Informatica → Snowflake Migration

The challenge

The existing reporting warehouse combined Oracle, Informatica PowerCenter and Control-M. It comprised 412 tables, 431 Informatica mappings across 396 workflows, and 94 scheduled jobs. It was costly to run and required regular attention after overnight failures; some incoming files were still managed manually.

The challenge was not just moving tables. Informatica mappings are metadata graphs, not plain code: transformations execute in order, sources and targets connect through ports, and some expressions retain state from one row to the next. A superficial translation can produce SQL that runs successfully while returning different business figures.

What Kovift built

Kovift created an automated conversion pipeline that parsed every mapping into a resolved dataflow graph before generating a Snowflake procedure. Rather than asking a model to infer relationships from raw metadata, it received structured sources, targets, transformation types, ports and execution order.

• Expanded reusable transformations and mapplets to expose full business logic.
• Generated ordered Snowflake procedures for aggregates, update strategies and stateful expressions.
• Replaced Informatica workflows with SnapLogic ingestion and Snowflake procedures.
• Retained Control-M where enterprise scheduling dependencies extended beyond the warehouse.

Validation before sign-off

Generation was never treated as proof. Structural checks verified compilation, column coverage, transformation coverage and step order. A second framework compared source and target results directly—row counts, null counts, column checksums and numeric totals—so every procedure had to agree with the original before approval.

The outcome

The Oracle and Informatica footprint was retired. Annual run cost fell by 50%, manual intervention dropped by 80%, and query performance improved by 50%. The team also gained a reusable validation framework for future pipeline changes.