Verifying Existence and Composition of Workflow Activity
Patterns in Real Workflow Processes
Carolina Chiao1, Lucinéia Heloisa Thom2, Cirano Iochpe1, Manfred Reichert3
1
Institute of Informatics – Federal University of Rio Grande do Sul
Av. Bento Gonçalves, 9500, 91501-970 – Porto Alegre, RS, Brazil
2
Institute of Database and Information Systems – University of Ulm
Oberer Eselsberg, Geb. O27, 89069 – Ulm, Germany
3
Faculty of Electrical Engineering Mathematics and Computer Science –
University of Twente – 7500 AE Enschede, The Netherlands
{cchiao, ciochpe}@inf.ufrgs.br, [email protected],
Abstract. In this paper we present a collection of high level workflow activity
patterns based on the semantic of specific business functions (e.g., notification,
task execution request, approval). In particular we discuss three pattern
samples (approval, unidirectional and decision patterns). Moreover we gather
the results of an analysis of their adoption on a wide set of real workflow
processes (models). The analyses showed that the patterns are not only enough
but also necessary to model all the 190 workflow processes which were subject
of the investigation. We also show and discuss specific sequences or
combination of patterns which were more often in the workflow processes
analyzed. In larger research we apply these patterns as well as the analyses
results in the development of a suite for process modeling and normalization.
1. Introduction
During the last years, companies have been exploring numerous techniques for
business process management (BPM) in order to align their information systems in a
process-oriented way and to stay competitive in their market. Accompanying this trend
the significance of BPM has increased and new quality standards have emerged.
According to the quality standard ISO 9001:2000, for example, an organization should
be mainly represented by its core business processes rather than by its organizational
chart. If BPM is associated with Information Technology (IT), it will become possible to
offer additional benefits to the organization, such as: (a) precise and unambiguous
description of the existing business processes; (b) improvements regarding the definition
of new processes; (c) effectiveness regarding the work coordination between different
agents; (d) real time gathering of precise information about process executions; and (e)
standardization of business processes.
For (computerized) business processes there exists a variety of fragments which
can be understood as self-contained activity blocks with a well-defined semantics
[Thom 2006a], [Thom 2006b]. In particular, a certain process fragment (or recurrent
business function) may occur several times within one (or different) process
definition(s). As an example, consider the evaluation process for price adjustment as
depicted in Figure 1. This process includes activities with the following partial order: (a)
a decision activity (to fix whether the input is a shopping order or not) (b) activity ‘send
e-mail to manager informing about price adjustment’; (c) activity ‘evaluate request of
price adjustment’; (d) activity ‘notify managers about conclusion of evaluation; (e)
activity ‘notify managers about automatic approval’; and (f) activity ‘prepare request to
be sent’. Altogether this process comprises fragments having generic semantics that can
be described as patterns such as decision (activity a), notification (activities b, d and e),
and task execution request (activities c and f).
Figure 1: Evaluation process for price adjustment
Recently, research on workflow patterns has emerged in order to increase the
reuse of recurring workflow structures. More precisely, different workflow patterns have
been proposed for control flow modeling [Aalst 2003], resource management [Russell
2004], data modeling [Russell 2005], service interaction [Barros 2005], workflow
exception handling [Russell 2006]. All these pattern sets have in common that they are
relevant for implementing a workflow modeling tool, or for defining or evaluating a
particular process description language. However, these structural patterns provide only
a partial answer to the question what business functions a modeler has to consider
repeatedly in various process models.
Usually, such process fragments [Flores 1998], [Medina-Mora 1992], [Malone
2004], [Muehlen 2002] are re-designed for each workflow application. Of course, this
lack of reusing model fragments and process knowledge has resulted in high costs and
error rates regarding the modeling and maintenance of process-oriented applications.
While there is some research reported on how metadata can be organized to manage
large-scale modeling projects (see [Thomas and Scheer 2006]), to our best knowledge
there exists no (empirical) work evidencing the existence of recurrent patterns in real
workflow applications. Furthermore, there is no work on which patterns are needed and
how good they may support the modeling of at least specific kinds of business
processes. Beyond that, contemporary workflow modeling tools do not provide
functionalities that enable users to define, query, and reuse such patterns in a proper
way.
Concerning this problematic in an earlier work, on [Thom 2007a] we presented a
first approach towards the implementation of workflow activity patterns based on an
Event-driven Process Chains (EPCs) tool [Keller 1992]. Recently we proposed a suite
for both process modeling and normalization based on the reuse of semantic process
patterns (see [Thom 2007b]). In this paper we gather samples of workflow activity
patterns. However, the main contribution of this paper is the description of a case study
where we analyzed 190 real workflow processes (models and not instances) from
different organizations as well as application domains. Taking the results of the case
study we show that the workflow activity patterns do not only exist in real workflow
applications, but are also necessary and sufficient to model all the workflow processes
which were subject of our investigation. We believe that the use of these patterns
together with other existent patterns (e.g., control flow patterns [Aalst 2003]) might not
only reduce design effort (e.g., it is a small set of patterns that seems to require little
effort to learn) but also optimize and improve the quality of it (e.g., the user can reuse
design solutions stored in a knowledge database). This database is part of a larger
research where we propose a suite for workflow design based on patterns reuse (cf.
[Thom 2007a], [Thom 2007b]).
Against this background, the outline of this paper is organized as follows:
Section 2 gives an overview of the workflow activity patterns. In particular, we discuss
the approval, notification and unidirectional as three examples. Section 3 gathers the
results of an extensive case study where we investigated the existence of the patterns in
190 workflow processes (models). In this section we also show how complete is the set
of patterns for the design of the 190 workflow processes. In Section 4 we present some
patterns combinations by dividing the set of workflow processes on System-Intensive
and Human-Intensive –oriented process. This classification is useful to obtain more
precise results concerning the kind of processes the patterns are more suitable to be
found. Moreover, this information will be used in the development of a knowledge
database of patterns. Finally, Section 5 concludes the paper and gives an outlook on
future research.
2. Workflow Activity Patterns
A
W
ORKFLOW
A
CTVITY
P
ATTERN
refers to the description of a recurrent business function
frequently found in business processes (e.g., notification, decision, approval). We
derived a set of 7 patterns from an extensive study based on the literature (see [Thom
2006a] and [Thom 2006b]). Examples of patterns are approval, question-answering,
unidirectional and bi-directional performative, information, notification and decision
patterns.
A block activity is suitable to represent each pattern according to [WfMC 2005].
The block activity concept is particularly suited because it allows to encapsulate the
well-defined semantics and to represent their atomic characteristics. This means that all
activities defined inside a block activity pattern must be completed before the
superordinated workflow can continue its execution.
We describe three example patterns with the Business Process Modeling
Notation (BPMN). The complete set of patterns can be found in [Thom 2006a]. For each
pattern we describe “context”, “problem”, “forces” and “Solution”. The solution
includes one “design choice”. Currently we are working to improve the patterns
documentations (e.g., examples of the patterns use, how they can be implemented and
additional design choices).
Pattern1:
A
PPROVAL
Context: A document must be evaluated by one or more organizational role.
Problem: How to model a human decision-making in the workflow process?
Forces:
- In case of multi-approvals (concurrent), the number of times that the
decision-making activity is repeated may vary depending on the level of
centralization of authority (less or more) as well as the direct supervision of
work existent in the organizational unit(s) where the process is executed.
- The decision-making activity is generally performed by a human.
- The decision-making activity must have more than one kind of response (e.g.
approval and reproval).
Solution: To include in the workflow, a human activity that characterizes a point of
decision-making on the sub-product in question (e.g. a document requiring approval).
This decision-making activity will be repeated according to the level of centralization
existent in the organizational units where it is executed. Figure 3 shows a single
approval.
Figure 3: Approval pattern
In Fig. 3 an organizational role reviewer performs a document review either
resulting in an approval or disapproval. In case of multi-approvals, it would be
necessary concurrent activities. The “Make final decision”(cf. Figure 3) would be
executed only when all reviewers had performed their revisions The revisions would be
then performed multiple times in parallel (concurrent) or in sequence (iterative)
according to the number of organizational roles specified or until disapproval occurs.
Generally, the number of organizational roles is connected to the level of centralization
(in high positions of the organization) with respect to decision-making.
Pattern 1:
U
NIDIRECTIONAL
P
ERFORMATIVE
Context: In a workflow, there is a moment that the process must request the execution
of an activity to the system or to an organizational role involved in the process.
Problem: While modeling the workflow, how to send a request without waiting for the
result of the activity execution?
Forces:
- The receiver’s response is not required.
- The process must keep its execution without waiting for the activity to be
completed.
- The request can be done to the system or a human.
Solution: To include in the workflow model the sequence of activities (see Fig. 4)
representing the unidirectional performative message. The sequence of activities
comprises the generation of a work item in the receiver worklist. However, the
workflow does not wait for a receiver response to continue execution.
Figure 4: Unidirectional Performative Pattern
A sender uses unidirectional performative messages to request the execution of an
activity from a receiver. As shown in Fig. 4, an activity execution request results in a
work item being assigned to a receiver (i.e., a specific workflow participant responsible
for the activity execution). After that, the process may continue execution without
waiting for a response.
3.
N
OTIFICATION
Context: During the process execution, some specific events have more relevancy such
that the process must inform some organizational roles about them.
Problem: While modeling the workflow, how can we keep the involved roles informed
about some process instance events?
Forces:
- The notification must be sent by an electronic way.
- The process does not have to wait for a reading response to continue execution.
- The notification generally contains the status of a process activity (e.g., completed,
document approved, rejected).
Solution: To include in the workflow the structure concerning the notifying and do not
wait for response. Figure 5 shows how this structure works. There’s the sending of the
notification and its receiving. These activities must inform the involved roles in the
process about news inherent to the respective workflow, such as, for example, the
approval (or reproval) of a document, or a task that has achieved its timeout.
.
Figure 5: Notification Pattern
3. Evidencing the Existence of Workflow Activity Patterns through the
analyze of Real Workflow Processes
With the goal to search for the existence of the workflow activity patterns in real
applications we analyzed 190 workflow processes. These workflows have been modeled
with the Oracle Builder tool and have stemmed from 12 different organizations related
to different application domains. Notice that we analyzed workflow models and not
instances or logs generated by the execution of them. Specially because the semantic of
the activities were important to the identification of the processes. Table 1 characterizes
the workflows which were subject of the analyses.
Size of the
company Kind of decision-making Examples of workflow processes
(models) we analyzed Number of workflow
processes analyzed
1 small Decentralized Management of internal activities 17
1 large Decentralized TQM and management of activities
11
6 large Centralized TQM; control of software access;
document management 133
4 large We had no access to
information about these
companies
Help Desk, User feedback;
document approval
29
Table 1: Core characteristics of the analyzed workflow processes
We have obtained the following results from the case study, i.e. the workflow
process analyses:
a) evidence with high probability that the workflow activity patterns exemplified
in this paper exist in real workflow applications;
b) evidence that the set of patterns is both necessary and sufficient to model all
190 workflows analyzed; and
c) identification of sequence of patterns based on the classification of the
workflow processes in Human-Intensive and System-Intensive [Le Clair 2007].
3.1. Method Used to Analyze the Workflow Processes
For each process pattern we calculated its support value (S). In the context of this paper,
S represents the number of occurrences of each pattern (P) in a set of 190 workflows.
For those workflows comprising more than one occurrence of the same pattern just one
was considered. The following formula was considered to calculate the support:
S = F (P)
T
T
Where:
F(P) = frequency of a specific pattern in the
total set of workflow models
T
T
= total number of workflow models
Initially, we identified and annotated semantic process patterns in all workflows
we analyzed. Afterwards, for all workflows we counted the number of occurrences of
each pattern. The obtained result then was divided by the total number of analyzed
workflows (i.e. 190 in our case). Accordingly, the (P) for this calculation corresponds to
a specific pattern while T
T
means the set of workflows.
3.2. Frequency of Workflow Activity Patterns in Real Workflow Processes
The
UNIDIRECTIONAL
and
B
I
-D
IRECTIONAL
P
ERFORMATIVE
P
ATTERNS
,
DECISION PATTERN
,
NOTIFICATION
PATTERN
and
I
NFORMATIVE PATTERN
are not dependent on specific application domains or
organizational structure aspects. This fact mainly explains why they were identified with
high-probability in practically all workflows analyzed. The same applies to the
APPROVAL
PATTERN
. This can be explained by the high degree of centralization on decision-making
existing in the organizational units for which we analyzed their workflows. This high
centralization implies the use of approval activities. Besides that, several workflows
belong to applications related to approval contexts. By contrast, most of the workflows
analyzed do not comprise
QUESTION
-
ANSWERING
activities. Figure 6 graphically illustrates
the frequency of each pattern in the set of workflow processes analyzed.
Frequency of each pattern in the set of 190 workflow models
83%
65%
54%
64%
16%
2%
60%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Unidirectional
Performative
Bi-directional
Performative
Notification Decision Informative Question-
answering
Approval
Support
Figure 6: Frequency of the workflow patterns in real workflow processes
3.3. Discussing the Completeness of the Semantic Process Patterns for Workflow
Modeling
The main goal of the study case presented in this paper was the measurement of the
frequency which each one of the workflow activity patterns happens in the set of
workflow processes that has been analyzed. This was done in order to verify whether
these business functions (e.g., task execution request, notification) could really be
considered as patterns with high probability of reuse in business as well as workflow
process design at least those with similar characteristics of the processes we analyzed.
While some patterns were identified only by the analyses of the activity
description (e.g., decision, approval and notification patterns), others required a more
detailed analysis. For instance, the
INFORMATIVE PATTERN
(see [Thom 2006a]) was
identified in activities where the user provides an information to the system (e.g., by the
fulfillment of a field in the context of an activity). In the case of the unidirectional and
bi-directional performative patterns, both the activity description and its execution result
(i.e., mandatory or not to trigger the next activity in the process) were important to
measure how often the patterns occur.
What really surprised us was the fact that all analyzed workflow processes can
be defined as a composition of the investigated patterns (see Figure 7 for an example).
That is, the set of semantic process patterns is necessary and sufficient to design all 190
real workflows that were subject of the mining effort. In each process, a specific process
pattern may appear zero or more times combined with other patterns.
This fact can be considered as a very important one which points out to new
questions to be investigated as part of a future work. For instance, how much could this
set of patterns be helpful if it was to be integrated into a workflow design tool? One
could think of an intelligent software module which relies on both a workflow activity
patterns repository in order to help designers to complete their workflow design. First
initiative in this approach we present in [Thom 2007a] and [Thom 2007b]. Figure 7
shows a workflow process sample where all activities match either a semantic process
pattern.
Figure 7: A payment process built up exclusively from the combination of
workflow activity patterns and control flows
Informative
pattern
Notification
pattern
Bi-directional
pattern
Notification
pattern
Unidirectional
Performative
pattern
Notification
pattern
Decision
pattern
Approval
pattern
Unidirectional
Performative
pattern
4. Identifying Sequences of Workflow Activity Patterns in Real Workflow
Processes
Some process fragments can occur many times in the same process definition [Thom
2006a]. Each time a specific fragment occurs, there may have successive process
fragments that can also occur with more frequency than other fragments.
With the objective of evidencing which are the workflow activity patterns that
succeed with more frequency one specific pattern, we analyzed 151 of the 190 workflow
processes, which were subject of our case study. Again we analyzed the workflow
models and not the execution logs.
Initially we made a preliminary investigation where we analyzed, for each
process activity and its respective workflow activity pattern associated, what was the
subsequent(s) pattern(s) that followed them. By doing that, we noticed that most of the
patterns pairs had low support value and confidence. Figure 8 illustrates one of the
obtained results of this first analysis. Notice that even the most frequent pair of this
example (
DECISION PATTERN
NOTIFICATION PATTERN
) has a low probability (29%).
Figure 8: Decision Pattern Subsequent Patterns on the preliminary mining
By analyzing the workflow processes, we observed that some kinds of patterns,
and their respective pairs, appear with more frequency in determined types of processes.
We notices that process including decision-making activities (i.e., approvals), there
exists bigger probability that the notification pattern be the follow construction after
those decision activities. This fact is mainly explained because the organizational roles
involved to the processes want to be informed about the result of the approval process.
In order to raise the support and confidence of the pattern pairs, we decided to
classify the processes into business process categories. For that, we studied some of the
main classifications found in the literature [Hammer 2001], [Harrington 1991],
[Dowson 1987], [Leymann 1999]. However, most of them classify processes by
application domains. Accordingly, those approaches were not feasible to our analyses
because the set of processes we were investigating do not cover all the categories
described on these works. We chose then the classification proposed in [Le Clair 2007]
where processes are divided into system-intensive and human-intensive.
The system-intensive processes are characterized by being handled on straight-
through basis, which means that there is minimal or no human touch and few
exceptions. The human-intensive processes require people to get work done by relying
on and interacting extensively with business applications, databases, documents and
other people. This type of process requires human intuition or judgment for decision-
making during individual steps.
Having this clear separation between the processes, we divided the set of studied
processes into processes that have human interaction and processes that don’t have
human interaction. The result of this classification is 31 processes system-intensive and
120 processes human-intensive. Figure 9 shows the results of this investigation over
decision patterns in system-intensive processes. Notice that the probability of having a
DECISION PATTERN
NOTIFICATION PATTERN
pair has increased to 50%.
Figure 9: Decision pattern subsequent patterns by mining only the system-
intensive processes
5. Conclusions
In this paper we presented samples of workflow activity patterns which can be used to
design business process and workflow models respectively. In particular we reported the
results of empirical studies we had performed with the main of to measure the frequency
with which each workflow activity pattern occurs within a set of 190 workflow
processes. This analysis was accomplished in order to verify whether specific business
functions frequently found in business processes (e.g., task execution request,
notification, approval) may be considered as patterns with high probability for reuse. We
also showed that by dividing the processes we analyzed into system-intensive and
human-intensive it was possible to identify pairs of patterns which were more frequently
present in the processes.
The main advantages of this approach can be summarized as follows: (a) the
completeness and necessity of the workflow activity patterns has been evidenced at least
for the design of the workflow processes subject of our analyses; (b) the patterns are
tool-independent and which make them easier to be adapted for any business process
modeling tool; (c) it is a small set of patterns which may reduce complexity in user
learning;
The result of the case study we presented in this paper will be used in the
development of a suite to the analysis and properties verification in workflow
specification (e.g., correctness, completeness, deadlock, processes equivalence, livelock,
model checking). This suite will have a knowledge database that will store the activity
workflow patterns as well as the results of our case study. We consider that this
knowledge database will help on matching the patterns on process that will be extracted
from legacy code and will help the user on designing the business process from scratch.
By using the pattern pairs, we can help the user by suggesting which pattern is better
combined with the one that he/she has already modeled.
As future work we intend to perform additional analyzes considering workflows
from different application domains (e.g., health insurance and automotive). Our goal is
to identify not only pairs of patterns but also sequences of workflow patterns, including
more than three patterns in sequence. In this context we also intend to continue studying
the workflow classifications so that we can find more specific classification and with
smaller granularity to divide the set of processes. A less generic classification will be
useful when we try to converge on the user needs using just a few steps Finally we
consider making an experiment for comparing process modeling with and without
pattern support.
Acknowledgements
The authors would like to acknowledge the Coordination for the Improvement of
Graduated students (CAPES), the Institute of Databases and Information Systems
of the University of Ulm (Ulm, Germany) and the Informatics Institute of Federal
University of Rio Grande do Sul (Porto Alegre, Brazil).
References
Aalst, W.M.P. van der. (2003) “Workflow Patterns”. In Distributed and Parallel
Databases, 14(3), pages 5-51.
Bernstein, P. A.; Hadzilacos, V.; Goodman, N. (1987) “Concurrently Control and
Recovery in Database Systems”. Reading: Addison-Wesley. 370p.
Barros, A.; Dumas, M.; Hofstede, ter A. (2005) Service Interaction Patterns. In W.M.P.
van der Aalst, editor, Proceedings of the 3rd International Conference on Business
Process Management, v. 3649 of Lecture Notes in Computer Science, pp. 302-318.
Springer-Verlag.
Dowson, M. (1987) Interation in the Software Process Review of the 3rd International
Software Process Workshop. In Proceedings of the 9th international Conference on
Software Engineering (Monterey, California, United States). International
Conference on Software Engineering. IEEE Computer Society Press, Los Alamitos,
CA, 36-41.
Flores, F. (1998) “Computer Systems and the Design of Organizational Interaction”.
Hammer, M.; Champy, J. (2001) “Reengineering the Corporation: A Manifesto for
Business Revolution”. HarperBusiness. 257p.
Harrington, H. J. (1991) “Business Process Improvement: The Breakthrough Strategy
for Total Quality, Productivity, and Competitiveness”. McGraw-Hill. 274p.
Keller, G.; Nüttgens, M.; Scheer; A.-W. (1992) Semantische Prozessmodellierung auf
der Grundlage “Ereignisgesteuerter Prozessketten (EPK)”. Heft 89, Institut für
Wirtschaftsinformatik, Saarbrücken, Germany.
Le Clair, C.; Teubner, C. (2007) The Forrester Wave: Business Process Management
For Document Processes, Q3 2007.
Leymann, F.; Roller, D. (1999) “Production Workflow: Concepts and Techniques”.
Prentice Hall. 479p.
Malone, T. W.; Crownston, K.; Herman, G. A. (2004) “Organizing Business
Knowledge”.
Medina-Mora, R.
(1992) “
The action workflow approach to workflow management
technology”.
Muehlen, M. zur. (2002) “Workflow-based process controlling”: foundations, design,
and application of workflow-driven process information systems. Berlin: Logos
Verlag.´299 p.
Russell, N. (2004) “Workflow Resource Patterns”. Brisbane: Queensland University of
Technology.
Russell, N.; Hofstede, A. H. M Ter; Edmond, D. (2005) “Workflow Data Patterns”. In:
Proceedings of the 24th ER, pp. 353-368.
Russell, N.; Aalst, W.M.P. Van Der; Hofstede, A. Ter. (2006) “Workflow Exception
Patterns”. In: Proceedings of 18th CAiSE, 18, p.288-302.
Thom, L. H.; Iochpe, C.; Amaral, V. L. do; Viero, D. M. (2006a) “Toward block
activity patterns for reuse in workflow design”. In: Workflow Handbook of WfMC,
pp. 249-260.
Thom, L. H. (2006b) “A Pattern–based Approach for Business Process Modeling”.
PPGC: Porto Alegre. Available at: http://www.biblioteca.ufrgs.br/bibliotecadigital/.
(Thesis).
Thom, L. H; Lau, J.M.; Iochpe, C.; Mendling, J. (2007a) “Extending Business Process
Modeling Tools with Workflow Pattern Reuse”. In: Proceedings of 9th International
Conference on Enterprise Information Systems, Funchal, Madeira, Portugal.
Thom, L. H., Chiao, C. M., Iochpe, C., Hess, G. N., Nascimento, G. S., Reichert; M
(2007b): “Towards an Intelligent Designer based on the Reuse of Workflow
Patterns”. In: 1
st
Brazilian Workshop on Business Process Management to be held in
conjunction with WEBMEDIA 2007, Gramado, Brazil.
Thomas, O.; Scheer, A.-W. (2006) “Tool Support for the Collaborative Design of
Reference Models - A Business Engineering Perspective”. In: 39th HICSS-39, CD-
ROM / Abstracts Proceedings. .
Workflow Management Coalition, 2005. “Process Definition Interface”: XML Process
Definition Language. Doc. Number: WFMC-TC-1025. (2005).
