Evidence first
Scores are anchored to answer-specific evidence and turn IDs, not impressionistic prose.
Back to HireEGG
Scoring methodology · v2.0
A score an HR team can question—and we can answer.
HireEGG converts observable interview evidence into role-weighted competency estimates. The number is reproducible, the uncertainty is visible, and every conclusion can be traced back to the candidate's answers.
The short answer for enterprise teams
AI finds evidence. Versioned code calculates the score. People make the hiring decision.
Reproducible
The same evidence, weights, and methodology version produce the same score.
Uncertainty shown
Thin evidence widens intervals and can withhold a rating instead of creating false precision.
From answer to score
Five reviewable steps.
No hidden end-to-end model decides employability. Each layer has a defined job and a reviewable output.
Capture observable evidence
Each answer is stored as a turn-level evidence item: the question, intended competency, a short evidence quote, structured strong and weak signals, and evidence strength.
Map evidence to competencies
Evidence is mapped to technical depth, communication, problem solving, and domain knowledge. One answer can contribute to more than one competency, but its relevance is explicit.
Weight for evidence quality
Specific, relevant answer evidence receives more weight. Thin or missing evidence receives less weight; it does not become a confident low or high score.
Calculate in deterministic code
Versioned code applies configured competency weights and Bayesian shrinkage. The language model cannot set, edit, or override the score.
Publish with uncertainty
Every score is paired with evidence coverage, reliability, cited turn IDs, and a 90% interval. If coverage is too low, HireEGG withholds the final readiness rating.
The mathematical core
Confidence before precision.
For each competency, HireEGG combines evidence strength and relevance into an effective evidence weight. The observed estimate is then shrunk toward a neutral prior of 50 according to reliability.
competency = reliability × observed evidence
+ (1 − reliability) × 50
overall = Σ(normalized weight × competency)
This structure prevents a sparse interview from appearing more certain than the evidence supports. The report also publishes a 90% interval around the estimate.
Default role profile
Competency weights
Technical depth
35%
Role-relevant technical judgment and implementation detail
Communication
25%
Clarity, structure, precision, and stakeholder awareness
Problem solving
20%
Trade-offs, failure analysis, validation, and decision quality
Domain knowledge
20%
Applied understanding of the role's working context
Enterprise role profiles can change these weights. HireEGG normalizes them to 100% and records the applied values in every report.
What the report exposes
The number never travels alone.
0–100 estimate
A role-weighted performance estimate, or withheld when evidence is insufficient.
Cited evidence
Turn IDs, evidence counts, and answer excerpts behind each competency.
90% interval
A visible range communicating uncertainty rather than a falsely exact number.
Coverage and reliability
Separate measures showing how much relevant evidence was captured and how stable it is.
Integrity is separate
Telemetry is evidence—not a verdict.
Browser and companion-app engines produce their own event timeline, graph briefs, sample counts, flagged rates, data-quality labels, limitations, and cross-device correlations.
No score contamination
Integrity risk never changes a performance or competency score.
No cheating declaration
A sensor event can warrant review; it cannot establish intent by itself.
Corroboration visible
Single-device events are distinguished from time-correlated browser and phone evidence.
Human review
Elevated risk routes the session to evidence review rather than automatic rejection.
Enterprise questions
Answers procurement and HR can audit.
Who decides the score—the AI or your code?
Deterministic, versioned code decides every number. AI is limited to extracting and summarizing evidence from the interview. It cannot alter competency weights, thresholds, reliability, intervals, or the final score.
What does 72/100 actually mean?
It is the weighted estimate of demonstrated competency evidence for the configured role and level—not a personality score and not a guarantee of future performance. HR also sees its interval, coverage, reliability, and the exact cited turns behind it.
Why do scores move toward 50 when evidence is weak?
Uncertainty should not masquerade as confidence. Bayesian shrinkage pulls thin-evidence estimates toward a neutral prior, while stronger independent evidence allows the estimate to move farther away from neutral.
Can an anti-cheat signal lower the performance score?
No. Performance and integrity are separate tracks. Integrity telemetry can route a session to human review, but it cannot change an answer score, competency score, or overall performance score.
Do you show candidate percentiles?
Not yet. HireEGG will not publish a percentile until it has a sufficiently large, role-and-level-specific cohort linked to validated outcomes. A fabricated or mixed-cohort percentile would be misleading.
Can the report reject a candidate automatically?
No. HireEGG provides decision support. Low evidence, low performance, or elevated integrity signals require human review of the cited evidence before an employment decision.
Validation status
Transparent about what is—and is not—validated.
This is HireEGG's internal evidence-based methodology. HireEGG does not currently claim independent predictive-validity certification, IO-psychology certification, or validated cross-employer percentiles. The validation program includes frozen rubrics, double-scored calibration sets, inter-rater agreement, role-specific outcome studies, and subgroup monitoring where lawful and ethical.