Applicant cards, kanban badges and list rows all show the same match score from 0 to 100. It is built from three parts:
Skills (60%) — overlap between the candidate's declared and extracted skills and the keywords in your job. Words in the title count most, then qualifications, then responsibilities, then the description.
Experience (25%) — the candidate's years of experience against the seniority your posting implies ("5+ years", "senior", "new grad").
Location (15%) — same city scores highest, then same region, then same country. Remote-friendly jobs skip this part.
When a part is unknown — a candidate who has not shared their years of experience, say — it counts as neutral (60), not zero, so a thin profile is not punished for being thin. The score comes with a short explanation ("strong skills · mid experience · good location") so you can see why.
The score is deterministic: the same inputs give the same number every time, and no AI model is consulted. Protected characteristics do not exist in the schema, so they cannot enter it.
Sorting. /admin/applicants offers five orders: most recent, best fit, most experience, cleanest (lowest fraud score first) and furthest stage. Best fit needs a job selected; without one it falls back to most recent.
Matches. A job's Matches tab looks across the network for people who have not applied but whose skills overlap your qualification text, and shows exactly which terms matched. Their names and contact details stay masked.
Posting and reviewing are free. Every stage, e-mail and export described here is in the console from the first sign-in.