Revamp Your Lead Scoring: Three Independent Scores for Better MQLs

Most HubSpot lead scoring setups are built on a single number. One blended score. Add points for a job title. Add points for a page view. Subtract points for inactivity. Watch the number go up and down and call it intelligence. It is not intelligence. It is noise with a threshold attached to it. And the reason your MQL queue is full of people who will never buy is that you treated fit and intent as the same signal — added them together — and routed on the total.

The architecture in this post is different. Three independent scores. No blending. Lifecycle transitions that depend on the combination of scores, not the sum of them. And a disqualification layer that overrides everything else when a hard stop condition is met.

A client came to us with exactly this problem. Their HubSpot lead scoring was generating MQLs that the sales team had stopped trusting. Not because the score was low — because the score was meaningless. A CEO at a two-person startup scored the same as a Marketing Director at a 40-person technology company because both had visited the pricing page. One of those is pipeline. One of those is noise. Their single blended score could not tell the difference.

Here is exactly how we rebuilt it.

A lead score that cannot distinguish between who someone is and what they are doing right now is not a scoring system. It is a weighted contact list. Build the scores separately. Combine them at the decision point. Never before. Arsalan Faysal — Revenue Systems Architect

The Decision: Three Scores, Not One

Before any scoring table was built, we made the foundational architecture decision. It drives every configuration that follows.

A blended score fails because it treats orthogonal signals as substitutes for each other. A high intent score from someone who is the wrong company size does not mean the lead is half-qualified. It means you have an interested prospect you cannot sell to. Routing that contact to sales wastes a rep's time and, more importantly, conditions the sales team to distrust the scoring model entirely. Once trust is gone, you are manually reviewing every MQL and the automation is decoration.

Three independent scores solve this by measuring three genuinely separate things:

Lifecycle transitions are gated on combinations. A contact needs a qualifying ICP Fit score AND a qualifying Intent score to become an MQL. No shortcut. No workaround. And if the Disqualification score fires, the conversation is over regardless of how the other two look.

That is the architecture. Now the detail.

Score 1: ICP Fit — Who They Are

The ICP Fit Score is explicit and structural. It measures company shape and buyer authority. It does not change based on what a contact does on the website. It changes when we learn something new about who they are.

Maximum score: 100. Minimum viable fit to become MQL eligible: 60. Approved ICP threshold: 75.

Two design principles govern the entire scoring table.

Company size is the primary anchor. Title scoring does not exist without it. A Marketing Director at a 5-person startup scores differently than a Marketing Director at a 40-person company — because at 5 people, there is no marketing budget, no team, no demand infrastructure. The title is the same. The context is completely different.

ICP fit is additive only. No negative points in this layer. Negative scoring belongs in the disqualification layer where it can trigger a hard stop. Subtracting points from a fit score creates a noise problem — you end up with scores that look moderate and cannot tell you why. Keep the fit score clean and structural.

Company Size — The Anchor

Buyer Role — Size-Aware Title Scoring

No title score applies without a company size condition being met first. This is not optional. It is the rule that prevents a solo founder with a Gmail address from scoring into the buyer tier.

Firmographic Qualifiers

ICP Fit Score Thresholds

A score of 75 requires both company shape and buyer authority to be present. You cannot reach 75 on company size alone. You cannot reach 75 on title alone. Both conditions have to be true.

Score 2: Intent and Timing — What They Are Doing

Intent scoring is where most teams get it wrong in the other direction. They track every page view, assign points to every session, and end up with an intent score that reflects general curiosity rather than buying behaviour. A contact who reads three blog posts is not demonstrating intent. A contact who views the checkout page after visiting the integrations page after initiating a live chat is telling you something completely different.

The intent score rewards specificity and recency. It is capped at 100 but structured so that the highest-value signals — demo requests, checkout page views, account creation attempts — dominate the score and cannot be outrun by accumulated low-value activity.

High-Value Signals — Explicit Buying Motion

Engagement Signals — Active Evaluation

Depth and Sustained Interest Signals

All page-view signals require 75% scroll depth or meaningful time-on-page. This prevents idle tab inflation — a browser left open on your integrations page for 40 minutes because the contact forgot to close it should not score the same as someone who read it actively. Blog pages are excluded from all intent scoring signals.

Intent Score Thresholds and Lifecycle Actions

The SQL Override Rule

There is one exception to the standard threshold logic. And it exists for a specific reason.

When ICP Fit Score is 85 or above — meaning the company shape and buyer authority are exceptionally strong — and any two of the three signals below each score above 15 points, the contact is SQL-eligible regardless of the total intent score:

• Live Chat Initiated (+15)
• Sales Email Reply or Click (+15)
• Inbound Cold Email to Sales (+15)

This rule exists because the best time to engage a perfect-fit buyer is not when they have completed a full evaluation cycle. It is before their internal consensus has formed. Before a competitor has gotten there first. Two direct engagement signals from a near-perfect ICP contact is a stronger buying indicator than a moderate-fit contact who clicked everything on the website.

The override is not a loophole. It is a deliberate mechanism to surface high-fit, early-stage buyers before the standard scoring logic would catch them.

Score 3: Disqualification — The Override Layer

The disqualification score does not add to or subtract from the other two scores. It overrides them. Completely. When a disqualification trigger fires, the contact cannot become an MQL or an SQL — regardless of what the fit score and intent score say. The conversation is over.

This is the most important layer in the system. Without it, a CEO of a one-person company who submits a demo request looks like a hot lead. The fit score is high. The intent score is high. But there is no budget, no team, and no viable sales opportunity. The disqualification layer is what catches this before a sales rep spends 45 minutes on a discovery call going nowhere.

One configuration decision worth flagging explicitly: the personal email domain disqualification does not auto-resolve. If a contact provides a personal email on a form and then later provides a corporate email through a different touchpoint, HubSpot will not automatically remove the disqualification flag. This has to be reviewed manually — and a workflow alert has to fire when a disqualified contact updates their email to a corporate domain so that the review actually happens. Without that workflow, the contact sits permanently suppressed even after the disqualifying condition is gone.

How the Three Scores Drive Lifecycle

The scores do not combine into a single number. They combine at the decision gate. Here is the full lifecycle logic:

LIFECYCLE TRANSITION LOGIC
------------------------------------------------------------------
SUBSCRIBER / LEAD
  |
  Check: Disqualification Score triggered?
  |-- YES → Suppressed or Recycled. No further progression.
  |-- NO  → Continue
  |
  Check: ICP Fit Score?
  |-- Below 60  → Nurture only. Cannot MQL.
  |-- 60–74     → Borderline. Requires strong intent (40+) to MQL.
  |-- 75+       → Approved ICP. Eligible for MQL.
  |-- 85+       → Approved ICP. Eligible for SQL Override Rule.
  |
  Check: Intent Score?
  |-- Below 25  → Curiosity. Keep in nurture.
  |-- 25–39     → MQL (high interest). Optional SQL if sales capacity.
  |-- 40+       → SQL (strong). Auto-promote and route to sales.
  |
SQL OVERRIDE EXCEPTION
  ICP Fit 85+ AND any 2 of 3 direct signals above 15 points each:
  → SQL-eligible immediately, regardless of total intent score

LIFECYCLE OUTCOMES
  |-- MQL: ICP Fit 60+ AND Intent 25+
  |-- MQL → SQL: ICP Fit 75+ AND Intent 40+
  |-- Direct SQL: ICP Fit 85+ AND Override Rule triggered
  |-- Disqualified: Disqualification flag present (overrides all)
  |-- Suppressed: Disqualified + no path to recovery
  |-- Recycled: Disqualified with potential to re-enter on email update

The system is designed so that every possible state a contact can be in has a defined outcome and a defined next action. There is no grey zone where a contact sits unrouted because nobody is sure whether it qualifies. The logic is explicit. The routing is automatic. The exceptions require human review — and the workflow alerts make sure that review actually happens.

Three Ways Lead Scoring Breaks Down

Every lead scoring system that fails breaks down in one of three ways. Building this architecture was partly about solving the problem — and partly about designing against the failure modes from the start.

The blended score problem. Fit and intent are not the same type of signal. Adding them together destroys the information they contain. A contact who is perfect fit with zero intent is a different sales conversation than a contact who is borderline fit with strong intent. A blended score of 65 tells you nothing about which one you are looking at. Separate scores preserve the distinction so sales knows exactly what they are working with before they pick up the phone.

The inflation problem. Without caps, a contact who browses the website across four sessions over three weeks accumulates intent points passively until they look like a buyer. The cap structure prevents any single low-value signal cluster from substituting for a genuine high-value signal. If someone has never submitted a demo request, viewed the checkout page, or initiated a chat — they should not score the same as someone who has. The caps enforce that hierarchy.

The disqualification blindspot. Most scoring models apply negative points. The problem with negative points is that they can be overcome. Enough positive intent signals and the negative is buried. A personal email address should not be a −10 that can be overcome by three page views. It should be a hard stop. The disqualification layer exists because some conditions are disqualifying — not penalising. Design them accordingly.

If Your MQL Queue Is Full of the Wrong People

A single blended score. Sales reps who have stopped trusting the routing. MQLs that sit unworked because the team knows from experience that half of them are not real. A pipeline that looks full and converts at a rate that does not make sense.

This is not a volume problem. It is an architecture problem. The scoring model is not separating the signals that matter from the signals that do not — so everything looks equally qualified and nothing is.

The three-score architecture described in this post is not complex to implement. It requires one decision: commit to separating fit, intent, and disqualification before you configure a single workflow. Everything else follows from that commitment.

I design HubSpot revenue architecture for B2B companies with non-standard sales motions — two-sided platforms, high-velocity inbound, founder-led pipelines, and anything where the default HubSpot lead scoring assumptions do not match the actual commercial reality. The engagement starts with an architecture sprint: a written scoring model, lifecycle logic, property list, and workflow map before a single automation is touched.