Multilevel Business Processes: Modeling and Data Analysis

The value creation processes of a hierarchical organization are inherently multilevel. At each hierarchy level, actors conduct value-creating activities using data objects at the appropriate level of abstraction. The different hierarchy levels within an organization are interdependent and interact with each other. Consider, for example, a company involved in various rental businesses. The board makes strategic decisions such as which commodities the company rents to its customers, and devises global business policies. Different divisions manage individual rental businesses, organize the corresponding sub-hierarchy within the company, and devise local business policies. Operational units then handle the actual rentals to customers. Depending on various factors, such as the type of the rented commodity and the renter type, an arbitrary number of additional hierarchy levels with their own business processes may exist in a particular sub-hierarchy. In this thesis, we introduce an artifact-centric modeling approach for the conceptual and logical representation of multilevel business processes. A business artifact describes an object-oriented data model along with the legal execution order of the manipulation operations, also known as the life cycle model. A multilevel business artifact (MBA), on the other hand, describes data and life cycle models at various levels of abstraction along with the synchronization dependencies between these levels. In this context, the employed notion of abstraction level differs from related work on business process model abstraction insofar as the data objects at a particular abstraction level are full-fledged entities with their own identity and life cycles rather than mere views at a coarser granularity of some business process. The task of business process model abstraction is orthogonal to multilevel business process modeling. The MBA-based business process modeling approach is hetero-homogeneous. An MBA describes the homogeneous data and life cycle model of an entire hierarchy of data objects. Multilevel concretization allows for the introduction of heterogeneities for individual sub-hierarchies, thereby following rules for behavior-consistent specialization of life cycle models. Each specialization then becomes the homogeneous model for the particular sub-hierarchy which may again be specialized for a sub-hierarchy of the sub-hierarchy, and so forth. The hetero-homogeneous modeling approach resolves the dichotomy between tailored solutions for individual business entities and the demand for rigorous compliance with company-wide standard operating procedures. An XML-based logical representation of MBAs allows for the (semi-)automated execution of multilevel business processes using information technology which produces event log data as the basis for performance analysis. The hetero-homogeneous nature of MBA models is beneficial to data analysis. Due to multilevel concretization, individual sub-hierarchies may capture performance measures at a finer granularity or record additional measures, thus bearing information which would be lost for analysis when restricted to homogeneous models. At the same time, each sub-hierarchy is guaranteed to provide at least the data defined by the homogeneous model.