What Is Lead Scoring and How to Set It Up

Lead scoring is a way of ranking your prospects by assigning each one a number that captures how valuable and how ready they are, so your sales team works the best leads first instead of treating a flat database as if every record were equal. The score blends two things: fit — how closely a lead matches your ideal customer — and intent — how engaged the lead is based on what it has done. A lead that fits your profile perfectly and has just requested a demo scores high and goes to a rep now; a lead that fits but has done nothing scores lower and goes into nurture; a lead that does not fit at all scores low or negative and stays out of the queue. This guide explains the two halves of scoring, how to build a model, where to set the thresholds that hand a lead from marketing to sales, why negative scoring matters, and how to keep the whole thing honest by iterating with your sales team.

It builds on how to qualify leads with website data and assumes the records you are scoring have been enriched — see how to enrich your CRM data.

Why score leads at all

Without scoring, every lead looks the same in the database, and a team either works them in the order they arrived or burns through them indiscriminately — both of which waste selling time on prospects that were never going to buy. Scoring imposes a priority order grounded in evidence: the leads most likely to convert rise to the top, and the rest wait or drop out. The payoff is focus. A rep with a scored queue spends the day on the handful of prospects that genuinely warrant a call, rather than dialling through a list where the good and the hopeless are mixed together. Scoring also creates a shared, objective definition of a "good lead" between marketing and sales, which defuses the perennial argument about lead quality — the score is the contract.

The two halves: explicit and implicit scoring

Every sound scoring model has two components, and understanding the split is the key to building one that works.

Explicit scoring rates fit using attributes — the things a lead tells you on a form or that you append through enrichment. Firmographics (industry, company size, region) and technographics (the platform and tools the company runs) are the staples. Explicit scoring answers the question: is this the right kind of company for us? It is stable, because a company's profile changes slowly.

Implicit scoring rates intent using behaviour — the things a lead does. Pages visited, emails opened and clicked, content downloaded, pricing page viewed, demo requested, webinar attended. Implicit scoring answers: how interested is this lead right now? It is volatile, rising and falling with activity.

The two are multiplicative in spirit. High fit plus high intent is a hot lead that should reach sales immediately. High fit but low intent is a good company that is not engaged yet — perfect for nurture. Low fit but high intent is a curious browser who will waste a rep's time — exactly what negative scoring exists to catch. A model that scores only behaviour will send enthusiastic but unqualified leads to sales; a model that scores only fit will sit on good-fit companies until a competitor engages them first. You need both.

Building a scoring model

A model is just a set of weighted signals summed into a number. Build it in four steps.

First, list the signals that predict a good customer — both fit and intent. Lean on what you already know: which attributes do your best customers share, and what did they do before they bought?

Second, assign each signal a weight reflecting how strongly it predicts revenue. A demo request is worth far more than a single blog visit; a perfect platform match is worth more than a roughly-right industry.

Third, define how to score each signal from your data, so scoring is consistent and automatable.

Fourth, sum the weighted signals into a single score per lead.

The table below is an illustrative starting model. The weights are deliberately round numbers — yours should reflect what actually predicts revenue for your product, which you will learn by calibrating against real deals.

Signal	Type	Example points
Industry matches ICP	Explicit (fit)	+15
Company size in target band	Explicit (fit)	+15
Runs a platform/tool you extend or replace	Explicit (fit)	+20
Missing a tool in your category	Explicit (fit)	+15
Visited pricing page	Implicit (intent)	+20
Requested a demo / contacted sales	Implicit (intent)	+30
Downloaded a deep-funnel asset	Implicit (intent)	+15
Opened/clicked recent emails	Implicit (intent)	+5
Job-seeker / student title	Negative	−20
Free-email or competitor domain	Negative	−15
Company far too small to buy	Negative	−20
Unsubscribed	Negative	−30
No activity in 90 days	Negative (decay)	−10

Sum the rows that apply to a lead and you have its score. Keep the model simple at first — a dozen well-chosen signals beats fifty noisy ones — and resist the urge to over-engineer before you have data to justify the complexity.

Setting thresholds and the MQL-to-SQL handoff

A score is only useful if it triggers an action, and that is what thresholds do. Define bands that map the score to a stage:

• Below the lower threshold: the lead stays in marketing nurture — too early or too unqualified for sales attention.
• Crossing the middle threshold: the lead becomes a marketing-qualified lead (MQL) — fit and engagement have reached a level worth a closer look.
• Crossing the higher threshold, often combined with a strong intent signal like a demo request: the lead becomes a sales-qualified lead (SQL) and is handed to a rep.

The right threshold is the score at which sales agrees the lead is genuinely worth their time. That is a negotiation, not a calculation — set it with the sales team, not for them. If reps complain that handed-over leads are not ready, the threshold is too low; if good leads are languishing in nurture, it is too high. The handoff is where scoring earns its keep, because it is the moment a number becomes a sales action.

Why negative scoring is non-negotiable

Teams new to scoring often build only positive signals, and the result is predictable: an unqualified but curious visitor — a job-seeker researching the company, a student, a competitor poking around — racks up engagement points by reading content and reaches sales looking hot, only to waste a rep's time. Negative scoring prevents this by subtracting points for anti-signals: a job-seeker or student job title, a free-email or competitor domain, a company far too small to buy, an unsubscribe, or a long stretch of inactivity that should decay an old score back down. The principle is symmetry — if positive signals can push a lead up, anti-signals must be able to pull it down, or the score drifts upward with mere activity and stops meaning anything. Negative scoring is what keeps the queue honest and the threshold meaningful, and it is the half that beginners most often skip and most regret skipping.

Using website and tech signals in the score

Some of the most reliable fit signals are observable on a prospect's own website, which makes them ideal explicit-scoring inputs because they are consistent and do not depend on the lead filling in a form. The platform a company runs, the tools it has installed (a budget and maturity signal), the absence of a tool in your category (often the strongest buying signal of all), and a documented weakness in a performance or SEO audit when that is the problem you solve — all of these can feed the explicit half of the score. Because they come from public, observable data, they fill the gap when a lead's self-reported attributes are thin, and they are detailed in how to qualify leads with website data and what technographics are and how to use tech-stack data to qualify leads. Folding technographic signals into the model is one of the most effective upgrades available, because it grounds the fit score in what a company actually runs rather than what it claims.

Calibrating against real deals

The weights in any starter model are educated guesses until your own outcomes correct them. The single most valuable calibration exercise is to look back at leads that became customers and leads that did not, and ask whether your scores actually separated the two. If leads that scored low ended up closing and leads that scored high went nowhere, the weights are wrong and the model is misleading you. Maybe the pricing-page visit predicts conversion far more strongly than you assumed, while a particular firmographic predicts almost nothing. Feed those findings back into the weights, and the model stops being generic and starts reflecting your actual market. This is a loop — score, hand off, win or lose, re-weight — and over a few cycles the model becomes a genuine predictor of revenue rather than a tidy-looking guess. The discipline of checking the score against reality is what separates a scoring model that helps from one that merely exists.

Iterating with sales feedback

Marketing owns the model, but sales lives with its output, so the two must iterate together. The reps working the handed-over leads are the richest source of truth about whether the score is right: they can tell you the leads that arrived ready and the ones that were nowhere close, the signals that correlated with real interest and the ones that were noise. Build a regular feedback channel — even a simple "was this lead ready? yes/no" on every SQL — and use it to adjust signals, weights and thresholds. When sales trusts the score because their feedback shapes it, the handoff stops being a battleground and becomes a shared, improving process. A scoring model maintained this way compounds in value; one built once and never revisited slowly drifts out of step with the market and quietly loses the team's trust.

The workflow

1. List the fit and intent signals that predict a good customer for you.
2. Weight each by how strongly it predicts revenue, and add negative signals for anti-fit and inactivity.
3. Sum the signals into a score and set thresholds for nurture, MQL and the SQL handoff — with sales, not for them.
4. Fold in observable website and tech signals to strengthen the fit half of the model.
5. Calibrate against real won and lost deals and iterate with sales feedback continuously.

A compliance note

Lead scoring itself is an internal prioritisation exercise on data you already hold, so the privacy questions sit upstream, in how you gathered and enriched the data, and downstream, in how you contact the leads the score surfaces. Score on lawfully held, business-relevant data; under GDPR keep it accurate and minimised; and when a high score sends a lead to a rep, the resulting outreach still owes honest identification and an easy opt-out under CAN-SPAM and CASL. Scoring should also never become opaque profiling of individuals on sensitive attributes — keep the signals business-relevant. Done on clean, business data, scoring is simply a way to spend your team's time where it counts.

Go deeper

• Feed the model good data: how to enrich your CRM data.
• Strengthen the fit score: how to qualify leads with website data.
• Act on the segments behind the score: how to segment your prospect list.
• The technographic foundation: what technographics are and how to use tech-stack data to qualify leads.

Want observable fit signals to score on? StackOptic detects any site's tech stack, performance and SEO so you can feed real technographic data into your scoring model — start free.