What a Snowflake migration covers

AK-DataGeek migrates legacy Oracle, Teradata, SQL Server and PostgreSQL warehouses to Snowflake (or, where it fits better, Amazon Redshift or Google BigQuery). Ownership is end-to-end: discovery, schema redesign, ELT pipelines, cutover and post-migration validation. The same person who designs the Snowflake schema writes the dbt models, configures the Airflow DAGs and explains the trade-offs to your CTO, which removes the lossy handoffs that make migrations expensive.

How the work is sequenced

Every migration follows the same five phases, because they map to decisions that have to be made in order: discovery (2–3 weeks), schema redesign (3–4 weeks), data movement (2–8 weeks, in parallel with pipeline rebuild), pipeline rebuild (4–6 weeks) and cutover and validation (1–2 weeks). Compressing phases by running them concurrently is one of the more reliable ways to blow up a migration. The full reasoning is in the 2026 migration playbook.

Start with a fixed-fee assessment

The Snowflake Migration Readiness Assessment is a fixed-fee, two-to-three week diagnostic of an Oracle, Teradata or SQL Server estate. You receive a migration inventory with risk scores, a dependency map, a do-not-migrate list, a cost-benefit model (Oracle TCO versus cloud, year one and steady state) and a phased roadmap with an effort estimate. It is priced at £9,000 fixed and credited in full against your migration if you proceed within 60 days. There are no open-ended day rates.

Tooling and delivery discipline

Transformations are built in dbt, orchestration runs in Airflow (or Cloud Composer), and infrastructure is managed with Terraform. Every architectural decision is recorded in an ADR, every schema change has a migration script and every data contract has a test, so your team owns a system it can run, debug and extend after the engagement ends. AI tooling such as Claude Code is used heavily to shorten delivery, but no code ships unreviewed.

Where this has been done before

Delivered migrations include HR datasets for a major aviation enterprise moved from a legacy environment to Snowflake, dbt and Airflow, with deployment owned across dev, test and production within a Scrum framework. Other work includes a multi-billion-record geospatial warehouse migration to AWS Redshift with zero downtime. Client names are confidential under NDA; see the case studies.

Frequently asked questions

Which source systems can be migrated to Snowflake?

Legacy Oracle, Teradata, SQL Server and PostgreSQL warehouses. Targets are Snowflake, Amazon Redshift or Google BigQuery.

How long does a migration take?

It depends on the estate. The phases are discovery (2–3 weeks), schema redesign (3–4 weeks), data movement (2–8 weeks), pipeline rebuild (4–6 weeks) and cutover and validation (1–2 weeks), with data movement and pipeline rebuild running in parallel.

What is the Snowflake Migration Readiness Assessment?

A fixed-fee, two-to-three week diagnostic that tells you what a migration will cost, how long it will take and where it will hurt. It costs £9,000 fixed, credited in full against your migration if you proceed within 60 days.

Do you work in regulated industries?

Yes. The work has been delivered in aviation, insurance, transport and fintech, where data correctness, audit trails and compliance are not optional. Client names are kept confidential under NDA.

How do we start?

Email hello@ak-datageek.com. The first conversation is free and exploratory, and AK-DataGeek is now booking Q1 2027.