Every organization can be conceived as a system where value is created by means of business processes. In large organizations, it is common for business processes to be represented by means of process models, which are used for a range of purposes such as internal communication, training, process improvement and information systems development. Given their multifunctional character, process models need to be captured in a way that facilitates understanding and maintenance by a variety of stakeholders.
This thesis proposes an integrated decomposition-driven method for modeling business processes with variants. The core idea of the method is to incrementally construct a decomposition of a business process and its variants into subprocesses. At each level of the decomposition and for each subprocess, we determine if this subprocess should be modeled in a consolidated manner (one subprocess model for all variants or for multiple variants) or in a fragmented manner (one subprocess model per variant). In this manner, a top-down approach of slicing and dicing a business process is taken. The process model is sliced in accordance with its variants, and then diced (decomposed). This decision is taken based on two parameters. The first is the business drivers for the existence of the variants. All variants of a business process has a root cause i.e. a reason stemming from the business that causes the processes to have differences in how they are executed. The second parameter considered when deciding how to model the variants is the degree of difference in the way the variants produce their outcomes. As such, the modeling of business process variations is dependent on their degree of similarity in regards to how they produce value (such as values, execution order and so on).
The method presented in this thesis is validated by two real-life case studies. The first case study concerns a case of consolidation existing process models. The other deals with green-field process discovery. As such, the method is applied in two different contexts (consolidation and discovery) on two different cases that differ from each other. In both cases, the method produced sets of process models that had reduced the duplicity rate by up to 50 % while keeping the degree of complexity of the models relatively stable.