Agile Quality Requirements Elaboration: A Proposal and Evaluation
Wasim Alsaqaf
a
, Maya Daneva
b
and Roel Wieringa
School of Computer Science, University of Twente, Enschede, The Netherlands
{w.h.a.alsaqaf, m.daneva, r.j.wieringa}@ utwente.nl
Keywords: Agile Projects, Quality Requirements, non-Functional Requirements, Large-Scale Distributed Projects,
Requirements Engineering, Goal-Oriented Modelling, Design Science, Focus Group Study.
Abstract: The increasing success and user satisfaction of agile methods’ application in their original context (e.g. small
co-located teams), motivated large organizations to utilize agile methods to deal with the rapidly changing
markets and the distributed global workforce. Several studies have reported a variety of quality requirements
(QRs) challenges in large-scale distributed agile (LSDA) context, so a recent empirical study has identified
15 QRs challenges in LSDA projects. This paper proposes an approach based on the concept of goal
documentation to deal with the 15 QRs challenges reported previously. Our proposal, the Agile Quality
Requirements Elaboration (AQRE) approach, introduces a new organizational role and a two-step process to
elaborate high-level goals(s) into epics and user stories alongside QRs. The fitness and the usefulness of
AQRE are evaluated by using a focus group with eight practitioners in the IT department of a large Dutch
government organization. The evaluation indicated that 12 of the 15 QRs challenges could be mitigated by
the AQRE. Our main contribution is two-fold, (i) we proposed a solution approach to deal with QRs
challenges in LSDA context, and (ii) our evaluation provided empirical evidence about its usefulness in real-
world context.
1 INTRODUCTION
Agile software development methods have become
increasingly popular in the last decade. In 2021, the
15th annual State of Agile report (Digital.ai, 2021)
indicated the explosive increase of agile adoption.
This is in contrast to the traditional non-agile
approaches of software development which are
characterized as heavyweight rigid processes (De
Lucia and Qusef, 2010). Due to their rigidity,
traditional non-agile approaches fail to keep up with
the rapidly changing business environments (Helmy
et al, 2012). Agile on the other hand presents itself as
the way to deal with rapidly changing markets. In
fact, responding to change is one of the values stated
in the Agile Manifesto (Agile Alliance, 2001). Based
on these values, agile practices have been developed
to increase the involvement of the customers and to
deliver software products faster. The agile practices
include short iterations that end up with delivering a
new incremental release, addressing as many
requirements as needed to start an iteration, simple
a
https://orcid.org/0000-0002-0253-428X
b
https://orcid.org/0000-0001-7359-8013
incremental designs that will be evolved during the
iterations and peer reviews (Shore and Shane, 2007).
The necessity of reacting quickly to the rapidly
changing market, pushes large organizations to
believe that the success stories of agile methods'
application in the context they originally were
designed for (e.g. small co-located teams) are
repeatable in large-scale distributed agile (LSDA)
context. The 15th annual State of Agile report
(Digital.ai, 2021) states that the rapid changing
economical market and the distributed global
workforce do force organizations to take on their
distributed agile chances. In line with this, software
engineering scholars (Calefato and Ebert, 2019;
Smart, 2018) describe two trends which are
demanded in large-scale projects, namely, agile
transformation and embracing team distribution. The
application of agile methods – which were originally
designed for the context of small and co-located
teams – in LSDA however does not go smoothly
(Smart, 2018; Conboy and Carroll, 2019). In
particular, many studies have reported the neglect of
quality requirements (QRs) in agile context (Maiti
668
Alsaqaf, W., Daneva, M. and Wieringa, R.
Agile Quality Requirements Elaboration: A Proposal and Evaluation.
DOI: 10.5220/0012143100003538
In Proceedings of the 18th International Conference on Software Technologies (ICSOFT 2023), pages 668-679
ISBN: 978-989-758-665-1; ISSN: 2184-2833
Copyright
c
2023 by SCITEPRESS Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
and Mitropoulos, 2015; Alsaqaf, 2019; Inayat et al,
2014). In a recent empirical study (Alsaqaf et al,
2019) the authors have identified 15 challenges that
LSDA projects cope with when it comes to the
engineering of QRs. Given this background, in the
present paper we propose and evaluate an approach to
engineer QRs in LSDA where the identified
challenges (Alsaqaf et al, 2019) could be mitigated.
Our proposal was created by using a process
grounded on Wieringa’s Design Science (2014) and
drawing on concepts from goal-oriented requirements
engineering (GORE) (Pohl, 2010). Our very first
evaluation with practitioners indicates that the
approach may achieve its goal and could be useful. In
what follows, we first present background and related
work (Section 2). In Section 3 and 4, we introduce our
proposed approach and its very first evaluation
respectively. A discussion of the findings is in Section
5. We reflect on the limitations of our approach in
Section 6 and conclude in Section 7.
2 BACKGROUND AND RELATED
WORK
Three streams of research form the related work for
this paper: (i) publications on challenges and possible
solutions in the engineering of QRs in agile context;
(ii) publications on the use of GORE, and (iii)
publications on requirements modelling.
2.1 QRs Engineering in Agile
As already indicated, numerous empirical studies
revealed the neglect of QRs in agile context (Maiti
and Mitropoulos, 2015; Alsaqaf et al, 2019; Inayat et
al, 2014). A recent one (Alsaqaf et al, 2019) has
identified 15 challenges of five categories that LSDA
cope with. For clarity, we list these QRs challenges in
Table 1, where the first column shows the categories
of the challenges as reported in (Alsaqaf et al, 2019)
and the second column reports the specific challenges
of each respective category in the first column.
These authors (Alsaqaf et al, 2019) have reported
nine practices that agile practitioners use to cope with
those challenges. However, the study indicated as
well that the used practices could incur other
challenges such as adding hierarchies to the agile
teams which would take the flexibility of agile a step
back towards waterfall.
Table 1: The reported QR challenges.
Category Challenges
1.Teams
coordination and
communication
challenges
-Late detection of QRs
infeasibility
-Hidden assumptions in inter-
team collaboration
-Uneven teams maturity
-Suboptimal inter-team
organization
2. Quality assurance
challenges
-Inadequate QRs test
specification
-Lack of cost-effective real
integration test
-Lengthy QRs acceptance
checklist
-Sporadic adherence to quality
guidelines
3.QRs elicitation
challenges
-Overlooking sources of QRs
-Lack of QRs visibility
-Ambiguous QRs
communication process
4.Conceptual
challenges of QRs
-Unclear conceptual definition
of QRs
-Confusion about QRs
specification approaches
5. Architecture
challenges
-Unmanaged architecture
changes
-Misunderstanding the
architecture drivers
A 2017 systematic literature review (Alsaqaf et al,
2017) surveyed an array of proposed solutions to
counter the neglect of QRs in agile context (e.g.
(Kumar et al, 2013; Domah and Mitropoulos, 2015;
Farid and Mitropoulos, 2012; Farid and Mitropoulos,
2013)). Those proposals introduced several new
artefacts and roles to include in the engineering of
QRs in agile processes. A 2022 study (Rahy and Baas,
2022) introduced as well two new artefacts e.g.
Documentation Work Item and Safety Critical Work
Item to treat QRs in an agile way. However, creating
new artefacts or new roles and integrating them into
the existing agile processes could be experienced by
practitioners as ‘heavyweight’ additions that
unexpectedly cause new problems (Alsaqaf et al,
2019). In the same vein, the review of Abheeshta et
al. (Abheeshta et al, 2018) on the adoption of heavy-
documented agile frameworks reported the “moving
away from agile” as an important challenge among
others. Another recent study (Sherif et al, 2022)
explored published solutions to cope with QRs in
agile context (e.g. (Maiti and Mitropoulos, 2015; Jeon
et al, 2011)). However, the study (Sherif et al, 2022)
also revealed the lack of empirical evidence that these
proposed solutions work, which agrees with the
Agile Quality Requirements Elaboration: A Proposal and Evaluation
669
observation made by Inayat et al. (Inayat et al, 2015)
indicating a lack of evidence about the effectiveness
of new proposed solutions. Finally, Kopczyńska et al.
(Kopczyńska and Nawrocki, 2014) described the
SENoR method as response to the neglect of QRs in
agile. SENoR includes several brainstorm sessions
where QRs are elicited based on ISO25010 (ISO/IEC,
2011) and documented using predefined QRs
templates. However, Sherif et al. (Sherif et al, 2022)
pointed out that the narrow focus of SENoR uniquely
on QRs in isolation of other requirements, is actually
a weakness of this method.
2.2 Goal-Oriented Requirements
Engineering
Requirements engineering (RE) is the process of
elaborating stakeholders’ intentions into
specifications of the desired system or services
(Lamsweerde et al, 1998). Therefore RE needs to
understand those intentions that have to be fulfilled
by the desired system of services. In RE, those
stakeholders’ intentions are referred to as goals (Pohl,
2010). Using goals to drive requirements has been
popular (Horkoff, 2019) due to, among others, the
following benefits (Pohl, 2010): (i) clarifying the
context and the value of the system, (ii) driving and
guiding the identifying of the system requirements,
since for each goal a set of requirements can be
defined, (iii) giving the possibility for identifying and
evaluating solution alternatives, (iv) giving
guidelines to identify irrelevant requirements, (v)
giving guidelines for requirements completeness, (vi)
giving rational for the relevance of requirements, (vii)
giving guidelines for identifying and resolving
requirements’ conflicts and (viii) giving stability
since goals are not subject for frequent changes while
requirements are. In alignment with Pohl (2010),
Daneva et al. (2007) highlighted the importance of
assessing stakeholders goals early in de system
development phase to achieve a clear scope definition
which would guide the identification of the most
significant requirements. Whether these benefits have
been observed in LSDA projects and to what extent
Goal-Oriented Requirements Engineering (GORE)
adds value in that context has not been empirically
researched in much depth. However, one might think
of the potential usefulness of GORE in LSDA, from
the following perspective: in their work on QRs
challenges in LSDA projects, Alsaqaf et al. (2019)
reported several common root errors behind these
challenges: (i) suboptimal priorities assigned to
conflicted QRs, (ii) focusing too much on system’s
parts and losing the big picture, (iii) the emerging of
relevant QRs late in the development phase. The
authors (Alsaqaf et al, 2019) acknowledge that while
these mechanisms were observed in LSDA projects,
they are not unique to agile. As GORE was
introduced to cope with such situations in ‘traditional
contexts’, we thought that there would be no reason
to assume that GORE would not work for agile
projects. In fact, we believe that implementing GORE
concepts could eliminate several of the mechanisms
reported in (Alsaqaf et al, 2019), which in turn means
that it can serve to mitigate the challenges in Table 1.
In GORE, goals can be of different levels of
abstraction; for example, Cockburn (2000) reported
the following levels: (i) Cloud level a high-level
business goal that need to be decomposed into sub-
goals, (ii) Kite level a decomposed goal from the
Cloud level that represents an end-to-end system
process, (ii) Sea level – a user goal decomposed from
the Kite level that can be achieved by one person
within the end-to-end system process, and (iv) Fish
level a task (not a goal by itself) carried out along
with other tasks to achieve a user goal. Agile software
development (ASD) however does not use these
abstraction levels of goals. Instead, ASD uses the
terms themes, initiatives, epics and user stories.
Noreika et al. (2021) defined those agile terms as
follows: (i) Theme is a logical organization and
aggregation of related user stories to show they have
something in common and is managed by business
representatives in a project. No software development
activities are required to achieve themes. (ii) Initiative
is a composition of epics that drive toward a common
business goal which should not span more than one
year. Initiatives provide also the needed context to
help companies make decisions regarding the course
of direction. Software development activities are
required to achieve an initiative. (iii) Epic is a set of
related user stories that need no longer than a quarter
to be completed. To achieve epics, software
development activities are needed as well. (iv) User
Story is the lowest level of granularity that means
work to be completed within one to four weeks.
Similarly to epics and initiatives, user stories need
software development activities to be implemented.
Based on the aforementioned description, throughout
this paper, we treat agile initiatives as high level
business goals (e.g. Cloud), epics ̶ as sub-goals (e.g.
Kite) and user stories ̶ as user goals (e.g. Sea).
2.3 Modelling Requirements
Models have been used widely in software
development. They provide an abstract representation
of a particular complex problem to simplify the
ICSOFT 2023 - 18th International Conference on Software Technologies
670
process of understanding the problem (Muller et al,
2012; Pastor et al, 2022). In more detail, Muller et al.
(2012) and Pastor et al. (2022) explained modelling
as a simplification of a system to be built with an
intended goal in mind and an abstraction of a relevant
part while ignoring irrelevant aspects. Moreover,
Girvan and Paul (2017) reported the following
benefits of using models, namely: models provide (i)
an effective manner for discussion and collaboration,
and (ii) an effective medium for communications.
These benefits are in line with the agile way of
working where individuals’ interaction and customer
collaboration are highly valued (Agile Alliance,
2001). The RE literature, e.g. (Horkoff, 2019),
introduced many GORE frameworks where goals are
used to identify significant requirements and are, in
turn, modelled. Specifically, the i* framework (Yu,
1997) has been recognized as one of the most popular
to model goals (Horkoff, 2019). Despite its broad use,
the i* framework was experienced as not so easy to
learn which hampered its spreading outside of the
originating community (Dalpiaz et al, 2016). This
experience forced the GORE community to come up
with a response, which was in the form of the iStar
2.0 goal-based requirements modelling language
(Dalpiaz et al, 2016). Throughout this paper, we will
use iStar to refer the iStar 2.0 as described in (Dalpiaz
et al, 2016). iStar is designed to provide means to
model (i) different types of actors and their
boundaries, (ii) independent elements e.g. goals,
qualities, tasks and resources, and (iii) different types
of relationships (Dalpiaz et al, 2016). iStar has an
open-source online goal modelling tool
c
, which we
use throughout this paper as a documentation tool to
explain our proposed approach. Our choice for the
online tool of iStar was driven by the free of charge
availability of the tool. Of course other
documentation tools such as Microsoft Visio could be
used for the purpose as well.
3 THE AQRE APPROACH
3.1 Our Process to Construct the
AQRE Approach
For the purpose of this work, we followed a research
process grounded on the design science guidelines of
Wieringa (2014). We chose this approach as it is
suitable for developing artefacts (i.e. methods) to
solve problems in practical contexts. The design
science inspired process (Wieringa, 2014) starts with
c
https://www.cin.ufpe.br/~jhcp/pistar/
formulating an objective that is expected to be
achieved with the construction of the artefact (in our
case, the proposal for an approach to engineering QRs
in LSDA projects).
The overall objective of our proposed AQRE
approach, is to help agile teams to deal with the QRs
challenges reported in (Alsaqaf et al, 2019). For our
approach to work, it needs to be embedded into the
larger software development process of an
organization (Pfleeger and Atlee, 2009). In our
research context, this is the process of delivering a
software system in a LSDA project. In line with this,
we start on the premise that our proposed AQRE
approach should be executed before the start of the
first sprint to help create the initial version of the
product backlog of the project and then enable the
start of this first sprint. In a nutshell, our proposal,
AQRE, represents a workshop session where
participants discuss and elaborate software
development initiatives into epics and user stories by
using a goal-based requirements modelling language
such as iStar. The AQRE approach consists of one
mandatory role and two steps which are described in
the next subsections.
3.2 The AQRE Role
Drawing on the industrial practice of McKinsey
(Bucy et al, 2017) and other large companies
(Sutherland et al, 2022), AQRE introduces the
organizational concept of Initiative Owner (IO).
While this term has been used in industrial experience
reports (Bucy et al, 2017; Sutherland et al, 2022), to
the best of our knowledge the term IO has not been
elaborated in sufficient depth in scientific studies. For
example, although Bucy et al. (2017) refer to this
term, these authors did not provide a clear description
of what they mean with it. Besides, Bucy et al. (2017)
use the term IO in the context of organization’s
transition in general, and not related to agile in
specific. Furthermore, Sutherland et al. (2022)
describe the role of IO as synonym to the role of
Product Owner (PO). To avoid any confusion, in
AQRE we use the term IO to refer to the one who is
responsible for achieving the initiative (e.g. the high-
level goal(s)) of the organization and coordinating the
activities needed to implement the initiative in
question. This initiative will be decomposed into
epic(s) and user stories and assigned to one or more
agile teams to be implemented. In case of multiple
agile teams, each would have its own PO. Each PO is
then responsible for coordinating the work of his/her
Agile Quality Requirements Elaboration: A Proposal and Evaluation
671
own team based on the customer’s values his/her
team wants or deliver.
3.3 The Two AQRE Steps
Our proposed approach includes preparation of the
AQRE workshop and its execution. Below, we
describe the steps of AQRE in terms of activities that
the AQRE participants would go through.
3.3.1 Preparation
The IO begins the initiative by providing a short
description. The IO determines further who will be
invited to the AQRE workshop (e.g. step two below)
to discuss and elaborate the initiative. We note that
the workshop participants should be chosen based on
the needed knowledge and not based on their role in
the organization. The types of needed knowledge are:
(i) domain knowledge, (ii) QRs knowledge, (iii)
enterprise architecture knowledge, (iv) infrastructure
and maintenance knowledge. Based on the nature of
the initiative, other particular types of knowledge
could be needed (e.g. security, regulations and
compliance, usability). In that case, the IO invites the
people with that particular knowledge as well.
Besides, the IO also ensures that the workshop’s
participants have sufficient knowledge of goals
modelling techniques. If a participant has no prior
exposure to these techniques, then the IO organizes a
training session as part of the preparation step to
educate the participants on goal documentation by
means of models.
3.3.2 Workshop
The IO starts the workshop by giving background
information such as organization’s vision and goals,
the initiative’s description and how it fits within the
organization’s vision and goals. Hereafter, the
participants start decomposing the initiative using the
AND/OR goals modelling and decomposition
technique described in (Pohl, 2010) and a modelling
tool to visualize the models. Prior to the workshop,
the IO assured that the participants understand the
goals decomposition technique (see step one on the
previous page). The objective of this process is to
break down the initiative into smaller goals (e.g.
matching the possible epics and user stories) based on
the expected value of the customer (e.g. the who),
defining possible alternative (sub)goals or solution’s
directions, defining QRs associated with the
identified (sub)goals, identifying and resolving
conflicts between (sub)goals in general, and QRs in
specific, and defining and agreeing upon the scope of
the initiative. The workshop could take hour(s) or
day(s), depending on how complex the initiative is.
Further, the workshop resulted in a model of
decomposed (sub)goals similar to the one depicted in
Figure 1. The IO is responsible for creating the
resulting model. However, she could ask a participant
to draw the model or she can draw it all by herself.
Since our study focuses on the mitigation of the
reported QRs challenges in (Alsaqaf et al, 2019), in
the next subsection we elaborate further on that
subject. We note that decomposing an initiative may
result in subgoals (e.g. epics) that are big and vague
to be technically implemented within the next sprint.
In that case the AQRE workshop could be conducted
again to elaborate each subgoal whenever it is needed.
3.3.3 QRs Elaboration
During the process of breaking down (sub)goals, the
participants have to identify those QRs associated
with the identified (sub)goals. The identified QRs can
be further broken down into other related QRs. For
example, a security quality attribute could be
decomposed into e.g. Confidentiality, Integrity and
Authenticity. Performance, for example, could be
broken down into e.g. Capacity and Resource
Utilization. To guide this process optimally we advise
the participants to use a QRs framework of their
choice, e.g. the ISO 25010 quality standards
(ISO/IEC, 2011), the NFR framework (Chung et al,
2000), or the Sustainable Catalogue (Albuquerque et
al, 2021). After identifying and elaborating the QRs,
the participants have to provide those QRs with
enough details in order to enable the development
teams to make the right design decisions (concerning
these QRs) at the right time. The details should at
least include: (i) the estimated impact of (fully)
having or (partially) missing the QR on the related
(sub)goal, and (ii) broad specification of the QRs.
This means that the participants should provide the
development team with just enough QR’s boundaries
to be able to implement and test the right QRs
correctly (Lauesen, 2002). For example, if the
performance of collecting user data after logging in
the system is an important QR, we can then specify
what is an unacceptable performance like “user data
should be collected and shown on screen in no longer
than 7 seconds”. Instead of “user data should be
collected and shown on screen in 5 seconds”. The last
option will limit the decisions which the development
teams can make, while the first option will give the
development teams space to make their consideration,
especially if there are conflicting performance and
ICSOFT 2023 - 18th International Conference on Software Technologies
672
usability requirements, e.g. the amount of data that
should be shown after logging in the system.
4 OUR FIRST EVALUATION OF
THE AQRE APPROACH
As stated earlier, the AQRE approach was developed
to help mitigate QRs challenges experienced by agile
practitioners (Alsaqaf et al, 2019). Therefore, to
evaluate our proposed approach, we chose to design
and conduct a focus group (FG) study (Krueger and
Casey, 2014) to collect and analyze the opinions and
perceptions of experts regarding the AQRE approach.
We chose the FG-based evaluation strategy as
recommended in (Wieringa, 2014; Lauesen, 2002).
We chose this strategy, because (i) design science
relies on feedback sessions from practitioners in the
field where the proposed artefact is supposed to be
used, and (ii) the FG approach lends itself naturally
for the context of our evaluation efforts.
Our evaluation was set out to address two
evaluation research questions (RQs): RQ1. Does the
goal model created by means of the AQRE workshop
fit the project initiation context, according to the
practitioners working in this context? and RQ2. In
what way is this model useful, according to the
practitioners in the FG study?
As preparation for the FG, the first author
developed a research protocol which was discussed
and improved in collaboration with the second author.
The final FG-based study consisted of the following
two steps:
(i) Conducting a AQRE workshop within a project
organization to provide a goal decomposition model
of a real-world initiative,
(ii) Conducting the FG session itself with
participants from the same organization (8
participants) to collect their opinions of the resulted
goal decomposition model and their perceptions of
the QRs model obtained by using AQRE.
Following FG research methodologists (Krueger
and Casey, 2014; Morgan, 1996) we think that
collecting experts’ opinions based on an initiative
from their own domain would provide us with a rich
and meaningful feedback which would help us to
improve the AQRE approach and use its improved
version in further evaluation studies. We note that as
this is our first evaluation, eight practitioners form a
big enough focus group for the purpose of very early
feedback-giving. Also, as per FG methodologists
(Krueger and Casey, 2014; Morgan, 1996), the
overall idea is to generate evaluative feedback from a
variety of practitioners’ perspectives and not strive
for statistical analysis of the responses or for
achieving consensus among the participants. To
recruit FG participants, the first author used his
professional network to first find a suitable
organization which was willing to provide us with a
real-world initiative and then to host the AQRE
workshop and also to spend time in our FG-study
afterwards. The AQRE workshop as well as the FG
session were conducted in Dutch and they were
performed at the IT department of a big government
organization in the Netherlands.
The IT department has more than 300 employees
and provides European business entities with the
needed tools and software to communicate business-
related issues with the Dutch government agencies as
well as with other European business entities. The IT
department of the host organization provided us with
an initiative that had to be elaborated. The responsible
employee for the initiative is a senior business analyst
with more than eight years of professional experience
in this job. In AQRE terminology, this practitioner
will be the IO.
4.1 Using the AQRE Workshop to
Create a Goal Model
For the first step of our evaluation, the first author and
the IO hold an AQRE workshop to elaborate the
initiative. The PO of the potential agile team(s) that
would implement the initiative has also attended the
AQRE workshop. The participating PO has about
fourteen years of experience as information manager
and PO.
The AQRE workshop lasted about two and half
hours and resulted in a goal decomposition model
created by iStar (Dalpiaz et al, 2016) (Figure 1). In
the first 30 minutes of the workshop, the iStar tool
was introduced to the IO and the PO. Thereafter the
actual goal modeling started. The IO as well as the PO
were already familiar with goal decomposition
technique.
4.2 Our FG: Data Collection and
Analysis Process
As already mentioned, we chose for a FG-based
evaluation strategy (Wieringa, 2014) to assess the
fitness and the usefulness of the model resulting out
of the application of the AQRE approach. Therefore,
we set up a FG with experts to collect feedback on the
deployment of the AQRE approach. As the AQRE is
to be deployed on high-level goals, we considered for
Agile Quality Requirements Elaboration: A Proposal and Evaluation
673
Figure 1: Goal decomposition by iStar.
inclusion practitioners possessing knowledge in one
or more of the following: domain expertise,
architecture knowledge and requirements knowledge.
We approached relevant contacts of the host
organization with those types of knowledges to our
FG-study. Eventually eight experts accepted the
invitation and attended the FG. We note that the
participant of the workshop in Sec. 4.1 were not part
of the FG, i.e. the FG-based evaluation includes
completely different people. Table 2 presents the
profiles of or FG participants. The first column
describes their roles in the IT organization and the
second describes the number of participants for each
role. The average years of IT experience and agile
experience of the participants for each role are
described in the third and the fourth columns,
respectively. We note (as per Table 2) that the
participants have gathered more IT experience in
‘heavyweight’ (or traditional) way of working than in
Agile.
Table 2: Roles of participants.
Role
Number of
participants
Average
years of
experience
in IT
Average
years of
experience in
Agile
PO 2 25 6
Architect 1 26 3
Domain
expert
2 22 1
Business
analyst
3 23 4
The FG was guided by the first author, was 45
minutes long and was video-recorded. At the
beginning of the FG, the QRs challenges reported in
(Alsaqaf et al, 2019) were explained in a nutshell. The
first author then explained the initiative used in the
AQRE workshop and presented the resulting model
as depicted in Figure 1. Thereafter, the participants
were given the time to express their opinions and
thoughts. To trigger the conversation, the first author
asked the participants what they think of the model on
Figure 1 in regard to whether it fits the context,
whether the participants think it is useful and in what
way they perceive it as useful. The video-recording
was analysed immediately after the FG meeting. The
process of data analysis consisted of two phases and
started by having the first two authors watching the
recorded video of the FG and analysing the content
(Phase 1). The grounded theory analysis techniques
as described in (Charmaz, 2006) were followed to
extract knowledge from the contents. Contrasted
interpretations made by authors during the data
analysing process were resolved by conducting an
argumentative discussion (Hitchcock, 2002) to reach
a shared interpretation. In the next phase (Phase 2),
the concept-mapping practices (Conklin, 2003) and
argumentative discussion (Hitchcock, 2002) were
applied to identify which QRs challenges reported in
(Alsaqaf et al, 2019) could be mitigated by using the
AQRE approach. In Section 4.3, Table 3 summarizes
the results of Phase 2. The first column of the table
shows the reported categories and their related
challenges, while the second indicates how the AQRE
approach could mitigate the related challenge in the
ICSOFT 2023 - 18th International Conference on Software Technologies
674
first column. A dash “-” in the second column means
that no remedy has been identified by the
practitioners, which also means that the AQRE
approach has no answer yet for the related challenge.
4.3 Findings
In this section, we summarize our findings based on
the analysis process we described in the previous
section. We first report the extracted knowledge
resulting from Phase 1. Then, Table 3 summarizes the
result of Phase 2.
By analysing the shared feedback by the
workshop participants and the FG participants, we
derived three themes: requirements visualization,
requirements documentation and requirements
analysis. These are described below.
4.3.1 Theme: Requirements Visualization
The participants indicated that they experienced the
AQRE approach as a way to visualize requirements
in general, and QRs in particular. They felt that the
AQRE approach could establish a visualized
discussion to elaborate goals into requirements be it
functional requirements, or QRs. A participant of the
FG, expressed her opinion as “This visualization tool
will help conduct the right discussion with the right
persons to elaborate the requirements, however, we
need somehow to document the outcomes of the
discussion”. Another FG-participant agreed with the
need of visualizing requirements and suggested the
use of Rich Picture (Berg and Pooley, 2013) instead
of a goal-based requirements modelling language
tool. While AQRE does not recommend one
visualization tool above another, the choice of a
visualization tool should not result in multiple of
vague interpretations of the requirements, which is
the case in Rich Picture (Berg and Pooley, 2013).
4.3.2 Theme: Requirements Documentation
A heavy documented requirements process was
recognized by the participants as a source of
annoyance. A participant reacted with: “I will be
happy if this approach (e.g. AQRE) results in less
documentation. Our process of elaborating
requirements from initiative to epics to user stories
results usually in long documents. The same
participant expressed her concerns and emphasized
the feeling of losing control, which people who are
used to heavy documented requirements process
might develop by implementing the AQRE approach.
Another participant elaborated further: We have ever
started with an initiative of two pages, nowadays we
write initiatives of more than 10 pages. Specially the
requirements for publishing tenders need to be very
concrete, detailed and SMART, something I do not
see happening when using the AQRE approach”.
However, the AQRE approach does not advocate “not
documenting the requirements at all”. It provides only
instruments to elaborate the requirements in a face-to-
face conversational way and documenting only what
is needed while leaving enough space for decisions to
be made at the right time by the right people.
4.3.3 Theme: Requirements Analysis
A FG-participant has recognized the need for
delaying the decisions to the right moment and giving
software developers the space to make the right
decision when enough detailed information is
emerged I see the AQRE approach as a discussion
instrument to elaborate and capture the requirements
(e.g. functional and quality requirements) while
giving the chance for trade-offs to be made at the
right time”. Another participant proposed to add
priorities to the defined (sub)goals to guide later the
development process. The same participant proposed
as well to assign the (sub)goals to individual PO’s to
be able to elaborate and coordinate the requirements
further within and between individual teams.
4.3.4 Mapping the Practitioners Perceptions
of AQRE Against the Challenges
Once the themes crystalized, our next step was to map
the practitioners’ feedback against the QRs
challenges from Table 1. (For interested readers, we
suggest them refer to ref. (Alsaqaf et al, 2019) which
explains in detail how the QRs challenges have been
identified based on case study research.) The
mapping is presented in Table 3. It indicates that
AQRE addresses 12 out of the 15 QRs challenges.
The three QRs challenges that are not addressed are
Uneven teams maturity, Suboptimal inter-team
organization and Lack of cost-effective real
integration test. We should note, however, that these
three challenges are organizational in nature and
therefore might demand an approach that is grounded
more on social and organizational behavior and
respective theoretical models than on RE concepts
(such as GORE).
Agile Quality Requirements Elaboration: A Proposal and Evaluation
675
Table 3: QRs challenges could be mitigated with AQRE.
QRs challenges
(Alsaqaf et al, 2019)
How AQRE helps
Late detection of QRs
infeasibility
AQRE workshop elaborates
QRs related to identified
(sub) goals
Hidden assumptions in
inter- team collaboration
Early Face-Face
communications and
discussion will point
participants to the same
direction
Uneven teams maturit
y
-
Suboptimal inter-team
or
g
anization
-
Inadequate QRs test
s
p
ecification
Elaborating QRs and
making decision just in time
Lack of cost-effective
real inte
g
ration test
-
Lengthy QRs
acce
p
tance checklist
Elaborating QRs and
making decision just in time
Sporadic adherence to
q
ualit
y
g
uide-lines
Elaborating QRs and
making decision just in time
Overlooking sources of
QRs
AQRE workshop involves
essential knowledge
Lack of QRs visibility
AQRE approach visualize
requirements
Ambiguous QRs
communication
p
rocess
AQRE workshop enforce
face-to-face communication
Unclear conceptual
definition of QRs
AQRE approach explicitly
mentions QRs
Confusion about QRs
s
ecification a
roaches
Elaborating QRs and
making decision just in time
Unmanaged architecture
changes
AQRE approach involve
architectural knowledge
early in the process
Misunderstanding the
architecture drivers
AQRE approach involve
architectural knowledge
early in the process
5 DISCUSSION
The results of this research indicates the complexities
of confronting the QRs challenges in LSDA projects
by using one only method. Although our proposal
seems to address 12 out of the 15 QRs challenges
identified in (Alsaqaf et al, 2019), there are three
more that remain to be dealt with. We consider this as
a clear signal that it might not be realistic to expect
that a single method might exist to deal with all QRs
challenges at once. As the three challenges are
organizational in nature, we think that these could be
most effectively delt with presumably by introducing
some project management actions that target the
social behavior of project team members and their
level of professional maturity. In line with this, it
seems worthwhile rethinking our AQRE approach in
terms of possible extensions regarding what it could
be supplemented with from organizational
perspective, in order to address the full spectrum of
QRs challenges.
Furthermore, it is important to note that in our
case, educating the two practitioners, the IO and the
PO, took about 30 minutes. We think that this is not
too much, knowing their solid professional
experience (i.e. the IO had 8 years while the PO had
14 years of experience in their fields, though not all
years were in agile projects). This indicates the
possible cost-effectiveness of AQRE in terms of
learnability and ease of adoption. Of course, as this is
our very first evaluation, more evaluations through
case studies are needed in order to have an accurate
understanding of the cost-effectiveness of our
approach. This is our next step. At the time of writing
this paper, the authors are already in a conversation
with another project organization willing to host a
case study.
One could argue that using AQRE to elaborate
QRs from high-level goals before starting the first
sprint looks more waterfall since it contrasts with the
agile “just-in-time” identification of the
requirements. We think that QRs as they impact the
software architecture should be identified as soon as
possible while leaving enough space to specify them
just-in-time. That is precisely what the AQRE does, it
identifies the QRs as early as customer’s intentions
have been decomposed. The exact specification of
those QRs are left to be just-in-time identified.
Did our practitioners feel that AQRE made their
agile project less agile and more heavyweight? The
FG results did not suggest this. One possible
explanation could be that the practitioners had long
experience in “heavyweight” projects in a large
government organization that values documentation.
As their organization was experimenting with agile
and figuring out how to achieve a good balance
between agility, flexibility and documentation focus,
the participants might not have perceived our goal
model as a source of additional bureaucratic and
documentation efforts. In their views, the benefits of
AQRE clearly outweighed the 2 hours invested time
ICSOFT 2023 - 18th International Conference on Software Technologies
676
for creating the AQRE model (Figure 1). Of course,
we might have obtained other results if the AQRE
approach would have been tried out in an agile
organization where a large-scale framework had been
adopted for many years and a history of many agile
projects existed.
Based on the feedback from the FG participants,
we recognized an interesting contradiction: on one
side, there was willingness for less documentation
while on the other side, there was the fear of losing
control by documenting less. It seems that the FG
participants associate the sense of control over a
project with the presence of documentation. More
documentation means demonstrating control and vice
versa. We think that the culture of the organization
may play a significant role in promoting more or less
documentation. We note that the evaluation of the
AQRE approach occurred in a government
organization and the practitioners involved in the FG
had far more experiences in traditional software
development than in agile (see Table 2).
6 LIMITATIONS
This study had some limitations. First, our newly
proposed approach addresses 12 of the 15 QRs
challenged. While this indicates important progress,
in order to cope with the remaining three challenges,
one needs to supplement it with some management
actions. We consider this an important aspect and it is
our intention to look into those three QRs challenges
in our follow-up research.
Second, there are some validity threats (Wieringa,
2014) concerning our very first evaluation. Because
all practitioners involved in our FG study operate in a
large-scale government organization in the
Netherlands, the themes that came out of the FG
might differ from those that might possibly have
come out if the demonstration of the AQRE approach
and its evaluation had happened in a private agile
company, for example one that is incorporated in the
past decade and has always been agile (i.e. never
worked with heavy-weight approaches).
Next, the FG moderator (i.e. the first author) had
experience in the host organization, so some bias
could be passed to the data analysis process.
However, we employed disciplined steps and
systematic techniques, e.g. coding, reflexivity,
memoing, and discussion of the analytic process with
the other researchers, all of which helped us reduce
bias.
Finally, we also would like to mention that while
we cannot claim universal generalizability of our very
first evaluation results, it could possibly be realistic
to expect that if we use the AQRE approach in similar
type but physically different organizations, we might
find some similar observations. This is due to the
similarities in the contexts of our organization with
others in the same sector (Ghaisas et al, 2013). For
example, other Dutch large-scale government IT
departments who embark on LSDA and experience
similar challenges like those described in Table 1.
7 CONCLUSION AND
IMPLICATIONS
In this paper we proposed the AQRE approach for
elaborating requirements in general and QRs in
specific from high-level goal(s) in LSDA projects.
The proposed approach consists of one role (e.g. IO)
and two steps (e.g. Workshops preparation,
Workshop execution). Adhering to the agile
principles is heavily taken into consideration since
our proposed approach leans on collaborations and
individuals interactions. The primary objective of
AQRE is to identify a remedy to the QRs challenges
reported in (Alsaqaf et al, 2019). The fitness and
usefulness of the proposed approach is further
evaluated using a FG as recommended in (Wieringa,
2014). This very first evaluation of the AQRE
indicated that the proposed approach could mitigate
12 of the 15 QRs challenges reported in (Alsaqaf et
al, 2019). The other three challenges are
organizational in nature (e.g. Uneven teams maturity,
Suboptimal inter-team organization, Lack of cost-
effective real integration test) and may need
additional organizational practices to deal with them.
Our highest priority is to evaluate the AQRE
approach further in other real-world projects in order
to improve its effectiveness and usefulness beyond
the context in which we did our very first FG
evaluation. To this end, we plan two FG studies in
two different contexts: one in a consulting company
with professional consultants providing agile project
management services to agile project organizations in
a wide range of business sectors, and a second FG
study in a national professional association whose
members are requirements engineering practitioners,
some of which work primarily in agile contexts. This
evaluation effort would help us better assess the
applicability of AQRE and its generalizability across
contexts.
Agile Quality Requirements Elaboration: A Proposal and Evaluation
677
REFERENCES
Abheeshta, P., et al. (2018). “Benefits and Challenges of
Adopting the Scaled Agile Framework (SAFe):
Preliminary Results from a Multivocal Literature
Review”. PROFES(9459): 334–351.
Agile Alliance. (2001). “Manifesto for Agile Software
Development”. http://www.agilemanifesto.org.
Albuquerque, D., et al. (2021). “A Sustainability
Requirements Catalog for the Social and Technical
Dimensions”. ER(1): 381–394.
Alsaqaf, W., et al. (2017). “Quality requirements in large-
scale distributed agile projects - A systematic literature
review”. LNCS 10153: 219–234.
Alsaqaf, W., et al. (2019). “Quality Requirements
Challenges in the Context of large-scale distributed
agile: An Empirical Study.” Information and Software
Technology.
Berg, T., Pooley, R. (2013). “Rich Pictures: Collaborative
Communication Through Icons”. Systemic Practice and
Action Research 26(4):361–376.
Bucy, M., et al. (2017). “Keeping transformations on
target”. McKinsey & Company(March): 1–17.
[Online]. Available: https://www.mckinsey.com/ca
pabilities/rts/our-insights/keeping-transformations-on-
target.
Calefato, F., Ebert, C. (2019). “Agile Collaboration
Distributed Teams”. IEEE Software 36(1): 72–78.
Charmaz, K. (2006). “Constructing grounded theory: a
practical guide through qualitative analysis”. SGAE. 1
st
edition.
Chung, L., et al. (2000). “Non-Functional Requirements in
Software Engineering”. Springer. NY, 1
st
edition.
Cockburn, A. (2000). “Writing Effective Use Cases”.
Addison Wesley. 1
st
edition.
Conboy, K., Carroll, N. (2019). “Implementing Large-Scale
Agile Frameworks: Challenges and Recommendations.”
IEEE Software 36(March/April):1–9.
Conklin, J. (2003). “Dialog Mapping: Reflections on an
Industrial Strength Case Study”. Visualizing
argumentation:1–1.
Dalpiaz, F., et al. (2016). “iStar 2.0 Language Guide”.
[Online]. Available: https://sites.google.com/site/
istarlanguage.
Daneva, M., et al. (2007).“Exploiting a Goal-
Decomposition Technique to Prioritize Non-functional
Requirements”. 10th WER.
De Lucia, A., Qusef, A. (2010). “Requirements engineering
in agile software development,” Journal of Emerging
Technologies in Web Intelligence 2(3): 212–220.
Digital.ai Software, (2021). 15
th
State of Agile Report.
Digital.ai.
Domah, D., Mitropoulos, F, J. (2015) “The NERV
Methodology : A Lightweight Process for Addressing
Non-functional Requirements in Agile Software
Development”. Southeastcon: 1-7.
Farid, W, M., Mitropoulos, F, J. (2012). “Novel lightweight
engineering artifacts for modeling non-functional
requirements in agile processes,”. Southeastcon: 1-7.
Farid, W, M., Mitropoulos, F, J. (2013). “Visualization and
scheduling of non-functional requirements for agile
processes”. Southeastcon: 1-8.
Ghaisas, S., et al. (2013). “Generalizing by Similarity :
Lessons Learnt from Industrial Case Studies”. CESI:
37–42.
Girvan, L., Paul, D. (2017). “Agile and Business Analysis
Practical guidance for IT professionals”. Bcs Learning
& Development Limited. 1
st
edition.
Helmy, W., et al. (2012). “Requirements engineering
methodology in agile environment,” International
Journal of Computer Science Issues 9(5): 293–300.
Hitchcock, D. (2002). “The Practice of Argumentative
Discussion”. Argumentation(16): 287–298.
Horkoff, J., et al. (2019). “Goal-oriented requirements
engineering: an extended systematic mapping study”.
RE 24:133–160.
Inayat, I., et al. (2014). “A systematic literature review on
agile requirements engineering practices and
challenges”. Computers in Human Behavior
51(October): 915–929.
Inayat, I., et al. (2015). “Reflection on Agile Requirements
Engineering: Solutions Brought and Challenges
Posed”. XP Workshops.
ISO/IEC. (2011). “ISO/IEC 25010:2011 - Systems and
software Quality Requirements and Evaluation
(SQuaRE)”. [Online]. Available: http://www.iso.
org/iso/iso_catalogue/catalogue_tc/catalogue_detail.ht
m?csnumber=35733.
Jeon, S., et al. (2011). “Quality attribute driven agile
development”. SERA: 203–210.
Kopczyńska, S., Nawrocki, J. (2014). “Using non-
functional requirements templates for elicitation: A
case study”. RePa: 47-54.
Krueger, R, A., Casey, M, A. (2014). “Focus Groups: A
Practical Guide for Applied Research”. SAGE. 5
th
edition.
Kumar, M., et al. (2013). “A Hybrid Approach of
Requirement Engineering in Agile Software
Development”. ICMIRA: 515–519.
Lamsweerde, van, A., et al. (1998). “Managing Conflicts in
Goal-Driven Requirements Engineering”. IEEE
Transactions on Software Engineering 24(11): 908–
926.
Lauesen, S. (2002). “Software Requirements: Style and
Techniques”. Addison-Wesley. 1
st
edition.
Maiti, R, R., Mitropoulos, F, J. (2015). “Capturing ,
Eliciting , Predicting and Prioritizing ( CEPP ) Non-
Functional Requirements Metadata During the Early
Stages of Agile Software Development.”
SoutheastCon:1-8.
Morgan, D, L. (1996). “Focus Groups as Qualitative
Research” . SAGE. 2
nd
edition.
Muller, P., et al. (2012). “Modeling modeling modeling”.
Softw Syst Model 11(3): 347–359,
Noreika, K., Gudas, S. (2021). “Modelling the alignment
between agile application development and business
strategies”. BIR-WS 2991: 59–73.
ICSOFT 2023 - 18th International Conference on Software Technologies
678
Pastor, O., et al. (2022). “Teaching Modeling in the time of
Agile Development”. Computer (Long Beach Calif)
55(June): 73–76.
Pfleeger, S., Atlee, J. (2009). Software Engineering:
Theory and Practice”. (2009). Pearson. 4
th
edition.
Pohl, K. (2010). “Requirements Engineering
Fundamentals, Principles, and Techniques”. Springer
Berlin. Heidelberg, 1st edition.
Rahy, S., Baas, J, M. (2022). “Managing non-functional
requirements in agile software development,” IET
Software 16(1): 60–72.
Sherif, E., et al. (2022). “Managing non-functional
requirements in Agile Software Development”.
ICCSA: 205–216.
Shore, J., Shane, W. (2007). “The Art of Agile
Development”. O’Reilly Media. 1st edition.
Smart, J. (2018). “To Transform to Have Agility, Dont Do
a Capital A, Capital T Agile Transformation”. IEEE
Software 35(6): 56–60.
Sutherland, J., et al. (2022). “Embracing Agile: How to
master the process that’s transforming management”.
Harvard Business Review 94(5): 40–50.
Wieringa, R, J. (2014). “Design Science Methodology for
Information Systems and Software Engineering”.
Springer. Dordrecht, 1st edition.
Yu, E, S, K. (1997). “Towards Modelling and Reasoning
Support for Early-Phase Requirements Engineering,”.
ISREI: 226–235.
Agile Quality Requirements Elaboration: A Proposal and Evaluation
679