ATS scores: what a real 6-component rubric measures (and what keyword counters miss)

Paste a job description, paste your resume, receive a number. The ritual feels scientific, and the number feels like a verdict. Before you reorganize your professional life around it, it is worth asking the question nobody asks the meter: what exactly is being measured?

For most free checkers, the honest answer is word overlap. Tokenize the job description, tokenize the resume, count the intersection, draw a progress bar. It is cheap to run and satisfying to watch go up. It is also a measurement of camouflage, not of fit — and the difference matters, because people make real decisions off these numbers.

What a keyword counter cannot see

• Context: it counts that you mentioned a tool, not whether you ever shipped anything with it. A sidebar stuffed with nouns scores like a career built on them.
• Level: a counter has no concept of seniority — five months and five years of the same keyword are the same token.
• Weight: it treats every term as equally important, when the job description plainly does not.
• Gaming: stuffing terms raises the score while making the document worse. Any meter you can cheat in five minutes is measuring the cheating, not the resume.

What a weighted rubric measures instead

A score worth trusting has to read the resume the way the hiring process does: skills first, in context, with structure as the price of admission. When we built ours, we made it six components with declared weights — partly for accuracy, partly so members can see exactly what moves the number and why:

• Hard skills match — 30%. The heaviest weight, because the named tools and competencies of the role are still the spine of any screen.
• Experience alignment — 20%. Whether your trajectory and level actually fit the role, not just its vocabulary.
• Soft skills match — 15%. Evidenced in your bullets, not listed in a sidebar.
• Keyword density — 15%. Yes, keywords still count — as one component of six, not the whole meter.
• Section completeness — 10%. Whether the resume has the structure a parser expects to find.
• Format and readability — 10%. The unglamorous failure mode: a strong resume in a layout that parses badly.

The weights are a judgment call, and we would rather declare them than hide them behind a percentage that looks like physics. Reasonable people could argue for a different split. What is harder to argue for is a single undisclosed formula that turns out, on inspection, to be a word-overlap count with a confident font.

A resume that scores 95 on a keyword counter and 60 on a structural rubric is not a 95. It is a 60 wearing a 95’s clothes.

A word about the robot-rejection myth

You have probably heard that most resumes are auto-rejected by ATS robots before a human ever sees them. We have never found a version of that statistic that survives attribution, so we will not repeat it. The duller truth: an ATS is mostly a filing-and-ranking system, and recruiters search and skim what it surfaces. The risk is not execution by robot — it is being filed where nobody looks. A good score should measure exactly that: not whether a machine will reject you, but whether the right human will ever be shown you.

How to use any score honestly

As a diagnosis, not a destination. The dishonest move is editing for the meter — synonym-swapping until a counter applauds. The honest move is closing real gaps: surfacing the experience that matches, restructuring what parses badly, cutting what buries the signal. That is why our score arrives attached to the documents themselves — a number with nothing to act on is a horoscope. And if a simple match-rate gut check is genuinely all you need, we keep an honest comparison of the tools that do that well.