Real-time analytics at Bilt: architecture and approach
Bilt runs real-time analytics on dbt with under one second of end-to-end data freshness.
Overview
This session covers how Bilt's data team used dbt and Materialize to build a real-time analytics foundation, applying analytics engineering practices to streaming data and the production tradeoffs involved.
What you'll take away
- 01
Materialize lets dbt compute on streaming data using differential dataflow, delivering under one second of end-to-end freshness, from a Postgres change to a consuming application, with no orchestration required.
- 02
The day-to-day workflow barely changes: version control, testing, and lineage all carry over from standard dbt practice, with one new wrinkle, a four-phase blue-green deploy built for a live database.
- 03
Bilt is now live with thousands of merchant-defined, real-time audiences, and has separately used the dbt Semantic Layer to cut analytics costs 80%.
Session speakers
James Dorado
BiltMarat Biyashev
BiltNick Heron
Bilt