Business Process Improvement (BPI) involves discovering weaknesses in a business process and addressing them in subsequent passes. This paper aims at establishing existing BPI frameworks, their status in terms of validation, tool support and gaps that require further investigation; including the potential of Visual Analytics (VA) application in bridging the gaps. A survey of relevant journals and other works is carried out that involves examination of the sources for BPI frameworks, tool support, validation, application of VA and any significant results of the research. So far no tools have been realized to deal with diagnosis and design in workflow systems or attempted to use VA. BPR and Process Mining lack techniques and tools to support automatic/intelligent redesign hence improvement. Though VA holds promise in solving many of the current frameworks limitations, research so far carried out does not addressed redesign of processes. We conclude that existing BPI frameworks have an information gap between the diagnosis and design phases of the BPM cycle hindering automatic or intelligent process redesign that would result in process improvement; and lack validated tool. A number of issues such as abstraction and complexity in the current frameworks can be solved using Visual Analytics.
Keywords- Business Process Improvement, Empirical research, Frameworks, Tools, Visual Analytics
Most modern organizations start off with well laid out operational plans or process. After a short while, the processes may have changed due to business dynamics, unforeseen scenarios, changes to product/services offered, preferred working relations and so on. In some cases such process are very many and may not always be adhered to. Some may be inefficient, yet not known to management. This may require updating of the operational process models hence need for tools that provide feedback on the processes and subsequent improvement .
This survey examines research carried out under a number of Business Process Improvement (BPI) related frameworks. The paper aims at establishing existing BPI frameworks, their status in terms of validation, tool support and gaps that require further investigation. It also examines the potential of Visual Analytics (VA) application in bridging the gaps. A survey of relevant journals and other works is carried out. The sources are examined for application of VA, tool support, validation and any significant results of the research. The findings form the basis of conclusion and recommendations of the paper.
The rest of this paper is organised as follows: Section 2 outlines the background of the subject, existing BPI frameworks in section 3, discussion in section 4 and conclusion in section 5.
Process improvement has long been practiced in the manufacturing sector as 'lean manufacturing'. Lean thinking is used with the aim of providing products or services in the most efficient manner by improving flow and eliminating waste from processes, hence holds promise for service sectors such as healthcare . Business Process Improvement (BPI) involves discovering weaknesses in a business process and addressing them in subsequent passes . A business process has at least three perspectives of control, data and resources operating in a particular environment/context necessitating an all round approach . BPI is a subset of a wider discipline called Business Process Management (BPM), mostly employed to improve, re-design or re-engineer existing business operations so as to improve overall effectiveness or efficiency of an enterprise . BPM involves a number of concepts, methods and techniques to support five distinct phases of design, implementation, enactment, control and diagnosis [3, 6].
Starting with design phase, business processes are identified, moderated and represented as operational process models for the organization. The processes could be spelt out on paper, charts or embedded in operational software. Configuration phase, involves definition of business model in sufficient detail for implementation. The chosen alternative process models are configured in detail and checked for completeness before being passed over for execution. In the execution phase, the process is enacted as per the configured system. The context data is related to each specific instance of the process model and this combination of definition and context data is used for the control of the process. Process execution is monitored at the control phase. A monitoring tool is used to give information about individual processes as well as overall or collective processes. Such information could be current or from the past three months. Diagnosis, which is the final phase, is used to reveal weaknesses/bottlenecks in the process , which can be addressed in the design phase of the next pass, yielding an improved process.
The major handicap encountered in the BPM cycle is that there is no support for automatically transferring the improvements to address weaknesses/bottlenecks discovered in diagnosis phase to the design phase. This resulted in researchers at University of Aarhus, Denmark to call for new techniques in Process Mining and intelligent redesign for the redesign and diagnosis phases, so as to close the BPM life-cycle .
3 Existing BPI Frameworks
Process improvement related research has been undertaken in areas such as Workflow Technology, business process redesign and process mining. Some research has also been carried out in the specific area of healthcare process improvement, and finally this section reviews research involving VA and healthcare as outlined in the paragraphs below.
3.1 Workflow Technology frameworks
In a dynamic task assignment environment, researchers suggest adoption of a Swam intelligence approach/framework in emergency response situations. However inability to accommodate interrupts still poses challenges as is the case with all workflow based technology. Colored Petri Net (CPN) tools was used to build a fictional case model which was verified to be 99% correct, but could not be validated using real life or historical data . There is need for abstraction simplification as well as aggregation of performance in workflows using event logs . A process model simplification framework, which is control flow focused, was developed and demonstrated using a prototype. Among their suggestions is that, integration of Oryx, ProM and AProMoRe will contribute towards model simplification. A number of equally important limitations were highlighted in the research that include; lack of visualization of abstraction and different information aspects; the static nature of the models as opposed to animation which could help users understand simplified models; need for interaction with users while developing the model and luck for real time capabilities for purposes of handling large models. In the research, a tool was produced and validated .
Inflexibility and adaptability have been identified as major weaknesses of workflow systems in the literature [13, 14]. Workflow systems are ineffective in dealing with interrupts as well as complex exceptional handling in unstructured situations such as emergencies. This is attributed to their inherent focus on process routing or control flow while oversimplifying resource and task attributes as well as limitations in representation of dynamic aspects of an organizational environment.
3.2 Business Process Redesign Frameworks
In validation of their earlier redesign framework, Mansar and Reijers conclude that their framework is indeed helpful in supporting process redesign and argue that its core elements are recognised and have been put in practice by the BPR practitioner community . However, no tool was realised and no mention of VA.
3.3 Process Mining
Application of different discovery algorithms on the same log in ProM generates different models of which is difficult to select the best . Other challenges include: accuracy of discovered process model in reflecting observed behaviour; and how well the discovered model describes reality. Evaluation of an event-log and process model can take place in different orthogonal dimensions that include fitness, precision, generalization and structure. In conclusion they call for adequate validation techniques to evaluate and compare discovered models and their quality . The research is not VA based, does not realise a tool hence no validation.
Algorithms used in Process Mining require well formed/structured process models for projection of performance information as opposed to complex environments. The researchers also developed and validated a tool that was implemented as Fuzzy Miner plug-in in the Process Mining Tool (ProM) . However the research does not address re-design and VA.
Process Mining is limited to the control flow perspective and only pinpoints the bottlenecks in the current processes . The framework cannot suggest alternative processes or how the bottlenecks identified in the current processes can be surmounted. Due to its rigidity alternative process models can only be generated by changing the data (event-logs). ProM does not incorporate VA. Many mining techniques assume that all events in an event log are logged at the same level of abstraction, contrary to reality, and the algorithms used produce results that are hard to understand . The researchers proposed a simple clustering algorithm to derive a model from an event log. The model only contains a limited set of nodes and edges, each node representing a set of activities performed in the process, but many nodes can refer to many activities and vice versa. The research produced and validated a tool.
Traditional process mining approaches have problems dealing with unstructured processes and recommend development of new techniques and on using existing techniques in an innovative way to obtain understandable high level information instead of the spaghetti-like models showing all details. The research did not produce a tool and was not validated.
3.4 Healthcare Business Processes
In the development of YAWL4Healthcare framework , the researcher argues that Workflow technology provides strong support for configuration and execution phases; very basic support for design and control phases and no support for diagnosis phase. He further argues that interoperability issues also often arise despite the reference framework. Though he realised and validated a tool, the framework does not address the BPM phases of execution, diagnosis and design .
Research focusing on the reduction of throughput time by exploiting Business Process redesign heuristics in a mental healthcare case resulted in the so called structured framework. In their conclusions, they argue that the application of best practices show potency in specific and very similar context; and recommend a more structured method on how to combine an effective set of best practices for a specific medical context . Limitations in this work include: lack of life testing; and reliance on intuition as part of the redesign procedure. The approach is also revolutionary as opposed to evolutionary or does not use historical process data.
A Healthcare system was analyzed from three different perspectives: (1) the control flow perspective, (2) the organizational perspective and (3) the performance perspective . They argue that traditional process mining approaches have problems dealing with unstructured processes and recommend development of new techniques and on using existing techniques in an innovative way to obtain understandable high level information instead of the spaghetti-like models showing all details. The research did not produce a tool and was not validated.
Research by Fitzgerald and Dadich  reveals how complex processes within a hospital can be understood, the way areas for improvement can be identified and simulation of viable options for improvement. The research did not exploit event-logs since most of the processes were captured by on site-observation. It was therefore limited in the sense that historic data was not available to automatically generate earlier processes for evaluation. Availability of earlier processes facilitates conformance checks and identification of improvement options. A number of researchers have argued against the use of simulation in complex processes as in healthcare, given that they assume abstract situations with many assumptions about reality that do not hold for long [2, 3].
Table 1 compares the various BPI related frameworks in Workflow technology, BPR, Process mining and those that are VA based. The focus of comparison is on goal of study, whether they are VA based, have tool support, if they were empirically validated and the recommendations or conclusions made from the study.
Table 1: Summary of BPI frameworks
Author Framework Tool support Validation Significant results
Reijers et al. (2007) Swam intelligence
Yes None None
Nugteren (2010) Process model simplification Yes Yes i) Built a prototype for Model simplification and aggregation (abstraction handling)
ii) Suggest integration of Oryx, ProM and AProMoRe; visualization, animation and interaction with dynamic models
Mans (2011) YAWL4Healthcare framework Yes Yes i) Workflow technology provides strong support for configuration and execution; very basic support for design and control, and none for diagnosis phase.
ii) Use of process mining to provide support for all phases of BPM cycle
Jansen-Vullers & Reijers (2005) Structured approach None None None
Mansar & Reijers (2005) Redesign framework None Yes i) Concluded that the framework is indeed helpful in supporting process redesign
ii) Its core elements are recognised and put in practice by the BPR practitioner community
Rozinat et al. (2008) Process Mining evaluation None None None
Mans et al. 2009).
Application of Process Mining in healthcare None None None
Gunther & Aalst (2007) Process simplification Yes Yes i) Algorithms used in Process Mining require well formed/structured process models for projection of performance information as opposed to complex environments
ii) Implemented as Fuzzy Miner plug-in in ProM
Author Framework Tool support Validation Significant results
Hornix (2007) process performance analysis Yes Yes i) Process mining is limited to control flow perspective only
ii) Changes to the model can only be effected through change of the input data (event-logs)
iii) Implementation of the Performance Analysis with Petri net and Performance Sequence Diagram Analysis plug-ins in ProM
Dongen, and Adriansyah (2009) Performance Visualization
Yes None None
Fitzgerald & Dadich (2009) VA based scheduling and patient flow improvement None None None
Riemers (2009) Process improvement using VA None None None
4 Visual Analytics in Business Process Improvement
One of the aims of Visual Analytics is to make data and information processing transparent. It can handle complexity in data and information overload. Such Complexity can be in form of: massive amounts of data, high dimensionality, heterogeneity, multiple facets, time variance, incompleteness, uncertainty, inconsistency . The technique also has capacity to portray processes at different levels of abstraction [26, 27]. VA has features common to information visualization, data mining and process mining properties that present the possibility of developing VA based tools for process mining in addition to other possibilities .
VA techniques are visual and interactive resulting in four important attributes. First, they can help users to understand complex data and situations where models alone are inadequate. Second, they readily detect trends and anomalies, evaluate hypotheses, and uncover unexpected connections. Third, through the use of contextual cues, they help the user to interpret the information he/she is presented. Finally, they encourage users to engage with and explore large datasets that might otherwise be daunting [30, 31].
One of the challenges in Process Mining complex models such as in healthcare, which have been created from mining algorithms such as ProM, is the higher levels data abstraction that summarizes complex relationships within large datasets. Though the abstraction is intended to make it easier for understanding, it, on the other hand makes it difficult to interpret behaviour of individual nodes or relations relative to the original data space. In such cases, Visualization can be used to drill down to the data level to link an observed relation to its data. It can also be used to visualize modelling alternatives, User view-based process visualization and to understand modelling results through better model-data linking and presentation [29, 32].
Abstraction handling is the main challenge in BPM. Likewise, Complexity and exceptional handling are the main challenges in Workflow systems. So far no tools have been realized to deal with diagnosis and design in workflow systems or attempted to use VA. This can be attributed to the nature workflow technology that is confined to control flow; a static procedure. The absence of VA application especially to deal with abstraction can only be attributed to the time difference in the evolution of the two techniques, a gap that can be crossed in future.
BPR though dedicated to design has yet to realize tools for automatic process redesign. Historical data (event logs) presents a special challenge in abstraction handling since it is not possible to tell which activities are aggregated and which are not. New tools such as in intelligent re-design and Visual analytics can be deployed to bridge this gap.
Despite attempts by a number of researchers, Process mining lacks techniques to support automatic or intelligent redesign [2, 3, 18, 23]. Like in workflow technology, Process Mining is also limited to control flow, only pinpointing bottlenecks in current processes. While it can generate the various control flows from event-logs, these are too many resulting in spaghetti-like flows. The complexity associated spaghetti-like flows in healthcare could be better handled using VA.
Research on BPI involving VA is limited. Research that focused on use of VA in scheduling of facilities and patient flow mainly dwelled on processes that represented collecting and assembling information about room capacity, room use, patient-scheduling practices, staff capacity, and equipment availability . Though VA based, the approach does not guide or assist the researcher to redesign new more efficient processes, he needs to use his/her ingenuity. Being a methodological research, no tool was realised nor any validation carried out. In another research that focused on Control and diagnosis phases of the BPM cycle , the researcher argues that, in the two phases, Process mining and visual analytics individually do not provide sufficient process insight; rather, a combination of both approaches is required. Riemers recommends an integrated solution which also exploits colouring and ability to filter out events directly in the process model . The research scope excluded the design phase such that there is a gap between identification of improvement requirements and their actualization through design, to realize an improved process [12, 3, 24]. The research did not produce a tool. A tool to integrate VA and Process Mining will be able to close this gap.
Evident from Table 1, existing frameworks lack validated tool that are also VA based. There is therefore a need for a new validated framework to support business process improvement. VA techniques are visual and interactive, can help users handle complexity, interpret information, understand models faster among other advantages. Visualization can also be used to visualize modelling alternatives and to understand modelling results through better model-data linking and presentation. VA holds promise in solving the problems of redesign, and abstraction/complexity handling in an effort to bridge the gap in the BPM cycle that will result in process improvement.
While Process Mining can be used to discover, confirm and extend the processes within an organization, it is limited to the control and diagnosis phase, excluding the redesign phase of BPM cycle. The extension stage in Diagnosis phase, only adds analyzed information such as bottleneck and execution times of activities to the old discovered process model without modifying it or giving suggestions for alternative models. Process redesign for improved processes is carried out in the design phase. However, due to an information gap between the two phases, there is no direct or automatic input from diagnosis to design phase for purposes of redesign. Currently redesign is done intuitively and manually.
There is a need for a framework that can support business process improvement by automatically transferring the identified improvements from the diagnosis phase to the design phase of BPM cycle, as well as intelligently support redesign of better processes. Such framework can be validated by development and testing of an artefact in form of a prototype.
i) Establish and evaluate current BPI frameworks, business process practices in healthcare and Visual Analytics techniques.
i) What is the status of, and gaps in existing BPI frameworks and how can Visual Analytics be applied to address challenges in improving healthcare business processes?
This study has revealed an information gap between the diagnosis and design phases of the BPM cycle hindering automatic or intelligent process redesign that would result in process improvement. VA attributes can help in bridging the gap. However, studies in the area do not report an automatic or intelligent tool or framework to bridge this gap and those that have attempted are either not VA based or lack empirical testing. Limited research involving VA does not address redesign. This is a major shortcoming because current redesign is majorly manual depending on ingenuity as opposed to intelligence based on diagnosis from the processes being used.
- Propose a framework for healthcare Business Process Improvement (What constitutes a BPI framework in healthcare?)
- Develop a BPI prototype tool with Visual Analytics extensions (How can a tool with Visual Analytics extensions be used to implement a BPI framework in healthcare?)
- Evaluate the software prototype (How effective is the realised tool and hence the framework?)
This review is part of a continuing research that aims at developing and empirically evaluating a framework and tool for improving healthcare business processes using Visual analytics.
This work is supported by the National Commission for Science, Technology and Innovation ' Science, Technology and Innovation Grant of the Republic of Kenya.
 D.I. Sjoberg, T. Dyba, and M. Jorgensen, The future of empirical methods in software engineering research, in IEEE Conf. on Future of Software Engineering, 2007,358-378.
 J. A. Fitzgerald, and A. Dadich, Using Visual Analytics to improve hospital scheduling and patient flow, Journal of Theoretical and Applied Electronic Commerce Research, 4(2), 2009, 20-30.
 W.M.P. Aalst, M. Netjes and H.A. Reijers, Supporting the full BPM life-cycle using process mining and intelligent redesign, in K. Siau (Ed.), Contemporary Issues in Database Design and Information Systems Development, (Hershey- USA, IGI Global, 2007) 92-100.
 C. Avgerou, The significance of context in information systems and organizational change, Information Systems Journal, 11(1), 2001, 43-63.
 J. Vom Brocke, J. Recker and J. Mendling, Value-oriented process modeling: integrating financial perspectives into business process re-design, Business Process Management Journal, 16(2), 2010,333-356.
 M. Dumas, W.M.P. Aalst, and A.H.M. Hofstede (Eds), Process-Aware Information Systems; Bridging people and software through process technology, (Hoboken, New Jersey: Wiley & Sons, 2005)
 P. Green, and M. Rosemann, Perceived ontological weaknesses of process modeling techniques: further evidence, in 10th European Conference on Information Systems, Gdansk, 2002, 312-321.
 M. Jansen-Vullers, and M. Netjes, Business process simulation'a tool survey, in Workshop and Tutorial on Practical Use of Coloured Petri Nets and the CPN Tools, Aarhus, Denmark, 2006.
 R. Lu, and S. Sadiq, A survey of comparative business process modeling approaches, In Business Information Systems,(Springer Berlin Heidelberg, 2007) 82-94.
 Reijers, H. A., Jansen-Vullers, M. H., Muehlen, M., & Appl, W. (2007), Workflow Management Systems + Swarm Intelligence = Dynamic Task Assignment for Emergency Management Applications, in G. Alonso, P. Dadam, and M. Rosemann, (Ed.), BPM 2007, LNCS 4714 (Berlin Heidelberg, Springer-Verlag, 2007) 125'140.
 G.M. Nugteren, Process Model Simplification, MSC thesis, Eindhoven University of Technology, Netherlands, 2010.
 R.S. Mans, Workflow support for healthcare domain, PhD thesis, Eindhoven University of Technology, Netherlands, 2011.
 A. Agostini and G. De Michelis, A Light Workflow Management System Using Simple Process Models, Computer Supported Cooperative Work (CSCW), (9), 2000, 335'363.
 I. Vanderfeesten, H.A. Reijers, and W.M. van der Aalst, An evaluation of case handling systems for product based workflow design. In Proc of the 9th International Conference on Enterprise Information Systems (ICEIS 2007), Funchal, Madeira ' Portugal, 2007, 39-46.
 W. Janssen, H. Jonkers, and J. Verhoosel, What makes business processes special? An evaluation framework for modeling languages and tools in Business Process Redesign. In Proc. 2nd CAiSE/IFIP (Vol. 8). Barcelona Spain, 1997.
 M. H. Jansen-Vullers and H.A. Reijers, Business process redesign in healthcare: Towards a structured approach. INFOR-OTTAWA-, 43(4), 2005, 321.
 S.L. Mansar and H.A. Reijers, Best practices in business process redesign: validation of a redesign framework, Computers in Industry, 56(5), 2005, 457-471.
 Rozinat, A., de Medeiros, A., G??nther, C., Weijters, A. J. M. M., & van der Aalst, W. (2008). The need for a process mining evaluation framework in research and practice. In Business Process Management Workshops, Heidelberg, 2008, 84-89.
 R.S. Mans, M.H. Schonenberg, M. Song, V.D.W. Aalst, and P.J.M. Bakker, Application of process mining in healthcare'a case study in a Dutch hospital, Biomedical Engineering Systems and Technologies, 2009, 425-438.
 C. G??nther and W. van der Aalst, Fuzzy mining'adaptive process simplification based on multi-perspective metrics, Business Process Management, 2007, 328-343.
 P.T.G. Hornix, Performance Analysis of Business Processes through Process Mining, MSc thesis, Eindhoven University of Technology, Netherlands, 2007.
 B.V. Dongen and A. Adriansyah, Process mining: fuzzy clustering and performance visualization, in Business Process Management Workshops, Berlin Heidelberg, 2010, 158-169.
 P. Riemers, Process improvement in Healthcare: a data-based method using a combination of process mining and visual analytics, MSc thesis, Eindhoven University of Technology, Eindhoven, Netherlands, 2009.
 W. M. P. Van Der Aalst, Challenges in business process management: Verification of business processes using Petri nets, Bulletin of the EATCS, 80, 2003, 174-199.
 Keim, D. A., Kohlhammer, J., Ellis, G., & Mansmann, F. (Eds.). (2010). Mastering the information age-solving problems with visual analytics. Florian Mansmann.
 Holten, D. (2006). Hierarchical edge bundles: Visualization of adjacency relations in hierarchical data. Visualization and computer graphics, IEEE transactions on 12(5), 741 - 748
 van Wijk, J. J., & Van Selow, E. R. (1999). Cluster and calendar based visualization of time series data. In Information Visualization, 1999.(Info Vis' 99) Proceedings. 1999 IEEE Symposium on (pp. 4-9). IEEE.
 van der Aalst, W.M.P (2011). Process Mining: Discovery, Conformance and Enhancement of Business Processes. Springer Verlag, (ISBN 978-3-642-19344-6)
 Bertini, E. & Lalanne, D., (2009). Surveying the complementary role of automatic data analysis and visualization in knowledge discovery. In K. Puolam??ki (Eds), Proceedings of the ACM SIGKDD Workshop on VAKD (pp.12-20).
 Cook, K., Earnshaw, R. & Stasko, J. (2007). Discovering the unexpected, IEEE Computer Graphics and Applications, 27(5), 15-19.
 Wong, P. C., Rose, S. J., Chin Jr, G., Frincke, D. A., May, R., Posse, C., ... & Thomas, J. (2006). Walking the path: a new journey to explore and discover through visual analytics. Information Visualization, 5(4), 237-249.
 Bobrik, R., Reichert, M. & Bauer, T. (2007). View-based process visualization. In Alonso, G., Dadam, P. & Rosemann, M. (Eds): BPM 2007, LNCS 4714 (pp.125'140). Springer-Verlag Berlin Heidelberg.