Every discovered lead gets a fit score from 0 to 100 — a judgment of how well this specific person matches your ICP, with written reasoning you can read on the lead card.
The bands
- 90–100 — excellent fit on role, company, and profile evidence alone.
- 70–89 — good fit with some uncertainty or thinner evidence.
- 50–69 — adjacent or uncertain fit.
- below 50 — poor fit.
The scorer is explicitly told to be honest, to tie the score to concrete evidence, and to name the band and the reason. A few rules keep it strict:
- A matching title keyword is not a fit — a "fractional advisor" doesn't fill a role that needs an owner with a team and budget.
- Seniority and brand never substitute for your ICP's core requirements — a VP at a famous company who lacks the core is capped as adjacent.
- Location matters: for an offer tied to a city or market, an out-of-market lead with no relocation signal is capped as adjacent at best.
- Your "Don't target" exclusions are hard rules — a match scores below 50.
Two passes, increasing evidence
Search results carry only name, title, company, location. Candidates are first ranked cheaply, then the promising slice gets a quick scoring pass. The best of those are enriched — full profile, bio, recent posts — and re-scored on the sharper model with the extra evidence. Your credits are spent where they matter: deep evaluation only for candidates that earned it.
Timing signals — a recent relevant post, a job change — are scored on a separate intent axis, so "how well they fit" and "why now" never blur. Right before the first message, leads are re-scored once more with their latest posts.
It learns from your decisions
Accepting and rejecting suggestions feeds the next run: queries and scoring are biased toward what you accepted and away from what you rejected — as soft guidance that never overrides the evidence rules above. Calibration anchors from your approved ICP keep scores consistent from run to run.