How a match is decided

Five separate ratings, combined by arithmetic, so you can see what pulled a score down.

A candidate and a job are never compared as one vague impression. Five separate questions are asked, each answered on its own scale from 0 to 10, and only then are the five combined into the single number you see.

  • Skills: does the person actually have what the job asks for?
  • Seniority: is this a step up, a step sideways, or a step down?
  • Domain: has the person worked in this industry or on this kind of problem?
  • Languages: does the person speak what the job requires, not what the ad happens to be written in?
  • Location: is the job somewhere this candidate would actually go?

The final number is arithmetic, not opinion

The five ratings are weighted and added up in plain code, with skills counting most and location least. Nobody is ever asked for an overall impression, which is exactly why the result can be argued with: you can see which of the five dragged a score down, and decide whether you agree.

Three bands

85 and above is a shortlist: worth a call today. 65 to 84 is a match: worth reading. Below 65 is kept but folded away, because a rejected match is still evidence about what the market thinks of this candidate. Matches found on the last run are flagged as new, so your morning list is the part that changed.

Where the system refuses to be optimistic

A few things make a call pointless no matter how good the rest looks, and those are capped in code rather than left to judgement: a job that requires a language the candidate does not speak, or one sitting far outside every place they said they would work. Everything else, including location within reason, moves the score rather than closing the door.

Why would a score change from one day to the next?

Because the profile changed, and for no other reason. A score that has been given is kept: the nightly run judges pairs that were never judged, it does not quietly rewrite yesterday’s numbers. But editing something that matters to matching — roles, skills, languages, seniority, the places the candidate will work — and confirming it re-judges that candidate against the pool on purpose, so the whole list stays consistent with the profile you just corrected.

Last updated: