Build a Lead Scoring Model

    You are a marketing-operations strategist building a scoring model that sales will actually trust.

What a genuinely good customer looks like: {{ideal_profile}}
What my best closed-won deals did before they bought — pages, content, actions, timing: {{winning_behaviors}}
What leads that never converted also did: {{non_converting_behaviors}}
Signals and data I can actually capture: {{available_signals}}
Current scoring, if any, and what's wrong with it: {{current_state}}
Sales capacity — how many leads they can genuinely work: {{sales_capacity}}

Produce:

**Two scores, not one.** Separate **fit** — is this the right kind of company and person — from **intent** — are they showing signs of a live problem. A single blended score hides the difference between a perfect-fit account browsing idly and a poor-fit visitor who read everything, and those two need completely different treatment. Define the dimensions and points for each, drawn from {{ideal_profile}} and {{winning_behaviors}}.

**Score intent on evidence, not engagement.** The core failure of most models is rewarding accumulated activity. Someone who reads twelve blog posts is a reader; someone who visits pricing twice and looks at the comparison page is a buyer. Using {{winning_behaviors}} against {{non_converting_behaviors}}, identify which actions actually distinguish buyers, and weight those heavily while giving low-signal engagement almost nothing. If both groups did something equally, it's worth zero points no matter how much of it there is.

**Negative scoring and disqualifiers.** Points off for signals of a bad fit, and hard disqualifiers that no amount of activity overrides — wrong company size, wrong geography, a competitor, a student or job seeker, a free-mail domain where that matters. A hard disqualifier is more valuable than any positive rule.

**Decay.** Intent is perishable. Define how quickly scores decay, because a lead that was hot three months ago and quiet since is not a hot lead, and models without decay slowly fill the top of the list with stale records.

**Thresholds tied to capacity.** Set the qualifying threshold so the volume of leads passing it matches {{sales_capacity}}. A threshold that produces three times what sales can work means reps cherry-pick and the model gets ignored — which is how scoring dies. State the expected volume at your recommended threshold.

**Tiers and routing.** What happens at each level: immediate sales contact, a lighter touch, nurture, or nothing. Note that speed on the top tier matters more than anything else in the model — contact inside the first hour dramatically outperforms same-day.

**Validation.** How to test this model against the last several months of closed-won and closed-lost before switching it on. If the model wouldn't have scored my actual wins highly, it's wrong, and it's better to find that out now.

**Review cadence.** What to check monthly, and the signal that the model has drifted — most commonly, sales quietly working leads the model didn't flag.