Introduction to Dimensional Modeling for Data Warehousing Part 2, Dimensional Modeling Principles

In part 1 of this article series, we described theThe level of detail captured should be the lowest
general structure of a dimensional model. In thepossible (atomic level). The lowest level of detail,
present article we shall describe the basic designincludes the full scope of informational dimensions
principles of dimensional modeling. Dimensionalrelated to an event.
modeling follows the four steps defined below. A.As soon as a higher level of detail is selected, by
Selection of the business process (or processes), theaggregating atomic data, certain event dimensions
performance of which shall be monitored. Businessare lost.
processes the performance of which is consideredDimensional models aim to capture measurements,
critical, and relevant data are sufficient (e.g.according to the way an Analyst views data:o Events
operations data derived from these processes),that took place at a certain moment in time o
should be selected with priority. The selectedPeriodic measurements which provide a snapshot of
business process, may relate to a single organizationalthe situation at a given moment in timeo Complete
unit, or spanning more than one organizational unit.view of a transaction which had more than one steps
The capture of overlapping information by different(did not start and complete at a single event)
departments which can lead to many versions ofC. Selection of the dimensions which form the event
truth, is avoided through the capture of a single dataframework, within which the measurements were
stream for an 'end-to-end' process. B. Determinationmade. Common examples of dimensions are: date (or
of the level of detail at which the process shall betime) at which the event took place, Customer,
monitored (also called grain statement). The grainproduct, branch office. The concrete definition of the
statement is the first step in a dimensional modellevel of grain facilitates the selection of dimensions.
design. Examples of grain statement are:·The lower the level of detail, the richer the set of
Each product sold (meaning: an entry shall be createddimensions which accompany the facts. D. Preliminary
in the fact table for each product sold)o Each newdetermination of the analysis methods to be
service contract (e.g. insurance contract) o The dailyimplemented. Selection of the key performance
snapshot of the stock in a pharmacyo Theindicators (KPI) for each monitored business process.
accumulated capture of all facts of a transactionIdentification of the facts needed in order to derive
which has been completed in more than one stepsthese indicators. Given that conditions change,
(e.g. the lifecycle of a tax transaction: tax statementadditional facts may be selected to be captured.
submission - statement control - tax clearance -These facts should relate to the same level of detail.
payment - final payment)The dimensional model should be flexible enough, in
Based on the grain statement, one can derive theorder to allow its future enrichment with new facts in
facts which should be stored in the fact table as wellthe fact table and new dimensional attributes.
as the 'surrounding' dimensions.