Towards a Semantic Framework for Business Activity Monitoring and Management

Process performance management and continuous improvement initiatives can be significantly enhanced by real-time performance, quality and traceability information. For example, the focus of a Six Sigma quality programme is to reduce variability using statistical methods to highlight variance. Current process modelling languages such as XPDL and BPEL provide little or no support for the inclusion of detailed process performance metrics. This paper describes a generic framework using event-based process modelling to support the definition and inclusion of performance metrics and targets within process models. The iWISE implementation of this framework is an XML and Web services-based infrastructure that uses this event-based model for integrating distributed processes and enhancing process visibility using real-time process metrics. Users can adjust alert thresholds on key process metrics in real-time. It uses an integrated rules engine, leveraging semantic technologies such as OWL and SWRL to write rules which are tested as process-related events occur in real-time. Introduction and Motivations Business Process Management (BPM) is the set of methods and tools required to manage the business processes making up an organization. A Six Sigma process improvement approach comprises five key phases: Define, Measure, Analyze, Improve and Control (DMAIC) (Adams, Gupta, & Wilson 2003). This life cycle comprises a set of phases which together form a closed loop of activities. The life cycle is depicted in fig. 1. The DMAIC approach can be summarized as follows: Define Capture the process requirements in a definable and manageable format. Include in this definition key measurements required for analyzing process execution times or other related performance information. Measure Continuously calculate key process metrics as processes are executing using event-based model defined in previous step. Metrics definitions are based on cycle times and Six Sigma calculations. Analyze Analyze enterprise processes for critical changes based on acceptable limits for key parameters, for example, cycle time measurements exceeding a given target. Copyright c © 2008, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. Provide analysis using applicable tools or techniques such as correlation graphs, pareto charts and cause-and-effect (fishbone) diagrams. Web-based dashboard portal technologies are used as a central point for process monitoring and visualization. Improve Use dashboards to identify bottlenecks and inefficiencies in the process and propose improvements. Simulate suggested process improvements to evaluate effect on process design and implement as approved. Control Use control charts and other techniques to verify predictable process states. Figure 1: Phases of the DMAIC process improvement life cycle. Each activity feeds into the next phase of the life cycle. It is difficult to maintain one model of a process where activities are managed by independent business functions. Lack of an integrated process view is one of the limitations of current Business Intelligence (BI) and reporting solutions (Morris, Vesset, & Fleming 2007). Processes, process metrics or Key Performance Indicators (KPIs), and even enterprise events are intrinsically related, but current practice is to handle them using separate software components. Such an approach leads to separation of definitions making it difficult to logically manage business events within their process context. Furthermore, the number of applications and databases used throughout an organization makes it difficult to create a single view of process performance, hence the need to create a common framework to describe process performance measures. (Sheina 2005) summarizes this by explaining that “highlighting the benefits of (near) real-time BI metrics and KPIs is one thing. But correlating them to process data is not a simple matter.” This paper presents a layered framework for process improvement that is based on a common underlying model. The remainder of the paper is organized as follows. A brief overview of process performance management and various related technologies follows this introduction. Following this, an approach for integrating the definition of processes, events and metric thresholds for (near) real-time alerting using a rules-based approach is presented. Final sections cover the software designed to support this capability through leveraging a single model for process modelling, business event management and metric analysis and reporting. Methodologies and Technologies Methodologies and technologies currently used for creating the agile enterprise include BPM, Business Activity Monitoring (BAM), Event-driven Architecture (EDA), and a plethora of modelling and definition standards. In conjunction with these, Semantic technologies are a maturing research area which will also play an ever increasing role. The proposal is that a BAM framework for process improvement will leverage all of these technologies. Figure 2 will serve as the basis for structuring a framework of contributing methodologies and technologies used to manage processes, events, and Key Performance Indicators (KPIs), or metrics, throughout this paper. It also highlights the fragmented nature of various methods or initiatives which could be consolidated. For example, metrics, whilst aligned to processes in management science literature, usually live independently from the process model as part of Data Warehousing (DW) projects. In addition, the framework is not exhaustive. Increasingly, Service Oriented Architectures (SOAs) are gaining prominence in practice as a stable approach to integration. The Operational Systems level represents the various systems supporting daily transactions and functions of the business. Examples of such systems are ERP, CRM and SC systems. An eventdriven BI/BAM solution providing full process visibility is achieved using a full or partial implementation of these layers. Process Performance Management There are three levels of organization structure concerned with processes and metrics: strategic level, tactical level, and operational level. Strategic level goals are an aggregation of tactical level metrics. Tactical level metrics are an aggregation of operational level metrics (Junginger, Kuhn, & Bayer 2004). Key Performance Indicators can be classified into basic KPIs (for all levels) and aggregated KPIs (which are built on basic KPIs) (Junginger, Kuhn, & Bayer 2004). For Figure 2: Methodologies and technologies ‘stack’ required for process, event and metric management. example, at the strategic level, the business goal is service leadership in the market. Service leadership is in turn realized by many sub-goals, among them, guaranteed service time. A strategic KPI might be the ratio of the company’s cycle time to that of the market. At a tactical level, business processes are used to carry out business goals. Here, average cycle time KPIs for various steps in a complaints handling process will determine the performance of the company with respect to its goal to be a service leader. Operational systems used to manage the incident management process might include CRM and ERP software. KPIs at this level may be the number of incidents filed per day without errors. These KPIs are aggregated up from the operation layer, through the tactical layer to ultimately indicate if the overriding strategy goal is achieved. Companies can track their organizational effectiveness using this top-down approach to performance management (Melchert, Winter, & Klesse 2004). Process performance management is implemented using BI and BAM technologies. Business Process Modelling Central to each phase of the process life cycle in fig. 1 is a process model. Process modelling languages (PMLs) define notations for capturing business processes graphically. Business Process Modelling Notation (BPMN) and the Unified Modelling Language (UML) are examples of PMLs. Event-driven Process Chains (EPCs) model business functions, data and events. Events are created by processes or by external actors of the model. An EPC model does not have a machine processable representation. EPCs are used in the ARIS process platform. Integrated Definition (IDEF) is a set of modelling methods that can be used to capture business operations. Some PMLs have an underlying XML format, for example, BPMN. Process execution or definition languages (PELs/PDLs) are specifications understood by process execution engines. One such example is the Business Process Execution Language (BPEL). (Mendling & Neumann 2005) provide a comparison of a set of process languages using 13 metamodel concepts. What emerges from this comparison is that support for defining or gathering metrics and data is not present. Without support for the measurement aspect of a process, monitoring processes remains an independent activity in the overall life cycle. A common model is necessary to support and link the various stages of a process life cycle (see fig. 1). In conclusion, current process modelling approaches lack a performance view required for activity monitoring and reporting.