Friday, July 8, 2016

Conformed dimensions and Stovepipe data marts

Long time back, I wrote a small note on Stovepipe data marts, thought make a blog post again on it as it is still unknown to many. The simplest way of remembering What is Stovepipe Data Mart, is remembering a  data mart that cannot be either extended to or integrated to Data Warehouse. This happens if your data mart does not contain Conformed Dimensions.

How do we know whether we design conformed dimensions?

Most of the entities identified during the requirement gathering process are common to entire organization. Example, Product, Employee, Customer are some of the entities. Although these entities have many attributes, only few attributes will be taken into the account during the design of data marts. Not only that, some attributes may not be correctly identified and designed wrongly though the wrong design satisfy the current requirement up to some extent. This does not lead you to conformed dimensions.

In order to make sure that the dimension designed is a conformed dimension, make sure that all related and known attributes are correctly added. There can be a situation that you need only few attributes to the current data mart, but if you have identified many attributes, make sure everything is added. When you add them, make sure that correct business logic related to them is implemented. This makes the dimension as a conformed dimension. Example, if yo u have designed a dimension called Product for Sales data mart and you have added all possible attributes and implemented all possible business logic, then it is a conformed dimension and you will be able to use it when you design Marketing or Finance data marts.

Coming back to Stovepipe data marts, simply if you do not have conformed dimensions, it can be considered as a Stovepipe data mart. You will not be able to integrate the data mart with data warehouse without doing modifications to the data mart, which will be costly, or which will be impossible considering factors like resource availability, system usage and budget constraint. Therefore make sure that you are not leading to a Stovepipe data mart

No comments: