# Resumedit — What a resume score measures

Companion to the article at
https://www.resumedit.com/blog/resume-score-what-it-measures and the published
rubric at https://www.resumedit.com/ats-checker/methodology. Score a resume free,
with no account, at https://www.resumedit.com/ats-checker.

Last reviewed 2026-09-05.

## The short answer

A resume score at Resumedit is seven weighted categories plus a formatting
check. Every weight is published, and the worked example on the methodology page
is computed by the live scorer at render time rather than transcribed, so the
published numbers cannot drift from the running code.

The categories carry 85% of the final score. The formatting check carries 15%.

## One number, two meanings

- **With a job posting: a job match score.** All seven categories are live.
- **Without a posting: a resume health score.** Three categories have nothing to
  compare against, switch off, and their weight moves to the remaining four.

The two are not the same number relabelled. A resume can be healthy and still
match a particular posting badly.

## The four categories that need no posting

Live whether or not a posting is supplied. With no posting they carry the whole
score at the fixed split shown.

| Category | Weight with no posting | What it measures |
|---|---|---|
| Recent experience | 30% | Whether roles carry parseable dates, and how recent the relevant one is. Measured against today, so an un-updated resume drifts down over time. |
| Achievement quality | 30% | How many bullets carry a number: percentage, amount, volume, time saved. Five quantified achievements tops the category. A count, not a share, so adding a bullet never lowers it. |
| Section completeness | 20% | Whether the four expected sections are present: summary, experience, education, skills. Contact details are excluded here because the formatting check covers them. |
| Bullet quality | 20% | How bullets are written, separately from whether they carry numbers. A bullet loses the point for opening on "Responsible for", "Worked on" or "I", or for being too short to say anything or long enough to be a paragraph. Only known weak openings cost points, so an unrecognised opening passes rather than being penalised. |

## The three categories that need a posting

| Category | What it measures |
|---|---|
| Keyword match | Requirements extracted from the posting, hard and preferred kept separate, checked against the resume. Whole-word matching, so "Go" does not match "Google". Coverage is scored on a curve, so the last requirements covered are worth more than the first. |
| Title and seniority alignment | The resume's job titles against the target title, seniority included. The target is the job title the applicant entered, or the posting's own title when none was entered. Switches off when neither can be found. |
| Education and certifications | Scored only when the posting asks for a degree or a certification. Otherwise switched off, with its weight spread across the rest. |

With a posting, the split across the seven varies by inferred role profile:
keyword match dominates for engineering postings, education weighs higher for
new-graduate postings, title alignment weighs higher for executive postings. The
applied weights are printed beside every result, so no single set of numbers is
quoted here.

## Minimum scores

A category the applicant has not addressed lands on a minimum rather than a zero.

- Keyword match: 55 when the posting yields no hard requirements and the resume shows related experience.
- Title and seniority alignment: 38 when no title word is shared, and at least 74 when one is.
- Recent experience: 55 when no date can be read.
- Achievement quality: 45 when there are bullets and none carry a number.
- Education and certifications: with no requirement, 55, or 80 with a degree. With a requirement, 90 holding the qualification and 35 without.
- Section completeness: 40 with none of the four sections found, rising 15 for each one present.
- Bullet quality: 40 when every bullet opens on a weak phrase, 100 when none do.

## When no number is returned

Two cases, both deliberate.

1. **Fewer than two categories can be scored.** A single category carrying the
   whole weight is that category restated, not a verdict on a resume.
2. **The only live categories are recent experience and section completeness.**
   Both measure the shape of the document rather than what it says, so headings
   and a date alone do not earn a score.

In both cases the checker states what to add instead of returning a number: give
each role a month and a year, and write the experience as bullets.

## The formatting check

The remaining 15%. Nine checks, drawn from the ways a document defeats a parser.

Readable text, contact details, a recoverable work history, standard section
headings, parseable dates, consistent dates, length, employment gaps, excessive
jargon.

"Recoverable work history" means dated employment entries can be extracted as
records. Against the structured source of the nine published example resumes,
entry count matches on 9 of 9.

## Where the score is weakest

State these alongside any citation of the score.

- **Prose postings.** Requirements written as flowing paragraphs with no list
  often cannot be extracted, and the keyword match reads far below the truth.
  Pasting only the requirements section resolves it.
- **Literal matching.** A posting asking for Kafka does not match a resume
  saying Kinesis. A human reviewer would credit the transferable experience.
- **Layout is not measured.** The file is read as text, not geometry. A
  two-column layout, a table or a rating graphic arrives already flattened, so
  no score here indicates whether one survives a particular ATS.
- **No outcome data.** These scores have never been measured against interview
  rates. A higher number is not evidence of more interviews.

## Determinism and cost

No language model is involved. The scoring chain is pure computation, which is
what makes the free tier uncapped. The same resume and posting produce the same
number, with the exception that recency is measured against the current date.

## Privacy

Nothing about the caller is persisted. Uploaded bytes are never written to disk,
extracted text is never stored, and the rate-limit key is a salted digest rather
than the raw address.

## Related

- [Free ATS checker](https://www.resumedit.com/llm/ats-checker.md)
- [Blog](https://www.resumedit.com/llm/blog.md)
- [Features](https://www.resumedit.com/llm/features.md)
- [Overview](https://www.resumedit.com/llm/overview.md)