SUPPORT FOR BUSINESS PROCESS REDESIGN

Transcription

1 SUPPORT FOR BUSINESS PROCESS REDESIGN SIMULATION, HYPERMEDIA AND THE WEB Anton Eliëns Frank Niessink Bastiaan Schönhage Jacco van Ossenbruggen Paul Nash(*) Vrije Universiteit, Department of Mathematics and Computer Science De Boelelaan 1081, 1081 HV Amsterdam, The Netherlands eliens, frankn, bastiaan, jrvosse, (*) University of Manchester KEYWORDS Business Process Redesign, Logistics-based Business Modelling, Discrete Event Simulation, Hypermedia, Active Documents, Web. ABSTRACT A critical success factor of business process redesign projects is the extent to which the organisation members themselves - both management and staff - are actively involved in the redesign project. Assessing alternative process designs with respect to the redesign project goals is an important aspect of such projects. One way to support the assessment of different redesign alternatives is to model these alternatives and subject them to a simulation study. Visualization and animation of the simulation models can provide insight into both the structure and the dynamics of the modelled alternatives. Effectively presented, visualization and animation enhance the involvement of the project members in the assessment and validation of the proposed redesigns. To support communication among all participants in the BPR project, we advocate an integrated presentation of redesign alternatives, visualized simulation experiments and the results of simulation studies in a single active document. In this paper we describe a framework enabling such integrated presentations by allowing the embedding of simulation models and experiments in hypermedia documents. These documents can be presented and executed on the Web. INTRODUCTION Business Process Redesign (BPR) is a method to perform organizational change. Instead of functions and departments it takes business processes as the main subject to redesign the way work is done in a particular organization (Davenport and Short 1990, Davenport 1993, Davenport and Nohria 1994, Hammer 1990, Wastell et al. 1994). Davenport and Short (Davenport and Short 1990) argue that business process redesign should be combined with the capabilities of Information Technology (IT). They show that IT can play two roles in BPR: on the one hand it is an enabler of new process structures. An example would be the use of expert systems to assist bank clerks handling loan requests. On the other hand IT can provide support for the implementation of new process designs. By, for example, simulating the logistic aspects of the redesigned loan request process. By modelling different redesign alternatives and subjecting them to simulation studies the participants in the business process redesign project can assess the alternatives with respect to the redesign project goals. Possible goals include cost reduction, lead time reduction, raising output quality and improving the quality of worklife. Visualization and animation of the simulation models can provide insight into both the structure and dynamics of the modelled alternatives and thus improve decision making. In this paper we advocate an integrated presentation of redesign alternatives, visualized simulation experiments and the results of simulation studies in a single active document. Our approach allows authors to write hypermedia documents with em-

2 bedded simulation models and experiments. These documents can be viewed on the Internet, or on a company Intranet, using an adapted Internet Web browser. Our system has been implemented using the hush (Eliëns 1995) hypermedia framework developed at the Vrije Universiteit. This framework provides support for multimedia user interfaces, a discrete event simulation library (Bolier and Eliëns 1994) and embedding of applications in the Web (van Doorn and Eliëns 1995, Eliëns et al. 1996). To model business processes we use Logisticsbased Business Modelling (LBM) (Gerrits 1995). The class library BPSIM (Business Process Simulation) facilitates the modelling of business processes using LBM and provides additional support for embedding and running simulations on the Web. Structure In the following sections we will introduce the concepts underlying logistics-based business modelling and we will describe the software support for simulating business processes, in particular the classes provided by the BPSIM library. We will present an example illustrating the construction of a business process model and its visualisation in a hypermedia environment. Then, we will briefly characterize the hypermedia environment and its relation to the World Wide Web. And, finally, we will discuss the merits and limitations of our approach. BUSINESS PROCESS MODELLING Dramatic improvements in business performance and productivity may be achieved by critically examining some of the rules that govern a business process. Business Process Redesign (BPR) is the generic label for many emerging methodologies aimed at producing these improvements. Redesign implies that the current state of affairs is no longer acceptable and can no longer be refined or evolved. The BPR perspective demands that its users understand the nature of the business they are in, and the processes that can be re-engineered. In order to achieve this understanding, part of the BPR approach is to map out an organisation s processes. The model created during this phase is used to communicate the BPR team s ideas and will also be used to plan changes to the current processes. Following Davenport and Short (Davenport and Short 1990) we define a business process as a set of logically related tasks performed to achieve some well-defined business outcome. It is important to note that business processes have customers that, somehow, receive the outcome of these processes. Also, to stress the functional nature of business processes, it may be remarked that business processes often transgress departmental boundaries within an organization, and even the boundaries between organizations. Despite the importance of qualitative assessments in BPR (Hammer 1990), for modelling we favor a more quantitative approach for which we provide support by means of a simulation library (BPSIM) based on the Logistics-based Business Modelling method (LBM) presented in (Gerrits 1995). Logistics-based modelling allows for analysing the time spent in executing a business process. The product lead time is defined as the time that passes between the moment a customer orders a product and the moment a product is delivered. In more detail, we can distinguish between processing time (the time actually worked on a job or operation), queue time (the time a jobs waits for a resource to become available), setup time (the time that passes between the moment a resource becomes available and the moment work on the job is started), wait time (the time that is spent waiting for another job to complete), and transport time (the time that is needed to move a job from a resource at a certain location to a resource at another location). For a particular model, measurements may be obtained by running a series of simulations. Based on an analysis of the simulation results alternative models may be proposed. For example, when the setup time for a job is relatively large, combining jobs into a single task for an employee may be more efficient. Logistics-based Business Modelling The LBM method offers a number of primitives, with associated graphic icons, from which a business process model may be constructed as a network of resources connected by transport arcs. The primitive entities offered by LBM are: operation processing component task a series of consecutive operations transport transport of information choice affects flow of information

3 Operations are atomic in the sense that wait time, queue time and transport time may not be part of an operation. Only setup time and process time are part of an operation. Tasks are introduced to allow for a series of jobs or operations to be processed, for example by one employee, in order to reduce the setup time needed. Transport entities represent the time it takes for information to flow from one resource (that is operation or task) to another. Transport implicitly defines the sequential structure of a process. However, duplications of information, and consequently parallel operations, are allowed. In addition to the primitives mentioned above, LBM allows us to characterize organizational units to represent departmental boundaries, external agents to represent opaque information producing or consuming entities, for example customers or other organizations involved in the process under examination, and archives to represent paper-based storage facilities. Also, employees may need additional means to engage in an operation or task. BUSINESS PROCESS SIMULATION The BPSIM library is an extension of the simulation library SIM (Bolier and Eliëns 1994). SIM is a C++ library offering classes supporting discrete event simulation, based on standard simulation techniques (Watkins 1993). In discrete event simulation, the components of the model consist of events, which are activated at certain points in time and in this way affect the overall state of the system. The simulation library consists of the following classes: simulation the scheduler event representing the events entity process consisting of events generator random distributions resource to model passive objects queue to hold waiting events histogram to plot the results analysis for statistical analyses The SIM library is integrated with the hush library, which may be used for defining a script interface to the simulation package, for developing a graphical user interface and for visualising simulation models. The library BPSIM The classes provided by the business process simulation library BPSIM reflect the entities provided by the logistics-based business modelling method LBM. On a somewhat more abstract level, a business process simulation consists of data, flowing through the process, and different handlers datahandlers performing some action on the data, such as transportation or specific operations. Accordingly, BPSIM provides two base classes underlying the classes corresponding to LBM entities: data which represents the product or case, i.e. the information, that flows through the process. datahandler which is the basis for all classes that handle information. Note that LBM contains no symbol for data as it only depicts the objects that handle the information. The following classes are derived from the datahandler class: operation, transport, waitqueue, choice, archive, and external agent, which realize the corresponding entities in LBM. An operation takes time, and is executed by an employee. Sometimes an operation results in more than one outgoing dataflow, for instance when it issues a request for additional information from a different department. A waitqueue functions as a regular queue if it has one incoming dataflow, i.e. transport. When there are more incoming flows, it functions as a synchronized queue. Data from one flow is not passed on to the next datahandler until the data from the other flow has arrived. This happens for instance when work on a case cannot continue until additional information has arrived. An external agent can either generate data according to some random number distribution, or take data, process it and pass it on to the next datahandler. In the latter case the agent functions as a black box: we only care how long the processing takes, not how it is exactly performed. In addition, BPSIM offers the classes means (which can be used to model resources that are necessary to perform certain operations), and employee

4 (which models the different people that perform the operations). As one can see, the entities task and organization unit from LBM have no specific counterpart in BPSIM. The reason for this is that having no other datahandler between two operations already implies those operations belong to the same task. Consequently they are executed with no time in between and by the same employee. Also, the fact that tasks are executed in different organization units does not add any information that changes the behaviour of the simulation. If it takes time to transport information between different units, then that time can be represented by the object of class transport between those units. For the gathering and analyzing of results, the SIM classes histogram and analysis are used. The class agent, for example, can be given a histogram to track the lead time of data it has generated. A script interface for BPSIM Employing the facilities of the hush library a script interface has been defined for BPSIM that allows the user, that is the designer of business models, to construct and run business simulation models with a short turn around time. Each class in the BPSIM library corresponds to a command in the script language. As illustrated in the next section, script commands result in a graphic representation of the model, which may be displayed and executed in a Web page as an applet. REQUESTS FOR LOANS Now let us take a look at an example business process model based on the objects made available in BPSIM. The example will detail possible situations before and after business process redesign has taken place. Current Situation: Our first model, which represents the current situation, consists of a client producing, for example, requests for a loan at a bank. The requests are initially handled by a clerk whose task is to send requests for amounts greater than 10,000 dollars to the boss. The boss, whose task is to approve or deny the loan, sends the result back to the clerk for processing. If the amount is less than 10,000 dollars, the clerk has authority to process the request himself. The current situation is unsatisfactory because the boss has very little time available for the task of approving the loans. This results in a high lead time for the loan requests as can be seen in Figure 1. The goal of the redesign project will be to lower the lead time substantially. Redesign Alternative: In our second model, which is the result of a redesign effort, requests are handled by a clerk who enters the data into a computer. The computer now makes the decision as to whether the loan is approved - for loans of value less than and greater than 10,000 dollars. By using an expert system we now no longer need the boss to approve requests for large loans. Figure 1 shows the visual representation of the current situation while Figure 2 shows the model resulting from the redesign effort, both embedded in a Web page. In addition to the model, both pages contain some results obtained by running a simulation. The histogram depicts the distribution of the lead times (in minutes) of incoming loan requests, that is the time that passes between receiving a request and giving an answer. The Web pages further contain buttons to control the simulation run, a description of the model and links to alternative models. The source for the Web pages looks as follows:... <h1>request for loan - Current Situation</h1> <p>in this model, a client makes requests for loans to a bank.... <hr> <applet class=alternative1> The applet tag specifies by means of its attribute class where the source for the applet, i.e. the script, is to be found. The Web browser then fetches the source and executes it, resulting in the display of the model as in Figures 1 and 2. The script Using the script interface to the BPSIM classes we built models of the current situation and of redesign alternatives. In general, the scripts consist of two parts, one part in which the objects employees, operations, waitqueues, resources are created and one part in which the process flow is specified. Objects can be created by means of the following script command:

5 Figure 1: Presenting the current situation class instance name parameters The process flow can be specified using the format: object -next object(s) First we take a look at the script for the current situation and then we examine what has to be changed to model the redesign alternative. The part of the script that models the current situation looks as follows: histogram leadtimes agent client -generate yes -histogram leadtimes -duration normal In the above fragment we have created the histogram for recording lead times and an agent representing the clients of the process which generates requests for loans. By providing the histogram leadtimes as an parameter to agent we specify that the agent should register the lead times of the requests it generates into the histogram. In addition, we have specified that the interval times between generated requests should be distributed according to a normal distribution with and, i.e. with mean 60 minutes and standard deviation 50 minutes. Next, we create some other needed objects. We need two employees: employee clerk employee boss Furthermore, we need to model four different operations. One operation which handles the incoming requests, one operation for handling small loans, one operation for handling large loans and one operation that passes the result of the process back to the client. operation receive -employee clerk -duration normal operation large loan -employee boss -duration normal 100.0

6 Figure 2: Presenting a redesign alternative 50.0 operation small loan -employee clerk -duration 20.0 operation finalise -employee clerk -duration normal To model the different flows of requests for large loans and for small loans, we use choice. The -fraction parameter of the choice specifies that 50% of the requests will concern small loans and 50% will be large loans: choice small or large amount -fraction 0.5 We also need four queues to model the fact that requests sometimes have to wait before they can be processed: queue receive queue queue small loan queue queue large loan queue queue finalise queue Furthermore, we need transport objects to model the fact that it takes time to bring the loan requests from one employee to the next. transport t1 -duration 20.0 transport t2 -duration 20.0 transport t3 -duration 20.0 transport t4 -duration 20.0 After the necessary objects for the model have been created the only thing left to do is to specify the order in which the objects are connected. The next fragment shows just that: client -next t1 t1 -next receive queue receive queue -next receive receive -next small or large amount

7 small or large amount -next large loan queue small loan queue large loan queue -next large loan small loan queue -next small loan large loan -next t2 t2 -next finalise queue small loan -next t3 t3 -next finalise queue finalise queue -next finalise finalise -next t4 t4 -next client Having built a model of the current situation we now turn to the redesign alternative. In this alternative situation we no longer need the boss because the clerk can now decide about both small and large loans with the support of a computer system. This means we need to incorporate the computer system in our model, and that we have to adjust the receive operation to use the computer: means computer operation receive -employee clerk -means computer -duration normal Further, we can delete the operations small loan and large loan, their queues, the choice and the boss from the script. The only thing left to do is to specify the new model structure: client -next t1 t1 -next receive queue receive queue -next receive receive -next t2 t2 -next finalise queue finalise queue -next finalise finalise -next t3 t3 -next client From Figure 2 we can see that this redesign alternative shortens the lead time of loan requests substantially, thus satisfying our redesign goal. THE ENVIRONMENT Given the importance of participant communication in a business process redesign effort, embedding business simulation models in hypermedia provides in a natural way the information context needed for exploring alternatives and taking decisions. With Intranets becoming more popular, the World Wide Web has proven to be a popular access point for many applications. The hush library and its extensions offer a rich environment for developing hypermedia applications and dynamic Web documents (van Doorn and Eliëns 1995). The hush library originated as a C++ interface to the GUI scripting environment provided by Tcl/Tk (Eliëns 1995). Tcl/Tk offers an extensible script language with powerful graphics and window programming utilities. A number of extensions to hush have been written, including the SIM and BPSIM libraries as well as libraries encapsulating multimedia facilities. The hush Web component provides a collection of (inline) data viewers for the hush Web browser. It supports full SGML and style sheets to specify dynamic layout properties in a declarative way (Eliëns et al. 1996). Also applets are supported, which are Tcl/Tk scripts augmented with the functionality of hush applications. Applets are the building blocks of active Web documents. By defining a script interface for applicationspecific C++ classes, application functionality may be embedded in a Web page as an applet. As illustrated in the example, our approach to providing hypermedia and Web support employs this feature, both to reduce the time involved in modelling and to present a graphical representation of the model and its execution to the user. CONCLUSIONS As observed by Wastell (Wastell et al. 1994) organisational change is a highly threatening and stressful experience for many participants and that high levels of stress can have a pernicious effect on individuals, group processes and organizational learning. Hence, directly involving the users in the modelling phase of the BPR project may be an important step towards capturing the human aspects that are necessary for the production of an optimal model. It may also help to lessen the anxiety of employees. Our contribution in this area is to provide support for an information context that encompasses both technical and social dimensions of the reengineering effort. On a technological level, we advocate the use of business process simulation. Nevertheless, to accomodate the social aspects, we support the visualisation of such models and their integration in an arbitrarily complex information context, such as the World Wide Web. As concerns the limitations of our approach, we wish to remark that we have only realized part of the visualisation and animation facilities that we consider desirable. Furthermore, in our cur-

8 rent realization the user is required to use the hush Web browser, instead of a popular browser such as Netscape. REFERENCES D. Bolier and A. Eliëns. Sim a C++ simulation library. Technical Report IR-367, Vrije Universiteit, T.H. Davenport and N. Nohria. Case Management and the Integration of Labor. Sloan Management Review, Winter T.H. Davenport and J.E. Short. The New Industrial Engineering: Information Technology and Business Process Redesign. Sloan Management Review, Summer T.H. Davenport. Process Innovation, Reengineering Work through Information Technology. Harvard Business School Press, A. Eliëns, J.R. van Ossenbruggen, and S.P.C. Schönhage. Animating the Web an SGMLbased approach. In Proc. Int. Conf. on 3D and Multimedia on the Internet, WWW and Network, Bradford, April Britisch Computer Society. ABOUT THE AUTHORS Anton Eliëns is Assistent Professor at the Software Engineering section of the Computer Science Department of the Vrije Universiteit, Amsterdam. He is coordinator and leader of the DejaVu project. His research interests include object orientation, hypermedia and distributed logic programming. Frank Niessink is PhD. student at the Software Engineering section of the VU. His research interests include business process modelling and redesign, software metrics and maintenance. Jacco van Ossenbruggen and Bastiaan Schönhage are both PhD. students at the Software Engineering section of the VU. They both participate in the DejaVu project. Jacco is doing research on time-based hypermedia, his research interests cover SGML, HyTime and the formalization of hypermedia models. Bastiaan is currently working at the development of high-level components for Web-aware applications. His research interests include simulation and 3D animation. Paul Nash is a MSc. student from the University of Manchester. He participated in the project during a visit funded by Erasmus. A. Eliëns. Hush a C++ API for Tcl/Tk. The X Resource, 14: , April J.W.M. Gerrits. Towards Information Logistics An Exploratory Study of Logistics in Information Production. Ph.D. thesis, Vrije Universiteit, M. Hammer. Reengineering Work: Don t Automate, Obliterate. Harvard Business Review, July- August M. van Doorn and A. Eliëns. Integrating WWW and Applications. In Third Int. World Wide Web Conf. Technology, Tools and Applications, Darmstadt, April D.G. Wastell, P. White, and P. Kawalek. A methodology for business process redesign: experiences and issues. Journal of Strategic Information Systems, 3(1):23 40, K. Watkins. Discrete event simulation in C. McGraw-Hill, 1993.