CONCEPT Cited by 1 source
Data governance tiering¶
Data governance tiering is the discipline of classifying every table in a data warehouse into a small number of quality tiers — typically three — with progressively lighter documentation, ownership, and retention requirements. The tier is published as first-class metadata (not an ad-hoc convention), and downstream systems (AI agents, search, cleanup jobs) treat the tier as a load- bearing input, not a hint.
Pinterest's three-tier scheme¶
(Source: sources/2026-03-06-pinterest-unified-context-intent-embeddings-for-scalable-text-to-sql.)
| Tier | Scope | Requirements |
|---|---|---|
| Tier 1 | Cross-team, production-quality tables | Strict documentation + data-quality requirements; human-in-the-loop doc review |
| Tier 2 | Team-owned tables | Lighter but still enforced standards; LLM-drafts-human-reviews for documentation |
| Tier 3 | Everything else (staging, temporary, legacy) | Aggressive retention + deprecation policies |
Why it matters¶
Pinterest had "hundreds of thousands of tables, most with no clear owner or documentation." Governance roadmap: reduce table footprint from ~400K to ~100K via standardization + cleanup. Tiering is the enforcement surface — without an authoritative tier, cleanup scripts have no mandate to delete, AI rankers have no trust signal, and every AI-generated query is a coin-flip over staging vs. production.
Pinterest's explicit thesis: "Governance and AI reinforce each other. A disciplined tiering and documentation program made AI assistance viable; the AI systems, in turn, made large-scale governance and documentation tractable."
Where tiers flow into production systems¶
- PinCat is the system of record for tier tags.
- Governance-aware ranking in the Analytics Agent — Tier-1 tables rank higher than equally similar Tier-2/3 alternatives.
- AI documentation prioritization — Tier-1 gets human review; Tier-2 gets LLM-first.
- Retention + deprecation automation — Tier-3 gets aggressive cleanup.
Related patterns in the wiki¶
Tiering as a design primitive is separately documented at the infrastructure and storage levels:
- concepts/storage-media-tiering — hot / warm / cold storage.
- concepts/tier-based-instance-sizing — capacity planning by tier.
- concepts/disaster-recovery-tiers — DR posture by tier.
This page is specifically about the data warehouse table governance instance.
Seen in¶
- sources/2026-03-06-pinterest-unified-context-intent-embeddings-for-scalable-text-to-sql — canonical wiki instance (Pinterest PinCat).