Bus Matrix

Bus Matrix is a design tool introduced by Ralph Kimball to explicitly align an organization’s business processes with the shared analytical dimensions that will describe them in the data warehouse.

What it looks like #

A two-dimensional matrix:

• Rows: business processes (sales, returns, campaigns, invoicing, inventory movements, policy issuance, premium collection, etc.)
• Columns: candidate conformed dimensions (customer, product, time, geography, channel, employee, etc.)
• Cells: an X if that process uses that dimension

Reading the matrix vertically tells you how many fact tables touch a given dimension: the more X’s, the more critical conformity is for that dimension.

What it’s for #

The bus matrix generates no code, creates no tables and tunes no queries. It exists for one purpose: to force the stakeholders (IT, business, finance, marketing) to stare at the same sheet of paper and agree, explicitly, on what they mean by “customer”, “product”, “date”. It’s an organizational alignment exercise that precedes the physical design.

When to do it #

At the very start of the project, before any CREATE TABLE. Kimball recommends it as the first step of the DWH lifecycle. Doing it later, when data marts have already been built independently by departments, costs orders of magnitude more: you need governance workshops, after-the-fact matching processes, mapping tables (xrefs) and time to renegotiate existing definitions.

Relationship with conformed dimensions #

The bus matrix is the diagnostic tool, conformed dimensions are the solution. Wherever two processes share a dimension in the matrix, that dimension must be conformed — same key, same structure, same semantics — otherwise cross-process analyses will return inconsistent numbers.