Score contacts with Jev (Jev judges the person, code counts the rest)

The insight: a contact score is two different kinds of fact, and only one of them is about the person. "Would this title own the budget?", "is this a recruiter?", "how senior is 'Head of Fleet Ops'?" are judgments about the person, and titles are messy enough that only a model reads them well. "Is their company an A?", "did they start this role in the last four months?", "how many points is that?" are lookups and arithmetic. Ask a model for "a lead score from 0 to 100" and it blends both into a number nobody can decompose, and quietly lets a VP at a poor-fit company outrank a manager at your best account.

The evidence is Jev's own documentation. TypeSafe publishes what Jev 1.13 is bad at, and the top of the list is numbers, counting and comparing dates; it reads literally; it degrades when the record carries fields the question does not need. Its recommended pattern for scoring is to break the judgment into atomic questions and combine them with weights in code. And Jev returns a probability for every option, so a score can use how sure it was.

What follows is the whole design: