Databricks migration

Databricks migration, in waves, with every number proven before you switch.

We move your warehouse, ETL and reports to Databricks one business domain at a time, govern everything in Unity Catalog from day one, and retire the old platform on a date you can budget for.

Book an architecture review See the Data Platform Assessment

The problem

What we usually find

Converted code that doesn't match

Conversion tools handle most SQL, but procedures, macros and ETL logic still produce different numbers. Without reconciliation, the first wrong report ends trust in Databricks.

A lakehouse that copies the old warehouse

Lifting tables one-to-one skips the medallion layers, governance and clustering, so costs rise and nothing gets simpler.

Costs that surprise finance

All-purpose clusters running jobs, no cluster policies and untagged compute make the first Databricks bill hard to explain.

Our approach

Four steps, each with a deliverable

01
Assess with evidence
We inventory the estate from query logs and job history, run the Lakebridge analyzer to size conversion, retire unused objects and plan waves by business domain.
02
Land and convert
Ingest with Lakeflow Connect or batch loads into bronze, silver and gold layers. Convert SQL and procedures with Lakebridge, rewrite heavy procedural logic as pipelines, and put everything in Git with CI/CD.
03
Govern and control cost
Unity Catalog with permissions on groups, cluster policies, budgets and tags from the first wave, so the platform is secure and its cost is attributable.
04
Reconcile and cut over
Run old and new in parallel, compare row counts, totals and row-level differences on business keys, repoint reports, then decommission the source domain by domain.

Reference architecture

Deliverables

What you receive

Inventory, conversion sizing and a wave plan with decommission dates
Databricks workspace, Unity Catalog and cluster policies defined as code
Converted, tested pipelines and SQL in your repository with CI/CD
Reconciliation reports for every migrated table
Cost tagging, budgets and a handover to your platform team

Typical timeline

How the work runs

Assess2–3 weeks
Inventory, Lakebridge analysis, target architecture, wave plan and cost model.
Pilot wave4–6 weeks
One business domain migrated to Databricks, reconciled and in production.
Scale waves8–20 weeks
Remaining domains migrated in parallel streams.
Decommission2–4 weeks
Final reports repointed and the source platform archived and switched off.

Step-by-step guides

Migrating to Databricks from

All migration guides, including Snowflake and Microsoft Fabric targets.

Frequently asked questions

Which platforms can you migrate to Databricks?

Teradata, Netezza, Oracle and Exadata, SQL Server, Azure Synapse, Amazon Redshift, Greenplum, Vertica, Snowflake, BigQuery, Hadoop and Hive, SAS, and ETL tools including Informatica PowerCenter, IBM DataStage, Talend, SSIS, Pentaho and Alteryx. Each has a step-by-step guide linked on this page.

Do you use Lakebridge?

Yes, where it fits. Lakebridge, the open-source migration toolkit from Databricks Labs, provides an analyzer to size the work, transpilers to convert SQL, and a reconcile tool to compare data. We review and test everything it produces, and rewrite the patterns it cannot convert.

Are you a Databricks partner?

Our Databricks partner application is in progress. We are independent and take no reseller margin, so our recommendations on architecture and spend are not tied to licence sales.

How do you prove the migrated data is correct?

Every table in a wave is compared between the source and Databricks on row counts, totals and row-level differences matched on business keys. Reports switch over only after the owners sign off the reconciliation.

How is a Databricks migration priced?

The Data Platform Assessment is a fixed fee and ends with a fixed price per migration wave, based on the inventory rather than an estimate.

Have 30 minutes? Get a senior architect’s view of your data platform.

No sales deck. You talk to the engineer who would lead the work.

Book an architecture review