 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
| Search BI/DW articles/white papers |
|
|
|
Overview:
Creating a data warehouse and business intelligence application can be a complex, lengthy and challenging process. A subtle wrong turn can change the compass of the roads ahead. You need to be careful about what you are trying to accomplish and what is actually required. It is very easy to get carried away. You must define a clear clarification between ‘must have’, ‘should have’ and ‘could have’ demands.
A recent study by the Data Warehousing Institute (DWI) found that methodology was the third biggest menace – technology and education were number one and two. The three represented 87 percent of the problems identified. Interestingly, solving the third biggest problem--methodology--often minimizes technology and education issues as well. (Source - DM Review Magazine)
So by now we know that we must follow a methodology. Every vendor claims to have one of its own and no matter which one you choose, they all follow two distinct ways. The Big Bang theory way – where you build a data store with all the possible data from the legacy system and then build several functional or subject area repositories from it or, the incremental way – where you build the repositories using an incremental approach strictly based on the reporting requirements and select only required data items from the legacy systems and then integrate them together to create a virtual data warehouse.
Our Methodology:
We have learned from our own experience that there is no such thing as perfect methodology. We believe the best one is what meets the requirements best. We generally construct systems with the 80:20 rule where eighty percent of the data items in the warehouse are based on the current business requirements, and the remaining twenty percent are based on the future use. This way we meet all the current business requirements and avoid being too rigid for not anticipating the future. Of course, we set the priority for meeting the business requirements first.
How do we do it?
We avoid the Big Bang theory. We take incremental approach and design the data warehouse strictly based on the business requirements. For these requirements, we pull the necessary data items from the legacy systems. We implement it for a quick success and then we take the next set of business requirements. We gradually bring all the required data items from the legacy systems. All these data items are integrated to facilitate the basic foundation of an Enterprise Data Warehouse.
Why not bring everything?
We have been asked on several occasions - why not dump everything from the legacy systems? Our response to this question follows an analogy. Consider data warehouse as your kitchen and legacy system as the grocery store. Will it make sense to go to the grocery store and buy everything? No. You make a wish-list of the items and then you buy only those items from the grocery store. You may buy some additional items, but certainly not the entire store! And this is exactly what should be done while building a data warehouse.
|
|
