Lead scoring in your CRM: how to build a model your sales team trusts

What a lead score is, and what it is not

A lead score answers one question: who should we talk to next? It combines two kinds of signal. Fit describes who the lead is, such as industry, company size, role or location. Engagement describes what they have done, such as visiting your pricing page, booking a call or replying to an email.

A score is a priority signal, not a verdict. A perfect-fit lead with no engagement is a prospect to nurture, and a highly engaged student downloading every guide is not a buyer. Keeping fit and engagement visible separately, rather than only as one blended number, is what stops those two cases from looking the same.

Start with your won and lost deals, not a template

Generic scoring templates give points for things that may mean nothing in your business. Export the last six to twelve months of deals and compare the ones you won with the ones you lost or that never replied. Look for the traits and actions that appear far more often in the wins.

Then sit down with sales and agree, in writing, what a sales-ready lead looks like and what sales will do when one arrives. That agreement is what makes them trust the number. If your CRM data is too messy to compare deals, fix that first: our guide to a CRM audit before you automate covers the clean-up.

Step 1: Build the fit score

Fit points come from properties you already store on the contact or company. Give the most points to the traits that most clearly separated wins from losses, and keep the list short so everyone can explain it.

• Industry or niche you serve best: +15
• Company size or revenue in your sweet spot: +10
• Decision-maker job title, such as owner, founder or head of sales: +10
• Location you can serve: +5, and outside your service area: -20
• Budget question answered at or above your minimum: +15
• Personal email domain on a B2B offer: -5

Step 2: Build the engagement score

Weight actions by how close they are to a buying decision. Booking a call or asking for a quote should outweigh ten newsletter opens. Count email opens lightly or not at all, because Apple Mail Privacy Protection loads tracking pixels automatically, so many recorded opens never involved a person reading the email.

• Booked a call or requested a quote: +40
• Visited the pricing or services page twice in a week: +15
• Replied to an email or SMS: +15
• Submitted a form or downloaded a guide: +10
• Clicked a link in an email: +5
• Opened an email: +1, or leave it out

Step 3: Subtract points and let old activity fade

Negative scoring matters as much as positive scoring. Subtract points for unsubscribes, bounced emails, visits to your careers page, competitor domains and replies that say not interested. Without them, a lead who clicked everything two years ago and then went quiet still looks hot.

Decay solves the same problem over time. Reduce engagement points as they age, for example halving them after 90 days, so the score reflects what a lead is doing now. Fit points usually do not decay, because a company's size or industry does not change because they went quiet.

Step 4: Set thresholds and decide what happens at each

A score only earns its keep when it triggers an action. Pick two or three bands and agree what each one means. A simple version is hot, warm and cold, or an A to C grid that crosses fit with engagement.

Hot leads should reach a person within minutes, which is where routing comes in: our guide to lead routing and speed to lead covers how to assign and alert. Warm leads go into a nurture sequence, and cold leads stay in marketing until their score changes. Our list of CRM automation workflows has the follow-up flows that sit behind each band.

• Hot (for example 70+ with good fit): create a task, alert the owner, send a booking link
• Warm (40 to 69): move into a nurture sequence with case studies and an offer to talk
• Cold (under 40): stay in newsletters and retargeting audiences
• Poor fit at any engagement level: route to a lower-touch offer or politely disqualify

How to set up lead scoring in HubSpot

HubSpot's lead scoring tool is included with Marketing Hub or Sales Hub on Professional and Enterprise plans. You create fit scores from property values, engagement scores from actions such as page views, form submissions and email clicks, or a combined score that uses both. Contacts and companies support all three, and deals support combined scores.

Each score can have an overall points limit and limits per group, so one busy lead cannot run away with the total. Engagement groups can use decay, reducing points every 1, 3, 6 or 12 months. Thresholds label records High, Medium or Low, and combined scores use a grid from A1 to C3, where the letter is fit and the number is engagement. AI-created contact scores need Marketing Hub Enterprise.

• Build fit and engagement as separate groups so you can see why a lead scored high
• Set thresholds, then use them as enrolment triggers in workflows and as filters in sales views
• Add the score and threshold columns to the contact views sales use every day

How to set up lead scoring in GoHighLevel

In GoHighLevel, create a number custom field called Lead Score, plus a dropdown for the band if you want one. Then build workflows that react to each signal: a form submitted, an appointment booked, a link clicked or a trigger link visited. In each, use the Math Operation action to add or subtract points from Lead Score. HighLevel replaces negative results with zero, so a score cannot drop below nothing.

A final workflow watches the Lead Score field and uses If/Else branches to set the band, add a hot-lead tag, create an opportunity in the right pipeline stage and notify the owner. Decay needs a scheduled step, such as a workflow that subtracts points after 30 or 60 days with no activity. If you are deciding between the two platforms, our GoHighLevel vs HubSpot comparison covers the wider trade-offs.

Where AI fits in lead scoring

Rules-based scoring is easy to explain, which is why we start there. AI adds two things on top. Predictive scores, such as HubSpot's AI-created scores, look for patterns across your historical data. AI agents read free-text messages, such as a form comment or a chat, and judge intent that no fixed rule can see.

A practical mix is a rules-based score for fit and actions, plus an AI step that reads the enquiry and adds or removes points with a written reason sales can check. We cover the conversation side in AI agents for lead qualification, and how to build that step yourself in how to build an AI agent with n8n.

Review the model every quarter

A lead scoring model is a hypothesis, so test it. Each quarter, compare the meeting and close rates for each score band. If warm leads close as often as hot ones, your thresholds are wrong. If sales keeps ignoring hot leads, ask why and change the inputs, not just the numbers.

Revisit the model whenever you launch a new offer, enter a new market or change your forms, because the signals that predicted a sale last year may not predict one now.

Frequently asked questions