A Process Variant Modeling Method Comparison Exper
A Process Variant Modeling Method Comparison Exper
net/publication/301518188
CITATION READS
1 1,081
3 authors, including:
Banu Aysolmaz
Eindhoven University of Technology
53 PUBLICATIONS 318 CITATIONS
SEE PROFILE
All content following this page was uploaded by Banu Aysolmaz on 30 October 2017.
1 Introduction
In enterprises, business process modeling (or process modeling for short) is of great
importance to reveal processes and develop business process management systems
(BPMS). In process modeling, one of the problems that analysts encounter is the need to
deal with process variability. Due to the diversity in business contexts, variants of the
same process may be modelled and used in multiple cases in the same organization [1].
This diversity may be caused by various factors such as differences in delivered prod-
ucts, customer types, and divergent business requirements in countries. When such
factors are present, consideration of process variants during process modeling is
inevitable [2]. However, in the design of a process model, it is a challenging task to
either maintain variants of the same process separately while managing the relations
between them or integrate the process variants into a single model while preventing
complexity and redundancy [2].
exploit process knowledge obtained from previous similar companies for new cus-
tomers. Based on the problem, the need for using a process variant modeling method for
4S can be summarized as follows:
• When they start to work with a new customer, 4S analysts need to combine their
knowledge on previous customers as a baseline for understanding the new as-is
process and suggesting improvements. Analysts would be better off if they would
have an integrated model, which they can practically use as a jump start in the
project initiation phase.
• Through the steps of process analysis, improvement and design, 4S analysts design
various processes for customers. Even when they start developing a process based
on a previously encountered process, the knowledge of such related processes and
the connections hereto are soon lost. Analysts cannot benefit from one another’s
experiences as it is hard for them to go over each process to find out if it is relevant
for a new case. The same problem persists through process enactment phase; as
developers cannot easily find out similar automated processes and activities for
example, to reuse their form design and flow logic.
• When an improvement or update is needed, 4S needs to go over each customer’s
processes to find out which ones are affected and where updates are needed. This
requires a lot of effort and can introduce errors due to manual review process.
not inclusive of all activities – it is rather a brief process model including must-have
activities. For this reason, we needed a method that has more flexibility to define
process variants. Another criteria for method selection was on the need of a variant
modeling tool. For some approaches, a tool that has specific process variant modeling
features are required to properly benefit from the approach [7]. Due to the concern of
increase in effort by adding a new tool to company repository and training needs, 4S
eliminated the methods that need a specific tool.
Considering these needs and limitations, we identified an initial list of process
variant modeling methods based on existing literature reviews [3] and our review of the
related work. We discussed potential pros and cons of these methods with the 4S team
and made a joint selection. As stated, we selected the Decomposition Driven method
and the Provop method.
The Decomposition Driven method was selected because it provides flexibility for
certain parts of the model. By means of step-wise decomposition, users can choose to
model some of the sub-processes together and some others separately [5]. Moreover,
the team specifically appreciated how the method does not only approach identification
of variants mechanically but it considers the wider business environment via business
drivers and syntactic drivers inherent to the processes. In turn, the Provop method was
selected due to its robust mechanism to treat all variants equally and create a big model.
The usage of the list of options to mechanically end up in new variants in a plug and
play logo-like feeling was seen as another advantage. The team focused on selecting
methods that have a different approach for variant modeling. In this way, a comparison
of the benefits of different process variant modeling approaches for future use would be
feasible.
international company. For all of them, their software project management processes
are defined based on PMI’s PMBOK guide [8]. One of the authors of this paper who is
affiliated with 4S, worked as a leader of the team in the company that implements the
methods and evaluates the results.
Although 4S uses the PMBOK guide as the baseline, the best practices provided
just the essential steps of a project management process. The processes were defined as
workflow definitions on HP PPM, but process models were not developed for analysis
purposes. We converted the low level workflow models to process models in BPMN
notation through discussion sessions with the experts. We aggregated workflow tasks to
higher level activities in BPMN. The experts found it easier to observe and define
relations between process variants using the BPMN models. For this reason, we
decided to use these models in variant modeling activities. For each variant, we
developed a high level software project management process. We created a relation
table for the corresponding workflow tasks for each BPMN activity. In this way, we
achieved more comprehensible process definitions where the experts could better
observe the relations between the variants. Still, we are able to map process model
activities to workflow tasks via the relation table. This enabled the experts to analyze
workflow definitions together with variant models after the study is completed.
A summary of the companies and their process metrics can be found in Table 1.
Process models for all process variants can be seen in [9].
The described research project was initiated based on the need in 4S as defined in
previous sections. After the analysis of related work, elimination took place of multi-
model approaches and approaches that require usage of a specific variant modeling tool
and that conduct automated process discovery. Subsequently, the approaches men-
tioned earlier were selected. Upon the selection of the methods, the following steps
were planned:
290 B. Aysolmaz et al.
• Identify the process to apply selected process variant modeling methods: The team
selected the software project management process, which is the most frequent
process that they provide consultancy for their customers.
• Identify the context for application: Five customers were identified that are repre-
sentative for different industries. Two of the customers implement two variants of
software project management process.
• Define process models for each variant: We developed process models in BPMN
notation for each customer as described at the beginning of this section.
• Apply the Decomposition Driven and the Provop methods to develop process
variant models: The team conducted the relevant steps for applying the two methods
as described in the following sections. Two methods were applied in parallel to
prevent the effect of the learning curve.
• Evaluate the process of method application and outputs: The team collected data on
the effort spent on each method and compared the outputs. In addition to comparing
the outputs and facts, we conducted interviews to understand how the experts
interpret the usability, complexity and efficiency.
• List the guidelines to choose proper method: The team identified the benefits and
disadvantages brought by the two methods and how can one select the proper
method with respect to priorities and benefits expected.
The method starts with the definition of a main top-level process [5]. Then, each
activity in the main process is defined in detail in a sub-process. Later, the
sub-processes is further decomposed into sub-processes until there is no meaningful
decomposition possible. At every level, the so-called variation map is created which
contains activities and relations necessary to configure every variant. In the following
sections, we describe the conduct of each step as prescribed by the method [5].
them to flexibly develop the models [5]. Business drivers are determined based on
factors such as: resources used, products and services produced, customers, countries.
In our case, we focused on how the high level activities in the main process are
performed and possible causes of variation. We observed that the main cause of
variation is the variety of customers. Another driver is identified as location of the
services. This driver is used to differentiate the processes of Company 4, leading to
variations of national and international services.
Syntactic drivers are the second type of drivers which diversify the way multiple
variants produce their outcomes. They are defined based on the similarity of the process
models of the variants. The method allows consolidation or separation of variant
models due to syntactic drivers. In 4S, we manually assessed the similarity of process
variant models with respect to the main process modeled in Fig. 1 [9]. We conclude
that there is no explicit syntactic driver, as the main process can be used to reach the
variant models by mostly adding nodes and alternative paths to the main process.
variant map [5]. For example, Moderate Initiation and Detailed Initiation variants were
merged as they were assessed to be similar. For both Plan Resources and Close Project
activities, a different subprocess was defined for each variant. Only two variants among
six were assessed to be similar for both activities. Rest of the subprocesses had very
strong drivers and were assessed to be not similar. Thus, there were five variations of
these activities as seen on the variation map. The details on activities in the subpro-
cesses, the similarity decisions for activities and the merging of subprocesses can be
seen in [9].
After this step, the Decomposition Driven method suggests the iteration of all steps
for the subprocesses of the main process. We applied the Decomposition Driven
method completely in the first level of decomposition in 4S. Moreover, we identified
the activities to be placed in each subprocess and discussed a sketch of the variation
maps with the experts. In this way, the experts were able to observe how the
Decomposition Driven method provided a flexible way of variant modeling in different
granularity levels. For example. For “Implement” process, variation in the high level is
not found necessary. However, it is observed that variants of this subprocess need to be
handled considering other business drivers such as project type.
294 B. Aysolmaz et al.
The Provop method focuses on creating a single base process model which includes
adjustment points and their related sets of options [6]. The options include a set of
atomic operations such as insert, delete, move and modify; which are used to configure
the base model to reach a certain variant. Defining the set of operations options pro-
vides a reusable mechanism to define common operations for multiple variants. This
mechanism decreases the complexity and increases controllability to configure a
variant. Moreover, the Provop method can support automated variant configuration by
defining context-aware configuration options.
Fig. 5. Base process model with adjustment points (Color figure online)
done by asking users to manually choose specific variants, which is hard if there are a
lot of options and specialized knowledge is required. To overcome the problem, the
Provop suggests the definition of context rules by identifying, for each option, the
context in which the options are applicable. In our case, the available knowledge on
business drivers became useful to define the context. For each option, we identified the
set of variants that are to be configured via this option. This can be seen in Fig. 6 as
context rules.
Another point to be considered while applying the options is the possible con-
straints with the options. For example, there may be implication relation between
options, an option implying the usage of another one [1]. We had an order constraint
for options 5 and 6, as option 5 always needs to be applied before 6. We observed that
the modelers need to pay special attention for constraints especially for options
effective on the same adjustment point pairs.
In conformance with the constraints, we manually apply the set of options shown in
Fig. 6 to the base process as indicated with red markers on Fig. 5 to achieve the variant
process of company 2 as in Fig. 7. In the following section, we evaluate the results of
applying the two methods and provide a guideline for selection.
After the definition of the base process model with adjustment points for the
software project management process, we needed to analyze variants for subprocesses
of the activities in the base process model. We were not able to identify specific
guidelines for applying the Provop method in a hierarchical process structure. We plan
to define subprocesses for each variant and conduct the same set of steps to develop
base process models of each subprocess. However, we need to consider that new
activities may be added to the high level process via options. In this case, we plan to
define base process maps for the subprocesses under those activities as well. This will
introduce problems in reading, as the user is not able to see and associate such a
subprocess in the base process model. We also observed that special attention is needed
to prevent conflicts among options for different levels of granularity.
As 4S is evaluating two variant modeling methods to implement in all its projects in the
future, it is important to identify the method that is practical to apply and meets the
needs for reusing process knowledge and maintaining multiple variants. For practical
reasons, we evaluated the effort spent on applying the two methods, structure of the
outputs and flexibility of using the outputs in new projects and maintaining them when
there is an update in one of the variants.
22 h were spent in 5 sessions for the Decomposition Driven method, whereas 15 h
were spent in 4 sessions with the Provop method. The experts appreciated the idea of
incorporating the business context to identify sources of variation. However, the variety
of customers was already an explicit business driver for 4S from the start of the study.
The experts think that their extra effort for the Decomposition Driven method will pay
off when they implement the method for low level subprocesses and other process
types with potentially more varied business drivers.
Comparing the structure of the resultant models, variation map of the Decomposi-
tion Driven method has 25 activities, 10 gateways and 50 edges. The Provop method
produced a simpler model with only 9 activities, 2 gateways, 13 edges and 11 adjust-
ment points as customized elements. The Decomposition Driven method seems to
produce a bigger and more complex model (due to edges/activities ratio). However there
is an extra artefact, list of options, required to read and customize the Provop base
process model. The experts indicated that it was easier for them to read the Provop’s
base process model and “picturize” how the adjustments may be conducted even
without seeing the option list. They found it non-intuitive to interpret the variation map
of the Decomposition Driven method, e.g. in particular with respect to finding out where
to start reading the process and how to configure a specific variant. This point reduces
the flexibility to maintain existing variants. On the other hand, the experts found it more
flexible to use the variation map for defining a new process, as they can see all options
together with constraints on the map.
The experts appreciated the flexibility of the Decomposition Driven method for
modeling variants in different granularity levels. It is conventional to develop separate
models when there are variants of subprocesses which are very different from each
other although the higher level process is similar. In this way, it is possible to balance
298 B. Aysolmaz et al.
We observe that the experts can benefit from the merits of the two methods even
when they use another method specifically in the following ways:
• The guidelines of the Decomposition Driven method may be used to extensively
evaluate why process variants emerge in your business context.
• The approach of the Decomposition Driven method for modeling variants of
hierarchical process models may be implemented in other methods as well.
A Process Variant Modeling Method Comparison: Experience Report 299
• The policies of the Provop method to define a base model, such as usage of
reference models, most frequent variant or minimal distance, may be used while
defining a main process in any single-model variant modeling approach.
• When the generated single-model does not include the steps of configuring a
specific process variant, option list approach can be used.
6 Conclusion
References
1. Hallerbach, A., Bauer, T., Reichert, M.: Capturing variability in business process models:
the Provop approach. J. Softw. Maint. Evol. Res. Pract. 22, 519–546 (2010)
2. Döhring, M., Reijers, H., Smirnov, S.: Configuration vs. adaptation for business process
variant maintenance: an empirical study. Inf. Syst. 39, 108–133 (2014)
3. Ayora, C., Torres, V., Weber, B., Reichert, M., Pelechano, V.: VIVACE: a framework for
the systematic evaluation of variability support in process-aware information systems. Inf.
Softw. Technol. 57, 248–276 (2015)
4. Enterprise, H.P.: Project and Portfolio Management – PPM. http://www8.hp.com/us/en/
software-solutions/ppm-it-project-portfolio-management/
5. Milani, F., Dumas, M., Ahmed, N., Matulevicius, R.: Modelling families of business process
variants: a decomposition driven method. CoRR abs/1311.1 (2013)
6. Hallerbach, A., Bauer, T., Reichert, M.: Configuration and management of process variants.
In: Brocke, J., Rosemann, M. (eds.) Handbook on Business Process Management 1 SE - 11,
pp. 237–255. Springer, Heidelberg (2010)
7. Conforti, R., Dumas, M., Rosa, M.La., Maaradji, A., Nguyen, H.H., Ostovar, A., Raboczi,
S.: Analysis of Business Process Variants in Apromore (2015)
8. Project Management Institute Inc: A guide to the project management body of knowledge
(PMBOK® guide) (2000)
9. Yaldiz, A.: Evaluation of process variant modeling approaches: a case study, Ankara,
Turkey (2016). http://expertjudgment.com/publications/METU_II_TR_2016_YILDIZ.pdf
10. Reichert, M., Rechtenbach, S., Hallerbach, A., Bauer, T.: Extending a business process
modeling tool with process configuration facilities: the Provop demonstrator. In: Proceedings
of BPM 2009 Demonstration Track (2009)