JD-Resume Fit Scorer

What it does

The JD-Resume Fit Scorer is not a keyword-matching ATS; it is an analysis-and-scoring engine grounded in the validated predictive-validity literature of personnel selection. It applies five frameworks in sequence:

1. KSAO job analysis — decomposes the JD into Knowledge (K), Skills (S), Abilities (A), and Other (O: license, shift, certification) components; re-resolves generic phrases like "5 years experience" by asking "which KSAO does this prove?"
2. Sackett, Zhang, Berry & Lievens (2022) hierarchy — selection methods are not equally valid: structured interview r≈0.42 > work sample r≈0.33 > GMA cognitive test r≈0.31 > unstructured interview r≈0.19 > job tenure r≈0.06. Score weight therefore shifts to demonstrated behavior, not years on a resume; tenure is capped at 10%.
3. BARS-anchored scoring — each criterion's resume evidence is scored 0–4: 4 = quantified result (e.g. "AHT 6:10→5:05"), 3 = role evidence without metric, 2 = transferable adjacent skill, 1 = keyword only, 0 = no evidence. Cluster score = (Σpts / (criteria×4))×100; Fit Score = Σ(cluster×weight)/100.
4. STAR + gap-targeted interview — generates behavioral questions probing each candidate's 3-4 lowest-scoring criteria; never produces vague "tell me about yourself" prompts.
5. Adverse-impact / job-relatedness guard — EEOC four-fifths rule logic + Turkish KVKK Art.4/6/22, Labor Law Art.5, Law 6701. Job-irrelevant signals (age, gender, name association, school prestige, employment gap, special-category data) are removed if they shift the score, and the raw→corrected delta is reported.

When to use it

You have a JD + 1–N resumes and want "who fits, why, and what to ask." It produces a fit profile for a single candidate and a defensible, legally traceable ranked shortlist for many. Suited to bulk screening, shortlist justification, interview prep, and discrimination-risk audit. Not suited if you want a yes/no automaton — the output is a rationale, not a decision.

Method / frameworks

Sackett, Zhang, Berry & Lievens (2022, Journal of Applied Psychology) — revision of Schmidt & Hunter (1998), correcting range-restriction overcorrection · EEOC Uniform Guidelines (1978) four-fifths rule & job-relatedness/business-necessity · SIOP/SHRM Principles — structured assessment & BARS standard · KVKK 6698 (Art.4 proportionality, Art.6 special-category data, Art.22 right to contest automated decisions) · Turkish Labor Law Art.5 + Law 6701 (anti-discrimination).