A 92% match on a job listing means your CV shares a lot of vocabulary with that job ad. That is the whole job match score meaning. It is not an assessment of whether you will get an interview, whether your experience is senior enough, or whether the company can legally hire someone who needs a residence permit. No board computes that last part, and for a non-EU candidate it is the only variable that decides whether the application was ever worth sending.

This matters because the number feels like feedback. After six weeks of applying and hearing nothing, a high percentage looks like confirmation that the strategy is right and the market is just slow. Usually it means the opposite: you have been spending your applications on roles that score well on text overlap and fail on the one filter nobody shows you.

So the score is worth reading. Just not the way the interface invites you to read it.

Job boards show a match percentage because it increases the chance you click apply. That is its commercial purpose. A candidate who sees 89% applies more often than a candidate who sees nothing.

It is calculated before any human at the employer has seen your name. In most cases it is calculated before the employer's own system has received your application at all, because the board's matching engine and the employer's applicant tracking system are two different pieces of software owned by two different companies. The board scores you against the text of the advert. The employer, later, scores you against the role as the hiring manager actually described it in the intake meeting, which is frequently not what ended up in the advert.

Those two assessments routinely disagree. You can hit 95% on a board and get rejected the same afternoon, because the recruiter's first screening question was "can we sponsor this person and does the budget clear the salary criteria", and the answer was no.

The number you're staring at isn't a hiring decision

The mechanics vary by platform, but the family of techniques is narrow. Nobody is doing anything mysterious.

- Keyword and phrase overlap. The advert mentions Kubernetes, Terraform, CI/CD and "stakeholder management". Your CV mentions four of those. Score goes up. This is the bulk of most scores.

- Title similarity. "Data Engineer" against "Senior Data Engineer" is a near match. "Data Engineer" against "Analytics Engineer" is weaker, even when the work is the same in practice.

- Structured field comparison. Years of experience, education level, location, sometimes language. If you filled in a years-of-experience field and the advert says five plus, that is a clean numeric comparison.

- Semantic similarity. Newer systems use embeddings, so "led a team of six" and "people management experience" can register as related without sharing words. This is a real improvement, and it is also why scores shifted on several platforms without anyone's CV changing.

- Behavioural signals. Some platforms fold in what people with profiles like yours applied to and got hired for. That is recommendation-engine logic, not fit assessment.

None of those inputs includes sponsor status. There is no field in the advert's structured data that says "this employer holds recognised sponsor (erkend referent) status with the IND and has budget approval to file". Even where an employer ticks a "visa sponsorship available" box, that box is self-declared marketing copy, and the gap between what those boxes claim and what the employer can actually do is wide enough that it needed its own breakdown of real versus fake sponsorship ads.

Dutch employers mostly run standard international applicant tracking systems, and ats keyword matching netherlands works no differently here than anywhere else: parse the document, extract fields, rank against required and preferred criteria. What is specific to the Netherlands is the screening logic layered on top. A recruiter at a recognised sponsor working through a stack of applications for one senior backend role is filtering on work authorisation early, because the candidates who need a permit carry extra weeks of process and a cost the hiring manager has to approve.

The ATS ranks you on text. The human ranks you on whether you can start.

Upload one PDF to four platforms and you will see four numbers for the same vacancy. People assume one of them is broken. All four are working as designed.

The differences come from parsing first. A two-column CV, a skills section rendered as a graphic, a header inside a table: each platform's parser handles those differently, and a parser that misses your skills block scores you on whatever it did manage to read. That is the unglamorous reason the CV format advice for the Dutch market keeps insisting on single-column, plain-text-extractable layouts. It is not aesthetic preference, it is machine readability.

Then the weighting differs. A linkedin job match percentage leans heavily on your profile's structured data (skills you have listed, your stated title, your connections at the company) rather than on an uploaded document, which is why your LinkedIn number can be high while a board that reads your actual CV gives you 60%. Niche technical boards weight the skills list hardest. Aggregators that scrape adverts from other sites often score against a truncated version of the advert text, so they are matching against half the information.

And the advert itself changes. The same role posted by an agency and by the employer directly contains different wording, so one CV scores differently against two postings of a single job.

A cv match score explained honestly is this: a similarity measurement between two documents, mediated by a parser, weighted by a commercial product decision you cannot see. Treat it as a signal about your CV's readability, not as a verdict on your candidacy.

Why the same CV scores differently on every board

For a highly skilled migrant application, the employer must hold recognised sponsor status with the IND and must be willing to use it for this specific role, at a salary that clears the applicable criteria for your age bracket. Three conditions. A match score checks none of them.

This is why high-scoring applications disappear. A 90% match at a 40-person Dutch agency that has never filed an IND application is a worse prospect than a 65% match at a registered sponsor with a visible track record of international hires. The 65% role has a path. The 90% role has a wall behind it, and no interface will draw that wall for you.

Register presence alone is not sufficient either. A company can sit on the IND public register and have no intention of sponsoring this particular opening, which is the distinction the piece on registered versus actually hiring exists to make. But register presence is necessary, it is checkable in a few minutes, and it filters out most of what you are currently wasting applications on. Working from the register towards vacancies rather than the reverse is the whole premise of the sponsor list to applications workflow.

Low scores are not all the same, and the difference is worth ten minutes of diagnosis before you dismiss or chase one.

A low score is probably wrong when the vocabulary differs but the work does not. You did forecasting in Excel and SQL; the advert says "financial planning and analysis". You ran platform reliability; the advert says SRE. Your internal title was "Consultant" and the market title is "Business Analyst". These are translation problems, fixable in your CV rather than by applying harder. The same mismatch is what keeps profiles invisible in recruiter search, covered in the piece on the skills recruiters actually search for.

A low score is also probably wrong when it is a parsing failure. If every platform gives you an unexpectedly low number on roles you are clearly qualified for, open your CV in a plain text viewer. If your skills and dates do not survive the conversion, no amount of applying fixes it.

A low score is telling you the truth when the gap is structural: the role requires a licence you do not hold, a language level you do not have, or a seniority band five years above you. Also when the advert requires Dutch at professional level and you have none, which boards weight inconsistently and employers do not.

The useful move is to stop treating the score as a ranking and start using it as one of two filters applied in a specific order.

- Filter one, binary: can this employer file for you? Check the entity name against the IND public register for the labour category. If no, the score is irrelevant. Discard.

- Filter two, graded: among employers that pass filter one, does the role plausibly fit? Here the score is genuinely informative, because you are comparing within a set where sponsorship is at least possible.

Run them in that order and your shortlist shrinks hard, which feels like bad news for about a day. Then the response rate changes, because every application you send goes to an employer that could legally say yes. Several of the reasons applications get no response collapse into this one sequencing error.

The order matters more than the precision of either filter. A rough fit assessment at a confirmed sponsor beats a precise one at a company that cannot hire you.

Per-listing scores have a structural blind spot beyond sponsorship: you only ever see a score for a job you already found. The board shows you what its algorithm surfaced, which skews towards recent postings, popular employers and roles matching your previous searches. Vacancies at sponsors you have never heard of, in Eindhoven or Enschede rather than Amsterdam, with titles you would not have searched, never generate a score because you never see them.

Scoring one CV against the full set of active vacancies at once inverts that. You get a ranked list rather than a reaction to whatever appeared in a feed, and the ranking can treat sponsor status as a primary dimension instead of ignoring it.

Whatever tooling you use, the output you want is small. Three to five roles per week, each one at an employer you have verified on the register, each application tailored enough that the recruiter can see quickly why you fit. That beats forty applications sorted by a percentage that never checked whether the company could hire you.

Practically: take your ranked shortlist, cut everything that fails the register check, cut anything where the salary band obviously sits below the criteria for your age bracket, then rewrite the top three CVs against the actual advert wording. An hour per application, not five minutes.

It means your CV and the advert share a lot of language. It says nothing about your odds. For a candidate needing sponsorship, a 90% match at a non-sponsor is worth less than a 60% match at a recognised sponsor, because only one of the two can result in a permit.

Change it so it parses correctly and uses the market's actual terminology for your work. Do not stuff it with advert keywords you cannot discuss in an interview. Recruiters notice, and a CV engineered for one advert's scoring usually reads worse to the human who opens it next.

LinkedIn scores your profile data rather than an uploaded document, and weights its own structured signals (listed skills, title, company connections) differently from boards that parse a PDF. Different inputs, different weights, same CV.

No. The score is a feature of the board's interface for you. The employer sees your application inside their own applicant tracking system, ranked by their own criteria, which include work authorisation.

No. Sponsor status is not an input to any matching algorithm. The only reliable check is the IND public register of recognised sponsors, matched against the legal entity name on the vacancy or the contract.

If sorting through that per-board noise yourself is taking more time than the applications, a Match Sweep runs your CV against all active jobs rather than just the newest ones, for €14.99 one-off. You can do the same work manually from the register and the current vacancies; the sweep mainly saves you the weeks of reading percentages that were never measuring sponsorship.