Skip to content
Sign in

Candidate Match Score

See an example

Score a CV against a job description with the Expertini Candidate Match Score (CMS): personal identifiers are stripped, AI weighs every requirement in the advert, and a published formula turns that into a 0–100 score you can check line by line.

Try it now — to save your work to your account (and see it in the Expertini app), create a free account or sign in.

CV and job

Names, email addresses, phone numbers, addresses and similar identifiers are removed from the CV text before the AI reads it. The file itself is not kept.

Each score uses one AI request from your plan and is saved to your history — the score and requirements only, never the CV.

Match score

How the score works

CMS = Σ(CSS × JRIS) ÷ Σ JRIS

JRIS · importanceHow much the employer needs each requirement (0–100), read from the advert's wording.
CSS · evidenceHow well the CV shows that requirement (0–100). A missing must-have scores 0.

The AI fills in the two numbers for 5–9 requirements; the formula — not the AI — produces the score.

Your scores 0

Keep every scoreSign in and each score is saved to your account and the Expertini app — never the CV.

What is the Expertini Candidate Match Score?

The Candidate Match Score (CMS) measures how well the professional evidence in a CV meets the requirements of one specific job description, on a scale of 0 to 100. Job seekers in Kagua Erave use it to see how a screener might read their CV before they apply; recruiters use it as a first, explainable filter. It is one measure among many, not a hiring decision.

The two numbers behind it

  • JRIS — Job Requirement Importance Score (0–100). How critical each requirement is to the employer, read from the advert's own language: “must have” or “essential” ≈ 90–100, “preferred” ≈ 50–70, “a plus” ≈ 30–50.
  • CSS — Candidate Skill Score (0–100). How well the CV evidences that requirement: 0 for no evidence, 100 for strong, documented evidence. A mandatory requirement with no evidence at all is a hard blocker and scores 0.

The formula

CMS = Σ(CSSi × JRISi) ÷ Σ JRISi — a weighted average of the evidence scores, weighted by importance. Because both inputs run from 0 to 100, the result does too.

Worked example. Seven requirements with (CSS × JRIS) of (100×100) + (90×100) + (85×95) + (80×90) + (50×90) + (60×60) + (0×100) = 42,375, and importances summing to 565: CMS = 42,375 ÷ 565 = 75.0. The one missing must-have (a security clearance, JRIS 100, CSS 0) adds 100 to the bottom of the fraction and nothing to the top — which is why a single hard blocker pulls a strong profile down sharply.

Score bands

85+ exceptional · 70–84 strong · 55–69 moderate · 40–54 partial · below 40 low. Always read the score with the requirement table: two CVs can reach the same number for very different reasons.

What it can and cannot tell you

It reads what the CV says against what the advert asks, including whether must-have requirements are evidenced at all. It cannot verify that qualifications, dates or achievements are genuine, whether work was individual or shared, or a candidate's availability. Removing identifiers reduces — but does not eliminate — bias: employers, schools, phrasing and career gaps remain in the text, and a language model still does the reading.

Which Expertini score is this?

The resume score, ATS score and job score compare the vocabulary two documents share (cosine similarity) — no model, so the same inputs always give the same number, but they cannot tell that two phrasings mean the same thing. CMS is the separate system that uses a language model to read meaning, then a fixed formula to score it.

Is the CV stored?

No. The file is read on the server, identifiers are removed from its text, the anonymised text is sent to the AI, and nothing from the CV is kept. Your history keeps the score, the requirement table, the job title and the first 300 characters of the job description.

Why are the AI parts not exactly repeatable?

The importance and evidence numbers come from a language model, which can vary slightly between runs. The formula applied to those numbers is fixed, and every number it uses is shown to you, so you can check the result yourself.

Can adding keywords raise the score?

Less than with keyword-matching tools: the AI is asked to look for evidence — achievements, responsibilities, results — rather than words. No screening method is immune to a determined rewrite, so never add claims you cannot back up in an interview.

What does it cost?

Each score uses one AI request from your plan's monthly allowance, which resets on the 1st. Reading saved scores, printing them and the explanation here are free.

Where is the method published?

Syed, A. H., Habeebi, S. A., & Habibi, S. M. M. (2026). From Stochastic to Deterministic: A Multi-Criteria Decision Analysis Framework for Bounded Semantic Parsing in AI-Driven Recruitment Screening. Expertini Research. This is Expertini's own research, not an independent evaluation — the formula above is the one it describes, so you can check the method against your own inputs.