One engine to rule them all: why Obie left the shire
A two-person data engineering team at Obie Insurance rebuilt a fragmented data platform on dbt and Apache Iceberg, and cut a critical monthly process from over 25 hours to 2.
Overview
Obie Insurance rebuilt its ELT infrastructure using Apache Iceberg, the dbt v2 engine, the dbt Semantic Layer, and AI tooling. This session cover the rebuild decision, resulting architecture, and production lessons.
What you'll take away
- 01
A two-person team rebuilt four duplicated repositories and scattered crontabs, systemd timers, and Prefect jobs into a single dbt project on a lakehouse architecture built on Apache Iceberg.
- 02
A real tuning lesson: setting evaluate_volatile_sql to true on clock-driven logic keeps values correct but forces a rebuild every run, so the team moved clock-derived columns into a view instead.
- 03
The results: a 99% production build success rate, roughly 40% cost savings from dbt State reuse, and a revenue operations month-end close that dropped from over 25 hours to about 2.
Session speakers
Tyson Doberneck
Obie InsuranceMatt Karan
Obie Insurance