Resumes
Resume Scoring
What resume match scores actually measure, how to read a 67/100 without stuffing, and which fixes to make first so the number moves for the right reasons.
Resume scores feel official. A progress bar is a powerful object. It is also usually a similarity score against one pasted job description, plus some format heuristics, sold as if it were the employer's ATS.
It is not. Greenhouse does not import your Jobscan number. A recruiter does not see "84." They see a preview and a search hit. The score is a flashlight, not a verdict.
This page is how to use the flashlight. It sits in the Ultimate ATS Resume Guide after parsing and keywords.
What typical scores mix
Consumer tools quietly blend different things:
| Ingredient | Useful? | Failure mode |
|---|---|---|
| Parse / structure checks | Yes | Ignoring them and stuffing instead |
| Exact keyword overlap with the JD | Sometimes | Copying the posting to 100 |
| Semantic similarity | Better | Still cannot check truth |
| Length, tense, "measurable bullets" | Weak heuristic | Punishing senior short resumes or honest non-metrics |
| Hard-coded "banned words" | Noisy | False panic over normal English |
If the tool will not tell you why you lost points, you cannot act. A single number without a gap list is a slot machine.
Employer-side ranking is a different machine anyway: How Applicant Tracking Systems Rank Candidates. Your job is still parse + honest overlap + human skim—not a third-party bar.
Read a 67 as a punch list
Suppose a score decomposes (Kyrolane-style) as:
- Parse: contact in header, two-column skills → fix tonight (format)
- Must-haves present without proof: Salesforce listed, no Salesforce bullet → write the bullet or drop the skill
- Must-haves missing that you actually have: posting says "Gainsight," you wrote "CS platform" → rename once
- Must-haves missing that you do not have: required RN license and you are not an RN → do not add it; consider skipping
- Nice-to-haves: extra points for a tool you touched once in 2019 → optional, not a stuffing target
That 67 can become a 78 with format + two honest renames and still be a better document than a stuffed 91.
What should move first
Order of operations, same as the pillar:
- Parse. A score that cannot see your skills will tell you they are missing. They are not missing. They are unextracted.
- Must-have coverage you can defend. Exact term once + proof bullet. Keywords.
- Human skim. If a friend cannot name the job after seven seconds, the score is optimizing the wrong reader.
- Nice-to-haves. Only if they are true and cheap (reorder skills, not a new autobiography).
Score session — 25 minutes
- Pick one real posting in your actual lane
- Run parse checks first; ignore keyword points until paste works
- List must-haves: have with proof / have without proof / do not have
- Fix wording and bullets only for the first two buckets
- Re-score once
- Stop if must-haves you possess are covered and skim still names the role
- Log interview rate for this resume version, not the integer
Related reading
- Ultimate ATS Resume Guide — Parse → Proof → Pitch
- Resume Keywords Guide — coverage without stuffing
- Resume Parsing Explained — why scores lie when extraction fails
- Resume Mistakes — stuffing and unverifiable metrics
- Get started with Kyrolane