This hub covers the architectural decisions that define modern BI: where the semantic layer lives, how metrics are governed, and how tightly the BI tool couples to the transformation layer. The reading order builds from foundational architecture to practical tool selection.
Prerequisites
- Familiarity with dbt (what it does, why teams use it)
- Basic understanding of data warehouse concepts (tables, views, SQL)
- Experience with at least one BI tool (any generation)
Reading Order
1. Semantic Layer Architecture
The three competing implementations (dbt MetricFlow, Snowflake Semantic Views, Databricks Metric Views), the OSI interoperability initiative, and why the semantic layer determines the accuracy of AI-powered analytics.
2. Metrics as Code
Defining business metrics in version-controlled YAML reviewed via pull requests and tested in CI/CD. Covers the define-review-test-deploy-discover workflow, MetricFlow syntax for derived and cumulative metrics, and the governance gap that opens without centralized definitions.
3. Headless BI Pattern
Decoupling the semantic layer from visualization so metrics become an API any frontend can consume. Covers Cube.dev vs dbt Semantic Layer API, when headless BI fits, and the embedded analytics market ($19.8B and growing).
4. dbt as the Center of Gravity for BI
The spectrum of dbt integration (built-on vs connected vs parallel), the Fivetran-dbt Labs merger, and the tradeoffs of choosing a BI tool before or after establishing a dbt layer.
5. BI Tool Selection Framework
Four decision questions (dbt investment, governance needs, budget, user sophistication), a player map grouped by philosophy, and a Lightdash vs Looker vs Metabase comparison.