A reproducible data platform for global trend intelligence

The client

WGSN is a global authority on consumer and design trend forecasting, used by brands and retailers to decide what to make next.

The challenge

The platform behind WGSN’s trend products had grown hard to change and could not be rebuilt reproducibly: run the same build twice and the answers could differ. Moving to something new also meant protecting the history customers already see in the product.

What we did

• Ran a discovery with the data and product teams and set out findings and recommendations.
• Designed the new platform: orchestration on AWS-managed Apache Airflow, a Data Vault 2.0 core in Snowflake built with dbt, and product-facing data marts on top.
• Managed Snowflake objects and access as code, with CI and automated review on every change.
• Piloted the platform end to end with a social-media trend product, from data collection to the data the application serves.
• Planned the cutover so the history customers already see is carried across intact.

The result

Two identical builds now produce identical data, mart builds run about a third faster, and the pilot ran unattended in pre-production for two weeks without a single failed query.