Reporting warehouse moved from SQL Server to Snowflake and dbt
The challenge
Client reporting ran on an ageing SQL Server virtual machine reached over VPN. Every dashboard depended on one server and a set of hand-maintained views, and no one could say with confidence that the numbers were right.
The solution
- dbt running inside Snowflake, with Power BI reading governed reporting tables
- Every staging and reporting view rebuilt as a tested dbt model
- Each model reconciled against the legacy system on every row and column, with the checks kept as automated tests
- A cost model and a phased plan with a parallel run before cutover
Outcome & impact metrics
- 100%models matched the legacy system row-for-row
- ~$500/moprojected platform run cost
- $50K/yrin hosting charges saved after cutover
Architecture: SQL Server → Azure Blob → Snowflake + dbt → Power BI