Browsing by Author "Niedermann, Florian"

Now showing 1 - 1 of 1

Open Access
Deep Business Optimization : concepts and architecture for an analytical business process optimization platform
(2015) Niedermann, Florian; Mitschang, Bernhard (Prof. Dr.-Ing. habil.)
Businesses find themselves today in a highly demanding world: The proliferation of Information Technology combined with highly globalized markets and increasingly distributed value creation has created an environment of unprecedented volatility and complexity. To be able to compete in such an environment, businesses need to be able to rapidly adapt both their service and product offerings and continuously improve their internal efficiency. For many businesses, this implies the need to be able to continuously refine and optimize their business processes. The ability to efficiently and effectively optimize business processes has hence become critical success factor for many businesses and industries. This challenge has been well-recognized already in the 1990s under the umbrella term of Business Process Reengineering (BPR). To overcome the then-dominant split into functional silos, BPR advised business executives to engage in large-scale, revolutionary process changes - usually taking a clean-sheet approach to process design that designed the target state regardless of the status quo. While this approach has been successful in many situations, it has also proven to be highly risky and associated with significant implementation cost. Further, a reengineering project can take considerably longer to implement than today's business cycles would allow. Hence, both business and research have over the past years shifted their focus towards more gradual, evolutionary process optimization methodologies. Compared to revolutionary, clean-sheet process optimization, tool support is much more important in evolutionary optimization. As the status quo is taken as the starting point, the success of the optimization is contingent on understanding it as well as possible and hence depends on three optimization capabilities: First, the optimization needs to take into account as much data about the process and its context as possible. Second, the analyst conducting the optimization needs to thoroughly analyse the data and discover core, often non-obvious, insights that are relevant to the optimization goals. Third, the analyst needs to translate these insights into concrete optimizations of the process that are ideally based on best practices in the respective application domain. In practice, businesses often struggle to excel in these three capabilities, which is at least partially attributable to the (lack of) tool support for evolutionary optimization: Current tools typically offer none or insufficient data integration capabilities, possess only limited analysis support and leave it up to the subjective abilities and judgement of the analyst to spot and properly apply optimizations. As a result, optimization is often both inefficient, i.e., takes longer and is more costly than necessary and ineffective, i.e., does not yield the full potential with regards to the optimization goals. To address this challenge, this thesis presents the deep Business Optimization Platform (dBOP) that combines data integration and advanced analysis capabilities with formalized optimization best practices (so-called patterns) to enhance both the efficiency and effectiveness of Business Process Optimization (BPO): The deep Data Integration (dDI) layer of the dBOP integrates flow-oriented process execution data with subject-oriented operational data sources. While the data integration layer greatly builds on existing schema and data integration techniques, it utilizes its own set of matching rules that take advantage of the specific properties of process data (such as the propagation of matches through assignment between different variables). The deep Business Process Analytics (dBPA) layer builds on the integrated data layer and generates optimization-relevant insights through the computation of key metrics and the application of data mining techniques. The dBPA layer of the dBOP manages to make the application of data mining both powerful and accessible to novice users by tying data mining techniques to certain optimization use cases, effectively reducing required user inputs to a minimum. The results of the dBPA layer are stored in the so-called Process Insight Repository (PIR), a process repository that augments the process model with optimization-relevant insights. In doing so, optimization results can be shared between analysts and accessed in different contexts. Finally, the deep Business Process Optimization (dBPO) layer combines the insights contained in the PIR with formalized optimization best practices and a comprehensive execution strategy to present the analyst with concrete optimization proposals, including a preview of the expected effects. For these proposals that the analyst confirms, the dBOP automatically and correctly rewrites the process. Next to introducing the main concepts of the dBOP, the thesis provides a rigorous evaluation of its capabilities through a qualitative case study, its prototypical implementation and finally, an empirical experiment involving 24 graduate students applying the dBOP to a set of different optimization tasks. Particularly the empirical experiment highlights that the dBOP is a viable approach to increasing the efficiency and effectiveness of evolutionary BPO. Finally, it discusses currently ongoing extensions and future work of the dBOP approach - both within the scope of classical business processes and in the application to other domains, such as manufacturing.