Most of your dbt models aren't in a hurry: how to stop paying for the whole DAG
Most models in a dbt DAG aren't on the critical path. Espresso AI and Authentic Insurance on turning that slack into real savings.
Overview
Every dbt run has a critical path—a small chain of dependent models determining when it finishes—yet the whole run is often costed as if every model were on it. Shares production data and three actionable checks for finding where compute spend exceeds actual need.
What you'll take away
- 01
Most models in a dbt DAG aren't on the critical path, so downsizing their warehouse compute saves money without slowing the job down overall.
- 02
Warehouse sizing follows an Amdahl's Law pattern: each size step roughly doubles cost without doubling speed, making aggressive downsizing off the critical path a straightforward lever.
- 03
Authentic Insurance applied this in production and saves 40% every quarter by routing dbt model runs to right-sized warehouses automatically.
Session speakers
Ben Lerner
Espresso AISaravana Kumar
Authentic