Kimball Dimensional Modeling Techniques

What it is

Ralph Kimball introduced dimensional modelling to the data warehousing industry with his 1996 book The Data Warehouse Toolkit. The Kimball Group's website publishes the official list of Kimball dimensional modelling techniques, drawn from the third edition of the Toolkit, as short pages and a downloadable PDF.

What it covers

• Fundamental concepts: the four-step design process (choose the business process, declare the grain, identify the dimensions, identify the facts), star schemas and OLAP cubes.
• Fact tables: additive, semi-additive and non-additive facts; transaction, periodic snapshot and accumulating snapshot fact tables; factless fact tables.
• Dimension tables: surrogate keys, natural and durable keys, degenerate dimensions, role-playing dimensions, junk dimensions, snowflaked dimensions and the calendar date dimension.
• Integration: conformed dimensions and the enterprise bus matrix.
• Slowly changing dimensions, types 0 to 7, each with a clear definition.
• Advanced patterns: bridge tables for many-to-many relationships, late arriving facts and dimensions, audit dimensions.

Why it still matters

Tools changed completely since the 1990s, but most analytics data models are still stars: a fact table of events at a declared grain, surrounded by descriptive dimensions. dbt projects, Power BI semantic models and the gold layer of a medallion lakehouse are usually dimensional models. Knowing the vocabulary lets you design tables analysts can use and discuss trade-offs clearly, for example why a type 2 dimension needs a surrogate key and effective dates.

How to use it well

Start with the four-step process and grain, then the fact table types, then slowly changing dimensions types 1 and 2, which cover most real cases. Model one business process from a public dataset (orders, rides or library loans) on paper before writing SQL. Come back to the advanced pages when you hit a specific problem.

The book

The Data Warehouse Toolkit, Third Edition (Kimball and Ross) is paid and goes much deeper, with a case study per chapter. The free pages are enough to start; buy or borrow the book if you do a lot of warehouse design.

Is it worth your time?

Yes. It is the most concise, authoritative reference on dimensional modelling, and it is free. The pages predate cloud warehouses and lakehouses, so they do not mention them; the modelling ideas apply unchanged.

Common questions

Kimball or data vault? Kimball models are built for querying; data vault is built for integrating and auditing many changing sources, and often feeds Kimball-style marts.

Do cloud warehouses still need star schemas? Wide denormalised tables are sometimes used, but stars remain the default for clarity and reuse.