
The Kimball data warehouse bus matrix is a compact planning grid that maps business processes to the dimensions used to analyze them. It supports design, communication, prioritization, and incremental delivery. This article covers its structure, how it relates to conformed dimensions and bus architecture, how to build it, and how to keep it current.
Key Takeaways
- Rows are business processes, columns are dimensions, and marked cells show where they relate.
- Conformed dimensions let users compare measures from different processes using one definition.
- The matrix is a planning tool, not a substitute for logical models, schemas, mappings, or tests.
- Teams can deliver one process row at a time while keeping an enterprise-wide design.
What Is a Kimball Data Warehouse Bus Matrix?
What the matrix represents
Kimball Group calls the matrix an essential tool for designing and communicating the enterprise data warehouse bus architecture. Business processes appear as rows. Dimensions appear as columns. A mark in a cell means that dimension applies to that process.
It is technology-independent at the planning level. Kimball describes it as a big-picture view regardless of database or platform preference.
Here is a conceptual example (illustrative only, not any specific company's warehouse):
| Business process | Date | Product | Customer | Location | Supplier |
|---|---|---|---|---|---|
| Sales | X | X | X | X | |
| Inventory | X | X | X | X | |
| Shipments | X | X | X | X | X |
Bus matrix versus bus architecture
The architecture is the overall incremental, integrated design approach. The matrix is the visual blueprint that documents and communicates it.
The matrix guides detailed dimensional modeling. It is not an entity-relationship diagram, a source-to-target mapping, or a physical implementation spec.
Why conformed dimensions matter
Shared, consistently defined dimensions let users analyze measures from different fact tables together, such as sales and inventory by the same product and month. Kimball says dimensions conform when attributes share the same column names and domain contents. That consistency lets separate fact results align in a drill-across report.

Before marking a dimension as conformed, ask:
- Are definitions and attributes consistent across processes?
- Do the keys and hierarchies match?
- Does someone own and steward it?
How to read the matrix
Scan across a row to test whether a process has the dimensions its questions require. Scan down a column to find where a dimension should be reused and conformed.
Empty cells carry meaning. A blank may be a valid absence, an unresolved requirement, or a design gap that needs stakeholder validation.
What Goes Into a Bus Matrix?
Business process rows and fact-table subject areas
Start with business processes and analytical needs, not source-system tables or report names. Typical process rows include:
- Ordering
- Sales
- Fulfillment
- Production
- Inventory
Kimball's process-event examples include taking an order, processing an insurance claim, and registering students.
Link each row to its eventual fact-table subject area. Document the grain separately before deciding which dimensions belong.
Dimension columns
Common columns include:
- Date
- Customer
- Product
- Employee
- Location
- Account
- Supplier
Your business processes decide the actual list.
Role-playing dimensions need clear handling. One Date dimension can serve as order date, shipment date, and delivery date. Kimball exposes each role as its own view with distinct attribute names, so name roles clearly in the detailed documentation.
Cells, detail levels, and metadata
An early conceptual matrix needs only a mark. A detailed implementation matrix can expand each process into fact tables and add grain, measures, and dimension roles.
Teams often also track fields Kimball does not prescribe:
- Ownership
- Status
- Dependencies
- Delivery estimates
Keep the enterprise matrix readable. Link to deeper definitions instead of cramming attributes, transformations, or SQL rules into cells.
How to Build a Kimball Bus Matrix Step by Step
Kimball's four-step design process is the backbone: select the process, declare the grain, identify dimensions, identify facts.
Start with business decisions, processes, and stakeholders
Identify the decisions the warehouse must support and the processes that generate measurable events. Find stakeholders who can validate terminology, metrics, and priorities.
Review existing reports, source systems, data dictionaries, and KPI definitions as inputs. Then rationalize duplicates and conflicts. Ten overlapping reports about orders are one process, not ten rows.
Declare the grain before selecting dimensions and facts
Write a one-sentence grain statement for every row:
- One row per order line
- One daily inventory snapshot per product and location
- One shipment event per package
Kimball treats grain as a binding contract. Mixing transactions, periodic snapshots, and accumulating snapshots in one row creates ambiguous measures and misleading dimension relationships.

Identify candidate dimensions and define conformance
List the entities users filter, group, and drill by. Decide which can be shared without losing detail or creating conflicting meanings.
Document the exceptions:
- Role-playing dates
- Degenerate identifiers, such as an invoice number kept in the fact table
- Slowly changing attributes (Type 1 overwrite, Type 2 new row, Type 3 alternate column)
- Security restrictions
- Similarly named dimensions that should stay separate
Identify facts and map the relationships
List measurable facts at the declared grain. Separate additive, semi-additive, and non-additive measures, since that determines how users can aggregate them. Account balances, for example, can't be summed across time.
Mark each valid fact-dimension relationship and name the role where needed. Flag many-to-many relationships that may need a bridge table.

Validate the matrix from both directions
- Row reviews: with process owners and analysts, confirm each process has the dimensions its questions need and nothing unrelated.
- Column reviews: with architects and data stewards, confirm shared definitions, spot reusable dimensions, and expose missing relationships that block cross-process analysis.
Prioritize delivery and operationalize the artifact
Pick the first row using business value, dependencies, data readiness, complexity, and conformed-dimension reuse. Plan later increments instead of building everything at once.
Keep version history, owners, definitions, status, decisions, and dependencies in a governed repository or code-assisted workflow, so the matrix stays aligned with models, tests, and delivery plans.
How to Use the Bus Matrix for Incremental Delivery and Governance
Turn the matrix into an implementation roadmap
Build a core set of conformed dimensions, then deliver one high-value process at a time. Kimball recommends implementing one row at a time, with business management setting priorities.

The matrix also balances quick wins against foundation work. A process sharing most dimensions with an existing one is easy to add. A process that needs many new dimensions requires more analysis first.
Use it to communicate with different audiences
- Executives and project managers: coverage, scope, priorities, dependencies, and status, with no physical table design.
- Analysts and business users: which measures they can analyze together and which dimensions are available.
- Engineers and architects: a high-level contract linking requirements to fact models, conformed dimensions, source mappings, tests, and semantic-layer definitions.
Use it as a design-quality checkpoint
Review the matrix whenever someone proposes a new report, metric, source system, or process. Does it extend an existing row, need a new one, or expose a flaw in a dimension?
Empty or inconsistent cells should trigger questions about missing data, unclear ownership, incompatible grain, duplicate dimensions, or unsupported requirements.
Connect the matrix to modern data-platform delivery
The matrix stays platform-independent. Implementation teams then map it to warehouse tables, transformation models, orchestration, data quality tests, documentation, and a semantic or BI layer. In dbt, for instance, model contracts can enforce the shape of a conformed dimension.

Organizations moving from fragmented reporting to a governed production warehouse can use AICG's senior-led advisory, architecture, engineering, platform implementation, and managed support to turn the matrix into production systems their own teams can own. AICG also teaches the bus matrix in its Kimball and data modeling courses.
Keep the matrix current after launch
Set up a lightweight review for new dimensions, changed KPI definitions, new processes, and shifts in ownership or security requirements.
Tie matrix updates to model changes, release notes, data contracts, and stakeholder approvals. Otherwise it decays into an outdated planning document.
Common Mistakes to Avoid
- Starting from source tables or report layouts. This reproduces existing silos, duplicate metrics, and inconsistent dimensions. Start from processes and grain.
- Calling dimensions conformed because the names match. Require agreement on meaning, grain, keys, attributes, hierarchies, and stewardship. Kimball ties conformance to a commitment to data stewardship and governance.
- Going too shallow or too deep. A bare grid with no definitions or status isn't useful. Embedding every column and transformation rule isn't either. Start high-level and drill down as each process is implemented.
- Treating it as one-time. Revisit it as requirements change, especially for role-playing dimensions, many-to-many relationships, missing relationships, and scope decisions.
Frequently Asked Questions
What is a bus matrix?
A data warehouse bus matrix is a grid mapping business processes to the dimensions used to analyze them. In Kimball dimensional modeling, it guides design and communicates the enterprise plan.
What is a bus admittance matrix?
It's an electrical power-systems concept. It relates network bus currents to bus voltages (I = Ybus V). It has nothing to do with data warehouse modeling.
What is an enterprise data warehouse bus matrix?
It's the enterprise-level version of the matrix. It shows business-process coverage and shared dimensions across the whole organization.
What should the rows and columns of a Kimball bus matrix represent?
Rows represent business processes or fact-table subject areas. Columns represent dimensions such as Date, Customer, Product, or Location.
How do conformed dimensions work in a bus matrix?
Conformed dimensions use consistent definitions and structures across multiple fact tables. That lets users compare measures from different business processes.
When should a team create a bus matrix?
Create an initial matrix during warehouse planning, once major business processes and requirements are identified. Then maintain it throughout incremental development.