Phased migration from SQL Server and hand-built ETL to a lakehouse pattern with validated cutovers for claims and policy systems.
Years of SQL Server jobs and hand-built ETL meant the same metrics were calculated three different ways. We staged a lakehouse-shaped target, validated every cutover with reconciliation, and retired legacy servers on a schedule the business trusted.
How we approached it
Wave planning
Claims and policy domains migrated in waves with dual-write windows and explicit exit criteria before widening scope.
Contracts & tests
Schema mapping with data contracts, regression packs, and signed-off parity checks between old and new pipelines.
Cutover discipline
Zero-downtime playbooks, rollback drills, and on-call bridges so operations knew exactly who to call during migration night.
What we delivered
- Migration waves with dual-write and reconciliation gates
- Schema mapping, data contracts, and regression test packs
- Zero-downtime cutover playbooks and rollback drills
- Decommission plan for legacy servers and jobs
Outcomes
-
One reconciled path for claims metrics across channels
-
Faster incident diagnosis with lineage to source systems
-
Decommissioned legacy footprint with archived audit trail
Representative stack: