Governance

Purpose

Define a pragmatic governance model that improves speed, trust, and compliance for data and analytics products built on Open Data Platform.

Scope

  • Domains: odp_staffing_demand
  • Data lifecycle: ingest, transform, serve, retain, archive, delete
  • Analytics scope: BI metrics, semantic models, dashboards, AI extraction and model outputs

Principles

  • Guardrails over gates: automate policy checks, escalate only high-risk exceptions
  • Federated ownership: domains own data products; platform team provides common standards and controls
  • Metadata by default: no production dataset without owner, steward, SLA, and classification
  • Quality is a product requirement: data quality SLOs are defined and monitored
  • Compliance is engineered: privacy and security controls are embedded in pipelines

Roles and Accountability

RoleCore Accountability
Domain Data OwnerBusiness accountability for dataset quality, definitions, and access approvals
Domain Data StewardMetadata quality, glossary alignment, retention and classification maintenance
Data Platform TeamTooling, policy automation, lineage, CI/CD governance checks, observability
Security and Privacy LeadPII policy, access model, retention/deletion policy, audit readiness
Analytics LeadCertified semantic definitions and dashboard lifecycle governance
AI LeadModel and agent release controls, human-in-the-loop design, monitoring standards

Decision Rights (RACI)

DecisionOwnerStewardPlatformSecurityCouncil
Data contract changeARCII
Sensitive data classificationCRIAI
Access policy (row/column)CRCAI
Canonical KPI definitionARCIC
AI model promotion (regulated)CCRCA

Governance Control Flow

  1. Domain opens change in Git (schema, pipeline, metric, model)
  2. CI runs contract, quality, security, metadata, and lineage checks
  3. Low-risk changes merge automatically after checks and owner approval
  4. High-risk changes trigger explicit Security/Privacy or Council review
  5. Production release publishes metadata and monitoring hooks
  6. Failures route to domain ownership with defined SLA and escalation

Required Artifacts per Data Product

  • Product spec (*_product.yaml) with owner, steward, classification, retention, SLAs
  • Metric definitions in schema/metrics.yaml where applicable
  • DBML model with notes, keys, refs, and validation in CI
  • Data quality rules and thresholds in pipeline code
  • Access policy and audit trail for sensitive datasets

Cadence

  • Weekly: domain data triage (quality incidents, SLA breaches, drift)
  • Bi-weekly: architecture and schema review across domains
  • Monthly: governance council (policy changes, risk exceptions, KPI conflicts)
  • Quarterly: access recertification, retention validation, control evidence review

Framework Alignment

  • DAMA-DMBOK: ownership, metadata, quality, security, lifecycle controls
  • TOGAF: architecture decisions and exceptions are explicit, traceable, and reviewable
  • ISO 27001: least privilege, secrets handling, logging, and information lifecycle controls
  • EU AI Act: use-case inventory, risk tiering, traceability, oversight, and monitoring