How We Score — And Why You Can Trust It

Every TalentProof score is calculated using transparent, auditable rules — no black boxes. Here's exactly how we measure skill, integrity, and fit.

Our Scoring Principles

Consistent & Fair

Every candidate is evaluated by the same AI rubric — no human bias, no mood swings, no inconsistency between interviewers.

Transparent

Scores aren't hidden behind proprietary algorithms. We publish our formulas, thresholds, and deduction rules right here on this page.

Cheat-Resistant

Multi-layered anti-fraud measures — from AI text detection to live proctoring — make it extremely hard to game a TalentProof assessment.

The Three Core Metrics

Every assessment produces three standardized scores on a 0–100 scale.

1. Performance Score (0–100)

Measures how well the candidate actually performed on the assessment. This is the core skill measure.

Standard Q&A

AI evaluates correctness, depth, and relevance of open-ended answers. Raw score 0–1000, normalized to 0–100.

Rapid-Fire MCQ

Percentage of correct answers after negative marking (−0.25 per wrong answer). Includes a hard-question gate requiring 40% of hard questions correct.

Hands-On Coding

Weighted combination: 50% functional correctness, 30% code quality, 20% task completion.

90–100
Expert
80–89
Advanced
70–79
Proficient
60–69
Developing
<60
Needs Work

2. Confidence Score (0–100)

Also called the Integrity Score — this tells you how confident we are that the results are authentic. It starts at 100 and deductions are applied for detected integrity issues.

What Causes Deductions:

Paste abuse detected

−20

Camera permission denied

−15

Screen share denied

−10

Multiple faces detected

−20 per event (max −40)

Face missing from camera

−3 per event (max −40)

Tab switching / alt-tab

−2 per 3 blurs (max −12)

AI hints are allowed — candidates can request hints during assessments. This is logged for context but does not reduce the Confidence Score. We test problem-solving ability, not memorization.

3. Job Fit Score (0–100)

Measures how closely the candidate's skills and experience align with the specific job requirements. This score is generated when the assessment is tied to a job description.

  • Keyword matching: Required skills from the JD matched against demonstrated abilities
  • Experience alignment: Years and depth of relevant experience vs. role requirements
  • Assessment performance: How they scored on the role-specific questions
85–100
Excellent Fit
70–84
Strong Fit
50–69
Partial Fit
<50
Limited Fit

The Overall Score Formula

The Overall Score is a weighted combination that gives recruiters a single number for quick comparison.

With Job Match

Overall = (Performance × 0.6) + (Confidence × 0.2) + (Job Fit × 0.2)

Without Job Match

Overall = (Performance × 0.7) + (Confidence × 0.3)

Performance always carries the most weight because skill is what matters most. Confidence (integrity) is always factored in so cheaters are penalized even if they score well on questions.

Why Scores Are Trustworthy

Multiple layers of integrity verification ensure authentic results.

Live Proctoring

Optional photo capture during assessments verifies who is actually taking the test. Face detection flags when:

  • No face is detected (candidate stepped away)
  • Multiple faces are detected (someone helping)
  • Camera permissions are denied

AI Text Detection

Our AI analyzes response patterns to detect content generated by ChatGPT and other AI tools. Proportional deductions are applied based on detection confidence.

Copy & Paste Monitoring

Clipboard activity is tracked. Excessive pasting from external sources triggers a significant integrity deduction, ensuring candidates answer from their own knowledge.

Tab Switch Detection

When candidates switch away from the assessment tab, each blur event is logged. Frequent tab switching suggests external resource usage and is penalized progressively.

Proof Score — Candidate Credibility (0–100)

Beyond individual assessment scores, each candidate builds a cumulative Proof Score that reflects their overall credibility on the platform. This is calculated across six pillars:

Identity
max 10

Profile completeness — photo, name, location, contact email

Foundation
max 10

Completed resume analysis with AI skill extraction

Self-Awareness
max 5

Completed psychometric / personality assessment

Verified Skills
max 50

Passed skill assessments — proctored skills earn bonus points

Engagement
max 20

Public scorecard, job matches, practice interviews, copilot usage

Excellence
max 5

Average confidence score across assessments ≥ 80%

Proof Levels:

Top Talent: 86–100
Verified: 61–85
Intermediate: 36–60
Basic: 16–35
Starter: 0–15

How AI Evaluation Works

Our AI evaluator (powered by Google Gemini) uses structured, schema-validated prompts to ensure consistent scoring. Here's what makes it reliable:

  • Schema-first prompts: Every AI flow uses Zod schemas to define exactly what outputs are expected — the AI can't return freeform, unstructured opinions.
  • Rubric-based evaluation: Questions are graded against specific criteria (accuracy, depth, relevance) — not subjective impressions.
  • Consistent across candidates: Unlike human interviewers who vary by mood, fatigue, or bias, the AI applies the same rubric every single time.
  • No demographic bias: PII (names, emails, locations) is encrypted before AI processing. The AI evaluates skills, not identities.
  • Job-specific questions: Assessments are dynamically generated from actual job descriptions, making them role-relevant rather than generic.

Pass Thresholds by Assessment Type

Standard Q&A

70%

(700 / 1000 raw score)

Rapid-Fire MCQ

70%

+ 40% hard question gate

Hands-On Coding

60%

(lower threshold due to difficulty)

Privacy by Design

All personally identifiable information is encrypted with AES-256 before storage. The AI never sees real names — it works with anonymized placeholders. Read more on our Data Security page.

See It in Action

Explore a sample scorecard to see how all three metrics come together, or start verifying candidates today.

View Sample ScorecardHow It Works