"ATS score" tools are everywhere and mostly unexplained. Ours is documented — this is the full walkthrough of what a 0–100 resume score measures, what it can't measure, and what real resumes most often get wrong, measured from our own product data.
What the score actually is
Three measured categories — format, content, keywords — rolled into a 0–100 signal (full methodology public: ATS Compatibility Index). The analyzer's first job is mechanical: can the text and critical fields be extracted cleanly from your file at all? If parsing fails, nothing downstream matters.
What real resumes get wrong most often
We aggregated the findings our analyzer raised across 301 analyzed resumes — 1,190 findings, an average of 4.0 per resume. (Anonymized category counts only; we never retain or aggregate resume text for research.) The distribution surprised us:
| Finding category | Share of all findings |
|---|---|
| Keyword gaps (skills section vs listing language) | 44% |
| Job-fit mismatches (experience vs the target role) | 33% |
| Writing quality (summary, responsibility bullets) | 13% |
| Mechanical ATS checks (format, fields, structure) | 10% |
Two honest surprises versus the folk wisdom:
- Parsing is rarely the villain. Mean parse confidence in this cohort was 0.92, and only about 2% of analyzed resumes fell below 0.7. The "multi-column layout broke the ATS" horror story exists — but among people actively working on their resume, it's the rare case, not the common one.
- The common failure is vocabulary, not formatting. Nearly half of everything the analyzer flags is the skills section not speaking the listing's language — the exact literal-match problem quantified in our companion study, What a Million Job Listings Reveal About Resume Keywords.
Severity data backs this up: only 18% of findings were high-severity, and 53% were directly patchable edits. Most resumes aren't broken — they're mistranslated.
One distribution caveat, stated plainly: these are resumes analyzed inside the product's draft workflow (median score 93; 84% scored 85 or above), so the cohort skews toward people already mid-improvement. First-upload scores run lower.
What a score can't tell you
- Whether a human will like the resume. Taste isn't parseable.
- Whether you're qualified. It measures presentation against listing language, not your career.
- Any specific company's private ATS configuration. Scores are calibrated against our own job-listing research corpus, not vendor internals.
The honest use of a checker
Run it before every application cycle. Fix the mechanical failures first — parse, fields, format — then tailor keywords to the specific listing's dialect. Our keyword study shows which dialects and compounds real listings use; the free checker measures your resume against the same corpus.
Methodology
Scoring categories and limits as published on the methodology page. Finding-frequency stats measured July 28, 2026 over 301 analyzed resumes and 1,190 findings in the product draft store, aggregated by finding category, severity, and target field only — resume text, messages, and identities are never read or aggregated for research (see our privacy policy). Cohort limit disclosed above: draft-workflow resumes skew toward already-improved states.
