Platform administration

Assessment governance

Prompts, rubrics and assessment versions are tested against benchmark interviews before release.

Versions

VersionScopeStatusBenchmarkOwner
v2.5Assessment engineIn benchmark22 / 40 reference interviewsProduct governance
v2.4Assessment engineReleased40 / 40 passedProduct governance
v1.8Healthcare operations rubricReleasedApproved 02 JulDomain review
v1.2Engineering rubricDraftNot startedDomain review

Assessment policies

Fixed rules that apply to every interview on the platform.

PolicySettingWhy
Answer formatSpoken onlyTyped and pasted answers are disabled platform-wide so results stay comparable.
Verification cap2 attempts per subjectOne question and at most one follow-up. The subject is then recorded as an anomaly, never pursued further.
Client questionsAsked verbatimClient-supplied must-ask questions are never reworded, and are assessed against the requirement they target.
Candidate recordsTenant-scopedRecords, interview evidence and pool matches stay inside the tenant that created them. There is no cross-tenant candidate index.
RetentionSet per tenantDefault 24 months from last activity. Records are deleted with their evidence when the window closes or the candidate asks.
Session monitoringConsent requiredTab, focus, keyboard and paste activity is recorded for human review only. It cannot trigger an automatic outcome.

Audit trail

Assessment versions, questions, evidence and human overrides are auditable.

TimeActorActionDetail
30 Jul 10:26N. Morley (Morley Talent)Dismiss anomalyMarcus Oyelaran · delegation claim · reason recorded
30 Jul 09:14N. Morley (Morley Talent)Override recommendationMarcus Oyelaran · reason recorded
29 Jul 16:02Product governanceRelease assessment v2.4Benchmark pass 40/40
29 Jul 11:47S. Ellery (Morley Talent)Exclude question from overall viewQ5 · audio quality
28 Jul 08:20PlatformRetention job12 assessments deleted at end of retention period
A model or prompt change cannot be released to customers until it has been re-tested against the benchmark interview set.