B potential pitfalls_of_process_modeling_part_b-2

REGULAR PAPERS
Potential pitfalls of process
modeling: part B
Michael Rosemann
Queensland University of Technology, Brisbane, Australia
Abstract
Purpose – This second part of the paper summarizes typical pitfalls as they can be observed in
larger process modeling projects.
Design/methodology/approach – The identified pitfalls have been derived from a series of focus
groups and semi-structured interviews with business process analysts and managers of process
management and modeling projects.
Findings – The article continues the discussion of the first part. It covers issues related to tools and
related requirements (7-10), the practice of modeling (11-16), the way we design to-be models (17-19),
and how we deal with success of modeling and maintenance issues (19-21). Potential pitfalls related to
strategy and governance (1-3) and the involved stakeholders (4-6) were discussed in the first part of
this paper.
Research limitations/implications – This paper is a personal viewpoint, and does not report on
the outcomes of a structured qualitative research project.
Practical implications – The provided list of intotal 22 pitfalls increases the awareness for the
main challenges related to process modeling and helps to identify common mistakes.
Originality/value – This paper is one of the very few contributions in the area of challenges related
to process modeling.
Keywords Modelling, Organizational processes, Process management
Paper type Viewpoint
The first part of this paper introduced the characteristics of contemporary process
modeling initiatives. It also discussed six typical characteristics of unsuccessful
process modeling related to strategy and governance and the involved stakeholders.
This second part continues this discussion with a focus on pitfalls related to tools and
related requirements (7-10), the practice of modeling (11-16), the way we design to-be
models (17-19), and how we deal with success of modeling and maintenance issues
(19-21).
7. Lack of realism
Companies tend to underestimate the number of relevant process models they have to
design. A CIO of an Australian insurance company asked me once, how many
processes do I have, 5, 50, 500, 5,000? It did not take long, and within a few weeks a
number of analysts designed 50 þ models. Think bigger, globally, allow more time,
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1463-7154.htm
The author thanks Brad Power, Babson College, Chris Taylor, BP, Roger Tregear, Leonardo
Consulting, Andrew Spanyi, Spanyi International Inc., and Jan Recker, Queensland
University of Technology, for their critical review and valuable contributions to this list of
pitfalls.
Potential pitfalls
of process
modeling
377
Business Process Management
Journal
Vol. 12 No. 3, 2006
pp. 377-384
q Emerald Group Publishing Limited
1463-7154
DOI 10.1108/14637150610668024

and we have hundreds and often thousands of models in the repository. This
complexity driver demands scalability in everything, i.e. the capabilities of the tool,
methodologies, modelers’ capabilities, communication strategies, model maintenance,
etc. Thus, it is important that scalability finds its way into selection and evaluation
procedures:
Do not under-estimate the number of models which you will have to maintain in your
repository over the next three years.
8. The chicken and egg problem
The chicken is the modeling tool and the egg is the modeling language or framework.
On the one side, organizations might have a desire to model in one specific way, but no
tool in the world exactly supports this. An illustrative example comes from an
American producer of chemicals. As part of their Enterprise Architecture modeling
activities, this organization first consolidated the perceived advantages of a number of
frameworks including Zachman, TOGAF, FEAF, Index and DoDAF. They
consolidated all this into one framework and then approached the market in order to
identify a tool which could be customized in a way that it was able to facilitate such a
framework. However, we also see cases, in which organizations stick to a known
approach (e.g. IDEF) and seek support for this methodology, without sufficiently
considering more recent developments. On the other side, organizations tend to select
tools based on recommendations of analysts and market studies, and then adapt the
methodologies as facilitated by the tool. Such an approach often works better when the
modeling maturity is rather low, and there is no capability to develop a
tool-independent approach. It also minimizes the required tool modifications and
benefits from an early exposure to actual practice. Whichever way you start, tool or
method, you will eventually have to compromise:
Be aware of the Catch 22 related to selected tools and methods.
9. Lack of details
While the scope of processes which can be modeled seems to be endless, there are often
annoying constraints about what parts of a process can be modeled. The most recently
proposed Business Process Modeling Notation (BPMN), for example, has only limited
capabilities to cater for modeling risks in the context of Sarbanes-Oxley, or tacit
knowledge for knowledge management or cost drivers in an activity-based costing
project. In many cases, such limitations derive from the history of these modeling
techniques. IDEF and UML, for example, were originally developed for the purpose of
systems analysis and design. Historically, their focus has never been on business
modeling. An increasing number of application areas such as business continuity
management or business rules management pose new challenges, which can hardly be
satisfied by any comprehensive modeling suite right now:
Be aware of the limitations of the selected modeling language and tool.
10. Lost in translation
This pitfall has nothing to do with the English version of SAP’s reference model or Bill
Murray’s “Enjoy your fright” (you will know what I talk about, if you saw the movie
BPMJ
12,3
378

“Lost in Translation”). It is about translating business models into system models, and
let us assume we are really modeling for an IT purpose. Currently we observe a hype
related to the translation of “easy-to-understand” business models into executable
process models. A number of one-to-one as well as generally standardized interfaces
between various business process management tools exist. Asking for such interfaces
is a typical one line item in every significant tool selection process. However, there is a
significant discrepancy between demanding an interface and actually using it.
Different purposes (business or system models?) still vary in their information needs
and we are far way from one standard way of modeling:
While an automated translation of business models to system models is a nice feature, the
capabilities of related interfaces but also the actual opportunities for a 100% translation are
often (still) limited.
11. Lost in a drawing tool
The most popular tool to model is Visio (Davies et al., 2004). And the question most
often asked in the process of selecting a more advanced modeling tool is “Do you have
an interface to Visio”? (the answer by the way is “no”). Visio is a representative
example for a sophisticated drawing tool. These tools have the advantage that they are
often already a part of the standard operating environment. Users tend to be familiar
with the simple drag-and-drop approach. Increasingly, these tools additionally provide
at least a pre-defined set of templates for the most common modeling languages. They
are perceived as easy to use and can easily be customized. They are quite appropriate,
when a few process models with a limited lifespan have to be designed for a specific
purpose. However, have you ever tried to model the largest bank, utility provider or
insurer of your country in Visio? Well, a number of attempts have been made. The lack
of an advanced repository, analysis and reporting functionality, among others,
significantly limits the scalability of such an approach. Sooner or later, every larger
modeling application will reach the limits of drawing tools and look for a more
advanced solution:
Drawing tools have their raison d’eˆtre; however they might just not be appropriate for larger
business process modeling activities.
12. Lack of complementary methodologies
A representative from an American brewery told me that they felt they picked the
world-leading modeling tool, but were overwhelmed by its capabilities. How do I start?
How do I use it in a modeling workshop? What functions are relevant for me?, etc.
Unfortunately, many tool vendors do not provide a comprehensive, detailed, accepted
and tested methodology which helps to use the tool and its plethora of modeling
techniques. An American chemical company even went to the point to characterize a
previous comprehensive modeling initiative as a failure, because its methods were
limited to the tool capabilities. It was only until they derived comprehensive conceptual
support for the entire business process lifecycle, that they were able to succeed.
Consequently, it is necessary to find a methodology, which supports the entire business
process lifecycle and together with the modeling tool facilitates sound model lifecycle
management. This includes modeling conventions, guidelines for modeling
workshops, quality assurance procedures, release cycle management, etc. One
Potential pitfalls
of process
modeling
379

related aim is to mitigate subjectivity in the modeling process, i.e. to make the actual
model design as independent as possible from the person who is doing the modeling:
Complementary methodologies are required to fully utilize the capabilities of modeling tools
and techniques.
13. L’Art pour l’Art
When the artists in the process modeling team dominate, there is the inherent danger
that modeling gains its own momentum. Rather than working for the purpose of
understanding a process or satisfying the information needs of a specific project, it is
the modeling process itself, which absorbs all energy. These are the cases in which
organizations aim for enterprise-wide models and modeling is perceived as a very
time-consuming activity. Completeness matters in these cases more than relevance. An
Enterprise Architect of a Western Australian utility provider phrased it as “Modeling
just-in-case that somebody might need the model.” The opposite would be modeling
just-in-time when the model is really required. A just-in-case approach might be
beneficial to an organization, which is rather risk-adverse. Process modeling, however,
is hardly an end in itself and is always conducted for a specific purpose:
Process models have to be relevant, not necessarily complete.
14 Lost in syntactical correctness
On the other side, a dominating scientific approach towards modeling can lead to an
over-engineered modeling initiative. I talked to owners of modeling guidelines who
could argue for half an hour if a certain attribute of an activity in a process model
should be mandatory or optional. Such an approach may be interpreted by some as
forcing us into speaking Shakespearean English. It is important that both the selection
and customizing of the modeling techniques consider the underlying current and future
objectives. In the academic world this is known as striving for feasible correctness
(Lindland et al., 1994):
Customizing of the modeling technique should strive towards applicability, not perfection.
15. Focus on models and not on modeling
If you ask participants of a modeling workshop at the end of the event what they
enjoyed most, the typical answer is not that they will rave about the beauty of the final
deliverables, i.e. the actual process models. Rather they will tell you how much they
enjoyed the discussions which occurred while they modeled the current or future
processes. They will tell you how stimulating these conversations have been and how
much they learned. What do we learn from this? The very act of modeling triggers a
change reaction and increases process awareness, even if only that those involved in
the modeling will think differently about the processes and related organizational
issues. The process of modeling is most likely more important (if this can be compared
at all) than the final models. As a consequence, process modeling should be conducted
in a highly interactive fashion. However, I recently met a business analyst from an
American chemical company. They conducted some interviews with individual
stakeholders and then consolidated all the feedback in process models sitting on their
own in front of their modeling tool. . .:
The experiences during the journey are part of the overall outcomes of process modeling.
BPMJ
12,3
380

16. Lost in detail
A large oil and gas company is currently conducting significant modeling efforts.
Approximately 100 people access the globally centralized repository of more than 3,000
models. At a recent conference, a representative from this company reflected on the
lessons learnt and stated “We did probably too much detailed process modeling.” It is a
common trap to go deeper and deeper when it comes to modeling. Involved modelers
and business users tend to be driven by the desire to capture all scenarios. They also
might be used to “the old way” of documenting step-by-step standard procedures.
However, the more detailed a model is the longer it takes to design, review and
maintain it. It will outdate more quickly and will often lack relevance as the detailed
steps are intuitive for the involved employees. To avoid such situations, it is essential
to agree on a number of conventions. The most important one being that a focus on the
80 percent case is often sufficient, if it is not the purpose for specifying processes to be
automated. 80 percent means here both probability and resource consumption.
Another possible approach is to not model activities which are conducted by just one
organizational unit. Finding the right level of detail is one of the core challenges in
process modeling. Being able to identify this appropriate level is a core capability of an
experienced modeler. General process improvement, for example, requires high-level
models, while executable models have to be much more detailed:
Define an appropriate level of detail in light of the underlying objectives.
17. Lack of imagination
I have seen a number of process improvement methodologies and techniques, which
were designed around the classical three step methodology “understand the current
process – find ways to improve it – action planning”. Without any doubt the current
performance of many business processes provides typically a number of ideas for
business process improvement. However, the danger of only focusing on the
shortcomings of the existing process is that the entire project concentrates very much
on “overcoming problems” rather than achieving inspirational, new, strategic goals. As
such, it has its constraints. We now observe an increased interest in business process
innovation, i.e. scenarios without a corresponding as-is model. These projects require
“out-of-the-box thinking” a good understanding of completely new ways of conducting
a business process and dealing with uncertainty:
A good understanding of the existing process is important, but it should never be the only
source of ideas for the new process.
18. Lost in best practice
Claudia Schiffer is one of the few recent German export success stories. She is without
any doubt beautiful and consequently features in various advertising campaigns. As
such she could be seen as an example for best practice. However, only one man in the
world can be married to Claudia. What do we learn? Not every best practice is
accessible for everyone. Let me give you another example. As part of a research project,
the faces of the final male and female contestants of a beauty contest were artificially
merged (www.beautycheck.de). The assumption was that the artificially derived final
face expressed the ultimate beauty. However, it looked rather androgynous and was
not beautiful at all. What do we learn? Be careful with best practices which consolidate
a number of features from existing “good cases”. Sometimes 2 þ 2 is 3. Let me give you
Potential pitfalls
of process
modeling
381

a third example. Toyota is well known for its efficient operations management. Many
companies have tried to adopt Toyota’s secrets. The problem has been that Toyota is a
huge, self-contained organism with a complex set of values, behaviors, assets, people,
processes, policies, etc. It is not sufficient to just copy one part of Toyota. However, it is
hard to know what parts have to be copied, and how they relate to the rest. Just a single
focus on best practices (such as Toyota) can be dangerous, if selection bias matters
(Denrell, 2005). With the “blinding light of success” as Denrell phrases it comes the
increased difficulty to distinguish between important and unimportant factors. In our
context: A successful company might have great processes, but it is not because of
these processes that it is so successful. Finally, with best practices we are often only
exposed to the final result, but we do not see the process which led to this
recommended case. All these examples show that it comes down to applicable best
practice. It is not about the existence of good cases. It is about pathways for the wider
roll-out of these ideas. A number of reference models try to capture the current state of
the art in the areas of IT service management, supply chain management, project
management or customer relationship management. In most cases, however, they are
not the complete cookbook to establish relevant best practice in your organization.
They tend to exclude references to actual case studies and do not consider relevant
context factors:
So called best practice models can be useful in terms of structure, content, overall guidance
and opening up more possibilities. The notion of best practice is, however, typically
over-rated.
19. Design to-be models solely centered on new IT
Some of our modeling-related projects had a timeframe for the to-be model of one
year, sometimes even two years. In discussions about how we can overcome the
current problems in such a timeframe, the most popular answer is – “With new IT”.
These people also believe that SAP as an abbreviation means “Solutions to All
Problems” and with enough patience, IT can fix anything. It is a bit like the strong
believe of my son in the endless capabilities of Bob the Builder. Such an attitude can
lead to a number of problems. First, it can be used as an excuse to not search for
non-IT related solutions. I may have great solutions for routing all my incoming
phone calls within my call centre, but why do I receive these phone calls in the first
place? Second, it can lead to an attitude that nothing gets done until the new system
is in place. Third, it might express the naı¨ve believe that the IT vendor really meant it
when (s)he said that the expected feature will be available in the next version. A CIO
described this to me as the “process management dilemma of IT”: IT is not in charge
of the process, but if the deployment of IT fails in this process, it is their fault. In light
of the classical application development backlog in many organizations, it is in fact
quite healthy to ask for process improvement ideas which explicitly do not utilize IT
changes at all:
Business process models stimulate an integrated organizational and IT view on process
change. An exclusive focus on IT solutions ignores other potentials resulting from non-IT
improvements.
BPMJ
12,3
382

20. Modeling success is not process success
In literally all process modeling sessions, we see an enormous satisfaction with the
newly designed business process blueprints, the so-called to-be (or should-be) models.
There is a great tendency to be too satisﬁed with the created process models. A ﬁrst
wave of enthusiasm can often be observed when a to-be model is created. However, this
remains just a model. It looks great, but does not change the world. Or, as Tom
Davenport phrases it: “After all, it was easier to create detailed models than it was to
create real change within organizations” (Davenport, 2004):
Appreciate the ideas which went into the new process design, but be aware that only the
implementation matters.
21. Lost in model maintenance
The next issue is related to modeling in the large. The oil and gas company mentioned
earlier has 3,000 models in their repository, the Australian utility provider has 4,800
models. How do those organizations keep the models up-to-date? Do they have to? How
many of these models are not current anymore? We call outdated models in a
repository, “pollution”. With increasing size of the model repository, it becomes
essential to establish sound practices for model lifecycle and review management as
part of an overall quality assurance. Such an approach must be scalable and should be
owned by a central process group. The responsibility for the actual correctness of the
model, i.e. its semantic quality, however, should rest with a business representative
and has to include a clear model ownership, which should rest with the business (see
also governance):
Establish sound, but appropriate procedures and ownership for maintaining an increasing
model repository.
22. Lack of measuring modeling performance
We model processes to improve process performance in the hope that this positively
contributes to increased business performance. It is important, therefore, to have, right
from the start, a good understanding of the parameters we are seeking to change and
the nature of the change we seek. Such parameters must be measurable and the
measurement process should be well understood before we start so that we can create a
baseline against which our changes can be assessed. Understanding the nature of the
improvements that we are trying to make and maintaining a credible record of
outcomes as we progress, can only happen if we develop appropriate plans for process
performance measurement before we start making changes. Companies have different
attitudes in terms of the evidence they require before they believe in actual process
improvement. The actual measures may vary and go far beyond the obvious
processing time. In a recent project in an insurance company in the area of processing
claims related to personal injuries in car accidents, it was the minimization of the
average payout, which was far more important than the processing time. In a loan
application example, the focus was on “maximizing face-to-face time spent with the
customer.” Measuring the cost of a process is another interesting challenge. The
discussions related to activity-based costing have shown that any measurement of
process costs has to be very much aware of the characteristics of direct and indirect
Potential pitfalls
of process
modeling
383

costs, and the correlation of ﬁxed and variable costs with the timeframe of the project.
Something not every business process analysts truly understands:
If you can’t measure it, you can’t manage it – and, more importantly, you can’t claim it as a
success.
23. Summary
In the light of 22 pitfalls, the reader might ask for the actual motivation to do process
modeling at all. Process modeling has proven essential for achieving business beneﬁts
in a large number of cases. In the American brewery mentioned above, for example, a
related case study states that a project which utilized modeling saved the organization
more than $3 million. The list above, however, hopefully increases the awareness for
the current challenges we face when we approach process modeling to the extent that it
is required in larger organizations.
References
Davenport, T. (2004), “Modeling in moderation”, Process Innovation. A Business Process Trends
Column, June, available at: www.bptrends.com
Davies, P.G., Rosemann, M. and Gallo, S. (2004), “Conceptual modelling – Why do we do it?”, in
Atzeni, P. et al. (Eds), Proceedings of the International Conference on Conceptual Modelling
(ER 2004), Shanghai, 8-11 November, pp. 30-42.
Denrell, J. (2005), “Selection bias and the perils of benchmarking”, Harvard Business Review,
Vol. 83 No. 1.
Lindland, O., Sindre, I. and Sølvberg, G. (1994), “Understanding quality in conceptual modelling”,
IEEE Software, Vol. 11 No. 2, pp. 42-9.
BPMJ
12,3
384
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints

B potential pitfalls_of_process_modeling_part_b-2

More Related Content

What's hot (18)

Viewers also liked (12)

Similar to B potential pitfalls_of_process_modeling_part_b-2 (20)

More from Jean-François Périé (10)

Recently uploaded (20)

B potential pitfalls_of_process_modeling_part_b-2