Every tool that compares a resume against a job posting hands back the same two things: a percentage, and a list of terms you are missing. The percentage is the least useful part of what it just produced. The comparison sitting underneath it is doing something quite specific, and knowing what that is tells you which of the differences it found are worth acting on and which are noise.
What a comparison is actually doing
It is resolving nouns. The tool reads the skills named in the posting, looks for each one in your document, and records a result per skill. The one I can describe precisely is our own, because I can read it: it sorts each skill the posting names into matched, partial or missing, and partial means a significant word of a multi-word skill turns up while the full name does not. Your resume says pipelines, the posting says data pipelines.
It also treats known aliases as the same thing, in both directions. Kubernetes and k8s, Amazon Web Services and AWS, continuous integration and CI/CD. That is a design choice rather than a law of matchers: a comparison only knows the aliases somebody put into it, so a synonym nobody thought of reads as a gap even when the work is identical.
The three results are not equally worth your time
- Partial is where nearly all the value is. A partial says the substance is already in your document and only the naming is off. You did the work, you described it in your own words, and the posting uses a different phrase for it. Fixing one takes a minute and costs you nothing in honesty.
- Missing splits into two very different cases. If you genuinely do not have the thing, that is not an editing problem, it is information about whether this application is worth the hour. If you do have it and never wrote it down, add it, which is the single most common real gap.
- Matched is finished. There is nothing to collect here. Ours settles each skill named in the posting exactly once, so a fourth mention of a term already counted as matched moves nothing at all.
That triage is the whole job. Everything else a comparison shows you is commentary on it.
What a comparison cannot tell you
It is bounded by its own vocabulary. A skill the posting names but the tool has never heard of does not appear on either side of the ledger, so a clean comparison is not the same as a complete one.
More importantly, it reads nouns and not fit. It does not know whether your experience is close to the role, whether your bullets show an outcome or only a responsibility, or whether your titles sit at the level being hired for. Our scorer weighs all of those alongside coverage, and coverage is the largest single input while still being a minority of the total. That is why a resume can carry every term in the posting and still not score well. Ours goes further and caps the result outright when overlap with the posting is high but the bullets behind it carry no measurable outcome and the experience is not close to the role, on the grounds that a document in that state is describing the posting rather than the person.
The number you are told to aim for
Tools on this search publish a score to target. Notice who sets it: the company that also built the scorer producing it. A target defined that way measures your agreement with one vendor's model, not your agreement with the employer, and two tools can read the same resume and the same posting and disagree about both the score and the gaps.
Treat the percentage as a progress bar for the triage above, not as a grade. When the partials are resolved and the real gaps are named, you are done, whatever the bar says.
Running one properly
Paste the posting into our resume keyword generator to get the extraction step done first, then work the list against your document rather than staring at both texts side by side. When the edits are in, run the file through the ATS resume checker to confirm the terms you added are actually being read and the formatting still parses.
If the comparison turns up more to fix than you expected, the mechanics of the edit are covered in tailoring a resume without rewriting it, and the question of how far to take it before the effort stops paying is covered in where tailoring pays off.