Investigating the Process of Process Modeling and its Relation to Modeling Quality - The Role of Structured Serialization

Jan Claes
Supervisors: Geert Poels (UGent), Frederik Gailly (UGent), Paul Grefen (TU/e), Irene Vanderfeesten (TU/e), p. 242, 2015

Abstract. Lately, the focus of organizations is changing fundamentally. Where they used to spend almost exclusively attention to results, in terms of goods, services, revenue and costs, they are now concerned about the efficiency of their business processes. Each step of the business processes needs to be known, controlled and optimized. This explains the huge effort that many organizations currently put into the mapping of their processes in so-called (business) process models.

Unfortunately, sometimes these models do not (completely) reflect the business reality or the reader of the model does not interpret the represented information as intended. Hence, whereas on the one hand we observe how organizations are attaching increasing importance to these models, on the other hand we notice how the quality of process models in companies often proves to be insufficient.

The doctoral research makes a significant contribution in this context. This work investigates in detail how people create process models and why and when this goes wrong. A better understanding of current process modeling practice will form the basis for the development of concrete guidelines that result in the construction of better process models in the future.

The first study investigated how we can represent the approach of different modelers in a cognitive effective way, in order to facilitate knowledge building. For this purpose the PPMChart was developed. It represents the different operations of a modeler in a modeling tool in such a way that patterns in their way of working can be detected easily. Through the collection of 704 unique modeling executions (a joint contribution of several authors in the research domain), and through the development of a concrete implementation of the visualization, it became possible to gather a great amount of insights about how different people work in different situations while modeling a concrete process.

The second study explored, based on the discovered modeling patterns of the first study, the potential relations between how process models were being constructed and which quality was delivered. To be precise, three modeling patterns from the previous study were investigated further in their relation with the understandability of the produced process model. By ?comparing the PPMCharts that show these patterns with corresponding process models, a connection was found in each case. It was noticed that when a process model was constructed in consecutive blocks (i.e., in a structured way), a better understandable process model was produced. A second relation stated that modelers who (frequently) moved (many) model elements during modeling usually created a less understandable model. The third connection was found between the amount of time spent at constructing the model and a declining understandability of the resulting model. These relations were established graphically on paper, but were also confirmed by a simple statistical analysis.

The third study selected one of the relations from the previous study, i.e., the relation between structured modeling and model quality, and investigated this relation in more detail. Again, the PPMChart was used, which has lead to the identification of different ways of structured process modeling. When a task is difficult, people will spontaneously split up this task in sub-tasks that are executed consecutively (instead of simultaneously). Structuring is the way in which the splitting of tasks is handled. It was found that when this happens consistently and according to certain logic, modeling became more effective and more efficient. Effective because a process model was created with less syntactic and semantic errors and efficient because it took less time and modeling operations. Still, we noticed that splitting up the modeling in sub-tasks in a structured way, did not always lead to a positive result. This can be explained by some people structuring the modeling in the wrong way. Our brain has cognitive preferences that cause certain ways of working not to fit. The study identified three important cognitive preferences: does one have a sequential or a global learning style, how context-dependent one is and how big one's desire and need for structure is. The Structured Process Modeling Theory was developed, which captures these relations and which can form the basis for the development of an optimal individual approach to process modeling. In our opinion the theory has the potential to also be applicable in a broader context and to help solving various types of problems effectively and efficiently.