The principals' framework for consistent dbt at scale
Two principal analytics engineers explain the three layers of dbt governance they chose to centralize, and the 70% faster project setup that followed.
Overview
At Kaizen Gaming, ~50 analytics engineers across six domain teams had each adopted dbt differently. Principal engineers standardized practices without new bottlenecks via centralized foundational models, shared macros/tests packaged as an internal package with automated propagation, and dbt platform config managed through Terraform with scoped CI.
What you'll take away
- 01
Rather than build a full platform team, Kaizen Gaming centralized exactly three layers, a foundational models project, a shared utilities package, and infrastructure as code, and left everything else to domain teams.
- 02
The results: new project lead time down 70%, $10,000 in annualized savings from templated setup, and AI onboarding for new hires down from a multi-day process to about 30 minutes.
- 03
Their decision rule for what to centralize: pull something into the shared layer when no single domain truly owns it or every team would otherwise solve the same problem independently.
Session speakers
Thomas Antonakis
Kaizen GamingStefanos Nikolaou
Kaizen GamingAkis Zervas
Kaizen Gaming