DataWarehouse The data which an OLAP aplication uses are normally stored in a DataWarehouse. A DataWarehouse is a relational database that contans the data obtained from the operational systems databases, or from other data sources. The DataWarehouse can be divided into various DataMarts, which normally contain data which refers to one department, Marketing or Logistics, for example. To create a DataWarehouse the accepted procedure is to start by creating various DataMarts, and then link them. There are many reasons to make a DataWarehouse instead of reading the data directly from the operations systems databases: Performance It is much quicker to access the data of the sole DataWarehouse database than to execute a query against various different databases. To execute complicated queries against operations systems' databases can severly affect performance for other users of these systems. Multiple datasources Combining the data from diferent datasources is quite a complicated task. Normally it is necessary to homegenise the data in one way or another. For example, it is probable that the same criteria are not being used in the different databases. The data should be homogenised during the process which loads data into the datawarehouse. Data Cleansing It is very probable that operations systems' databases contain data which is either incorrect, superfluos or missing. This data can be corected during the process which loads data into the datawarehouse. Adjustments If you are combining financial information from different countries, for example, it will be necessary to ajust the data so that it conforms to just one accounting standard in order for it to be comparable. Another example could be confidential human resources data which should be hidden. These adjustments can be made during the process which loads data into the datawarehouse. Periodicity Data periodicity often varies from database to database, one could be weekly whilst another could have data organised on a monthly basis, for example. The periodicity can be modified during the process which loads data into the datawarehouse. Historical data Historical data is not normally stored in operations systems' databases, but they are an essential part to any meaningful analysis. The datawarehouse is the right place to store this information. Any modifications to the historical data which have to be made to make it comparable can be done during the process which loads data into the datawarehouse. Aggregated data Often it is not necessary to see the most detailed line of information in a BI application analysis. In this case it is possible to keep only the necessary aggregated data in the datawarehouse. The necessary calculations can be made during the process which loads data into the datawarehouse.