aggregation layer

The aggregation level is an important feature for linkingdata in data integration and analysis systems in general, e.g. in data warehouses and in business intelligence.

In order to be able to link data with each other, it must be known on which aggregation level the data is located. Technically speaking, this means which key values make a data set unique. For example, in an order, the order number and the total price are on the level of the order header, the article number and the sales price of an article are on the level of the order item, and the description of an individual part is on the level of the parts list of the article. The data of each level cannot be easily transferred to other aggregation levels. For example, the sales price of the item cannot simply be assigned to the individual parts.

The correct assignment is important to ensure that the correct relations are maintained in relational, hierarchical data structures such as in XML and in OLAP data structures, Online Analytical Processing (OLAP). For example, the price of the article must be entered in a table with aggregation level Article and not in a table for the individual parts of the article.

The aggregation level plays a special role in the integration of data from different source systems. For example, the individual items of an order posting cannot be linked to the individual items of the associated order and must be aggregated to the order header aggregation level. The aggregation level is equally central to the creation of star schemas, snowflake schemas, and OLAP structures in business intelligence systems. These are built entirely on the lowest aggregation level intended for them.

The OLAP cube also determines how the key figures are calculated at the various aggregation levels. Typical functions here are the sum, the average or also percentage calculations. Certain key figures cannot be displayed on all levels. For example, the contribution margin of an article does not make sense on the level of individual parts of the article.

The choice of the lowest aggregation level, i.e. the maximum possible level of detail, within a data structure has a considerable influence on the size, the processing speed and the access speed. For this reason, only higher aggregation levels are often provided, especially in evaluation systems.