Modelo de Lead Scoring B2B 2026: Plantilla + Umbrales MQL/SQL

The Lead Scoring Problem Most Companies Get Wrong

Here is a statistic that should concern every B2B marketing team: according to Forrester, fewer than 10% of leads generated by marketing are ever contacted by sales. The reason is not laziness -- it is a lack of trust. Sales teams have been burned too many times by leads that marketing labeled as qualified but turned out to be students, competitors, or people who downloaded a whitepaper with no buying intent.

Lead scoring is supposed to solve this problem. In practice, most lead scoring models fail because they conflate activity with intent. A prospect who downloads five content pieces may be a researcher with no budget, while a prospect who visits your pricing page once may be a VP with purchasing authority and an urgent need. A scoring model that cannot distinguish between these two scenarios is worse than useless -- it actively damages marketing-sales alignment.

Defining MQL and SQL: The Handoff That Makes or Breaks Your Funnel

Before building a scoring model, marketing and sales must agree on definitions. A Marketing Qualified Lead (MQL) is a lead that has demonstrated sufficient fit and engagement to warrant sales attention. A Sales Qualified Lead (SQL) is a lead that sales has accepted and confirmed as a genuine opportunity through direct conversation.

The MQL definition should be jointly developed by marketing and sales leadership, documented in writing, and reviewed quarterly. It should include both a minimum fit score (does this lead match our ICP?) and a minimum engagement score (has this lead shown enough interest?). Neither dimension alone is sufficient.

The handoff process matters as much as the definitions. When a lead crosses the MQL threshold, it should be routed to a specific salesperson within 5 minutes (not 5 hours or 5 days). The salesperson should have 48 hours to accept or reject the lead, with mandatory feedback on rejected leads. This feedback loop is the single most important mechanism for improving your scoring model over time.

The Four Dimensions of Lead Scoring

An effective lead scoring model evaluates leads across four dimensions: demographic fit, firmographic fit, behavioral engagement, and negative signals.

Demographic fit scores the individual: job title, seniority level, department, and decision-making authority. A C-suite executive scores higher than a junior analyst. A department head scores higher than an individual contributor. Use LinkedIn data enrichment to fill gaps in your CRM records.

Firmographic fit scores the company: revenue, employee count, industry, technology stack, and geographic presence. Weight these attributes based on your ICP analysis. A company that matches your ideal firmographic profile on all dimensions should receive 30-40% of the total possible score from firmographics alone.

Behavioral engagement scores actions: page visits (weighted by page -- pricing pages score higher than blog posts), content downloads, email opens and clicks, webinar attendance, demo requests, and free trial signups. Recency matters: actions taken in the last 7 days should score 3x higher than actions taken 30 days ago.

Negative signals reduce scores: generic email domains (gmail, yahoo) for B2B products, competitor company names, unsubscribes, bounced emails, and prolonged inactivity. A lead that was highly engaged 6 months ago but has gone silent should see their score decay over time.

Implementing Lead Scoring in Your CRM

Most modern CRMs (HubSpot, Salesforce, Pipedrive) support native lead scoring. The implementation steps are straightforward but the calibration requires discipline.

Start with a 100-point scale. Allocate points across your four dimensions: 25 points for demographic fit, 25 for firmographic fit, 40 for behavioral engagement, and 10 points of possible deductions for negative signals. Set your MQL threshold at 65 points -- this should be roughly the top 15-20% of your lead database.

Build the scoring rules in your CRM, test them against your last 100 closed-won and 100 closed-lost opportunities, and verify that the model correctly scores won deals higher than lost deals. If the model cannot distinguish between historical wins and losses, your scoring criteria need adjustment before you go live.

Once live, review scoring performance monthly. Track three metrics: MQL-to-SQL acceptance rate (target: above 60%), SQL-to-opportunity conversion rate (target: above 30%), and average time from MQL to first sales contact (target: under 4 hours). If any metric is off target, adjust scoring weights or thresholds accordingly.

Moving to Predictive Lead Scoring

Once you have 6-12 months of scoring data and at least 200 closed opportunities, you have enough signal to build a predictive lead scoring model. Predictive scoring uses machine learning to identify patterns in your historical data that human-designed rules might miss.

The simplest approach is logistic regression using your CRM data. Export all leads from the past year with their scores, engagement history, and outcomes (won, lost, or stale). Train a model that predicts the probability of conversion based on all available features. The resulting model will typically outperform rule-based scoring by 20-40% in terms of conversion rate among top-scored leads.

Tools like HubSpot Predictive Lead Scoring, Salesforce Einstein, and standalone platforms like MadKudu can automate this process. But even with automation, the feedback loop remains essential. Predictive models degrade over time as market conditions change. Plan for quarterly retraining and continuous monitoring of model accuracy against actual outcomes.