Migrating a retail BI estate from Redshift to Snowflake

The challenge

A premium homeware retailer with online and physical stores was moving its data warehouse from Amazon Redshift to Snowflake. Dozens of BI datasets, dashboards and analyses had accumulated on the old warehouse over years — some business-critical, some abandoned — and nobody wanted to migrate dead weight or discover a broken executive dashboard after cutover.

What we did

Before touching any SQL, we audited the entire BI estate: cataloguing every Redshift-dependent dataset and classifying each by real field-level usage — which were queried live, which were cached, which could be ported, dropped or were blocked on upstream work. Migration effort then went only where it mattered.

For the datasets worth keeping, we translated the SQL systematically, building a reusable mapping of table renames, column changes and dialect idioms between the two warehouses, with a per-dataset verification loop comparing results against the original before switchover.

The outcome

The migration proceeded dataset-by-dataset with verified parity rather than a big-bang cutover. A significant share of the estate turned out to be unused and was retired instead of migrated — reducing both the migration bill and the ongoing warehouse spend. The mapping guide remains as documentation for every future query the team writes.