title: "Lead Scoring That Actually Works: Beyond MQL Theater" slug: "lead-scoring-actually-works-beyond-mql-theater" date: "2026-04-19" excerpt: "Most lead scoring models are theater -- arbitrary point systems that sales ignores and marketing defends. Here is how to build a scoring model grounded in real buying behavior that sales actually trusts." featuredImage: null category: "article" tags: ["fractional-head-demand-gen", "fractional-vp-revops"]
Your marketing team is celebrating. MQL volume is up 30% this quarter. The lead scoring model is working -- more leads are hitting the threshold, more leads are getting passed to sales, and the marketing dashboard shows a beautiful upward trend.
Meanwhile, your sales team is ignoring most of those MQLs. The SDRs cherry-pick the ones that look interesting based on company name or job title and let the rest sit in the queue until they expire. When you ask the VP of Sales about it, they shrug: "Most of those leads are junk. They downloaded a whitepaper and got enough points to trigger the MQL, but they are not actually in a buying cycle."
This is MQL theater. The lead scoring model creates the appearance of marketing-sales alignment while the actual alignment is broken. Marketing optimizes for MQL volume because that is what they are measured on. Sales ignores MQL volume because they have learned through experience that the score does not predict buying intent.
The result is waste -- wasted marketing spend generating leads that go nowhere, wasted SDR time sorting through unqualified leads, and a growing rift between marketing and sales that makes collaboration harder with each passing quarter.
A fractional Head of Demand Gen who has seen this pattern at multiple companies knows how to fix it. The solution is not to abandon lead scoring. It is to build a scoring model that is grounded in actual buying behavior, validated against real outcomes, and continuously refined through a feedback loop with sales.
Why Most Lead Scoring Models Fail
Lead scoring fails for predictable reasons. Understanding these failure modes is essential to avoiding them.
Arbitrary Point Values
The most common approach to lead scoring is to sit in a room and assign point values to activities based on gut feeling. Downloading a whitepaper: 10 points. Visiting the pricing page: 20 points. Attending a webinar: 15 points. Filling out a contact form: 50 points. MQL threshold: 100 points.
These point values are almost always wrong. They are based on assumptions about what indicates buying intent, not on data about what actually predicts conversion. Maybe webinar attendees at your company convert at twice the rate of whitepaper downloaders. Maybe pricing page visits are actually a weak signal because your pricing page ranks well in organic search and attracts tire-kickers. You will not know unless you validate the model against actual outcomes.
No Distinction Between Fit and Interest
A VP of Engineering at a 500-person SaaS company who downloads a whitepaper is a very different lead than an intern at a 10-person agency who downloads the same whitepaper. But if your scoring model only measures behavioral signals (activities), both leads accumulate points at the same rate.
Effective lead scoring separates two dimensions: fit (does this person match your ideal customer profile?) and interest (are they exhibiting buying behavior?). A lead needs to score well on both dimensions to be a genuine MQL. High fit but low interest is a target account that needs nurturing. High interest but low fit is someone who likes your content but will never buy. Only high fit plus high interest is worth passing to sales.
No Negative Scoring or Decay
Most scoring models only add points. They never subtract them. A lead who was highly engaged six months ago but has gone silent still carries their historical score. They might even trigger an MQL months after their last interaction.
Effective models include score decay (points diminish over time if there is no new activity) and negative scoring (points are subtracted for signals that indicate low buying intent, like using a personal email address, being in an excluded industry, or unsubscribing from emails).
No Validation Against Outcomes
The most critical failure: most companies never check whether their lead scoring model actually predicts conversion. They set up the model, turn it on, and assume it works. They never run the analysis that compares MQL scores to actual opportunity creation, pipeline value, and closed-won revenue.
Without this validation, you are flying blind. You might be passing leads at exactly the right threshold, or you might be flooding sales with noise. You literally do not know.
The Two-Dimensional Scoring Model
The foundation of a scoring model that works is separating fit from interest and evaluating each independently.
Dimension 1: Fit Score (Firmographic and Demographic)
The fit score measures how closely the lead matches your ideal customer profile. It is based on attributes, not behavior.
Factors to score:
- Company size. If your ICP is companies with 50-500 employees, a lead from a 200-person company scores higher than one from a 5,000-person company or a 10-person company.
- Industry. If you sell primarily to SaaS companies, a lead from a SaaS company scores higher than one from a law firm.
- Job title and seniority. If your buyer is typically a VP or Director, those titles score higher than Manager or Individual Contributor.
- Geography. If you only sell in North America, leads from outside the region score zero or near-zero.
- Technology stack. If your product integrates with Salesforce, leads from companies that use Salesforce score higher.
How to assign fit scores:
Instead of guessing, analyze your closed-won deals from the past 12 months. Which firmographic and demographic attributes correlate with conversion? Weight your fit score based on actual data. If 80% of your closed-won deals come from SaaS companies with 100-1,000 employees where the buyer is a VP or above, those attributes should carry the highest fit scores.
Dimension 2: Interest Score (Behavioral)
The interest score measures the prospect's level of engagement and buying intent based on their actions.
Behavioral signals from strongest to weakest:
- High-intent actions (direct buying signals): Requesting a demo, requesting pricing, filling out a contact form, starting a free trial. These signal active buying intent.
- Mid-intent actions (research behavior): Visiting the pricing page, viewing case studies, comparing your product to competitors (if you can track this), attending a product-focused webinar. These signal active evaluation.
- Low-intent actions (awareness behavior): Downloading educational content, reading blog posts, attending thought-leadership webinars, following on social media. These signal awareness and interest in the topic but not necessarily in your product.
How to assign interest scores:
Again, analyze actual data. Look at the behavioral patterns of leads that eventually became opportunities and compare them to leads that did not. Which actions are truly predictive? In most B2B companies, demo requests and pricing page visits are strong signals, while content downloads are weak signals. But your data may show different patterns.
Score decay: Behavioral scores should decay over time. A pricing page visit from yesterday is a strong signal. The same visit from six months ago is irrelevant. Implement a decay curve that reduces behavioral scores by 50% every 30 to 60 days.
Combining the Two Dimensions
Instead of a single MQL threshold, use a two-dimensional matrix:
| | Low Interest | Medium Interest | High Interest | |---|---|---|---| | High Fit | Nurture | MQL (Fast Track) | MQL (Priority) | | Medium Fit | Monitor | Nurture | MQL (Standard) | | Low Fit | Exclude | Monitor | Disqualify |
This matrix ensures that only leads with both adequate fit and adequate interest get passed to sales. It also provides clear guidance for what to do with leads that score well on one dimension but not the other.
Building the Feedback Loop with Sales
A scoring model without a feedback loop is a scoring model that will drift into irrelevance. The feedback loop is what keeps the model calibrated to reality.
The Disposition Requirement
When an SDR or AE receives an MQL, they must disposition it within a defined timeframe (typically 48-72 hours). The disposition options should be simple:
- Accepted: The lead is qualified and the rep is working it as an opportunity.
- Rejected -- Bad Fit: The lead does not match the ICP (wrong industry, wrong size, wrong geography, etc.).
- Rejected -- Bad Timing: The lead matches the ICP but is not in a buying cycle (no budget, no project, no urgency).
- Rejected -- Bad Data: The contact information is wrong, the person has left the company, or the record is a duplicate.
The Monthly Calibration Meeting
Once a month, the demand gen leader and the sales leader sit down together and review:
- MQL volume and acceptance rate. What percentage of MQLs were accepted vs. rejected? If the acceptance rate is below 60%, the scoring model needs adjustment -- the threshold is too low or the scoring weights are wrong.
- Rejection reasons. Are most rejections for bad fit (scoring model problem) or bad timing (nurturing problem)? The distinction matters because the fix is different.
- MQL-to-opportunity conversion. Of the accepted MQLs, what percentage became opportunities? What percentage closed?
- Score distribution of won deals. What scores did the leads have that eventually became closed-won deals? This tells you whether your MQL threshold is set correctly.
A fractional VP of RevOps often facilitates this meeting because they sit between marketing and sales and can broker the data-driven conversation without it devolving into finger-pointing.
Quarterly Model Recalibration
Every quarter, revisit the scoring model. Run the analysis again: which fit attributes and behavioral signals actually predicted conversion in the last 90 days? Adjust weights accordingly. Move the MQL threshold if needed. Add new signals or remove ones that turned out to be noise.
The model is never done. It is a living system that gets smarter over time as you feed it more data and more feedback.
Measuring Scoring Effectiveness
Once the model is in place, track these metrics to assess whether it is working:
MQL acceptance rate. Target: 60-80%. Below 60% means the model is too loose -- you are passing unqualified leads. Above 80% might mean the model is too tight -- you might be leaving good leads in the nurture pool too long.
MQL-to-opportunity conversion rate. What percentage of accepted MQLs become qualified opportunities? This measures whether the leads that pass the threshold are genuinely in a buying cycle.
MQL-to-closed-won conversion rate. The ultimate test. Do leads that score as MQLs actually buy? If MQL-to-closed-won conversion is below 5%, the model is not predicting buying intent effectively.
Average score of closed-won deals. Are your best deals coming from leads that scored high in the model? If closed-won deals have similar scores to closed-lost or disqualified leads, the model is not differentiating.
Time from MQL to opportunity. How quickly are scored leads converting to opportunities? If MQLs take 90 days to become opportunities, you might be identifying interest too early and should consider a higher threshold or additional scoring criteria.
Moving Beyond Traditional Lead Scoring
The two-dimensional model described above is a significant improvement over the single-score approach that most companies use. But for companies that want to go further, there are advanced approaches worth considering.
Account-Based Scoring
Instead of scoring individual leads, score accounts. Aggregate the fit scores and behavioral signals across all contacts at a target account. An account where three people from different departments have all visited your pricing page in the last two weeks is a much stronger signal than one person who did the same thing.
Intent Data Integration
Third-party intent data providers (Bombora, 6sense, G2, TrustRadius) track buying signals that happen outside your website -- research behavior across the web. Incorporating intent data into your scoring model can surface accounts that are actively evaluating solutions in your category before they ever visit your site.
Predictive Scoring
Machine learning models can analyze patterns in your historical data that human-designed scoring models miss. If you have enough data (typically 1,000+ leads and 100+ closed-won deals), a predictive model can identify non-obvious correlations and assign scores that outperform manually weighted models.
Each of these approaches adds complexity, and the basics -- two-dimensional scoring, feedback loop, quarterly recalibration -- should be solid before you layer on advanced techniques. But they represent the direction that lead scoring is evolving, and a fractional Head of Demand Gen can help you assess when your organization is ready to make the leap.
The goal of lead scoring is simple even if the execution is not: pass the right leads to sales at the right time, and stop wasting everyone's time with leads that were never going to buy. Build the model on data, validate it against outcomes, and keep it calibrated through continuous feedback. That is lead scoring that actually works.