What Is Lead Scoring? Models, Criteria & How to Set It Up

Lead scoring is a system for ranking prospects by assigning point values to the traits and behaviors that predict whether someone will buy — so your reps spend time on the leads most likely to convert, not the ones who downloaded a whitepaper once and disappeared. Done well, it turns a chaotic inbound funnel into a prioritized queue. Done poorly, it creates false confidence and lets real buyers slip through.

What is the lead scoring definition in B2B sales?

Lead scoring is the process of assigning a numeric value to each lead based on how closely they match your ideal customer profile and how much buying intent they've demonstrated. The total score determines how urgently a rep should follow up — or whether the lead should stay in a nurture sequence until they're ready.

In practice, a lead might earn 10 points for matching your target company size, 15 points for being a VP-level contact, and 20 more points for requesting a demo. That same lead loses points if they work in an industry you rarely close or if they're outside your target geography. The net score tells your team where to focus.

The core purpose is triage. Most B2B funnels generate far more leads than reps can work manually. Without scoring, reps default to recency (whoever came in last gets called first) or volume (whoever opened the most emails). Neither correlates well with revenue. Scoring routes effort to likelihood.

What criteria should you use to score leads?

Lead scoring criteria fall into two categories: fit and intent. Fit tells you whether this company could ever be a customer. Intent tells you whether they're ready to buy now.

Fit criteria (who they are)

• Company size: Does the prospect's headcount or revenue match the segment you close? Assign higher points to your sweet spot, lower to outliers.
• Industry: If you win 80% of deals in SaaS and FinTech, those verticals score higher than manufacturing or retail.
• Job title and seniority: Economic buyers (VP, C-suite) score higher than end users. Decision-maker contact = more points.
• Geography: If you only serve North America, EMEA leads score lower — not zero, but deprioritized.
• Tech stack: Companies using tools that integrate with yours (or that you replace) are a stronger fit signal than companies with no visible stack overlap.
• Competitor usage: A company actively paying a competitor has already validated the problem and the budget. This is one of the most underused fit signals in B2B lead scoring models.

Intent criteria (what they've done)

• Demo or trial request: The highest-intent action in most funnels — score this heavily (20–30 points).
• Pricing page visit: A strong buying signal. Anyone who looks at pricing is comparing options.
• Multiple content downloads: One download is curiosity. Three downloads over two weeks is research.
• Email reply or response: Engagement beats passive open rates as an intent signal.
• Return visits: Someone who visits your site four times in a week is not casually browsing.

Most teams also apply negative scoring — subtracting points for signals that indicate poor fit or low intent: personal email addresses, competitor domains, student accounts, or contacts who unsubscribed from email.

What are the main lead scoring models?

There are three models in common use. Which one fits your team depends on your data maturity, CRM capabilities, and how much time you're willing to invest in setup.

Rule-based lead scoring

You define the rules manually: this job title gets 15 points, this page visit gets 10, this company size gets 20. It's simple to set up and easy to explain to stakeholders. The weakness is that the weights are based on assumptions, not evidence — your team might heavily weight email opens when, in reality, pricing page visits are what actually predicts conversion.

Demographic + behavioral (two-axis scoring)

A more structured version of rule-based scoring that separates fit from intent explicitly. You score each lead on two axes — how well they match your ICP and how active they've been — then plot them in a 2×2 matrix. High fit + high intent = call immediately. High fit + low intent = nurture. Low fit + high intent = low priority. Low fit + low intent = remove from sequence. This model is more actionable than a single score and easier to operationalize across a team.

Predictive lead scoring

Machine learning models trained on your historical CRM data — specifically, patterns in closed-won vs. closed-lost deals. Instead of guessing which signals matter, the model identifies which combinations of attributes and behaviors actually correlated with revenue in your specific business. According to Gartner, organizations using AI-driven lead scoring report up to 30% higher conversion rates than those using static rule-based models, because the model self-corrects as your customer base evolves.

The catch: predictive scoring requires clean, historical CRM data — typically 500+ closed deals minimum — and a platform that supports it (Salesforce Einstein, HubSpot's predictive scoring, or dedicated tools like MadKudu). It's not a starting point for early-stage teams.

"The biggest mistake I see in lead scoring is teams assigning points based on what feels important rather than what their data shows. We rebuilt our model from closed-won data alone and cut our average sales cycle by three weeks because reps stopped working the wrong leads."

— VP of Sales, 80-person B2B SaaS company

How do you set up a B2B lead scoring model from scratch?

The fastest way to build a model that actually works is to start with your closed-won deals, not with a blank spreadsheet of attributes you think should matter.

Step 1: Audit your last 50 closed-won deals

Pull every closed-won deal from the last 12 months. Look for patterns: What company sizes appear most often? Which job titles were the economic buyer? Which pages did they visit before converting? Which content did they download? What was their time-to-close? These patterns are your scoring inputs — they're derived from evidence, not assumption.

Step 2: Define your scoring categories and point values

Map the patterns you found to scoring categories. A simple starting model might look like this:

Step 3: Set your MQL threshold

Your MQL (Marketing Qualified Lead) threshold is the score at which a lead gets routed to a sales rep. Set it by calculating the average score of your closed-won deals using your new model, then position the threshold slightly below that. If your average closed-won deal scored 68 points, a threshold of 55–60 captures similar leads before they've gone through the full buying process.

Step 4: Run it for 90 days, then recalibrate

No model is right on day one. After 90 days, compare the score distribution of leads that converted vs. those that didn't. Adjust weights accordingly. The model gets sharper with iteration — the first version just needs to be good enough to be better than no model at all.

What are the strongest intent signals in B2B lead scoring?

Not all signals carry equal weight. Salesloft's outreach benchmarks consistently show that leads who engage with bottom-of-funnel content (pricing pages, case studies, ROI calculators) convert at 3–5x the rate of leads who only interact with top-of-funnel content like blog posts or social ads. That should be reflected directly in your point allocation.

The three signals that most reliably predict near-term conversion in B2B are: a demo or trial request, a pricing page visit, and a direct reply to a sales email. If a lead has done all three, they should be at the top of the queue regardless of other factors.

One signal that's often missing from standard lead scoring models: whether the prospect's company currently uses a direct competitor. This tells you they have budget allocated, they've already decided the problem is worth solving, and they're in an active vendor relationship that can be disrupted. Tools like Stealery surface exactly this — you search a competitor name and get a list of companies actively using it, which you can then cross-reference against your inbound leads to boost the score of any that match.

When should you use predictive lead scoring instead of manual models?

Predictive lead scoring becomes worth the investment when three conditions are met: you have at least 500 closed deals in your CRM with consistent data, your manual model is producing MQLs that reps are consistently ignoring or deprioritizing, and you have the technical resources to implement and maintain a machine learning model.

Before those conditions are met, a well-built manual model outperforms a poorly configured predictive one. The advantage of predictive scoring is not that the algorithm is smarter — it's that it removes the cognitive bias baked into manually assigned weights. Humans consistently overweight the signals they're most aware of (email opens, for instance) and underweight signals that are harder to see (return site visits, specific page sequences).

If you're a team of two to five SDRs working a defined ICP, a two-axis manual model with quarterly recalibration will serve you better than a predictive system you don't have the data or tooling to maintain properly.