SLIDE 1: What resume parsers actually read
RESUMEGET PARSING LAB
A 45-minute, evidence-first workshop using synthetic PDF and DOCX resumes.
- No participant resumes are needed.
- Every fixture, raw extraction, method, and limitation is public.

SLIDE 2: One disclosed parser, not a universal compatibility score
THE BOUNDARY
The lab runs ResumeGet's production text extractor. It does not certify or reproduce any named applicant tracking system.
- Observe what the release proves.
- Name what it cannot prove.
- Keep negative results visible.

SLIDE 3: 10 layouts. 20 files. 11 risk patterns.
THE CORPUS

- PDF and DOCX versions of every case
- Headers, footers, columns, tables, symbols, page breaks, decorative spacing, and mixed-language labels
- Reserved synthetic identity, employers, institution, URLs, and achievements

SLIDE 4: Rendered layout is not extracted reading order
TWO VIEWS
A person sees position, alignment, weight, and proximity. A text extractor returns a sequence.
- Which block is read first?
- Which essential value disappears?
- Which heading becomes separated characters?

SLIDE 5: Observed results, including the failures
RELEASE 2026-07-v2

- 14 files preserved every disclosed marker in order.
- 1 file preserved markers but changed reading order.
- 5 files lost at least one disclosed marker.
- 0 files were unreadable by the production extractor.

SLIDE 6: Essential content in DOCX headers and footers can disappear
PATTERN 1
In this release, the body extractor omits synthetic identity and contact values stored only as document furniture.
- Keep name and essential contact details in the normal body flow.
- Use repeating furniture for secondary context, not the only copy.

SLIDE 7: Columns and tables can change sequence
PATTERN 2
Visual adjacency does not guarantee that extracted education, skills, and experience retain the intended order.
- Inspect the raw text, not only the rendered page.
- Prefer a clear single reading flow for essential information.

SLIDE 8: One person reads the page. One reads only the text.
PAIRED EXERCISE

- Describe intended hierarchy and actual extraction order.
- Mark missing or moved content.
- Sketch one lower-risk revision and state the visual tradeoff.
- Share the fixture ID, never a participant's personal detail.

SLIDE 9: Preserve meaning before decoration
REDESIGN RULES

- Use ordinary text for section labels.
- Keep essential identity and contact data in body content.
- Use one predictable reading flow.
- Retest after export because PDF and DOCX can differ.

SLIDE 10: Download the corpus, raw outputs, method, and workshop kit
INSPECT THE EVIDENCE
resume-get.com/research/resume-parsing-lab
- The QR code points to the same canonical page.
- Citation is welcome; a backlink is never required for reuse.
