|
In a way- ALL. A data warehouse when designed should be as granular as possible, and also should have foundation dimensions and measures, which can be used for various purposes. Data Warehouse is like a repository, which provides data for different users through different end-user tools. Ideally one should always design database for at least medium term for the given set of business themes. A Data Warehouse typically is in a relational database format and stores the data to the transaction level detail. The OLAP layers and End User Tools (enterprise reporting, Analytic Applications, Business Performance Management, Data Mining tools..) are created to meet specific purposes by crunching the data available in Data Warehouse. We strongly recommend for you to refer to the section BI-Architecture Components.
Data Warehouse in other words is a kind of 'system of record' for all of your data and information needs. In an ideal world, a Data Warehouse will be congruent to the data lying in the source system, with an added advantage of being cleaner, consistent and integrated. Once you have this data you can use it for million different purposes. Here are some demystifying statements, which will further provide the clarity:
- Data Warehouse not only serves analytics: You can run operational and enterprise reports out of a data warehouse.
- Data Warehouse not only serves the reporting and analytics: You can actually use data Warehouse for operational reasons, like a contact centre executive looking at customer single view, while doing up sell or cross-sell to the customer.
- Data Warehouse does not serve BI only: Over the time, we have started seeing data warehouse as being a platform which can support metadata and master data management initiatives, and not only, what we traditionally know as BI. For example- You can use a data warehouse for your Customer Master Data initiative. Please refer to Master Data Management vs. Metadata Management vs. BI.
- Data Warehouse is not mandatory to achieve some of the objectives like reporting on key performance indicators and Enterprise reporting (which you can do from source systems as well or from an offline reporting database). You can also do data mining, by picking up data directly from the ERP. However, the solution will be much less capable (if you ask 'why' it will take a long answer, so take it at face value).
Given the above, we always recommend Data Warehouse to have as much granular data (in other words, detailed transactional and master data), due to million different applications. |
