Most "ATS scores" you see online are calculated by a single keyword-matching script. Resuma takes a different approach. We run every resume through a 4-stage AI pipeline that mirrors how a careful recruiter would actually read it — extract the role's signals, profile the candidate, draft a tailored response, then evaluate the result.
This post is a behind-the-scenes look at the five signals that pipeline scores you on. If you understand them, you can fix the ones working against you before you ever paste your resume into a tool.
The 4-stage pipeline in one paragraph
Stage 1 reads the job description and extracts what the role actually wants — keywords, seniority, real responsibilities. Stage 2 reads the resume and extracts the candidate's strengths, skills, and experience. Stage 3 produces the optimized resume document. Stage 4 evaluates the result and reports a match score with feedback. The five signals below are the inputs to that final score.
Signal 1: Keyword coverage (weighted by importance, not count)
Why "keyword density" misses the point
Older ATS tools counted how many times each keyword appeared in your resume. That metric is easy to game — and easy to fail. A resume that crams every keyword in once usually beats a resume that mentions one keyword ten times.
What Resuma's pipeline actually does is split keywords into tiers based on how prominent they are in the job description (job title repetition, requirements bullets, "must have" framing). A keyword that appears in a "required skills" bullet is weighted heavier than one mentioned in a passing sentence.
What this means for your resume
Don't pad. Cover the heavy-weighted keywords once, in a meaningful sentence, and stop. If you're missing a heavy keyword and you actually have the experience, the mirror trick post walks through how to add it without rewriting from scratch.
Signal 2: Role and seniority fit
The pipeline checks whether your most recent role actually maps to the role being applied for — and whether the seniority is plausible.
A "Senior Engineer" applying for a "Staff Engineer" role passes this check. A "Marketing Coordinator" applying for a "VP of Marketing" role fails it, regardless of how strong the resume is in isolation.
If you're switching levels (or industries), this signal is the one most likely to drop your score. The fix isn't to lie about your title — it's to use your summary section to bridge the gap explicitly. Our career change resume guide walks through that bridging in detail.
Signal 3: ATS-friendly structure
This is the boring one, and it's the most likely to silently kill your score.
The pipeline expects standard sections (Summary, Experience, Skills, Education), reverse-chronological ordering, single-column layout, and parseable bullet points. If your resume hides experience inside text boxes, uses tables for the skills section, or mixes your dates into a footer, the structure score drops sharply.
If you've got a designed-template resume and your scores keep coming back lower than they should, seven common formatting mistakes is worth a read.
Signal 4: Outcome density (bullets that prove value)
Responsibilities aren't outcomes
A bullet that says "Responsible for managing the customer support team" is a responsibility. A bullet that says "Restructured the support team's escalation flow, cutting average resolution time from 14 hours to 3" is an outcome.
The pipeline treats outcomes — bullets with quantified results, before/after framing, or measurable change — as much stronger evidence of role fit than bullets that describe what you were assigned to do.
How the score weights this
A resume with 12 outcome-style bullets across the most recent two roles scores significantly higher than one with 30 responsibility-style bullets. Quality compounds; quantity plateaus quickly. Our bullet point guide is the most direct way to fix this if your bullets read flat.
Signal 5: Recency-weighted relevance
The pipeline weights your most recent role much more heavily than older ones — which matches how recruiters actually scan. Strong bullets in a 2026 role count more than equally strong bullets in a 2018 role.
This has two consequences for your resume:
- The first bullet of your most recent role is the highest-leverage real estate on the entire document. Make it the most relevant outcome to the role you're applying for.
- Older roles can be condensed. Three tight bullets on a 2018 role beat seven bullets that nobody is going to read carefully anyway.
What the final score actually means
Resuma reports a single 0–100 match score. Internally, that number is a weighted aggregate of the five signals above plus a small adjustment for the resume document's overall coherence (no orphan sections, clean formatting, proper metadata).
A score of 85+ means you're in the strong-match zone for the JD as written. A score of 65–84 means there's relevant fit but missing keyword or structure work. A score under 65 usually means the resume needs more than tweaks — either there's a real gap to address (signal 2) or the resume needs structural rework (signal 3).
What this changes for you
The practical takeaway: if you understand the five signals, you don't need a tool to tell you why your resume isn't landing. You can audit your own resume against the JD using the same logic — heavy keywords covered, role/seniority match, ATS-friendly structure, outcome-dense bullets, recency-weighted relevance.
Tools like Resuma make the audit instant and the fixes faster, but the underlying signals are public. Use them.
If you want to see your own score against a specific job description in seconds, paste both into Resuma — it's free for the first three generations, no card required.