Why match percentages on job sites are meaningless

You paste your CV into a job site and it lights up: 87% match. You feel a small lift. You spend your evening on a cover letter. Three weeks later there is no reply, and you never find out which of the two of you was wrong.

Here is the thing worth knowing about that number: it was never trying to be accurate. It was trying to make you apply.

Who profits from your hope

Ask what the tool showing you the percentage gets paid for. Job boards earn from applications and from employers who want a full funnel. Recruiters earn per placement, so more candidates means better odds for them, not for you. Application tools measure success in applications sent.

Now ask what a truly honest number would do to those businesses. If a site told you the truth on most postings, which is you are one of four hundred and you are missing the one thing they will not bend on, you would apply to fewer jobs. That is the number nobody has an incentive to build.

So the percentage drifts into a comfortable band. It is rarely 30%, which would make you close the tab. It is rarely 98%, which would be a promise. It sits around 80, just high enough to justify one more evening.

What the number actually measures

Most match scores are keyword overlap with some weighting. Your CV says React, the posting says React, that is a point. It looks like understanding. It is counting.

Counting cannot see the things that decide the outcome:

• Recency. React five years ago and Angular for the last three is a very different candidate from React last month. A keyword list sees both as "React".
• Hard versus soft. "Must have a security clearance" and "affinity with our mission" both count as one requirement. Only one of them can end your application before a human reads it.
• Depth. Listing Kubernetes because you once fixed a manifest reads the same as running clusters for four years.
• Marketing. Half of most postings is language written to attract, not to filter. Counting treats it as requirements.

A number cannot tell you which of these applies to you, because a number has thrown away the reasoning that would.

What to read instead

When you open a posting, ignore the score and answer four questions honestly. They take two minutes and they beat any percentage.

1. What will they not bend on? Usually one or two things: a clearance, a language, a specific stack, years in a domain. If you miss one of those, the rest of the posting does not matter.
2. What is recent, not just present? Look at what you actually did in the last two years, not what your skills list claims. That is what an interviewer will dig into.
3. What is decorative? Cross out every line that is there to sell the job. What remains is the real posting, and it is usually short.
4. What is the honest gap, and can you name it? If you can say out loud what you are missing and why you would still be good, you have something to write. If you cannot, no cover letter will save it.

An honest verdict has a shape

A useful answer is not a number. It is a decision with a reason attached: apply, wait, or skip, and the one factor that decided it. Something you can argue with.

That is why Shinsa has no percentage to inflate. It says GO, HOLD, or SKIP, tells you what tipped it, and it says SKIP as easily as it says GO, because it has nothing to gain from your application. Not applying to the wrong job is not a failure. It is the evening you get back.

← All notes