Business Intelligence Dictionary: Data Warehouse (DWH) Explained | NSBI
What is a Data Warehouse (DWH)?
We define data warehousing as a database infrastructure used to integrate data from various sources, provide standardized view and infer knowledge. A data warehouse is often used to implement a business intelligence architecture.
NSBI Tutorials aim to make Business Intelligence (BI) and Data Warehousing (DWH) attractive to non-technical people as well as to those who are now entering the field and are excited by the numerous ways data is changing our world. NSBI Tutorials are written and delivered by Nick Shopov, (BI Software Developer & DWH Consultant).
Benefits of a Data Warehouse (DWH)
As to align with our data warehouse definition we have grouped some common data warehouse benefits in three categories according to the stage of data workflow:
(a) Data Warehouse Integration
A Data Warehouse is used to integrate data from multiple operational databases and apply historization for slowly changing dimensions. It does not serve operational needs, but is rather used for analysis purposes over a much larger volume of data (millions of records, terabytes of data).
(b) A Data Warehouse Standardized View
A data warehouse is used to apply consistency across data from various operational databases, i.e. provide a common definitions language of the data coming from different systems.
(c) Data Warehouse Inferred Knowledge
A data warehouse serves as a single point of entry for executive reporting with data from multiple operational databases. These reports may also include complex statistical analysis applied on the data repository as well as various pre-composed aggregations, etc.
A Data Warehouse Example
Manager wants to create a CRM (Customer Relationship Management) system and give away discounts for the most loyal clients. This will first require that we integrate those two operational databases into a single data warehouse that includes all time sales per client per order source (in-call or in-venue purchases).
It turns out that the call attendee abbreviated a customer's name as "P. Jackson", while the credit record for the same person says "Peter Jackson". A data warehouse developer needs to make sure that "P. Jackson" and "Peter Jackson" are integrated as one and the same person, i.e. provide a standardized view in the data warehouse.
Last but not least we run an algorithm over all time records for Mr. Jackson as to analyze whether he qualifies for the discount promotion or not and repeat the same procedure for all clients of the pizza place. It turns out that the data warehouse has inferred knowledge, rather than only storing and retrieving user data.