Scoring methodology · contract v2.21.6
How the Guul ATS Score Is Calculated
Resume scores disagree wildly between tools, and almost none of them publish how they are produced. This page publishes ours: the exact category weights, the score bands, the versioned contract that stops the numbers moving quietly, the 6 parser families we simulate — and, in as much detail, the things this score cannot know about your job search.
The category weights
The scorer runs in two modes. Without a job description it asks whether the document works; with one, it also asks how well it answers that specific posting. Both modes total 100.
Without a job description
| Category | Weight | What it measures |
|---|---|---|
| File readability | 35% | Whether the file itself can be read: text that is actually text rather than an image, a layout that survives being flattened into a single reading order, contact details in the body rather than in a header, and characters that do not mangle on the way in. |
| Structure | 30% | Whether the document is shaped the way resume parsers expect: recognisable section headings, dates in formats that index, entries with the fields a candidate record needs, and a length that does not get truncated. |
| Content quality | 35% | Whether the writing carries evidence: achievements rather than duty lists, bullets with enough substance to be informative, and skills that appear in the experience that supposedly demonstrates them. |
Swipe to see more →
With a job description
| Category | Weight | What it measures |
|---|---|---|
| File readability | 25% | Whether the file itself can be read: text that is actually text rather than an image, a layout that survives being flattened into a single reading order, contact details in the body rather than in a header, and characters that do not mangle on the way in. |
| Structure | 15% | Whether the document is shaped the way resume parsers expect: recognisable section headings, dates in formats that index, entries with the fields a candidate record needs, and a length that does not get truncated. |
| Content quality | 20% | Whether the writing carries evidence: achievements rather than duty lists, bullets with enough substance to be informative, and skills that appear in the experience that supposedly demonstrates them. |
| Job description match | 40% | How much of what a specific posting asks for is genuinely present in the resume. Only scored when you supply a job description, which is why the other three categories carry less weight in that mode — the total is still 100. |
Swipe to see more →
The score bands
Bands are a reading aid for one underlying 0–100 number, not separate scores.
90–100
Excellent
Nothing structural is standing between the document and a parser. Remaining work is editorial.
80–89
Strong
Parses cleanly with minor issues that are worth fixing but unlikely to lose information.
70–79
Good
Readable, with specific fixable problems — usually formatting or thin evidence in one section.
60–69
Fair
Something is reliably going wrong: a section that does not index, dates a parser cannot read, or bullets with little to extract.
0–59
Needs work
Expect material loss of information on the way into a candidate record. Start with the file readability issues.
Scores are versioned and regression-tested
A score is only meaningful if it means the same thing tomorrow. Ours is treated as a versioned contract: the scoring logic carries a version number — currently v2.21.6 — and when the scoring changes, the version changes with it, along with a public release note recording what moved and why.
Behind that sits a set of golden tests: fixed sample resumes pinned to fixed expected scores, checked automatically on every change. If an edit anywhere in the system would shift a score, those tests fail and somebody has to decide whether it was a regression or a deliberate recalibration. Silent drift is not an available outcome. The same version is locked across the parts of the product that compute and display scores, so two Guul surfaces cannot quote you different numbers for the same resume.
The practical consequence for you: a score from one version and a score from another are not directly comparable, which is why the version is published rather than hidden.
The 6 simulated parser families
Each family has a quirk profile: bullet length limits, which date formats index, how tables and multi-column layouts tend to be flattened, how headers and footers are handled, and how reliably PDF is read compared with Word.
What “simulated” honestly means
These profiles apply Guul’s configured risk rules. They are not connected to any vendor’s live system, and no vendor endorses or reviews this simulation. We do not send your resume to those ATS vendors for testing. A per-family result is a prediction under Guul’s assumptions, not a verdict from that company’s software. Check the extracted fields against your original document and the employer’s application form.
What this score cannot know
The most useful thing a scoring page can publish is its own boundary. Here is ours. For a longer plain-language take on scores versus interviews, see ATS scores vs interviews.
Whether a human ever opens it
A resume can parse perfectly into a candidate record and still never be read. Volume, timing, referral routing, and how a recruiter searches their pipeline decide that, and none of it is visible from your document.
Recruiter judgement
A recruiter spends a handful of seconds forming an impression from a few anchors — names, titles, companies, dates. That judgement is not a checklist, it varies by reader, and no score models it.
Whether you fit the role
Job-description matching measures overlap between two documents. It cannot tell whether your three years of that work were the right three years, whether the team wants a specialist or a generalist, or what the hiring manager actually meant by the posting.
How a specific employer has configured their system
Employers can configure different required fields, knockout questions and screening rules. Our family profiles are configuration-derived predictions, not measurements of a particular installation.
Your odds
We do not publish an interview probability, a callback rate, or a multiplier on your chances. Our checks do not measure those outcomes.
Things the scorer deliberately will not do
Keyword stuffing does not raise the score
Content quality asks whether a skill is evidenced in the experience that claims it, so repeating a term, hiding text, or pasting a keyword wall does not move the number the way it moves some scores. Recruiters discount keyword-perfect resumes on sight; optimising you into that bucket would be a disservice dressed up as a feature.
No adversarial framing
An ATS is not an opponent to be defeated — it is an indexing system that can lose parts of your resume. So the checker reports what got garbled and what got lost, and the fix list is about recovering information, not about outsmarting software.
The parse preview is display-only
The field-by-field reconstruction of "what the ATS sees" is derived from an already-computed parse result. It can never feed back into the score, so the picture you are shown and the number you are given cannot disagree.
No score without an explanation
Every issue that moves the number maps back to a location in your document. A point deduction you cannot act on is a complaint, not a check.
Related pages
Turn feedback into a specific document check
A parsing result, a keyword match and a hiring decision are different things. Inspect missing fields before changing accurate content to chase a score.
Frequently Asked Questions
How is the ATS score calculated?
Without a job description, the score is a weighted blend of three categories: File readability 35%, Structure 30%, Content quality 35%. When you supply a job description, a fourth category is added and the weights change to File readability 25%, Structure 15%, Content quality 20%, Job description match 40%. Both modes total 100.
Why does my score change when I add a job description?
Without a posting, the score combines Guul’s parsing, structure and content checks. Adding a posting includes a job-match category and changes the weights shown above. A resume can be clearly written while matching few requirements in a particular posting. Neither score is an employer’s assessment.
Do you test against real Workday or Greenhouse systems?
No, and it matters that we say so. We simulate 6 parser families — Workday, Greenhouse, Taleo (Oracle), iCIMS, Lever, SAP SuccessFactors — using Guul's configured risk rules for text, dates, layout and contact fields. These are configuration-derived predictions, not independently verified descriptions of vendor internals or tests of an employer's live installation. Compare the preview with your original document and the actual application fields.
What does the version number on my score mean?
Scores are a versioned contract, not an implementation detail. The current version is v2.21.6. Any change that would move a score requires the version to change, a changelog entry explaining what moved and why, and updated expectations in a set of golden tests that run automatically. Those tests hold fixed resumes at fixed expected scores, so a scoring change cannot happen quietly — an unintended shift fails the build. It also means an old score and a new one are comparable only when they carry the same version.
Is a high score a prediction that I will get an interview?
No. The score summarizes Guul's parsing, structure, content and optional job-match checks. It is not a score from an employer's system. Interview decisions depend on the role, other applicants and recruiter judgement, which these checks do not measure.
Can I raise the score by adding more keywords?
Not usefully. Content quality is scored on whether skills are evidenced in the experience that claims them, so a keyword list that is not backed up by anything does little for the score and actively hurts you with the human who reads it next. The productive move is the opposite: make the evidence you already have easier to find.
What are the score bands?
Excellent: 90–100; Strong: 80–89; Good: 70–79; Fair: 60–69; Needs work: 0–59. The bands are a reading aid for the same underlying number, not separate scores.
Do I need an account to check my resume?
You can check a resume without signing in to see its Guul score, top findings and a modeled field preview, within usage limits. Sign in to see the complete basic findings and save your scan. Paid analysis and resume downloads have separate access requirements shown on the pricing page.