Requirements-driven data warehouse design based on enhanced pivot tables

Abstract

The design of data warehouses (DWs) is based on both their data sources and users’ requirements. The more closely the DW multidimensional schema reflects the stakeholders’ needs, the more effectively they will make use of the DW content for their OLAP analyses. Thus, considerable attention has been given in the literature to DW requirements analysis, including requirements elicitation, specification and validation. Unfortunately, traditional approaches are based on complex formalisms that cannot be used with decision makers who have no previous experience with DWs and OLAP. This forces a sharp separation between elicitation and specification. To cope with this problem, we propose a new requirements analysis process where pivot tables, a well-known representation for multidimensional data often used by decision makers, are enhanced to be used both for elicitation and as a specification formalism. A pivot table is a two-dimensional spreadsheet that supports the analyses of multidimensional data by nesting several dimensions on the x- or y-axis and displaying data on multiple pages. The requirements analysis process we propose is iterative and relies on both unstructured and structured interviews; particular attention is given to enable the design of irregular multidimensional schemata, which are often present in real-world DWs but can hardly be understood by unskilled users. Finally, we validate our proposal using a real case study in the biodiversity domain.