Investigating Process Elicitation Workshops using Action Research

Investigating Process Elicitation Workshops
using Action Research
Alexander Luebbe and Mathias Weske
Hasso-Plattner-Instittute, 14482 Potsdam, Germany,
alexander.luebbe,mathias.weske@hpi.uni-potsdam.de,
WWW home page: http://bpt.hpi.uni-potsdam.de
Abstract. We develop a workshop technique for process co-creation
with domain experts called tangible business process modeling. After assessing
the idea in a laboratory experiment, we started workshops with
professionals in the field. This paper illustrates how we used action research
in two subsequent studies in which groups modeled BPMN and
EPCs using tangible tiles on a table.
The practical result is best practice guidance for moderators applying
tangible process modeling. Our research interest is to investigate the
differences between tangible modeling and other techniques. In the lab
experiment we found that tangible modeling supports user engagement
and validated results. In the field, we compare the workshop performance
and outcome to software-supported workshops. We found tangible modeling
to be competitive in speed and result.
Key words: business process elicitation, workshops, process modeling,
tangible business process modeling, field research, action research
1 Introduction and Background
Business process modeling is the act of capturing domain knowledge about work
activities, their interdependencies, and responsibilities, in a graphical representation.
Process models are used to analyze current business operations, communicate
requirements for IT systems, or discuss alternatives to the current way of
doing business. Typically, domain experts share their knowledge with a process
analyst in interviews or workshops. The model is either created after the interview
or in parallel during the workshop by a dedicated tool expert [1]. Domain
experts review the results but do not actively participate in process model creation.
This practice detaches the domain expert from the process model with
problems such as misunderstandings and limited commitment [2].
We have created a modeling approach to engage the domain expert with the
act of process modeling. The approach covers a toolkit and a method to apply it.
In the first phase we developed the toolkit as a set of transcribable plastic tiles
to be used on a table for process modeling [3], see Fig. 1. In the second phase
we assessed the effect of modeling with tangible tiles by comparing individuals
in interview situations with tangible process modeling sessions. Subjects were

2 Alexander Luebbe, Mathias Weske
t.BPM Introduction
more engaged with the task. Moreover, they reviewed and corrected the model
more often leading to more validated results [4].
Fig. 1. tangible modeling tool set: transcribable plastic tiles reflecting the BPMN
iconography
The lab experiment showed positive effects of tangible modeling on individuals.
The current research interest is the efficiency of group modeling workshops
in reality. We aim to assess the applicability of tangible modeling in real environments
and compare tangible modeling workshops with other techniques in
place today.
The practical goal is to derive guidance for professionals who want to leverage
the theoretical potential [4] of tangible business process modeling in their
workshops, namely stronger engagement and validated results. Since practical
guidance can only be derived from practice, we choose action research as a
method to guide our cooperation with professionals in the field.
In Section 2 we introduce the idea of action research. Afterwards, we apply
action research in Section 3 and Section 4 in two subsequent iterations and
discuss our learnings in Section 5. Finally, we look at related empirical research
in Section 6 and conclude the paper in Section 7.
2 Action research – an iterative learning cycle
Action research (AR) is a class of research approaches in which researchers collaborate
with practitioners to act in or on a social system [5]. The goal is to
solve practical problems and generate scientific knowledge. Action research assumes
that complex social problems cannot be reduced to meaningful controllable
studies. They need to be investigated within the context in which they
appear. Therefore the system is studied, changed, and the effect of change is
studied again. It is an iterative learning cycle. We adopt a five stage process
proposed by e.g. Baskerville [6], see Figure 2. We also follow the principles for
AR defined by Davison et al. [7] such as multiple iterations, the principle of
theory, and learning through reflection.
In the diagnosing step the context and problem statement are described. In
action planning the desired future state and proposed changes are introduced

Investigating process elicitation workshops 3
diagnosing
specify learning
action research
cycle
action planning
evaluating
taking action
Fig. 2. Action research cycle used to guide the field studies in this paper
based on theoretical background. The first two steps are carried out in close
collaboration with practitioners. In taking action the practitioners act alone
while the researchers observe and collect data for the evaluating phase. This
phase analyses whether the proposed changes realized the intended effect and
whether the practitioners’ problems were solved. The results are used to specify
learning as more general knowledge, practically and scientifically. In this last
step, we also develop research questions for the next iteration.
3 BPMN process modeling workshop with clinical staff
In mid 2009, a consultant used the tangible toolkit (see Fig. 1) to capture clinical
paths in a hospital with medical staff. A clinical path is a treatment process for
a class of patients with a similar disease, in this case for a certain type of cancer.
It was the first application of our tangible modeling approach in the field. As
researchers, we observed the situation and framed it as a case study [2]. Within
one week, twenty BPMN processes were modeled together with three doctors
using the tangible modeling toolkit. The workshop was considered a success by
the consultant. In mid 2010, the same consultant approached us again with the
same request. The setting was very similar, only the participants and the clinical
path were changed. This was the starting point for the first action research cycle.
Diagnosing. When reviewing the workshop from 2009 together with the consultant,
we agreed that the introduction was not optimal. Originally, a one-day
introduction to BPM and BPMN was given. Modeling only started on the second
day. A further problem was that the workshop in 2009 got confused with
different media. Tangible tiles, paper printouts, and software were all present
during the workshop. Jumping between media created different embodiments
with different versions of the same process model. A clear guideline was needed
that matches the media with the purpose of the modeling phase. Finally, as researchers,
we wanted to investigate to which extend data collection is feasible in
such a research environment and what can be traced from the data. Up to this
point, we were not sure which data to trace in order to characterize modeling
sessions and results.

4 Alexander Luebbe, Mathias Weske
Action planning. We developed a shorter BPMN introduction with the consultant
and proposed warm-up excercises to enable a quicker start with process
modeling. This was based on the idea of learning by doing [8]. We collected
modeling exercises from consultants and literature such as ‘withdrawing money
from an ATM’ inspired by Rittgen [9]. These scenarios enable participants to
get familiar with process modeling without getting lost in domain specific discussions.
We also defined a media framework that proposes to (1) complement
discussions with drawings, (2) use tangible tiles on a table to generate models,
(3) use printouts for reviews, and (4) use software to store models. This was
inspired by the research of Edelman [10] on the role of media in steering conversations.
For data collection, we sent an observer to take photos and detailed
notes during the workshop.
Action taking. Again, a one-week workshop was conducted. Three participants
from the hospital worked out the clinical paths for liver transplantation together
with the consultant. The introduction to BPM and BPMN was shortened to
half a day. The second half of the first day was used to model a pizza-ordering
process and discuss modeling decisions. Jena2011-1 In the following four days, the doctors
modeled thirteen processes together with the consultant. Fig. 3 shows some
impressions from the workshop. In total we collected more than six hundred
pictures, nine pages with observer notes and interviews with the consultant and
each participant. From the modeling sessions we collected six snapshots of the
project to document the progress during the week.
Fig. 3. Modeling (left) and reviewing (right) clinical processes using BPMN
Evaluating. In interviews at the end of the week, the participants qualified tangible
modeling as well suited to adopt process thinking. The modeling exercise in
the beginning was perceived as a logical part of the introduction. The consultant
liked the way in which the modeling exercise complemented the introduction and
activated participants early. In his opinion the contribution to learning BPMN
is not very high, but the modeling exercise reduces the barrier to start creating.
All together, the introduction and modeling exercise consumed the first day of
the workshop.
The consultant also classified the media framework as a nice idea to keep
in mind. However, it is not a golden rule and does not necessarily match the
practical situation. For example, on day three people added role information to

Investigating process elicitation workshops 5
all existing models. They did this by coloring process steps in printed models.
Strictly following our media framework would have implied recreating tangible
models to add this information. Participants said to have experienced the media
choices as natural.
The observer notes and photos were analyzed to identify patterns. We distinguished
four working modes during the workshop week. People got introduced
to new modeling knowledge, participants modeled, the consultant digitalized the
models in the evenings and everybody reviewed results. These modes happened
in a cyclic pattern: (1) introduce, (2) model, (3) digitalize, and (4) review.
Additionally, we used process model metrics by Mendling [11] to explore
characteristics of the models and the modeling progress. Model metrics that
showed a high dynamic were considered as candidate KPIs. We concluded the
amount of information pieces to be characteristic for the modeling progress of
the project. The amount of information mapped per hour in a modeling session
seemed to be a good indicator for productivity. These metrics do not tell all the
truth about a workshop. They are an abstraction to compare workshop outcomes
in relation to the time spent. However, in this special case there was no data to
compare with.
Finally, we asked the consultant for his preferred alternative to tangible modeling
and he responded: “There is no alternative. I knew it would work from our
previous workshop. If I did not have this option then I would have to do interviews.
But if you are not an expert in the domain – which I am not – good luck
understanding your client.”
Specify learning. From this field study, we learned that participants and the
consultant perceived the warm-up exercise as positive. It did shorten the time
to start modeling but overall it did not save time in comparison to the previous
workshop. The actual modeling of content started in both cases at the second
day.
We also learned that the media framework is a good idea to guide consultants
in the selection of media for the workshop. But it is not a golden rule and the
media must be chosen specific to the situation. Observation notes and photos
enabled us to trace phases, productivity and modeling progress. From the transition
of phases, we identified a natural modeling cycle. The productivity and
process data itself, however, is meaningless without a proper data set to compare
with.
By reviewing the workshops in 2009 and 2010 we identified themes that
worked well. From that we derived the first practical guidance for moderators
and iterated it with the consultant. The guidance was on topics such as the
right setting, the media framework and the warm-up exercises. To broaden our
results, we identified the following research questions to be most pressing for a
subsequent AR study:
1. Is the idea specific to BPMN as a process notation
2. Does tangible modeling create results different from other techniques
3. How productive is tangible modeling compared to other group workshops

6 Alexander Luebbe, Mathias Weske
4 EPC modeling at an energy provider
In early 2011, an in-house consultant working for a subsidiary of a large energy
provider in Germany contacted us. The company is responsible to develop and
test new ideas for the energy market such as smart home devices and e-mobility
concepts. Within this company, our contact captures and improves organizational
processes in workshops together with domain experts. He approached us
because he saw a need for tangible modeling in specific workshop situations.
Diagnosing. Two people typically run the process elicitation workshops at this
company as a team, a moderator and a software tool expert. While the moderator
elicits information in a conversation with the participants, the tool expert
translates the information into a process model in the background. The computer
screen is simultaneously projected to a wall. The participants can see the
process model evolving and review it. In this case the process notation is EPC
and the modeling tool is ARIS [12]. A modeling workshop lasts three hours.
The process modeling team worked on more than six hundred process models
together in the last years.
The modeling team faced some problems, which they wanted to tackle with
tangible models. For one, the projector limits the overview. A typical screen
resolution for projectors is 1024 by 768 pixels but even a full HD resolution
would not enable to see the large models all at once. Thus, participants can lose
the overview. A further aspect is the very limited involvement of the participants,
especially, if the model is not that clear but has to be discussed and developed
together. The moderator felt that some conflicts are not expressed because the
participants have no means to do that. These conflicts pop up in subsequent
workshops when the previous workshop result is reviewed. For us, as researchers,
the setting was excellent to tackle our open research questions.
Action planning. We wanted a similar setting but change the modeling tool.
Instead of ARIS and a dedicated tool expert, we proposed to use a tangible
version of EPCs, see Fig. 4 (left). The moderator insisted on the software and
the tool expert as a backup. Thus, the model was created on the table and
digitalized for documentation by a dedicated ARIS modeler in parallel.
It was clear that not all workshops would equally benefit from tangible modeling.
We looked for processes with lots of need for discussion and people openminded
towards active workshop participation. The department for idea management
was identified within the company. This department was formed one
year earlier and just grew from two to three knowledge workers. The processes
were not modeled yet but the department looked for software to support their
work. Two workshop sessions were scheduled at two subsequent days. The goal
was to create an overview of the existing processes in this department and to
model the core process that would benefit from software support.
To collect data, we sent two observers to the workshop. Furthermore, we
got access to the organizational handbook of this company with 22 processes
of a similarly structured department. It serves as a reference point for typical

Investigating process elicitation workshops 7
processes captured by this team of process modeling experts in this company. To
trace productivity, we collected effort estimations for a recently finished modeling
project from a different department. This project spanned 18 models and served
as a reference point for the productivity of the process modeling team.
Action taking. The first workshop started late and ended early due to other
commitments of the participants. Two out of three hours actually remained.
Within that time, the moderator gave a standard ten-minute introduction about
the need for process modeling. In the first step a map of processes for this
department was created. In a brainstorming manner, all processes were collected
on Post-Its at a flip-chart. Subsequently, the workshop participants consolidated
the process landscape by re-arranging the Post-Its. Here, it already turned out
that the scope of this department was not entirely clear. Creating this map
consumed the first workshop hour. In the remaining time, the tangible EPC set
was introduced and the first steps for the ‘idea creation and evaluation’ process
were mapped.
Fig. 4. tangible EPC toolkit (left) and modeling idea management (right) at energy
provider
The next day started with a review of the existing process snippet. The main
path was completed and two alternative paths were modeled. Iteratively, each
path was enriched with roles, documents and IT systems. Three explicit reviews
were done during the workshop, after each completed path. Many more small
reviews happened in the form of mental leaps during discussions. We collected
three pages of observer notes and 112 photos including four snapshots of intermediate
modeling results. We interviewed the moderator, the software tool
expert and one of the participants after the workshops.
Evaluating. The interviewees expressed satisfaction with the workshop result.
The software tool expert attributed the stronger activation of participants to the
non-technical, puzzle-like nature of tangible modeling. The moderator classified
the stronger participation as an expected result. For him, standing, pointing,
and talking already adds value to the workshop. The additional option to move
around tiles and try out alternatives replaces review discussions at later stages,
he said. The participant liked playing an active role in the workshop. For him,
being an active workshop participant was not a new experience. When asked

8 Alexander Luebbe, Mathias Weske
about the next steps of reviewing the digitalized EPC model, he said ”I won’t
look at it. There is nothing to gain. I was here. I know what is in the model.”
Limited review interest might apply !"#$%"& to all types of workshops making active
participation and corrections during the workshop even more relevant. )'&
The digital EPC printed on A3 paper was hardly readable. A graph (& with 126
nodes in total. The same model as a tangible representation was overlooked on
'&
a three by one meter table. Both observers took notes of events where
*+!&
participants
leaped from one end of the model to the other in order to add or correct
+,-.%&)&
information.
We evaluated the workshop result by quantifying the information in the
model. In Figure 5 (left) we compare this with a box-plot derived by evaluating
22 processes modeled by the same BPM experts in a similar department.
For this statistic we counted the amount of node
!"#$%&"'('#")*+,-./0(+"1'2#3)4(
type occurrences in the models.
%"5!2#3(%*''"#/3)(!"#$%&"'%(-%6(
The values from tangible modeling are in range except for role information.
/278.9:3(1")3:.78(
amount of elements per type
100
80
60
40
20
0
Amount of elements
35
54
20 17
control Control Flow flow roles Roles documents Documents IT-Systems IT-systems
elements 5$%6.&6-77%8&7%9&,#:9&
mapped per hour
4'&
3'&
2'&
('&
1'&
0'&
/'&
.#>-9%@.:77#9$%8&#9A.,#7.&
)'&
$-"B5C

Investigating process elicitation workshops 9
was an unexpected finding. We see model size as a candidate to determine the
tool choice for modeling workshops.
Specify learning. For the practitioners the modeling workshop was successful.
As researchers, we realized that standing, pointing and discussing are important
aspects, no matter which media is in place. Tangible media adds additional
value because it enables the participants to express ideas by directly interacting
with the model. We have chosen the idea management department because
they are familiar with interactive workshops and they had a need for clarifying
discussions. We learned that those situations welcome the use of shared tangible
media.
Our measures indicate that tangible modeling creates similar results when
compared to the established software-supported modeling technique. In our data
set, tangible modeling is also faster for smaller models. Larger models benefit
from a digital tool because it allows copying, pasting, and adapting model parts.
Despite this restriction, we think the tangible modeling productivity is astonishing.
Especially because the additional software modeling tool expert is not
required during tangible workshops. But we are also aware that the numbers we
compare with represent a range of cases from a different department. They were
modeled by the same team of process experts but with other participants.
Although we provided our existing best practice knowledge from the first AR
study, the moderator in charge at the energy provider did not look at them prior
to the workshop. As a result, there was e.g. no warm-up exercise. Instead the
moderator improvised a variant of his existing workshop technique. We reviewed
both AR studies and complemented our best practice knowledge with new insights
about setting and group facilitation. As examples, the modeling session
should be well scoped. Especially, the modeling goal to create an as-is or a to-be
model must be defined upfront.
5 Result discussion
As a practical result of this work, we identified guidelines in the categories of
setting, media framework, modeling cycles, warm-up exercises and group facilitation.
Experienced modeling experts might ignore the advice entirely and run
the workshop by their intuition. But some good practices, like scoping the modeling
session to as-is vs. to-be models, have also been forgotten by the experts in
our studies. Thus, these guidelines can serve as a reminder even for experienced
modeling professionals. The guidelines for setting and media framework are of
particular importance to tangible modeling.
The workshops were able to deliver the expected strong engagement of participants.
The benefits are inline with findings from our lab experiment [4]. For
example, participants said that this type of modeling helped them to learn process
thinking. This is strongly related to ‘more insights into process modeling’
measured in the lab experiment when comparing interviews with tangible modeling.
We also found in the same experiment that participants do many more

10 Alexander Luebbe, Mathias Weske
reviews and corrections when the process is mapped on the table. In the second
AR study, reviews and corrections were noticed as mental leaps by the workshop
moderator and by both observers independently.
We answer our research questions as follows:
1. Tangible process modeling is not specific to a process notation but to the
people and the process to be discussed. People need to be open for the
technique and see a need to discuss the process.
2. Tangible modeling can be used to produce models competitive to softwaresupported
modeling workshops.
3. Tangible modeling productivity is competitive to software-supported modeling
for models that cannot take advantage of re-use through copy and paste.
We want to draw the attention to some limitations of this research. One
particular aspect to point out are the measurement instruments used to compare
productivity and outcome. They are new instruments which need further
testing. Our testing compared workshops from different departments with different
participants. Nevertheless, all data comes from the same team of process
modeling experts that ran the AR study with us. Therefore we think the data
is valuable to build insights and hypotheses. Further investigations with larger
and more homogeneous data sets are needed to strengthen or disprove our conclusions
with statistical expressiveness. For now, we can only make transparent
the path taken for readers to follow or challenge our conclusions.
The two field studies reported in this paper were the main drivers for insights
into the technique and the basis to answer our research questions. But this paper
can only tell a subset of the overall story. As an example, we discussed our
guidelines with more practitioners and researchers to form a stable opinion on
relevance.
6 Related action research
Action research is quite well adopted in information systems science to guide
researchers working with professionals [14]. As examples, it was used to improve
the process of software development in coordination with the client organization
[15], increase the value of existing information systems [16] or use software
to create social impact [17].
A typical action research publication mentions the research method but focuses
on problem descriptions and findings from multiple iterations. This is complemented
by theoretical papers on the action research cycle [6] or general AR
principles [7]. In this paper, we use these theoretical guidelines and show their
operationalization in two subsequent action research cycles.
In business process management, action research is not mainstream yet. Practitioners
are typically involved in research through surveys and case studies [18].
If solutions are developed together with practitioners, this is framed as or combined
with design science research (DSR) [19]. Indeed both research approaches

Investigating process elicitation workshops 11
have a lot in common [20], in particular the iterative refinement and assessment
of the research artifact, in our case the method.
The nearest relative to our work is the research done by Rittgen [9, 21]. He
designed and implemented software to engage people in group modeling workshops.
The fundamental difference to our work is the assumption that people
can be engaged with computer-based modeling. Our tangible approach targets
people who do not want to work with computers in group workshops.
Rittgen describes his research method as design science research(DSR) in [21]
or a combination of DSR and action research in [9]. In [21], he presents three
years of research as a single execution of the design science cycle. We illustrate
our research journey in this paper using two concrete AR iterations to spread
knowledge about the operationalization of action research.
7 Conclusion
We created a modeling approach to engage domain experts in the creation of
their process models. In previous work [4], we determined the effect of tangible
media on individuals in a lab experiment. In this paper, we present field research
with groups. We opt for a five-step action research model to guide our cooperation
with practitioners. We successfully applied tangible modeling with doctors
and businessman using BPMN or EPC as process notations.
In cooperation with the professional partners, we developed guidance for
moderators facilitating tangible modeling workshops. This guidance spans the
areas of setting, media framework, modeling cycles, warm-up exercises and group
facilitation. The knowledge we bring into the scientific discussion evolves from
our case data. We claim that (1) tangible process modeling is not restricted to
a process notation but to the people willing to do it. Furthermore, (2) tangible
modeling does not produce a different type of process model and (3) it is not
slower than software-supported modeling as long as model re-use is not common
practice.
Our conclusions are based on qualitative research meaning they are the result
of interpretation. Further investigations with larger and more homogeneous data
sets may strengthen or disprove our claims. In this paper, we present the observations
that led to these interpretations. We also provide a sample for other BPM
researchers that want to work with practitioners to generate both, practical and
scientific knowledge.
Acknowledgement
We are grateful to the BPM practitioners that fearlessly tried out new ideas
at work. We also acknowledge the support by Markus Guentert who helped to
collect and evaluate data.

12 Alexander Luebbe, Mathias Weske
References
1. Stirna, J., Persson, A., Sandkuhl, K.: Participative Enterprise Modeling: Experiences
and Recommendations. LNCS 4495 (2007) 546
2. Luebbe, A., Weske, M.: Designing a tangible approach to business process modeling.
Electronic Colloquium on Design Thinking Research (2010)
3. Grosskopf, A., Edelman, J., Weske, M.: Tangible business process modeling -
methodology and experiment design. In Mutschler, B., Wieringa, R., Recker, J.,
eds.: ER-BPM’09. Volume 1., Springer Verlag (2009) 53–64
4. Luebbe, A., Weske, M.: Tangible media in process modeling – a controlled experiment.
In Mouratidis, H., Roland, C., eds.: CAiSE. Volume 6741. (2011) 283–298,
5. Brydon-Miller, M., Greenwood, D., Maguire, P.: Why action research Action
Research 1(1) (2003) 9
6. Baskerville, R.: Investigating information systems with action research. Communications
of the AIS 2(3es) (1999) 4
7. Davison, R., Martinsons, M., Kock, N.: Principles of canonical action research.
Information Systems Journal 14(1) (2004) 65–86
8. Anzai, Y., Simon, H.: The theory of learning by doing. Psychological review 86(2)
(1979) 124
9. Rittgen, P.: Success factors of e-collaboration in business process modeling. In
Pernici, B., ed.: CAiSE. (2010) 24–37
10. Edelman, J.: Hidden in plain sight: Affordances of shared models in team based
design. Proceedings of the 17th International Conference on Engineering Design,
ICED’09 (2009)
11. Mendling, J.: Metrics for process models: empirical foundations of verification,
error prediction, and guidelines for correctness. Springer-Verlag (2008)
12. Scheer, A.: ARIS-business process modeling. Springer Verlag (2000)
13. Chiasson, M., Germonprez, M., Mathiassen, L.: Pluralist action research: a review
of the information systems literature*. IS Journal 19(1) (2009) 31–54
14. Baskerville, R., Stage, J.: Controlling prototype development through risk analysis.
Mis Quarterly (1996) 481–504
15. Lindgren, R., Stenmark, D., Ljungberg, J.: Rethinking competence systems for
knowledge-based organizations. European Journal of IS 12(1) (2003) 18–29
16. Wastell, D., Kawalek, P., Langmead-Jones, P., Ormerod, R.: Information systems
and partnership in multi-agency networks: an action research project in crime
reduction. Information and organization 14(3) (2004) 189–210
17. Bandara, W., Tan, H.M., Recker, J., Indulska, M., Rosemann, M.: Bibliography
of process modeling: An emerging research field. Technical report, QUT (2007)
18. Hevner, A., March, S., Park, J., Ram, S.: Design science in information systems
research. MIS Quarterly 28(1) (2004) 75–106
19. Cole, R., Purao, S., Rossi, M., Sein, M.: Being proactive: where action research
meets design research. In: 26th ICIS. (2005) 325–336
20. Rittgen, P.: Collaborative Modeling–A Design Science Approach. In: Proceedings
of the 42nd Hawaii International Conference on System Sciences. (2009) 5–8