This thesis proposes a novel Business Process Architecture (BPA) design method for overcoming limitations of currently available alternatives. The proposal { named domain-based BPA (dBPA) method { uses domain models as a starting point for building BPA models. The evaluation of the dBPA method revealed that it was perceived as useful and likely to be used in practice.
The thesis focuses on Business Process Architecture (BPA) design. A BPA consists of an organized overview of business processes and their relations. An analysis of the literature of the BPA research eld revealed problems regarding BPA design methods (see Sect.3) that ultimately challenge the most relevant quality metrics { and predictors of use { of BPA methods, namely, ease of use and usefulness.
The aforementioned issues led to identifying the need for a new BPA design
method: the domain-based BPA (dBPA) method. The research, which followed
the Design Science Research (DSR) paradigm [
Data models represent information handled
by organizations. A domain model represents
the structure of a domain by showing its main
concepts and how they relate to one another
as classes and associations, respectively (see
Fig. 1). An object lifecycle (OLC), can be
built for a class to represent the dynamic
aspects of data, namely states and transitions.
Unlike traditional activity-centric approaches
to process modeling (e.g. [
The selection of a suitable architecture description language for the dBPA method
was based on requirements regarding abstract syntax and cognitive e ectiveness
of Business Process Model and Notation (BPMN) [
The dBPA method is based on a number of foundations. First, regarding the identi cation of business entities and relationships: { Each domain model class de nes, at
most, one business entity. { Each domain model association de nes one or more relationships between the OLCs of the partaking classes.
Second, integration of structure and dynamic data models involves specializing the domain model classes with dynamic hierarchies such that: { A dynamic hierarchy is built by specializing the original class with sub-classes that hold a bi-univocal relationship with the OLC states. { The properties of the hierarchy are de
ned in a way that they are consistent with the organization of OLC states.
Third, formal de nitions used for the method include the concepts of: association a, BPA model , business entity b, business process ', class c, data dictionary dd, domain model w, and object lifecycle l. 6
The dBPA method has six steps that are { generally speaking { sequential:
1. Prepare domain model. Some transformations are applied to the original
domain model to produce a semantically-equivalent domain model w where
all associations are directed and power types are replaced by hierarchies.
2. Identify business entities. The set of business entities B corresponds to
the set of all classes c 2 w that are not sub-classes.
3. Identify states of domain model classes. The set of states for each
business entity b 2 B is identi ed by using two sources of information: the
domain model w and its complementary information (e.g. its data dictionary
dd ), and domain expert knowledge.
4 Materials reported in this chapter are partially published in [
The conceptual evaluation of the proposed method was done using a novel
framework based on the works by [
G1 (Artifact design): The dBPA method was designed for overcoming
issues of currently available BPA design methods. The reuse of structured
business knowledge was implemented as using domain models as an input for the
method. The method also considers four types of process relations that, though
mentioned in the literature, are rarely considered together in BPA design
methods. Including them allows building more complete BPA models. The use of an
industry-standard language was done by proposing an ArchiMate viewpoint that
would improve cognitive e ectiveness of the resulting BPA models. The proposed
method has some limitations that could be addressed in future research: (i) it
works with a simpli ed version of the domain model; (ii) it assumes that the
granularity of all OLC transitions is such that it can constitute one business
process; and (iii) it is domain model dependent. Future works include: partial
automation of the method, and development of tools for BPA analysis.
G2 (Instrument design): Two instruments were tailored for evaluation of the
proposed method. The framework for qualitative analysis was found useful for
the description, evaluation, and comparison of (other) BPA design methods at
a conceptual level. Future lines of work might include validating this framework
with a panel of experts. An instantiation of the MEM was successfully used
for quantitative analysis. Future research could include improving some issues
6 Materials reported in this chapter are partially published in [
G3 (Knowledge): The evaluation provided signi cant evidence that the proposal was perceived as useful and not easy to use. The user study also provided preliminary evidence that the proposed method is an improvement regarding ease of use, usefulness, and intention to use. These results are consistent with the main design decisions of the proposed method. In the end, however, further research is needed to gather signi cant evidence.
G4 (Prediction): In terms of adoption in practice, the study provided preliminary evidence in favor of the proposal. Again, future research is needed in this regard. Also, the study provided signi cant evidence stating that usefulness seems to be the main predictor of adoption in practice followed by ease of use.