How scoring works
No black box. Here’s what happens to your resume text, and why you’ll get the same score if you run it again.
The five categories
The five categories are not weighted equally. What you accomplished (Impact & Quantification) counts most, and the things a reader resolves first — recency, timeline, contradictions (Red Flags) — count next; formatting and brevity matter, but they are the cheapest to fix and weigh the least. We don’t publish exact percentages, because the split can shift as the checks evolve — what stays fixed is the order: substance outweighs formatting.
- ATS Readability & Format
- Can an ATS parser and a human reader extract this resume's structure cleanly? Table/column artifacts, missing sections, inconsistent dates, missing contact info all hurt this.
- Machine-readable textContact info presentExperience section foundEducation section foundSummary or profile foundNo table/column artifactsStandard section headingsConsistent formatting
- Impact & Quantification
- Do bullets show impact with numbers, outcomes, and specifics rather than vague duties?
- Bullet points usedBullets carry numbersStrong bullet openersOutcomes over dutiesConcrete metrics
- Keywords & Skills
- Does the resume cover skills/keywords relevant to the roles it's clearly targeting?
- Skills section foundSkills backed by resultsRole-relevant keywordsSkills coverageIndustry terminology
- Brevity & Clarity
- Is it concise, well-organized, and free of filler, jargon, or first-person pronouns?
- Length fits experienceNo first-person voiceFree of generic phrasingConcise bulletsProfessional tone
- Red Flags
- Unexplained gaps, inconsistent formatting, typos, or anything that would make a recruiter pause.
- No outdated conventionsConsistent date formatRecent dated activityTenure reads steadyMonth and year on datesSpelling & grammarNo unexplained gapsInternally consistentNothing that makes a recruiter pause
What actually computes your score
Your score is calculated by rule, not by a model. Every one of the five category scores comes from deterministic checks run over your extracted text — length, detectable sections, contact info, date consistency, layout artifacts that scramble parsers, how many bullets carry a number, how many describe a result rather than an assignment, how long each one runs, and how your vocabulary compares to the role the resume is targeting.
Two checks are the exception, and they’re the two no rule can settle honestly: spelling & grammar and anything that makes a recruiter pause. A spellchecker that flags Kubernetes or your employer’s name is worse than no spellchecker, and there is no rule for “a reader would hesitate here.” Those two go to a language model. Nothing else does.
This is also why a check can come back unscored. If we can’t tell which role your resume targets, we don’t guess — the keyword checks say they didn’t run rather than passing you by default. A checker that quietly passes what it never looked at is a checker you can’t use.
The word lists
Two of our checks used to be judgment calls by the model. They aren’t any more — weak bullet openers and empty buzzwords are finite lists, so we match them exactly instead of asking an AI for an opinion. That makes those checks reproducible, free to run, and possible for you to argue with.
We publish a sample of each list below rather than the whole thing — the full lists are part of what you’re using us for. What we promise instead is better: wherever one of these checks flags your resume, your report quotes the exact term and the exact line it came from, so you’re never told “weak wording” without being shown the word.
One thing worth being straight about: no applicant tracking system penalises these words. Parsers don’t care. Human readers discount them, because a phrase every candidate uses tells a reader nothing about you. That’s the only reason we flag them.
Weak bullet openers (17 on the list) — a sample
Flagged only when a bullet starts here — these words are ordinary English mid-sentence, and flagging them there would just be noise.
- was responsible for
- duties included
- tasked with
- participated in
- involved in
- helped with
- worked on
- worked with
Empty buzzwords (50 on the list) — a sample
Self-descriptions a reader can’t check. Matched anywhere in the text, once per term.
- results-driven
- detail-oriented
- team player
- hard worker
- hard-working
- thought leader
- self-starter
- go-getter
Same resume in, same score out
Not approximately — exactly. Because the score is computed by rule, the same text scores identically every time, and every point you lost traces to a named check on your results page. If a number moves, it’s because your resume changed.
The two model-judged checks don’t feed the five category calculations — they can only deduct from Red Flags when they actually find something, and every response is validated against a strict schema first. If that pass fails, your score and every other check still stand: those two are marked unscored and we say so on your results page. We’d rather show you a report with two honest holes in it than a confident guess.