How to Build a Data Warehouse?
summary
Argues that startups build a practical warehouse with simple SQL and scripts instead of big BI vendors, using a fact table and simple aggregations as the core, loading via batch SQL jobs stored in git, and modeling occasional dimensions only for performance. Recommends Postgres + SQL scripts for ETL and storage, exposes KPIs via Mondrian MDX cubes (definitions in XML) and a lightweight UI (Cycle/Saiku), and publishes a small helper Monza war file combining Mondrian and the UI.
AI-written summary · about
chapters (15)
- 0:00Data warehouses intro
- 2:54Startups need data warehouse
- 5:45Problems with BI vendors
- 7:32Building warehouse yourself
- 8:59Fact tables in analytics
- 11:30Beyond simple fact tables
- 13:21Data integration basics
- 16:37Automating ETL scripts
- 17:40Querying with MDX/Mondrian
- 22:37Visualizing with Cyclops/Cycle
- 25:55Putting it all together in the stack
- 27:03Hiring BI team
- 27:47Product management in BI
- 29:15Q&A: Elasticsearch vs PostgreSQL
- 30:45Q&A: Handling business questions
related talks
01h 01mModern Data Warehousing Strategy
01h 00mBuilding an Effective Data Warehouse Architecture
39mBuilding a Modern Data Warehouse Without Breaking the Bank
42mBuilding Tomorrow's Warehouse Today: Modern Data Warehouse Modelling Insights
35mData Warehousing with Python
43mData warehousing reinvented for today's needs
43mData Modelling for Startups: Domain Models & the Modern Data Stack
24mData Pipelines to Dashboards