Data Warehouse – The two ism’s

Decisions, Decisions, Decisions & Decisions – you need to make lot of decisions for making a Decision support system. So, Prior to desinging/building a Data Warehouse(DSS), several descisions need to take from the initial stage itself. Making a wrong decision at the initial stage, makes it more expensive to fix.

While building a DWH, the initial decision that you need to make is how you structure your Data Warehouse. Basically there are 2 ‘ism’ s that one can follow and choose to. They are –

1. INMONISM
2. KIMBALLISM

INMONISM is proposed by Inmon, who is considered as Father of Data Warehousing. Inmon follows top-down approach, where an entire enterprise data warehouse is designed first and through which data marts are created and feeded with. Inmon proposed that the Data Warehouse should be developed using an ER model. The architercute proposed by Inmon for implemneting Data Warehouse is ‘Corporare Information Factory’ (CIF).

Corporate Information Factory (CIF) is a technical architecture that gets its data from business operations. This data is been transformed into usable format and loaded into Data Warehouse. Departmental users access the information, manipulate it and assimilate it into their own environments (Data Marts). Once this data is passed on to the appropriate data marts, it can be accessed and analyzed by the decision makers within a company.

The nice thing about relationally designed tables as a basis for a data warehouse is that in a relational format the relational data can be reshaped and reformed into any configuration that is needed. Inmon advocates that once the relational foundation is in place, then it is possible to denormalize the data, which has the flexibility to support multi-dimensional data marts and other data structures like exploration ware-houses and data mining data bases etc.,

KIMBALLISM is proposed by Ralph Kimball, who is considered as Father of Business Intelligence. Kimball’s approach(Kimball Data Bus architecture) is bottom-up data warehousing methodology, in which individual data marts for each departmental data could be created and later combined into a larger one that encompass the data warehouse. Data Warehouse should always be modeled using a Dimensional Model/star schema. As per Kimball, an ER model can be represented as an equivalent set of Dimensional Model/Star Schemas without loss of information.


Data Marts are often an attractive alternative to the enterprise wide Data Warehouse, as they are easier to build and manage. Data Marts can be quickly implemented and offers fast payback for the users. But where the difficulty arises with the Data Marts is that, if they are implemented with no fore-thought for a future Data Warehouse that serves the entire enterprise. This happens when each department select different hardware, software and neglecting to standadise and integrate information. Use of Data Marts with a standard infrastructure can offer unsurpassed business analysis and management capabilities and ultimately it leads to smooth transistion from Data Mart’s to an enterprise wide Data Warehouse.

Both approaches have its own pros and cons, it is up to us to decide which methodology to follow for a given business problem.

Even tough both experts follow different strategies, but they do agree on certain points like –

* Success of warehouse/marts depends on how effectively the 
  business requirements are gathered.
* Granularity and atomic level of data is required for their 
  designs i.e CIF & DATA BUS.
* Stand-alone or independent(silo) data marts or DWH do not 
  satisfy the users at enterprise scale.

So, before concluding this post, it is good to summarize which one to choose for a specific business environment -

Inmon’s CIF if ….

        * DWH is viewed as Strategic.
	* need of integration between departments
	* information requests vary
	* information requests are very ad-hoc
	* have more time to build DWH

Kimball’s Data BUS if …

        * DWH is viewed as Tactical.
	* build a departmental DWH(SILO)
	* information request dont change much
	* scope of warehouse very limited.
        * need to build DWH in less time.

Thus we conclude the post here and will discuss a new topic in next post.

Posted in Data Warehouse | Leave a comment