Marketing
Lead Scoring: A Guide for Teams Who Hate Lead Scoring
Most lead-scoring models are elaborate machines that predict nothing. Here is how to build one that earns its keep.
Score behavior over demographics
Many models load up on firmographic points: industry, company size, job title. These describe whether a lead could buy, but not whether they want to. Intent lives in behavior. A director who visited your pricing page three times this week is a hotter lead than a vice president who downloaded one ebook six months ago.
Weight your model toward recent, high-intent actions and decay old points over time. A lead's interest is perishable, and a score that never forgets treats a stale lead as if they are still raising their hand.
Validate the model against closed deals
The only proof that a scoring model works is that high-scoring leads close at higher rates than low-scoring ones. Pull your last six months of deals, look up the score each lead had at handoff, and check the correlation. If there is none, your model is a random number generator with extra steps.
This validation should be a recurring ritual, not a one-time exercise. The behaviors that predicted buying last year may not predict it this year, and a model nobody revisits slowly drifts into noise.
Keep sales in the loop
A scoring model built by marketing in isolation will not be trusted by the reps who have to act on it. Bring sales into the design, ask which signals actually correlate with good conversations, and let them flag leads the model got wrong. That feedback is the fastest path to a model people believe in.
When reps trust the score, the whole system works: marketing sends fewer but better leads, sales stops complaining about quality, and the handoff finally stops being a fight.