Research & Data
What we've learned testing resumes against real ATS engines.
We've tested resume samples against the parsing patterns of 8 widely-deployed Applicant Tracking Systems. These are the patterns we've documented — what works, what fails, and what surprised us.
Headline numbers
Failure patterns by frequency
From the resumes we've tested, these failure modes appear in this rough order of frequency:
| Pattern | Frequency | Most affected ATS |
|---|---|---|
| Two-column / sidebar layouts | Very common | Workday, Taleo |
| Modern fonts (Avenir, Proxima Nova, Montserrat) | Common | All — font substitution |
| Contact info in document header | Common | Workday, Taleo, iCIMS |
| Image-based PDFs (Canva, Photoshop exports) | Common | All — extracts nothing |
| Glued tokens (e.g., 'SAPOracle') | Common | Greenhouse, Lever |
| Tables for layout | Moderate | Workday, Taleo |
| Custom section names ('My Journey', 'Career Story') | Moderate | Lever, Greenhouse |
| Special characters (curly quotes, em dashes) | Moderate | Taleo (encoding) |
| Date formats other than 'Mon YYYY' | Moderate | Workday, Taleo |
| Skill bars / progress bar graphics | Common in Canva templates | All — invisible |
Most common parsing failure: the two-column resume
Of every parsing failure pattern we document, two-column layouts are the most damaging. Here's why:
- Most ATS engines parse PDFs/DOCXs left-to-right, top-to-bottom
- When you have a sidebar, the parser reads the entire sidebar first, THEN the main column
- Result: chronology is destroyed. Skills appear before contact info, education appears before name, etc.
- The recruiter receives a candidate profile where fields are cross-wired or empty
- This is the #1 reason "qualified" resumes get auto-filtered with no feedback
Parsing engine differences
Each ATS engine has its own quirks. Patterns we've documented:
Workday
Behavior: Strictest of all — penalizes any non-standard formatting
Weak with: Two-column layouts, modern fonts, document headers
Strong with: Standard Word docs with single-column layout
Greenhouse
Behavior: Field-level extraction — focuses on Name/Email/Title
Weak with: Image-based contact info, custom section names
Strong with: Hyperlinks (LinkedIn, portfolio) — better than Workday
Taleo
Behavior: Plain-text only — formatting is irrelevant
Weak with: Special characters (em dash, curly quotes), custom dates
Strong with: If your content makes sense as plain text, Taleo handles it
iCIMS
Behavior: Most modern parser — better error recovery
Weak with: Industry-specific jargon may not surface as skills
Strong with: OCR fallback for image PDFs (still imperfect)
Methodology & honesty
We're a small team. The patterns documented here come from:
- Public ATS documentation from Workday, Greenhouse, Lever, etc. when available
- Empirical observation — feeding diverse resume samples through ATS parsing engines and recording outputs
- Existing industry research from ATS vendors and resume-tool ecosystem
- Our own product testing — every pattern we detect corresponds to a real failure observed in our pipeline
What we're not: we don't have inside access to proprietary ATS engines. Where industry numbers like "70-80% of employers use ATS" are cited, those come from public surveys and vendor disclosures, not our own measurements.
Test your resume against real ATS patterns
See exactly which of the patterns above your resume hits. Free, 30 seconds.
Free ATS scan →