=Paper=
{{Paper
|id=Vol-1612/paper16
|storemode=property
|title=Consistent Integration of Decision (DMN) and Process (BPMN) Models
|pdfUrl=https://ceur-ws.org/Vol-1612/paper16.pdf
|volume=Vol-1612
|authors=Laurent Janssens,Ekaterina Bazhenova,Johannes De Smedt,Jan Vanthienen,Marc Denecker
|dblpUrl=https://dblp.org/rec/conf/caise/JanssensBSVD16
}}
==Consistent Integration of Decision (DMN) and Process (BPMN) Models==
https://ceur-ws.org/Vol-1612/paper16.pdf
Consistent Integration of Decision (DMN) and
Process (BPMN) Models
Laurent Janssens1,2 , Ekaterina Bazhenova3 , Johannes De Smedt1 , Jan
Vanthienen1 , and Marc Denecker2
1
KU Leuven, Leuven Institute for Research on Information Systems, Leuven,
Belgium
{laurent.janssens,johannes.desmedt,jan.vanthienen}@kuleuven.be
2
KU Leuven, Declarative Languages and Systems, Leuven, Belgium
{laurent.janssens,marc.denecker}@kuleuven.be
3
University of Potsdam, Hasso Plattner Institute, Potsdam, Germany
ekaterina.bazhenova@hpi.de
Abstract. Organizations use business process and decision management
techniques to run their businesses more efficiently. Modeling the business
processes and decisions is a vital part of this. The recently proposed De-
cision Modeling and Notation (DMN) standard introduces a declarative
approach for modeling decisions and aims at the separation of decision
logic from business processes. Decoupling decisions and process models
is crucial for the flexibility and maintainability of business processes.
However, this aspect has received little attention, except in straightfor-
ward situations. In this paper we identify five integration scenarios and
provide a formal basis for DMN models. Using these scenarios and for-
malization we illustrate how to achieve consistent integration of decision
and process models.
Keywords: Decision Modeling, Process Modeling, BPMN, DMN
1 Introduction
Business process management (BPM) and decision management (DM) are being
used to improve the efficiency and effectiveness of organizations. Companies
are interested in running effective and competitive processes, and use BPM to
describe and improve these processes. The BPMN standard [12] allows processes
to be described visually in a structured, and executable way. However, while
decision making is an important aspect of BPM, BPMN supports no clear-cut
way to represent this. Similarly decision management is being used to map the
decisions made within the business, and to evaluate, and improve these decisions.
However, there is little emphasis on how these decisions are made in a process
context. Moreover, complex decisions are often modeled as processes to guide
the business through the decision making process. This can result in difficulties
to maintain and redesign decisions, as mentioned in [4].
Copyright c by the paper’s authors. Copying permitted only for private and academic
purposes.
In: S. España, M. Ivanović, M. Savić (eds.): Proceedings of the CAiSE’16 Forum at
the 28th International Conference on Advanced Information Systems Engineering,
Ljubljana, Slovenia, 13-17.6.2016, published at http://ceur-ws.org
122 Laurent Janssens et al.
Recently the emphasis has been on integrating decision and process man-
agement. The DMN standard [13] was developed to support decision modeling
complementary to processes represented using BPMN. In this paper we describe
five integration scenarios, which allow us to identify situations where integration
plays an important role. For two of these scenarios we discuss the challenge of
achieving consistent integration, i.e. integration between a decision and process
model which preserves a consistency with respect to information requirements.
To support our approach we introduce a formalization of decision requirements
in DMN.
2 Background
Recently the separation of processes and decision logic has become an evident
trend. In response to this trend the OMG group developed the new Decision
Model and Notation (DMN) [13] standard as a standard approach for modeling
decisions within organizations [15]. At the same time emphasis is placed on
integrating decision management and process management.
The DMN standard allows to model and describe decisions in a declarative
way on two levels, the requirements level and decision logic level. For the first
level decisions requirement diagrams (DRD) are used to represent the informa-
tion requirements of the decisions in the model. These diagrams can consist
of several types of elements, decisions, input data, business knowledge models,
and knowledge sources. In this paper we take abstraction of knowledge models
and sources, since our approach relies solely on the decisions and input data.
Information requirements in the DRDs represent the requirements of decisions
in terms of subdecisions and input data, depicted using arrows going from the
DMN_Credit_Eligibility
requirement to the decision. An example of a DRD is given in Figure 1.
Eligibility
Employment
Medical
history Health risk Financial risk
Smoking
habits Salary Homeowner
Lifestyle
Household Working
situation environment
Fig. 1. Decision requirement diagram for the credit eligibility decision
The second level uses the declarative FEEL expression language to describe
the decision logic behind every decision. DMN provides no guarantees for ef-
ficient resolution of decisions, this is left to the invoking context. Additionally
Consistent Integration of Decision (DMN) and Process (BPMN) Models 123
decisions have no side-effects and storage of outputs and intermediate results is
unconstrained and vendor-dependent. DMN is designed to be complementary to
the BPMN standard, as discussed in [5].
In the trend towards integration several situations can be identified. Basic
solutions see processes represented using only BPMN, or decisions using only
DMN. In other cases decisions are often emulated using intricate process con-
trol flows, which can result in cascading gateways. These hidden decisions must
be identified in the process. After identifying and modeling these decisions the
resulting model must be integrated consistently with the process model.
In more complex processes several decisions might influence the flow and
result. Correct representation and invokation is crucial for a proper understand-
ing. Since the decisions’ context can have an impact on their results information
requirements have to be taken into account when designing the process. This
contrasts the declarative nature of DMN. Here true integration of process and
decision models is required. However, these situations have received little atten-
tion. By integrating decision models and process models, new challenges arise.
One of these challenges is ensuring the consistency between the decision models
and associated process models.
3 Consistent Integration of DMN and BPMN Models
3.1 Integration Scenarios
When considering decisions in processes five scenarios can be identified. This sec-
tion outlines each of these scenarios, discussing their occurrence and the possible
relation between them. Challenges for each scenario are identified and existing
solutions discussed where appropriate.
Scenario 0. Processes without Decisions In cases where a process consists
of a series of predetermined activities and the control flow is simple. Thus, there is
no opportunity for integration with a decision model, since there are no decisions
to be made. Often this will be subprocesses of more complex business processes.
It is important to note, processes where decisions are hard-coded using com-
plex control flows, are bad examples of this scenario. Typically, changing these
decisions is difficult as process paths have been fixed, making it hard to re-
evaluate the outcomes and factors for invoking the decision [16].
Scenario 1. Local Decisions In this scenario local decisions ensure simple
separation of control flow and decision logic. These decisions, characterized by
their atomic outcome and local effect, can still be complex and require multiple
subdecisions. However they have no impact on the rest of the process and can
hence be inserted safely. Decision management is handled by providing decision
outcomes as inputs for the process control flow. This simplifies and increases the
flexibility of the process. These models can be obtained from mining techniques
[1, 9].
124 Laurent Janssens et al.
Scenarios 2 & 3. Processes with Interrelated Decisions The second and
third scenarios include situations where multiple interrelated decisions are made
during the process. Often both the flow and result of the process will depend
heavily on the outcomes of these decisions.
In the second scenario the decisions are placed directly before the business
activities requiring their output, thus Here the information pertaining to the
relation of the decisions is kept externally in the decision model. This constitutes
an inefficient ordering of the decisions and requires additional data management,
to store results of subdecisions.
In the third scenario the structure of the decisions is used while modeling
the process. By ordering related decisions according to their requirements no
decision has to be made twice, increasing the efficiency of the process.
For process models fitting either of these scenarios it is crucial to have a de-
cision model, since information on the relation between the different decisions is
required for understanding and executing the process. In both of these scenarios
it is important that the process model is consistent with the decision model,
discussed further in Section 3.4.
Scenario 4. Knowledge Intensive Processes The last scenario constitutes
the opposite situation as Scenario 0, here the process is actually a single decision.
This scenario occurs when a process is highly knowledge intensive, i.e. when the
process is used to decide a single output. In these cases only a decision model is
required. As in Scenario 0 integration offers little benefit in this situation.
3.2 Example Case
In our example, customers can apply for credit at the company, thus starting a
new process instance. After receiving the necessary information a decision has
to be made about whether the customer is eligible for a loan. This decision is
based on the customer’s health and financial situation, to minimize the risk of
non-payment. The requirements for this decision are illustrated using DMN in
Figure 1. If the customer is eligible for a loan an insurance is filed and the loan
accepted, if not the loan is rejected.
The credit application process is depicted using BPMN in Figure 2. The
process contains two business activities, the eligibility decision, and filing the
insurance. However, the decision requirement diagram in Figure 1 shows the
two data objects Health risk category and Financial risk category are
actually the respective outputs of the Health risk and Financial risk de-
cisions. Using these intermediate results in the process model poses problems
for flexibility and maintainability. If the structure of the Eligibility decision
changes and these decisions are no longer needed, the process model will nec-
essarily need to change. To resolve this issue, the decisions determining these
intermediate results should be added to the process model.
BPMN_Credit_Eligibility_Use_Intermediate_Results
Consistent Integration of Decision (DMN) and Process (BPMN) Models 125
Not eligible
Reject application
Determine Eligible File
Eligibility insurance
Credit application Accept application
Financial Health risk
risk category
category
Fig. 2. Business process where intermediate decision results are used
3.3 Formalization
To support our approach we introduce a formal basis for decisions and require-
ments in DMN models.We take abstraction from the use of Business Knowledge
Models and Knowledge Sources. However, all definitions and theorems provided
can be readily extended to include the use of these concepts.
We adopt the definition of a decision requirement diagram from [2] and
slightly extend it to identify an important constraint, i.e. that no cycles can
occur in the information requirements.
Definition 1. A decision requirement diagram DRD is a tuple (DDM , ID, IR)
consisting of a finite non-empty sets of decision nodes DDM , input data nodes
ID, and directed edges IR representing the information requirements such that
IR ⊆ DDM ∪ ID × DDM , and (DDM ∪ ID, IR) is a directed acyclic graph.
In the service-oriented approach a decision is implemented as a service offer-
ing a single decoupled point of entry to the business logic of that decision. This
has benefits for automation and flexibility, since the process only needs informa-
tion about the service’s interface. [11]. To define the interface of a decision we
define the input requirement set and a decision’s output set in Definition 2.
Definition 2. The interface of a decision D is a tuple (dirsD , OD ). Where the
decision input requirement set dirsD is the set of input data directly or indirectly
required by D. The output set OD is the set of all possible outputs of D.
While the input is clearly indicated in the DRD, as in Figure 1, outputs are
not specified explicitly. They are made explicit on the decision logic level. In this
case we assume the output is identified by the decision’s name.
To identify incorrect uses of decisions in process models it is important to
know their structure and meaning. As in the example in Figure 1 decisions are
often structured to use the results of other intermediate decisions, called subde-
cisions. The subdecision outputs are defined as intermediate results in Definition
3.
Definition 3. An output O is an intermediate result of decision D if and only
/ OD and there exists a subdecision D0 of D for which O ∈ OD0 .
if O ∈
126 Laurent Janssens et al.
In the context of DMN, the execution of a decision is called invoking that
decision. Definition 4 defines when a decision can be invoked, based on the
available data.
Definition 4. A decision D is invokable from a set of data elements S if it
contains all the required inputs, dirsD ⊆ S.
This formal framework will allow us to identify inconsistencies between a
decision model and an associated process model. The process of identifying and
resolving these inconsistencies is described in the following subsection.
3.4 Consistent Integration for Scenarios 2 and 3
Using the formalization from the previous subsection, we can define when a
process model fitting integration scenario 2 or 3 is consistent with an associ-
ated decision model. Definition 5 defines consistency based on the availability of
required input and the use of decision outputs.
Definition 5. A business process model is consistent with a decision model iff
no intermediate results of invoked decisions are used, and each decision invoked
in the process, must be guaranteed to be invokable at that process stage.
When addressing inconsistencies we assume changing a decision model will, in
most cases, change the decisions’ results, while changing the order of activities
of a process in a systematic way will have little or no impact on the result.
Under this assumption our approach uses the decision model as a reference for
consistent integration.
Resolving the Use of Intermediate Results The process model in Fig-
ure 2 violates the first condition of Definition 5. Intermediate results of the
Eligibility decision are used, i.e. the process model is inconsistent with the
decision model in 1. We identify the Health risk category and Financial
risk category as being intermediate results, produced by the Health risk
and Financial risk decisions. Thus, these decisions must be added to the pro-
cess model. The correct decision order differs based on the scenario. Conformance
with scenario 2 requires the two decision activities producing the intermediate
results to be placed directly in front of the File insurance activity. Confor-
mance with Scenario 3, is achieved when the requirement structure of the decision
model is used to determine the order of the decisions. The main benefit of invok-
ing these decisions explicitly in the model is flexibility. Should the Eligibility
decision change, so that the subdecisions are no longer relevant, this will not
affect the process. The resulting process, conformant with scenario 3, is shown
in Figure 3.
BPMN_Credit_Eligibility_Model_Unavailable_Data_Solved
Consistent Integration of Decision (DMN) and Process (BPMN) Models 127
Health risk
category Financial
risk
category
Determine Determine Determine Eligible
Financial File
Health Risk Eligibility Insurance
Risk
Credit application Accept application
Not eligible
Reject application
Fig. 3. Consistent process model, conformant with scenario 3
4 Related Work
Ensuring the consistency of process models using data has been investigated in
numerous works already. Many data-aware process modeling approaches have
been proposed that use different types of data representation and input, such
as the ontology-based knowledge-intensive approach in [14], enhancing DMN
and declarative process models [10] and all works concerning colored Petri nets
[8]. The latter offers a complete formalized approach to deal with integration,
sticking mainly to local data. Approaches such as [3] use Petri nets as they offer
ways to ensure consistency between data and process model. However, the focus
on data often downplays the holistic view that should be achieved to support
decisions, and does not deal with DMN.
The separation of concerns has enjoyed plenty of attention, mainly in the do-
main of software modeling and design [7]. They offer firm motivation for keeping
multi-perspective modeling tasks, such as control flows and decision making, iso-
lated. The externalization of business logic was already discussed in [6] in the
context of business rules. This need for separation carries over to decision logic.
In the field of process mining this attention to separation is equally important.
In [1] a method is described to extract the underlying decisions and simplify the
corresponding process models.
5 Conclusion and Future Work
In this paper we identify five modeling scenarios for the combination of decision
and process models. Two scenarios are identified where consistent integration
of decision and process models is vital to support the separation of concerns,
and related benefits to flexibility and maintainability. These scenarios include
situations where interrelated decisions are invoked within the process. For these
two scenarios we outline an approach which allows to identify and resolve in-
consistencies between the decision and process model. To support our approach
we introduce a formal basis for decisions and information requirements in DMN
models. Using this formalization our approach offers a systematic way to inte-
grate process and decision models, even for complex combinations of decisions
and process activities.
128 Laurent Janssens et al.
In our approach only the models are used, no translations or additional data
are required. This would allow it to be be introduced in modeling tools and used
as a guide while modeling, to guard the user from introducing inconsistencies
between process and associated decision models. As future work, we plan to
extend the introduced formalization of DMN and use this to further examine
the integration of business process and decision models, with added emphasis on
the data perspective.
References
1. Batoulis, K., Meyer, A., Bazhenova, E., Decker, G., Weske, M.: Extracting decision
logic from process models. In: Advanced Information Systems Engineering, pp.
349–366 (2015)
2. Bazhenova, E., Weske, M.: Deriving decision models from process models through
enhanced decision mining (2015)
3. Cabanillas, C., Resinas, M., Ruiz-Cortés, A., Awad, A.: Automatic generation
of a data-centered view of business processes. In: Advanced Information Systems
Engineering, pp. 352–366 (2011)
4. De Roover, W., Vanthienen, J.: On the relation between decision structures, ta-
bles and processes. In: On the Move to Meaningful Internet Systems: OTM 2011
Workshops. pp. 591–598. Springer (2011)
5. Debevoise, T., Taylor, J.: The MicroGuide to Process Modeling and Decision in
BPMN/DMN. CreateSpace Independent Publishing Platform (2014)
6. Goedertier, S., Vanthienen, J.: Compliant and flexible business processes with busi-
ness rules. In: 7th Workshop on Business Process Modeling, Development and
Support at CAiSE’06. pp. 94–104. Presses Universitaires de Namur (2006)
7. Gordijn, J., Akkermans, H., Van Vliet, H.: Business modelling is not process mod-
elling. In: Conceptual modeling for e-business and the web, pp. 40–51. Springer
(2000)
8. Jensen, K.: Coloured petri nets. Springer (1987)
9. de Leoni, M., van der Aalst, W.M.: Data-aware process mining: discovering de-
cisions in processes using alignments. In: Proceedings of the 28th annual ACM
symposium on applied computing. pp. 1454–1461. ACM (2013)
10. Mertens, S., Gailly, F., Poels, G.: Enhancing declarative process models with dmn
decision logic. In: Enterprise, Business-Process and Information Systems Modeling,
pp. 151–165. Springer (2015)
11. Mircea, M., Ghilic-Micu, B., Stoica, M.: An agile architecture framework that
leverages the strengths of business intelligence, decision management and service
orientation. Business Intelligence-Solution for Business Development (2011)
12. OMG: BPMN (2011), http://www.omg.org/spec/BPMN/2.0/
13. OMG: DMN (2015), http://www.omg.org/spec/DMN/1.0/
14. Rao, L., Mansingh, G., Osei-Bryson, K.M.: Building ontology based knowledge
maps to assist business process re-engineering. Decision Support Systems 52(3),
577–589 (2012)
15. Taylor, J., Fish, A., Vanthienen, J., Vincent, P.: Emerging standards in decision
modeling. BPM and Workflow Handbook series (2013)
16. Vanthienen, J., Caron, F., De Smedt, J.: Business rules, decisions and processes:
five reflections upon living apart together. In: Proceedings SIGBPS Workshop on
Business Processes and Services (BPS’13). pp. 76–81 (2013)