varBPM
A Product Line for Creating Business Process Model Variants
Andreas Daniel Sinnhofer, Peter P
¨
uhringer and Christian Kreiner
Institute for Technical Informatics, Graz University of Technology, Austria
{a.sinnhofer, christian.kreiner}@tugraz.at, p.puehringer@inode.at
Keywords:
Software Product Lines, Feature Oriented Modelling, Business Processes, Tool Integration.
Abstract:
Business processes have proven to be essential for organisations to be highly flexible and competitive in
today’s market. To manage the life-cycle from modelling such business processes over the execution and
the maintenance, Business Process Management Tools are used in the industry. In many cases, different
business processes do only vary in few points. This leads to the situation that new business process variants are
formed through copy or clone of previous solutions leading to a high number of instantiated process templates.
However, this means that changes to a template affects many processes, where all of them need to be manually
updated, which can lead to a considerable amount of work and money for a bigger company. In this paper,
we will present a framework for the integration of business process modelling tools and software product line
engineering tools to provide a systematic way to reuse and trace process variations of whole process families.
1 INTRODUCTION
Business Process (BP) oriented organisations are
known to perform better regarding highly flexible de-
mands of the market and fast production cycles (Mc-
Cormack and Johnson (2000); Hammer and Champy
(1993); Valena et al. (2013); Willaert et al. (2007)).
These goals are achieved through the introduction of
a management process, where business processes are
modelled, analysed and optimised in iterative ways.
Nowadays, the business process management is also
coupled with a workflow management, providing the
ability to integrate the responsible participants into
the process and to monitor the correct execution of
the business process in each process step. To adminis-
ter the rising requirements, so called business process
management tools are used (BPM-Tools) which cover
process modelling, optimization and execution. In
combination with an Enterprise-Resource-Planning
(ERP) system, the data of the real process can be in-
tegrated into the management process.
In many cases, business processes do only vary in
some points, which leads to the situation, that new
process variants are created through a copy and clone
of old solutions (often called as templates). As a
result, such templates are instantiated in many vari-
ous processes which makes the propagation of pro-
cess improvements time and cost intensive for a big-
ger company. Also the consistency of the documenta-
tion of this huge number of process variants is a chal-
lenging task.
Software Product Line Engineering (SPLE) tech-
niques have been successfully applied for almost any
domain, providing a technique for the systematic
reuse of domain artefacts. Although the topic of prod-
uct line techniques in the domain of business process
modelling is not new (e.g. Gimenes et al. (2008);
Rosa et al. (2008); Fantinato et al. (2012); Derguech
(2010)) only little work is found for the issues related
to the correct configuration of whole process families
(e.g. Hallerbach et al. (2009a,b)), the integration into
existing toolchains and the reuse throughout various
production plants. Thus, our approach is focused on
developing a framework for the integration of a SPLE
Tool and a BPM Tool, to provide a generic way to
generate process variants of whole process families.
In particular, we use the SPLE Tool for a system-
atically reuse of expert knowledge in form of valid
process variations, designed in an appropriated BPM
Tool. The integrity of the process variations is secured
by the capabilities of the BPM Tool and a rich con-
straint checking in the SPLE Tool. Furthermore, our
proposed approach enables the abilities to automati-
cally trace all process variants for an automatic prop-
agation of changes and process improvements and
the systematic integration into the capabilities of the
BPM Tools such as documentation generation, work-
flow engines, process optimisation tools, etc.
184
Daniel Sinnhofer A., PÃijhringer P. and Kreiner C.
varBPM - A Product Line for Creating Business Process Model Variants.
DOI: 10.5220/0005886901840191
In Proceedings of the Fifth International Symposium on Business Modeling and Software Design (BMSD 2015), pages 184-191
ISBN: 978-989-758-111-3
Copyright
c
2015 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
A B C
B_1
B_2
B_3
B_4
Sub-Process B
Sub-Process B
Figure 1: Principal structure of a business process accord-
ing to
¨
Osterle (1995) which is used for our approach. Start-
ing with an abstract description of the process, the tasks are
further described in sub-processes until a complete work
description is reached (microsopic level).
BPM Tool SPLE Tool
Business
Processes
Feature
Model
creates
modifies
derive
update
Feature
Selection
Transformation
Business
Process
Variant
derive
update
add
knowledge
add
input
influences
Process
Designer
Production
Experts
Production
Experts
Figure 2: Overall conceptual design. The upper side of the
Figure describes the domain engineering part and the lower
side of the Figure the application engineering part.
This work is structured in the following way: Sec-
tion 2 gives an overview over the concepts of tool in-
tegration and the design paradigm for business pro-
cesses which is needed for our framework. Section
3 presents the conceptual design of the framework
and states construction rules of the according feature
models and some design rules for the BPM Tool. In
section 4 we will introduce our case study regard-
ing some metrics and implementation details. Section
5 summarizes the related work and finally section 6
concludes this work and gives an overview of open
issues.
2 BACKGROUND
2.1 Tool Integration
According to the work of Karsai et al. (2005), two
possible patterns exists for tool integration. The first
approach is named ”Integration based on integrated
models” and is based on the idea of a common data
model which is shared between each participating
tool. This means that each tool needs two model
converters, one for the conversion of the native data
model into the common data model and one for the
opposite direction. The data is shared over a so called
integrated model server where each tool can publish
or consume data. For obvious reasons this approach
is meaningful applicable if each participating tool has
a similar data model. A drawback of this approach is
that it does not scale very good with the number of
connected tools.
The second pattern is named ”Integration based
on process flows” and is based on the idea of a point
to point message based communication. Each partic-
ipating tool registers itself at a backplane providing
information about what data is shared and what data
is intended to be consumed. As a result, this approach
scales better with the number of participants since po-
tentially fewer model transformations are needed at
which each model can be better optimized regarding
the communicating tools. This approach is used in
those situations, where the data is processed in a spe-
cific sequence.
For our framework both patterns would be possi-
ble. Due to the fact, that the number of participants is
small (in most circumstances there is only one SPLE
Tool and one BPM Tool) and since the representation
of business processes is very similar throughout vari-
ous tools, the first approach is more applicable.
2.2 Business Processes
A business process can be seen as a sequence of
tasks/sub-processes which needs to be executed in a
specific way to produce a specific output which is of
value to the costumer (Hammer and Champy (1993)).
According to
¨
Osterle (1995) the process design on the
macroscopic level (high degree of abstraction) is split
up into sub-processes until the microscopic level is
reached. This level is reached, when all tasks are de-
tailed enough, so that the process employees can use
it as work instructions.
In other words, a complete business process is
designed in layers, where the top layer is a highly
abstracted description of the overall process, while
the production steps are further refined on the lower
levels. As a result, the lowest level is highly de-
pendable on the concrete product and production en-
vironment, providing many details for the employ-
ees. In fact the top layers are mostly indepen-
varBPM: A Product Line for Creating Business Process Model Variants
185
dent from the concrete plant and the supply chain
and could be interchanged throughout the production
plants, whereas the lower levels (the refinements) of
the processes would need to be reconsidered. Figure
1 gives an overview of such a structure. Variability
of such a process structure can either be expressed
through a variable structure of a process/sub-process
(e.g. adding/removing nodes in a sequence) or by
replacing the process refinement with different pro-
cesses. The current version of our developed proto-
type focuses on the second method but the framework
is not limited to it.
2.3 Informal: Feature Model
A feature is defined by Kang et al. (1990) as a ”promi-
nent or distinctive user-visible aspect, quality, or
characteristic of a software system or system”. In
context of a Software Product Line, a feature model is
a model which defines all these features and explicitly
states their relationships, dependencies and additional
restrictions between each other. It enables the ability
to visually represent the variable parts of a system and
the options available for all products of a product line.
3 VARIABILITY FRAMEWORK
3.1 Conceptual Design
The overall conceptual design is based on a feature
oriented domain modelling approach and is displayed
in Figure 2. It is intended, that the domain experts
(process designer) design process templates in the ac-
cording BPM Tool, providing also all needed infor-
mation for e.g. a workflow engine. Based on these
templates and the abstract process model (the top
level description of the process) a feature model is
partially automatically created/updated with the guid-
ance of the domain experts. This is done by identifica-
tion of the variation points and the linkage of the ac-
cording variations on every process level. Each varia-
tion can contain additional variability by either defin-
ing new variation points where further refinement can
be linked to or by a variable process structure. Ad-
ditional constraints regarding the possible combina-
tion of features are intended to be modelled in the
SPLE Tool, but are not limited to it. In application
engineering the domain experts (not necessarily a pro-
cess designer, but someone who knows the current
needs of the production) adds his knowledge and se-
lects the needed features. The found description is
then automatically transformed in a real business pro-
cess which can be executed by the workers. During
Table 1: Needed process information within the SPLE Tool.
property
name
description
id A unique id to identify the pro-
cess/task
category id A unique id of the category of the
process
display
name
A human readable and understand-
able name of the process/task
children A list of ids which references the
processes/tasks of the process itself
(empty if the microscopic level is
reached)
additional
data
A list of additional data which is
needed for the concrete instantia-
tion of the process. E.g. data for
a workflow engine, the responsi-
ble workers, etc. This data can be
provided through the BPM Tool
(almost static) or can be design
variable in the SPLE Tool.
the process execution, loads of data is generated re-
garding the performance and efficiency of the process.
Thus, it is possible that some additional information
is added to the derived processes which leads to a
possible influence of the according business process
templates or a possible influence of the feature selec-
tion. This flowback mechanism is an important task
and needs to be considered for the maintenance and
the evolution throughout the lifecycle of a process.
For illustration a short example for the flowback
mechanism is given: Task B of the process displayed
in Figure 1 could be dependent on the logistic chain of
a supplier of a specific part or material. If during the
execution of the process the supplier is changed, it is
also likely that the overall control of the logistic chain
is changed due to the fact that the newly integrated
supplier can only deliver goods in a specific way. If
the process of the logistic chain was not already mod-
elled, then the process designer would need to create
a new process variant and would need to update the
existing feature models first. After this is done, our
proposed toolchain needs to update the feature selec-
tion of the respective process in the SPLE Tool. This
automatically updates the structure of the overall pro-
cess variant, leading to an almost on the fly update of
the complete process workflow especially when the
process variation was already modelled.
Summarizing this concept means that the SPLE
Tool is responsible for the following points: It needs
to assist the domain experts (process designer) during
Fifth International Symposium on Business Modeling and Software Design
186
settings root process variation point variation
addNode
deleteNode
1 1...*
1
0...*
1
0...1
1
0...1
1
0...*
0...*
1
1
0...*
1 1...*
1
1
0...*
0...*
Figure 3: Feature model construction rules. The white parts
cover the variability on the process level itself and the dark
grey parts cover the variability between each layer of the
overall process structure.
the (partially) generation of the feature models and
during the selection and creation of the concrete prod-
uct variant. Furthermore, it has to keep track of all
generated process variants to automatically apply pro-
cess template improvements or changes to the overall
process structure and to flow back information added
during the execution in the BPM Tool. To do so, the
tool needs information about the properties of the pro-
cess displayed in Table 1. The BPM Tool is respon-
sible for the creation of semantically correct process
variants and to provide capabilities which are of value
for the developing company e.g. automatic documen-
tation generation, workflow engine, etc. For obvious
reasons, each Tool must have rich import/export capa-
bilities or the ability to integrate user defined plug-ins
to extend the functionality.
3.2 Type Model
To model the variability within the SPLE Tool in a
structured way, the feature model should support the
following type models:
settings: Is a data model, containing information
for the tool adaptors to identify the right datasets
(e.g. identifier of the database of the BPM Tool
from which the process structure is imported; log-
in settings, etc.)
root process: A process model of the top level
process and therefore an abstract description of
the overall process sequence. It consists of var-
ious nodes where some of them deal as variation
points.
variation point: A node in a process where at
least one variation can be linked to.
variation: A process model for a task or a sub-
process which can be linked to a number of vari-
ation points. If it is a process, it can also con-
tain variation points or a variable process structure
(addition/deletion of nodes).
addNode: Adds a node (task or process model) at
a specific location of the process structure. This
added node can also be a variation point for fur-
ther refinements. The addition of the node can be
dependent on the feature selection.
deleteNode: Deletes a node (task or process
model) at a specific location of the process struc-
ture. The deletion of the node can be dependent
on the feature selection and hence it is also possi-
ble to link further refinements to this node.
The construction rules for this type model can be
seen in Figure 3. The settings node is only instantiated
once in such a model. The root process is somehow
very similar to a variation, but with the difference that
it must contain variation points (to prohibit a ”Blob”
anti pattern [Brown et al. (1998)]).
3.3 Design Rules for Business Processes
Aforementioned, the processes should be designed
as stated by
¨
Osterle (1995). Secondly we have no-
ticed, that almost every bigger BPM Tool supports
the assignment of specific group identifiers to groups
of processes, providing a more structural design of
the processes. Thereby it is possible to automatically
map specific groups of processes to a specific varia-
tion point. This leads to the situation that new varia-
tions can be automatically detected and can be advo-
cated for an integration into existing feature models.
Furthermore, the structuring in groups of processes
enables the ability to introduce a constraint check so
that variation points are limited to specific groups of
processes. This increases the assurance in the creation
of semantically valid processes.
4 INDUSTRIAL CASE STUDY:
VARBPM
In this section an overview over our industrial case
study is given, which describes the domain of our in-
dustry partner and the developed toolchain.
4.1 Industrial Project Partner
Our project partner Magna Cosma
1
is an interna-
tional company in the metal stamping and assembly
industry – specialised on class-A car body panels and
closure parts (e.g. doors) with several plants all
over the globe. The implemented business processes
1
http://www.magna.com/de/kompetenzen/karosserie-
fahrwerksysteme
varBPM: A Product Line for Creating Business Process Model Variants
187
are mostly controlled by an SAP infrastructure and
are designed with the BPM-Tool Aeneis
2
. Although
some plants are specialised on the same production
parts, almost every plant develops and maintains their
own business processes, which makes it difficult to
compare processes, mark bottlenecks, optimise the
processes and publish the changes to other plants.
4.2 Tool Integration
As mentioned before, the tool integration of the SPLE
Tool (pure::variants
3
) and the BPM Tool (Aeneis) is
based on the Pattern ”Integration based on Integrated
Models”. The integrated data model contains the rel-
evant data enumerated in Table 1 where all fields are
of type String respectively an array of Strings for the
children and data field. This means that the native
data model of the SPLE Tool is the integrated data
model and hence only the BPM tool adaptors need
to implement a data conversion. To support an up-
date mechanism without sending the complete pro-
cess, each published dataset can be assigned to a spe-
cific type indicating what should happen with this
dataset. Possible types are:
New: Indicating that this dataset was not pub-
lished before and hence it should be integrated di-
rectly just as is.
Update: Contains the id of the according dataset
and the data which shall be updated. Non speci-
fied attributes are not affected.
Remove: Contains the id of the according dataset
which should be deleted out of the system. Linked
variations of such a process are not affected.
The developed tool adapters are also applicable to
get notified when data is added/updated so that such
changes are processed almost immediately. If this
mechanism is not supported by the participating tool
connector, an operator needs to trigger this update
mechanism manually. In the current development, the
SPLE Tool needs to check manually for updated data
since this task is intended to be supported by an do-
main expert, whereas the BPM Tool uses the benefits
of the immediate notification system. The communi-
cation of the tools is done by an XML based file ex-
change and due to some consistency issues the com-
munication is only possible if both tools are running.
I.e. deriving process variants is only possible if the
BPM Tool is running too.
2
http://www.intellior.ag
3
http://www.pure-systems.com
4.3 pure::variants
pure::variants is a feature oriented domain modelling
tool and is based on Eclipse. As such, it can easily be
extended based on java plug-in development. During
the implementation of this project, five different plug-
ins where developed:
An import plug-in, which is capable of importing
the process structure - including the definition of
variation points and the according variations - and
converting it into a feature model compliant to the
construction rules displayed in Figure 3 without
the white parts.
An extension to the internal model compare en-
gine so that different versions of created feature
models can be compared with each other.
An update mechanism to automatically search
for deleted / added variations or updated process
structures, providing graphical assistance for the
domain expert.
An extension to the internal model transformation
engine so that the feature selection is automati-
cally converted into a business process in the com-
mon data model; This process is then delivered to
the attached BPM Tools, so that a native version of
the process can be created/updated and executed.
Additions to the internal model check engine
to model and create only valid processes (e.g.
checks related to the feature selection, the consis-
tency of the feature model, etc.).
To keep track of all generated business process
variants, a list including all ids of the processes is
stored and maintained in a file located in the same
directory as the variant description model (feature se-
lection), but hidden from the user perspective. This
list is automatically updated when a process variant is
deleted in the BPM Tool or created with the SPLE
Tool. pure::variants also provides a framework for
the comparison of different models, which enables the
ability to compare different process variants in an ef-
ficient way.
4.4 Studied Use Cases and Results
In the list below, use cases can be found which we
investigated during the development of our approach
regarding the performance of our toolchain (time sav-
ing). The according results are displayed in Figure
5, where the white bars are related to the manual
case and the grey bars to our approach. Although the
varBPM approach automatically integrated each de-
rived process into the capabilities of the BPM Tool
Fifth International Symposium on Business Modeling and Software Design
188
Root
VP 1 VP 2
Var
01
Var
02
Var
03
Var
04
Var
05
Var
06
VP 3 VP 4
Var
07
Var
08
Var
09
Var
10
Var
11
Figure 4: The variability structure of the process for the
evaluation examples.
(documentation generation, workflow engine), the
manual approach only reflects the time used for the
creation of the bare process structure. The used pro-
cess structure is part of a bigger process used for an
on-demand manufacturing of spare parts for different
car manufacturer.
Use case 1: For this use case, the time was mea-
sured that a domain expert needs to create a new
process variant manually or using an existing feature
model. The process setup consisted of four variation
points (two top-level variation points and two varia-
tion points on lower levels) where a total number of
twenty-seven different process variants were possible
(for illustration, the variability structure can be seen
in Figure 4). The number of other processes in the
database of the BPM tool was considered to be low
(20 other processes). The experiment was repeated
with different experts and different process setups (the
overall variability structure was kept the same but the
process structures changed). The results of this use
case are divided with the number of variation points
to get a rough estimate for the time saving per varia-
tion point.
Use case 2: This use case is an addition to the
first one, with the difference of a high number of other
processes in the database (200 processes).
Use case 3: Is related to the topic of maintenance.
In this scenario a process template was changed and
all process variants should be updated. The domain
expert was told that there is a number of six variants
he needs to update providing only the name and the
id of the changed process template. The size of the
database was limited to 50 processes. The change to
0 10 20 30 40 260
Use case 1
Use case 2
Use case 3
Use case 4
Use case 5
Use case 6
Time in seconds
Manual Manipulation
Using varBPM
Figure 5: The results of the evaluation use cases. The grey
parts are the time spent using the varBPM approach and the
white parts states the time taken if a manual manipulation is
used.
the template was a deletion/addition of a node. The
results displayed in Figure 5 are normed to the update
of one process.
Use case 4: This use case is an addition to the
previous case, but with the difference that the domain
expert was now told how the process variants where
called (assuming that the domain expert was responsi-
ble for the creation of the process variants and exactly
knows the variants).
Use case 5: In this situation, a new process varia-
tion was created by a process designer and the domain
expert should now derive a new product variant out of
it (in fact, a new variation for the Variation Point ”VP
3” was developed). It is assumed that the processes
are designed according to our proposed design rules.
The number of other processes in the database was
considered as low (20 processes). This use case also
gives a good estimate on how much overhead is pro-
duced, to update an existing feature model.
Use case 6: In this case, a new feature model was
developed according to the variability structure dis-
played in Figure 4 to generate a metric on how much
overhead is produces for the initial creation of the fea-
ture model. To get a rough estimate on the overhead
per variation point, this number is divided by four.
The results of ”Use case 2” were surprising, since
our developed approach performs worse in relation
to the previous scenario. The reason for this is that
the project now consists of ten times more processes,
leading to a more time demanding search for the right
processes. For humans, it was still quite easy to find
the right processes since they were organised in a
clear ”human understandable” manner. As a result the
varBPM: A Product Line for Creating Business Process Model Variants
189
increasing number of processes do not have a high in-
fluence to a manual manipulation if the processes are
structured in a clear and meaningful way. Otherwise
the time would increase more significantly.
To get a rough estimate of the break-even point,
the following equation is used:
p
Overall Overhead
Average time saving
=
259
4 · 29
2.2 (1)
The overall overhead is the time spent to create
the feature model (”Use case 6”) and the average time
saving is calculated using the average time saving
per variation point multiplied by the number of varia-
tion points. Interestingly, when software product line
techniques are applied to pure software systems, the
break-even point is also located at around three sys-
tems (according to Pohl et al. (2005)).
4.5 Restrictions
Depending on the API of the used BPM Tool, your
approach can be limited in terms of the available fea-
tures. This means that if the BPM Tool only provides
access to the basic process structure, our framework
is limited to the creation of derived processes with-
out the ability to automatically integrate the models
into the capabilities of the BPM Tool (e.g. workflow
engine).
5 RELATED WORK
As stated in the survey of Fantinato et al. (2012), ma-
jor challenges in the field of business process variabil-
ity modelling are related to the reaction time of pro-
cess changes and of the creation and selection of the
right business process variants, which are also main
topics in our approach.
Derguech (2010) presents a framework for the
systematic reuse of process models. In contrast to
our approach, it captures the variability of the pro-
cess model at the business goal level and describes
how to integrate new goals/sub-goals into the existing
data structure. The variability of the process is not
addressed in his work.
Gimenes et al. (2008) presents a feature based
approach to support e-contract negotiation based on
web-services (WS). A meta-model for WS-contract
representation is given and a way is shown how to in-
tegrate the variability of these contracts into the busi-
ness processes to enable a process automation. It does
not address the variability of the process itself but en-
ables the ability to reuse business processes for differ-
ent e-contract negotiations.
While our approach reduces the overall process
complexity by splitting up the process into layers with
increasing details, the PROVOP project (Hallerbach
et al. (2009a,b) and Reichert et al. (2014)) focuses
on the concept, that variants are derived from a basic
process definition through well-defined change oper-
ations (ranging from the deletion, addition, moving
of model elements or the adaptation of an element at-
tribute). In fact, the basic process expresses all possi-
ble variants at once, leading to a big process model.
The work of Gottschalk et al. (2007) presents an
approach for the automated configuration of workflow
models within a workflow modelling language. The
term workflow model is used for the specification of
a business process which enables the execution of it
in an enterprise and workflow management system.
The approach focuses on the activation or deactiva-
tion of actions and thus is comparable to the PROVOP
project for the workflow model domain.
Rosa et al. (2008) extends the configurable pro-
cess modelling notation developed from Gottschalk
et al. (2007) with notions of roles and objects provid-
ing a way to address not only the variability of the
control-flow of a workflow model but also of the re-
lated resources and responsibilities.
The work of Leitner and Kreiner (2010) addresses
the process variability through a bottom up approach
by examining the possible configurations through the
scan of the according ERP System (SAP). In contrast
to this approach, we focus on an top down method to
abstract the complexity of the underlying ERP Sys-
tem.
The Common Variability Language (CVL Haugen
et al. (2013)) is a language for specifying and resolv-
ing variability independent from the domain of the ap-
plication. It facilitates the specification and resolution
of variability over any instance of any language de-
fined using a MOF-based meta-model. A CVL based
variability modelling and a BPM model with an ap-
propriate model transformation could lead to similar
results as presented in our work.
6 CONCLUSION AND OUTLOOK
The reuse of business process models is an important
step for an industrial company to survive in a com-
petitive market. With our work we have proposed a
way to combine the benefits of software product line
engineering techniques with the capabilities of a busi-
ness process modelling tool to provide a framework
for the systematic reuse of business processes. With
the proposed design rules, our approach results in an
automatic detection and propagation of new and/or
Fifth International Symposium on Business Modeling and Software Design
190
changed business process variations. On the other
hand it leads to an automatic integration of new as-
sembled process variants into the BPM capabilities
such as an automatic integration into a workflow en-
gine, integration of responsibilities and resources, etc.
Our developed framework covers the variability of
the process in two different ways: Through the link-
age of different process variations to variation points
and through a variable process structure (deletion / ad-
dition of nodes) in each layer. Due to the fact that
our developed framework is in an early stage of us-
age, further research efforts would address the col-
lection and evaluation of data regarding the evolution
and maintenance of the process models. In this con-
text, an integration of Six Sigma
4
into our framework
is aimed to provide a complete framework from mod-
elling and improving process models. Additionally
the customization of the ERP system of the underly-
ing system (in this case SAP) is an interesting topic,
providing a complete framework for the topics of pro-
cess modelling, execution and maintenance including
the planing of the resources of the concrete produc-
tion facility.
ACKNOWLEDGEMENT
We want to gratefully thank Magna Cosma for spon-
soring this project, Danilo Beuche from pure::systems
and Intellior AG for their support.
REFERENCES
Brown, W. J., Malveau, R. C., McCormick, H. W. S., and
Mowbray, T. J. (1998). AntiPatterns: Refactoring
Software, Architectures, and Projects in Crisis: Refac-
toring Software, Architecture and Projects in Crisis.
John Wiley & Sons.
Derguech, W. (2010). Towards a Framework for Business
Process Models Reuse. In The CAiSE Doctoral Con-
sortium.
Fantinato, M., Toledo, M. B. F. d., Thom, L. H., Gimenes, I.
M. d. S., Rocha, R. d. S., and Garcia, D. Z. G. (2012).
A survey on reuse in the business process management
domain. International Journal of Business Process In-
tegration and Management.
Gimenes, I., Fantinato, M., and Toledo, M. (2008). A Prod-
uct Line for Business Process Management. Software
Product Line Conference, International, pages 265–
274.
4
Six Sigma is a set of techniques and tools for a system-
atic management of process improvements based on quality
management methods, including some statistical methods.
Gottschalk, F., van der Aalst, W. M. P., Jansen-Vullers,
M. H., and Rosa, M. L. (2007). Configurable Work-
flow Models. International Journal of Cooperative
Information Systems.
Hallerbach, A., Bauer, T., and Reichert, M. (2009a). Guar-
anteeing Soundness of Configurable Process Variants
in Provop. In Commerce and Enterprise Computing,
2009. CEC ’09. IEEE Conference on, pages 98–105.
IEEE.
Hallerbach, A., Bauer, T., and Reichert, M. (2009b). Issues
in modeling process variants with Provop. In Ardagna,
D., Mecella, M., and Yang, J., editors, Business Pro-
cess Management Workshops, volume 17 of Lecture
Notes in Business Information Processing, pages 56–
67. Springer Berlin Heidelberg.
Hammer, M. and Champy, J. (1993). Reengineering the
Corporation - A Manifesto For Business Revolution.
Harper Business.
Haugen, O., Wasowski, A., and Czarnecki, K. (2013). Cvl:
Common variability language. In Proceedings of the
17th International Software Product Line Conference,
SPLC ’13.
Kang, K., Cohen, S., Hess, J., Novak, W., and Peterson,
A. (1990). Feature-oriented domain analysis (foda)
feasibility study.
Karsai, G., Lang, A., and Neema, S. (2005). Design pat-
terns for open tool integration. Software & Systems
Modeling, pages 157–170.
Leitner, A. and Kreiner, C. (2010). Managing erp configura-
tion variants: An experience report. In Proceedings of
the 2010 Workshop on Knowledge-Oriented Product
Line Engineering, KOPLE ’10, pages 2:1–2:6.
McCormack, K. P. and Johnson, W. C. (2000). Business
Process Orientation: Gaining the E-Business Com-
petitive Advantage. Saint Lucie Press.
¨
Osterle, H. (1995). Business Engineering - Prozess- und
Systementwicklung. Springer-Verlag.
Pohl, K., B
¨
ockle, G., and Linden, F. J. v. d. (2005). Soft-
ware Product Line Engineering: Foundations, Princi-
ples and Techniques. Springer.
Reichert, M., Hallerbach, A., and Bauer, T. (2014). Lifecy-
cle Support for Business Process Variants. In Jan vom
Brocke and Michael Rosemann, editor, Handbook on
Business Process Management 1. Springer.
Rosa, M. L., Dumas, M., ter Hofstede, A. H. M., Mendling,
J., and Gottschalk, F. (2008). Beyond control-flow:
Extending business process configuration to roles and
objects. In Li, Q., Spaccapietra, S., and Yu, E., editors,
27th International Conference on Conceptual Mod-
eling (ER 2008), pages 199–215, Barcelona, Spain.
Springer.
Valena, G., Alves, C., Alves, V., and Niu, N. (2013). A
Systematic Mapping Study on Business Process Vari-
ability. International Journal of Computer Science &
Information Technology (IJCSIT).
Willaert, P., Van Den Bergh, J., Willems, J., and De-
schoolmeester, D. (2007). The Process-Oriented Or-
ganisation: A Holistic View - Developing a Frame-
work for Business Process Orientation Maturity.
Springer.
varBPM: A Product Line for Creating Business Process Model Variants
191