Every CV runs through a deterministic rules engine: 36 rules, three categories, fixed weights. The same CV always gets the same score. Each rule below is explained with its own inspector card and has a dedicated detail page.
[ACTION: Led] a team of 5 engineers migrating a monolith serving [SCALE: 1.4M active users] to Go microservices, delivering [OUTCOME: 38% faster responses and 20% cost savings].
Understand the automated filter between your application and the recruiter's desk. Insights from Harvard Business School research on recruiting software.
A comprehensive walkthrough on crafting a professional resume that passes automated Applicant Tracking Systems and impresses recruiters.
Transform a mediocre resume into an elite application using Google's X-Y-Z principle and the Action-Scale-Outcome (ASO) methodology.
Sub-scores start at 100; every failing rule subtracts its own weight.
30%
Mechanical
45%
Content
25%
ASO
final = mechanical * 0.30 + content * 0.45 + aso * 0.25 subScore = clamp(0, 100, 100 - Σ(weight × severityMultiplier × failRatio))
Eight steps from upload to score.
The full list grouped by category. Open any card to read how the rule is measured, why it matters and how to fix it.
The Tier 1 mechanical filter: can an ATS read the file at all? Two rules here are hard blockers — when they fail, the score cannot exceed 45 no matter how good the content is.
MEC-00 checks whether the PDF contains a real text layer. If character density per page area is near zero, the CV is a scanned image: the ATS reads nothing and your application is stored as an empty record. It is a hard blocker — when it fails, the final score is capped at 45.
Read the rule→MEC-01 looks for a column corridor that breaks ATS reading order. Line start x values are clustered into 15pt bins; if a secondary cluster sits at least 22% of the page width to the right of the primary one, holds 5+ lines and overlaps it vertically by half its own range, the layout is multi-column. This is a hard blocker.
Read the rule→MEC-02 checks whether body text uses a font that ATS engines decode reliably. Font names are normalized by stripping subset prefixes and style suffixes; if non-whitelisted fonts cover more than 20% of characters, the rule fails.
Read the rule→MEC-03 measures whether body text drops below 10pt. Point size is read from each text run’s transform matrix; the size carrying the most characters counts as body size. Body text under 10pt — or a large share of text under 9pt — fails the rule.
Read the rule→MEC-04 looks for skill progress bars, star ratings and icon glyphs mixed into the text flow. It catches both real block characters (████░░) and the repeated symbol runs (ˆˆˆˆ‘‘) they turn into when the font is not embedded.
Read the rule→MEC-05 checks, field by field, whether email, phone, LinkedIn and GitHub sit inside the page header or footer band. The header is the top 6% of the page, the footer the bottom 6% — many ATS engines process those bands separately from the main flow and skip them.
Read the rule→MEC-06 judges CV length against your career level: one page for graduates, one to two pages mid-level. Documents over three pages, or ones that end halfway down page one, fail — and the rule also measures meaningful content density per page.
Read the rule→MEC-07 inspects the uploaded file name. Generic names such as "cv.pdf" or "document-final(2).pdf" get lost in a recruiter’s download folder, and some ATS engines pre-fill the candidate name field from the file name.
Read the rule→Sections, contact details, dates and job-posting fit — the information layer of the CV.
CON-01 checks section headings against the heading vocabulary ATS engines are trained on. Experience placed under a creative heading such as "My Journey" lands in an "other" field instead of the experience field, so the system shows you with zero years of experience.
Read the rule→CON-02 verifies that GitHub, LinkedIn and portfolio addresses are genuinely clickable in the PDF annotation layer. If a URL appears as text with no matching link annotation, the link is dead — nobody retypes a 40-character address, so your portfolio is never opened.
Read the rule→CON-03 looks for personal data that does not belong on a CV: date of birth, marital status, military service, national ID, full street address and — in sectors where photos are not expected — an embedded portrait.
Read the rule→CON-04 checks whether the CV opens with a professional summary and whether that summary is buried in clichés. Adjective stacks like "dynamic, dedicated team player" match no job-posting keyword, so the first three lines a recruiter reads are wasted.
Read the rule→CON-05 counts how many distinct date patterns the CV uses. When "03/2022", "March 2022", "2022" and "Mar ’22" mix in one document, the ATS date parser leaves what it cannot recognise empty and your total years of experience are computed wrong.
Read the rule→CON-06 finds gaps longer than six months between consecutive roles and checks whether the CV explains them. The gap itself is not the problem — leaving it unexplained is.
Read the rule→CON-07 checks the canonical spelling of technology names, first-person pronouns in bullets and punctuation consistency. A CV that says "Phyton" is invisible to an ATS searching for an exact "Python" match.
Read the rule→CON-08 looks for the "References available upon request" line and for referee contact details printed on the CV. The first carries no information; the second shares a third party’s data without consent.
Read the rule→CON-09 checks, in sectors where a portfolio is expected, whether a projects section exists and whether each project carries a verifiable link. If the sector profile sets projectsExpected to false, the rule never runs.
Read the rule→CON-10 measures keyword overlap between the job posting you supply and your CV. Technical terms are extracted from the posting and searched in the CV; below 50% overlap most ATS rankings push you to the bottom of the list.
Read the rule→Volume, grouping and signal quality of the skills section.
SKL-01 measures how many technologies the skills section lists and how often terms repeat. Modern ATS engines flag keyword-density anomalies as spam, and human readers assume none of the 40 listed tools is deep.
Read the rule→SKL-02 checks whether skills are grouped under category labels. A flat, uncategorised list hides the technology a recruiter is scanning for and pushes every skill into the parser’s "other" bucket.
Read the rule→SKL-03 measures the share of unproven soft-skill claims in the skills section. Technical postings filter on technical terms, so "communication, leadership, teamwork" produces no match. In sectors where soft skills are central the rule is skipped.
Read the rule→Document-wide rules: ordering, section layout and bullet structure.
GEN-01 checks that experience and education entries run newest to oldest. Some parsers take the "current role" field from the first entry in the list, so a broken order makes an old internship look like your current job.
Read the rule→GEN-02 evaluates whether the section order matches your career level. Education on top of an experienced engineer’s CV signals stalled progression; for a graduate, education and projects belong first.
Read the rule→GEN-03 measures bullet word counts and how much of the experience section is written as paragraphs. Bullets over 30 words go unread in a six-second scan, and paragraph blocks are often ingested as a single field, so the achievements inside them are never parsed.
Read the rule→GEN-04 checks that every experience entry carries all three of title, company and dates. An ATS opens an experience record from that triple; with one missing, the record is incomplete or never created.
Read the rule→GEN-05 measures whether bullets describe duties or achievements. Patterns like "was responsible for" copy the job description back at the reader and make you indistinguishable from 200 other candidates with the same title.
Read the rule→Action–Scale–Outcome: whether each experience bullet carries an action, a scale and a measurable outcome. 25% of the score comes from here.
ASO-01 is the first leg of the ASO (Action–Scale–Outcome) model: does every experience bullet open with a strong, active verb? Phrases like "was involved in" or "assisted with" leave it unclear whether you did the work, and they cost you a level of perceived seniority.
Read the rule→ASO-02 checks whether bullets state the size of the work: users, request volume, data size, budget, team size. Without scale, "optimised the database" could describe a hobby project or a billion-row OLTP system — and the reader assumes the smaller one.
Read the rule→ASO-03 is the heaviest leg of the ASO model: does the bullet answer "what changed?" with a number? Without a percentage, duration, amount of money or count, the bullet is a job description. A measurable outcome is the single concrete argument a recruiter passes to the hiring manager.
Read the rule→ASO-04 measures the variety of verbs opening your bullets. The same verb three or more times makes bullets look like copies of one another and suggests a narrow range of work.
Read the rule→ASO-05 looks for hollow corporate phrasing — "value-added solutions", "synergy", "360-degree perspective". Such phrases take space, match no keyword and read as a red flag to technical reviewers.
Read the rule→Verification of the GitHub profile in sectors that write code. Never runs when the sector profile sets usesCodeRepos to false.
GIT-01 checks, in sectors that write code, whether the CV carries a clickable GitHub address. If the sector profile sets usesCodeRepos to false, no GIT-* rule runs at all.
Read the rule→GIT-02 checks for a README in the username/username repository. The profile README is the first page a hiring manager lands on; without it, visitors cannot tell which of your repos matters.
Read the rule→GIT-03 measures the originality of pinned and recently updated repositories: fork ratio, coursework/clone naming patterns, and missing READMEs or descriptions. A profile made of homework and clones reads as "has never built anything alone".
Read the rule→GIT-04 searches public repositories for committed API keys, tokens and credentials. Files committed by convention — .env.example and friends — are excluded. A finding makes this a hard blocker: the score is capped at 45, because in many companies a leaked key alone ends the process.
Read the rule→GIT-05 looks at the last push date on your profile. A profile idle for months contradicts a CV that calls you a continuous learner and raises doubts about how current your skills are.
Read the rule→Sector filter: depending on the sector profile you pick, some rules never run or return `skipped` — a skipped rule never costs points.
MEC-00 checks whether the PDF contains a real text layer. If character density per page area is near zero, the CV is a scanned image: the ATS reads nothing and your application is stored as an empty record. It is a hard blocker — when it fails, the final score is capped at 45.
MEC-00 — Read the ruleMEC-01 looks for a column corridor that breaks ATS reading order. Line start x values are clustered into 15pt bins; if a secondary cluster sits at least 22% of the page width to the right of the primary one, holds 5+ lines and overlaps it vertically by half its own range, the layout is multi-column. This is a hard blocker.
MEC-01 — Read the ruleASO-03 is the heaviest leg of the ASO model: does the bullet answer "what changed?" with a number? Without a percentage, duration, amount of money or count, the bullet is a job description. A measurable outcome is the single concrete argument a recruiter passes to the hiring manager.
ASO-03 — Read the ruleCON-10 measures keyword overlap between the job posting you supply and your CV. Technical terms are extracted from the posting and searched in the CV; below 50% overlap most ATS rankings push you to the bottom of the list.
CON-10 — Read the rule