Evaluating A Novel Agile Requirements Engineering Method: A Case
Study
Tanel Tenso, Alex Norta and Irina Vorontsova
Department of Informatics, Tallinn University of Technology, Tallinn, Estonia
Keywords:
Visualization, Agile, Requirements Engineering, Modelling, Goals.
Abstract:
The use of agile methods during software development is a standard practice and user stories are an established
way of breaking complex system requirements into smaller subsets. However, user stories do not suffice for
understanding the bigger picture of system goals. While methods exists that try to solve this problem, they lack
visual tool support and are too heavy for smaller projects. This article fills the gap by evaluating a novel agile
agent-oriented modelling (AAOM) method for requirements engineering. The AAOM-method comprises a
visual approach to agile requirements engineering that links goal-model creation techniques taken from agent-
oriented modelling and connects goals intuitively to user stories. A case study based evaluation explores the
applicability of AAOM for requirements engineering in an agile software development process.
1 INTRODUCTION
Requirements engineering (RE) is an important
software-development activity and traditionally con-
sidered one of the first phases in software develop-
ment. RE is a process of formulating, documenting
and managing the requirements for software and com-
prises requirementsidentification, analysis, documen-
tation and validation (Hull et al., 2010). Since RE is
the first software-development phase, late detected er-
rors are very costly (Carlson and Matuzic, 2010) and
produce incorrect software that does not satisfy cus-
tomer needs.
Agile development is the most widely used
method (Version One, 2016) for developing software
systems that adhere to theagile manifesto (Beck et al.,
2001). Agile development methods are time-boxed,
iterative and incremental. Characteristic is also a
frequent delivery of usable software and collabora-
tion with customers. Additionally, the agile method
supports self-organizing cross-functional teamwork.
These factors play a role in the ability to quickly re-
spond to changes (Cao and Ramesh, 2008). For agile
affiliated RE, several variations exist that affect the
use of RE, e.g., Scrum (Schwaber, 2004), XP (Beck,
2000), Lean (Poppendieck and Poppendieck, 2007),
Kanban (Kniberg and Skarin, 2010). Common for
these agile variations are of a lacking intuitive align-
ment (Cao and Ramesh, 2008) between engineered
requirements and intuitive visual system development
support.
To address this gap we define a novel require-
ments engineering method, namely the agile agent-
oriented modelling (AAOM) method (Tenso and
Taveter, 2013). AAOM emerges through multiple ex-
periments in various projects varying from small scale
projects, for example adding release management ca-
pability for a task management system, to large scale
distributed projects like crisis simulation software de-
velopment. AAOM is derived from agent-oriented
modelling (AOM) (Sterling and Taveter, 2009), a
holistic method for analysing and designing socio-
technical systems consisting of humans and technical
components. More concretely, AAOM focuses on a
specific model type out of a larger model set that are
part of AOM, namely goal models comprising func-
tional goals, quality goals that are also knows as non-
functional goals, and affiliated roles these goals af-
fect.
To evaluate the applicability and usefulness of
AAOM, we choose a case study based research
methodology (Runeson et al., 2012; Yin, 2013). We
use as a running case a project for developing a
lost&found (l&f) mobile app. The business idea of
the mobile app is to reunite lost objects of any type
with their rightful owners. Instead of having to rely
on lost&found offices at police stations, airports, cin-
emas and so on, the l&f-app is a simple and quick mo-
bile solution to report findings by using smartphone
capabilities: a phone camera allows to instantly take
156
Tenso, T., Norta, A. and Vorontsova, I.
Evaluating A Novel Agile Requirements Engineering Method: A Case Study.
In Proceedings of the 11th International Conference on Evaluation of Novel Software Approaches to Software Engineering (ENASE 2016), pages 156-163
ISBN: 978-989-758-189-2
Copyright
c
2016 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
a picture of a found item while simultaneously the
photo-shot location is traced and stored via smart-
phone GPS. The app is also beneficial for people who
lose something: the app announces the loss and re-
ceives notifications when an item with a similar de-
scription and attributes is found and entered into the
database. The l&f-app either establishes the associ-
ation between object and owner because of provided
descriptions, or the object is tagged with a visual iden-
tification code a mobile phone can read. The l&f-
app also offers a so-called giveaway section for found
items that nobody claims.
The remainder of the paper is structured as fol-
lows. Earlier studies and theory of AAOM are in Sec-
tion 2. Research questions, case setup, data collec-
tion, analysis and validity procedures are described in
Section 3. In Section 4, we present results and find-
ings of the analysis of collected data. Finally, Section
5 concludes findings and gives open issues for future
work.
2 RELATED WORK
To provide a basis for the evaluation, in Section 2.1,
we give more details about the application of AAOM.
In Section 2.2, we discuss similar methods to AAOM
for comparison.
2.1 AAOM-Method Explanation
Face-to-face communication in agile over written
specifications facilitates embracing change and apply-
ing iterative cycles for RE. Figure 1 depicts (Tenso
and Taveter, 2013) how AAOM activities fit into the
agile development lifecycle. After an initial discov-
ery phase, the result is a set of further elaborated goal
models and affiliated user stories for which a prelim-
inary backlog is established. Next, in iterations, the
goal models and their user stories evolve by chang-
ing, eliminating, updating the latter. Once the main
development phase of a project ends, the system-
maintenance phase commences where the goal mod-
els and user stories are the foundation for exploring
how to add, modify, or remove features.
The AAOM method links goal models from AOM
(Sterling and Taveter, 2009) to user stories in accor-
dance with Figure 2. In the depicted model, goals are
shaped as parallelograms and quality goals shaped as
clouds represent functional- and non-functional sys-
tem requirements respectively. Goal models also con-
tain roles as sticky men with relationships between
roles and goals/quality goals. User stories are at-
tached in Figure 2 at the leave-level of sub-goals for
Figure 1: AAOM activities in agile development lifecycle.
The Notation For Goal Model
Reunite Item
Trustworty,
using
protocols
Lost Item
User
Enter lost item info Search lost item
User Stories
User Stories
As an app user I want to be able to see list of all lost
items from database
As an app user I want to be able to see list of found items
from database
As an app user when clicking on item in a list of found/
lost items I want to see all available item information on a
screen (picture + all the info)
Symbol Meaning
Goal
Quality Goal
Role
Relationship between
goals
Goal and quality goal
relationship
Figure 2: Example of a goal model with user stories at-
tached.
establishing and tracing a connection to a system’s top
goal.
User stories are simple artefacts for agile software de-
velopment and requirements documentation (Cohn,
2004). A user story is a written sentence or two
that describe functionality from a system’s user point
of view. There are several formats and concepts, in
which Cohn’s (Cohn, 2004) definition is as follows:
As a <role/type of user>, I want <goal/desire> so
that <benefit/reason> .
Examples:
As a user, I want to reserve a hotel room.
As a frequent flyer, I want to rebook a past
trip, so that I save time booking trips I take
often.
A user story must be small enough for implementa-
tion within one development iteration (Cohn, 2004).
This limits implementation work per user story, pro-
viding a fast feedback and verification of require-
ments for system development.
Similar to AAOM, other methods exist, that pro-
vide structure to agile requirements by organizing
Evaluating A Novel Agile Requirements Engineering Method: A Case Study
157
user stories. These methods are briefly discussed in
the next section.
2.2 Earlier Studies
Similar to AAOM, there are methods for organiz-
ing user stories into structures to mitigate visibility
problems, for example Cohn (Cohn, 2004) suggests
Epics, that are bigger user stories grouping smaller
ones. Epics covering different levels of abstrac-
tion are also used in Scaled Agile Framework(SAFe)
(Leffingwell, 2016). Large-Scale Scrum (LeSS) (Lar-
man and Vodde, 2008) concentrates on splitting re-
quirements into small Product Backlog Items, usu-
ally user stories. Disciplined Agile Delivery (DAD)
(Ambler and Lines, 2012) sums up many agile prac-
tices and introduces the term Portfolio management
that is requirements management in a hierarchical list
of work items. Scrum of Scrums (Sutherland et al.,
2007) includes team level planning and requirements
tracking between teams. The lean approach to ag-
ile requirements (Leffingwell, 2010; Leffingwell and
Aalto, 2009) divides requirements according to de-
tails onto team-, program- and portfolio level. De-
spite having several positive impacts and influenc-
ing AAOM theory aspects, these approaches are not
truly visual and, according to our experience not gras-
pable without special training by non IT stakehold-
ers. Furthermore, these methods are meant for enter-
prise scale usage and are too heavyweight for smaller
projects where it is important to establish a conversa-
tion with clients and align everybody to the same set
of goals.
On the other hand, goal modelling techniques ex-
ist exactly for depicting system goals in a visual way.
Goal-based requirements engineering is well estab-
lished, for example Hull et al. (Hull et al., 2010) sug-
gest representing use scenarios as a sequence of goals.
Lamsweerde approaches goal modelling from a for-
mal point of view, providing a mathematical proof and
meta model for goal based requirements engineer-
ing. His effors culminate in (Van Lamsweerde et al.,
2009), covering high-level system analysis with goal
techniques. One prominent method that includes goal
modelling is i* (Yu, 2009) that provides both tooling
and principles for the problem domain. Finally, an
example of more social goal-modelling techniques is
described in agent-oriented modelling (AOM) (Ster-
ling and Taveter, 2009) theory that is the basis for
AAOM. However, the aforementioned approaches
are too heavyweight to be included into short ag-
ile development feedback cycles and are meant for
model driven development. While AOM provides
the most lightweight goal modelling technique for re-
lating roles, functional- and non-functional require-
ments, its other models focus on the agent paradigm
and are thus too specific for wider system-design use.
Connecting agile user stories with goal models is a
novel approach introduced by AAOM.
3 CASE STUDY DESIGN
We choose a case study research method to conduct
the evaluation of the AAOM-method. First we de-
fine research questions in Section 3.1 for guiding the
AAOM evaluation, followed by a justification of case-
and subject selection in Section 3.2. Data sources of
evidence are discussed in Section 3.3, followed by an
analysis procedure in Section 3.4. A validity discus-
sion finalizes the case study design in Section 3.5.
3.1 Research Questions
Based on previous experiments with the AAOM
method and feedback gathered during them, the fol-
lowing main research question is devised:
How does AAOM help to improve software engineer-
ing requirements engineering activities?
This question can be refined in various sub-questions:
RQ1: What are the benefits of using AAOM from a
user perspective?
RQ2: What effect has project setup and tooling on
AAOM usage?
RQ3: What aspects of the AAOM-method usage
need further refinement?
Aforementioned questions establish the basis for se-
lecting appropriate case, data sources and analysis
methods.
3.2 Case Selection
The l&f-app development project follows an agile
software development cycle, specific scrum tech-
niques such as planning with user stories, back-
log management and iterative development are used.
Three development iterations take place in which
work is visualized on a task board and meetings with
clients take place at the end of iterations.
For the l&f-project research, we set up as a single
case study with holistic design (Runeson et al., 2012).
The unit of analysis and the case is an AAOM RE
method application for iterative requirements gather-
ing. Four people participate in this project carrying
the labels role and experience. Role helps us to evalu-
ate the AAOM usage by different team members, ex-
perience provides reliability for gathered information.
Three participants fill the role of client who order the
ENASE 2016 - 11th International Conference on Evaluation of Novel Software Approaches to Software Engineering
158
l&f-app development, whereas one of them has a high
level of experience as an ICT expert while the other
two are from different domains. One person acts as
an analyst and developer, being a beginner in the first
role and highly experienced in the second. A high
level of experience means a person acts at least for
two years in a specific role.
The research team consists of three researchers
working in co-operation, providing peer-review to
each other. The procedures to observe a case include
taking part in all meetings between case subjects that
are modelling, demo and retrospective sessions while
video recoding them all. We act as silent participants
taking notes of the AAOM usage throughout differ-
ent meetings. Based on the research questions and
meeting notes, we devise interview questions and ses-
sions for gathering qualitative data. The analysis of
the gathered data by researchers provides answers to
the research questions.
3.3 Data Sources
Interviews are the most valuable data source for
the running case study. Based on recommendations
from (Runeson et al., 2012), interview planning com-
mences with selecting interviewees. There are four
people and three roles in the running case as men-
tioned in Section 3.2 and all participants are inter-
viewed. Since there is one person in both roles of
analyst and developer, different questions are posed
to her addressing both roles. Next follows the plan-
ning of interview sets. Ideally, interviews take place
multiple times, for example after elaborating a first
branch of requirements, after every development iter-
ation and at the end of the project. As a limitation,
for this l&f-app project, we conduct only one set of
interviews after completing the planning session and
three development iterations.
For conducting interviews, we choose a semi-
structured format (Runeson et al., 2012). The re-
search questions stated in Section 3.1 and the meet-
ing notes guide the preparation of interview questions.
Different sets of questions target each respective role
without offering predefined answers. Thus, intervie-
wees can not answer, e.g., ”yes” or ”no”, instead they
must express their own opinions. Questionnaires for
client, analyst and developer roles are depicted in a
longer version of this article
1
.
The interviews adhere to a time-glass model
(Runeson et al., 2012) so that an interview begins
with broad questions first and continues with more
specific questions. At the end of an interview, again
broad questions are presented. Thus, four interviews
1
http://digi.lib.ttu.ee/i/?4064
include three with clients and one combined for the
developer/analyst. The interview structure is similar
in all cases and the interviewees are informed about
the interview structure during the process.
Each interview session lasts roughly one and a
half hours and starts with an introduction, followed
by role specific questions. The interviews are audio
recorded into MP4 files for subsequent post-interview
activities and analysis. In case an interviewee re-
sponds to a question briefly, we ask additional ques-
tions on the same topic to gather more insight. At
the end of an interview, a participant learns that ev-
ery interview is transcribed and sent for verifying the
captured ideas are correct.
We also gather work artefacts, such as goal mod-
els, user stories and source code. Since a dedicated
development toolkit for AAOM does not exist yet, we
employ provisionally a set of freely available tools to
host our case. The first tool, Draw.io
2
, is an online
diagramming tool to draw goal models for the l&f-
app. Another used tool is Trello
3
, a collaboration tool
to organize project tasks on boards. Trello visualizes
tasks in the form of user stories similar to post-it notes
in status columns to observe the progress during soft-
ware development, e.g., to do, pending, in progress,
completed, and so on.
The generated goal modelsand user stories are rel-
evant for investigating the evolution during the project
while the history of changes is recorded. Project
source code for the l&f-app in a version control sys-
tem allows to observe lines of code (LOC), changes
in LOC, time between LOC changes and links to
user stories. After evaluating these quantitative data
sources, the main finding is that there is no existing
body of knowledge for analysing them as needed for
the AAOM method usage validation. Thus, we use
these anecdotal data sources only to subjectively eval-
uate some statements received from interviewees.
As our main data source is interviews, we focus
in the next section on interview analysis. The latter
yields answers to the research questions.
3.4 Analysis Procedure
Interview analysing requires first transcribing and
then coding (Salda˜na, 2015) the results and for both
we use the tool NVivo
4
. We transcribe the interviews
and for corrections and clarifications, the interviewees
review the transcripts.
To code the interviews, we determine first a list
of so-called a priori codes (Salda˜na, 2015) that are
2
https://www.draw.io
3
https://trello.com
4
http://www.qsrinternational.com/
Evaluating A Novel Agile Requirements Engineering Method: A Case Study
159
deduced from the research questions. Upon labelling
interviews with predefined labels, we detect sentences
and blocks of text that do not fit into existing codes.
The contents of these exceptions lead to the creation
of new so-called called grounded codes (Salda˜na,
2015).
To summarize coding, first themes are introduced
(Salda˜na, 2015) for grouping codes. This axial coding
(Runeson et al., 2012) identifies connections between
the themes and codes. Themes are also divided into
two sets, either based on relations to research ques-
tions (Section 3.1), or for not having any relation.
Additionally, we perform theoretical coding for
devising a set of attributes to codes for evaluating the
latter during analysis. The chosen attributes attached
to codes are polarity and type. Polarities show the
opinion of the interviewee and has three possible val-
ues, namely positive, neutral or negative. Types show
if a code and thus, the interviewee references an ex-
isting situation, or a need for change/addition. Type
has two values, namely statement or suggestion.
Polarity- and type values provide combinations
that describe whether a code proves, contradicts, or
has no relation to research. The first combination of
six, negative suggestion, indicates a need to change
an existing part of case under investigation. Second,
positive suggestion is an additional idea, or improve-
ment put forward by the interviewee. Next, neutral
suggestion is not related to the AAOM method evalua-
tion, negative statement denotes a flaw in the AAOM-
method while the interviewee has no improvement
suggestion. The final two combinations are positive
statement for indicating a participant’s satisfaction
with AAOM and neutral statement for representing
an opinion, or a statement of affairs not related to the
research question.
We use a simple formula to evaluate which codes
have more value for analysis. Three components play
a role in determining code validity, namely first ref-
erences that show how many times a code is men-
tioned in interviews. The second element is sources
denoting how many different interviewees mention
one code. Finally, role experience expresses an in-
terviewee’s experience in a role. Each component has
numeric values and a higher value increasing code va-
lidity. The formula to sort codes based on mentioned
components is as follows:
codevalue = (references sources) + experience
3.5 Validity Procedure
To asses the validity of our research, we use criteria
proposed by A. K. Shenton (Shenton, 2004), namely
credibility, transferability, dependability and con-
firmability. To increase credibility, we employ the fol-
lowing strategies:
We use a well established body of knowledge of
conducting case studies, mainly Runeson et al
(Runeson et al., 2012). To set up and receive
useful data from interviews, we use guidelines
by Robson (Robson, 2002) and for best-practices
to analyse gathered data, we consider Saldana
(Salda˜na, 2015).
We conduct interviews with all participants in the
project covering all different roles. This is consid-
ered as a form of triangulation since we capture
viewpoints of all informants (Van Maanen, 1979).
Unfortunately, we are not able to triangulate via
other types of data sources, since there is a lack of
methods for analysing goal models.
Before starting the l&f-app development, re-
searchers study AAOM as the unit of analysis and
also the context where it is applied, the field of
finding and losing assets.
To help ensure honesty in subjects, we inform
them that the data use is anonymous and that
their voice is recorded. All interviewees agree
with recording. The interviewees review the tran-
scribed documents to assure a valid transferral of
ideas.
Three researchers participate in the case study
and provide constant peer reviewing to each other.
With that setup and constant debriefing among
each other, the researchers’ vision is wider than
working alone.
The researchers participating in the case study
have relevant related backgrounds. Thus, credi-
bility is assured by the extensive research experi-
ence.
The second criteria, transferability, can not be
demonstrated since the case covers only a specific
project and a specific set of individuals (Shenton,
2004). Still, the reporting context of this case study is
helpful for other researchers to compare their results
with ours (Shenton, 2004). To test transferability, we
conduct another case study with a similar setup to ver-
ify whether the same results reoccur. One more factor
assuring transferability is the fact that agile teams are
limited by definition to a size of 3 to 9 (Cockburn,
2006).
To address dependability, Shenton et al. (Shenton,
2004) recommend a case study report that includes
sections devoted to:
the research design and its implementation
the operational detail of data gathering
ENASE 2016 - 11th International Conference on Evaluation of Novel Software Approaches to Software Engineering
160
reflective appraisal of the project
In this article, the research design is described in Sec-
tion 3 and data gathering details are depicted in Sec-
tion 3.3. We do not evaluate the effectiveness of pro-
cesses undertaken in the current case study. This will
be a evaluated by a future case study.
According to (Jensen, 2008), confirmability is an
accurate means through which to verify the two basic
goals of qualitative research:
to understand a phenomenon from the perspective
of the research participants and
to understand the meanings people give to their
experiences
This research contains a threat to researcher bias be-
cause one researcher is the inventor of AAOM, a
method under investigation. We are aware of this
threat and avoid it with the same means as for credi-
bility - by providing detailed descriptions. According
to (Shenton, 2004), in a qualitative study, researchers
biases are inevitable. We also acknowledge the lim-
itation of one person acting in two roles possibly re-
moving one friction point between analyst and devel-
opers. This could have been mitigated by assigning
another person to that role for which we lack a suffi-
cient project budget.
4 RESULTS
The main goal of the analysis is to understand whether
theories about the AAOM method are valid by finding
answers to questions specified in Section 3.1. We also
take into account any non-expected data found during
the interviews. All themes and codes discussed fur-
ther on, are presented in longer version of this arti-
cle
5
.
4.1 Benefits of using AAOM From a
User Perspective (RQ1)
In order to find answers to RQ1, we use five themes
that cover direct benefits the participants state about
how AAOM improves communication by collabora-
tive modelling by including the participants. Addi-
tionally, a method comparison along with visual rep-
resentation provides insights to AAOM.
To evaluate results, we refer to a formula devised
in Section 3.4. The first theme Benefits shows all
codes are positive statements for using AAOM. All
the codes in this theme are positive statements and
adhere to the case-study research question about what
5
http://digi.lib.ttu.ee/i/?4064
the benefits are of using AAOM. The highest ranked
codes are a secure feeling for the project direction,
mutual communication and discovering new angles in
requirements for the project. Four codes have a lower
evaluation score and thus, are more unreliable to draw
confident conclusions from.
Next, theme Collaborative Modelling relates to
the AAOM-theory of improving communication be-
tween clients and the development team by working
together on requirement elicitation. The highest rated
codes are all positive suggestions and thus, confirm-
ing the expectations set by AAOM. Having everyone
“on the same page”, improves understandability and
pinpointing problems represents a benefit for collab-
orative modelling. There is one negative suggestion
about composing goal models should be more struc-
tured.
The Method Comparison theme gathers the par-
ticipants’ experience with similar methods compared
to AAOM. Unfortunately, the experience with simi-
lar methods is low among participants in the running
case study. Consequently, a comparison with other
methods is not sufficient and does not provide enough
results.
The final two themes have less codes than Benefits
and Collaborative Modelling while the codes have a
high formula value that shows same opinions from all
participants. The Participation theme gathers objec-
tive opinions about how the method includes everyone
in the project and it shows the ICT-expertise is an ad-
vantage. The theme Visual Representation consists of
only one strongly referenced code denoting that the
goal-model representation is a definite benefit.
4.2 Effects of the Projects Setup and
Tooling on AAOM (RQ2)
Three themes in Section 3.1 cover RQ2 starting with
explaining the setup of elaboration sessions and find-
ing the effects for the AAOM-method application.
The next theme is related to temporal measures as
well, explaining time usage to manage AAOM mod-
els. The final theme covers effects of software based
tooling usage on AAOM application.
Which practices are favoured and which need im-
provement, are ranked again by formula results (Sec-
tion 3.4). The Elaboration Sessions theme covers
the AAOM-methods application-session content, -
duration and -suitability. The highest ranked code
suggests the selected session length that is 1.5 hours,
is selected correctly even though one low ranked con-
tracting code is gathered to have shorter sessions. The
remaining codes are ranked relatively low with incon-
clusive findings.
Evaluating A Novel Agile Requirements Engineering Method: A Case Study
161
The theme Modelling Time Usage answers to
questions related to spent time on modelling activities
carried out using the AAOM method. Positive state-
ments are gathered for system requirements fast cap-
ture, fast development based on models and overall
effective time usage. One neutral statement is added
that moderate time is spent until an idea is formed
as a user story. The unit of measurement is the par-
ticipants’ subjective feelings about the time spent on
method activities.
In order to answer aspects of tool importance in
using AAOM, we find a theme Tools Usage. There are
several different codes revealing several viewpoints.
The two highest ranked codes show that using freely
available tools is provisionally satisfactory while hav-
ing an integrated suit would reduce the amount of
work needed for completing tasks. Available com-
mercial tools are considered better while they are pro-
hibitively expensive, especially in smaller projects.
We conclude interest exists to use new tools that are
better tailored than the chosen free tools in the run-
ning case.
4.3 Further AAOM-refinement Meeds
(RQ3)
For further improvementswe collect codes under only
one theme called Method Clarification that addresses
what practices of AAOM are clear and which need ex-
planation, or redefining. On the positive side, we find
that the sequence of activities for goal models com-
posing is clear. In top ranked codes we also find pos-
itive statements being clear about concepts of quality
goals, user stories and roles.
On the negative side, we gather contradicting in-
formation if creating user stories for the lowest level
goals is clear. Both codes are highly ranked while
statements about the process being unclear are ranked
slightly higher. Studying this contradiction is future
research work in addition to the exact usage of qual-
ity goals in relation to user stories that is currently un-
clear. As a counterpart for top ranked positive state-
ments, there are few contradicting negativestatements
about same aspects. Analysing these contradictions
deeper, we find that a lack of experience causes these
results.
4.4 Emerged Results
The emerged codes result from grounded coding that
is explained in Section 3.4. The participants ex-
press their feelings in the theme Drawbacks about the
project setup that is not directly related to AAOM
modelling while it still affect its usage. The most
mentioned code states that experienced participants
are needed to fully benefit from the AAOM-method.
Thus, AAOM is not intuitively fully understandable
to all participants in requirements engineering. To
aid understandability, a solution could be a better user
guide for the AAOM method, as suggested by inter-
viewees in the second top mentioned code. Further-
more, the analyst is identified as carrying a vital role
in AAOM usage by having the biggest responsibility
in modelling the activities. As a final statement worth
noting a doubt exists pertaining to the method suit-
ability in smaller projects due to a possible method
overhead.
The Expectations theme shows the participants’
positive impression. Working results and extensible
implementations are expected, goal models and user
stories are expected to be updated. A negative point
is that more participants are needed to accomplish the
goals pertaining to the project setup and not the im-
plementation phase of AAOM.
The results captured under the theme Modelling
Suitability support the theory about the method help-
ing to focus on objectives elicitation and organizing
thoughts to express a client’s feelings. Statements
gathered under the theme New Ideas give novel sug-
gestions stemming from the study participants such
as assigning financial values to the goal, to use goal
models as a system documentation, and so on.
5 CONCLUSION
In this article, we evaluate the AAOM-method as
a novel method for requirements engineering in a
small-scale project that employs agile software de-
velopment for developing a mobile app from scratch.
A case-study based research approach is instrumen-
tal for an evaluation. Interviews with project partici-
pants provide the most relevant input for the AAOM-
method’s evaluation. Interviews are coded and ana-
lyzed to answers research questions, also new knowl-
edge outside the scope of research questions we gen-
erate.
The benefits of using the AAOM-method are a
secure feeling for project direction, mutual commu-
nication and discovering new angles in requirements
for the project. The visual representation provided by
goal models is intuitively comprehensible. Further-
more, the AAOM-method enforces communication
via collaborative modeling, improving understand-
ability, pinpointing problems and involving partici-
pants. Time spent on AAOM activities for gathering
requirementsis found to be adequate and the overhead
during that process is marginally low. For projects
ENASE 2016 - 11th International Conference on Evaluation of Novel Software Approaches to Software Engineering
162
on a small budget, free tools with manual integration
suffice while there is a desire for AAOM-tailored in-
tegrated tool support. Unclear procedures, that need
better guidance or redefining, pertain to the way of
finding the lowest level of goal models with the cor-
responding deducing of user stories, and quality-goal
use. One of the discovered factors outside the scope
of research questions is that before using AAOM ef-
fectively, longer experience in ICT is needed or addi-
tional guidance should be available. Finally, the ana-
lyst role has been identified as too heavy in applying
the AAOM method.
The first limitation is that the collected informa-
tion is mostly based on the client-side opinions be-
cause three participants out of four fill the client role.
Only two of the four participants have experience par-
ticipating in software developmentprocesses and con-
sequently, feedback also stems from inexperienced
participants. Another limitation is a lack of inter-
view feedback with respect to other methods for agile
and requirements/engineering methods. This can re-
veal how well the AAOM-method is combinable with
other agile methods. A final limitation is related to re-
search bias towards supporting factors of AAOM us-
age since one researcher is an inventor of AAOM.
Further studies and investigation must apply the
AAOM-method in diverse domains for demonstrat-
ing its universal applicability. Based on the feedback
from participants of the runningcase, an improvement
of the AAOM-method must include better explana-
tion of the role of quality goals. Since the analyst
plays an important role during applying the AAOM-
method, we need a deeper understanding of this es-
sential role.
REFERENCES
Ambler, S. W. and Lines, M. (2012). Disciplined agile de-
livery: A practitioner’s guide to agile software deliv-
ery in the enterprise. IBM Press.
Beck, K. (2000). Extreme programming explained: em-
brace change. Addison-Wesley.
Beck, K. et al. (2001). The agile manifesto.
Cao, L. and Ramesh, B. (2008). Agile requirements engi-
neering practices: An empirical study. IEEE Software,
25(1):60–67.
Carlson, D. and Matuzic, P. (2010). Practical Agile Re-
quirements Engineering. Technical report.
Cockburn, A. (2006). Agile software development: the co-
operative game. Pearson Education.
Cohn, M. (2004). User Stories Applied: For Agile Software
Development. Addison-Wesley.
Hull, E. et al. (2010). Requirements engineering. Springer.
Jensen, D. (2008). Confirmability, page 113. Sage.
Kniberg, H. and Skarin, M. (2010). Kanban and Scrum-
making the most of both. Lulu.com.
Larman, C. and Vodde, B. (2008). Scaling lean & agile
development: thinking and organizational tools for
large-scale Scrum. Pearson Education.
Leffingwell, D. (2010). Agile software requirements: lean
requirements practices for teams, programs, and the
enterprise. Addison-Wesley.
Leffingwell, D. (2016). Scaled agile framework.
http://scaledagileframework.com.
Leffingwell, D. and Aalto, J. (2009). A lean and scalable re-
quirements information model for the agile enterprise.
https://scalingsoftwareagility.wordpress.com/.
Poppendieck, M. and Poppendieck, T. (2007). Implement-
ing lean software development: from concept to cash.
Pearson Education.
Robson, C. (2002). Real word research. Blackwell.
Runeson, P., Host, M., Rainer, A., and Regnell, B. (2012).
Case study research in software engineering: Guide-
lines and examples. Wiley.
Salda˜na, J. (2015). The coding manual for qualitative re-
searchers. Sage.
Schwaber, K. (2004). Agile project management with
Scrum. Microsoft Press.
Shenton, A. K. (2004). Strategies for ensuring trustwor-
thiness in qualitative research projects. Education for
Information, 22(2):63–75.
Sterling, L. and Taveter, K. (2009). The art of agent-
oriented modeling. MIT Press.
Sutherland, J. et al. (2007). Distributed scrum: Agile project
management with outsourced development teams. In
System Sciences. IEEE.
Tenso, T. and Taveter, K. (2013). Requirements engineering
with agent-oriented models. In Evaluation of Novel
Approaches to Software Engineering, pages 254–259.
Van Lamsweerde, A. et al. (2009). Requirements engineer-
ing: from system goals to UML models to software
specifications. Wiley.
Van Maanen, J. (1979). The fact of fiction in organizational
ethnography. Administrative Science Quarterly, pages
539–550.
Version One, I. (2016). 9th annual state of ag-
ile survey. http://info.versionone.com/state-of-agile-
development-survey-ninth.html.
Yin, R. K. (2013). Case study research: Design and meth-
ods. Sage.
Yu, E. S. (2009). Social modeling and i*. In Conceptual
Modeling: Foundations and Applications, pages 99–
121. Springer.
Evaluating A Novel Agile Requirements Engineering Method: A Case Study
163