Most advice on tailoring a resume assumes you will rewrite the document for every application, which is a large enough job that it tends to get skipped on the applications that matter. The useful version is narrower: a small number of things carry nearly all the signal, and everything else can stay as it is.
Start from the posting, not from your resume
Read the job description once and pull out the nouns. Tools, systems, certifications, methodologies, the name of the thing the team actually does all day. Those nouns are what someone types into a search box later. Greenhouse, to take one applicant tracking system that documents the feature, lets a recruiter switch on full text search across candidate resumes and returns the candidates whose resume text matches the terms entered. A term that is not in your document cannot come back in that result. Adjectives about culture and ambition are not what you are mining for.
Our resume keyword generator does this extraction for you if you would rather paste the posting in and get the list back.
Change three things, not thirty
The parts of a resume that reward tailoring are the ones a reader sees first:
- The headline or summary. This is the one line that should sound like it was written for this posting. If you change nothing else, change this.
- The top two or three bullets of your most recent role. Reorder them so the work closest to the posting sits at the top. You are not inventing anything, you are choosing what leads.
- The skills list. Match the posting's vocabulary where it is honest to do so. If the posting says one thing and your resume says a synonym, the parser may not connect them.
Everything below your most recent role can usually stay untouched.
Match the vocabulary, do not stuff it
Repeating a phrase until it looks unnatural does not help you. Take the one scorer I can speak for, the one behind our own ATS resume checker: it settles each skill named in the posting exactly once, as matched, partial or missing, so a fourth mention of the same term moves nothing. It also caps the result outright when keyword overlap is high but the demonstrated work behind it is thin. Other systems are not obliged to behave the same way, which is the point: you cannot know, so writing for an imagined parser is a bet you cannot check. A human reading a bullet packed with keywords registers something you can predict, which is that you were writing for the machine.
The honest test is whether a person who knows the job would read the bullet and recognise the work. If yes, the keyword is earned. If not, cut it.
Check it before you send it
Once you have made those changes, run the result through our ATS resume checker to confirm the file still parses cleanly and the terms you added are being read. Formatting that survives one parser sometimes breaks in another, and that is worth knowing before you apply rather than after.