Databases Reference
In-Depth Information
same. The only difference is the inclusion of specific data quality assess-
ments, which will lead to specific data quality guarantees.
If the data warehouse ETL does not already assess the quality of the data
on its way to the data warehouse, that would be another reasonable cause
to pause the creation of the Market Basket ETL until the data warehouse
ETL has included data quality assessments in its processes. This may seem
rather picky, but it is not. The time and energy expended in guaranteeing
the data quality within the data warehouse should be directly associated
with the data warehouse and not the Market Basket application. If that
time and energy expenditure to create data quality assessments in the data
warehouse is associated with the Market Basket application, it will dimin-
ish the perceived ROI and the Market Basket application and increase the
perceived responsibilities of the Market Basket ETL. In that situation, the
Market Basket ETL will be perceived to be the consumer and owner of
the data warehouse ETL. While that may be true insofar as the human
team members for the data warehouse and the Market Basket ETL may
indeed be the same persons, in terms of management's perception of bud-
gets, roles, and responsibilities, the Market Basket application can only be
best served by incurring only those expenses directly associated with the
Market Basket application and owning only those responsibilities directly
achieved by the Market Basket ETL.
mArket BAsket etl DesIGN
ETL is often thought to be an application, or set of applications, that
retrieves data from an operational source system, conforms it to the
design of a data warehouse, and then loads that conformed data into the
data warehouse. So, it may seem strange to think of ETL as a set of appli-
cations that pull data from a data warehouse, conform that data to the
design of a Market Basket application (or any other datamart), and then
load that data into the Market Basket application (or, again, any other
datamart). But, in the world of data warehousing, that is exactly what
happens. ETL is a generalized name given to applications that move data
in a controlled and intelligent fashion from a source platform to a desti-
nation platform.
Typically, ETL pulls data from a data warehouse and loads that data into
a separate datamart for one of two primary reasons. The first reason is the
