How Fanatics manages dbt metrics with Snowflake CoCo for natural language analytics
Fanatics Betting and Gaming fixed metric chaos with dbt, then layered Cortex Code and Snowflake agents on top.
Overview
Explores how Fanatics built a governed metric framework using dbt's metadata-driven approach (YAML definitions, testing, documentation) feeding Snowflake's semantic capabilities, enabling trusted natural-language analytics via Snowflake CoCo.
What you'll take away
- 01
Three teams asked for the same metric, Gross Gaming Revenue, and got three different numbers; a medallion architecture and a dbt Mesh-style setup gave the company exactly one governed definition.
- 02
Cortex Code, run with full project context, remodeled a 28-model, $300-a-day setup into 4 medallion-style models costing about $60 a day, with humans still validating every change.
- 03
None of the natural-language analytics layer worked until the modeling and governance came first — "garbage in, garbage out" was the team's own framing for why sequencing mattered.
Session speakers
Christine Kwon
Fanatics Betting & GamingBrett Pendleton
Fanatics