Data Warehousing & BI (KA 9)

Data Warehousing & BI (KA 9)

The Inmon vs Kimball debate, the modern lakehouse compromise, and the BI delivery loop.

Why it matters

The KA the business actually sees. Everything else is invisible until the dashboard loads. DMBOK organises this KA around delivery, not technology.

Going deeper

The classical debate, in one table:

Approach	Build order	Strength	Cost
Inmon (top-down)	Enterprise model first → marts	Enterprise consistency	Slow first delivery
Kimball (bottom-up)	Marts first → conformed dimensions	Fast first value	Risk of inconsistent dimensions
Lakehouse + dbt	Raw → staging → marts in SQL, in git	Engineering rigour	Requires SQL+git discipline

Modern stacks converge on dbt + lakehouse: Kimball-style marts on top of Inmon-style staging, all version-controlled. The argument about ‘which one wins’ is mostly historical.

A data warehouse is the cafeteria, the lake is the loading dock.

Trucks unload raw produce at the loading dock (lake) — boxes, crates, sometimes still muddy. The cafeteria (warehouse) cleans, portions, plates and serves meals — predictable, fast, ready to eat. A diner who'd rather rummage in the loading dock is welcome to; everyone else expects the cafeteria.

The lakehouse is the cafeteria and the loading dock under one roof, with the kitchen visible to both — fewer trucks, less spoilage, but the discipline of the cafeteria still has to live somewhere.