74%
CVault win rate on contested field comparisons.
Benchmark
Parser benchmark evidence, without the hype page.
In this benchmark, CVault won 74% of contested fields and 8 resumes, with 2 ties. This page keeps the result visible while avoiding old oversized benchmark styling.
Benchmark pages should support scrutiny. Review the methodology and known weaknesses before treating any number as product proof.
Summary
8 won
2 resumes tied and 0 lost in this benchmark set.
34 - 12
CVault field wins compared with Affinda field wins.
Category results
10 / 0 / 0
CVault wins / Affinda wins / ties.
6 / 0 / 3
CVault wins / Affinda wins / ties.
5 / 0 / 4
CVault wins / Affinda wins / ties.
5 / 1 / 3
CVault wins / Affinda wins / ties.
3 / 0 / 6
CVault wins / Affinda wins / ties.
2 / 0 / 7
CVault wins / Affinda wins / ties.
1 / 0 / 8
CVault wins / Affinda wins / ties.
2 / 5 / 2
CVault wins / Affinda wins / ties.
0 / 1 / 9
CVault wins / Affinda wins / ties.
What went well
Structured responsibility extraction — 10/10
Every resume produces clean JSON arrays of individual bullet points. Affinda returns unstructured text blobs with \n separators, requiring downstream re-parsing. This is the single biggest structural advantage for ATS integration, search indexing, and evidence-backed candidate matching.
Education extraction — 6 wins, 0 losses
CVault captured all education entries on every resume, including edge cases: General Studies at a community college, a 3-entry section on a 3-page executive resume, and 5 entries across 3 countries. Affinda missed entries on multiple resumes and misclassified certifications as education.
Phone extraction — 5 wins, 0 losses
Affinda dropped phone numbers on 5 of 10 resumes. CVault captured every phone number present on every resume.
Certification classification — 5 wins
CVault correctly separates certifications from education. Affinda folds certifications into education entries with a "Course/Certificate" level tag, conflating two distinct candidate data categories.
Job title accuracy — 3 wins, 0 losses
CVault preserves full job titles including geographic scope ("HEAD OF HR, INDIA"). Affinda splits geo-scoped titles incorrectly, assigns the geographic region as the employer, adds trailing commas, or drops titles entirely.
Where CVault lost or both systems struggled
Email extraction — 0 wins, 1 loss
On one resume where no email was present, CVault hallucinated one. Affinda correctly returned nothing. A real edge case to fix.
Inferred skill quantity — 2 wins, 5 losses
Affinda generates more skills by inferring aggressively from job descriptions (e.g., 138 skills on a 1-page resume). CVault extracts skills closer to what is explicitly listed. More inferred skills means more noise alongside more signal — a design tradeoff, not an accuracy gap.
OCR-damaged text
Both parsers produce identical errors on OCR-damaged PDFs ("SACETY COORDINATOR", "DADA/DACA SCIENTIST"). This is an upstream text extraction issue, not parser logic.
Missing company names
When resumes omit company names entirely, neither parser can invent them.
DOB / Nationality / Driving License
Both miss these non-standard fields.
Methodology
10 publicly available sample/template resumes with varied layouts, industries, and complexity. Not proprietary candidate data.
Methodology note from the benchmark packet.
Ground truth established by human review of each source PDF before running either parser.
Methodology note from the benchmark packet.
Both parsers run via standard pipeline with no manual review or correction.
Methodology note from the benchmark packet.
"Wins" counted only on contested fields where one parser is correct and the other is incorrect or missing. Ties (both correct or both wrong) excluded from win rate.
Methodology note from the benchmark packet.
Skill quantity comparisons favor the parser with more skills — acknowledged as a design tradeoff, not a pure accuracy metric.
Methodology note from the benchmark packet.
Resume 10 (Accountant) audited separately with full per-field comparison; results consistent with the 9-resume batch. Including resume 10: CVault 38–13 Affinda, 75% win rate.
Methodology note from the benchmark packet.
Tested against Affinda specifically. Not a claim of superiority over all resume parsers.
Methodology note from the benchmark packet.