The framework presented by Jarvus Innovations (Session 2, May 2026) for operationalizing data governance through product management. The core insight: design the workflow so the proof of governance is produced as the work happens.
The 8 Stages
- Intake — Requests land in a queue. Visibility into demand, roles, and resource equity.
- Discover — Write specs for what exists. Understand the consumer set.
- Design for Reuse — Choose patterns (batch ELT, streaming, reverse ETL). Define owner, consumer contract, SLA, and data classification BEFORE building.
- Build Through the Warehouse — Layer: Raw/Bronze → Staging/Silver → Marts/Gold → Semantic/Serving. Layer determines rules, owners, and access.
- Quality as Build-Time Evidence — Schema, volume, freshness, business, reconciliation, and anomaly tests. Tests as code = automated quality dashboard. See dbt-data-processing for a live example of what happens when this stage is skipped — tests disabled since October due to alert fatigue.
- Publish — Catalog entry, published contract, role-based access, runbook, deprecation policy.
- Defend (Operate) — Freshness & volume alerts, incident log, quarterly review. Operating the product produces the audit trail automatically.
- Evolve/Retire — DETECT → DECIDE → NOTIFY → MIGRATE → ARCHIVE. RETIREMENT IS A STAGE — a product can be deleted only after consumers know the replacement path.
Governance Payoff
Every lifecycle action produces governance evidence:
- Intake log → demand visibility
- Product spec → metadata standards, data classification
- Warehouse layers → architecture standards, access controls
- Contracts & SLAs → stewardship proof, compliance-by-design
- Tests as code → automated quality dashboard
- Incident logs → audit trail
Phased Adoption (for Caltrans DDS)
Phase 1 — Foundational: Stand up central intake queue. Inventory top 5 outputs. Score against 5-Attribute Test. Phase 2 — Productize: Write specs for top 5. Add schema/freshness/volume tests. Register in data catalog. Phase 3 — Operationalize: Publish SLAs and runbooks. Start incident log. Present quality metrics to Council.
Key Insight (from Case Study)
The same team in lifecycle mode can answer a request in 1 hour that took 2 weeks in ad-hoc mode. The difference: intake queue → existing products identified → contracted metric → internal alert → regression test.
Relationship to Governance
This lifecycle operationalizes the data-governance-framework defined in Session 1. Governance asks: “Who decides, what rules apply, and how do we prove it?” The lifecycle answers through everyday artifacts — specs, contracts, tests, and incident logs. Each stage produces the evidence the governance council needs as a natural byproduct of doing the work.