Lesson 3 — Kimball, Facts, Dimensions & the DW Bus
The Kimball organizing idea
Kimball organizes analytical data around business processes using dimensional structures designed for understandable and efficient querying.
A fact is a quantitative measurement of a business process: revenue, quantity, cost, clicks, balance movement.
A dimension supplies descriptive context: Date, Product, Customer, Region, Channel.
The deciding test is analytical role, not datatype. A ZIP code can be numeric yet still be a descriptive dimension attribute.
Grain comes before measures
Grain states what one fact row represents. “One row per order line” is different from “one row per customer per month.” If grain is unclear, measures can be added together incorrectly and dimensions may not relate consistently.
Good exam reasoning: define the business process and grain before choosing facts or aggregations.
Conformed dimensions
A conformed dimension is governed consistently for reuse across fact areas/marts. If Sales and Inventory both use the same Date and Product definitions, those processes can be analyzed together reliably.
A reused name is not enough. Two marts both calling a table Product does not make it conformed if category definitions, keys, history handling, or values differ.
DW bus and bus matrix
The DW bus is the enterprise integration mechanism created by conformed dimensions/facts across dimensional increments.
The bus matrix is a technology-independent planning tool that crosses business processes/fact areas with dimensions. It helps teams see which dimensions must be shared and how incremental delivery fits an enterprise analytical plan.
Do not confuse the bus matrix with: - a physical star schema; - a batch job sequence; - server/network topology; - an access-control matrix.
Kimball does not mean independent marts
A common distractor says Kimball is just a collection of isolated departmental marts. That misses the entire enterprise integration mechanism. Independent marts with incompatible Customer and Date dimensions create analytic fragmentation; conformed dimensions and the bus are what make the increments work as an integrated warehouse environment.
Inmon vs Kimball
Both approaches pursue trustworthy integrated analytical data.
Inmon/CIF: enterprise integration is centered on a corporate/central warehouse, traditionally normalized, from which marts can be derived.
Kimball: dimensional process areas/marts are the primary analytical structures and integrate through conformed dimensions/facts and the bus.
Do not answer “Inmon” merely because marts exist. Do not answer “Kimball” merely because enterprise data exists. Look for the integration mechanism.
Stop and check
Fact vs Dimension: Revenue = fact; Product/Date/Region = dimensions.
Conformed Dimension: Sales and Inventory use the same governed Product dimension with compatible keys, definitions, and values.
Scenario: Incremental Sales and Inventory star schemas share Product and Date dimensions. Answer: Kimball/conformed-dimension integration.
Source anchor: pp. 368–370.