AIViral logoAIViral
← Back to blog

Lead Scoring Models: Types, Examples and How to Build One

Analyst reviewing prospect scoring records

A lead scoring model ranks prospects with numbers so your sales team knows who to call first. It combines who the lead is (fit) with what the lead does (behavior), adds points for good signs, subtracts points for bad ones, and sends anyone above a set threshold to sales. This guide covers the main model types, sample point values, how predictive scoring differs from rules-based scoring, and how to build and maintain a model that reps actually believe.

Aiviral
Skip Scoring and Reach Interested Leads
Aiviral uses AI-driven outreach to deliver B2B leads who actively request meetings or pricing, so interest is already clear.
Explore Aiviral

TL;DR

  • A lead scoring model assigns numerical values to attributes and actions, so reps work the most sales-ready leads first instead of going by gut feeling.
  • Good models mix explicit fit data (title, industry, company size) with implicit behavior (pricing page visits, demo requests, email replies), and they subtract points for disqualifiers.
  • Start simple: a points-based model on a 0 to 100 scale with one threshold. Move to predictive scoring only when you have enough clean CRM history.
  • Review the model every quarter against closed deals. A score nobody validates drifts away from reality.
  • If you would rather skip scoring and buy only leads who already asked for a meeting or pricing, that is what we do at AIViral.

Table of Contents

What a lead scoring model does

A lead scoring model is a systematic way of putting a number on how likely each prospect is to become a customer. A ZoomInfo guide describes it as assigning values, typically 1 to 100, to prospect attributes and behaviors so B2B teams can prioritize the most sales-ready people. HubSpot’s guide frames the same idea as a value reflecting how likely a lead is to buy, which lets a company focus its effort where conversion is most probable.

The practical payoff is time. Salesforce’s State of Sales findings, as summarized on its blog, say reps spend 9% of their time researching, 8% prospecting and 8% prioritizing leads. Scoring compresses the prioritizing part: the list arrives already sorted.

A second payoff is shared language. Oracle’s guide calls scoring an objective ranking of one lead against another, which lets teams match follow-up to buyer readiness rather than rely mainly on subjective judgment. When marketing and sales agree on what a 70 means, handoff arguments get shorter. Twilio’s guide adds that scores sort leads into priority groups, so sales can focus first while marketing optimizes campaigns for qualified leads instead of raw volume.

The main lead scoring models

Most models you will see are variations on five building blocks. You can combine them, and most working models do.

Explicit (fit) scoring

Explicit data is information you know about the person or company: industry, company size, revenue, tech stack, job title, seniority and geography. These attributes tell you whether a lead resembles your ideal customer profile (ICP). A VP of Operations at a 500-person manufacturer scores high if that is who buys from you. A freelance student does not.

Implicit (behavioral) scoring

Implicit data is what the lead does: pricing page visits, demo requests, content downloads, email engagement, webinar attendance and navigation patterns. HubSpot’s lead scoring guide lists these as indicators of buying interest. Weighting matters: a pricing page visit should be worth more than a blog view, and a demo request more than a webinar signup.

Signals can start well before a visit. Paid search is one source: clicks on high-intent terms in B2B Google Ads campaigns are an early behavioral signal you can feed into the score alongside fit data.

Negative scoring and decay

Positive points alone inflate every score over time. Adobe’s lead scoring guide recommends deducting points for disqualifiers such as unsubscribes, entry-level titles, students, personal email domains and prolonged inactivity. Decay does the same job gradually: a pricing visit from six months ago should not count like one from yesterday.

Scoring thresholds

A threshold is the number at which a lead changes status. Below it, the lead is nurtured. At or above it, the lead becomes a marketing-qualified lead (MQL) and moves toward sales. Our piece on MQL vs SQL handoff rules goes deeper on what should happen after a lead crosses that line.

Multiple scores

One number does not fit every situation. According to Clearbit’s lead scoring basics, companies may keep separate scores for different products, customer segments or go-to-market motions, and one approach is a letter grade for fit alongside a numeric score for intent. The same source cautions that scoring data from one market may not carry over to a new market or motion.

That last point is the difference between lead scoring and lead grading. Scoring usually measures interest and activity. Grading measures how well the lead matches your ideal profile. A lead graded A with a score of 10 is a great company that has not engaged yet. A lead graded D with a score of 90 is an enthusiastic student. Seeing both tells reps what to do with each.

Lead scoring examples with point values

There is no universal point table, and anyone who sells you one is guessing. What you can borrow is the structure.

A tiered scale. The HubSpot guide cited above suggests 10 to 15 points for critical attributes, 5 to 9 for important ones, 1 to 4 for influential ones, and negative points for disqualifiers. It proposes a sales-ready threshold such as 75 or higher.

Explicit negatives. The Adobe guide’s sample model takes 50 points off for an unsubscribe and 10 off for an entry-level title, while giving 15 points for a Google search of a product name.

A common range. A ZoomInfo guide notes that points-based models often use a 0 to 100 scale where something like 70 points defines an MQL, with lower scores triggering nurturing.

Here is an illustrative model for a B2B software company selling to mid-market operations teams. The numbers are our own example built on the tiered structure above, not a benchmark. Adjust everything to your data.

Signal Type Points
Title is director level or above Fit +15
Company in a target industry Fit +12
Company size within ICP range Fit +10
Visited pricing page Behavior +15
Requested a demo or meeting Behavior +25
Replied to outbound email with interest Behavior +25
Opened three or more emails in 30 days Behavior +5
Attended a webinar Behavior +8
Personal email domain Negative -10
Student or job seeker Negative -20
Unsubscribed Negative -50
No activity for 60 days Decay -15

A director in a target industry at the right-size company starts at 37 from fit alone. Add a pricing visit and a meeting request and she reaches 77, past a threshold of 75, and goes to a rep. A student who downloads three ebooks and attends a webinar might reach 30 after negatives and stays in nurture.

Notice what dominates: a direct request for contact outweighs everything passive. That matches the weighting advice from the research: intent signals count for more than casual engagement. If you want a starting split for a balanced model, one 2026 expert guide suggests roughly 40% fit, 35% intent and 25% engagement, then adjusting from quarterly reviews of which attributes correlate with closed deals. Treat it as a hypothesis to test, not a rule.

Lead scoring examples with point values — overview diagram

Fit or behavior: which should weigh more?

Teams disagree on this, and both camps have a point. Fit-heavy models protect reps from chasing people who will never buy. A DataOps Group article argues the other side: recent pricing visits, demos and sustained engagement can show buying readiness better than seniority alone, so behavior should not be buried under static profile data.

A workable resolution is to use fit as a gate and behavior as the accelerator. A lead below a minimum fit level never reaches sales regardless of clicks. Among leads that pass the gate, behavior decides who goes first and how fast. Our guide to intent data for B2B sales and marketing explains how to match different signals to different sales moves, and lead qualification frameworks like BANT, CHAMP and MEDDIC show what a rep should confirm once a high-scoring lead picks up the phone.

Fit threshold filters leads before behavior prioritizes follow-up

Predictive lead scoring

Predictive scoring replaces hand-picked points with a machine learning model. Instead of you deciding that a pricing visit is worth 15, the model studies historical CRM data, finds which attributes and behaviors preceded closed deals, and weights them accordingly. The ZoomInfo guide says it needs clean, substantial historical data and ongoing training, while rules-based scoring depends on human-defined points and constant manual refinement.

The tradeoffs are straightforward:

Rules-based Predictive
Who sets weights You and your sales team The model, from past outcomes
Data needed Little; can start with judgment Substantial, clean conversion history
Transparency Easy to explain to a rep Harder to explain why a lead scored high
Maintenance Manual reviews Retraining and data hygiene
Best for Early teams, new markets Mature teams with many closed deals

If you have a few dozen closed deals, a predictive model has little to learn from. Start rules-based, record outcomes carefully, and graduate when the data supports it. Even then, keep a simple explainable layer on top. Reps trust a score they can read.

How to build a model your sales team trusts

  1. Define the ICP with sales in the room. Oracle’s guidance is to jointly define fit and engagement with both sales and marketing. List the firmographics and titles of your best ten customers and the ones who churned fast.
  2. Pull the real patterns. Look at what your closed-won accounts did before buying: which pages, which emails, how many touches. Ignore what you assume matters.
  3. Pick 8 to 15 criteria. More than that and nobody can explain the score. Use the tiered scale: few critical items, several important ones, a handful of minor ones.
  4. Add negatives and decay. Unsubscribes, personal domains, students and inactivity all need subtraction.
  5. Set one threshold and map it to action. Oracle recommends mapping score or rating bands to actions such as priority CRM follow-up or continued nurturing. Below the line, leads move into lead nurturing strategies rather than being discarded.
  6. Agree on response times. A high score loses value if the lead waits two days. See our note on speed to lead for why the first minutes matter.
  7. Track acceptance. Measure how many scored leads sales accepts and why it rejects the rest. Our write-up on sales accepted lead metrics covers the intake numbers that expose handoff gaps.
  8. Review quarterly. Rocket’s guide to building a trusted scoring model recommends checking each quarter whether top-scoring leads are truly the ones converting, then updating weights to match current sales performance. Adobe’s guide similarly suggests reviewing won and lost opportunities and whether top prospects have top scores.

If you also need to know which channels deserve credit for the leads that scored well, three lead attribution models matched to your data pick up where scoring stops.

Common mistakes

  • Scoring every click equally. A blog view and a pricing page visit are not the same intent.
  • No negative points. Scores only climb, and the threshold stops meaning anything.
  • Built by marketing alone. Reps ignore a number they had no part in shaping.
  • Never recalibrated. Products, markets and buyers change. A model built for one segment can mislead in another.
  • Treating the score as a verdict. It is a priority queue. A rep’s judgment on a call still counts.

Our perspective: scoring predicts interest, replies prove it

We run lead generation for a living, so we see scoring from the other side. A score is an estimate. It says this lead looks like someone who might buy. It cannot tell you that a decision-maker just wrote back asking for pricing.

That gap is why we built AIViral around verified replies rather than predicted ones. Our system runs outreach across cold email, LinkedIn, social channels and search ads, checks real-time replies from decision-makers, and a human reviews every response before a lead reaches your dashboard. Only prospects who asked for a meeting, call, pricing, info or a demo count. In scoring terms, we deliver leads at the top of the scale, where the strongest behavioral signal, a direct request, has already happened.

This does not make scoring obsolete. If you generate inbound traffic, run webinars or maintain a large database, you need a model to sort it. But for outbound, we think teams over-engineer scores for leads who never expressed anything. A smaller, simpler model plus a stream of leads who raised their hands usually beats a sophisticated model run over cold records.

Our advice: score what you must, keep it simple enough to explain in one sentence, and give heavy weight to explicit requests for contact over anything passive.

— Max

Skip the scoring for outbound: pay only for interested leads

If your bottleneck is not sorting leads but getting interested ones, AIViral handles the outreach and the ad spend, and you pay only when a prospect replies with clear interest. Pricing is per lead with no setup fees or contracts: Warm leads are $5 to $15, Hot leads $15 to $30, and On fire leads $30 to $50, depending on the level of interest. If a lead does not meet the interest standard, you can dispute it and get a credit.

Aiviral

We also build targeted programs by sector, including IT services and MSPs, web and software development and financial advisors. You can read how reply verification works in our guide to AI lead generation for B2B.

FAQ

What is the best lead scoring model for a small B2B team?

A simple points-based model with fit and behavior criteria, a handful of negative scores and one threshold. It is cheap to build, easy to explain and needs no historical data volume. Add complexity only after quarterly reviews show where it fails.

What is a good lead score threshold?

There is no universal number. Published examples include 70 for an MQL on a 0 to 100 scale and 75 or higher as a sales-ready level. Choose yours by looking at where past leads started converting in your own data, then test it each quarter.

How is predictive lead scoring different from a points-based model?

Points-based scoring uses weights people choose. Predictive scoring uses machine learning on past conversion data to find the weights. It needs substantial clean CRM history and ongoing retraining, so it suits teams with many closed deals.

How often should I update my lead scoring model?

Review it at least quarterly. Check whether the highest-scoring leads are the ones that actually close, compare won and lost opportunities, and change weights where the score and the outcomes disagree.

Do I need lead scoring if I buy interest-verified leads?

Less than with raw lists. Leads who already requested a meeting or pricing have shown the strongest signal, so most can go straight to a rep. You may still score for fit, so your best accounts are called first.

Sources

← Back to blog