Home / Data Warehousing

Data Warehousing

It's not easy to give a clear definition of data warehouse " Data Warehousing, DWH ". The original description of Bill Inmon, the father of the data warehouse, reads as follows: " The data warehouse consists of the collection of time- oriented, integrated, time-oriented data that are the basis of management decisions " " Inmon, 1992 - own translation ". Today it's still the most used description for this concept.

In its original meaning, the data warehouse purely means the storage of information. However, Inmon immediately gives a goal to data collection: making business decisions. This means that data storage is the basis of business intelligence " BI ". The close relationship between the two concepts has made the terms data warehouse and business intelligence coexist, but in practice they are often used as synonyms.

Why do companies set up a data warehouse?

In a data warehouse, information is stored from all types of business applications: from document management to human resources and ERP. In principle, a data warehouse doesn't gather this information by itself. The data can't be checked in the data warehouse either. However, by making relevant information readily available, it makes all types of reports efficient.

Of course, reports can also be made directly through the ERP. The reporting modules " limited " that many ERP systems offer are also satisfactory for some companies. However, a data warehouse can save a lot of time for those companies that want to gather large-scale data and want to get an overview of their business operations. The central data collection ensures that more holistic analysis can be created. This allows real decisions to be made at the level of business policy and optimize strategies. This general approach is also useful for providing general reports to shareholders.

Finally, the combination of a general image and detailed information in the data warehouse also helps companies by showing them if they comply with a certain law. For example, when a legislator receives a complaint of poor business management or privacy, he may request certain information. A company that has its data and activities in the data warehouse useful in advance can be better justified and therefore less likely to be wrong.

Different work models

A data store can be configured in many different ways. The best known models are those of the original Inmon method and the Kimball method. In addition, there is also a newer method called Data Vault. The most suitable method will differ depending on the company.

The Inmon method

The father of the data warehouse, Inmon, uses a top-down approach. Depending on your model, the design of a data warehouse begins with the overall structure. First, the entire standardized data model is configured, and then the data marts.

Data marts contain specific information for a specific department or for a specific application. Like the data model, these data marts have been standardized. The Inmon method is specially suitable for companies that work following really strict and standardized business processes. In addition, it's a very holistic and structured model. Smaller data marts merge seamlessly into a larger data model. The design and commissioning of the entire model requires more time and investment than the Kimball method, but with this clear classification, the system has relatively little maintenance.

The Kimball method

Kimball offered the first alternative to the traditional Inmon method: a bottom-up approach. This doesn't initially involve a standardized data store. Instead, it first focuses on the collection of real data. Then, the data is divided into data marts. The structure of both data marts and larger data models depends, therefore, on the type of data that a company wants to gather. The Kimball method is specially chosen by companies that want to be fast at the operational level and who don't want or can face a large investment. In addition, it's a more flexible working method, since the superior model is influenced by data marts. The biggest drawback that companies find with this method is that a general structure is missing.

Data vault

Data vault is the newest and perhaps the most complicated way to carry out a data warehouse. This model combines all forms of data collection and also connects them in various ways between them. The data vault model consists of three components:

- Hubs
- Links
- Satellites

To clarify each component, we will proceed to explain it using data from an ERP system for trade.

The hubs found within the data vault are tables that represent the business entity. For example, the commercial entity may be a " customer ", " product " or " warehouse ". The entity can be identified through a unique code number and with its different names.

The links represent relationships or transactions between the hubs. In this way, the relationship between a product and the warehouse can indicate the level of inventory. The transaction that occurs between a product and the customer is a buying action.

The satellites complete the data model. They add very relevant information about the hub or link. For example, this may refer to customer location data, special discounts, etc.

What makes the data vault even more complicated is that the data comes from different sources and comes in different versions. All data is saved as recorded. Therefore, the responsibility for the reliability of the data rests with the source. Historical data is also saved. Therefore, an update of some data doesn't eliminate its previous versions.

Often, the data vault method is implemented by companies that want to offer their data in a very dynamic way and that attach great importance to the underlying relationships. Actually, Data Vault goes far beyond the data warehouse. The information design is responsible for directly providing interpretations that are usually more technically covered by business intelligence " BI ".

See also:
Share on Facebook Share on Twitter Share on LinkedIn
Back to top

Home | About Us | Contact | Privacy Policy | Terms of Use

Copyright 2011 - 2020 - All Rights Reserved