goes stale. Enroll contacts whose last engagement was more than 30 to 90 days ago, deduct a portion of their behavior score, and re-enroll on a schedule so dormant leads cool off and don't sit above your MQL threshold forever.

Last updated: 14 June 2026