An applicant tracking system does two jobs. First it extracts text from your file into fields — name, contact, employers, dates, skills. Then it ranks what it extracted against the job, usually on keyword overlap. Most CVs that never get a reply fail at step one, not step two.
Extraction reads in a straight line
The parser walks the document in reading order and tries to slot each block into a field. It has no eyes. A layout that looks obviously structured to you — a sidebar, a two-column skills grid, a header band — is just a stream of characters to the parser, and the order it receives them in is often not the order you see.
- Two-column layouts interleave. The parser can read left-column line one, right-column line one, left-column line two — your summary and your skills end up spliced together.
- Contact details in the page header are frequently skipped entirely. A perfect CV that nobody can phone.
- Tables used for layout are read cell by cell. Nested tables are often dropped.
- Icons instead of labels — a phone glyph with no word “phone” — leave the field with nothing to extract.
- An image or a scanned export has no text layer at all. There is nothing to index.
What survives
Single column. Standard section headings the parser recognises — Professional Summary, Work Experience, Education, Skills. Dates in a consistent MMM YYYY format. Contact details as plain text in the body. A standard font. Real selectable text, exported straight to PDF, not a picture of a document.
Every layout Roleweave exports follows those rules. The variation between them is type, spacing and section order — nothing that changes how a parser reads the file.
Then, and only then, keywords
Once your text is extracted cleanly, ranking is about vocabulary. If the posting says stakeholder management and you wrote keeping people informed, you described the same work in words nobody searches for. That is the part tailoring fixes — and the part the Quick check diagnoses for free.