Glossary
What is Candidate Scoring?
Candidate scoring is the process of assigning a numeric match score to each applicant against a job description or rubric. The score drives ranking — highest-fit candidates surface first, reducing the CVs a human needs to review closely from hundreds to a handful.
Three common scoring approaches
- ✓Keyword matching: Count how many JD keywords appear in the CV. Brittle — misses semantic matches.
- ✓Weighted rubric: Hiring manager defines categories (skills 40%, experience 30%, education 20%, other 10%). Score each category, sum. Transparent but labor-intensive.
- ✓LLM-based semantic scoring: Language model reads CV + JD in context, outputs a fit score with explanation. More accurate, less brittle.
What a good scoring system outputs
- ✓Numeric score (usually 0–100 or 0–10)
- ✓Per-criterion breakdown (skills match, seniority fit, industry experience)
- ✓Evidence quotes — which lines in the CV drove the score
- ✓Caveats — what's missing or unclear
Scores without explanation are black boxes. Avoid them — they're hard to calibrate and hard to defend in a bias audit.
Avoiding bias in scoring
- ✓Never score on name, photo, gender, age signals, or university prestige
- ✓Weight skills and concrete experience, not demographic proxies
- ✓Keep humans in the loop — scoring is triage, not final decision
- ✓Audit outputs periodically — check if top-scored shortlists match your team's actual hires
Legal note (EU AI Act / NYC Local Law 144)
In regulated jurisdictions, automated hiring tools face disclosure and audit requirements. Keep scoring systems that expose their reasoning (not opaque scores), document your rubric, and be able to show a candidate why they were or weren't advanced. This is both a compliance and a fairness concern.