Custom filters with MDX subcubes

Omniture Discover is not general purpose OLAP client tool. Rather it is specialized application that lets users to explore data from the Omniture’s Web Analytics product. But it does look and feel like an OLAP browser – which is not surprising, since “A” in OLAP stands for “Analysis”, and the product is all about analyzing web site performance. Omniture model feels like typical cube too - there are dimensions (called segments), and there are measures (called metrics), and there are drilldowns, pivots, aggregations etc. However, there is something not typical about this client tool – the way it handles slicers. Usually, OLAP client provides a UI element called slicer or filter or page field or something similar, where user can choose how the current view should be sliced. UI for slicer does differ between applications, from classic drilldown list to calendar control for Time dimension to something else, but there is always something to choose the current slice. Omniture Discover doesn’t have this. Instead, it uses different paradigm. User is expected to define “custom segments”, which really are dimension filters, but they are given names and list of these segments is stored in special folder. Then user can apply these predefined filters to any report view. The definitions of the segments are very flexible, they really can be arbitrary expressions – something from “Last 3 days” (filter on Time) to “Users referred by www.live.com through paid search campaign who bought product in specific category” (filter on Referer, Action and Product dimensions) etc. Dimensions used in the custom segment definition can be included in the report, and can be omitted from it – the filter is applied in either case. The only limitation is that only one such segment can be applied to the report at a time – if the user want to combine two custom segments, he needs to create a new one. First I was taken aback by how unusual this approach was, but then I saw that it could come in handy sometimes.