You might have seen product providers’ sales team as they walk into their clients board-room, open-up their presentations carrying the first slide on what we call a ‘dream DW architecture’. A nice assortment of polygons and arrows talk about the every possible data source being integrated into a central data warehouse, which is provider of accurate and reliable information. This is followed by the ‘ideal’ benefits of an enterprise data warehouse with statements like ‘right information at the right time at the right place’, ‘Get all what you need to know on a click of a button’…etc
What intriguing is that why service and product providers make this pitch, when they perhaps know best that building a true enterprise data warehouse in one-go is next to impossible for any medium or large size organization. Surely intent is positive! they might be selling the ‘point of arrival’, but at the same time, they need to be educating their clients that this POA may be many years away.
Let’s talk about what a big-bang Data-Warehouse may involve:
- Having all possible business dimensions
- Addressing all possible business themes
- Picking data from all source systems
- Creating an enterprise data warehouse metadata repository…etc…
You must refer Data warehouse has unique challenges, to understand uniqueness of DW initiatives.
- Fuzziness of business requirements
- Lack of focus from the users, as life can go on without DW.
- Unpredictable use of DW environment…etc…
Set a right foundation and go phase wise:
If you feel that what we are talking make sense, one can adopt million different ways on how to phase out a DW initiative. This depends a lot on your Organization Readiness, how much you have invested into DW initiatives already, the level of success (or failures), you have created in building DW etc…What we have seen in our experience is that:
- There are million different business intelligence ‘current states’
- There are generally similar issues related to BI. If we talk to 50 different companies on their BI issues, there is more than 75% chance that they will come out with the similar set of answers.
- The solutions to these similar problems can be very different depending on an organization.
One more aspect one has to be clear about it that BI point of arrival (POA) is generally the same (enterprise integration of data, virtualization of information, high-availability of information…), in terms of BI capabilities. However the actual architecture implementation could be very different. It depends a lot on the journey to POA and also on how large or complex an organization is. Very few architectures will be anywhere close to (even in POA) the dream DW architecture
Coming to DW Journey to Point of Arrival
My professional network keeps on asking me, if there are standard rules for designing your journey to the point of arrival? Here is my ‘very simple’ tentative answer:
- Start with stand-alone data marts, and let users taste the blood.
- Combine the data marts at a functional level. Fix your functional level agenda first.
- Move gradually towards an enterprise level integrated data warehouse. You can refer integrating data-marts
There is a rider to the above approach. It is must that you create a central governance structure around the Business Intelligence. You may call it Business Intelligence Competency Centre or something else. This central practice can set common standards, disciplines, oversight to ensure that you are not creating the stuff, which you may have a hard time integrating later. Moreover, too many independent data-marts can overload your source systems, if they are not managed well. This central practice can help you a lot in terms of providing inputs on how to have efficient design etc…
The contents of this page are referred from Big-bang Enterprise Data Warehouse is a pipe-dream
