Why Two CV Checkers
Give Different Scores
Last updated 30 September 2026
Because they are not measuring the same thing, and almost none of them say what they are measuring. Some scorers add points upward for features they recognise; others, Careery included, start every dimension at full marks and deduct only for defects they detected and showed you. The same CV reads far lower on the first kind. Neither number is wrong, and neither can be compared with the other.
A Score Is a Reading From One Instrument
Two thermometers disagreeing tells you something is wrong. Two CV scorers disagreeing usually tells you nothing at all, because they are not two readings of one quantity. One may be counting keywords against a job advert you pasted in. One may be checking whether a parser can find your sections. One may be adding a point for every feature it recognises. Put the same document through all three and you get three numbers that share a scale, a maximum, and nothing else.
The useful question is therefore not which score is right. It is which score you can read the working for. Everything below is ours.
The Mechanic, In One Line
dimension points = max(dimension floor, weight - sum of capped deductions)That is the whole thing. Every dimension starts at full marks. Points are lost only for a defect that was actually detected and that is actually shown to you in the issues list. A CV with no detected defects scores 100, and the absence of a flag never silently costs anything. This is the inverted mechanic, and it is how the comparable products work too: VMock's SMART Score, Resume Worded, Rezi, Enhancv and Kickresume all score down from a ceiling rather than up from zero.
The metric this replaced accumulated upward from zero and was called employability, which claimed something a CV cannot support. A competent CV routinely landed in the forties under it. If a tool you are using rates a CV you know to be decent in the forties, that is the likeliest reason, and it is a property of the scale rather than of your CV.
The 6 Dimensions and What Each One Weighs
Weights out of 100, rendered from the scoring module itself rather than written here, so this table cannot drift away from the product.
Evidence and Impact30 points
Quantified outcomes, strong verbs, results rather than duties, and the scope you worked at.
Relevance Signal15 points
A declared direction, checkable named skills, and content that supports both.
Clarity and Communication18 points
Bullet length, active voice, no filler, no wall of text, spelling and consistent tense.
Structure and Section Logic15 points
Section order, reverse chronology, complete dates and clear headings.
Machine Readability12 points
A parsable file, a single column body, no text locked inside images, headings a parser recognises.
Professional Hygiene10 points
A professional email, contact details that work, no typo in your own name or titles.
No dimension can fall below a quarter of its own weight, so one bad area cannot annihilate the total.
The Bands, and the Only Real Boundaries
The boundaries are 40, 55, 70, 85. There are no others. If you see a different cut-off quoted about a Careery score, including on another page of this site, it is wrong and we would like to know.
Where Our Own Number Has Limits
Three of them, because a scale published without its limits is not published.
- A very short document can score at or near 100. The content detectors only fire above 60 whitespace tokens, so a page that says almost nothing loses almost nothing. A high score means no defect was detected, not that the CV is persuasive.
- Deductions are capped. Criticals take at most 30 points from a dimension, majors at most 25, and suggestions at most 3 across the whole CV. Style problems alone cannot remove more than 70 points, a CV cannot pass below 40 without a genuine structural failure, and nothing lands below 15. So a long list of small faults moves the number less than the length of the list suggests it should.
- Above 82, a critical issue is treated as a calibration error on our side and downgraded to a major, because a document that scores that highly and also has a critical fault means the detector and the band disagree.
The scorer also reads English only. A CV written mostly in another script is refused rather than scored, and what happened when we measured ours on real Arabic CVs is the clearest illustration on this site of a confident score that meant nothing.
Five Questions to Ask Any CV Score
1What are the dimensions, and what does each weigh?
If a tool will not tell you, the number is a single opaque figure and you cannot act on it. You can only act on a dimension, because a dimension names the kind of fix.
2Does it accumulate upward or deduct downward?
This is the single biggest cause of two tools disagreeing about the same document. Upward scoring rates what it recognised; downward scoring rates what is wrong with it. The same CV reads far lower on the first than the second, and neither has made an error.
3Does it need a job description?
A keyword score with no target role is not measuring relevance to anything. If a tool gives you a keyword number without asking what you are applying for, ask what the keywords were compared against.
4What does it do when nothing is wrong?
A scorer that cannot return full marks has a ceiling below 100, and every number it gives you sits on a scale whose top you were never told.
5Which version of the tool produced this?
A scorer that changes and does not say so makes two different instruments look like one. Careery records the rubric version beside every score for this reason.
Five Comparisons That Are Not Comparisons
- A score from a tool that needs the job advert, against a score from a tool that does not. One is measuring fit to one posting, the other is measuring the document.
- A score out of 100 from a tool with a hidden ceiling, against a score out of 100 from a tool that can genuinely return 100.
- Two scores from the same tool taken weeks apart, if the tool changed in between and does not publish a version.
- An overall number against a single dimension. The ATS or formatting sub-score of one tool is not the overall score of another, and quoting one as the other is how a threshold nobody uses gets into circulation.
- Any score against a recruiter's opinion. Nothing here has measured what a human does with your CV, and a tool that claims otherwise has not measured it either.
Which Instrument Took Your Reading
A Careery score is produced by two things, and both can move it: the deterministic rubric that runs in your browser, and the model assessment that feeds it. A stored score is therefore a pair of versions rather than a bare number. The rubric version at the time of writing is 2026-09-17.1.
This matters for the most ordinary case there is: you fix your CV, re-upload it, and the number moves. That difference is a measurement of your editing only if the instrument did not change in between. When it did, the two numbers came from two different scorers, and comparing them tells you about us rather than about your CV.
Every figure on this page is read from src/lib/cvStrength.ts at build time, and a vitest guard fails the build if the page and the scorer ever disagree.
Where to Go Next
- What applicant tracking systems actually do. The other number people quote at each other, and what the parsers really reject.
- Can a CV checker read an Arabic CV?. A worked example of a scorer producing a confident number it had no right to.
- A good score and still no interviews. What a document score cannot tell you, and where to look instead.
- How to find the skill gaps on your CV. The dimension most scores punish hardest, and how to close it for nothing.
Frequently Asked Questions
Why do two CV checkers give my CV different scores?
Because they are not measuring the same thing on the same scale, and almost none of them tell you what they are measuring. A score is a reading from one instrument. Two readings are only comparable when both instruments are named, both scales are published, and both were taken on the same version of the tool. Careery publishes its dimensions, their weights, its band boundaries and the rubric version that produced your number, which is why a Careery score can be compared with another Careery score. It cannot be compared with a number from a tool that has published none of those things, and neither can that number be compared with ours.
Which CV score should I believe?
The one whose working you can read. Ask three questions of any score: what are the dimensions and what does each one weigh, does the tool need a job description to produce the number, and what happens to the score when nothing is wrong. A tool that accumulates points upward from zero for features it recognises will rate a competent CV in the forties, because it is scoring what it spotted rather than what is wrong. A tool that deducts from full marks for defects it detected and showed you will rate the same CV much higher. Neither is lying. They are different instruments.
How does Careery calculate CV Strength?
Every dimension starts at full marks and loses points only for a defect that was actually detected and is actually shown to you. The arithmetic is dimension points equals the dimension floor or the weight minus the capped deductions, whichever is larger. There are 6 dimensions weighted 30, 15, 18, 15, 12, 10 to a total of 100. A CV with no detected defects scores 100. The absence of a flag never silently costs anything, so there is nothing on the page you were not told about.
What is a good CV Strength score?
Careery bands the number rather than leaving you to interpret it. 85 to 100 is "Strong CV, ready to send to most employers"; 70 to 84 is "Solid CV, a short list of fixes will lift it"; 55 to 69 is "Developing CV, the substance needs work"; 40 to 54 is "Weak CV, rebuild the evidence and the structure"; 0 to 39 is "We could not read this document properly". The boundaries are 40, 55, 70, 85. Those are the only thresholds in the product: any other cut-off you see quoted about a Careery score is not one we use.
Can a bad CV score highly?
Yes, and this is the most useful thing to understand about any scorer. Careery's content detectors only fire on a document of at least 60 whitespace tokens. Under that, there is not enough text to judge, so nothing is deducted and a very short document that says almost nothing can score at or near 100. A high score means no defect was detected, not that the CV is persuasive. That is a limit of the instrument and we would rather print it than let you discover it.
Why did my score barely move when I have lots of problems?
Deductions are capped, deliberately. Within one dimension, criticals can take at most 30 points, majors at most 25, and suggestions at most 3 across the whole CV. No dimension can fall below a quarter of its own weight. Style problems alone can never remove more than 70 points in total, a CV cannot go below 40 without a genuine structural failure, and nothing lands below 15. The caps exist because an uncapped scorer turns a long list of small faults into a number that reads as a bug.
Does Careery score my CV against a job description?
Not for CV Strength. There is deliberately no generic keyword score in it, because a keyword count without a target is a measure of nothing: Jobscan and Careerflow both require a job description before they will score keywords, Rezi files keywords under optimisation rather than quality, and VMock has no keyword module. Careery computes a Relevance Signal from the CV alone. Matching against a specific advert is a separate number, and scoring a CV against a job is a separate tool.
Why does my Careery score change between uploads of the same CV?
Two reasons, and both are visible. A Careery score is produced by a deterministic rubric that runs in your browser and by a model assessment that feeds it, so a score is stored as a pair of versions rather than a bare number. The rubric version is 2026-09-17.1. If the rubric changed between your two uploads, the two numbers came from two different instruments and the difference between them is not a measurement of your editing. That is exactly why the version is recorded and published.
More on your CV and your job search
- Can a CV Checker Read an Arabic CV?No. What our English-trained scorer did when measured on real Arabic CVs, and the script rule that now makes it refuse instead of guessing.
- Job Adverts Are Not VacanciesAn advert count measures advertising, not jobs. ONS's own monthly figures for one occupation swing by more than three times inside a year.
- Best online courses to add to your CV or resume in 2026Which courses are worth the line, what employers value, and how to write the entry.
- How to Update Your CV: What to Add and What Comes OffUpdating a CV is an editing job, not a writing job. The order that makes it quick, plus a ninety minute version.
- How to Find the Skill Gaps on Your CVWhich skills you are actually missing, how to tell a real gap from a badly described one, and which to close first.
- How to Choose Which Job to Aim For After GraduatingWork backwards from real job adverts rather than forwards from your degree. Five steps to a target you can reach.
A number you can read the working for
Upload your CV and see the score broken into its 6 dimensions, with every deduction attached to the issue that caused it. Nothing is taken off for something you were not shown.