A conformed dimension is a dimension that means exactly the same thing across every fact table and data mart that uses it. The same dim_date, dim_customer, or dim_product - same keys, same attributes, same definitions - is shared by sales, support, and marketing facts alike. Conformed dimensions are the single most important idea in Kimball-style warehouse design, because they're what let you compare and combine data across business processes.

Why they matter

The payoff is cross-process analysis. If the sales mart and the support mart both reference the same conformed dim_customer, you can ask "what's the support ticket volume for our top-revenue customers?" by joining both facts to the shared dimension. If instead each mart has its own customer dimension with its own keys and its own definition of "customer," that question becomes a data-reconciliation project. Conformance is what turns a collection of isolated marts into an integrated warehouse where a customer is one customer everywhere.

The bus architecture

Kimball formalized this as the data warehouse bus: a matrix of business processes (rows) against conformed dimensions (columns), where each cell marks which dimensions a given fact uses. Build the conformed dimensions once, and each new business process plugs into the existing bus by reusing them rather than inventing its own. The bus lets a warehouse grow incrementally - add the marketing mart next quarter - while staying integrated, because every new fact connects to the same shared dimensional backbone.

Degrees of conformance

Dimensions don't have to be byte-identical to conform - they have to be consistent. Two dimensions conform if one is a subset of the other or they share a common set of attributes with identical meaning. A daily dim_date and a monthly dim_month conform if month is a true rollup of date; a detailed product dimension and a category-level one conform if category means the same thing in both. This lets facts at different grains still be analyzed together along the shared attributes. Achieving conformance is largely a master-data and stewardship problem: agreeing one definition of "customer" across the business is harder than building the table.

Frequently asked questions

What makes a dimension "conformed"? It has the same keys, attributes, and meaning across every fact and mart that uses it, so the same entity is identical everywhere.

Why are conformed dimensions so important? They enable cross-process analysis - joining sales, support, and marketing facts through one shared dimension - instead of leaving each mart an isolated silo.

What is the data warehouse bus? Kimball's matrix of business processes against conformed dimensions; building the shared dimensions once lets each new process plug in by reusing them.

Do conformed dimensions have to be identical? No - they must be consistent: one a subset of the other, or sharing attributes with identical meaning (e.g., month as a true rollup of date).