PERFORMANCE IMPROVEMENT BY

WORKFLOW MANAGEMENT SYSTEMS

Preliminary results from an empirical study

Hajo A. Reijers

Department of Technology Management, Technische Universiteit Eindhoven , Den Dolech 2, Eindhoven, The Netherlands

Keywords: Workflow management systems, performance measurement, case study, empirical research

Abstract: Workflow Management (WfM) s

ystems have acquired a respectable place in the market of enterprise

information systems. Although it is clear that implementation of a WfM system may shorten process

execution and increase efficiency, little is known about the extent of these effects on business process

performance. In this paper, we report on a running longitudinal multi-case study into the quantitative effects

of WfM systems on logistic parameters such as lead time and service time. We conclude that in most cases

significant decreases of lead time and service time will take place for the cases under consideration. In the

presentation of our research outline, we show how we use process simulation for the validation of our

measurements, the prediction of performance improvement, and the comparison of the pre- and post-

implementation situation. As a side effect of this study, we present some interesting characteristics of actual

business processes and the way WfM systems are implemented in practice.

1 INTRODUCTION

Workflow management (WfM) systems have been

around since the early nineties, while their

conceptual predecessors range back even further

(see e.g. Ellis, 1979). Although not as widespread as

for example Enterprise Resource Planning systems,

they have become "one of the most successful

genres of systems supporting cooperative working"

(Dourish, 2001). The worldwide WfM market,

estimated at $213.6 million in 2002, is expected to

redouble by 2008 (Wintergreen, 2003). Furthermore,

WfM functionality has been embedded by many

other contemporary systems, such as ERP, CRM,

and call-center software.

The alleged advantages of WfM systems are

ear. By having a dedicated automated system in

place for the logistic management of a business

process, such processes could theoretically be

executed faster and more efficiently (Lawrence,

1997). Yet, very little is known about the extent of

performance improvement an organization may

experience in practice. Single case studies are

available (e.g. Prinz and Kolvenbach, 1996; Goebl et

al., 2001), but do not lend themselves for

generalization. Few empirical studies that include

multiple implementations are known to us. What is

more, their focus is not on performance issues, but

on aspects such as implementation (Parkes, 2002) or

the appreciation of the technology by end-users

(Kueng, 1998). The study most related to our

research is that of Oba et al. (2000), who developed

a regression model on the basis of 20 cases to

predict the reduction of lead time as a result of WfM

implementation. Other available data on

performance improvement comes from WfM

vendors, who are perhaps not completely unbiased.

A study among 100 clients of Staffware, one of the

world's largest WfM vendors, indicates for instance

that 62.5% of their clients sees increased efficiency

as a result of WfM implementation (Staffware,

2000). Unfortunately, this outcome is not

accompanied by indications how this figure is

established or how much efficiency gains are

achieved.

This paper is an interim report on a longitudinal,

ulti-case study into the effectiveness of WfM

technology (see Yin, 1994). Its aim is to quantify the

contribution of WfM technology to improved

business process performance with respect to lead

time, wait time, service time, and utilization of

resources. In this way, it is an extension of the scope

of the study by Oba et al. (2000).

Our study, which is conducted in the

Neth

erlands, is a joint effort by Eindhoven

359

A. Reijers H. (2004).

PERFORMANCE IMPROVEMENT BY WORKFLOW MANAGEMENT SYSTEMS - Preliminary results from an empirical study.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 359-366

DOI: 10.5220/0002634303590366

 SciTePress

University of Technology and Deloitte & Touche

management consultants. It started in 2001 and is

planned to continue until at least mid 2005. So far, 6

organizations are involved who are in the process of

implementing WfM technology to support 17

different business processes. All organizations are

administrative organizations, both commercial and

not-for-profit, ranging from medium-sized to large.

Although few decisive insights can be reported

yet, we are able to publish our expectations on

performance improvement for each process. These

expectations can be tested when the implementation

have been completed and the WfM-enabled

processes are taken into operation. Moreover, this

study reveals interesting characteristics of

administrative business processes and the way that

workflow implementations are carried out. These

insights may be of interest to those who are active as

researchers or practitioners in the WfM arena.

Finally, we see this paper as an opportunity to

present our chosen research methodology and

generate feedback from the research community.

The structure of this paper is as follows. First we

will outline our research design in Section 2. In

Section 3, we will highlight our experiences on the

execution of the study. In Section 4, we will present

our preliminary results. Section 5 contains our

conclusion and outlook.

2 RESEARCH DESIGN

2.1 Objective

The aim of the effectiveness study is to determine

how the performance of the business processes is

affected by the implementation of workflow

technology. The four performance indicators

selected to investigate for each involved business

process are as follows:

- lead time, i.e. the time between the arrival of

a case and its completion (also known as

cycle time, completion time, and turnaround

time)

- service time, i.e. the time spent by resources

on the processing of a case

- wait time, i.e. the time a case is idle during

its life cycle,

- utilization of involved human resources, i.e.

the ratio of activity versus their availability.

For each of these indicators, we report in this

paper on the average values. We are aware of the

importance of other measures, such as service levels

and the degree of variance. Information on these

values will be included in our final report.

By introducing WfM technology, one may aim to

decrease each of the average values of the given

performance indicators. Because work is routed by

an automated system, work reaches people faster

and will not get lost. This decreases lead time and

wait time. It will allow people to spend less time on

coordination and on the transfer of work, which

means a decrease of service time. When the supply

of work and resources remain constant, work load

and utilization will decrease as a result. Therefore,

the hypothesis for this study is that the averages of

all four performance indicators will decrease

significantly as a result of the use of a WfM system.

2.2 Research steps

To determine the effects on process performance, at

the very least an initial measurement of the relevant

parameters is required (a) before the WfM

implementation and (b) afterwards. Three major

issues have further shaped the design of the

research:

- the validation of the measurements: how can

it be ensured that the collected data on the

process performance is correct?

- the prediction of results: can we try to

estimate the results of the WfM technology

before its actual implementation?

- the comparison of the measurements: how

can a proper comparison between various

situations takes place?

The major steps in the research that address these

issues are given in Figure 1. In this figure, two axis

can be distinguished. On the horizontal axis, we

have the situation before the WfM technology

implementation on the one hand and the situation

afterwards on the other. On the vertical axis, we

distinguish between the real data on the process on

the one hand and the data that follow from a

simulation of that process on the other. In the figure

it is shown that there are six research steps, which

take place in the order 0, 1a, 2a, 3, 2b, and 1b. We

will explain these steps in some detail and explain

how they address the issues we identified. For now,

it is sufficient to say that the a-measurements use the

initial circumstances, while b-measurement are

based on the final circumstances.

The basis of the research design is formed by

gathering real data on the process before and after

the implementation of the WfM systems. We

respectively refer to these measurements as the 0-

measurement and the 3-measurement.

To address the issue of validation (1.), a

computer model is build of each business process

subject to the study, both before and after the WFM

implementation. We refer to the simulation of the

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

360

simulation

using final

circumstances

real data

simulation

using initial

circumstances

after WfM

implementation

before WfM

implementation

1a-measurement

1b-measurement

0-measurement

2a-measurement

2b-measurement

3-measurement

Figure 1: Research steps

model of the initial and final situation as

respectively the 1a-measurement and the 2b-

measurement. Both simulation models are as

realistic as possible, including real data on the actual

structure of the business process, the actual arrival of

cases, the actual availability of resources, and the

actual routing probabilities of cases flowing through

the process, etc. Enactment of the simulation model

delivers results on, for example, the lead times of

the process and the resource occupation. These

simulation results can be compared with the

observations of the actual process. For example, the

average lead time following from the simulation can

be compared with lead time averages observed in

practice. Concordance of the real and simulated data

gives us some support for the validity of the

measurements, either of the initial situation or the

situation after the WfM implementation. Large

differences between these outcomes may indicate

that a part of the process is not understood or

modeled correctly.

To enable prediction (2.), we attempt to build a

simulation model that reflects the situation after the

implementation of the WfM system (the 2a-

measurement). This model is based on the

simulation model of the current process (used for the

1a-measurement) and captures both realistic and

estimated data. On the one hand, it incorporates the

aspects of the initial process that presumably will be

the same after implementation. On the other hand,

typical effects of WfM technology are incorporated

in it. For example, transportation activities that exist

in the current process are eliminated from the model,

because WfM technology will take care of

transportation. Furthermore, planned initiatives of

the organization to e.g. optimize the process

structure or change the resource staffing are also

incorporated in the model for the 2a-measurement.

In this way, its estimate of the future overall effect is

the most realistic. A comparison between the 1a- and

2a-measurement delivers insights in the expected

benefits of the WfM technology.

The issue of comparison (3.) is slightly more

sophisticated. As we are primarily interested in the

effect of the WfM technology, a straightforward

comparison between the initial and final

measurement (the 0- and 3-measurement) is

perilous. After all, various other variables may have

changed during WfM implementation that affect the

final measurement. For example, if WfM technology

is implemented while at the same time a staff

reduction takes place, the performance following

from the 0- and 3-measurement may be similar. It

would not be proper in such a case to decide that

WfM technology has not contributed anything.

Similarly, the supply of work may have changed

drastically.

To counter these effects, we build a new

simulation model, which is used for a so-called 1b-

measurement. It mixes elements from the 1a-model

and the 2b-model. More specifically, it is as close to

the 1a-model as possible, while incorporating all

changes apart from the workflow implementation

that have taken place between the 0- and 3-

measurement. In the previously example of staff

reduction, this would mean that it includes e.g. the

original process structure but a reduced number of

staff compared to the original, initial situation. A

comparison between the 2a and 2b-measurements

will therefore be much more meaningful.

In summary, the 1a- and 2b-measurements serve

as validation for respectively the 0- and 3-

measurement. A comparison between the 1a- and

2a-measurement gives an estimation of the effects of

PERFORMANCE IMPROVEMENT BY WORKFLOW MANAGEMENT SYSTEMS: PRELIMINARY RESULTS

FROM AN EMPIRICAL STUDY

361

WfM technology beforehand, while a comparison

between the 1b- and 2b-measurement is the most

reliable estimation of the actual net effect of WfM

technology.

2.3 Data gathering and analysis

Business processes contain a certain structure and

they show a certain behavior. For this study, the

most important categories of data to be determined

for each business process are as follows:

- process: tasks, milestones, business logic,

routing probabilities

- resource: types of resources, work

assignment policies, number and availability

of resources

- performance: service times, lead times,

arrival rate of new cases, work-in-progress,

resource utilization

For data gathering, the researchers used a multi-

method approach, combining interviews, existing

process descriptions, observations, management

reports, self-registrations by people involved in the

process, and automatically collected data by existing

information systems. For each measurement, a

careful consideration has been made for the most

suitable mix of instruments.

An important difference between the 0- and 3-

measurement with respect to data gathering concerns

the availability of data. Where possible, the use of

existing registrations on the processing of historic

cases were favored over conducting new, manual

registrations for reasons of representativity and ease

of extraction. For the 0-measurement it was

somehow inevitable that new data collection had to

take place, for useful administration of this data

within the organizations was often lacking. For the

3-measurement, the data gathered by the WfM

system itself is an obvious rich and accessible source

of this type of information.

Processes were modeled as Petri nets using the

commercial tool Protos (Pallas Athena, 1997). The

tool allows for efficient communication with end-

users and the organization’s management, thus

simplifying knowledge extraction and validation.

Protos models were automatically translated to

simulation models, which could be executed and

analyzed by the Petri-net based simulation tool

ExSpect (Van Hee et al., 1989). ExSpect provides a

rich environment for simulation and analysis (e.g.

confidence intervals, sensitivity analysis). For more

information on the interplay between these tools, the

reader is referred to a paper by Van der Aalst et al.

(2000).

2.4 Progress

The workflow study started in September 2001 and

is expected to continue until at least mid-2005. So

far, six Dutch organizations have been actively

involved in the study. The characterization of these

organizations are given in Table 1. Note that the

column ‘cases per year’ shows the typical number of

cases processed by the largest process under

consideration for that specific organization. The

respective processes of the involved organizations

typically spanned a dozen up to over hundred

activities. Both fully automated, semi-automated and

purely manual activities were part of almost all of

these processes. In the simplest case, only two

different resource classes were involved in the entire

workflow process, while this ranged to seven or

eight different resource classes for more complex

cases.

An organization could participate in the study

when it had already selected a WfM system, but did

not yet implement it. The actual WfM systems

involved in this study were three commercially

available WfM-systems (Staffware, COSA, and

FLOWer) and one proprietary system (VenWfm).

For all listed organizations, the initial 0-

measurement has now been completed. For two of

these, the final 3-measurement is in process. For two

others, the WfM project has been stopped by the

respective organizations. At the same time, three

new candidate organizations have applied to

participate in the study (not included in the list).

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

362

Table 1: Participants in the workflow study

organization

number

organization

description

number of

employees

turnover/

budget (x

millions €)

focus of

involved

processes in

study

number of

involved

process in

study

cases per

year

(x 1000)

1. governmental

agency

700 60 (b) debt

collection

1 7000

2. health insurer 2300 5200 (t) policy

maintenance

7 250

3. regional

public works

department

1000 250 (b) invoice

processing

1 20

4. local

municipality

300 210 (b) invoice

processing

2 25

5. insurance

intermediary

5000 29000 (t) policy

maintenance

4 2000

6. domiciliary

care agency

1450 50 (t) human

resource

management

2 1,5

3 ISSUES AND EXPERIENCES

3.1 Study

A disappointing event during the study was the

premature break off of two of the participating

organizations. In both cases, considerations on the

organizational level caused the termination of the

workflow projects. This involved a change of

management within one organization and IT budget

problems within the other. No indications were

found that the decisions to stop were related to the

WfM-technology itself. We suppose that the

implementation of WfM technology is as hazardous

as that of other large-scale information systems.

3.2 Simulation

The validation of the processes by means of

simulation has proved to be an effective validation

means, but a more difficult one than estimated.

Simulations did turn out to be helpful in finding

several mismatches and omissions. Yet, four major

modeling issues emerged.

The first issue concerns the problem that human

resources are often dedicated to more than one

business process at a time. The distribution of

attention over these processes is rather dynamically

determined and hard to model, e.g. some people

work on the process that is the busiest and others on

the ones they like most. This issue has also been

identified by Sierhuis (2001).

The second issue involves the modeling of part-

time resources. The problem is different than the

previous one. Even though a broken number of

available FTE's may be known, the simulation is

sensitive to the way this availability is implemented.

For example, 9 half-time people perform differently

than 4 full-time people and 1 half-time person.

Thirdly, the performance of human resources

proved to be rather elastic. We have seen people be

able to process 40 % more cases in busy periods. In

other words, average service times decreased when

the supply of work increased.

The final issue was related to a varying work

load throughout the seasons. For the insurance

intermediary, for instance, the month of December

proved to be a much busier period than any of the

summer months.

All four issues have been met in the simulation

models by extending the logic of the simulation

components, extending the number of observations,

and/or an extended analysis of work practice.

3.3 Data gathering

The activity of data gathering was hampered by the

unavailability of registrations of all kinds of data

within organizations, such as the receipt of triggers

or the completion of milestones. Observations and

registrations by workers themselves were used to

counter this problem. Furthermore, the available

data gathering period preceding the WfM

implementation was often shorter than the usual lead

time of a single case. So, instead of following single

cases flow through the business process to determine

the service time spent on this case, more often for

each task in the process its average service time was

determined by counting a number of executions for

different cases.

PERFORMANCE IMPROVEMENT BY WORKFLOW MANAGEMENT SYSTEMS: PRELIMINARY RESULTS

FROM AN EMPIRICAL STUDY

363

3.4 Process

An interesting side-effect of this study was that it

gave the researchers the opportunity to examine the

characteristics of business processes as executed,

monitored, and designed in practice. We will

highlight these insights here.

For each of the organizations, the distinguished

performance criteria as distinguished in this study

were mentioned as targeted by their own WfM-

implementation. This positively confirmed our ideas

on the importance of these notions. Additional goals

that were mentioned: increased service quality,

increased process flexibility, and a better integration

of stand-alone applications.

One of the striking observations was that out of

the 17 processes considered none of these processes

incorporated concurrent behavior, i.e. parallel

processing of single cases. Business processes

turned out to be completely sequential structures.

Their routing complexity was only determined by

choice constructs and iterations. Even more

remarkable is that for only one of these processes the

process owners indicated that they considered to put

more parallelism in the process once it would be

supported by a WfM-system. This contradicts with

the idea that parallel processing is an obvious next

step in improving the performance when adopting

WfM technology (see e.g. Van der Aalst and Van

Hee, 2003, p.93).

On a related note, the implementation of a WfM-

system did not prove to be a direct incentive to

redesign the structure of the business process

drastically. Without exception, each participating

organization favored in the short term the situation

to have a WfM-system supporting the current

process over a drastically improved version of the

process. This may be counter-intuitive. One of the

strengths of WfM-technology's is that it enables the

restructuring of the process structure. Moreover,

automating a "paper" process may not be the most

effective way to achieve decreased lead and service

times. On the other hand, this approach decreases

the risk of failure by lowering the project's

complexity. So, from a change management

perspective, the selected strategy of the

organizations may be wise.

3.5 Technology

For four out of the six participating organizations,

WfM technology was entirely new at the start of this

study. The two other organizations had experiences

with the implementation of WfM technology in

other business processes.

For all involved processes the implementation of

a document management system accompanied the

introduction of the WfM system. Considering the

tight integration between the two types of systems,

we have established that the outcomes of this study

can only relate to their combined use. In other

words, the digital storage of both structured and

unstructured data is crucial for a WfM system to be

useful in practice.

4 RESULTS

A summary of the most important results is shown in

Table 2. Table 2 gives the 1a- and 2a-measurements

of the lead time and service time, as well as the 1a-

measurement of the utilization. Also shown are the

expected gains from WfM systems for the lead time

and service time, as can be derived from their 1a-

and 2-measurements. Significant changes are

accentuated.

For 15 out of 17 business processes (88%), the

average lead time is expected to decrease

significantly. The gains range from 25% to 83%,

with an average of 48%. For the two other processes,

these results are not significant.

Process 14. is the furthest away from a

significant decrease of lead time. However, this

turns out to result from a redefinition of the ‘case’

concept. In the new situation, the WfM system

makes it possible to loop back a class of problematic

cases to the beginning of the process. Historically,

these cases could only be terminated and then re-

instantiated under a different id, which used to lead

to overly positive lead time outcomes.

With respect to service time, for 13 out of 17

business processes (76%) a significant change is

expected to take place. From these, 12 processes

show an expected decrease of service time between

4% to 47%. However, in the situation of process 5.

an increase of service time is expected to take place

(9%). On average, an expected decrease of service

time of 23% is expected for these 13 processes.

It is interesting to take a closer look at process 5.

It handles the simple mutations of health insurance

policies, such as caused by a change of address. It is

the process with the lowest complexity and the

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

364

Table 2: Main results study

Lead time Service time Utilization Org.

nr.

(see

Tab.1)

Proc.

nr.

1a-meas.

(average

value in

days)

2a-meas.

(average.

value in

days)

reduction

(%)

1a-meas.

(average

value in

minutes)

2a-meas.

(average

value in

minutes)

reduction

(%)

1a-meas.

(weighted

average %)

1. 1. 59,1 9,8 83

13,45 7,14 47

2. 3,83 2,13 44

16,01 9,01 44

3. 3,35 1,89 44

4,16 4,01 4 65

4. 3,40 1,96 42

8,54 8,14 5

5. 3,76 1,72 54

3,51 3,84 -9

6. 4,19 2,31 45

9,25 8,90 4

7. 3,37 2,01 40

10,75 8,19 24

8. 3,01 1,83 39

5,4 3,89 28

3. 9. 16,00 11,93 25

17,66 17,39 2 4

10. 6,50 1,81 72

42,00 22,11 47

3 4.

11. 13,08 6,82 48

19,45 13,21 32

12. 6,17 4,56 26

12,13 12,56 -4

13. 5,17 2,34 55

11,25 11,11 1 67

14. 8,68 9,48 -9 61,55 43,44 29

15. 5,18 2,36 55

12,06 11,03 9

16. 8,92 5,12 43

24,19 20,97 13

23 6.

17. 1,49 1,36 9 13,69 10,72 22

** = significant with two-sided 99% confidence intervals, * = significant with two-sided 90% confidence intervals

lowest initial average service time value (3,51

minutes). Clearly, the overhead caused by the use of

the WfM system – starting the system, registering

work to be completed, etc. – can in this case not be

compensated by less coordination work.

Note that some categories of data are not shown

in the table. In this phase of the study, they can still

be derived from the presented data as follows:

− the 0-measurements: All average values of

the 0-measurement are within the 99%

confidence interval of the values of the 1a-

measurement. In other words, the 1-

measurements accurately reflect the

situation at the 0-measurement.

− the 2a-measurement on the utilization:

Utilization will change accordingly to the

expected change of service time, because an

equal supply of work and workforce is

assumed after each WfM implementation.

− the measurement on the wait time: Because

of the almost complete lack of concurrency,

the wait time in each situation can be

accurately determined by subtracting the

service time from the lead time. The general

relation between these entities is discussed

by e.g. Reijers (2003, p.182).

In other words, the effects on utilization are

equal to the effects on service time and the effects

on wait time are similar to the effects on lead time.

Note that in general these similarities will not hold

when comparing the 0- and future 3-measurements.

5 CONCLUSION AND OUTLOOK

At this stage of the research, we have indications

that WfM systems in general will positively affect

the identified performance indicators averages. In a

large majority of cases we investigated, service time

and utilization are expected to decrease with 23%.

For an even larger majority, lead time and wait time

are expected to decrease with more than twice that

amount, namely 48%. Clearly, it needs to be seen

whether these results will be accomplished in

practice. On the basis of an almost completed 3-

measurement for organization 3., we are hopeful that

the actual gains are indeed in the range of the

predicted gains.

As a side effect, this empirical study has proved

to be a valuable source of information on actual

business process properties and their execution.

Also, simulation proved to be a good way of

validating the initial measurements, but a number of

challenges had to be faced. Unfortunately, we have

seen two organizations putting their WfM

implementations on hold, perhaps definitively. We

are still attracting new organizations to get involved

in the study to generate support for general

conclusions. Finally, it seems that the evaluation of

PERFORMANCE IMPROVEMENT BY WORKFLOW MANAGEMENT SYSTEMS: PRELIMINARY RESULTS

FROM AN EMPIRICAL STUDY

365

the performance of WfM systems cannot be

separated from the contribution of document

management systems.

Currently we are carrying out the first two 3-

measurements. As part of this work, we are

developing a general tool to derive from the event

logs of WfM-systems performance information. We

are studying industry solutions in this field, such as

the ARIS process performance monitor, and XML

data formats.

ACKNOWLEDGEMENTS

The following people from Eindhoven University of

Technology and Deloitte & Touche have greatly

contributed to this study: Wil van der Aalst, Kees

van Hee, Charo Narvaez, René Theunissen and Eric

Verbeek

REFERENCES

Aalst, van der, W.M.P., Crom, de, P.J.N, Goverde,

R.R.H.M.J., Hee, van, K.M, Hofman, W.J., Reijers,

H.A., Toorn, van der, R.A., 2000.

ExSpect 6.4: An Executable Specification Tool for

Hierarchical Colored Petri nets.

In ATPN’00, Application and Theory of Petri Nets

2000, Lecture Notes in Computer Science 1825.

Springer-Verlag.

Aalst, van der, W.M.P, Hee, van, K.M., 2002. Workflow

Management: Models, Methods, and System, MIT

Press. Cambridge.

Dourish, P., 2001. Process descriptions as organizational

accounting devices: the dual use of workflow

technologies. In GROUP'01, ACM Conference on

Supporting Group Work. ACM.

Ellis, C.A., 1979. Information Control Nets: A

Mathematical Model of Office Information Flow. In

SIGMETRICS’97, ACM Conference on Simulation,

Measurement and Modeling of Computer Systems.

ACM.

Goebl, W., Messner, K.J., Swarzer, B., 2001. Experiences

in introducing workflow management in a large

insurance group. In HICSS’01, 34

Hawaii

International Conference on System Sciences. IEEE.

Hee, van, K.M. Somers, L.J., Voorhoeve, M., 1989.

Executable Specifications for Distributed Information

Systems. In Proceedings of the IFIP TC 8 / WG 8.1

Working Conference on Information System Concepts:

An In-depth Analysis. Elsevier Science.

Kueng, P., The effects of workflow systems on

organizations: a qualitative study. In Business Process

Management: Models, Techniques, and Empirical

Studies, Lecture Notes on Computer Science 1806.

Springer-Verlag.

Lawrence, P., editor, 1997. Workflow Handbook 1997,

John Wiley and Sons. New York.

Oba, M., Onada, S., Komoda, N., 2000. Evaluating the

quantitative effects of workflow systems based on real

cases. In HICSS’00, 33

Hawaii International

Conference on System Sciences. IEEE.

Pallas Athena. PROTOS User Manual. Pallas Athena BV,

Plasmolen, 1997.

Parkes, A., 2002. Critical Success Factors in Workflow

Implementation. In PACIS’02, 6

Pacific Asia

Conference on Information Systems. Jasmin.

Prinz, W., Kolvenbach, S., 1996. Support for Workflows

in a Ministerial Environment. In CSCW’96, ACM

conference on Computer supported cooperative work.

ACM.

Reijers, H.A., 2003. Design and Control of Workflow

Processes: Business Process Management for the

Service Industry, Lecture Notes in Computer Science

2617. Springer-Verlag.

Staffware, 2000. Benefits, Progress and Competitive Edge

– The Customer’s Experience, Staffware. Maidenhead

UK.

Yin, R. K., 1994. Case Study Research, Design and

Methods, Sage Publications. Newbury Park, 2

edition.

WinterGreen, 2003. Business Process Management (BPM)

Market Opportunities, Strategies, and Forecasts, 2003

to 2008, WinterGreen Research. Lexington MA.

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

366