Case 04 — Legacy Warehouse to Modern Analytical Platform
Situation
A company plans to replace a 15-year-old on-premises warehouse with a cloud analytical platform.
Current facts: - 600 ETL jobs feed a normalized enterprise warehouse and multiple marts; - the team has no trustworthy end-to-end lineage for many jobs; - source systems use inconsistent terms and schemas; - several transformations silently correct source Data Quality defects; - nightly batch meets some needs, but selected fraud/operations use cases require much lower latency; - the target platform supports ELT and scalable raw-data landing; - architects have drawn a high-level “cloud target” but no transition-state architecture; - project teams have begun redesigning schemas independently before enterprise concepts/standards are agreed; - stakeholders assume every ETL job should become ELT because ELT is “modern”; - leaders want one big-bang cutover to avoid running two platforms; - users depend on historical reports whose semantics cannot change during migration.
Your task
Use the ten-layer worksheet.
Your response must distinguish: - current vs target vs transition architecture; - Architecture vs detailed Modeling & Design; - ETL vs ELT as context-dependent patterns; - mapping vs orchestration; - batch vs micro-batch/event-driven/synchronous needs; - profiling vs lineage; - source remediation vs target transformation; - migration delivery vs organizational adoption/change management.
Minimum retrieval requirements
State closed-book: 1. why “ELT is newer” is not a sufficient architecture decision; 2. why a transition-state architecture is required if both platforms coexist during migration; 3. how lineage helps prioritize and safely sequence the 600 jobs; 4. why silently correcting source defects during migration can preserve bad processes; 5. what kind of use case might justify lower latency without making every integration synchronous; 6. why independent project schemas should not define the enterprise target by accident.