Inside Our Fit-Score Engine: How We Score Every Application
We get asked a version of the same question by almost every client we introduce our technology to: "how does the score actually work?" Fair question — a match score is only useful if you trust what's behind it. Here's the honest breakdown.
The Four Pillars of a Fit Score
Every application is scored out of 100 across four weighted components. Sector Expertise (25 points) measures how strongly a resume's language matches the vocabulary of the role's sector — shipping, offshore, energy, logistics, trading, or professional services each have their own keyword universe. Role Alignment (20 points) checks how closely the candidate's stated title and experience mirror the job title itself.
Requirements Coverage (35 points) — the heaviest-weighted pillar — breaks the job's requirements and description into individual clauses and checks how many are substantively addressed in the resume, rather than just keyword-matched. Seniority & Experience (20 points) reads stated years of experience and seniority language (senior, lead, director, manager, associate) and compares it against what the role actually calls for, with a bonus or penalty depending on the direction of the mismatch. A small education bonus rounds out the total.
Why We Built It In-House, Not as a Black Box
The engine runs entirely as local logic — no resume text is sent to a third-party scoring API to produce the number. That matters for two reasons. First, confidentiality: candidate and client data for a live executive search should not be leaving our systems any more than is strictly necessary. Second, and just as important to us, explainability — because the scoring logic is ours, we can show a consultant (or a client, if they ask) exactly which requirements a candidate matched and which ones they didn't, rather than pointing at an opaque model and shrugging.
"If a consultant can't explain why a candidate scored the way they did, the score isn't doing its job. Every number our engine produces comes with a reason attached." — Jason Tay, Managing Partner
The Limits We Designed For
No scoring model reads between the lines the way a human does. A candidate who undersells a genuinely relevant achievement in plain language, or whose most relevant experience came from a role with an unconventional title, can score lower than they should on paper. That's precisely why the output is a ranked shortlist for a consultant to review, never an auto-reject list — nothing is filtered out of consideration purely on score.
We treat the model the same way we'd treat a very fast, very literal junior researcher: extremely useful for narrowing 400 applications down to the 40 worth a closer look, not a substitute for the judgement that turns 40 into a shortlist of four.