Lead Scoring: Models & AI Rules

Lead scoring is the practice of ranking prospects with a number that combines how well they fit your ideal customer profile with how strongly they've shown buying intent. It tells sales who to call first and tells marketing who still needs nurturing. 

Quick answer: A lead score adds points for fit (industry, company size, job title) and engagement (pricing page visits, demo requests, email replies), subtracts points for disqualifiers (student emails, competitors, inactivity), and compares the total against thresholds. Leads above the MQL threshold go to sales development; leads above the SQL threshold get a call right away. Point values are yours to set, and the model should be tuned against what actually closes.

What Is Lead Scoring?

Lead scoring is a method for ranking leads by their likelihood to buy, using points assigned to the attributes and actions that matter for your business. Instead of working leads in the order they arrive, sales teams work them in the order of their scores. Marketing uses the same scores to decide who's ready for sales and who needs more nurturing.

Why score leads at all

Most B2B leads aren't ready to buy, and some never will be. Without a scoring system, SDRs spend time on whoever filled out the latest form: students researching a paper, competitors checking prices, and companies far outside your target market, alongside the few real buyers. Scoring sorts that mix so the real buyers get a fast call and the rest get an email sequence instead of a rep's hour.

Scoring also gives marketing and sales a shared definition of a good lead. When both teams agree on what earns points, arguments about lead quality turn into a discussion about point values, which is a much easier problem to solve.

How is lead score calculated?

Most lead scoring systems use a simple formula:

Lead score = Fit points + Engagement points - Negative points - Decay

•         Fit points come from explicit data about who the lead is: industry, company size, revenue, job title, and location.
•         Engagement points come from implicit data about what the lead does: pages visited, content downloaded, emails opened and replied to, webinars attended.
•         Negative points remove leads that look active but can't buy, such as competitors or personal email addresses.
•         Decay reduces engagement points over time so a lead who was active six months ago doesn't look as hot as one active this week.

A hypothetical example for a marketing director at a 300-person software company who visited the pricing page twice and attended a webinar:

Fit:     Target industry +10, 200-1,000 employees +15, Director title +15  = 40
Engagement:  Pricing page x2 +20, Webinar attended +10                    = 30
Negative: None                                                             =  0
Total score:                                                                   70

If the MQL threshold is 60, this lead goes to an SDR. The sections below explain where those numbers come from and how to tune them. If you're new to the funnel stages, our guide to MQLs and marketing qualified leads covers the vocabulary.

Lead Scoring Criteria: Fit, Behavior, and Negative Points

A good lead scoring model measures two separate things: whether the lead could buy, and whether they seem to want to. Fit answers the first question with firmographic and demographic data. Behavior answers the second with engagement data. Scoring them separately, then combining them, keeps a highly active student from outranking a quiet but perfect-fit buyer.

Fit: explicit data about who the lead is

Fit criteria come from your ideal customer profile (ICP), the description of the companies and people that buy from you and stay. Common fit attributes:

•         Industry: Points for the verticals where you win most often; zero or negative for the rest.
•         Company size: Employee count or revenue bands that match your best customers.
•         Job title and seniority: Decision-makers and influencers score higher than individual contributors.
•         Location: Regions you sell to and can support.
•         Technology used: Tools that signal a fit, such as a CRM you integrate with.

Fit data comes from form fields, enrichment tools, and your CRM. It changes slowly, so fit scores don't need decay. Our guide to what an ICP is in business explains how to build the profile these criteria come from.

Behavior: implicit signals of buying intent

Behavior criteria track what the lead does across your website, email, and events. Weight them by how close each action sits to a purchase decision:

•         High intent: Demo or contact requests, pricing page visits, ROI calculator use, replying to a sales email.
•         Medium intent: Case study downloads, webinar attendance, product page visits, repeat visits within a week.
•         Low intent: Blog visits, newsletter opens, social media follows.

Velocity matters too. Three pricing page visits in two days say more than three visits spread across a year, which is why most lead scoring software lets you score an action more than once up to a cap.

Negative scoring and point decay

Negative lead scoring removes points for signals that a lead can't or won't buy. Without it, scores inflate and SDRs waste calls. Typical negative criteria:

•         Personal email domains such as gmail.com or yahoo.com, when you sell to businesses
•         Competitor domains
•         Job titles like student, intern, or consultant, depending on your market
•         Visits to the careers page, which usually signal a job seeker
•         Unsubscribes or spam complaints
•         Companies outside your size range or in countries you don't serve

Point decay handles inactivity. HubSpot's lead scoring, for example, lets you lower a score by a chosen percentage after a chosen period, such as 50% every three months, so stale engagement stops qualifying leads. A lead who hit 70 last spring and hasn't been back since shouldn't still be an MQL today.

Lead scoring criteria examples with point values

The table below is a hypothetical starting matrix. Treat the numbers as a first draft to adjust against your own closed-won data.

Note: Cap the points any single behavior can earn. Without a cap, one person refreshing your pricing page twenty times becomes your "hottest lead" of the month.

Lead Scoring Models: Rule-Based vs. Predictive AI

There are two ways to decide how many points each signal is worth. In a rule-based model, people decide, using judgment and whatever data they have. In a predictive lead scoring model, software analyzes your past converted leads and sets the weights itself. Rule-based works for most teams starting out; predictive needs enough history to learn from.

Rule-based lead scoring

Rule-based scoring is the manual approach: marketing and sales agree on criteria and point values like those in the table above, configure them in the CRM or marketing automation platform, and review them every quarter. Its strengths are transparency and control. Anyone can see why a lead scored 70, and you can change a rule in minutes when the business changes.

Its weakness is that the weights are guesses, informed guesses at best. Teams often overweight the signals they can see, such as content downloads, and underweight the ones that predict revenue. The fix is a regular review: pull the last quarter's closed-won deals, look at what those leads did and who they were, and adjust the points toward those patterns.

Predictive and AI lead scoring

Predictive lead scoring uses machine learning to find the attributes and behaviors that separated past converters from non-converters, then applies those patterns to new leads. The models are typically classification algorithms, such as logistic regression or gradient-boosted decision trees, trained on your CRM's conversion history. The output is usually a probability or a score, such as 1 to 99, rather than a hand-built point total.

Predictive models need data. Salesforce's Einstein Lead Scoring, for example, requires at least 1,000 leads created in the last 200 days, with at least 120 of them converted, to build a model from your own data; below that it falls back to a global model built from anonymized data across Salesforce customers. That requirement is a useful rule of thumb: if you don't have at least a few hundred conversions, a rule-based model built from your team's judgment will likely serve you better than an algorithm trained on too little.

Predictive scoring also has a transparency cost. When a rep asks why a lead scored 85, "the model said so" is a harder answer than "they're a VP at a target account who requested a demo." Most predictive tools now show the top contributing factors for each score, which helps.

Automated lead scoring in the CRM

Whichever model you choose, the scoring itself should be automated. Modern CRMs and marketing platforms calculate scores continuously as new data arrives and let the score drive actions:

•         HubSpot supports fit scores based on property values, engagement scores based on events, and combined scores, with threshold labels that grade records from A to C for fit and 1 to 3 for engagement. Scores can trigger workflows, such as assigning an owner and notifying a rep when a contact reaches the high-engagement range.
•         Salesforce offers rule-based scoring through its marketing tools and predictive scoring through Einstein, which assigns a 1 to 99 score based on how closely a lead matches past conversion patterns.
•         Other marketing automation platforms follow the same pattern, typically with separate behavior and demographic scores that feed campaign triggers.

The automation pattern is the same everywhere: score changes, threshold crossed, record routed, task created, rep notified.

Thresholds, Routing, SLAs and Nurturing

A score only matters when it triggers an action. Thresholds turn the number into a status, routing sends the lead to the right owner, and a service level agreement (SLA) sets how fast that owner must act. Leads below the threshold go into nurturing until their score rises.

Typical B2B lead scoring thresholds

There's no universal scale, but most teams using a 100-point model land on bands like these:

•         0 to 30, cold: Early research or poor fit. Stay in automated nurturing.
•         31 to 60, warm: Fit is fine and interest is building. Nurture with targeted content; marketing watches for a jump.
•         61 to 80, MQL: Marketing qualified lead. Route to an SDR for review and outreach.
•         81 to 100, SQL candidate: Strong fit and high intent, or a direct hand-raise. Immediate call.

Two refinements make these bands work better. First, require a minimum fit score to become an MQL, so engagement alone can't push an out-of-profile lead to sales. Second, let hand-raises such as demo requests bypass the score entirely and go straight to a rep.

Set the bands with data, not instinct. Look at the scores of the last 100 deals you won and the last 100 you lost at the SQL stage. If most wins scored above 65 at handoff and most losses scored below, 65 is your MQL line. If you can't see a difference, your criteria need work before your thresholds do.

Lead routing and SLA timelines

When a lead crosses a threshold, the CRM should assign it, create a task, and start a clock. The clock matters because leads cool quickly. A Harvard Business Review study of 2,241 U.S. companies found that firms contacting leads within an hour were nearly seven times as likely to qualify them as firms that waited even an hour longer, and more than 60 times as likely as firms that waited 24 hours or more. The same study found that only 37% of companies responded within an hour, and the average response time was 42 hours.

A practical SLA ladder:

•         SQL candidates and hand-raises: First call attempt within 5 minutes during business hours, and within the first hour of the next business day otherwise.
•         MQLs: Accept or reject within 4 business hours; first contact attempt within 24 hours.
•         All routed leads: At least 6 to 8 attempts across phone and email over two weeks before closing out.

Routing rules should also handle territory, language, and account ownership. A lead from an existing customer's company should go to the account owner, not the inbound SDR queue.

Aligning lead scoring with lead nurturing

Lead scoring and lead nurturing are two halves of one system. Nurturing raises scores by prompting the behaviors you score; scoring tells nurturing who to stop emailing and hand to sales. Design them together:

•         Below the MQL line: Enroll leads in content sequences matched to their fit tier and the topics they've engaged with.
•         Crossing the line: Pause marketing emails when an SDR takes over, so the lead doesn't get a promo and a cold call on the same morning.
•         Rejected or stalled: When sales rejects an MQL or an SQL goes quiet, recycle the lead to nurturing with the reason code, and reset the engagement score so it has to re-earn the handoff.
•         Decay: Leads whose scores fall below the warm band drop into a lighter-touch sequence.

Note: The 5-minute SLA is a target for inbound hand-raises during staffed hours, not a universal rule. If your team can't meet it, the honest fix is more qualification capacity or longer coverage hours, not a looser SLA.

How to Build Lead Scoring and Scale Qualification With Prime BPO

Building a lead scoring system takes a few weeks of analysis and configuration, then ongoing tuning. The steps below take you from your first ICP audit to a working model, and cover what to do when scored leads arrive faster than your SDRs can call them.

Step 1: Audit your ICP against closed-won deals

Pull the last 12 to 24 months of closed-won and closed-lost deals. For each, record industry, company size, title of the main contact, lead source, and the behaviors logged before the first sales conversation. Look for the attributes that show up far more often in wins than in losses. Those become your highest-weighted fit criteria, and the behaviors that preceded wins become your highest-weighted engagement criteria. If your CRM data is too thin for this, start with sales' judgment and plan to revisit after a quarter.

Step 2: Draft the scoring matrix

Using the audit, write a point table like the one earlier in this guide. Keep it simple at first: no more than 10 to 15 fit criteria, 10 to 15 behaviors, and 5 to 8 negative rules. Add caps to repeatable behaviors and a decay rule. Get sign-off from both marketing and sales leadership before building anything.

Step 3: Configure scoring in your CRM

Set up the fit and engagement scores as separate properties, then a combined score. Build the threshold bands, routing rules, task creation, and notifications. Test with a sample of known leads: a few recent customers, a few obvious non-buyers, and a few borderline cases. If the obvious non-buyers score as MQLs, fix the negative rules before launch.

Step 4: Set the SLA and measure it

Agree on response times for each band and build the reports to track them: time to first contact, MQL acceptance rate, MQL-to-SQL conversion rate, and SQL-to-opportunity rate. Review weekly for the first two months.

Step 5: Add qualification capacity when scored leads outrun your team

Scoring often exposes a capacity problem. Once the model is working, it surfaces more MQLs than in-house SDRs can call within the SLA, and the leads at the bottom of the queue go cold. At that point you can hire, narrow the thresholds, or add an outsourced qualification team.

Prime BPO's lead generation services cover outreach by phone, email, and social media, qualification based on interest, budget, and decision-making authority, appointment setting for qualified leads, and keeping lead details organized in your CRM. An outsourced team can take the first-touch and qualification work for MQLs, so in-house reps spend their time on SQLs and meetings. To set it up:

1.       Share your ICP, scoring matrix, and SQL criteria so the team qualifies against the same rules your model uses.
2.       Give the team CRM access to log every attempt, outcome, and rejection reason, which also feeds your scoring reviews.
3.       Agree on the SLA for first contact and attempts per lead.
4.       Pilot on one segment for 30 to 60 days and compare contact rate, acceptance rate, and meetings booked against your in-house baseline.
5.       Scale what works and feed rejection reasons back into the scoring model.

Prime BPO operates from six offices in Rawalpindi, Pakistan, and holds ISO 18295-1:2017 certification for customer contact services. Its lead generation page lists technology and SaaS, financial and insurance services, healthcare, and ecommerce among the industries it supports. For more on how qualification roles fit together, see our guides to what an SDR does in sales and SDR vs. BDR.

Note: Calling U.S. business leads is regulated. The FTC's 2024 Telemarketing Sales Rule amendments extended the ban on misrepresentations to business-to-business calls, and FCC rules under the TCPA require consent for autodialed or prerecorded calls to wireless numbers. Any calling team, in-house or outsourced, needs to follow them.

Conclusion

Lead scoring works when it measures two things separately, fit and intent, subtracts for the signals that disqualify, and connects the resulting number to a fast, defined action. Start with a rule-based model built from your own closed-won data, review it quarterly, and move to predictive scoring once you have enough conversions to train on. Then make sure someone can call the leads the model surfaces, within the SLA, every day.

Scale Lead Qualification With Prime BPO

If your scoring model is surfacing more qualified leads than your team can reach, scale lead qualification with Prime BPO. Prime BPO's team can handle first-touch outreach, qualification calls, and appointment setting against your scoring criteria, so your reps spend their time on meetings.

•         Explore the service: Prime BPO lead generation services
•         Business development support: Prime BPO business development
•         Request a pipeline review: Contact Prime BPO
•         Call: +92-51-8733177

Frequently Asked Questions

How do you score a lead?

Start with your ideal customer profile and list the attributes that describe your best customers: industry, company size, title, and location. Assign points to each. Then list the behaviors that show buying interest, such as pricing page visits, demo requests, and webinar attendance, and assign points weighted by intent. Add negative points for disqualifiers like competitor domains or personal email addresses, and a decay rule for inactivity. Total the points, set thresholds for MQL and SQL, and adjust the values every quarter against the leads that actually became customers.

How is lead score calculated?

Most systems use the formula: fit points plus engagement points, minus negative points and decay. Fit points come from who the lead is, engagement points from what they've done, and negative points from signals that they can't or won't buy. A lead at a target company with a director title who visited the pricing page twice and attended a webinar might score 70 on a 100-point scale, which would make them an MQL under a 60-point threshold. Predictive models calculate a probability instead, using machine learning trained on past conversions.

What is a lead scoring model?

A lead scoring model is the set of rules, or the trained algorithm, that converts lead data into a score. Rule-based models use point values that people assign. Predictive models learn the weights from historical conversion data. Many teams visualize the model as a two-dimensional matrix, with fit on one axis and engagement on the other: high-fit, high-engagement leads go to sales now, high-fit, low-engagement leads get nurtured, and low-fit leads are deprioritized regardless of engagement.

How is lead scoring used in the CRM?

The CRM stores the score as a property, recalculates it as new data arrives, and uses it to trigger actions. Common automations include assigning an owner when a lead crosses the MQL threshold, creating a follow-up task with a deadline, sending an alert for high scores, adding leads to segmented lists for nurturing, and pausing marketing emails once sales takes over. HubSpot, for example, lets scores trigger workflows and grades records with threshold labels such as A1 for high fit and high engagement, while Salesforce's Einstein assigns a 1 to 99 predictive score.

What are typical lead scoring thresholds in B2B?

On a 100-point scale, a common setup is cold at 0 to 30, warm at 31 to 60, MQL at 61 to 80, and SQL candidate at 81 to 100, with demo requests bypassing the score. The exact lines should come from your own data: compare scores of recent won and lost deals at the handoff point, and set the MQL threshold where the two groups separate.

Sources and 2026 Verification

All figures and product details were checked against the sources below in November 2026.

•       HubSpot Knowledge Base: build lead scores: fit, engagement, and combined scores, threshold labels, decay, and workflow triggers (updated August 2026)
•       Salesforce: Sales Cloud Einstein implementation guide: Einstein Lead Scoring data requirements, 1 to 99 scoring, and global model fallback
•       Harvard Business Review: The Short Life of Online Sales Leads: lead response time research by Oldroyd, McElheran, and Elkington
•       Prime BPO: lead generation services: service scope, industries, and ISO 18295-1:2017 certification
•       Federal Register: Telemarketing Sales Rule amendments (April 16, 2024): business-to-business misrepresentation ban
•       eCFR: 47 CFR 64.1200: TCPA rules on autodialed and prerecorded calls

This article is for general information and isn't legal advice. Point values, thresholds, and SLA targets shown are hypothetical starting points, not industry standards, and should be tuned to your own conversion data. Software features change, so confirm current capabilities with your CRM vendor.