On managing business processes variants

https://doi.org/10.1016/j.datak.2009.02.009Get rights and content

Abstract

Variance in business process execution can be the result of several situations, such as disconnection between documented models and business operations, workarounds in spite of process execution engines, dynamic change and exception handling, flexible and ad-hoc requirements, and collaborative and/or knowledge intensive work. It is imperative that effective support for managing process variances be extended to organizations mature in their BPM (business process management) uptake so that they can ensure organization wide consistency, promote reuse and capitalize on their BPM investments. This paper presents an approach for managing business processes that is conducive to dynamic change and the need for flexibility in execution. The approach is based on the notion of process constraints. It further provides a technique for effective utilization of the adaptations manifested in process variants. In particular, we will present a facility for discovery of preferred variants through effective search and retrieval based on the notion of process similarity, where multiple aspects of the process variants are compared according to specific query requirements. The advantage of this approach is the ability to provide a quantitative measure for the similarity between process variants, which further facilitates various BPM activities such as process reuse, analysis and discovery.

Introduction

There have been many efforts towards providing agile business process management (BPM) support in recent years. Business process management systems (BPMS) have been recognized as a substantial extension to the legacy of workflow management systems (WFMS). While a typical WFMS supports process design, deployment and enactment, an extension of WFMS functionality provided by BPMS is the facilitation of process diagnosis activities [1]. Furthermore, new requirements emerging from the flexibility and dynamism of business processes require support for instance adaptation, which further impacts on the design, execution and especially the diagnostic activities of BPMS, and eventually will contribute to process evolution and improvement (cf. Fig. 1).

The process diagnosis phase refers to a wide range of BPM activities, including business process analysis (BPA) and process mining and discovery [1], [2], [10]. These post-execution activities are intended to identify and resolve operational process problems, discover preferred work practices, and provide business intelligence. Instance adaptation is an emerging paradigm due to various reasons such as changes in underlying business objectives and operational constraints, and unexpected events that cannot be handled by pre-defined exception handling policies. Consequently, the execution of process instances needs to be changed at runtime causing different instances of the same business process to be handled differently.

Over the last several years of developments in BPM research and industry, we see two equally strong but often conflicting forces impacting on the developments. Where as one fundamental aspect of BPMS and its predecessor WFMS, is to provide control and coordination of business activities, there is another equally demanding aspect of ensuring that the control does not prohibit the operational flexibility, to unacceptable levels.

There are many use cases for such requirements. For example, in in-patient hospital administration processes, where patient admission procedures are predictable and repetitive, however, in-patient treatments such as X-rays, pathology tests, etc. are prescribed uniquely for each case, but none-the-less have to be coordinated and controlled. Another example can be found in higher education and professional training, where students with diverse learning needs and styles are working towards a common goal (degree program). Study paths taken by each student need to remain flexible to a large extent, time providing study guidelines and enforcing course level constraints is necessary to remain compliant with curriculum requirements and maintain a certain quality of learning.

Similarly, consider an engineering firm that provides maintenance and advisory services for telecommunication faults and inquiries. Service plans for individual customer inquiries will be quite diverse, even though basic administration may be the same. We will later introduce the last scenario in more detail as it is used as a running example to demonstrate various concepts and methods.

Many research prototypes (MOBILE [18], ADAPTflex [30], Pocket of Flexibility [34], Worklets [5], DECLARE [29]) have shown a variety of conceptually advanced solutions along this direction (see Section 6 for detail descriptions). In order to provide a balance between the opposing forces of control and flexibility, we have argued for [34], a modeling framework that allows part of the model that requires less or no flexibility for execution to be pre-defined, and part to contain loosely coupled process activities that warrant a high level of customization. When an instance of such a process is created, the process model is concretized by the domain expert at runtime. The loosely coupled activities are given an execution plan according to instance-specific conditions, possibly some invariant process constraints, and their expertise. Current BPM solutions only provide limited support for instance adaptation. For example, the de facto industrial standard for process modeling, business process modeling notations (BPMN) [28] offers a concept called ad-hoc sub-process (AHS) that provides certain level of support for instance adaptation requirements. AHS is a group of activities that have no pre-defined execution dependencies. A set of activities can be defined for the AHS, but the sequence and number of executions for the activities is completely determined by the performers of the activities and cannot be defined at design time. Based on the runtime conditions and their domain expertise, the performers determine how to execute the activities within the AHS, namely the order of execution (sequential or parallel). The contained activities can be executed multiple times until the pre-defined completion conditions are satisfied [28]. Fig. 2 shows an example process model with a AHS for a network diagnostics scenario. Fig. 3 shows three of many potential execution possibilities of the AHS.1

The sub-process illustrated in these figures can be considered part of a process that manages maintenance and advisory services for telecommunication faults and inquiries in a telecommunications company. Below we detail a typical scenario in this regard to provide further motivation and rationale for the approaches proposed in this paper. This scenario will constitute a running example throughout the paper.

A telecommunications company receives customer enquiries about network connection problems, where each complaint case is assigned to a system engineer who is responsible for designing a service plan and solving the problem. The inquiry logging and reporting procedures are predictable and repetitive, while diagnostic tests required to prepare a service plan will typically be case specific and potentially uniquely configured for each case, but none-the-less still have to be coordinated and controlled. The particular configuration of diagnostic tests specific to a given instance is expected to be determined dynamically by a domain expert, such as a senior engineer, based on case specific properties and the experts knowledge and experience.

There are eight diagnostic tests T1,T2,,T8 available in the process, as shown in Fig. 2:

  • Send Test Message (T1);

  • Test Hub201 (T2);

  • Test ExchangeA30 (T3);

  • Reboot Srv59 (T4);

  • Test ExchangeA37 (T5);

  • Loop1078 (T6);

  • Test Hub709 (T7);

  • Test Hub544 (T8).

Any number of these tests can be prescribed for a given request, in some preferred order. The network engineer has the flexibility to design a service plan that best suits the individual case. The design decisions can only be made at runtime when case specific conditions are available and thus cannot be fully anticipated at design time.

Although constructs such as BPMN AHS can be flexibly configured to execute contained activities sequentially or in parallel, there is no means for controls such as restricting the number of selectable activities, nor defining complex/partial dependencies among them. Techniques are required where part of the modeling effort is transferred to domain experts or knowledge workers who make design decisions at runtime under meaningful, domain-relevant constraints. In Section 3, we will provide an approach capable of capturing a large number of selection and scheduling constraints, thus providing a meaningful context for runtime instance adaptations.

At the same time, it can be observed that the typical consequence of an effective instance adaptation environment is the production of a large number of process variants. An executed process instance reflects a variant of realization of the process constraints, and provides valuable knowledge of work organization at the operational level. There is evidence that work practices at the operational level are often diverse, incorporating the creativity and individualism of knowledge workers and potentially contributing to the organization’s competitive advantage. Such resources can provide valuable insight into work practice, help externalize previously tacit knowledge, and provide valuable feedback on subsequent process design, improvement, and evolution.

Nevertheless, the way that domain experts reason about the situation during instance adaptation cannot be truly reconstructed using computational techniques. Building a repository to systematically capture, structure and subsequently deliberate on the decisions that led to a particular design is a more pragmatic way to approach the problem. We observe that a process variant at least contains information from the following dimensions:

  • Structural dimension contains the process model based on which the process instance is executed. For process variants, the structural dimension is represented by the process model that is adapted from the design time model for the particular process variant during instance adaptation.

  • Behavioral dimension contains executional information such as the set of tasks involved in the process execution (may differ from structural dimension due to choice constructs), the exact sequence of task execution, the performers and their roles in executing these tasks, the process-relevant data, and execution duration of the process instance.

  • Contextual dimension contains descriptive information (annotations) from the process modeler about the reasoning behind the design of a particular process variant.

There are various occasions in the BPM life-cycle when precedents of process variants need to be retrieved. For example, during instance adaptation itself, domain experts may refer to a list of precedent process variants designed for a similar situation. Using appropriate analysis techniques, a collection of sufficiently similar process variants could be generalized as the preferred/successful work practice, and consequently contribute to the design of a given instance and subsequently to process improvements.

In this paper we will address both the above issues, namely techniques for supporting instance adaptation, and the utilization of the direct consequence of instance adaptation – management of process variants. We will introduce a framework for instance adaptation to support flexible business process management based on the notion of process constraints. This approach transfers part of the process modeling effort to domain experts who make execution decisions at runtime. Instance adaptation is supported by techniques for specifying instance-specific process models and constraint checking in different variants of the business process. We will demonstrate how the specification of so-called selection and scheduling constraints can lead to increased flexibility in process execution, while maintaining a desired level of control. This aspect is detailed in Section 3.

We then introduce the key technique for managing process variants, namely a query formalization and progressive-refinement technique for process variant retrieval. In our previous work, we have developed a reference architecture for managing such process variants for effective retrieval [23]. The contribution of this paper is to provide an approach for utilizing the retained process variants. An essential concept in this regard is the definition of similarity between process variants in terms of their various dimensions. In other words, how to characterize the degree of match between two similar process variants. This is a hard problem in general due to the informal nature of commonly adopted process description languages, and more so due to the subjectivity in process model conceptualization. Questions such as how to measure the similarity between two process variants having different process models but same sequence of task execution can come forth. From the behavioral perspective two variants are equivalent since they have the same execution behavior. While from the structural perspective they may be dissimilar. Thus variants can share features in one dimension but be dissimilar in another dimension, making an objective evaluation of similarity rather difficult. At the same time, it is desirable that the similarity between the variants can be quantified, i.e., to be able to define a metric space to indicate the degree of similarity or dissimilarity.

The rest of the paper is organized as follows. Section 2 will provide background concepts for the underlying framework for supporting instance adaptation. In Section 3, we introduce the core concept of process constraints, based on which the instance adaptation framework is developed. Section 4 discusses how a repository of process variants manages a large number of executed process variants as an information resource. In particular, the schema for process variants in the repository will be defined. Queries applicable on process variants and their formalization will be discussed in Section 5. In this section, we also provide a quantitative measure for defining similarity between process variants, covering structural, behavioral and contextual dimensions, as well as a progressive-refinement technique for query processing. Related work is presented in Section 6, followed by the conclusion and future work in Section 7.

Section snippets

Framework for managing business process variants

The framework for constraint-based flexible business process management comprises of two major components, namely business process constraint network (BPCN), and process variant repository (PVR). In this section, the overall approach of the framework is presented, which shows how this framework can be applied to BPMS, and the relationship between BPCN and PVR. The functionality of the proposed framework (cf. Fig. 4) is explained with respect to different stages in the BPM life-cycle.

The

Business process constraint network (BPCN)

The foremost factor in designing business processes is achieving improvements in the business outcomes [17]. However, decisions at the strategic level need to be evaluated in light of constraints that arise from several sources. It has been identified that at least four sources of constraints have impact on a business process design:

  • Strategic constraints define the tactical elements of the process, e.g., approval of director required for invoices beyond a certain value.

  • Operational constraints

Process variant repository

It can be observed that BPCN facilitates the creation of “quality” process variants, since each variant conforms to a set of necessary constraints, but also represents a domain expert’s preferred approach to handle a particular case. In the proposed framework, process variants and their properties will be retained in a repository, called process variant repository (PVR). Over time, PVR can build into an immense corporate resource.

The fundamental goal of PVR is to provide an appropriate

Query processing

PVR uses queries to express requirements for process variant retrieval. Based on different retrieval requirements, a query is a collection of one or more process variant features (cf. Definition 3), describing some desired attributes for the targeted variants. Many of such features can be expressed by a typical structured query language, and can mostly be satisfied using well established query techniques like SQL. For example, in order to find all process variants in which execution duration is

Related work

The requirements for providing flexibility in process models and execution stem from the need for change in business processes, which have been recognized for over a decade [9], [16]. Instance level change is regard as the major strength of flexible workflows and has been receiving much attention in recent years. Industrial standard modeling language BPMN [28] provides a construct called ad-hoc sub-process to cater for such a requirement. The introduction of flexible components into process

Conclusion and future work

Variations in work practice often represent the competitive differentiation within enterprise operations. In this paper we have argued for acknowledging the value of variants in business process management platforms. We have presented how process constraints can be used to express minimal restrictions on the selection and ordering of tasks for variants of the targeted business process. The selection and scheduling constraints are specified at design time, through intuitive constraint notations.

Dr. Ruopeng Lu received his Ph.D. in computer science from The University of Queensland in 2008. He is currently working as a researcher at SAP Research Australia in the business process management and semantic interoperability program. Ruopeng is involved in a number of industrial and public funded research projects including the German lighthouse project Theseus/TEXO. His current research areas include advanced business process management technology on enterprise information systems scale,

References (38)

  • W.M.P. van der Aalst et al.

    Workflow mining: a survey of issues and approaches

    Data and Knowledge Engineering

    (2003)
  • T. Madhusudan et al.

    A case-based reasoning framework for workflow model management

    Data and Knowledge Engineering

    (2004)
  • R. Muller et al.

    AGENT WORK: a workflow system supporting rule-based workflow adaptation

    Data and Knowledge Engineering

    (2004)
  • W.M.P. van der Aalst, A.H.M. ter Hofstede, M. Weske, Business process management: a survey, in: W.M.P. van der Aalst,...
  • W.M.P. van der Aalst et al.

    Case handling: a new paradigm for business process support

    Data and Knowledge Engineering

    (2006)
  • W.M.P. van der Aalst, A.K. Alves de Medeiros, A.J.M.M. Weijters, Process equivalence: comparing two process models...
  • M. Adams, A.H.M. ter Hofstede, D. Edmond, W.M.P. van der Aalst, Implementing dynamic flexibility in workflows using...
  • J. Allen

    Maintaining knowledge about temporal intervals

    Communications of the ACM

    (1983)
  • P. Berens

    The FLOWer case-handling approach: beyond workflow management

  • E. Browne, M. Schrefl, J. Warren, Goal-focused self-modifying workflow in the healthcare domain, in: Proceedings 37th...
  • F. Casati et al.

    Workflow Conceptual Model

    (1999)
  • F. Casati, Industry trends in business process management: getting ready for prime time, in: Proceedings of DEXA’05...
  • R. Dechter

    Constraint Processing

    (2003)
  • P. Dias, P. Vieira, R. Rito-Silva, Dynamic evolution in workflow management systems, in: Proceedings of 14th...
  • R. Dijkman, A classification of differences between similar business processes, in: Proceedings of 11th IEEE EDOC...
  • B. van Dongen, M. Dijkman, J. Mendling, Measuring similarity between business process models, in: Proceedings of 20th...
  • M. Ehrig, A. Koschmider, A. Oberweis, Measuring similarity between semantic business process models, in: Proceedings of...
  • C.A. Ellis, K. Keddara, G. Rozenberg, Dynamic change within workflow systems, in: Proceedings of COOCS’95, 1995, pp....
  • M. Indulska, S. Chong, W. Bandara, S. Sadiq, M. Rosemann, Major issues in business process management: a vendor...
  • Cited by (106)

    • Product lifecycle management in knowledge intensive collaborative environments: An application to automotive industry

      2017, International Journal of Information Management
      Citation Excerpt :

      It is also observed that in many cases existing information and communication technologies tools are not prone to manage effectively daily activities due to their lack of flexibility regarding dynamic (Kammer, Bolcer, Taylor, Hitomi, & Bergman, 2000) and changeable environments and processes (Lu, Sadiq, & Governatori, 2009; Mangan & Sadiq, 2002). Large enterprise information systems impose structured and predictable workflows, whereas “real world” processes, in particular, those related with concept and design of new products and services are hardly predictable once they involve several human based decisions and collaboration (Hill, Yates, Jones, & Kogan, 2006; Lu et al., 2009). Moreover, many times, large enterprise information systems are not able to manage communication and information together with the knowledge-based process itself (Mangan & Sadiq, 2002).

    • SiDiTeR: Similarity Discovering Techniques for Robotic Process Automation

      2023, Lecture Notes in Business Information Processing
    View all citing articles on Scopus

    Dr. Ruopeng Lu received his Ph.D. in computer science from The University of Queensland in 2008. He is currently working as a researcher at SAP Research Australia in the business process management and semantic interoperability program. Ruopeng is involved in a number of industrial and public funded research projects including the German lighthouse project Theseus/TEXO. His current research areas include advanced business process management technology on enterprise information systems scale, and its enablement through SOA and SaaS methodology.

    Shazia Sadiq is currently working in the School of Information Technology and Electrical Engineering at The University of Queensland, Brisbane, Australia. She is part of the Data and Knowledge Engineering (DKE) research group and is involved in teaching and research in databases and information systems. Shazia holds a PhD from The University of Queensland in Information Systems and a Masters degree in Computer Science from the Asian Institute of Technology, Bangkok, Thailand. Her main research interests are innovative solutions for Business Information Systems that span several areas including business process management, governance, risk and compliance, data quality management, workflow systems, and service oriented computing.

    Dr. Guido Governatori received his Ph.D. in computer science and law at the University of Bologna in 1997. Since then he has held academic and research positions at Imperial College, Griffith University, Queensland University of Technology, the University of Queensland, and NICTA. He has published more than 150 scientific papers in logic, artificial intelligence, and database and information systems. His current research interests include modal and nonclassical logics, defeasible reasoning and its application to normative reasoning and e-commerce, agent systems, and business process modeling for regulatory compliance. Dr. Governatori is a member of the editorial board of Artificial Intelligence and Law.

    This research work has been conducted at The University of Queensland.

    View full text