The ATS Black Box

Enterprise Applicant Tracking Systems — Taleo, Workday, Greenhouse, iCIMS, and others — employ multi-stage document processing pipelines to convert your resume into structured data. Understanding each stage reveals exactly where most resumes break down.

1. 1Text Extraction — Raw text is pulled from your file using OCR (for scanned documents) or direct text extraction (for machine-readable PDFs and DOCX).
2. 2Section Segmentation — The parser identifies sections (Contact Info, Work History, Education, Skills) using pattern recognition and header detection.
3. 3Field Mapping — Extracted content is mapped into structured database fields that recruiters can search, filter, and sort.
4. 4Scoring & Ranking — The structured data is compared against the job requisition and assigned a composite match score.

The core ranking mechanism in most ATS platforms is keyword frequency analysis. The system compares extracted resume tokens against a weighted keyword list derived from the job requisition.

Not all keywords carry equal weight. Here is how most systems prioritize:

• Hard skills and certifications receive the highest weighting coefficients.
• Exact job titles matching the requisition are weighted heavily.
• Industry-specific terminology scores higher than generic equivalents.
• Soft skills typically receive the lowest weight — or are ignored entirely.

Semantic matching — understanding that "JavaScript" and "JS" refer to the same technology — varies significantly between platforms. Legacy systems perform strict literal matching, while modern platforms like Lever and Ashby incorporate limited synonym resolution.

After keyword extraction, the ATS assigns a composite score — typically ranging from 0 to 100. This score determines whether your application enters the active review pipeline or remains buried.

The composite score is a weighted sum of several factors:

• Keyword density — how many required keywords appear in your resume.
• Years of experience — compared against the stated requirement.
• Education level alignment — degree type and field of study.
• Geographic proximity — when location requirements are specified.
• Recency of relevant experience — recent roles weighted more heavily.

Not all file formats receive equal treatment from ATS parsers. Choosing the wrong format can cause extraction failures regardless of how well your content is written.

• DOCX — Highest parse fidelity across all major ATS platforms. Recommended as first choice.
• PDF (text-based) — Generally reliable, but compatibility varies by platform. Ensure the PDF was generated from a word processor, not scanned.
• PDF (image-based) — Frequently causes extraction failures. Scanned documents require OCR, which introduces errors.
• TXT — Universal compatibility but sacrifices all formatting.
• Google Docs links, Apple Pages, and rich media formats — Unsupported by the majority of enterprise ATS platforms.

Beyond keyword gaps, several structural elements trigger automatic deprioritization or complete parsing failure. Each of these is entirely preventable:

• Multi-column layouts — Parsers interleave content from different columns, producing garbled data.
• Tables — Cause similar extraction errors, especially in older ATS platforms.
• Headers and footers — Frequently ignored entirely. Placing contact info solely in a header means the ATS has no record of your name or email.
• Embedded images, charts, and icons — Invisible to text extraction engines.
• Custom fonts with glyph substitution — Can cause entire words to be misread.
• Excessive formatting — Heavy use of bold, italic, underline, and color changes can confuse parsers.