If you extract the data from different source systems, we recommend that you model a Data Warehouse layer for consolidating data as an intermediate layer. Use DataStore objects to do this.

If you want to store data on document level on a BW system, we recommend that you build a Data Warehouse layer from DataStore objects. The data is usually only stored as slightly denormalized with an appropriate granularity.

The granularity of the data stored in the Data Warehouse layer should have the same granularity as the data delivered from the extractor.

The data is not aggregated in the Data Warehouse layer, which means that relevant information is kept for operational analysis.

The extractors for transaction data and for master data should be delta compatible.

The data should contain the entire history.

The data in the layer of architected data marts should match the requirements for data analysis.

The data for supporting operational data analysis should be stored (in the highest granularity) in the Operational Data Store.

Use the same central InfoObjects in all layers. This is the only way to achieve consistency.

Do not model your data flow in a way that results in a stove pipe data model. Here the data flows from the extractor directly to the architected data mart layer.