Evidence-backed ranking
Evidence-backed ranking scores or orders candidates while exposing the evidence supporting each important requirement, along with missing evidence, inference, uncertainty, and verification needs.
Last reviewed: 2026-09-06
Why it matters in recruiting and sourcing
Ranking is useful only when a recruiter can challenge it. A single match percentage can hide whether the system matched a candidate because of an explicit skill, adjacent title, inferred experience, stale provider record, or unrelated keyword. Evidence-backed ranking keeps the score subordinate to the proof.
Example
A candidate ranks highly for a data-engineering role because work history supports Kafka, Spark, and production pipeline ownership. The dossier separately shows that healthcare-domain experience is missing and should not be inferred from one project name.
Common failure modes
- Opaque percentage scores
- Giving expansion terms the same weight as hard requirements
- Treating missing public evidence as proof of absence
- Using protected characteristics in ranking
- Failing to show the source supporting a match