[ FULL DISCLOSURE · EVERY WEIGHT · EVERY THRESHOLD ]
How our ATS resume score works
The scan is worth nothing you cannot verify, so here is the entire system: 14 structural checks worth 65 points, a keyword-match algorithm worth 35, and an honest list of what the score cannot tell you. The scanner is deterministic — the same input always produces the same report — and it runs in your browser, where you can watch it work in the developer tools.
The fourteen checks, with exact thresholds
Five buckets: structure (15), contact (10), impact (20), clarity (10), repetition & formatting risk (10). Under 40 words of text, the “absence of problems” checks (walls, pronouns, clichés, repetition, formatting) score zero rather than awarding points to an empty page.
| CHECK | POINTS | RULE |
|---|---|---|
| Resume length | 0–5 | Word count: 0 under 100 words, 2 at 100–149, full 5 at 150–900, 3 at 901–1200, 1 above. |
| Core sections | 0–6 | Heading lines detected: experience (2), education (2), skills (1), summary/profile (1). |
| Employment dates | 0–4 | Date references (Mon YYYY, YYYY–YYYY, MM/YYYY): 4 at three or more, 3 at two, 2 at one, 0 at none. |
| Email address | 0–4 | A valid email pattern anywhere in the text. |
| Phone number | 0–3 | A US-format 10-digit phone pattern. |
| Location | 0–2 | A 'City, ST' pattern with a real US state abbreviation. |
| LinkedIn URL | 0–1 | A linkedin.com/in/ or /pub/ URL. |
| Measurable results | 0–12 | Share of bullet lines containing a digit, % or $: 12 at ≥50%, 8 at ≥30%, 4 at ≥15%, 2 above zero. |
| Action-verb openings | 0–8 | Share of bullets opening with a known accomplishment verb (Led, Built, Reduced…): 8 at ≥60%, 5 at ≥40%, 3 at ≥20%. |
| Paragraph walls | 0–4 | Lines over 45 words: 4 at none, 2 at one, 0 at two or more. |
| First-person pronouns | 0–3 | I/me/my/we/our counts: 3 at none, 2 at one–two, 1 at three–five, 0 above five. |
| Clichés & buzzwords | 0–3 | Distinct empty phrases ('team player', 'synergy', 'results-driven'…): 3 at none, 2 at one, 1 at two, 0 at three+. |
| Repetition | 0–4 | Duplicated lines (30+ chars) and content words used 8+ times: 4 at none, 2 at one issue, 0 beyond. |
| Formatting risk | 0–6 | Starts at 6; deducts for all-caps lines (−2 each, max −4), table/column characters (−2), mixed bullet styles (−1), and bulletless long documents (−2). |
How keyword matching works
1. Extract the posting’s keywords.The job description is tokenized and counted in 1–3 word phrases. Stopwords and posting filler (“candidates”, “responsibilities”, “strong”…) are dropped. A phrase qualifies with two or more mentions, or one mention if it is a known hard skill — a tool named once still matters.
2. Prefer the phrase over its parts.If “project management” appears three times, the unigrams “project” and “management” at three each are the same three mentions, not extra signal — the phrase subsumes them. Each surviving keyword is weighted by mentions × word count.
3. Match closely related word forms.A keyword counts as present when the exact phrase appears in your resume, or when every stemmed token appears, so “managed teams” can satisfy “managing teams” in this checker. This is a transparent text-overlap rule, not a model of any employer’s system. The match score is the share of total keyword weight your resume covers, multiplied by 35.
4. Flag the hard skills. Keywords tagged as hard skills (tools, languages, licenses, frameworks) are marked ▲ because they are often important requirements in job postings, so missing ones go to the top of your review list.
The honest limitations
- 01No vendor emulation. Workday, Greenhouse, iCIMS, Lever and the rest differ by employer and configuration. We measure two broad signals used in many screening workflows — readable pasted text and keyword overlap — and nothing vendor-specific.
- 02No hiring prediction.A 95 means your pasted text scores strongly under this site’s published rubric. It does not mean an employer’s software will surface the resume, or that a person will read or advance it.
- 03No file parsing. The scanner reads pasted text, not PDF or DOCX files — because the failure modes live in that conversion, and you should see the text the parser will see. Copy-paste from your file first; what survives the paste is a good proxy for what a parser receives, though it is not a byte-exact replica of any specific upload pipeline.
- 04Keywords must be truthful. The tool will happily reward you for adding skills you do not have. That is a terrible idea — the interview checks what the ATS cannot. Add only what you can defend.
Questions or a check you think is miscalibrated? The scoring code is small and deterministic; tell us via the contact page and we will argue about the weights happily.