What is a JD Score?
A JD Score is a skill assessment generated from a specific job description, producing a single score from 0 to 10. It runs in four layers — rapid breadth questions, applied understanding, hands-on execution in real code and SQL, and senior-level scenario reasoning — weighted 20%, 20%, 30% and 30%. It is CareerID's alternative to screening a resume.
- The assessment is generated from the individual job description, so two roles requiring different skills produce different tests.
- Four layers: L0 timed breadth, L1 applied understanding, L2 hands-on execution, L3 scenario reasoning.
- Weighting is L0 20%, L1 20%, L2 30%, L3 30%, on a 0–10 scale.
- A layer with no questions is excluded and remaining weights renormalise, so an omitted layer is not penalised.
- The employer's cutoff is a soft ranking threshold: above it auto-advances a candidate to Shortlisted, below it never auto-rejects.
- Tests are proctored and generated in multiple parallel question sets, so answers cannot be shared between candidates.
Recruitment Efficiency & Filtration Metrics
Recruiter Screening Time
Hours / HireAverage hours engineering and talent acquisition teams spend sorting unvetted resumes, parsing keywords, and conducting initial phone screens.
Automated Candidate Dropout
% DisqualifiedPercentage of candidate profiles discarded by automated keyword parsers, tier-1 college filters, or arbitrary CGPA gates before a human evaluates work.
Filtered by CGPA < 7.5, non-IIT/NIT tags, or missing resume keywords
Filtered by 10th/12th marksheets and tier accreditation
Scoring below cutoff never rejects an application; all profiles stay reviewable
CareerID 4-Layer Assessment Architecture & Score Weights
Proctored 0–10 score generated dynamically from the employer's specific job description.
Timed rapid-fire coverage across every primary skill in the job description.
Nuance and edge-case trap questions that separate true understanding from memorization.
Actual code execution (Monaco) and in-browser SQL queries (PGlite) graded on outputs.
Ambiguous production trade-off scenarios scored the way senior engineering leads evaluate.
Why generate the test from the job description
A generic aptitude test measures something the job did not ask for. A test generated from the posting measures the skills that posting actually requires, which is what makes the resulting number comparable between applicants for that role and meaningless to copy between roles.
It also removes a decision from the employer. Rather than configuring a test matrix, the employer sets one value — the cutoff — and the layer counts, question mix, hands-on format per skill and proctoring profile are derived from the required skills.
The four layers
L0 is rapid timed multiple-choice across the role's skills, measuring breadth and speed. It is deliberately shallow: its job is coverage, not depth.
L1 asks per-skill micro questions, trap questions and short scenarios. The trap questions matter most — they are written so that a memorised answer and an understood one diverge.
L2 is hands-on. Depending on the skill this is code executed against real test cases, a SQL query run against a seeded database, a whiteboard exercise, or structured multiple-choice where hands-on assessment genuinely does not apply. It carries the joint-highest weight because it is the layer hardest to fake.
L3 presents real-world scenarios requiring judgement rather than recall, clustered by the skills involved. It also carries 30%.
What the cutoff does, and what it does not do
The employer sets a cutoff on the 0–10 scale, and its effect is deliberately asymmetric. A candidate at or above it advances to Shortlisted automatically, with a timeline entry and a notification to both sides.
A candidate below it is not rejected. Nothing happens automatically, and the application remains in the pipeline for the employer to review. This is the point of the design: automatic rejection at a numeric threshold would rebuild exactly the blind gate that CGPA cutoffs and resume-keyword filters create.
Integrity
Each test is generated with multiple parallel question sets, so two candidates sitting the same assessment do not receive an identical paper, and question selection serves least-recently-used items first so a candidate applying to two similar roles does not see the same questions twice.
Proctoring covers camera, microphone and screen capture, and logs tab-switch, paste and fullscreen-exit events. Those flags are shown to the employer next to the score rather than used to disqualify silently.
Common Questions
How is a JD Score calculated?
Four layers are scored and combined: L0 rapid breadth at 20%, L1 applied understanding at 20%, L2 hands-on execution at 30%, and L3 scenario reasoning at 30%, producing a single value from 0 to 10. If a layer has no questions it is excluded and the remaining weights are renormalised, so a test that omits a layer is not penalised for it.
Does a low JD Score reject my application?
No. The cutoff is a ranking threshold, not a filter. Scoring at or above it advances you to Shortlisted automatically. Scoring below it does nothing automatically — your application stays in the pipeline for the employer to review manually. There is no score at which CareerID rejects an application on its own.
Is the assessment the same for every job?
No. It is generated from each job description, so the skills tested, the hands-on format and the question mix differ by role. This is also why a score is meaningful for the role it was taken for rather than transferable as a general rating.
Can candidates share answers?
Each test is generated in multiple parallel question sets, so candidates sitting the same assessment receive different papers, and selection prefers least-recently-served questions. Sessions are also proctored, with integrity events logged and surfaced to the employer alongside the score.