Lead scoring examples: How to score, weight, and act on leads

Most sales teams do not have a lead problem. They have a prioritisation problem. Leads arrive, they enter a queue, and reps work through that queue in roughly the order contacts appeared, whichever name they recognise, or whichever account their manager mentioned last week. None of those signals reliably correlate with conversion probability.

A lead scoring model replaces that flat queue with a ranked list built on criteria that actually predict which accounts are worth calling first. Instead of treating every lead equally, it directs sales effort towards prospects with the highest likelihood of converting, helping teams spend their time where it produces the greatest return.

This article covers practical lead scoring examples, including the criteria to include, how to weight them, where most models break down, and what has to happen after a score is assigned for it to change anything.

Why lead queues break without a scoring model

When there is no scoring model, prioritisation defaults to recency or instinct. Both have the same structural failure: they treat all leads as roughly equivalent and leave reps to decide where to focus under time pressure and without the data to decide well.

The time cost compounds quickly. In a typical week, reps already spend roughly a quarter of their working hours across researching prospects, active prospecting, and lead prioritisation. A scoring model does not eliminate that time, but it concentrates the prioritisation piece on contacts that have already passed a fit and intent filter, rather than on whoever appeared in the queue most recently.

The cost on the buyer side is equally concrete. 73% of B2B buyers actively avoid sellers who send irrelevant outreach. Without a model filtering for fit and intent, irrelevant outreach is a routing problem, not a rep performance problem. Reps contacting the wrong accounts at the wrong time is the predictable output of a system with no structure telling them otherwise.

A lead scoring model fixes the routing layer by assigning a numerical value to each lead based on how closely they match the target buyer profile and how much interest they have demonstrated, then surfacing the highest-value contacts at the top of the queue. The criteria behind that score, and how they are weighted, determine whether the model reflects conversion reality or just reorganises a list.

Four scoring models and when each one applies

The right scoring model depends on how leads arrive and how much data exists about them. Using the wrong model produces a score that does not reflect actual conversion probability, which means the queue it generates is not meaningfully better than an unsorted one.

• Demographic and firmographic scoring: assigns points based on who the lead is: job title, company size, industry, and geography. This model works when the ICP is tightly defined and contact data is reliable. It does not capture timing or intent. For outbound-first teams where leads have not yet engaged with the company, this is often the only viable starting point. The score tells you who is worth contacting, not who is ready to buy.
• Behavioral scoring: assigns points based on what the lead does: pricing page visits, demo requests, email link clicks, form submissions, and event attendance. This model surfaces urgency and purchase intent but requires sufficient lead volume and engagement history to be meaningful. It fits inbound-heavy pipelines where behavioral signals accumulate before a rep contacts anyone.
• Predictive scoring: uses machine learning to analyse historical conversion data, identify patterns across many variables, and assign scores automatically. The model adjusts as new outcomes come in and requires enough closed-won and closed-lost CRM data to train reliably. It is best suited to teams at scale with several years of pipeline history.
• Hybrid scoring: combines firmographic fit with behavioral engagement signals. Hybrid scoring is the most practical starting model for B2B teams because it accounts for both who is worth contacting and when, using fit criteria to filter by ICP match and behavioral signals to surface timing, so reps receive a queue that reflects both dimensions rather than just one. A perfect-fit lead who has done nothing is not an MQL. A high-activity lead with poor fit is not one either.

‍

A note for outbound teams: when leads have not yet engaged, lead scoring engine criteria examples based on behavioral signals do not apply. Scoring operates on ICP fit alone, drawing on title, company size, industry, and technographic signals to prioritise which cold accounts to contact first. This functions more like territory prioritisation than traditional lead scoring, but the underlying logic holds: direct rep time toward the accounts most likely to convert once engaged.

Scoring criteria: what to weight, and by how much

The criteria behind a lead scoring model determine whether it reflects conversion reality or just organises a list by activity. Most teams underweight fit signals in favour of behavioral ones, or vice versa. Both dimensions need to be present, and point values assigned without reference to closed-deal history produce a model that rewards activity rather than conversion probability. The only reliable calibration input is data on which actions leads took in the 30 days before they closed.

Fit criteria (explicit scoring)

These signals come from what is known about the lead before they interact with the company.

• Job title and seniority: decision-makers score significantly higher than individual contributors. A VP of Operations might earn 25 points; a Sales Manager, 15; a coordinator, 5. The cut-off matters: if the buying decision requires VP-level sign-off, a manager-level contact should not score equivalently regardless of behavioral activity.
• Company size: score ICP-range companies highest. For a product targeting mid-market B2B, a company with 200 to 1,000 employees might earn 20 points. A company under 50 employees earns 5. A company over 5,000 may fall outside the current motion and earns 0 or negative.
• Industry match: weight ICP verticals higher than adjacent ones. A SaaS company might earn 15 points; professional services, 10; retail or consumer-facing businesses, 3. The weighting should reflect where the product actually closes, not where the team wants to expand.
• Geography: if the product is territory-restricted, out-of-territory contacts score 0 or negative. This prevents SDRs from spending time on accounts the company cannot serve.

Behavioral criteria (implicit scoring)

These signals come from what the lead does after entering the system.

• Pricing page visit (single): a direct signal of commercial intent. Typical weight: 10 points.
• Pricing page visit (two or more in one week): a near-term buying signal. The repeat visit indicates active evaluation, not casual research. Typical weight: 15 points.
• Demo request or contact form submission: the highest-intent action in most B2B models. A lead who fills in a demo form has self-identified as a buyer. Typical weight: 20 points.
• Email link click (not just an open): an email open is passive. A click, particularly to a product or pricing page, indicates active engagement. Typical weight: 8 points.
• Webinar or event attendance: signals research-stage intent. The lead is actively gathering information and is somewhere in an evaluation process. Typical weight: 10 points.
• Blog read or newsletter open: low intent. These actions indicate awareness, not purchase consideration. Assign minimal points (2 to 3) and cap the category so a lead cannot accumulate a high score from content consumption alone.

The point values above are illustrative. Calibrate them against historical closed-won data before treating them as operational. Lead scoring engine criteria examples built on assumptions about what should predict a close produce a queue sorted by activity level, not by conversion probability.

Negative scoring and score decay

A scoring model that only adds points will eventually produce a queue full of leads who look engaged on paper but stopped being relevant months ago. Negative scoring and score decay are the mechanisms that keep the model accurate over time.

Negative scoring subtracts points for signals that indicate poor fit or active disengagement.

• Personal email domain (gmail, yahoo, hotmail): -10 points. A business contact uses a business email. A personal address is a reliable signal the contact is not a professional buyer.
• Student or entry-level job title: -15 points. These contacts are not decision-makers evaluating a purchase.
• Email unsubscribe: -25 points. The lead has actively signalled they do not want contact. A score that ignores this is not measuring intent.
• Out-of-territory geography (where applicable): -20 points or full disqualification.

Score decay reduces point values over time. A lead who visited the pricing page six months ago and has since done nothing is not a warm prospect today. Treating them as one consumes SDR time on a contact that has already self-selected out of the evaluation window.

A practical decay structure:

• High-intent actions (demo requests, pricing page visits): reduce by 25% every 60 days.
• Low-intent actions (blog reads, email opens): reduce by 50% every 30 days.

A rep looking at a high-scoring lead should know that score reflects current activity. Without decay running, a scoring model becomes a record of past behaviour rather than a live indicator of purchase intent, and the queue it produces misleads rather than guides.

Setting your MQL threshold

The MQL threshold is the score at which a lead crosses from the marketing bucket into the SDR queue. Most B2B teams start between 60 and 80 points, then adjust based on which scored leads actually convert. The threshold is a dial calibrated to conversion data, not a fixed industry benchmark.

The calibration method: pull closed-won deals from the past 12 months and identify the score those leads held when they were handed to sales. That figure is the starting threshold. If the data does not yet exist, begin at 65 and recalibrate after 90 days of tracking which MQLs moved to opportunity and which did not.

A working framework for categorising lead temperature:

• Hot (70 to 100 points): immediate SDR outreach, same day.
• Warm (40 to 69 points): enters a nurture sequence; revisit in two to four weeks.
• Cold (0 to 39 points): deprioritised; no active SDR time allocated.

Teams carrying high lead volume raise the threshold to protect SDR time from being consumed by contacts that pass a basic score but have not yet demonstrated meaningful buying intent; a higher threshold concentrates rep effort on the contacts most likely to take a meeting. Teams with lower lead volume sometimes lower it to 50 to 60 to avoid a dry pipeline. Either way, the threshold only has operational value when it is tested against actual conversion outcomes and adjusted to reflect them.

What happens after the score: routing and response

A score sitting in a CRM dashboard has not changed anything. The value of a lead scoring model is entirely in what it triggers. Classification without action is administrative overhead, not pipeline management.

When a lead crosses the MQL threshold, two things should happen automatically: the lead routes to the appropriate SDR based on territory, product line, or company size, and the SDR receives an immediate task or notification. The rep then works the queue in descending score order, not by when the lead arrived.

A lead who visits a pricing page and receives no SDR contact for 48 hours gives competing teams time to engage first; the scoring system's value is only realised when threshold crossings trigger same-day action, not next-day review. Pairing scoring thresholds with a sales automation engine removes the manual step between a lead hitting threshold and an SDR receiving the task.

Account-level scoring adds a second dimension. In B2B, the buying decision is rarely made by one person. If three contacts at the same company visit the pricing page in the same week, the account is showing buying-committee activity, even if no individual contact has crossed the MQL threshold alone. Aggregating contact-level scores at the account level surfaces this signal and is particularly valuable for outbound sales and account-based motions where the goal is to identify which companies to prioritise, not just which individuals.

How Lilian handles ICP-fit scoring before the CRM

The lead scoring model described in this article applies to leads that have already arrived: contacts who came in through inbound channels with a score generated from behavioral and firmographic data. In an outbound motion, that picture is incomplete. The prospect has not interacted with the company yet. Behavioral signals do not exist. Scoring is fit-only, and whoever builds the prospect list applies that fit logic before any lead enters the CRM.

That prospecting layer is where most of the manual work sits. Building prospect lists, researching accounts, and scoring cold contacts before deciding who to contact first are tasks that consume SDR time without generating pipeline. This is the problem Lilian removes.

Lilian is a digital worker for outbound sales prospecting. It identifies ICP-fit accounts, applies firmographic and intent-signal logic to determine which contacts are worth engaging, and runs the outreach sequence without SDR involvement in the research and list-building phase. SDRs receive a queue of ICP-matched contacts who have already been engaged, rather than a spreadsheet to sort through before selling can begin. The pipeline that results reflects ICP fit from the start, not after a rep has spent time eliminating accounts that should never have been on the list.

How to know if your scoring model is working

A scoring model calibrated once and never revisited will drift. The criteria that predicted conversion 12 months ago may not hold today if the ICP has shifted, the product has changed, or the competitive context looks different.

Three signals the model is working:

• MQL-to-opportunity conversion rate is improving across rolling 90-day periods. If it is flat or declining despite consistent lead generation volume, the model is not selecting the right contacts.
• SDR-reported quality of MQLs is positive. The reps working the queue know fastest whether the contacts at the top of it are worth calling. Consistent negative feedback from SDRs is a signal the threshold is too low or certain behavioral signals are overweighted.
• High-scoring leads convert at higher rates than lower-scoring leads that reached sales through other means. This is the direct test of whether the score reflects conversion probability.

Three signals the model needs recalibration:

• MQLs consistently fail to convert despite high scores. A behavioral signal carrying significant weight, newsletter opens for instance, is likely not predictive of actual buying intent.
• Low-scoring leads are closing. The model is missing a signal. Find the common thread across those closures and add it to the criteria.
• Score decay is not running. Stale leads accumulate in the hot tier, inflating the queue and consuming SDR time on contacts who were once engaged but have since gone cold.

The review cadence: monitor monthly for early signs of drift; recalibrate thresholds and point weightings quarterly using closed-won and closed-lost data from the prior period.

Stop routing by arrival order

A lead scoring model replaces arrival-order prioritisation with a ranked queue where position reflects both ICP fit and current intent signals; the contacts a rep calls first are determined by data, not by when the lead arrived or which name they recognise.

That queue is only as reliable as the model behind it. Fit criteria filter by who is worth contacting. Behavioral criteria determine who is ready now. Negative scoring and decay keep the model accurate as time passes and lead activity changes. The lead scoring tool that produces this ranked queue only generates pipeline if it triggers routing logic, response SLAs, and SDR accountability. A score without those downstream mechanics is a number in a dashboard.

If the prospecting layer above your scoring model still relies on manual research and list-building, book a demo to see how Lilian removes that step entirely.

Frequently asked questions

What is a good MQL threshold for a B2B lead scoring model?

Most B2B teams start with an MQL threshold between 60 and 80 points, then adjust based on which scored leads actually convert. The right threshold reflects the score that leads typically held when they moved to opportunity in your historical data, not an industry standard. Start at 65 and recalibrate after 90 days.

What is the difference between explicit and implicit lead scoring?

Explicit scoring assigns points based on who the lead is: job title, company size, industry, and geography. Implicit scoring assigns points based on what the lead does: pricing page visits, demo requests, and email link clicks. Most B2B models combine both, using fit signals to establish baseline eligibility and behavioral signals to determine timing and intent.

How often should a lead scoring model be reviewed?

Monitor the model monthly for early signs that MQL quality is declining or that specific signal weights are producing poor-quality handoffs. Recalibrate thresholds and point weightings quarterly using closed-won and closed-lost data from the prior period to test whether current criteria still predict conversion.

What is score decay in lead scoring?

Score decay automatically reduces a lead's score over time based on inactivity. A pricing page visit from six months ago carries far less conversion signal than one from last week. Decay rates typically differ by action type, with high-intent signals decaying more slowly than low-intent ones. Without decay, stale leads hold high scores and consume SDR time.

How is outbound lead scoring different from inbound lead scoring?

Inbound lead scoring uses both fit and behavioral signals because leads have already engaged with the company. Outbound scoring is fit-only at the start because behavioral data does not exist until the prospect responds. Outbound models score on job title, company size, industry, and technographic signals to prioritise which cold accounts to contact first.

‍