Your CRM is full of leads. Your sales team complains that most of them are garbage. Your CMO asks why marketing keeps generating "MQLs" that close at 2%. The gap between "we have leads" and "we have buyers" is where lead scoring lives, and most B2B SaaS teams handle it badly or skip it entirely.
Only 44% of organizations use lead scoring at all, and many of those that do score leads on the wrong signals. This post walks through the practical setup: what to score, how to weight the criteria, and the mistakes that quietly kill the model. For the broader context on building a B2B lead engine end to end, see our complete B2B lead generation guide.
What Lead Scoring Actually Is (and Is Not)
Lead scoring assigns a numerical value to each lead based on how well that lead matches your ideal customer profile and how engaged they are. The score is not a guarantee, it is a probability signal. A lead with 85 points is statistically more likely to close than a lead with 20 points. That is the entire point.
What lead scoring is not: an excuse to ignore lower-scoring leads forever, a one-time setup, or a substitute for talking to your sales team. The score is a routing mechanism. High scores get instant sales follow-up, medium scores enter nurture, low scores get filtered or recycled.
40% of marketers in 2026 named lead quality and MQLs the single most important success metric - higher than any other KPI in the HubSpot State of Marketing Report. Without scoring, you cannot measure quality.
The Two Scoring Dimensions: Fit and Behavior
Every workable B2B scoring model uses two independent dimensions. Conflating them is the most common reason scoring fails.
Fit (firmographic and demographic): Does this lead match your ideal customer profile? Company size, industry, geography, role, seniority, technology stack. Fit signals do not change much over time. A 50-person fintech in Munich is a fit on day one and on day 90.
Behavior (engagement and intent): Is this lead actively buying? Pricing page views, demo requests, content downloads, repeat visits, webinar attendance, email replies. Behavior signals decay. A pricing page view from last week matters more than one from six months ago.
Score them separately. A high-fit lead with no behavior is someone to nurture. A high-behavior lead with no fit is probably a competitor or student doing research. Only leads with strong scores on both dimensions go to sales as MQLs.
How to Set Up Your Scoring Model in Practice
Pick five fit signals and five behavior signals. Assign points based on how predictive each one is for your business. Here is a starting template you can adapt:
Set your MQL threshold at the score where leads historically convert at 3x your average. Look at six months of closed-won deals, score them retroactively with this model, and find the cutoff. Most B2B SaaS teams land between 50 and 70 points as the MQL line. Below that, leads stay in nurture. Above it, sales gets the alert.
The negative scoring rule matters more than people think. Subtract points for free email domains, competitor company names, locations outside your serviceable geography, and student or researcher roles. Without negative scoring, your model accumulates false positives over time.
Common Mistakes That Kill Lead Scoring
Treating it as a one-time setup. A scoring model built in January will be miscalibrated by July. New marketing channels, product changes, and seasonality all shift what predicts a sale. Review the model quarterly: pull the last quarter of closed-won and closed-lost deals, score them, and check if the threshold still separates them cleanly.
Scoring without sales agreement. If sales does not believe the MQL definition, they will ignore the alerts. Sit down with sales every quarter and ask which "MQLs" they actually worked and which they ignored. The signals they trust are the signals worth scoring.
Conflating fit and behavior. A common failure mode is one combined score where a junior researcher with high engagement outranks a CMO who only viewed the pricing page once. Keep them separate, route based on the lower of the two scores.
Ignoring decay. Behavior signals should lose value over time. A pricing page view from last week is hot. The same view from four months ago is cold. Build decay into your model, or filter on recency directly. Companies that implement lead scoring properly see 138% ROI on lead generation, compared to 78% for companies without scoring - the difference is in the recalibration discipline, not the initial setup.
Conclusion
Lead scoring is not a CRM feature you switch on and forget. It is a routing system that has to be built, calibrated, and recalibrated against actual sales outcomes. Score on two dimensions separately, agree the threshold with sales, and review the model every quarter. Done well, scoring turns a noisy lead list into a prioritized queue. Done badly, it just adds a number to bad data. For the wider context on running a B2B lead engine, see our complete B2B lead generation guide.
Frequently Asked Questions
What is the difference between an MQL and an SQL?
An MQL (Marketing Qualified Lead) is a lead that crosses the scoring threshold and matches your ideal customer profile, indicating sales should follow up. An SQL (Sales Qualified Lead) is an MQL that sales has personally vetted and confirmed as a real opportunity. Lead scoring decides who becomes an MQL. The conversation with sales decides who becomes an SQL.
Do I need a CRM with built-in lead scoring?
You need a system that can apply rules to records and trigger actions on threshold crossings. HubSpot, Salesforce, Pipedrive, and most modern CRMs include this natively. If yours does not, a marketing automation tool like ActiveCampaign or Customer.io can sit alongside it. Avoid building scoring in spreadsheets, because it will not scale and the data will not stay clean.
How long does it take to see results from lead scoring?
Routing improvements show up immediately, sales gets the right leads first. Conversion rate improvements need at least one full sales cycle to measure properly. For B2B SaaS with a 30 to 90 day cycle, expect to see real numbers in three to six months. Plan a recalibration review at month three to catch any obvious miscalibration before it hurts pipeline.