Rediscovery rate
How many past candidates are worth recruiter review for the live role? Count candidates moved into the reviewable shortlist divided by the permitted CV pool size.
Benchmark methodology
Benchmark whether your CV database can still produce useful candidates.
This is the public benchmark framework behind CVault's agency rediscovery wedge: one live role, a bounded set of permitted CVs, recruiter review, and evidence-first scoring.
Methodology template. No result run has been published. This page defines what a permitted customer benchmark should measure.
Five metrics recruiters can review
Evidence acceptance
Do recruiters accept the cited CV evidence behind the match reasons? Recruiter review marks each evidence-backed reason as accepted, disputed, or insufficient.
Gap visibility
Does the shortlist expose missing must-haves before outreach? Track whether key gaps are visible on the candidate card and whether they change outreach decisions.
Shortlist rebuild time
How long does it take to rebuild a shortlist from old CVs? Compare recruiter time spent on the CVault sample against the agency's normal database-search workflow.
Recruiter actionability
Can a recruiter decide the next action from the output alone? Reviewers classify each candidate as contact, hold, no outreach, or needs more evidence.
Recommended benchmark process
- Choose one active role with clear must-haves and nice-to-haves.
- Select up to 100 existing CVs the agency is allowed to process for the sample.
- Run the CV pool through parsing, role matching, evidence extraction, and gap surfacing.
- Ask a recruiter to review the ranked shortlist and mark accepted, disputed, and insufficient evidence.
- Compare actionability and time spent against the agency's usual CV database search process.
Limitations
- This page is a methodology template, not a production benchmark result.
- Synthetic demos can show workflow shape, but real-CV readiness requires a permitted customer sample and recruiter review.
- The benchmark should not be used as an automated hiring decision; it is a decision-support review framework.
Inspect the workflow before commissioning a run
Run the synthetic rediscovery demo, then inspect the deterministic report preview. Both are labeled synthetic and use no customer CVs.
A real benchmark requires a permitted corpus, named reviewer, run date, version identifiers, and recorded disposition evidence.