Smilegate cuts SQL Server migration effort by 77% with AWS Transform and Claude Code

Migrated the MS SQL Server for the STOVE store and payment APIs to Aurora PostgreSQL, applying a hybrid strategy combining AWS Transform for SQL Server and Claude Code
Total project effort was reduced by approximately 77%, from an estimated 21.3 MM using the conventional method to an actual 4.87 MM, resulting in an approximate 4.4x improvement in productivity
AWS Transform generated conversion drafts for approximately 58,000 lines of T-SQL in a single Job run (8–10 hours), while Claude Code handled consistency verification between deliverables and redesign of manual areas
Freezing the schema before proceeding with code conversion, and cross-validation leveraging deliverables as context, were key success factors

Migration Targets and Scope

Targets: Store database with 98 tables, 342 stored procedures, and approximately 58,000 lines of T-SQL
Applications: Store API (.NET Framework 4.5 → .NET 8) and Payment API (.NET 8), with a code size of approximately 100,000 lines

AWS Transform: Consistent rule-based bulk conversion, generation of rich deliverables (evaluation/conversion reports, DDL)
Claude Code: Cross-validation between deliverables, detection of defects without warnings, and redesign of 256 Dapper call sites for PostgreSQL conventions
Validation Tool: Creation and reuse of an equivalence validation tool that executes procedures with the same input in both source and target to compare results

Database Track: 1.27 MM actual effort versus 9.8 MM estimated, achieving 7.7x productivity
Application Track: 3.2x productivity improvement for Store API, 3.2x productivity improvement for Payment API
Additional savings compared to Claude Code standalone execution: 29% for Store API, 27% for Payment API