What-if OLAP Queries with Changing Dimensions

In a data warehouse, real-world activities can trigger changes to dimensions and their hierarchical structure. E.g., organizations can be reorganized over time causing changes to reporting structure. Product pricing changes in select markets can result in changes to bundled options in those markets. Much of the previous work on trend analysis on data warehouses has mainly focused on efficient evaluation of complex aggregations (e.g., data cube) and data-driven hypothetical scenarios. In this paper, we consider hypothetical scenarios driven by changes to dimension hierarchies and introduce the notion of perspectives. Perspectives are parameters such as time or location that drive changes in other dimensions. We demonstrate how perspectives aid in capturing a whole suite of what-if analysis queries. We propose various semantics for OLAP queries under perspectives and develop techniques for the efficient evaluation of such queries. We have implemented our techniques on the Essbase OLAP engine which fundamentally supports changing dimensions, and conducted a comprehensive set of experiments. Our results demonstrate the feasibility, scalability, and utility of our techniques for evaluating what-if queries with perspectives.