How fit scoring works

Every application gets one match score from 0 to 100 — the same number on the applicants list, the pipeline cards, a job's applicant list, the Matches tab, and the percentile the applicant sees for themselves. It is built from four parts:

Skills (50%) — what your posting asks for against what the candidate shows. The posting is read for its requirements: must-haves (a Requirements or Qualifications list, "required", "must have"), nice-to-haves ("preferred", "bonus", "a plus") and skills it only mentions. Must-haves count twice as much as nice-to-haves. The candidate's evidence comes from their profile skills, the tools listed against past jobs, certifications, job titles, their résumé and their application-form answers. A skill they have counts in full; one shown by another (Django shows Python) counts 80%; a related tool (MySQL for a PostgreSQL job) half; a partial match 60%; one they're learning 30%. "ADP or UKG" is met by either. Soft skills such as communication count for little — a profile can't show them.

Role (20%) — how close their current or past job titles are to the job's: the same role, a very close one (Payroll Specialist for Payroll Administrator), the same kind of work, or related work. A past title counts 90% of a current one.

Experience (20%) — years of experience against the minimum your posting asks for ("5+ years"; a range such as "5–8 years" counts from 5; otherwise words like "senior" or "new grad"). Having those years — or being up to two years short — scores in full, and more years never lower the score: there is no "overqualified" penalty, because years of experience track age. A stated maximum ("no more than 5 years") is ignored for the same reason, and years that aren't experience ("18+ years old", "a 4-year degree") don't count.

Location (10%) — same city scores highest, then same state or region, then same country (higher if they're open to relocating). Remote jobs leave location out unless they are limited to one country.

Unknowns are neutral, not zero. A part the candidate hasn't shared anything for — no years, no location — counts as 60, so a thin profile isn't punished for being thin. A part your posting gives nothing to compare with (no skills named, no years asked, no location) is left out, and the other parts scale up to 100.

Why this score. Click Why? on any row (or the match badge on a pipeline card) for the full breakdown: the points each part earned, every requirement in your posting with whether and how the candidate meets it and where it was seen (profile, past job, résumé, application form), strengths, gaps, what isn't known, a confidence level (how much the candidate shared), and their rank among the job's applicants.

Percentile. An applicant's percentile is their rank among the job's applicants scored the same way — withdrawn and disqualified applications left out — and is shown once at least five applicants are scored. Fewer than that, and nobody is shown a percentile.

Kept current. When a candidate updates their profile, résumé or answers, or you edit the posting, the affected scores are recalculated within a minute or two.

Never used. Age or anything standing in for it (graduation years, "life stage"), gender, race, ethnicity, religion, disability, veteran status. Degrees aren't scored. The score is deterministic: the same candidate and posting give the same number every time, and no AI model is consulted.

Sorting. /admin/applicants offers five orders: most recent, best match, most experience, fraud-clear first and pipeline stage. Best match ranks each applicant by their score for the job they applied to.

Matches. A job's Matches tab finds people in the talent network who haven't applied and scores them exactly as applicants are scored, with the same reasons and where each would rank among your applicants. Their contact details stay masked until you unmask them; with Talent network access you see their names from November 25, 2026 (until then, names are masked too).