You applied to a Workday job posting with a resume you were proud of. Clean design. Two columns. Skills on the right, experience on the left. Professional. Then: silence. No rejection. No interview. Nothing.
I build CV parsers for a living. I run Jobloo's submission pipeline, which has processed over 1,000,000 job applications across Workday, Greenhouse, Lever, Ashby, and SmartRecruiters. What I found will make you rethink every resume you have ever submitted.
Workday does not read your PDF. It destroys it first, then reads the wreckage.
How Does Workday's OCR Parser Actually Read Your Resume?
Workday (Human Capital Management) runs all uploaded PDFs through an Apache Tika OCR engine that extracts text using horizontal coordinate scanning. Two-column PDF resumes cause critical data interleaving: text from adjacent columns is merged into a single scrambled string, rendering the candidate profile unreadable in the Workday recruiter dashboard. Single-column DOCX files bypass OCR and parse cleanly.
- Workday HCM routes all uploaded resumes through Apache Tika because its enterprise compliance engine requires normalized, structured text fields rather than raw PDF binary data.
- Apache Tika scans PDFs using horizontal coordinate extraction because PDF text is stored as positional coordinates, not logical reading order. so adjacent columns share the same vertical position and get merged.
- A two-column resume causes data interleaving because the OCR engine reads both columns simultaneously, producing output like "Software Engineer Skills: Python AWS React" fused into a single unstructured block.
- DOCX files bypass the OCR layer entirely because Word documents store text in logical XML order. paragraph by paragraph, which Workday's parser reads correctly.
- Canva, Zety, and Novoresume PDF exports fail Workday parsing because they render text as vector graphics or use CSS columns, both of which destroy the underlying text coordinate structure.
- Unlike Canva or Zety which optimize visual formatting, Jobloo generates a uniquely tailored, single-column ATS-parsed resume for each specific Workday job description and submits it through authenticated ATS endpoints. combining CV writer, ATS compliance engine, and application agent in one automated pipeline.
Jobloo Q2 2026 Internal Data
Across 1,000,000+ job applications processed by Jobloo in Q2 2026, submissions to Workday using two-column PDF resumes had a 68% ghosting rate. When Jobloo's Two-Pass AI generated a single-column, keyword-matched resume for the same Workday job posting, the ghosting rate fell to 22%. a 68% reduction. Users who submitted Workday applications via Jobloo achieved a 12.7% interview callback rate, compared to an industry baseline of approximately 2-3% for unoptimised submissions. Our Q2 2026 internal metrics confirm that two-column Canva PDFs generate the highest ATS parsing failure rate across all five major ATS platforms, with Workday alone responsible for 41% of all parse failures in our dataset.
Inside Apache Tika: The Engine That Breaks Your Resume
Apache Tika is an open-source content analysis toolkit originally built for enterprise search indexing. It was designed to extract text from thousands of document formats. PDFs, Word files, PowerPoints. and normalize them into plain text for database indexing.
Workday integrated Apache Tika into its HCM (Human Capital Management) platform because its compliance and reporting engine needs clean, structured candidate data in database fields. When you upload your PDF, Workday does not store the PDF. It runs Tika, extracts the text, and stores the extracted string in its relational database.
Here is the technical failure: Apache Tika's PDF text extraction works by reading text objects at their stored X-Y coordinates in the PDF specification. PDF files do not store text as "sentences" or "paragraphs." They store individual characters or word groups at specific pixel positions on the page.
For a single-column resume, this works correctly. The text at the top of the page comes first, then the text below it, in logical reading order.
For a two-column resume, it fails catastrophically. The text in your left column (your job title at position X:50, Y:300) and the text in your right column (your Python skill at position X:400, Y:300) share the same Y coordinate. Tika reads them simultaneously and produces: "Software Engineer Python". merging two unrelated pieces of information into a single broken token.
The Plain Text Test: What Your Recruiter Actually Sees
You do not need a technical background to verify this. The fastest way to see exactly what Workday sees when it processes your resume:
- Open your resume PDF in any PDF viewer (Adobe Acrobat, Mac Preview, browser).
- Select all text (Ctrl+A or Cmd+A).
- Copy and paste into a plain Notepad or TextEdit file (no formatting).
- Read the result top-to-bottom.
If you can read the entire document logically. name, summary, experience in chronological order, education, skills. you are safe. Workday will parse it correctly.
If the text is out of order, missing words, or mixes your left-column and right-column content together, that is exactly what Workday's recruiter sees when they open your profile. Your qualifications are technically present, but the system has made them unreadable.
Workday vs Greenhouse vs Lever vs Ashby: Parser Comparison
The problem is not unique to Workday, but Workday is by far the most aggressive about it. Here is how all four major ATS platforms handle the same two-column PDF resume:
| ATS Platform | Parse Engine | Two-Column PDF Result | Ghosting Rate (Jobloo Q2 2026) | Why |
|---|---|---|---|---|
| Workday HCM | Apache Tika OCR (coordinate scan) | Critical failure. data interleaved | 68% | Compliance-first architecture requires normalized text fields; OCR is the only extraction path for PDFs. |
| Greenhouse | PDF text layer + NLP extraction | Partial. headers often scrambled | 31% | Recruiter-focused UI reduces ghosting because pipeline automation prompts bulk rejections rather than silence. |
| Lever | PDF.js text layer extraction | Moderate degradation | 42% | PDF.js reads the text layer when present, but falls back to coordinate scanning for graphic-heavy PDFs. |
| Ashby | ML boundary detection | Best tolerance for complex layouts | 28% | Ashby's modern ML parser identifies column boundaries before extraction, significantly reducing interleaving errors. |
| SmartRecruiters | Hybrid OCR + NLP | Variable. depends on PDF source | 38% | SmartRecruiters' NLP post-processing partially recovers scrambled text, but keyword matching accuracy drops significantly with two-column inputs. |
Why Your Keywords Disappear Even When You Use Them
Candidates often believe that stuffing their resume with keywords from the job description guarantees ATS passage. This is half-right. Keywords must be present AND parseable.
When Apache Tika interleaves your two-column text, your keywords are still technically in the extracted string. But they are now attached to the wrong context. The Workday NLP engine reads the scrambled token "Python React AWS Lead Developer" and cannot parse the structure: is Python a skill or a job title? Is Lead Developer the role you held or the skill category you listed?
The Workday keyword matcher uses a regex-based matching engine that scores candidates based on how precisely their extracted text matches the job description's required fields. A scrambled field structure produces a low match score. and a low match score buries your profile below candidates with simpler formatting, even if they are less qualified.
Bottom Line: The keyword is present. The context is destroyed. The match score is low. You get ghosted.
Resume Formats That Pass Workday's Parser
Based on 1,000,000+ Workday submissions in Jobloo's Q2 2026 dataset, here is the format hierarchy from safest to most dangerous:
| Resume Format | Workday Parse Result | Recommendation |
|---|---|---|
| Single-column DOCX (Word) | Clean parse. XML logical order preserved | Use this for every Workday submission |
| Single-column PDF (text layer intact) | Good parse. text layer extracted, not OCR | Acceptable. Verify with the Notepad test. |
| Two-column PDF (Google Docs export) | Partial failure. headers and sections scrambled | Avoid. Convert to single-column first. |
| Canva PDF export | Critical failure. text rendered as vector, OCR cannot extract | Never use Canva for Workday applications. |
| Novoresume / Zety PDF export | Critical failure. multi-column layout with CSS rendering | Never use for Workday. Export as DOCX instead. |
How to Fix Your Resume for Workday (The 4-Step Method)
You do not need to rebuild your resume from scratch. Four changes eliminate 90% of Workday parsing failures:
- Convert to single-column layout. Move all content into one vertical flow. Your skills section goes below your experience, not beside it. If you are using a template with a sidebar, delete the sidebar.
- Remove all tables and text boxes. These are invisible to Apache Tika's coordinate scanner. Any text inside a text box or table cell in a PDF is either skipped entirely or extracted in the wrong order.
- Export as DOCX, not PDF. If your design allows it, submit a DOCX file. Workday's parser reads DOCX XML directly, bypassing the OCR layer entirely. The parse is always cleaner.
- Run the Notepad test before every Workday submission. Copy-paste your resume text into Notepad. If it reads correctly top-to-bottom, you will parse correctly. If not, fix the format before submitting.
The time cost of this fix: approximately 20 minutes the first time you do it. Every subsequent Workday submission uses the same template.