Why Job Match Percentages Are Misleading (And What to Look At Instead)

Short answer: Most match percentages are keyword overlap between your CV and the job ad, normalised into a number. They ignore seniority, mandatory qualifications, recency and logistics — so an "85% match" can be an automatic rejection while a "60%" can be a shortlist. What predicts an interview is requirement coverage, seniority alignment, hard filters and recency of the relevant experience.

What the number is usually made of

Under almost every consumer-facing match score sits some version of this: tokenise the job description, tokenise the CV, compute overlap, weight a few fields, scale to 100. Some products add embeddings so that "revenue growth" and "top-line expansion" count as related. That is genuinely useful for retrieval — deciding which 50 jobs out of 200,000 to show you — and close to useless as advice about whether to apply.

Three structural problems:

It treats all requirements as equal. Losing "SAP" and losing "must hold a valid UAE driving licence" cost the same points. One is a preference; the other ends your application.

It has no concept of level. A CV full of the right keywords at the wrong seniority scores highly. Regional Director experience against a Team Lead role produces a beautiful percentage and a fast rejection.

It ignores recency and depth. A skill you used once, six years ago, for three months, matches the same token as a skill you have led with for a decade.

The 82% that is a guaranteed no

A worked example. A telecoms network architect applies to a "Solution Architect — Digital Banking" role in Riyadh. Shared vocabulary: architecture, stakeholder management, TOGAF, integration, cloud, migration, governance. The overlap is high, so the platform shows 82%.

What the number cannot see: the employer needs core banking migration experience, SAMA regulatory familiarity and financial-domain delivery history. The candidate has none of the three. The real verdict is not a fit — and the honest version of that advice would have saved the candidate a tailored application and two weeks of waiting.

The inverse also happens. A candidate covers nine of ten requirements but describes them in their own company's internal language, so overlap is low and the score reads 58%. That is not a bad fit; it is a badly worded CV, and it is fixable in twenty minutes.

The four signals that actually predict a shortlist

1. Requirement coverage, itemised

How many of the stated requirements does your CV evidence — and, critically, are the missing ones genuinely absent or simply unstated? Those two cases demand opposite responses: skip, versus tailor and apply.

2. Seniority alignment on two axes

Individual scope (depth and complexity) and management scope (team, budget, reporting line) are separate. Matching one and missing the other by two levels is a rejection in both directions — under-levelled candidates get filtered, over-levelled candidates get filtered faster.

3. Hard filters

Licences, certifications, language, nationality quotas, regulatory experience, visa status. Binary, and they override everything else on the page.

4. Recency and duration

Employers read the last five to seven years. A requirement met only in your early career reads as unmet, whatever the keyword parser thinks.

Replace the score with an instruction

The useful output of a matching engine is not a number but a decision, with reasons. DexterCV's Find Jobs is built this way: every live GCC posting carries one of five verdicts — apply directly, tailor before applying, stretch, not a strong fit, no fit — plus the evidence behind it: which requirements you cover, which are missing, which you have but have not presented well, and any visa or logistics condition that would decide it anyway. The percentage still exists internally; it is just not what you are asked to act on.

Before any of that helps, your CV has to be legible to a parser at all. Run it through the free Resume Scanner first — a machine-unreadable CV produces both bad match scores and bad advice.

How to use scores you cannot get rid of

If you are on a platform that only shows a percentage:

• Treat it as a sorting aid, never as a verdict.
• Manually check hard filters first — they are usually in one line near the bottom.
• Compare the ad's title and stated scope to your last two roles for level fit.
• Ignore any score above 75% until you have confirmed the core domain matches; that band is where keyword inflation lives.

FAQ

Are job match scores completely useless?

No. They are good at ranking which jobs to look at out of thousands. They are poor at telling you whether to apply, because they weight all requirements alike and ignore level, recency and hard filters.

Why does my match score change between platforms?

Each platform uses a different parser, vocabulary and weighting, and some compare against a profile you filled in rather than your actual CV. The same CV and job can plausibly score 55% and 85% on two sites.

Can I game a match score?

You can raise it by copying the ad's keywords into your CV, and recruiters recognise the result immediately. Aligning genuine experience to the employer's vocabulary is legitimate; inserting skills you do not have fails at interview and damages your credibility.

What is a good match percentage to apply?

There is no reliable threshold, which is the whole problem. Use requirement coverage instead: apply when your CV evidences most of the stated must-haves and no hard filter blocks you.

Does the ATS itself score me like this?

Applicant tracking systems mostly rank on knockout questions, keyword presence and recruiter-defined criteria, not on a consumer-style percentage. Passing the ATS is largely about parseable formatting and using the employer's terminology accurately.