Next platform deadline: Amazon EKS Nov 26, 2026  

All 14 deadlines

Data Engineering & Analytics

Moving a data warehouse to Snowflake, Databricks or Fabric: a 7-step checklist

Warehouse migrations run late on the code and reports nobody listed, and stall when the numbers do not match. Seven steps that keep them on track, with notes for Synapse, Hadoop and Teradata.

By the SRASH team · · 6 min read

Key facts

  • Conversion tools such as Databricks Lakebridge and Snowflake SnowConvert speed up routine SQL; procedural code still needs engineers.
  • Microsoft publishes migration guidance from Azure Synapse dedicated SQL pools and Spark to Microsoft Fabric.
  • Reconciliation, not conversion, is usually what decides the cut-over date.
  • Retiring unused objects before migrating is the cheapest saving in the project.

1. Inventory what is really used

List every table, view, procedure, macro, script, job and report, then use query history to mark what has not been touched in months. Retiring unused objects before the move is the cheapest saving in the project and shortens every later step.

2. Choose the target and write down why

Snowflake, Databricks or Microsoft Fabric, native tables or open formats such as Delta Lake and Apache Iceberg, and how access will be governed. Record the reasons so the decision survives staff changes. Our decision guide covers the main trade-offs.

3. Pilot the hardest object first

Pick the ugliest stored procedure or ETL flow, convert it and reconcile it. Conversion tools such as Databricks Lakebridge and Snowflake SnowConvert help with routine SQL, but the pilot shows how much engineering the rest really needs, and that is what makes a fixed quote possible.

4. Plan waves by subject area

Group tables, code and reports by business domain and dependency, so each wave can be tested and signed off by the people who use it.

5. Move history, then keep it current

Bulk-load history, then run incremental loads or change data capture so the new platform stays current while the old one still serves the business.

6. Reconcile before you cut over

Compare row counts, aggregates and key hashes table by table, and have business owners compare the reports they rely on side by side. Agree tolerances for rounding and time zones up front.

7. Cut over with a way back, then switch off

Repoint BI tools and feeds, keep the old system read-only for an agreed window, then decommission it and update licence and hardware contracts.

Notes by source platform

  • Azure Synapse: Microsoft documents paths from dedicated SQL pools to Fabric Data Warehouse and from Synapse Spark to Fabric. Check feature differences such as distribution settings and unsupported T-SQL before committing.
  • Hadoop and Cloudera: Hive and Impala SQL usually convert well; Oozie workflows, Sqoop jobs and custom MapReduce need redesign rather than conversion.
  • Teradata and Netezza: BTEQ scripts, macros and stored procedures carry most of the effort; profile them early.

Download the printable migration readiness checklist, or see how our migration readiness sprint works.

Start with a free 30-minute review.

Databases, cloud or AI: talk to a senior engineer about what you run and what is coming. We reply within one business day and sign your NDA on request.