B2B Lead Scoring Models That Actually Work

Most lead scoring models fail because they score the wrong things. Teams assign 10 points for a whitepaper download, 5 points for a website visit, and 25 points for a demo request. Then they set a threshold at 50 points and call everything above it 'sales-ready.' The result: marketing celebrates a pile of MQLs, sales complains the leads are garbage, and nobody trusts the model.

The core problem is that traditional scoring treats all activity as buying intent. A prospect who downloads your pricing PDF is probably evaluating vendors. A prospect who downloads your annual industry report might just be curious. A competitor doing research will trigger every engagement score in your system. Activity does not equal intent, and scoring models that treat them as equivalent produce false positives at scale.

The second failure mode is over-indexing on demographic data. 'VP at a 200-person SaaS company' gets a high score because it matches your ICP. But that VP might be at a company that just signed a 3-year contract with your competitor. Fit without timing is wasted effort. A strong lead scoring model combines fit (do they match your ICP), intent (are they showing buying behavior), and engagement (are they interacting with your outreach).

The teams that get lead scoring right treat it as a prioritization tool, not a qualification gate. The score does not tell you whether a prospect will buy. It tells you which prospects deserve your limited selling time right now. That framing changes everything about how you build and calibrate the model.

Fit measures how closely a prospect matches your Ideal Customer Profile. This is the most stable dimension of your scoring model. A company's size, industry, tech stack, and funding stage do not change week to week. Fit scoring tells you whether a prospect is worth pursuing at all, regardless of timing.

Intent measures whether a prospect is actively evaluating solutions in your category. Intent signals include G2 or Capterra visits, searches for your competitors, job postings for roles your product supports, and content consumption patterns that indicate a buying journey. Intent is the most valuable dimension because it separates 'would buy someday' from 'is buying now.'

Engagement measures whether a prospect is interacting with your specific outreach and content. Email opens, link clicks, LinkedIn profile views, website visits, and webinar attendance all count. Engagement is the most volatile dimension, and it should be weighted accordingly. A prospect who opened three emails this week is more likely to respond than one who has been silent for 60 days.

The model works best when these three pillars are scored independently and then combined with explicit weights. A typical starting point: Fit 40%, Intent 35%, Engagement 25%. This means a perfect ICP-fit company showing strong intent signals but no engagement yet still scores high enough to pursue. Adjust the weights based on your data. If your best customers consistently show high intent before first contact, increase the intent weight.

Start with your ICP document. Every criterion in your ICP should map to a scoring attribute. Company size, industry, tech stack, growth rate, funding stage, geographic location, and organizational structure. Each attribute gets a weight based on how strongly it predicts deal success in your historical data.

Here is a concrete example for a sales engagement platform. Company size 80-400 employees: +30 points. Series B or C funding: +20 points. SaaS, FinTech, or MarTech vertical: +15 points. Uses Salesforce CRM: +15 points. Sales team of 8+ reps: +20 points. US or UK headquartered: +10 points. Headcount grew 20%+ in past year: +15 points. Under 50 employees: -40 points. No dedicated sales team: -30 points. Total possible fit score: 125 points.

Negative scores matter as much as positive ones. If companies under 50 employees churn at 3x your average rate, actively penalize that attribute. If companies in certain industries consistently stall in evaluation, penalize those too. Without negative signals, your model will over-score prospects that look good on paper but fail in practice.

Calibrate the model against your closed-won and closed-lost data quarterly. Run your top 20 wins and top 20 losses through the fit score. If the model cannot distinguish between the two groups, your weights are off. Adjust until the median fit score for wins is at least 40% higher than for losses. The GTMS ICP builder generates fit scores automatically from the criteria you define.

Intent data comes in two flavors: first-party and third-party. First-party intent is behavior on your own properties, like visiting your pricing page, reading case studies, or watching a product demo video. Third-party intent is behavior outside your properties, like searching for your category on G2, reading competitor reviews, or posting about a relevant problem on LinkedIn.

First-party intent signals and their typical point values: Pricing page visit: +25. Case study page visit: +15. Product documentation visit: +10. Blog post read (2+ posts in a week): +5. Webinar registration: +15. Webinar attendance: +25. Demo form submission: +40. Each signal decays over time. A pricing page visit from 3 days ago is worth more than one from 30 days ago. Apply a decay factor: full value within 7 days, half value at 14 days, zero at 30 days.

Third-party intent is harder to capture but more predictive. If a company is researching your category on review sites, they are likely in an active buying cycle. Providers like Bombora, G2 Buyer Intent, and TrustRadius provide this data. The challenge is signal-to-noise ratio. Third-party intent data at the company level does not tell you which person is doing the research. Combine it with your fit score and contact-level engagement to identify the right individual.

Job postings are an underrated intent signal. A company posting for an 'SDR Manager' or 'Revenue Operations Analyst' is building or scaling their sales function. That is a strong buying signal for sales tools. A company posting for a 'Head of Marketing' after their previous one left is likely re-evaluating their marketing stack. GTMS monitors 44 distinct buying signals, including job postings, funding events, tech stack changes, and leadership transitions, and incorporates them into lead scores automatically.

Engagement scoring measures how a prospect responds to your direct outreach. This is the most actionable dimension because it tells your reps exactly who to follow up with today. A prospect who clicked a link in your email this morning is a warmer conversation than one who has not responded to three touches.

Email engagement tiers: Open only (no click): +3 points per open. Link click: +10 points per click. Reply (any sentiment): +25 points. Positive reply: +50 points. Meeting booked: +100 points. LinkedIn engagement: Connection request accepted: +15 points. Message read (if visible): +5 points. Profile viewed your profile: +10 points. Liked or commented on your post: +20 points.

The key insight with engagement scoring is recency weighting. An email click from yesterday is worth 10x more than a click from three weeks ago. Implement aggressive time decay: full value for 0-3 days, 50% for 4-7 days, 25% for 8-14 days, and 10% for anything older. This ensures your scoring reflects current interest, not historical curiosity.

Watch for false engagement. Email auto-openers, security scanners, and link preview bots can inflate your open and click metrics. If a prospect registers 8 opens in 30 seconds, that is a bot, not a human. Filter these out or your engagement scores become meaningless. Most modern email tools flag automated opens; make sure your scoring system excludes them.

One pattern that works well: create a composite engagement velocity score. Instead of looking at total points, look at the rate of engagement over the last 7 days. A prospect who went from zero engagement to three interactions in a week is surging. That surge, not the absolute score, is what your reps should act on.

Scoring a company is only useful if you can identify and reach the right person. At most B2B companies, buying decisions involve 3-7 stakeholders. Your job is to find the economic buyer (who signs the check), the champion (who drives the evaluation), and the technical evaluator (who validates the product). See our Sales Navigator guide for the tactical search process.

Role-based scoring helps prioritize contacts within an account. CRO or VP of Sales: +30 points (likely economic buyer). Director of Sales or Revenue Operations: +25 points (likely champion). Sales Manager or SDR Manager: +20 points (likely user or champion). SDR or BDR: +5 points (end user, lower decision authority). Adjust these values based on where your deals actually close. If your champion is typically a Director, not a VP, weight Directors higher.

Seniority alone is not enough. A VP of Sales at a 50-person startup has different authority than a VP of Sales at a 5,000-person enterprise. Context matters. Cross-reference role with company size, department size, and reporting structure. A person who is the most senior sales leader at their company (regardless of title) carries more weight than someone three levels below the CRO.

Multi-threading is the tactical implication of decision-maker scoring. Do not put all your outreach effort into one contact per account. Sequence the economic buyer and the champion simultaneously. If your champion responds and your VP-level contact does not, the champion can make an internal introduction. GTMS supports multi-contact sequencing within the same account, so you can coordinate touches across stakeholders without duplication or conflict.

Lead scores become actionable when they drive segmentation. Instead of one generic outbound sequence, use your scores to assign prospects to different sequences with different messaging, cadence, and channel mix. This is where the ROI of lead scoring materializes.

A four-tier segmentation works for most teams. Tier 1 (score 80+): Highest fit, strong intent, recent engagement. These get your best sequence with maximum personalisation, multi-channel touches, and same-day follow-up on replies. Your top reps work this tier. Tier 2 (score 60-79): Strong fit, moderate intent or engagement. Standard multi-channel sequence with signal-based first lines. Tier 3 (score 40-59): Moderate fit, low or no intent signals. Email-only nurture sequence with educational content. Check back in 30 days. Tier 4 (below 40): Low fit or actively disqualified. Do not sequence. Add to marketing nurture or remove entirely.

The mistake most teams make is putting everyone into the same sequence. When a perfect-fit, high-intent VP gets the same generic 5-email drip as a barely-qualified coordinator, you waste your best opportunities and annoy everyone else. Segmentation is not optional; it is the mechanism that makes lead scoring useful.

Reassess segments dynamically. A Tier 3 prospect who suddenly shows intent signals (visited your pricing page, their company raised a round) should move to Tier 1 or 2 immediately. Static segmentation is barely better than no segmentation. The scoring model should run continuously, and segment changes should trigger automatic sequence transitions. GTMS handles this with real-time signal monitoring that automatically adjusts scores and re-routes contacts to the appropriate sequence.

Manual lead scoring works when you have 200 prospects. It breaks at 2,000. At 20,000, it is physically impossible. AI-powered scoring solves the scale problem by processing signals, enrichment data, and engagement events in real time and outputting a score that updates continuously.

The first generation of AI scoring was rule-based: if title contains 'VP' and company size is 100-500, add 30 points. That is just automation, not intelligence. Modern AI scoring uses your historical conversion data to learn which combinations of attributes and behaviors actually predict closed deals, including non-obvious correlations that humans miss. Maybe companies that use both Salesforce and Slack close 2x faster than Salesforce-only companies. A rule-based system would never discover that. A trained model will.

The risk with AI scoring is the black box problem. If your reps do not understand why a prospect scored high, they will not trust the model and will ignore it. The best AI scoring systems provide explainability: 'This prospect scored 87 because: ICP fit 92% (Series B SaaS, 180 employees, uses Salesforce), intent signal detected (2 G2 visits this week), engagement surge (opened 3 emails in 2 days).' Transparent scoring drives adoption. Opaque scoring drives skepticism.

GTMS uses a multi-layer scoring approach. Fit scores are calculated from your ICP definition. Intent scores are generated from 44 monitored signals, processed through AI to assess confidence and recency. Engagement scores are computed from outbound interaction data in real time. The three layers combine into a single priority score that your team can sort by, filter on, and trust. Try the free lead scorer to see how it works with your own data, or visit pricing to explore full platform access.