Multi-abstraction layered business process modeling

Abstract

Business Process Management (BPM) acceptance requires a modeling framework in which all abstraction layers (from the high-level enterprise view to the low-level task view) are unified and kept consistent at all times. This paper presents a simple yet effective multi-layered multi-view business process modeling framework that integrates organizational aspects, behavioral aspects, and input and output of business processes. The ultimate goal of this framework is to structure modeling efforts and to improve the usability of business process models in day-to-day decision making. A set of relationships between the layers are defined to support round-tripping as much as possible. A financial services based example is presented as a proof-of-concept in which the modeling framework was tested and assessed.

Introduction

Every organization consists of a large number (hundreds or thousands) of fundamental building blocks to perform their day-to-day work and reach their long-term goals. These work routines (procedures, best practices, etc.), are today also called business processes. Organizations are continuously looking for solutions to not only improve the efficiency and effectiveness of these processes, but also, in order to survive, to become an organization that can quickly adapt to the changes that come with our globalized economy [40]. Business Process Management (BPM) can be defined as the concepts, methods, standards and tools to design, implement, enact and evaluate business processes [37]. Today, BPM technology has reached a mature state. But we are, however, still a long way from the business process penetration and automation that Smith and Fingar proposed already almost a decade ago [31].

This paper focuses on one dimension of the BPM life-cycle [37], namely business process modeling (part of the first phase). Being one of the fundamental aspects of BPM, modeling has received tremendous attention, and a huge number of modeling techniques (and to a certain extend methods) for business processes have been proposed. The modeling techniques range from prescriptive to descriptive methods, and from state based (finite state machines) to task based (BPMN), or a combination of task-based and state-based (Petri Nets). Without going into detail about the adoption and use of these modeling languages in organizations, it is safe to say that modeling languages have proven their usability [25]. However, one of the key problems with BPM today, is that practitioners (and especially business people) are still reluctant to model all of their business processes (or even the key business processes) in full detail using these languages, because the resulting models offer little added (business) value. Consequently, business process models are often seen as a technical picture of the business processes, used as a basis for business process implementation after which they will be left in an inconsistent state from reality. The IT department is in charge of, and takes ownership of business process models, which hampers organization-wide BPM acceptance.

A first key difficulty in applying these notations in wide scale and realistic environments is that these notations are essentially a fine-grained description of the business processes at hand. Process modeling languages require us to think about processes as a set of detailed behavioral patterns (routing logic) between tasks. Therefore, it is currently not interesting for an organization to model all of their business processes and process variants using, for instance, BPMN (or any other modeling method), as it only provides a set of detailed diagrams. Furthermore, it is safe to say that existing modeling languages do not adequately support business people to use business processes in their day-to-day and long term decision making. Simple managerial questions are still almost impossible to answer (even with state-of-the-art process repositories as discussed in [11]): e.g., what is the impact of a change of business process P on processes X, Y and Z, or on process workers A, B and C; what are the task responsibilities of process worker D, etc.. If such questions remain unanswered, how can you demonstrate the added business value of modeling business processes? In addition, contemporary organizations do not have one, but a few hundred or even thousands of business processes, which even further complicates the issue. As a consequence, managing hundreds of (low-level) business process models is nowadays impossible. The authors believe that in our quest to come up with the semantically richest and feature-rich process modeling language, one of the most important aspects of BPM has been overlooked: business processes are (or should be) one of the strategic assets of an organization [5]. Therefore, a first research goal of our modeling method includes to be a genuine tool to manage business processes, i.e., to provide all the different BPM stakeholders with a diagram containing the information they require.

Secondly, the focus on behavioral patterns implicitly reveals a second issue: BPMN and many other business process modeling languages ¹ deal with the definition of sequence constraints between tasks (i.e., which task has to happen when). However, many authors have already identified (e.g., by creating a business process meta-model [15], [34]) that a business process model is much more than a (complex) BPMN or EPC diagram. This behavioral model should not only be complemented with business rule models, task instance allocation schemas, information models, or organizational charts, but also strategic information and performance measures. This makes the problem of managing business processes even more complex as not one model per process but possible multiple models per process should be managed and kept consistent. The true power of BPM is not the behavioral model, but the fact the business processes interconnect all those dimensions. A second research goal therefore consists of providing a modeling method able to represent these multiple dimensions, however, in such a structured way that the resulting business process diagram remains understandable.

The third problem with current BPM state-of-the-art modeling languages (and frameworks) deals with the abstraction levels. Almost all process modeling languages take a business process as its fundamental basis to start modeling by adding levels of detail. BPMN, for instance, uses the concept of sub-processes to introduce a hierarchical modeling approach in which high-level constructs can be hidden from the low-level implementation details. The degree of decomposition, i.e., the hierarchy depth, remains one of the most fundamental but until now unanswered problems. The third research goal is therefore to introduce traceability in the business process decomposition by relating the modeling artifacts, spanning different modeling levels, in a straightforward and semantic way to support round-tripping, not requiring the often too complex and ambiguous translation algorithms currently used in other approaches.

This paper therefore tries to solve some of these issues by introducing a multi-layer, multi-view processes modeling method realizing the three research goals aforementioned: able to provide a business process abstraction adapted to the target audience; expressing all dimensions, and not only sequence flow, in a comprehensible way; and a straightforward traceability between the different modeling artifacts residing on multiple modeling layers. The next section elaborates on some of the basic assumptions underlying the proposed modeling framework. The modeling framework itself will be discussed in Section 4. Section 5 continues with the application of the framework in a real-life example (proof-of-concept). Finally, we conclude the paper and discuss some topics for further research.