Technical & Operational Glossary
Platform terms & methodology,
defined with precision.
Standardized operational definitions for all statistical formulas, classification pipelines, data governance controls, and integrity metrics across the Yves platform.
- ActionabilityScoring
- Whether designated personnel can initiate an operational remediation today. Scored independently of sentiment, because a low-urgency submission can express acute anger while a severe compliance issue can be described in calm terms.
- Aggregate suppressionPrivacy
- Automated cell suppression withholding any aggregate reporting metric derived from fewer than a minimum threshold of submissions (default n=5) to mathematically prevent individual re-identification.
- Anonymity by architecturePrivacy
- Structural privacy enforced by declining to collect identifying telemetry at the network edge: zero user accounts, zero device identifiers, zero third-party tracking scripts, and self-hosted fonts.
- Attention auditGovernance
- Automated logging and auditing of operator interaction rates across triage categories, tracking team focus against independent risk severity.
- BaselineAnalytics
- Historical moving averages computed specifically for an individual operating site or category, providing localized comparative context rather than distorting trends through global averages.
- Brier scoreGovernance
- A strictly proper scoring rule measuring the accuracy of probabilistic classifications against empirical outcomes, establishing calibration benchmarks for triage models.
- BurstIntegrity
- An anomalous surge in submissions originating from matching network entropy signatures or rapid time distributions, triggering automated forensic review.
- Case briefInvestigate
- A synthesized investigation report reasoned strictly from pinned evidence records, detailing verified operational findings, confidence metrics, and evidentiary boundaries.
- CellAnalytics
- A single cross-tabulated coordinate in a reporting breakdown (e.g., category by site by week). Low-volume cells are explicitly flagged for statistical sparsity.
- CoachPortal
- On-device client guidance assisting respondents during draft composition: prompting for operational context (site, date) while intercepting accidental PII before transmission.
- Coordination graphIntegrity
- Network graph mapping anomalous submission clusters linked by near-duplicate semantic content and synchronized timing windows to identify coordinated campaigns.
- CorpusCore
- The immutable, encrypted repository of all feedback records held by a tenant, including verbatim text, normalized translations, and audit lineage. Fully exportable via API.
- Decision queueCore
- The real-time operational triage interface prioritizing unaddressed submissions by composite urgency, actionability, and risk indicators with attached source evidence.
- DriftAnalytics
- Temporal shift in semantic vector representations within an established taxonomy category, identifying emerging operational failure modes before schema reclassification.
- EmbeddingCore
- High-dimensional vector representations capturing semantic meaning across unstructured text, powering cross-lingual search, deduplication, and unsupervised clustering.
- EnrichmentCore
- The automated processing pipeline executed at ingestion: language identification, translation, multi-label taxonomy classification, sentiment polarity, urgency scoring, actionability index, PII redaction, and vector embedding.
- Escalation floorScoring
- A mandatory baseline urgency score assigned to designated statutory categories (e.g., clinical negligence, fraud, harassment) that enforces priority triage regardless of submission tone.
- Evidence packDeliver
- An audit-ready executive dossier containing sanitized verbatim submissions, Wilson score confidence intervals, and integrity attestations for regulatory and board review.
- FingerprintIntegrity
- An ephemeral, salted cryptographic hash generated client-side to detect automated bursts and ballot stuffing without recording persistent or identifiable device identifiers.
- Integrity attestationIntegrity
- A formal cryptographic metadata record attached to published reporting outputs certifying data-layer exclusions, filtration signals, and comparative raw totals.
- Integrity verdictIntegrity
- Automated classification assigning submission status as Verified, Suspect, or Quarantined prior to reporting dataset ingestion. Always reversible by authorized administrators.
- Language equityGovernance
- Continuous statistical auditing of triage resolution latency and action rates disaggregated by submission language, measuring institutional responsiveness across non-primary languages.
- LineageGovernance
- The complete, immutable audit trail tracing every metric from edge intake through sanitization, enrichment pipeline version, filtration gates, and formula computation.
- Meaning searchUnderstand
- Semantic vector search querying conceptual intent and situational themes across 40+ languages simultaneously without requiring exact keyword syntax.
- Metric registryGovernance
- The centralized platform directory defining every analytical calculation, underlying population sample, suppression rule, and dependent reporting view.
- Near-duplicate matchingIntegrity
- Semantic vector distance comparison identifying coordinated astroturfing and duplicate content across variant phrasings, calibrated to organizational risk tolerance.
- OntologyInvestigate
- The structured domain graph modeling relationships across operational entities (sites, incidents, categories, remediation milestones) with bounded traversal queries.
- ProposalAgents
- A structured triage recommendation produced by automated pipeline models: specifying analytical objectives, queried data schemas, confidence scores, and source evidence for mandatory operator approval.
- Purpose-based accessPrivacy
- Access control protocol requiring authenticated operators to declare a specific operational rationale before viewing sensitive records. Automatically logs access and expires after 8 hours.
- QuarantineIntegrity
- Forensic isolation of flagged submissions: segregated from aggregate analytics while preserved immutably for administrator review and campaign verification.
- Recurrence radarDeliver
- Continuous vector similarity tracking verifying whether completed remediations successfully resolved the underlying issue across subsequent incoming submissions.
- Redaction policyPrivacy
- Role-based masking rules applied dynamically at render time to scrub sensitive identifiers and special-category data prior to viewing or export.
- SentimentScoring
- A normalized continuous score measuring emotional polarity on a calibrated scale, avoiding distortive ternary bucketing (positive/neutral/negative).
- Silent failureAnalytics
- An operational discrepancy where an administrative action is marked closed on paper while incoming complaint volume and negative sentiment persist in the feedback stream.
- Site dimensionAnalytics
- The physical or organizational touchpoint where feedback originated (facility, department, terminal, branch), captured via localized intake touchpoints.
- SuspectIntegrity
- A submission flagged by secondary integrity rules. Excluded from primary executive metrics but inspectable within trust-weighted comparison views.
- Tail bucketCore
- The residual unclassified category in an institutional taxonomy. Monitored continuously by semantic clustering to surface emerging risks for formal schema addition.
- TaxonomyCore
- The formal classification schema establishing operational category boundaries, escalation floors, and routing paths approved by institutional governance.
- Trace IDSecurity
- A unique cryptographic correlation identifier tracking the complete chronological sequence of automated inferences, database queries, and operator actions for a single event.
- Trust weightingIntegrity
- Statistical weighting calibrating aggregate metrics by integrity confidence scores, displayed alongside unweighted figures with stated deltas.
- Under-attendedAnalytics
- High-severity, low-frequency risk signals masked by gross volume rankings, surfaced through independent urgency weighting and Wilson confidence intervals.
- Wilson intervalAnalytics
- The asymmetric binomial proportion confidence interval calculated for all proportion metrics, ensuring statistical defensibility at small sample sizes and extreme distributions.
Next step
Audit-ready definitions in every report.
Every metric and term documented here maps directly to the live Metric Registry inside the console, complete with underlying formulas, population parameters, and confidence intervals.