TL;DR: The B2B marketing industry keeps declaring lead scoring dead. It isn’t. The traditional framework — tracking engagement with owned assets, routing to sales at threshold — still captures real signal. The problem is that most scoring systems were built once, weighted by intuition, and never validated against actual revenue outcomes. That’s not a measurement methodology problem. It’s a discipline problem.
The Old Framework Is Not Dead
Every few years the industry declares lead scoring obsolete. The current version of the argument goes: buyers research in AI tools and private Slack communities, third-party intent signals are unreliable, and the behavior that actually predicts purchase decisions produces no events in your system. So why score at all?
The conclusion doesn’t follow from the premise.
Form fills, content downloads, email engagement, and repeated site visits still capture real signal. A buyer who has visited your pricing page three times in two weeks and opened four consecutive emails is exhibiting a different pattern than one who submitted a form eight months ago and hasn’t engaged since. That distinction has value — not because the scoring system can see everything, but because the behavior it can see still correlates with intent.
The fact that measurement is harder at the margins doesn’t mean the core framework is worthless. It means the core framework needs to be maintained and validated, which most teams aren’t doing.
The Decay Problem Is Real and Chronically Underaddressed
The most common scoring failure isn’t bad weights — it’s the absence of timed decay. A lead scoring model without decay applied to behavioral signals will systematically inflate the scores of contacts who engaged once, a long time ago, and have since gone cold. Those contacts aren’t more likely to buy; they’re occupying the top of your queue based on history that no longer reflects their current intent.
Applying a rolling window to behavioral scores — so that engagement from 12 months ago contributes meaningfully less than engagement from last week — fixes most of the inflation problem without requiring a new methodology. Most marketing automation platforms support this natively. Most teams haven’t configured it.
Directional movement also contains information that absolute score levels often miss. A contact whose score has been flat for six months and then triples in two weeks is more interesting than one whose score has been consistently high for a year. Surfacing velocity — not just level — to your sales team is a low-cost improvement most teams aren’t making.
Third-Party Intent Data: Treat It as a Hypothesis
Third-party intent data — aggregated behavioral signals from research platforms, review sites, and content syndication networks — has a credibility problem. According to Landbase’s 2026 intent signal research, most B2B companies are missing significant revenue opportunities by acting on intent signals that haven’t been validated against their own pipeline data. Quality varies dramatically by vendor and category. Some signals genuinely predict in-market behavior. Others reflect research done by competitors, consultants, or people with no intention of buying. Most vendors are not transparent about how they construct their scores, which makes independent validation difficult.
The right posture is skepticism plus experimentation. Don’t turn off intent feeds if your team is using them. But validate them before weighting them.
The test is simple: if this signal is predictive, accounts with high intent scores in a given week should convert to pipeline at a meaningfully higher rate than matched accounts without high intent scores over the following 60–90 days. Pull that analysis. If the hypothesis holds, the signal has value and you can weight it accordingly. If it doesn’t, reduce the weight or eliminate the feed. This is the same validation discipline you’d apply to any new paid channel. There’s no reason intent data should get a pass.
Build Hypotheses and Run Experiments
The discipline that separates functional lead scoring from theater is the willingness to generate specific, testable hypotheses and run experiments against them.
A hypothesis-driven approach looks like this: “We believe contacts who visit our pricing page more than twice in 14 days are significantly more likely to convert to an opportunity within 30 days than contacts who haven’t visited pricing, regardless of other engagement.” Test it. Pull historical data. Route contacts matching that pattern to a higher-priority sales queue for a quarter. Measure conversion to opportunity for that group against a control. If the hypothesis is validated, encode it in the model. If it isn’t, discard it and generate the next one.
This requires accepting that your current lead scoring model is a set of working hypotheses, not established facts. Most models aren’t built that way — they’re built on assumptions that felt reasonable at the time and then calcified into configuration. The result is a system that reports on activity with false precision while having no verified relationship to revenue.
A quarterly review of scoring performance — close rates by score tier, score distribution at opportunity creation, scoring threshold versus actual qualification rates — will surface whether the model is tracking toward revenue or toward its own internal logic. That review, done consistently, is more valuable than any new intent signal or AI-powered scoring layer added on top of a framework that hasn’t been validated.
FAQ
Q: Should we still use lead scoring if most of our pipeline comes through outbound?
Yes, though the inputs shift. Outbound-heavy teams should weight firmographic fit and account-level signals (job changes, funding rounds, technology signals) more heavily than behavioral engagement scores. The framework is the same — score for likelihood to convert, validate the weights against closed-won data — but the variables are different.
Q: How do we know if our lead scoring is producing MQLs or producing revenue?
Pull close rates and average deal value by score tier at the time of MQL designation. If there’s no meaningful difference in close rate between your top score tier and your second tier, your thresholds are miscalibrated. If your MQLs close at the same rate as your unscored inbound, the model isn’t doing anything. Both patterns are common and both are fixable with better threshold calibration and feature validation.
Q: At what scale does lead scoring actually matter?
Lead scoring produces the most value when volume exceeds what the sales team can manually review. If your sales team can look at every inbound lead individually, the operational urgency is low. As soon as inbound volume creates a prioritization problem, scoring becomes a leverage tool — and the quality of your scoring model directly affects how efficiently your sales team spends its time.
Additional Resources
From the Zaitz Marketing Knowledge Library:
- Why Your Attribution Model Is Lying to You — Why the signals that look most useful are often the most misleading
- What is Incrementality in Marketing? — The holdout experiment methodology for validating channel and signal effectiveness
- Your Attribution Platform Has No Idea the AI Funnel Exists — How to account for buyer behavior that produces no events in your system
- What Marketing Mix Modeling Actually Tells You — Channel-level measurement that goes beyond lead scoring to budget allocation
External Reading:
- How to Build B2B Lead Scoring Based on Buying Intent — FL0 Journal
- B2B Lead Scoring: The Complete Framework for 2026 — Breadcrumbs
- 15 Intent Signal Statistics That Prove B2B Companies Are Missing Massive Revenue Opportunities — Landbase (source for intent signal validation research)
- B2B Buyer Intent Data: What It Is & How to Use It in 2026 — Leadfeeder
Want a second read on your measurement setup?
Start with a Growth Architecture Review. We will map your channel mix, audit your attribution, and show you where the real leverage is.
Book a Conversation →