Customer story · Retail
Data on time, every day. That's the win.
Ruben Missinne, Data & AI Officer, Colruyt GroupData platform migration to Snowflake, more than 50 sources in under six monthsAll customer stories
A 9 a.m. promise under pressure
The challenge
Colruyt Group had made a commitment to the business: data ready at 9 a.m., every morning. That commitment was under critical pressure. Compute on the on-premises platform was at its ceiling and the easy performance gains were exhausted.
The warehouse held 21 TB across 2,600 tables and was growing by 45 GB a day and 300 tables a quarter. Behind it sit more than 100 million point of sale transactions a day, with individual Data Vault satellites above 8 billion records.
Scaling on premises meant a large hardware investment. Moving compute to the cloud meant extra development work, but a better business case.
Fast Track
The decision
The group called the program Fast Track: move compute from the on-premises platform to Snowflake, and guard the move as a lift and shift. No redesign along the way, one decision maker, and clear criteria for what counted as done.
How VaultSpeed made it possible
Metadata, not code
The raw vault and PIT pipelines did not move as code. Their VaultSpeed metadata was transferred to a fresh VaultSpeed instance, a few parameters were adapted, the existing sources were switched to the new platform, and the code was rebuilt in full for Snowflake.
VaultSpeed professional services built the tooling for it: a Data Vault cloner, metadata comparison, a business vault split and a template copier. The only real cost on the VaultSpeed side was two weeks of developer time out of the raw vault.
On the Snowflake side, an instance and the knowledge were already in house. A short tailored training and cost monitoring from day one completed the setup.
Results
What changed
More than 50 sources were migrated and live on Snowflake in under six months from decision to production. One source stayed behind for technical reasons. The 9 a.m. commitment is met every morning, and the business case was profitable after three months.
Adding a pipeline no longer disrupts the ones already running, so sharper delivery commitments per pipeline are now feasible and the teams building pipelines have a better experience. Next up: a 30 percent cost and performance improvement on Snowflake in the first year.
The partnership with our vendors outperformed any process.
Ruben Missinne, Data & AI Officer, Colruyt Group