BUSINESS PROCESS MODEL IMPROVEMENT BASED ON

MEASUREMENT ACTIVITIES

Laura Sánchez-González, Francisco Ruiz, Félix García and Mario Piattini

Alarcos Research Group, TSI Department, University of Castilla La Mancha

Paseo de la Universidad, nº4, 13071, Ciudad Real, Spain

Keywords: Business process, Measurement, Continuous improvement, BPMN.

Abstract: The current importance of Business Process improvement lies in the fact that it is a key aspect for

organizational improvement. Since business process improvement can be dealt from different perspectives,

we propose the use of measurement as a technique by which to collect information concerning the quality of

the process. We have specifically applied measures to the design stage of the business process lifecycle,

which signifies measuring conceptual models. Measurement in Design and Analysis lifecycle stage has

several advantages, principally in that it is a means to avoid the propagation of errors to later stages, in

which their detection and correction may be more difficult. We therefore propose certain steps for business

process model improvement, based on measurement activities (measurement, evaluation, and redesign).

These activities have been applied to a real hospital business process model. The model was modified by

following expert opinions and modelling guidelines, thus leading to the attainment of a higher-quality

model. Our findings clearly support the practical utility of measurement activities for business process

model improvement.

1 INTRODUCTION

In recent years, business process (BP) modelling and

improvement has become an important means of

ensuring changes in an organization’s structure and

functioning, thus leading to the creation of a more

competitive and successful enterprise (Damij, Damij

et al. 2008). BP influences product quality and

customer satisfaction, which are fundamental

aspects in a market environment, and enterprises are

therefore forced to improve their processes in order

to improve products and services (Cardoso 2006).

The first step towards improving business

processes is to collect any data regarding their

design, deadlocks, bottlenecks, etc. Measurement is

a good means of collecting this kind of data, and

serves at least the following three purposes:

understanding, control and improvement (Park,

Goethert et al. 1996). The use of measurement

information therefore makes it possible for

organizations to learn from the past in order to

improve their performance and achieve better

predictability over time.

A business process is a complex entity with a

characteristic lifecycle. In our work we consider the

approach defined by Weske (Weske 2007), who

organizes the lifecycle in a cyclic structure with

logic dependences between the design and analysis,

configuration, enactment, and evaluation stages. We

focus on the first stage, design and analysis, in

which the principal activity is that of process

modelling. The main purpose of design and analysis

is to capture the business schema and general

procedures (Sparks 2000). The conceptual models

produced in this stage are first required to be

intuitive and easily understandable in order to

facilitate communication among stakeholders.

Measuring and improving BP models has several

advantages, principally that of avoiding the

propagation of errors or bad-structures to later

lifecycle stages, in which corrections and

modifications may involve a high economic cost and

effort (Wand and Weber 2002).

Measures for conceptual models deal with the

static properties of BP and are defined upon the BP

model at the time of the design. Several initiatives

concerning the measurement model have recently

been published, owing to the advantages of

improving business processes in this stage. Most of

the measures published to date have been collected

104

Sánchez-González L., Ruiz F., García F. and Piattini M..

BUSINESS PROCESS MODEL IMPROVEMENT BASED ON MEASUREMENT ACTIVITIES.

DOI: 10.5220/0003462201040113

In Proceedings of the 6th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE-2011), pages 104-113

ISBN: 978-989-8425-57-7

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Improvement activities in BP lifecycle.

in (Sánchez-González, García et al. 2010b). This

work shows that there is no consensus among

researchers as to which measurable concepts it is

most interesting to measure (complexity,

structuredness, cohesion, coupling, etc). It also

highlights that most of the proposals have not been

empirically validated. This lack of validation

particularly emphasises the need for research in this

area. The work presented herein contributes to the

maturity of BP measurement through the collection

of measures and the demonstration of their practical

utility in an experience report.

The principal idea behind our proposal is to

apply measurement during the early stages of the

lifecycle, the design and analysis stage, in order to

obtain feedback controlled by measures and thereby

achieve a higher-quality implementation of the

process, with a lower value of complexity, therefore

making it easier to maintain (Mendling 2008). The

measurement process is divided into three activities:

applying measures, evaluating measurement results

and redesigning the model. The pragmatic idea of

these activities is to discover unsafe design,

hazardous structures or unexpected. Finally, one

critical aspect of the improvement activities is to

demonstrate that they are potentially useful in

practice. We therefore present an experience report

of the application of improvement activities to a real

hospital business process.

The remainder of the paper is as follows. In

Section 2 we describe the improvement activities in

which measures were applied, evaluate the

measurement results and redesign the model. In

Section 3 we present an experience report of the

application of these activities to a real business

process, specifically a hospital business process. In

Section 4, we describe some implications and

limitations of this research. Finally, Section 5 shows

our conclusions and presents topics for future

research.

2 BUSINESS PROCESS MODEL

IMPROVEMENT

In this article, we propose certain activities for

business process model improvement. The principal

idea is to collect as much information as possible

about the static properties of the business process.

The activities are: applying measures collected in

previous works, evaluating measurement results

against threshold values and redesigning the model.

These three activities can be executed in a cyclic

manner, signifying that multiple iterations can be

run to obtain a high-quality model. This idea is

depicted in greater detail in Figure 1.

In Figure 1 the lifecycle stages are represented as

a square and the improvement activities as ellipses.

The design and analysis stage initially produces a

conceptual model. This model serves as input for the

improvement activities. The improvement of the

model can be carried out in several iterations of the

3 activities (measurement, evaluation and redesign).

These activities can be introduced in the BP

lifecycle as an extended stage, which can enrich the

final product. After the configuration stage, the

execution model is enacted through the generation of

log files, which describe all the steps followed to

achieve the business goals. These log files can be

measured (processed) in order to discover certain

important aspects such as execution time, deadlocks,

etc. The measurement initiatives for improvement in

the execution stage are described in (Delgado, Ruiz

BUSINESS PROCESS MODEL IMPROVEMENT BASED ON MEASUREMENT ACTIVITIES

105

et al. 2009). The evaluation of these execution

reports implies the generation of new business

requirements which had not previously been

considered.

2.1 Measures for Business Process

Conceptual Models

In recent years, the number of measurement

approaches for conceptual models has grown

considerably owing to the advantage of improving

business processes in the early phases. BP model

measures are used to quantify structural aspects of

models, which signifies measuring their internal

quality. This internal quality is understood as the

model’s total number of characteristics from an

internal view, and this is measured and evaluated

against the internal quality requirements (ISO/IEC

2001). Internal quality (quality in general) can be

seen from different points of view, and should

therefore be quantified with more than one measure

in order to obtain as much information as possible

with regard to the model. For example, model

complexity cannot be measured solely with the

Control-flow Complexity (CFC) measure, because

this measure only takes into account decision node

elements.

As we mentioned, various measures are found in

literature (Sánchez-González, García et al. 2010b),

and Table 1 specifically shows references to their

measurement initiatives and provides a brief

description of them.

However, it is also important to consider external

quality in conceptual models. External quality refers

to the total number of characteristics in the model

from an external view (ISO/IEC 2001), such as how

understandable the models are, how difficult it is to

modify them, etc. From the point of view of a top-

down quality SEQUAL framework (Krogstie, Sindre

et al. 2006), understanding is an enabler of

pragmatic quality, which relates to model and

modelling and its ability to enable learning and

action. In order to clarify this idea, Figure 2 shows

the relationship between internal and external

quality and some examples of measurable attributes.

Most authors have carried out experiments

focused on the relationship between measures and

external quality attributes: understandability and

modifiability. These belong to the more general

concepts of usability and maintainability

respectively (ISO/IEC 2001).

To the best of our knowledge, very few articles

concerning the relationship between measures for

internal quality and measures for external quality

Table 1: Proposals of measures for business process

models.

Measure Description

Coupling, cohesion and

connectivity level

(Vanderfeesten, Cardoso et

al. 2007; Vanderfeesten,

Reijers et al. 2008)

Cohesion and coupling

between activities and cross

connectivity in the

relationship between nodes

and directed arcs.

Structural complexity

(Rolón, García et al. 2006)

Measures related to the

number of different elements

of BPMN models.

Error probability (Mendling

2008)

Number of nodes, diameter,

gateway mismatch, depth,

density, average and max

connector degree, cyclicity,

sequentiality and

separability.

Control flow complexity

(Cardoso 2006)

Related to the number of

OR-split, AND-split and

XOR-split

Entropy (Jung 2008) Uncertainty or variability of

workflow process models

Structuredness (Laue and

Mendling 2009)

Number of unstructured parts

Complexity (Meimandi

Parizi and Ghani 2008)

Activity, control-flow, data-

flow and resource

complexity

Goodness (Huan and Kumar

2008)

Goodness of models

regarding execution logs

have been published to date, although some research

has been published in (Rolón, Cardoso et al. 2009;

Rolon, Sanchez et al. 2009; Sánchez González,

García et al. 2010), and these works obtained a

subgroup of measures which can be considered as

good indicators for understandability and

modifiability. This subgroup of measures is shown

in Table 2. The application of this subgroup of

measures is produced in a pair (measure, result),

which should be reported in a document in order to

be used in next activity: evaluation.

Figure 2: Internal and external quality in conceptual

models.

ENASE 2011 - 6th International Conference on Evaluation of Novel Software Approaches to Software Engineering

106

2.2 Evaluation of Measurement Results

The evaluation of measurement results involves

providing an objective assessment of them.

Numerical results only offer information in terms of

comparison between models rather than an

independent interpretation. For example, given two

process models, it is possible to discover not only

which of them is best in the relative terms of a

specific measure, but whether the values are

acceptable or not. It is therefore necessary to

consider the threshold or limit values in order to

indicate for what specific value the measure’s

quality begins to decline.

Table 2: Empirically validated measures and their

relationship with understandability and modifiability.

Measure Description U

Measures of Rolón (Rolón, García et al. 2006)

TNSF Total Number of sequence flows X

TNE Total Number of events X

TNG Total Number of gateways X

NSFE Number of sequence flows from

events

NMF Number of message flows X

NSFG Number of sequence flows from

gateways

X X

CLP Connectivity level between

participants

NDOOut Number of data objects which

are outputs of activities

NDOIn Number of data objects which

are inputs of activities

CLA Connectivity level between

activities

Measures of Cardoso (Cardoso 2006)

CFC Control flow complexity. Sum

over all gateways weighted by

their potential combinations of

states after the split

X X

Measures of Mendling (Mendling 2008)

Number of

nodes

Number of activities and routing

elements in a process model

Gateway

mismatch

Sum of gateway pairs that do not

match each other, e.g. when an

AND-split is followed by an OR-

join

X X

Depth Maximum nesting of structured

blocks in a process model

Connectivity

coefficient

Ratio of total number of arcs in a

process model to its total number

of nodes

Density Ratio of total number of arcs in a

process model to the

theoretically maximum number

of arcs

Sequentiality Degree to which the model is

constructed from pure sequences

of tasks

X X

Various proposals for the extraction of threshold

values exist in literature, principally in the Software

Engineering field. Some proposals for thresholds are

derived from experience (McCabe 1976; Nejmeh

1988; Coleman, Lowther et al. 1995), but the lack of

scientific support has led to disputes about their

values. Some authors, on the other hand, have used

statistical techniques to obtain thresholds. For

example, Shatnawi (Shatnawi 2010) extracted

thresholds for Object Oriented (OO) code measures

in order to study the relationship between OO and

error-severity categories. This author also validated

the Bender method (Bender 1999) and found that

there are effective thresholds for the measures

analyzed.

With regard to business process measurement,

we have attempted to extract threshold values for

some measures in previous works. This is the case of

Control-flow complexity measure, structural

complexity and error probability measures, which

were used to apply the Bender method in order to

extract thresholds. These works were published in

(Sánchez-González, García et al. 2010a; Sánchez-

González, Ruiz et al. 2011). Table 3 shows extracted

thresholds for some empirically validated measures.

This table divides the domain of the measure into 4

different groups, depending on the level of

efficiency: “very efficient”, “fairly efficient”, “fairly

inefficient” and “very inefficient”.

2.3 Redesign of Business Process

Models

In this section, we focus on modifying some parts of

the model in order to improve its general quality.

Those parts that are candidates for alteration have

been identified through the use of measures. For

example, let us imagine that we are analyzing the

results of the CFC measure in a specific model, and

we obtain a numerical value which is higher than the

threshold: “If CFC is higher than 44, the model is

difficult to understand”. These results indicate that

the number of decision nodes must be reduced in the

model, since it may be difficult for stakeholders to

understand.

Nevertheless, modifying the model using only

the information collected from measures and

thresholds can be quite difficult. Some guidelines

therefore exist to assist modellers in this task. In

literature, it is possible to discover various

guidelines for inexpert modellers, whose purpose is

to obtain higher-quality models that can ensure a

more reliable execution. Mendling et al. (Mendling,

Reijers et al. 2010) proposed seven pieces of advice

BUSINESS PROCESS MODEL IMPROVEMENT BASED ON MEASUREMENT ACTIVITIES

107

for modellers (denominated as 7PMG) which are

built on strong empirical insight. This advice is

related to the maximum number of nodes before

decomposition, number of events, OR-routing

elements, routing paths per element or the use of a

verb-object activity label. On the other hand, Becker

et al. (Becker, Rosemann et al. 2000) define certain

guidelines of modelling (GoM), which are

specifically six general techniques for adjusting

models to the perspectives of different types of user

and purposes. To illustrate the used of these

guidelines, let us imagine the following example. If

the measure “total number of events” is higher than

20 (very inefficient), 7PMG advises that the use of

“one start and one end event” is the best way to

reduce the measure value.

Redesign therefore involves changing those

specific parts of the model with low quality detected

by measures. Modelling guidelines can also help to

ensure the quality of the model but a previous

measurement effort is necessary to identify any

potential problems.

Table 3: Thresholds for business process model measures.

1: very

inefficient

2: fairly

inefficient

3: fairly

efficientt

4: very

efficient

Understandability

Nºnodes 65 50 37 31

GatewayMismatch 29 16 6 1

Depth 4 2 1 1

Coefficient of

connectivity

1,7 1,1 0,6 0,4

Sequentiality 0,1 0,35 0,6 0,7

TNSF 72 49 34 20

TNE 20 12 7 2

TNG 17 10 5 0

NSFE 28 13 4 0

NMF 27 15 7 1

NSFG 40 22 11 0

CLP 7,5 4,23 2,2 0,2

NDOIN 31 44 4 0

NDOOUT 23 11 3 0

CFCxor 30 17 8 1

CFCor 9 4 1 0

CFCand 4 2 0 0

Modifiability

GatewayMismatch 46 22 4 1

Densitiy 0,6 0,22 0,00

Sequentiality 0 0,18 0,6 0,86

NSFG 25 13 9 0

CLA 0,53 0,875 1,1 1,3

CFCxor 27 16 8 1

CFCor 9 4 1 0

CFCand 6 2,3 0 0

3 EXPERIENCE REPORT:

HOSPITAL PROCESS

In order to demonstrate the practical utility of this

proposal, we describe an experience report which

was developed in the General Hospital of Ciudad

Real (GHCR) in Spain. First, a specific work group

was created, consisting of specialists in modelling

tasks (Software Engineers) and health professionals

at the hospital:

a) Those responsible for processes: the

assistant director of nursing and the person

responsible for hospital’s admissions units.

b) Collaborators: head of human resources and

finances, head of computer services and

head of out-patients’ healthcare.

The work group then modelled various processes

which had previously been selected by the hospital’s

managerial and quality staff, although in this paper

we shall focus on the “Incorporation of a new

employee” (INE) process, which includes the

training plan, information and suitability of those

people involved in the hospital in order to facilitate

their integration into the new job. The process model

is shown in Figure 3.

This process was selected as a low-complexity

process, although the services provided are very

important. It is a purely administrative process (it is

not related to patient care), but moves a large

number of users (in 2007, the hospital staff consisted

of 2.600 workers, and 6989 new contacts were made

with regard to substitutions and new incorporations).

This process involves different professional

categories: doctors, pharmacists, nurses,

psychologists, administrative and technical staff and

others. Specific process characteristics were the

following:

a) Mission: to promote the organization of the

INE process, which includes a plan for

training, information and adaptation of the

people involved to the hospital requirements

in order to facilitate their integration into the

new job.

b) Limits: the INE process starts when the

professional comes to the hospital and

finishes when he/she is incorporated into the

new job.

c) Clients: new professionals

d) People responsible: those responsible for

nursing, medical aspects and management.

e) Participants: new professionals in hospital,

human resources, computer services,

lingerie, pharmacy, prevention services,

nursing and management service.

ENASE 2011 - 6th International Conference on Evaluation of Novel Software Approaches to Software Engineering

108

Figure 3: BPMN model for the Incorporation of the New Employee (INE) hospital process.

f) Suppliers: human resources, provisions,

maintenance, training and information

systems.

The results of the application of the

improvement activities are described in the

following sub-sections:

3.1 Applying Improvement Activities

The design of the INE process model is represented

in BPMN (OMG 2006) (Figure 3), the de facto

standard for BP modelling. This conceptual model

was a candidate for improvement. We therefore

applied the three measurement activities previously

presented.

A) Measurement

We applied most of the measures published to date,

particularly those measures which had been

empirically validated. It was not possible to apply all

of them owing to the absence of certain elements in

this specific model. The results obtained are shown

in Table 4 (pair measure/result).

Table 4: Measurement results for the INE process.

Measure Result Understandability Modifiability

Nº of nodes 59 Fairly inefficient -

Density 0,02 -

Fairly

efficient

Sequentiality 0,396 Fairly inefficient

Fairly

inefficient

Connectivity

coefficient

1,54 Very efficient -

Mismatch

connector

16 Fairly inefficient

Fairly

inefficient

Control flow

complexity

22 Fairly inefficient

Fairly

inefficient

CLA 0,61 -

Very

inefficient

CLP 3 Fairly efficient -

TNE 5 Fairly efficient -

NSF 73 Very inefficient -

NMF 18 Fairly inefficient -

B) Evaluation

After obtaining the measurement results, we

evaluated them by following the threshold values

shown in Table 3. The conclusions were as follows:

 Number of nodes is 59, so the model is fairly

inefficient in understandability tasks

BUSINESS PROCESS MODEL IMPROVEMENT BASED ON MEASUREMENT ACTIVITIES

109

 Density is 0.02, so the model is fairly efficient in

modifiability tasks

 Sequentiality is 0.396, so the model is fairly

inefficient in understandability and modifiability

tasks

 Connectivity coefficient is 1.54, so the model is

very inefficient in understandability tasks

 Mismatch connector is 16, so the model is fairly

inefficient in understandability and modifiability

tasks

 Control flow complexity is 22, so the model is

fairly inefficient in understandability and

modifiability tasks

 CLA is 0.61, so the model is very inefficient in

modifiability tasks

 CLP is 3, so the model is fairly efficient in

understandability tasks

 TNE is 5, so the model is fairly efficient in

understandability tasks

 TNSF is 73, so the model is very inefficient in

understandability tasks

 NMF is 18, so the model is fairly inefficient in

understandability tasks

After the evaluation, we detected some potential

parts for alteration. For example, number of nodes

was a very high value, and could have compromised

the understandability of the model. The same applies

to connectivity coefficient, control-flow complexity,

CLA and TNSF, which obtained the worst results of

the measurement activity. On the other hand,

density, CLP and TNE obtained acceptable results

and did not need to be analyzed for further

improvement initiatives. These results guided us in

our definition of some proposals for redesign.

C) Redesign

After the selection of those parts of the INE model

that are potential elements for modification, the

redesign activity is carried out. This is the most

critical activity, since it depends on the successful

implementation of improvement activities.

Redesign was classified into two different

groups: changes proposed by specialists in

modelling tasks following guidelines of modelling

and changes proposed by health professionals.

Changes proposed by Health Professionals

Professionals at the hospital proposed certain

modifications which implied some differences in the

way in which some parts of the model were

designed.

The work group created to model tasks proposed

changes which produced several semantically

equivalent models. The Dephy method (Linstone

and Turoff 2002), was used to allow the work group

to select the most suitable changes. Each of these

changes produces a different version to the original,

specifically 4 different versions are generated:

A) The “belongs to a nursing unit with medical

dispenser” decision node was eliminated in

the immediately superior lane.

B) Some activities were added: “complete

pharmacy report” “send registration request

to pharmacy services”, “receive registration

in computer services”, “inform the

employee” in the immediately superior

lane.

C) The “belongs to a nursing unit” decision

node was eliminated and another decision

node was added in order to distinguish two

categories: planned or urgent in specific

superior lane.

D) Combination of version B and C.

The work group’s opinion and a first

application of the measures revealed that

version D is the best option, and we

Figure 4: Model of INE process applying changes proposed by health professionals.

ENASE 2011 - 6th International Conference on Evaluation of Novel Software Approaches to Software Engineering

110

selected it as the candidate for the improved

conceptual model. The results of these changes are

depicted in Figure 4. This change obtained better

results with regard to measures in comparison to the

original model. Table 5 shows the measures

analyzed and the results obtained. The measurement

values for the original model are shown in brackets

for the purpose of comparison. This comparison

shows an evident improvement in the model quality.

Table 5: Measure values of the improved model generated

by health professionals.

Measure Result

Mismatch connector 16 (15)

Control flow complexity 22 (21)

CLA 0,61(0,64)

NSF 73 (71)

Changes Proposed following Guidelines of

Modeling

On the other hand, the changes proposed by

modelling experts was based on the guidelines for

modellers published in (Mendling, Reijers et al.

2010). The following modifications were therefore

applied to the INE process model:

1. To reduce number of nodes:

a. Decompose a model with more than

50 elements.

b. Use one start and one end event.

2. To reduce TNF:

a. The elimination of some nodes

reduces the number of sequence

flows.

3. To reduce NMF:

a. The grouping of activities in a

subprocess reduces the number of

messages.

4. To reduce control-flow complexity and

mismatch connector:

a. Avoid OR routing elements.

5. A further improvement that is not taken into

account in the measures is “use verb object

activity labels”.

The proposed changes to the model are depicted in

Figure 5, and the measures’ improved results are

described in Table 6.

Table 6: Measure values of the improved model generated

by IT expert.

Measure Result

Nº of nodes 48(59)

Sequentiality 0,47(0,396)

Mismatch connector 10(16)

Control flow complexity 19(22)

TNE 3(5)

SF 63(73)

D) Selection of the improved Business Process

Design

The application of measures in both alternatives

allowed us to discover that the most acceptable

design is that obtained by professionals in

modelling. Specifically, 35% of the measures

analyzed improved their values when following

guidelines for modelling, as opposed to 23% of the

measures obtained when following the advice of

professionals in the health sector.

Figure 5: Version of INE process, including changes proposed following guidelines of modelling.

BUSINESS PROCESS MODEL IMPROVEMENT BASED ON MEASUREMENT ACTIVITIES

111

This signifies that the conceptual model depicted in

Figure 5 obtained better measurement results, thus

suggesting that the model is a good choice and can

increase the probability of obtaining a correct

process enactment.

4 IMPLICATIONS AND

LIMITATIONS

In this section we highlight some of the implications

and limitations of our research. In the previous

section, we described the process used to improve

conceptual models. In the first part, some measures

were applied to an INE process model, obtaining

certain measurement results. One limitation is

related to applied measures. Although more

measurement initiatives have been published, it is

not possible to apply them because of their lack of

empirical validation. This is an important

disadvantage in business process measurement and

may have limited our research.

On the other hand, measurement values were

assessed by following thresholds in order to guide us

in redesigning tasks. In a real situation

(Incorporation of a new employee) we had two

different initiatives for redesigning. One of them

was based on the opinion of health experts. After

seeing some business issues as a conceptual model,

represented in BPMN, they discovered that some

parts can be realised in a different way with the

same results. These changes to the original model

were made, and some improvements were made to

the measures (i.e. Control flow complexity was 21

rather than 22 in the original model). Nevertheless,

some improvement initiatives can be also be made

by following theoretical guidelines, with which even

better results are obtained (nº of nodes, sequentiality,

mismatch connector, control flow complexity, TNE

and NSF). These results reveal that theoretical

guidelines produce better modification proposals

than changes based on experience. Despite this

result, we believe that the changes proposed by

guidelines should not be applied in isolation, but

should be accompanied by the opinions of domain

experts. If the BP is modified by domain experts in a

controlled manner, it will be possible to avoid the

rejection of changes in the lifecycle enactment stage.

5 CONCLUSIONS AND FUTURE

WORK

We conclude this article by summarizing its

contributions and by providing an overview of future

research. We have discussed the importance of

measuring business processes, specifically in the

design and analysis stage, because it is known that

improving conceptual models in the first stage

implies several advantages in the case of avoiding

the propagation of errors to later stages, in which

their elimination might be more difficult and

expensive. This finding has a strong implication for

the way in which business process improvement is

confronted. A high-quality conceptual model can

therefore ensure an acceptable execution.

The experience report allows us to demonstrate

the practical utility of measurement activities,

obtaining a higher-quality model. The application of

measurement to conceptual models detected some

potential parts for alteration (number of nodes,

reducing sequence and message flow, reducing

decision nodes, or reducing number of events).

Guidelines of modelling also assisted us in making

these modifications. Finally, we obtained an

improved quality model which can ensure a better

execution.

As a future work, we wish to provide more

empirically validated measures in order to make the

measurement process more reliable. We also intend

to design more guidelines for inexpert modellers.

Finally, our idea is to apply measurement activities

in other real business processes at the hospital and in

other real organizations in order to ensure their

practical utility.

ACKNOWLEDGEMENTS

This work was partially funded by the following

projects: INGENIO (Junta de Comunidades de

Castilla-La Mancha, Consejería de Educación y

Ciencia, PAC 08-0154-9262); ALTAMIRA (Junta

de Comunidades de Castilla-La Mancha, Fondo

Social Europeo, PII2I09-0106-2463), ESFINGE

(Ministerio de Educación y Ciencia, Dirección

General de Investigación/Fondos Europeos de

Desarrollo Regional (FEDER), TIN2006-15175-

C05-05) and PEGASO/MAGO (Ministerio de

Ciencia e Innovación MICINN and Fondo Europeo

de Desarrollo Regional FEDER, TIN2009-13718-

C02-01).

REFERENCES

Becker, J., M. Rosemann, et al. (2000). Guidelines of

Business Process Modeling. Business Process

ENASE 2011 - 6th International Conference on Evaluation of Novel Software Approaches to Software Engineering

112

Management, Springer Berlin / Heidelberg. 1806:

241-262.

Bender, R. (1999). "Quantitative Risk Assessment in

Epidemiological Studies Investigatin Threshold

Effects." Biometrical Journal 41(3): 305-319.

Cardoso, J. (2006). "Process control-flow complexity

metric: An empirical validation." SCC '06:

Proceedings of the IEEE International Conference on

Services Computing: 167--173.

Coleman, D., B. Lowther, et al. (1995). "The Application

of Software Maintainability Models in Industrial

Software Systems." Journal of Systems and Software

29(1): 3-16.

Damij, N., T. Damij, et al. (2008). "A methodology for

business process improvement and IS development."

Information and Software Technology 50(11): 1127-

1141.

Delgado, A., F. Ruiz, et al. (2009). "MINERVA: Model

drIveN and sErvice oRiented framework for the

continuous business processes improVement &

relAted tools." Work. on Engineering Service-Oriented

Applications (WESOA’09), with ICSOC.

Huan, Z. and A. Kumar (2008). "New quality metrics for

evaluating process models." Business Process

Intelligence workshop.

ISO/IEC (2001). "9126-1, Software engineering - product

quality - Part 1: Quality Model."

Jung, J. Y. (2008). "Measuring entropy in business process

models." International Conference on Innovative

Computing, Information and Control 0: 246-252.

Krogstie, J., G. Sindre, et al. (2006). "Process models

representing knowledge for action: a revised quality

framework." Eur. J. Inf. Syst. 15(1): 91-102.

Laue, R. and J. Mendling (2009). "Structuredness and its

Significance for Correctness of Process Models."

Information Systems and E-Business Management.

Linstone, H. A. and M. Turoff (2002). "The Delphi

Method: Techniques and Applications." Addison-

Wesley.

McCabe, T. J. (1976). "A Complexity Measure." IEEE

Transactions on Software Engineering SE-2(4): 308-

320.

Meimandi Parizi, R. and A. A. A. Ghani (2008). "An

Ensemble of Complexity Metrics for BPEL Web

Processes." Ninth ACIS International Conference on

Software Engineering, Artificial Intelligence,

Networking, and Parallel/Distributed Computing:

753-758.

Mendling, J. (2008). Metrics for Process Models:

Empirical Foundations of Verification, Error

Prediction, and Guidelines for Correctness, Springer

Publishing Company, Incorporated.

Mendling, J., H. A. Reijers, et al. (2010). "Seven Process

Modeling Guidelines (7PMG)." Information and

Software Technology 52(2): 127-136.

Nejmeh, B. A. (1988). "NPATH: a Measure of Execution

Path Complexity and its Applications." ACM 31(2):

188-200.

OMG. (2006). "Business Process Modeling Notation

(BPMN), Final Adopted Specification." from

http://www.omg.org/bpm.

Park, R. E., W. B. Goethert, et al. (1996). "Goal-Driven

software Measurement: A Guidebook." HANDBOOK

CMU/SEI-96-HB-002.

Rolón, E., J. Cardoso, et al. (2009). "Analysis and

Validation of Control-Flow Complexity Measures

with BPMN Process Models." The 10th Workshop on

Business Process Modeling, Development, and

Support.

Rolón, E., F. García, et al. (2006). "Evaluation Measures

for Business Process Models." Simposioum in Applied

Computing SAC06.

Rolon, E., L. Sanchez, et al. (2009). "Prediction Models

for BPMN Usability and Maintainability." BPMN

2009 - 1st International Workshop on BPMN: 383-

390.

Sánchez-González, L., F. García, et al. (2010a).

"Assessent and Prediction of Business Process Model

Quality." CoopIS 2010 - 18th International conference

on Cooperative Information Systems: 78-95.

Sánchez-González, L., F. García, et al. (2010b).

"Measurement in Business Processes: a Systematic

Review." Business process Management Journal

16(1): 114-134.

Sánchez-González, L., F. Ruiz, et al. (2011). "Towards

Thresholds of Control Flow Complexity Measures for

BPMN Models." 26th Symposium On Applied

Computing SAC 10: in press.

Sánchez González, L., F. García, et al. (2010). "Assessent

and Prediction of Business Process Model Quality."

CoopIS 2010 - 18th International conference on

Cooperative Information Systems: 78-95.

Shatnawi, R. (2010). "A Quantitative Investigation of the

Acceptable Risk levels of Object-Oriented Metrics in

Open-Source Systems." IEEE Transactions on

Software Engineering 36(2): 216-225.

Sparks, G. (2000). "An Introduction to UML, The

Business Process Model." Enterprise Architect.

Vanderfeesten, I., J. Cardoso, et al. (2007). Quality

Metrics for Business Process Models. BPM and

Workflow Handbook 2007.

Vanderfeesten, I., H. A. Reijers, et al. (2008). "On a Quest

for Good Process models: the Cross Conectivity

Metric." International Conference on Advanced

Information Systems Engineering.

Wand, Y. and C. Weber (2002). "Research commentary:

Information systems and conceptual modeling–a

research agenda." Info. Sys. Research 13(4): 363--376.

Weske, M. (2007). Business Process Management:

Concepts, Languages, Architectures, Springer Verlag.

BUSINESS PROCESS MODEL IMPROVEMENT BASED ON MEASUREMENT ACTIVITIES

113