XP3 allows you to organize and summarize your data in two different ways: these are known as Hierarchy and Segmentation. Both types allow you to create insightful ways of looking at your data, but they are fundamentally different.

Hierarchy

Hierarchy is a means of ordering items of a particular dimension. It is usually displayed as an ordered tree. The main characteristic of a hierarchy is a parent/child relationship. Children of a particular item always belong to its parent. Consider the following example:

East, West, and Central belong to the Total US and are its children. The items lend themselves to a hierarchical representation. Total US would never be a child of East, since it is a summation of East, West and Central. Each measure for Total US is an aggregated value of all of its children.

Segmentation

Segmentation is a means of grouping items based on particular user defined characteristics. A segment defines a characteristic type. An item usually belongs to many different segments. Consider the following example of segments that could describe the Products dimension:

Brand, Flavor, and Size are properties that describe a Product. These are the segments: Apple, Apricot, Cherry are possible values of the Flavor segment. Each Product can be assigned to a value for each segment. If we had a product named “My Cola”, it could be assigned to the Brand X group and the Apple Flavor.

A user of segmentation could choose one or more segments to look at particular items. A user could compare Dollar Sales of Cherry to Cinnamon by choosing the Flavor segment. Aggregating the Dollar Sales numbers for all items in each of the segments will create a single value for each segment, Cherry and Cinnamon.

Furthermore, a user can look at multiple segments one at a time. By selecting more than one segment a user could compare Apple Brand X to Apple Brand Y. Notice that there is no natural order to segments. Flavor is not a child or parent of Brand; it is a completely different attribute.

When to use Hierarchy instead of Segmentation

Segmentation is a very flexible way of representing data. This flexibility is made possible by the fact that all of the aggregation is done on the fly. While this provides almost endless possibilities for data representation, it also has two major drawbacks. Since segmentation relies on aggregation instead of actual items present in the source data, certain non-additive measures such as ACV Wtd Distribution can never be represented accurately for segmented items. If an estimate based on aggregation rules is not acceptable, it is necessary to use hierarchy to select actual data points for your presentation.

The flexibility of segmentation relies on the ability to aggregate multiple items into new, unique line items. In very large data sets, this aggregation may be more than a desktop machine can handle in a reasonable timeframe. If this is the case, it is necessary to represent the data hierarchy to select existing subtotals. In an enterprise environment, hierarchy can be used to pre-aggregate totals on a more powerful server.