page 1  (10 pages)
2to next section

A Dynamic Modeling Toolkit to Add Rigor to Business Process Re-engineering

James R. Warren Paul J. MacArthur Robert L. Crosslin

School of Computer & Info. Sci.

University of South Australia

Dept. of CSIS

The American University

Dept. of CSIS

The American University

Abstract Business Process Re-engineering (BPR) is concerned with the identification and improvement of areas in a business where maximum productivity gains can be achieved. Davenport and Short [2] propose a five-step approach to redesigning business processes.

Business Process Re-engineering (BPR) can lead to huge savings and may be essential to survival in a competitive environment; but decisions to redesign business processes are often based upon little more than heuristics and gut feelings. Dynamic modeling techniques add rigor to BPR, however, off-the-shelf simulation software is not generally well-tailored to the BPR approach. We have developed a Dynamic Modeling Toolkit (DMT) utilizing pre- and post- processors to customize CACI?s SIMprocess as a BPR tool. A front-end data repository supports management of a large BPR analysis and aids formulation of SIMprocess models of business operations. A back-end processor combines results from simulations of system components into a hierarchy, and supports a ?black box? approach where component models can be replaced by alternative designs to assess the impact on the overall system. The DMT provides a comprehensive evaluation of system performance, yielding objective and rigorous economic justification for redesign.

1. Develop business vision and process objectives. 2. Identify processes to be redesigned (identify critical or bottleneck processes).
3. Understand and measure existing processes (identify current problems and set baseline).
4. Identify IT ?levers? (brainstorm new processes and approaches).
5. Design and build a prototype of the process. Traditionally, the decision to redesign business processes has been based on the recommendations of consultants and ?gut feeling.? Similarly, the thrust of the redesign effort is often driven by a collection of success factors and best business practices (heuristics) with little or no rigorous modeling to predict the outcome of the reengineering. Certainly, there have been stunning successes (e.g., Ford as reported by Hammer [4]), but the aggregate productivity numbers suggest that the wellpublicized successes are the exception and not the rule.

1. Introduction Computer simulation is well-suited to identifying processing bottlenecks, understanding the dynamics of existing systems, and predicting the performance of new processes -- activities which correspond to steps 2 through 4 of Davenport and Short?s BPR model. Simulation can provide the rigor to remove some of the guesswork from BPR and introduces an element of objectivity. In a complex business system there are many interdependent components, and many aspects of the environment to which the system may be sensitive. Tampering with one subsystem may have unexpected impacts on the business as a whole. Simulation is appropriate for tracking complex interactions and avoiding optimization of one component at the expense of overall system productivity.

Information technology (IT) has not been a panacea for US businesses despite their hearty investment in information systems. Consider these reports of marginal gains:
? In the service sector in the last decade, IT investment per white-collar worker has more than doubled to $12,000 per annum while their productivity has shown only a meager increase of approximately 0.5% per annum [8].
? In 1987 Business Week reported that almost 40% of all US capital spending went to IS, some $97 billion a year. Despite this, few firms have achieved major productivity gains; aggregate productivity figures for the US have shown no increase since 1973 [2].

In today?s environment of fierce international competition, achieving significant productivity gains is not just a desirable goal for organizations, but can be a matter of survival. Pressure from Japanese ?keiretsu? or The figures suggest that it is not sufficient to throw IT at problems if one expects real productivity gains.

business cartels provides an excellent example. The keiretsu specialize in exploiting and dominating the open market. Some markets that were once US dominated and are now falling (or have fallen) to the keiretsu include: popular media, machine tools and robotics, electronics, computer memory chips and semiconductors, automotive parts, banking and finance [7]. The keiretsu typically proceed to control and then own their foreign competitors, including, for instance: 7/11 stores, Dunlop, Columbia Pictures, and Epson.

modeling of information systems and their surrounding organizations; and,
3. Development of a custom toolkit around an existing simulation environment: To garner the benefits of a toolkit tailored to the BPR domain, without ?reinventing the wheel? and developing a new simulation environment, an existing simulation can be integrated with customized front-end and back-end tools.
Let us consider each of these alternatives in turn. Thus, we suggest that: There is a compelling logic to using existing generalpurpose simulation tools for BPR. Simulation languages such as GPSS and SIMSCRIPT were designed to support performance modeling to aid in the design of logistic and organizational systems. A standard introductory textbook on GPSS [9] gives case-study models of: ships unloading and loading at harbor (p. 211), a television assembly quality control system (p. 223), and a hospital emergency room (p. 258). Although potentially powerful, simulation languages in their raw form have many undesirable characteristics for BPR:

? BPR is a vitally important activity for today?s businesses; ? BPR is frequently practiced with minimal rigor, relying on ?gut feelings? and highly-subjective judgments; and,
? Simulation is an appropriate tool to aid in quantifying the value of design alternative, thus adding rigor to the BPR process.
Given this state of affairs, there is an apparent priority to explore the best tools and methods to bring simulation technology into the BPR process. In the next section, we explore three alternative approaches to applying simulation technology to BPR: (a) use of an existing simulation environment, (b) development of a custom environment, and (c) development of a custom toolkit around an existing simulation environment. Subsequently, we explore the third of these options and describe the architecture of the Dynamic Modeling Toolkit (DMT), which incorporates CACI?s SIMprocess as the simulation engine. The fourth and fifth sections describe two major components of the DMT: (a) a front-end data collection tool; and (b) SIMtree1, a back-end processor to SIMprocess that aids in managing model hierarchies. The conclusions discuss the merits of simulation in BPR and of the DMT approach in particular.

1. They require significant amounts of programming to model complex real-world systems;
2. The special programming language must be mastered by the BPR analysts (an additional, highlyspecialized demand on individuals who are already expected to have exceptional mastery of communications and analysis skills, as well as business practices); and,
3. The simulation programs require significant effort to verify, validate, and maintain, especially in light of a complex and dynamic business analysis.
Thus, simulation languages are not well-suited to the pace and scope of BPR, or to the skills of its practitioners.

Many shortcomings of simulation languages are rectified by simulation environments, which frequently possess a common simulation language at their core, but provide more user-friendly (usually graphical) interfaces to the capabilities. SIMprocess by CACI, for example, is developed in SIMSCRIPT, and supports the simulation of easily-drawn diagrammatic models. SMC?s ARENA, an environment built around the SIMAN language, allows the specification of objects of arbitrary complexity, which can then interact to form a performance model. These environments provide benefits beyond ease of specification (such as sophisticated animation capabilities), and offer the analyst the advantage of utilizing the same leading-edge tools as found in other disciplines such as engineering and computer-aided design (CAD).

2. Applying simulation technology to BPR

Given the desire to incorporate simulation technology into the BPR process, we perceive three distinct possible approaches:
1. Use of an existing simulation environment: General-purpose simulation languages have been around since the 1960?s, and environments or ?shells? around simulation engines have become increasingly sophisticated;
2. Development of a custom environment: Numerous efforts have been undertaken to development simulation environment tailored to the dynamic

Simulation environments, despite their ease of use and power, still have significant disadvantages for BPR. These disadvantages stem from the general-purpose na1SIMtree is a proprietary product of Technology Economics, Inc. (TEI). This article is produced with the explicit consent of TEI.

ture of the environments. ARENA is non-trivial to use, requiring specification of code for each of the objects. ARENA does not (yet) offer a library of pre-coded objects common in BPR, and thus the analyst is left to do significant coding to implement a BPR analysis. SIMprocess allows models to be specified more easily, but it does little to support the analyst in managing the complexity of a real-world organization, which would typically entail dozens of SIMprocess diagrams. In each of these environments, once the model has been coded (albeit, ?coding? may be through a graphical interface), the representation is not easily queried to examine its correspondence to the analyst?s intentions. That is, it is nontrivial to determine that a SIMprocess or ARENA model is an accurate representation of the dynamics of a real or proposed business process.

erate Siman/Cinema code from them, is under development.

We have developed a Dynamic Modeling Toolkit (DMT) using CACI?s SIMprocess as the simulation engine. Particular strengths of SIMprocess include: 1. An easy-to-use diagrammatic notation, which makes it relatively easy for an IS analyst to put together performance models of operational subsystems. No programming is involved.
2. An attractive presentation that allows the customization of icons and supports animation. These are features important to validating simulation models and gaining the confidence of end-users and managers.
We also perceive, however, two weaknesses in SIM- process as a stand-alone tool for BPR:
An alternative to making do with general-purpose environments is to develop custom environments tailored to the specific needs of BPR. Numerous custom environments have been developed, and many offer powerful features for BPR. Dur and Bots [3] present a graphical environment for dynamic modeling of organizations based on task/actor simulation. Their environment includes editors for tasks and actors, as well as a custom simulation engine. Jordan and Evans [6] utilize a custom simulation language, SIMIAN, to simulate IS strategy. Warren et al. [11, 12] describe a custom system simulator based on the popular data flow diagram (DFD) notation. Each of these environments is specialized for the modeling of business processes.

1. SIMprocess is an inadequate front-end for BPR data. SIMprocess models are not a convenient or complete description of a business. While activities and their required resources are represented, the totality of a business? personnel and equipment is difficult to extract from SIMprocess diagrams. Also, the transcription of field observations of business processes directly into SIMprocess models is non-trivial; an intermediate representation that organizes business resources is in order.
2. SIMprocess fails to support the hierarchical decomposition of business processes. In the analysis of business processes, it is typical to use an iterative decomposition process to produce a hierarchy of views ranging from the most high-level overall view of the system on through to highly detailed views of specific sub-system. The hierarchical decomposition processes continues until a level of detail has been reached that is sufficient for the analysis. Hierarchical decomposition is important in the analysis of complex systems because the human mind can only grasp so much detail at one time. The union of all the lowest level views would show the system in full detail, but in a large system this amount of detail presented all at once would present a daunting web of complexity. With hierarchical decomposition, the analyst can start with ?the big picture? (i.e., the highest level view) and proceed into further detail with whatever subcomponent is of interest at that time. This method allows one to manage a system even if it is too complex to apprehend its entire structure simultaneously.

A problem with completely custom environments is that they run counter to the basic software engineering principle of reuse of code. Many advanced features that have been implemented in general-purpose simulation environments, such as graphical interfaces and animation, must be re-implemented for custom environments. Development of a BPR toolkit based on existing simulation technology is an efficient strategy that strikes a compromise between making due with a general-purpose simulation environment that is not entirely appropriate for BPR, and developing new tools from scratch. Examples of this approach include the SASOS system [5] which utilizes the simulation capabilities of Design/CPN2, but uses a custom application developed in Apple Computer?s Hypercard 2 for its business information repository. Streng and Sol [10] present an approach where inter-organizational dynamics are represented in terms of Layered Actors, Networks, and Entities (LANE); the LANE representation can then be simulated (and animated) using SMC?s Siman/Cinema. A tool to manage LANE representations, and automatically gen-

To address the weaknesses of SIMprocess for BPR we have developed a Dynamic Modeling Toolkit (DMT) around SIMprocess. The DMT has two major components beyond SIMprocess, one to address each of the 2Design/CPN is a proprietary product of MetaSoftware Corporation.

Figure 1. Schematic of the Dynamic Modeling Toolkit (DMT) as a decision support tool for Business Process Re-engineering.

problems listed above. The next section describes the DMT.
are then developed until a desired level of detail has been reached. As the BPR analysis progresses, alternative models of one or more of the subsystems may be developed. 3. Architecture of the Dynamic Modeling Toolkit (DMT) SIMtree is used to integrate the SIMprocess simulation results from a set of diagrams. With SIMtree, performance statistics can be derived for models appearing anywhere in the hierarchy (i.e., for the context diagram, or one of its subsystems, or a subsystem?s subsystem, etc.). SIMtree supports the substitution of one or more alternative models into the hierarchy. SIMtree is described in section 5.

Figure 1 shows a schematic overview of the Dynamic Modeling Toolkit (DMT) as a decision support tool for business process re-engineering. The DMT is designed to support the quantitative assessment of design alternatives.

The first step in using the DMT is to capture the structure and dynamics of the business as it operates in the real world. Capturing the business data requires standard systems analysis practices such as observing the business processes, interviewing the users, and reviewing forms, memos, files, etc. The front-end data repository is used to store the captured information on organization structure, technology (i.e., equipment), activities, human resources. The repository is discussed in the next section.

4. Front-end data repository for BPR

Notes from field observations made during a BPR analysis are not in a form readily amenable to designing an improved business system design. In BPR, even to a greater extent than in traditional IS development, the fundamental objective is to perceive opportunities to reapportion resources and achieve a more productive system. To make this analysis the BPR analysts must have at their disposal a systematically organized and integrated collection of information on the equipment,

SIMprocess models are formulated based on the data in the repository. A top-level, or context, model is developed to show the whole system in its environment. Models representing decompositions of the subsystems

human resources, and activities
performed within a business. SALES_SYSTEM

T_NEG_DEAL_&_ T_SHIP_GOODS T_ACCT_

TMETAMOD

T1 T2 T3 T4

T_CUSOMER_

T41 T42

T_NEGATIVE

T421 T422

PLACE_O SETTLEMENT T_CUSTOMER_
SATISFACT

T_INITIAL_
CONTACT FEEDBACK

FEEDBACK
T_POSITIVE
FEEDBACK

* * *

*

* *

Figure 2. Hierarchy tree for the proposed sales system.

A front-end data repository
for BPR is being developed in
Toolbook. Information stacks
are maintained on each type of
business entity and links
between entities of different
types are supported. For
example, if an analyst is
viewing the record on a specific
activity (e.g., intake of a customer
order), then the analyst can
?click? a button and have
access to the records on the
human resources and/or equipment
involved in the activity.
From there, the analyst could
traverse, for example, to the
records of the other activities
that use the same minicomputer
as the customer order intake
process.
The repository is intended to

SIMtree improves the support for hierarchical decomposition of processes by allowing the simulation results from multiple SIMprocess models to be combined in an explicitly defined hierarchy tree. In particular, simulation results for ?parent? models are computed based on the results from their ?child? models in a hierarchy. At the ?leaves? of the hierarchy tree we have models that have been simulated with SIMprocess. SIMtree can produce reports on the performance of stations, queues, products, and resources at any level of the hierarchical decomposition.

act as a support for all aspects of the redesign effort. For example, analysts can peruse the database in an unstructured manner to gain the insight into the operation necessary to identify possible redesign opportunities. Also, later in the BPR effort, human resources and equipment can be matched to the activities in the redesigned system. The intended use of the repository, however, is to provide the information needed to formulate SIMprocess simulation models.

5. A modeling hierarchy manager: SIMtree

Figure 2 shows a schematic of a simple hierarchically decomposed process. A proposed sales system is divided into four components: negotiating the deal and placing the order (T1), shipping goods (T2), account settlement (T3), and customer satisfaction (T4)3. Figure 3 shows the SIMprocess diagram for the context view of the sales system (TMETAMOD). The customer satisfaction subsystem is further decomposed into the initial contact component (T41) and the customer feedback component (T42). Customer feedback decomposes to negative feedback (T421) and positive feedback (T422) subcomponents. Figures 4, 5 and 6 show the SIMprocess diagrams for the customer satisfaction subsystem, its customer feedback component, and the negative feedback subcomponent, respectively.

The purpose of SIMtree is to improve SIMprocess? ability to support hierarchical decomposition of processes. SIMprocess has the ability to chain a set of models together such that the simulation results from one model act as the input to the next; but this represents only a ?horizontal? view of the lowest (i.e., most detailed) level of the organization. This horizontal view runs across the traditional vertical view of organizations divided into functional departments. For BPR analysis, it is important to be able to combine the horizontal and vertical views into a two-dimensional view of the organization that depicts the organizational hierarchy and the relationships among the branches of the hierarchy. A view that models only the lowest level fails to explicitly address the performance of the larger components of the business. 3The ?T? prefix of the model names is denoting a design that is ?To be,? i.e., proposed as compared to existing.

Figure 3. SIMprocess diagram for the proposed sales system (TMETAMOD).
Figure 4. SIMprocess diagram for the Customer Satisfaction subsystem (T4) of the proposed sales system (decomposition of the

T_CUSTOMER_SATISFACT station in TMETAMOD).

Figure 5. SIMprocess diagram for the Customer Feedback component (T42) of the Customer Satisfaction subsystem (decomposition of the

T_CUSTOMER_FEEDBACK station in T4).

Figure 6. SIMprocess diagram for the Negative Feedback subcomponent (T421) of the Customer Feedback component (decomposition of the T_NEGATIVE_FEEDBACK station in T42). This diagram corresponds to the leaf node at level 3 of the hierarchy tree in figure 2.

A hierarchy of processes is mapped into SIMprocess following these guidelines:
3. Each non-leaf node has one child node for each station in its SIMprocess model. So, for example, TMETAMOD has four stations in SIMprocess and four children in the hierarchy.
1. A SIMprocess model is created for each node in the hierarchy tree.
2. Each node in the hierarchy is either a leaf or a nonleaf node (the leaf nodes in figure 2 are marked with asterisks in their upper-left corners).

4. SIMprocess simulates the performance of the leaf nodes (and only the leaf nodes).

5. SIMprocess? ?chaining? facility is used
to combine the horizontal simulation
results among the leaf nodes.

Figure 7. SIMtree's hierarchy specification screen.

6. SIMtree is used synthesize results for
the non-leaf nodes from the SIMprocess
results for leaf nodes.
Notice that branches of the hierarchy tree
need not all be decomposed to the same level of detail. In figure 2, for example, leaf nodes are found at levels 1, 2, and 3 of the hierarchy. This uneven level of analysis, focusing on customer satisfaction in our example, could reflect some processes being in reality more complex than others, or could reflect a BPR analysis that is centering on the customer satisfaction subsystem.
Figure 7 shows SIMtree?s hierarchy specification
screen. On the left are a list of all An array of synthesis rules is employed to formulate the aggregate (i.e., summary) statistics for each model. A model?s aggregate statistics (as shown in figure 8) are the numbers passed up to its parent model in order to synthesize the parent?s statistics. For example, the Station Units Processed statistic for T_NEGATIVE_FEEDBACK (440 units) in T42 is the aggregate of Station Units Processed from model T422. The aggregate Station Units Processed statistic is defined as the units processed of the model?s designated final product. Aggregate busy and idle percentages of human resources are computed as the average over all resources in the model. Some statistics require consideration of the topology. For example, if there is a branch in a model?s process flow, leading to multiple final products, then only those stations along a path leading to a given final product are considered in assessing the average processing time of that product. For each model (leaf or non-leaf) unit cost statistics are computed as the number of final products produced divided by the total resource costs -- this encourages objective evaluation of ?the bottom line? on system performance.

SIMprocess models in the project directory. The analyst chooses a model from the list as the context diagram. SIMtree then analyzes the structure of this top-level model and, for each station in the model, prompts the analyst to provide a SIMprocess diagram that specifies that station in more detail. The specification procedure is repeated for each of the subprocesses until the analyst indicates that the diagram is at the leaf (i.e., lowest) level of the hierarchy.

SIMtree supports a ?black box? approach wherein pieces of the business process hierarchy can be replaced with alternative designs and the effect on overall system performance evaluated. SIMtree allows any number of hierarchies to be created from a set of SIMprocess diagrams. Hierarchies can be developed which differ only on one specific model, or major branches of the hierarchy may differ. For instance, if we wish to consider the effect of an alternative model for account settlement (say, model T3B), we create a second hierarchy identical to figure 2 except that model T3B is substituted for T3. Performance statistics could then be developed for each of the design alternatives to determine the most effective design in terms of overall performance. 6. Conclusions Figure 8 shows output from SIMtree for model T421 (a leaf node of the hierarchy). For leaf models, SIMtree uses the performance statistics generated by SIMprocess directly to formulate a performance summary. For nonleaf models, SIMtree synthesizes performance statistics from the performance statistics of the model?s children. This synthesis process begins at the leaf nodes of the hierarchy tree and works its way up to the context level diagram. Figure 9 shows the SIMtree output for model T42 that is derived based on its two child models, T421 and T422.

Today?s businesses are facing tough competition and must, as a matter of survival, exploit all opportunities for increased efficiency and quality of goods or services. Business process re-engineering (BPR) is, virtually by definition, the process of re-structuring activities (particularly using new technologies) to face-up to these challenges. Three key activities of BPR are: (a) identifying critical or bottleneck processes to redesign, (b) understanding and measuring the existing processes, and (c) devising new processes and approaches. While these activities can be done in large part on the basis of intu-

ition and experience with good business practices, there exists a tremendous opportunity to introduce more rigor into the process via computer simulation. Blattberg and Hoch [1] found that the combination of decision models and managerial intuition led to significantly more accurate decisions than either source alone; there was benefit to the manager/model combination even when models were severely degraded by the intentional deletion of decision variables.

ance. Simulation helps to prevent optimization of subprocesses at the expense of overall performance by predicting the global effects of local changes.

We have developed a Dynamic Modeling Toolkit (DMT) using CACI?s SIMprocess as the simulation engine. A front-end data repository collects and organizes business information prior to preparation of SIMprocess models, and SIMtree, a modeling hierarchy manager, supports the synthesis of performance statistics at a relevant level of detail from the SIMprocess output. SIMprocess possesses excellent usability through its graphical, no-programming interface, offers animation, and produces a variety of reports readable by SIMtree. With the addition of back-end and front-end software tools to SIMprocess, the DMT is able to organize business process data, simulate the performance of existing and alternative designs, and present system performance

Dynamic modeling with computer simulation has many desirable characteristics as a tool to add rigor to BPR. Simulation provides objective and quantitative measurement of production levels, operating costs, utilizations, and response times for both the existing (base-line) system and for any alternative designs. Also, changes in one subprocess of a complex dynamic systems can have unforeseen impacts on overall system perform-

DIAGRAM T421: 07/14/93 16:34:03

UNITS
STATION PROCESSED BUSY(%) IDLE(%) ------------------------------------------------------------------------------------------------------------------------ T_COMPLAINT_PROCESSE 449 22.4 77.6 T_PROBLEMS_DETAILED 448 33.7 66.3 T_PROBLEMS_RESOLVED 448 43.1 56.9 ------------------------------------------------------------------------------------------------------------------------ AGGREGATES 448 33.1 66.9

UNITS % FOR TOTAL % FOR OPERATIONS STATION PROCESSED STATION TIME STATION ------------------------------------------------------------------------------------------------------------------------ OP_COMP_PROC T_COMPLAINT_ 449 100.0 224 100.0 OP_PROBS_DET T_PROBLEMS_D 448 100.0 337 100.0 OP_PROBLEMS_ T_PROBLEMS_R 448 100.0 431 100.0 ------------------------------------------------------------------------------------------------------------------------ AGGREGATES 1345 992

UNITS MIN Q MAX Q AVG Q STDDEV MIN Q MAX Q AVG Q STDDEV Q BUFFERS PROCESSED BUSY(%) IDLE(%) DELAY DELAY DELAY DELAY LEVEL LEVEL LEVEL LEVEL ------------------------------------------------------------------------------------------------------------------------ Q_PROBLEMS_D 449 0.0 100.0 0.0 0.0 0.0 0.0 0 1 0.0 0.0 Q_PROBLEMS_R 448 0.0 100.0 0.0 0.0 0.0 0.0 0 1 0.0 0.0 R_CUST_CMPLT 450 98.9 1.0 0.0 0.0 0.0 0.0 0 451 222.9 0.0 ------------------------------------------------------------------------------------------------------------------------ AGGREGATES 450 98.9 1.0 0.0 0.0 0.0 0.0 0 451 222.9 0.0

FINAL OUTPUT ------- PROCESS TIME/UNIT -------- ------- UNIT LEVEL ----------- PRODUCTS PROD UNITS BUSY(%) IDLE(%) MIN MAX AVG STDDEV MIN MAX AVG STDDEV ------------------------------------------------------------------------------------------------------------------- COMPLAINT 448 100.0 0.0 0.7 0.9 0.8 0.0 0 1 0.3 0.0 DETAILED_PROB 448 100.0 0.0 0.8 1.1 1.0 0.0 0 1 0.4 0.0 FEEDBACK * 448 N/A N/A 2.2 4.2 3.9 0.1 N/A N/A N/A N/A FORWARDED_INFO 449 0.1 99.9 0.5 2.4 2.2 0.1 0 452 223.9 0.0 ------------------------------------------------------------------------------------------------------------------- AGGREGATES 448 45.0 55.0 2.2 4.2 3.9 0.1 0 454 224.6 0.0

OPERATING COSTS: HUMAN RESOURCES BUSY(%) IDLE(%) BUSY($) IDLE($) TOTAL($) UNITS PER UNIT TOTAL ------------------------------------------------------------------------------ ------------------------------- CLERK521 99.3 0.7 14,888.58 0.00 14,888.58 448 114.58 51,333.56 CLERK521A 76.9 23.1 12,296.83 0.00 12,296.83 CLERK521B 43.1 56.9 6,471.45 0.00 6,471.45 SUPER521 76.9 23.1 17,676.70 0.00 17,676.70 ------------------------------------------------------------------------------ TOTAL H.R. COSTS 51,333.56 0.00 51,333.56 AVG H.R. COSTS 74.1 26.0 12,833.39 0.00 12,833.39

Figure 8. SIMtree report for the Negative Feedback model (T421).

statistics at a high or detailed level. SIMtree; for example, tracking of error rates in processes would allow the toolkit to be extended to address issues of product quality as well as efficiency.
We believe that utilizing the advanced features of an existing general-purpose simulation environment is a practical and efficient way to proceed in developing a high-powered dynamic modeling capability for BPR analysts. The details of developing a dynamic modeling toolkit would, however, be different using a different simulation environment. SMC?s ARENA, for instance, with its object-oriented capabilities, would have less need of a hierarchy manager such as SIMtree; there are, however, other aspect of the tool that may not be completely ?BPR-friendly? -- for instance, the issue of front-end usability may require more attention with ARENA than is the case for SIMprocess. Furthermore, while the hierarchical emphasis of SIMtree has been found appropriate in practice, not all systems fit well into hierarchies. SIMtree?s underlying relational model of simulation results could be used to synthesize other structures, such as a network.

References

1. Blattberg, R. C., and S. J. Hoch, ?Database models and managerial intuition: 50% model + 50% manager,? Management Science, 36(8), 1990, pp. 887-899.
2. Davenport, T. H. and J. E. Short, ?The new industrial engineering: Information technology and business process redesign,? Sloan Management Review, 31, Summer 1990, pp. 11-27.
3. Dur, R. C. J. and P. W. G. Bots, ?Dynamic modelling of organizations using task/actor simulation,? in Dynamic Modelling of Information Systems, II, (H. Sol & R. Crosslin, eds.), Elsevier Science Publishers, 1992.
4. Hammer, M. ?Reengineering work: Don?t automate, obliterate,? Harvard Business Review, July-August 1990, pp. 104-112.
5. Humphreys, P., D. Berkeley and F. Quek, ?Dynamic process modelling for organisational systems supported by SASOS,? in Proc., Third Int. Working Conf. on Dynamic Modelling of Information Systems (pp. 47-64), Noordwijkerhout, The Netherlands, June 9-10, 1992.

Future research directions are tied to increasing the integration and usability of the DMT. Further refinement and field testing of the front-end data repository is under way. An ultimate objective is to produce SIMprocess models directly from the data in the front-end repository with minimal user intervention. Field testing also reveals a demand for further varieties of output from

6. Jordan, E. and J. B. Evans, ?The simulation of IS strategy using SIMIAN,? in Dynamic Modelling of Information Systems, II, (H. Sol & R. Crosslin, eds.), Elsevier Science Publishers, 1992.
7. Leclerc, L. ?Does America say Yes to Japan?? 2nd ed.,

DIAGRAM T42: 07/14/93 16:37:47

UNITS
STATION PROCESSED BUSY(%) IDLE(%) ------------------------------------------------------------------------------------------------------------------------ T_NEGATIVE_FEEDBACK 448 33.1 66.9 T_POSITIVE_FEEDBACK 437 99.0 1.0 ------------------------------------------------------------------------------------------------------------------------ AGGREGATES 437 66.1 34.0

UNITS MIN Q MAX Q AVG Q STDDEV MIN Q MAX Q AVG Q STDDEV Q BUFFERS PROCESSED BUSY(%) IDLE(%) DELAY DELAY DELAY DELAY LEVEL LEVEL LEVEL LEVEL ------------------------------------------------------------------------------------------------------------------------ Q_POSITVE_F 438 94.5 5.5 0.0 0.0 0.0 0.0 0 10 5.5 3.2 R_CUSTOMER_ 450 98.9 1.0 0.0 0.0 0.0 0.0 0 451 222.9 0.0 ------------------------------------------------------------------------------------------------------------------------ AGGREGATES 450 98.9 1.0 0.0 0.0 0.0 0.0 0 451 222.9 0.0

FINAL OUTPUT ------- PROCESS TIME/UNIT -------- ------- UNIT LEVEL ----------- PRODUCTS PART UNITS BUSY(%) IDLE(%) MIN MAX AVG STDDEV MIN MAX AVG STDDEV ------------------------------------------------------------------------------------------------------------------- FEEDBACK 437 15.3 84.7 2.0 2.6 2.3 0.1 0 11 6.5 N/A FORWARDED_INFO 448 45.0 55.0 2.2 4.2 3.9 0.1 0 454 224.6 N/A SATISFIED_CUST * 437 N/A N/A 4.2 6.8 6.2 N/A N/A N/A N/A N/A ------------------------------------------------------------------------------------------------------------------- AGGREGATES 437 34.2 65.8 4.2 6.8 6.2 N/A 0 465 231.1 N/A

OPERATING COSTS: HUMAN RESOURCES BUSY(%) IDLE(%) BUSY($) IDLE($) TOTAL($) UNITS PER UNIT TOTAL ------------------------------------------------------------------------------ ------------------------------- T_NEGATIVE_FEED 74.1 26.0 51,333.56 0.00 51,333.56 437 144.66 63,215.33 T_POSITIVE_FEED 99.0 1.0 11,881.77 0.00 11,881.77 ------------------------------------------------------------------------------ TOTAL H.R. COSTS 63,215.33 0.00 63,215.33 AVG H.R. COSTS 79.1 21.0 31,607.66 0.00 31,607.66

Figure 9. SIMtree report for the Customer Feedback model (T42).

Internet ftp site monu6.cc.monash.edu.au, 1992. and Trends for the 1990's and Beyond (T. Bergin, ed.), Harrisburg, PA: Idea Group Publishing, 1993.8. Roach, S. S. ?Services under siege -- the restructuring
imperative,? Harvard Business Review, September-October 1991, pp. 83-91.
12. Warren, J., J. Stott, and A. Norcio, ?Stochastic simulation of information systems designs from data flow diagrams,? J. Systems & Software, 18(2), pp. 191-199, 1992.9. Schriber, T. An Introduction to Simulation Using
GPSS/H. New York: Wiley, 1992.
Acknowledgments
10. Streng, R. J. and H. G. Sol, ?A dynamic modelling approach to analyze chain dynamics on the inter-organizational level,? in Proc., Third Int. Working Conf. on Dynamic Modelling of Information Systems (pp. 1-35), Noordwijkerhout, The Netherlands, June 9-10, 1992.

Special thanks are owed to Debra Warren for her development work on SIMtree, and to Scott Ramoth and Anthony Wenig for their ongoing efforts in the implementation and integration of the toolkit.
11. Warren, J. ?Simulation for CASE: Use of a prototype and experimentation to assess the effectiveness of a CASE method,? in Computer-Aided Software Engineering: Issues