Must every business process be implemented at once?

No. The architecture supports incremental delivery within a shared plan.

Is the matrix a star schema?

No. It shows process - dimension relationships; individual dimensional models require more detail.

Can one process produce several fact tables?

Yes. Transaction, snapshot or summary requirements can lead to multiple implementations, which a detailed matrix can record.

AI Glossary / Data engineering and architecture / Intermediate

Enterprise bus matrix

Definition

An enterprise bus matrix is a data-warehouse planning grid that maps business processes to the dimensions used to describe them. In the Kimball approach, it helps teams plan shared, conformed dimensions while delivering analytical capabilities one business process at a time.

What it means

Rows represent activities such as taking orders, shipping goods or handling returns. Columns represent descriptive dimensions such as date, customer and product. A marked intersection shows that a dimension applies to that business process.

The grid gives teams a broad view of relationships without pretending to be the complete database design. Reading across a row reveals the process's descriptive context. Reading down a column reveals where common definitions are needed across processes.

The word bus is an architectural metaphor for integration through shared dimensions. This is not an enterprise service bus that routes messages between applications, nor a hardware bus. A customer dimension can support several analytical processes because its applicable attributes have consistent meanings, not because those processes communicate through a special network.

How it works

Identify the business processes that generate measurements or events. Discuss them with the people responsible for the work, rather than treating departments or existing reports as the only organizing units.

List the candidate dimensions and mark where each applies. Review every marked intersection: does the dimension describe this process at the required detail, and which role does it play? An order date and a shipment date both use calendar information but represent different events.

Review columns for conformance. Shared labels alone are insufficient if departments define customer, product category or geography differently. Agree on the attributes and meanings that can support comparisons.

Use the matrix to discuss delivery priorities and dependencies. A detailed implementation version can expand a process into specific fact tables and record their grain and measures. The matrix guides those designs; it does not replace decisions about keys, historical changes, source quality or aggregation.

Three practical examples

A retailer plans order and return analysis around compatible product and customer dimensions, making their common reporting context visible.
A training provider compares enrollments and course completions using shared course definitions while retaining different event dates and measures.
An AI analytics project uses the matrix to identify which process datasets can support a cross-process question before asking a model to generate an answer.

Think of it this way

Think of a planning chart that shows which shared reference books each team needs. The chart reveals where everyone must use the same edition and terminology, while each team still performs its own work.

The analogy is limited: conformed dimensions involve precise data definitions and mappings, not merely circulating identical documents. A marked box expresses an intended relationship, not proof that the source data already meets it.

Worked example

Building and interpreting a small matrix

An invented shop starts with three process rows: orders, shipments and returns. It proposes four columns: date, customer, product and carrier. Date, customer and product apply to all three rows. Carrier applies to shipments and returns, but not to the order-placement event in this design.

The grid has 3 × 4 = 12 possible intersections and eleven marked cells. Those eleven marks do not mean eleven physical tables or eleven completed integrations. They describe applicable relationships.

Looking down the product column highlights a shared definition needed in all three processes. Looking across shipments highlights the carrier context absent from orders. The team can implement orders first while designing product definitions for later reuse.

Suppose orders are measured per order line and shipments per shipped line segment. A line split into two shipments must not cause the original ordered quantity to be counted twice. The matrix signals shared context; separate grain statements and correct query design are still needed to compare the facts.

Illustrative case study

A fictional office-furniture supplier wants one dashboard covering orders, deliveries and returns. Three teams already have reports called “product performance,” but one groups products by current category while another uses the category recorded at sale time.

A bus-matrix workshop makes the shared product column explicit. The teams agree which historical and current perspectives each measure requires, document the distinction and assign stewardship for the common attributes. They choose orders as the first delivery because its source and definitions are sufficiently understood.

When deliveries are added, the team reuses the agreed product context but retains delivery-specific dates and carrier information. Its AI assistant is instructed through the reporting model's definitions, not expected to guess what “performance” means. The scenario illustrates planning; it establishes neither automatic agreement nor business improvement.

Common misconceptions and mistakes

Rows should always be departments. The core matrix organizes measurable business processes, which can serve several departments.
Every column must apply to every row. Missing relationships can be legitimate and informative.
Matching dimension names prove conformance. Definitions, values and meanings must also agree.
A completed matrix is a completed warehouse. It is a planning and communication tool that informs detailed implementation.

Business use

Use the matrix early enough to identify shared definitions before separate projects harden incompatible assumptions. It shows how incremental deliveries fit a broader analytical design.

For management, the grid helps reveal reusable work and dependencies. For data teams, it highlights the conversations needed about grain, dimensions and stewardship. A separate stakeholder matrix can show which business groups care about each process without confusing those groups with dimensions.

For AI-enabled reporting, its value is indirect but substantial: consistent context makes cross-process questions easier to define and validate. The matrix alone does not authorize data access or guarantee that generated queries aggregate correctly.

Check your understanding

Two rows both mark a customer column, but one source's customer means a household and the other's means an individual buyer. Can the teams assume the dimensions are already conformed?

Answer: No. The shared mark identifies an integration requirement. The teams must resolve or explicitly model the different meanings before making compatible comparisons.

Related terms

Fact table
Data Warehouse
Conformed dimension
dimensional modelling
grain
star schema
data stewardship

Learn next

Read conformed dimension and fact-table grain. Sketch three processes and their descriptive dimensions, then identify one column whose definition must be agreed across more than one process.

Fact table

Sources

Kimball Group, Enterprise Data Warehouse Bus Matrix: https://www.kimballgroup.com/data-warehouse-business-intelligence-resources/kimball-techniques/dimensional-modeling-techniques/enterprise-data-warehouse-bus-matrix/ and Enterprise Data Warehouse Bus Architecture: https://www.kimballgroup.com/data-warehouse-business-intelligence-resources/kimball-techniques/kimball-data-warehouse-bus-architecture/.

Margy Ross, The Matrix: Revisited: https://www.kimballgroup.com/2005/12/the-matrix-revisited/. Primary accounts of the method, conformance and detailed or stakeholder extensions. The shop matrix and supplier case are original illustrations.

Previous: Entailment Next: Enterprise cognitive system