Most lead scoring models are built on assumptions from three years ago. The buying committee has changed. The content that used to signal intent now signals casual curiosity. The demographic filters are targeting titles that no longer exist.
And yet the model keeps running, passing "qualified" leads to sales, burning trust between marketing and revenue teams one bad handoff at a time.
The Problem With Activity-Based Scoring
The most common mistake is scoring based on volume of activity rather than quality of intent. Opening five emails doesn't mean someone is ready to buy. It might mean the subject lines are good. Downloading a whitepaper doesn't indicate budget authority. It indicates interest in the topic.
Activity-based scoring conflates engagement with intent. They're related, but they're not the same thing.
What a Working Model Looks Like
A lead scoring model that actually predicts conversion does three things:
- Separates fit from interest. Demographic scoring (company size, industry, title) runs independently from behavioral scoring. Both must meet threshold.
- Weights recency. An action taken yesterday matters more than one taken three months ago. Score decay is essential.
- Validates against outcomes. The model is regularly tested against actual conversion data. Scores are correlated to closed-won, not just MQL creation.
The Rebuild Framework
Start with the end. Pull 12 months of closed-won data. Identify the common attributes and behaviors of accounts that actually converted. Build the model backward from there.
The best scoring models are boring. They don't try to be clever. They just reflect what the data already knows.
Then test it. Run the new model in parallel with the old one for 30 days. Compare which set of MQLs converts at a higher rate. Let the data decide.
The Bottom Line
If sales is ignoring MQLs, the scoring model is the first place to look. Not sales process. Not SDR effort. The model. Because if the model is wrong, everything downstream is wrong too.