About applying filters after aggregating multidimensional data

When you ask common business questions, you often filter values; that is, you try to find dimension values that meet certain conditions. For example, you may want to look at regions where Sales are greater than $2 billion. You likely want to examine values that contribute to the success of these regions.

With multidimensional data, the different levels in a hierarchy contain aggregations of the data at lower levels. The aggregation of data values happens first, then you filter those values. Thus, values for aggregate data are not dependent on the filter. For example, suppose you create a query in which you are displaying Total Company and region values where Sales are greater than $2 billion and Quota Variance is greater than 7%, as shown in "Example of applying filters after aggregating".

Table 3-1 Example of applying filters after aggregating

Regions	Sales Dollars	Quota Variance
- Company A	$10 billion (immutable)	10%
+ Northeast	$3.5 billion	8%
+ Southwest	$2.2 billion	12%

The value for Company A's Sales will not change when underlying dimension values have been filtered. So in our example, Company A's Sales remain $10 billion even though only the regions who Sales are greater than $2 billion are displayed in the query.

An aggregate value will meet the filter condition even if not all of its children do, which enables you to drill into underlying issues. For example, consider our example above in which Quota Variance is greater than 7% for Company A and for the Northeast and Southwest regions. When you drill into the Southwest region, you discover that the Quota Variance for the state of Arizona is -50%; a pressing problem that needs attention.

Aggregate values and their children are often at different scales. For example, a region might be 100 times bigger than any one of its children.