How to Build a Lead Scoring Framework from Scratch
Most B2B SaaS teams don't have modern lead scoring. Those that do often have a version built 2-3 years ago that no one maintains. In 2026, the highest-performing teams are shifting from manual, rules-based scoring to AI-assisted models that dynamically weight behavioral, intent, and product signals. If you're starting from scratch, this guide walks you through building a framework that combines rules-based foundations with AI augmentation, using product usage data as a first-party signal.
Before You Build: Alignment and Definitions
Before touching spreadsheets or CRM settings, align your team on definitions.
Define Your Ideal Customer Profile (ICP)
Your ICP is the bullseye. Lead scoring starts here.
Ask your sales team: - What industries do we sell best to? (e.g., financial services, B2B SaaS, healthcare) - What company sizes close fastest? (e.g., 50-500 employees, not micro, not enterprise) - What revenue range signals buying power? (e.g., [pricing varies, check vendor website]M-100M ARR) - What geographies are easiest to serve? (e.g., US + UK) - What job titles are decision-makers? (e.g., VP of Marketing, CMO)
Document this. Example ICP: - Industry: B2B SaaS companies (marketing tech, sales tech, analytics) - Size: 50-500 employees (mid-market, not enterprise, not startup) - Revenue: [pricing varies, check vendor website]M-100M ARR - Geography: US, UK, Canada - Decision-makers: VP of Marketing, CMO, VP of Revenue, Sales Director
Define MQL and SQL
These definitions matter more than they seem.
Marketing Qualified Lead (MQL): - A lead that matches your ICP AND shows buying intent (research, engagement, or purchase readiness) - Ready to enter a sales conversation - Marketing says: "This person should talk to sales"
Sales Qualified Lead (SQL): - A lead a sales rep qualifies as worth pursuing - May add criteria: budget confirmed, timeline mentioned, decision-maker confirmed - Sales says: "This is worth our time to pursue"
Your scoring framework gets prospects from lead -> MQL. Sales determines if MQL -> SQL (qualification call).
Example definitions:
MQL: Prospect matches ICP fit signals (company size, industry) + shows engagement signals (visited product pages, downloaded content, or requested demo) + passes ICP industry check.
SQL: Prospect is an MQL + has confirmed budget ("We have budget allocated") + has indicated timeline ("Evaluating through Q3") + prospect confirms they're a decision-maker or can get to one.
Map Your Sales Process
Understanding your sales process shapes your scoring.
Typical enterprise/mid-market B2B SaaS sales process: 1. Awareness: Prospect discovers you via ad, content, referral 2. Consideration: Prospect visits website, reads blog, downloads content 3. Evaluation: Prospect visits pricing page, watches demo video, requests demo 4. Negotiation: Prospect enters sales conversation, gets proposal 5. Close: Deal closes
Your scoring should reflect this journey: - Awareness + consideration stage leads: Low engagement score (just visiting) - Evaluation stage leads: High engagement score (pricing pages, demo request)
The Framework: Three Scoring Layers
Build scoring in three layers. Start with fit (foundation), add engagement (momentum), add intent (future signal).
Layer 1: Fit Scoring (Firmographic)
Fit scoring answers: "Does this company match our ideal customer?"
Create fit criteria:
Pick 3-5 firmographic signals. Each criterion is binary (yes/no) or tiered (points if in range, no points if not).
Example framework:
| Criterion | Points | Logic |
|---|---|---|
| Industry match | 30 | If financial services, SaaS, or healthcare: +30. Otherwise: +0 |
| Company size | 25 | If 50-500 employees: +25. If 500-2,000: +15. If 2,000+: +5. If <50: +0 |
| Annual revenue | 20 | If [pricing varies, check vendor website]M-100M ARR: +20. If [pricing varies, check vendor website]M+: +10. If <[pricing varies, check vendor website]M: +0 |
| Geography | 15 | If US/UK/Canada: +15. Other: +0 |
| Keyword in job title | 10 | If job title contains "VP", "CMO", "Director": +10. Otherwise: +0 |
Total fit score: 0-100 points
A prospect matching all five criteria gets 100 points (perfect ICP match). A prospect matching none gets 0 (not a fit).
Implementation in CRM:
In HubSpot, Salesforce, or Pipedrive: 1. Create a custom field: "Fit Score" (number field, 0-100) 2. Create automation rules: - If Industry = [Your target industries], add 30 points to Fit Score - If Company Size = [Your range], add 25 points - If Revenue = [Your range], add 20 points - Etc. 3. Set threshold: Fit Score 60+ is a baseline fit.
Layer 2: Engagement Scoring (Behavioral)
Engagement scoring answers: "Is this prospect actively engaging with our company?"
Engagement signals are time-based. Recent signals matter more than old ones.
| Signal | Points | Decay |
|---|---|---|
| Demo request | +30 | No decay (explicit intent) |
| Pricing page visit | +20 | 50% at 14 days, 25% at 30 days |
| Product page visit | +15 | 50% at 7 days, 25% at 14 days |
| Content download | +12 | 50% at 14 days, 25% at 30 days |
| Whitepaper/case study download | +10 | 50% at 30 days, 25% at 60 days |
| Website visit | +5 | 50% at 7 days, 0 after 14 days |
| Email open | +5 | 50% at 3 days, 0 after 7 days |
| Email click | +8 | 50% at 3 days, 0 after 7 days |
Signal decay example: A prospect downloads a whitepaper today: +10 points. In 30 days: +5 points (50%). In 60 days: +2.5 points (25%). After 60 days: +0 points (signals expire).
This prevents old engagement from inflating scores.
Total engagement score: 0-50 points (cap it so fit + engagement don't over-weight)
A prospect requesting a demo in the last 7 days scores 30 (demo) + 15 (recent product visit) + 10 (content download) = 55, capped at 50.
Implementation in CRM:
This requires automation because engagement changes daily.
In HubSpot: 1. Create workflows triggered by actions (page visit, content download, email open) 2. Increment "Engagement Score" field when actions occur 3. Create a "Last Activity Date" field 4. Use a periodic workflow (daily) to apply decay: - If last activity was 7+ days ago, reduce engagement score by 50% - If last activity was 30+ days ago, reduce by 75% - If last activity was 60+ days ago, reset to 0
Or use a third-party tool that automates decay (Marketo, Pardot, or scoring tools like Clearbit, Alchemer).
Layer 3: Intent + Product Signal Scoring (Future Readiness)
Intent scoring answers: "Is this prospect actively in-market, researching solutions?" In 2026, add first-party product signals (for PLG/freemium models) and dark social indicators.
| Signal | Source | Points | Cost |
|---|---|---|---|
| First-party website behavior (product pages, pricing) | Your analytics | +10 | Included |
| Product adoption / trial usage (for PLG/freemium) | Product analytics (Amplitude, Mixpanel) | +15 | [pricing varies, check vendor website] |
| Competitive research (visited competitor sites) | Third-party intent vendors | +8 | [pricing varies, check vendor website] |
| Keyword research (researching your category) | Bombora, 6sense | +10 | [pricing varies, check vendor website] |
| G2 review activity | G2 API / manual | +8 | pricing varies, check vendor website to [pricing varies, check vendor website] |
| LinkedIn hiring signals (relevant job postings) | +8 | Free (manual) or [pricing varies, check vendor website] (automated) | |
| Dark social mentions (Reddit, forums, Slack, Discord) | Sprout Social, social listening | +8 | [pricing varies, check vendor website] |
| Recent company news (funding, executive hire) | CrunchBase, ZoomInfo | +5 | Included in enrichment |
Total intent + product score: 0-35 points
Implementation:
Start with free intent signals: - First-party intent: Set up UTM tracking for product pages. Score anyone visiting a product page in the last 7 days. - LinkedIn hiring: Manually monitor job postings for 20-30 target accounts. When you see relevant postings, add +8 points. - G2 research: Use G2's API to track when companies from your target list are active on your product's G2 page.
For paid signals: - If budget allows, integrate Bombora or 6sense. They provide intent scores that you can sync to your CRM daily.
---Combining Layers: Total Score and Thresholds
Total possible score: 0-185 points - Fit: 0-100 - Engagement: 0-50 - Intent + Product: 0-35
Define MQL thresholds:
Decide: At what score does someone become an MQL?
Conservative approach (higher bar): - MQL threshold: 120 points - Requires: Strong fit (80+) + some engagement (30+) + some intent (10+) - Result: Fewer MQLs, but higher conversion quality - Use when: Sales team is small, you want high conversion rate
Balanced approach (typical): - MQL threshold: 100 points - Requires: Good fit (60+) + some engagement (30+) + minimal intent (10+) - Result: Moderate volume, reasonable conversion - Use when: Sales team is mid-size, you balance volume and quality
Aggressive approach (lower bar): - MQL threshold: 80 points - Requires: Some fit (40+) + any engagement (20+) - Result: High volume of MQLs, lower conversion rate - Use when: Sales team is large, you want to capture all early-stage opportunities
Adding Negative Scoring
Negative scoring disqualifies leads that don't fit.
| Disqualifier | Points | Logic |
|---|---|---|
| Non-target industry | -30 | If prospect works in manufacturing (not your target), subtract 30 points |
| Company too small | -20 | If company has <10 employees, subtract 20 |
| Recently implemented competitor | -40 | If prospect announced they just signed with your competitor, subtract 40 |
| Budget constraint stated | -15 | If prospect says "no budget this year", subtract 15 |
| Wrong title for purchase influence | -10 | If prospect is IC (individual contributor) with no management influence, subtract 10 |
| Unsubscribed from emails | -50 | If prospect unsubscribed, set score to 0 (lead is dead) |
| Free plan only (no commercial intent) | -20 | If prospect uses free version and has stated no plans to upgrade, subtract 20 |
Negative scores should cap the total at 0 (no negative scores allowed). A lead with negative signals gets a score of 0, not -50.
Implementation:
Create workflows: - If prospect unsubscribes: Set "Fit Score" = 0 - If prospect fills out form saying "Not in-market this year": Subtract 15 from total score - If prospect is on your "Do Not Call" list: Set total score to 0
Putting It Together: A Complete Example
Let's walk through a real prospect.
Prospect: Jane Doe, VP of Marketing at Acme Financial
Signals: - Company: Acme Financial (1,200 employees, [threshold] ARR, headquartered in New York) - Job title: VP of Marketing - Engagement: Visited product pages 3 times in last 7 days, downloaded a case study 2 days ago, opened 2 of your last 3 emails - Intent: Visited your pricing page 5 days ago, 2 of her colleagues at Acme have also visited your site in the last 30 days
Score calculation:
Fit: - Industry (Financial services): +30 - Company size (1,200 employees, in 500-2K range): +15 - Revenue ([pricing varies, check vendor website]M, in [pricing varies, check vendor website]M-100M range): +20 - Geography (US): +15 - Job title (VP): +10 - Fit subtotal: 90
Engagement (today): - Product page visit 3 days ago: +15 (fresh signal) - Case study download 2 days ago: +10 (no decay yet) - Email opens (2/3): +5 - Engagement subtotal: 30
Intent (today): - First-party intent (pricing page 5 days ago): +10 - Multiple company visitors (2 colleagues in last month): +5 - Intent subtotal: 15
Total score: 90 + 30 + 15 = 135
MQL threshold: 100 -> Jane is an MQL. Pass to sales.
---Skip the manual work
Abmatic AI runs targets, sequences, ads, meetings, and attribution autonomously. One platform replaces 9 tools.
See the demo →Implementation Checklist: 30/60/90 Day Plan
Week 1-2: Planning
- [ ] Align with sales on ICP definition
- [ ] Define MQL and SQL criteria
- [ ] Document your sales process
- [ ] List current lead sources (where do leads come from?)
- [ ] Get 12 months of historical lead data (spreadsheet or CRM export)
Week 3-4: Build Scoring (Fit + Engagement)
- [ ] Create Fit Score and Engagement Score custom fields in your CRM
- [ ] Document your fit criteria (industry, size, revenue, geography)
- [ ] Document your engagement signals and point values
- [ ] Build CRM workflows for fit criteria
- [ ] Build CRM workflows for engagement signals (demo request, page visit, content download)
Week 5-6: Test and Validate
- [ ] Run your scoring model on historical data (last 100 leads)
- [ ] Calculate MQL-to-SQL conversion for scored leads
- [ ] Compare to control group (non-scored leads)
- [ ] Ask sales: "Which of these scored leads looked good to you?"
- [ ] Adjust point values based on feedback
Week 7-8: Add Intent and Negative Scoring
- [ ] Add free intent signals (first-party website behavior, LinkedIn hiring signals)
- [ ] Add negative scoring (disqualifiers like unsubscribe, competitor announced)
- [ ] Re-run historical data with full three-layer scoring
- [ ] Set final MQL threshold
Week 9-12: Launch and Monitor
- [ ] Launch scoring to sales team
- [ ] Train sales on what an MQL is (they should understand the criteria)
- [ ] Monitor MQL-to-SQL conversion weekly
- [ ] Collect feedback from sales: "Which MQLs converted? Which didn't?"
- [ ] Monthly: Review signal performance. Adjust weights based on actual outcomes.
Common Implementation Mistakes
Mistake 1: Overly complex scoring. Creating 20+ signals makes scoring hard to maintain and difficult for sales to understand. Start with 5-8 signals. Add more only if you can measure them.
Mistake 2: Not communicating thresholds to sales. Sales gets MQLs without understanding what an MQL is. Result: Sales ignores scoring because "these don't look like good leads." Train sales on your criteria. Show them historical conversion by score tier.
Mistake 3: Setting it and forgetting it. You build scoring, launch it, and never update it. After 6 months, market changes, your ICP evolves, your product adds features. Your scoring becomes stale. Review monthly. Reweight quarterly. Rebuild annually.
Mistake 4: Not measuring impact. You implement scoring but don't track MQL-to-SQL conversion, win rate by score tier, or sales cycle length by score tier. Without measurement, you don't know if it's working. Measure everything.
Mistake 5: Conflating lead scoring with account scoring. Some businesses are account-centric (multiple buyers, one deal per account). For ABM, you should score accounts and leads separately. This guide assumes lead-centric. If you're ABM, apply similar logic to account-level signals instead.
Tools to Build Scoring
All-in-one (CRM-native): - HubSpot (workflows + custom fields, included in Hub) - Salesforce + Pardot (predictive scoring, [pricing varies, check vendor website])
Lightweight (spreadsheet + CRM integration): - Pipedrive + custom fields + Zapier - Freshsales + automation - Zoho CRM + workflows
Third-party scoring platforms: - Clearbit (data enrichment + lightweight scoring, [pricing varies, check vendor website]) - 6sense (predictive + intent, [pricing varies, check vendor website]) - HubSpot lead scoring tools (e.g., Privy)
For intent data integration: - Bombora (keyword intent, [pricing varies, check vendor website]) - G2 (review intent, free API to [pricing varies, check vendor website]) - 6sense (multi-signal intent, included in platform) - LinkedIn Sales Navigator (hiring signals, free or included in seat)
---Governance: Maintaining Your Scoring Over Time
Once you launch, maintain it.
Monthly: - Track MQL volume, SQL conversion rate - Note any spikes or dips - Review sales feedback
Quarterly: - Analyze which signals correlate with closes - Adjust point values - Test new signals
Annually: - Rebuild with latest closed-deal data - Update ICP if market changed - Evaluate third-party intent vendors
Trigger-based updates: - New product launch: Add signals for new product pages - New vertical entered: Add industry/vertical signals - New funding round: Adjust company size criteria - Competitor emerges: Add competitor signals to negative scoring
Bottom Line
Building a lead scoring framework takes 4-8 weeks from planning to launch. Start with fit (does this company match?), add engagement (is this person active?), add intent (are they buying?), and measure continuously. The payoff is significant: your sales team focuses on the highest-probability leads, conversion rates improve, and sales cycles shorten.
The most common mistake is launching and forgetting. Scoring requires quarterly reviews and annual rebuilds. But the effort is small compared to the payoff: a 20-30% improvement in sales efficiency, compressed deal cycles, and better pipeline predictability.
Ready to build your lead scoring framework? Schedule a demo with Abmatic AI to see how behavioral data and intent signals improve lead quality and sales productivity.
