THE PROBLEM
Why this pattern exists
A lakehouse can centralize data but does not automatically reconcile definitions, ownership, identity, historization or business processes.
DESIGN GUIDANCE
Principles to preserve
- Source contracts and ingestion health are observable.
- Business keys, history and reconciliation are explicit.
- Semantic products have named owners and approved definitions.
- Access controls and lineage span ingestion through consumption.
- Data quality is measured against decisions and downstream use cases.
THE REAL DESIGN WORK
Trade-offs, not just boxes
- Flexible raw ingestion is valuable, but defers transformation discipline.
- Central models improve consistency but can become bottlenecks without domain stewardship.
- Near-real-time pipelines add cost and operational complexity over scheduled batch.
REVIEW QUESTIONS
Before you implement
- Which system is authoritative for each business entity?
- How are ARR, customers and contracts reconciled across domains?
- Can a business user trace a KPI back to source records?
- What is the acceptable freshness, quality and recovery objective?