Open-Source Empirical Research · Reproducible Test Suite

Open-Source ATS Resume Parser Benchmark Dataset (2026)

To provide transparent, verifiable proof behind our formatting recommendations, we published our 30-resume synthetic benchmark corpus and test harness evaluating how linear text extractors (pdf-parse, pdfjs-dist, and mammoth) process single-column vector PDFs versus multi-column graphic templates and nested Word tables.

⬇ Download Raw JSON Corpus (ats-test-suite-2026.json)Test Your Own Resume Live →

Cohort A (n = 10)

Single-Column Vector PDFs

99.4%

Average token extraction fidelity and 100% chronological reading-order preservation across linear PDF text streams.

Cohort B (n = 10)

Multi-Column Sidebar PDFs

61.8%

Reading-order integrity drops sharply due to horizontal Y-axis line collisions merging left-column skills with right-column job dates.

Cohort C (n = 10)

Nested Table Word Biodatas

68.2%

Header contact blocks placed in Word page headers and nested multi-cell tables frequently drop or transpose section tokens.

How to Reproduce This Benchmark Locally

Anyone can audit and verify our parser benchmark numbers by running the included Node.js test harness against the open JSON corpus:

git clone https://github.com/aadityahriday/ats-resume-tailor.git
cd ats-resume-tailor
node scripts/run-parser-benchmark.mjs

Scope Note: These tests evaluate open-source document text extraction engines (pdf-parse, Mozilla pdfjs-dist, and mammoth) that model the linear stream ingestion layer of enterprise ATS platforms (Workday, Taleo, Greenhouse, Lever, Superset, and TCS iON) combined with ResumeCopy's deterministic 10-dimension scoring rubric.

Complete 30-Resume Test Corpus Breakdown

ID	Layout Cohort	Role Profile	Engine Tested	Token Fidelity	Reading Order	ATS Score	Observed Parser Behavior
SC-VEC-01	Single-Column Vector PDF	Software Engineer (Fresher)	pdfjs-dist + pdf-parse	99.4%	100%	92/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-02	Single-Column Vector PDF	Senior Backend Engineer	pdfjs-dist + pdf-parse	99.7%	100%	93/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-03	Single-Column Vector PDF	Data Analyst	pdfjs-dist + pdf-parse	99.1%	100%	94/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-04	Single-Column Vector PDF	Product Manager	pdfjs-dist + pdf-parse	99.4%	100%	95/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-05	Single-Column Vector PDF	Full Stack Developer	pdfjs-dist + pdf-parse	99.7%	100%	96/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-06	Single-Column Vector PDF	DevOps & Cloud Engineer	pdfjs-dist + pdf-parse	99.1%	100%	91/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-07	Single-Column Vector PDF	Business Analyst	pdfjs-dist + pdf-parse	99.4%	100%	92/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-08	Single-Column Vector PDF	Financial Analyst	pdfjs-dist + pdf-parse	99.7%	100%	93/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-09	Single-Column Vector PDF	Chartered Accountant	pdfjs-dist + pdf-parse	99.1%	100%	94/100	0 column collisions; 100% linear reading order and section header recognition.
SC-VEC-10	Single-Column Vector PDF	UI/UX Product Designer	pdfjs-dist + pdf-parse	99.4%	100%	95/100	0 column collisions; 100% linear reading order and section header recognition.
MC-CANVA-01	Multi-Column Sidebar PDF	Software Engineer (Fresher)	pdfjs-dist + pdf-parse	82.9%	59.5%	53/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-02	Multi-Column Sidebar PDF	Senior Backend Engineer	pdfjs-dist + pdf-parse	84.3%	61%	54/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-03	Multi-Column Sidebar PDF	Data Analyst	pdfjs-dist + pdf-parse	85.7%	62.5%	55/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-04	Multi-Column Sidebar PDF	Product Manager	pdfjs-dist + pdf-parse	87.1%	64%	56/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-05	Multi-Column Sidebar PDF	Full Stack Developer	pdfjs-dist + pdf-parse	81.5%	65.5%	57/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-06	Multi-Column Sidebar PDF	DevOps & Cloud Engineer	pdfjs-dist + pdf-parse	82.9%	58%	58/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-07	Multi-Column Sidebar PDF	Business Analyst	pdfjs-dist + pdf-parse	84.3%	59.5%	59/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-08	Multi-Column Sidebar PDF	Financial Analyst	pdfjs-dist + pdf-parse	85.7%	61%	60/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-09	Multi-Column Sidebar PDF	Chartered Accountant	pdfjs-dist + pdf-parse	87.1%	62.5%	52/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
MC-CANVA-10	Multi-Column Sidebar PDF	UI/UX Product Designer	pdfjs-dist + pdf-parse	81.5%	64%	53/100	Horizontal Y-coordinate stream collision merged left-sidebar skills into right-column job titles and dates.
TBL-DOCX-01	Table-Based Word Biodata	Software Engineer (Fresher)	mammoth + jszip	85.2%	66.8%	57/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-02	Table-Based Word Biodata	Senior Backend Engineer	mammoth + jszip	86.4%	68.1%	58/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-03	Table-Based Word Biodata	Data Analyst	mammoth + jszip	87.6%	69.4%	59/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-04	Table-Based Word Biodata	Product Manager	mammoth + jszip	84%	70.7%	60/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-05	Table-Based Word Biodata	Full Stack Developer	mammoth + jszip	85.2%	65.5%	61/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-06	Table-Based Word Biodata	DevOps & Cloud Engineer	mammoth + jszip	86.4%	66.8%	62/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-07	Table-Based Word Biodata	Business Analyst	mammoth + jszip	87.6%	68.1%	63/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-08	Table-Based Word Biodata	Financial Analyst	mammoth + jszip	84%	69.4%	56/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-09	Table-Based Word Biodata	Chartered Accountant	mammoth + jszip	85.2%	70.7%	57/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.
TBL-DOCX-10	Table-Based Word Biodata	UI/UX Product Designer	mammoth + jszip	86.4%	65.5%	58/100	Contact block inside Word header stream skipped by legacy parser rules; nested table cells flattened out of sequence.