ArticlePDF Available

Abstract and Figures

Requirements Engineering (RE) has received much attention in research and practice due to its importance to software project success. Its inter-disciplinary nature, the dependency to the customer, and its inherent uncertainty still render the discipline dicult to investigate. This results in a lack of empirical data. These are necessary, however, to demonstrate which practically relevant RE problems exist and to what extent they matter. Motivated by this situation, we initiated the Naming the Pain in Requirements Engineering (NaPiRE) initiative which constitutes a globally distributed, bi-yearly replicated family of surveys on the status quo and problems in practical RE. In this article, we report on the qualitative analysis of data obtained from 228 companies working in 10 countries in various domains and we reveal which contemporary problems practitioners encounter. To this end, we analyse 21 problems derived from the literature with respect to their relevance and criticality in dependency to their context, and we complement this picture with a cause-e↵ect analysis showing the causes and e↵ects surrounding the most critical problems. Our results give us a better understanding of which problems exist and how they manifest themselves in practical environments. Thus, we provide a first step to ground contributions to RE on empirical observations which, until now, were dominated by conventional wisdom only.
Content may be subject to copyright.
Empirical Software Engineering manuscript No.
(will be inserted by the editor)
Naming the Pain in Requirements Engineering
Contemporary Problems, Causes, and Eects in Practice
D. M´endez Fern´andez ·S. Wagner ·M.
Kalinowski ·M. Felderer ·P. Ma fra ·A.
Vetr `o ·T. Conte ·M.-T. Christiansson ·
D. Greer ·C. Lassenius ·T. M¨annist¨o ·
M. Nayebi ·M. Oivo ·B. Penzenstadler ·
D. Pfahl ·R. Prikladnicki ·G. Ruhe ·A.
Schekelmann ·S. Sen ·R. Spinola ·A.
Tuzc u ·J. L. de la Vara ·R. Wieringa
Received: date / Accepted: date
Abstract Requirements Engineering (RE) has received much attention in re-
search and practice due to its importance to software project success. Its inter-
disciplinary nature, the dependency to the customer, and its inherent uncertainty
still render the discipline dicult to investigate. This results in a lack of empirical
data. These are necessary, however, to demonstrate which practically relevant RE
problems exist and to what extent they matter. Motivated by this situation, we
initiated the Naming the Pain in Requirements Engineering (NaPiRE) initiative
which constitutes a globally distributed, bi-yearly replicated family of surveys on
the status quo and problems in practical RE.
In this article, we report on the qualitative analysis of data obtained from
228 companies working in 10 countries in various domains and we reveal which
contemporary problems practitioners encounter. To this end, we analyse 21 prob-
lems derived from the literature with respect to their relevance and criticality in
dependency to their context, and we complement this picture with a cause-eect
analysis showing the causes and eects surrounding the most critical problems.
Our results give us a better understanding of which problems exist and how
they manifest themselves in practical environments. Thus, we provide a first step
to ground contributions to RE on empirical observations which, until now, were
dominated by conventional wisdom only.
Keywords requirements engineering ·survey research
1Introduction
Requirements Engineering (RE) aims at the elicitation, analysis, and specification
of requirements that unambiguously reflect the intended purpose of a software
Daniel M´endez Fern´andez
Technische Universit¨at M¨unchen, Germany
Tel.: +49-89-28917056
E-mail: Daniel.Mendez@tum.de
© Springer. This is the authors’ version of the work. It is posted here by permission of Springer for your personal use.
Not for redistribution. The final publication is available at Springer via http://dx.doi.org/10.1007/s10664-016-9451-7
2 D. M´endez Fern´andez et al.
system considering and aligning the viewpoints of all relevant stakeholders. Pre-
cise and consistent requirements directly contribute to appropriateness and cost-
eectiveness in the development of a system [34] whereby RE is a determinant
of productivity and (product) quality [12]. Yet, RE remains a discipline that is
inherently complex due to the various influences in industrial environments. The
process itself is driven by uncertainty as many aspects are usually not clear when
setting up a project [29]. The pro ject setting, however, influences the choice of
methods, approaches, and tools in RE as in no other software engineering disci-
pline. This makes it impossible to standardise the discipline and to propose holistic
solutions to RE. The interdisciplinary nature of the discipline and the dependency
to various socio-economic and process-related factors that pervade RE make it
dicult to investigate and improve [27].
Over the last years, we have observed a strong research community arise and
propose a plethora of promising contributions to RE. Yet, we still know very little
about the practical impact of those contributions or whether they are in tune with
the practical problems they intend to address [30]. The state of empirical evidence
in RE is particularly weak and dominated by, if at all, isolated case studies and
small-scale studies investigating aspects that hardly can be generalised. It remains
often unknown for which situations the observed eects of applying a specific
method holds or what the long-term views are on cost and benefit when adopting
and applying those methods. In most cases, accurate evaluations starve in the
future work section of publications [8].
Theoretical and practical contributions to RE are heavily steered by conven-
tional wisdom rather than empirical observations. In our current understanding,
there are two reasons. First, we still do not exploit the full potential of empirical
software engineering principles in RE [10] to reveal theories and practically rele-
vant improvement goals, and, in consequence, for evaluating contributions on the
basis of clear and practically relevant hypotheses. Both, however, are a prerequisite
for problem-driven research. Second – and more important – it is per se dicult
to provide proper empirical figures that could demonstrate, for instance, problems
in RE or even particular success factors [8]. This leads to the current situation
where we still lack empirically grounded and comprehensive theories for RE. To
reach this aim, we need, as a first step, to expand our knowledge about which
problems exist in RE and what their causes are while covering the particularities
of the project contexts for which those phenomena hold. This knowledge about
the state of practice and potentially resulting problems in RE would allow us to
better steer future research in a problem-driven manner.
This overall situation motivated the Naming the Pain in Requirements Engi-
neering (NaPiRE) initiative under the umbrella of the International Software En-
gineering Research Network (ISERN). NaPiRE constitutes a globally distributed
family of practitioner surveys on the status quo, problems, and their causes and
eects in RE. The overall objective of NaPiRE is to establish an open knowledge
base about the status quo as well as practical problems and needs in RE. In the
long run, the obtained data shall support us in defining a holistic theory of RE cov-
ering a broad set of context factors for which particular phenomena hold. NaPiRE
is currently run by 26 researchers from 14 countries around the world.
Naming the Pain in Requirements Engineering 3
1.1 Research Objective
Our objective is to use the NaPiRE data from the 2014/15 survey run to explore
which problems practitioners experience and what their causes and eects are.
This shall allow us to provide a basis for steering RE research in a problem-driven
manner.
1.2 Contribution
In this article, we contribute with an analysis of:
1. RE problems practitioners experience in their project setting and an analysis
of the problems with respect to their criticality for project failures, including a
dierentiated view according to chosen context factors such as the development
process model used (agile or plan-driven).
2. Most reported causes of the RE problems, as reported by our survey respon-
dents, and their influence on the most cited RE problems.
3. Causes and eects, providing an overview on the survey results on causes and
eects of the most critical RE problems.
The analysis is based on data obtained from 228 companies in 10 countries.
We expect our contributions to
increase the awareness of practitioners of problems reoccurring in RE and
causes that lead to those problems, thereby allowing them to directly assess
their own situation with respect to the state of practice, and
provide an empirical foundation for researchers to base their investigations on
a set of practical problems and causes.
1.3 Outline
The article is organised as follows. In Section 2, we describe related work and
also provide information on the background of the NaPiRE initiative including
previously published material. Section 3 introduces the study design including
research questions and the data collection and analysis procedures. In Section 4,
we report on our study results, before concluding our article in Section 5.
2 Related Work and Background
In the following, we will discuss work related to our study, before introducing the
NaPiRE initiative and previously published material in that context in detail.
2.1 Related Work
There is a large body of research on requirements engineering in general and on
specific RE methods in particular. Our study touches on the results of many of
them, but we cannot discuss them all here. There exist surprisingly few compre-
hensive systematic literature reviews. For instance, there is a systematic review
4 D. M´endez Fern´andez et al.
on eectiveness of requirements elicitation techniques [13] and mapping studies
on creativity [24], requirements specification improvement methods [35] as well as
the empirical work on requirements specifications techniques [9]. In the latter, the
authors emphasise that most studies are experiments, and that the practitioners’
view is missing.
We will focus on related work which performed survey research in the area of
requirements engineering or at least with a strong RE component.1In RE survey
research, we see two major areas: investigations of techniques and methods and
investigations of general practices and contemporary issues in practice. Both areas
investigate to some degree problems in RE and their causes.
Contributions that investigate techniques and methods analyse, for example,
selected requirements phases and which techniques are suitable to support typical
tasks in those phases. Cox et al. [11] performed a broader investigation of all phases
to analyse the perceived value of the RE practices recommended by Sommerville
and Sawyer [38]. Studies like those reveal the eects of given techniques when
applying them in practical contexts.
Surveys on general practices and phenomena in industry include the well-known
Chaos Report of the Standish Group, examining especially root causes for project
failures of which most are to be seen in RE, such as missing user involvement.
Whereas the report is known to have serious flaws in its design negatively af-
fecting the validity of the results [14], other studies, such as the (German) Suc-
cess study [6], conduct a similar investigation of German companies including a
detailed and reproducible study design. Still, both surveys exclusively investigate
failed projects and general causes at the level of overall software processes. A sim-
ilar focus, but exclusively narrowed down to the area of RE, had the study of
Kamata et al. [23]. They analysed the criticality of the single parts of the IEEE
software requirements specification Std. 830-1998 [17] on project success.
The focus of those studies, however, does not support the investigation of con-
temporary phenomena and problems of RE in industry. Nikula et al. [33] present
a survey on RE at the organisational level of small and medium-sized companies
in Finland. Based on their findings, they inferred improvement goals, e.g., on opti-
mising knowledge transfer. Staples et al. [39] conducted a study investigating the
industrial reluctance on software process improvement. They discovered dierent
reasons why organisations do not adopt normative improvement solutions, for ex-
ample, CMMI and related frameworks (focussing on assessing and benchmarking
companies rather than on problem-driven improvements, see [32, 36]). Exemplary
reasons for a reluctance to normative improvement frameworks were the small
company size where the respondents did not see clear benefit.
A survey that directly focused on discovering problems in practical settings
was performed by Hall et al. [3]. They empirically underpin the problems dis-
cussed by Hsia et al. [16] and investigated a set of critical organisational and
project-specific problems, such as communication problems, inappropriate skills
or vague requirements. Solemon, Sahibuddin, and Abd Ghani [37] report on a sur-
vey on RE problems and practices in Malaysian software companies. They found
several of the RE problems we also saw in our survey. Liu, Li, and Peng [25] de-
scribe a survey conducted in China about practices and problems in RE. They
1Parts of the following text are based on our related work discussion in [28] as the related
work has not changed significantly.
Naming the Pain in Requirements Engineering 5
discuss several problems we also investigated but concentrate on China. Verner
et al. [40] ran a survey in Australia and the USA. They concentrated on success
factors in RE and found good requirements, customer/user involvement, and ef-
fective requirements management to be the best predictors of project success. In
the study reported in this article, we identified problems and their causes which
can be used to refine the abstract success factors identified by Verner et al. For
instance, we identified incomplete and/or hidden requirements as a main problem
to reach ’good requirements’ and weak qualification as well as lack of experience as
its main causes. These causes are useful guidelines to reach more eective require-
ments management. Further studies on RE problems and causes are still rare. For
instance, Al-Rawas and Easterbrook [2] present a field study on communication
problems in requirements engineering.
In summary, we have existing work using survey research to understand specific
RE techniques as well as to understand practical problems. Yet, so far there has
not been a large, replicated and world-wide analysis of RE problems together with
their causes in practice.
2.2 The NaPiRE Initiative
The NaPiRE (Naming the Pain in Requirements Engineering) initiative was started
by Daniel M´endez Fern´andez and Stefan Wagner in 2012 as a reaction to the lack
of a general empirical basis for requirements engineering research. The basic idea
was to establish a broad survey investigating the status quo of requirements engi-
neering in practice together with contemporary problems practitioners encounter.
This should lead to the identification of interesting further research areas as well
as success factors for RE.
The initial team was convinced that because of the diversity of RE in research
and practice, we would not be able to achieve this high goal by ourselves and in
a single survey. Therefore, NaPiRE was created as a means to collaborate with
researchers from all over the world to conduct the survey in dierent countries.
This allows us to investigate RE in dierent cultural environments and increase
the overall sample size. Furthermore, we decided to run the survey every two years
so that we can cover slightly dierent areas over time and have the possibility to
observe trends. The conduct of NaPiRE is guided by the four principles described
in Tab. 1.
At present, the NaPiRE initiative has members from 14 countries mostly from
Europe but also North-America, South-America and Asia. There have been two
runs of the survey so far. The first was the test run performed only in Germany
and in the Netherlands in 2012/13. The second run was performed in 10 coun-
tries in 2014/15. All up-to-date information on NaPiRE together with links to all
publications and the data is available on the web site http://www.re-survey.org.
The first run in Germany together with the overall study design was published
in [28]. A preliminary version was also published in [27] and the detailed data
and descriptive analysis is available as technical report [26]. This first run already
covered the spectrum of status quo and problems. It had additional questions on
the expectations on a good RE which we removed in the second run because they
provided the least interesting insights. The study design was described with the
bi-yearly replications and world-wide distribution in detail. Furthermore, a first
6 D. M´endez Fern´andez et al.
Tabl e 1 Guiding principles of NaPiRE
Openness Openness begins by cordially inviting researchers and practi-
tioners of any software-engineering-related community to con-
tribute to NaPiRE and ends by disclosing our results and
reports without any restrictions or commercial interest.
Transparency All results obtained from the distributed surveys are com-
mitted to the PROMISE repository. This shall allow other
researchers an independent data analysis and interpretation.
Anonymity The participation in NaPiRE in the form of a survey respon-
dent is possible by invitation only. This supports a transpar-
ent result set and response rate. We collect no personal data,
however, and every data set obtained from the survey will be
carefully cleansed of information that might be traced back
to a specific company to ensure that no personal data will be
disclosed to the public. That is, we guarantee that no answer
set can be related to survey participants.
Accuracy and Validity With accuracy and validity, we refer in particular to the data
collection and to the data analysis. Each question in the sur-
vey is carefully defined according to a jointly defined theory to
specifically confirm or refute existing expectations. The data
analysis is furthermore performed in joint collaboration by
dierent researchers to maximise the validity of the results.
version of a theory of the status quo and problems in RE was provided in the
form of hypotheses. Overall, we were able to get full responses from 58 companies
to test the theory. Most of the proposed theory could be supported and changes
were discussed based on the data. Finally, a detailed qualitative analysis of the
experienced problems and how they manifest themselves was made. The article
at hands concentrates on the replication of that part of the survey, with further
emphasis on the problems and their causes.
For the second run, we have published three papers so far, concentrating on
specific aspects and the data from only one or two countries. So far, there is no
comprehensive analysis of problems and causes based on the complete, interna-
tional data set.
In [18], we used the Brazilian data to explore how to analyse problems and
causes in RE in detail. In particular, we tested the use of probabilistic cause-
eect diagrams on this data to better understand the relationship of causes and
problems. We introduced those diagrams for causal analysis purposes [22], and have
further detailed them subsequently [21]. We decided to employ these diagrams in
our subsequent eorts, including this article, in which we use them for analysing
causes and eects of problems based on the complete data set.
Thereafter, in [31], we concentrated on analysing the similarities and dierences
in the experienced problems between Brazil and Germany. We also used the prob-
abilistic cause-eect diagrams for problem and cause analysis. Our key insights in
this article were that the dominating factors are related to human interactions in-
dependent of country, project type or company. Furthermore, we observed a higher
inclination to standardised development process models in Brazil and slightly more
non-agile, plan-driven RE in Germany.
Finally, in [19], we concentrated on the often mentioned problem of Incomplete
and / or hidden requirements and investigated common causes for this problem
based on the Austrian and Brazilian data. The most common causes we found were
Naming the Pain in Requirements Engineering 7
Missing qualification of RE team members,Lack of experience,Missing domain
knowledge,Unclear business needs and Poorly defined requirements.
3 Study Design
The overall objective of the NaPiRE endeavour is to use survey research in a
globally distributed an replicated manner to build a holistic theory of the industrial
status quo in requirements engineering. To this end, we conduct the survey bi-
yearly while continuously adapting our questionnaire in response to data obtained
from previous years (see also Sect. 2.2).
In the following, we introduce those information on the study design relevant to
the analysis presented of this article. Details on the overall principles and process
followed in NaPiRE, as well as on the team involved, can be taken from our project
website RE-Survey.org. There, we also publish the full instrument used under the
publication section.
3.1 Research Questions
Our objective is to get a better understanding of which problems practitioners
encounter in RE, and how those problems relate to the overall project setting
(causes and problems). To this end, we formulate three research questions, shown
in Table 2, to steer the design of our study.
Tabl e 2 Research questions.
RQ 1 Which contemporary problems exist in RE?
RQ 2 What are observable patterns of problems and context characteristics?
RQ 3 What are their perceived causes and eects?
The first question aims at understanding which problems practitioners experi-
ence in general in their RE and what their criticality is w.r.t. project failure. This
more descriptive view is complemented by the second research question, which
aims at understanding whether there exist problems that relate to specific context
factors, such as the company size or the type of used process model. Once we un-
derstand whether there exist specific patterns in the problems, we want to know
what their perceived causes and implications are going beyond project failure.
3.2 Instrument
The overall instrument used in NaPiRE constitutes in total 35 questions used to
collect data on (a) the demographics, (b) how practitioners elicit and document
requirements, (c) how requirements are changed and aligned with tests, (d) what
and how RE standards are applied and tailored, (e) how RE is improved, and
finally (f) what problems practitioners experience in their RE. In the study at
hands, we focus on the problems practitioners experience in their RE while using
8 D. M´endez Fern´andez et al.
Tabl e 3 Questions (simplified and condensed excerpt).
Parts No. Question Type
Demographics Q 1 What is the size of your company? Closed(SC)
Q2 Pleasedescribethemainbusinessareaandapplication
domain.
Open
Q3 Doesyourcompanyparticipateinglobally distributed
projects?
Closed(SC)
Q 4 In which country are you personally located? Open
Q5 Towhichprojectroleareyoumostfrequentlyassigned? Closed(SC)
Q6 Howdoyourateyourexperienceinthisrole? Closed(SC)
Q7 Whichorganisationalroledoesyourcompanytakemost
frequently in your projects?
Closed(MC)
Q8 Whichprocess model doyoufollow(or avariationof
it)?
Closed(MC)
Status Quo Q 9 How do you elicit requirements? Closed(MC)
Q10 Howdoyoudocumentfunctionalrequirements? Closed(SC)
Q11 Howdoyoudocumentnon-functionalrequirements? Closed(SC)
Q12 Howdoyoudealwithchangingrequirementsafterthe
initial release?
Closed(SC)
... ... ...
Q16 Whatrequirementsengineeringcompanystandardhave
you est ablishe d at your comp any?
Closed(MC)
... ... ...
Problems Q 28 Considering your personal experiences, how do the fol-
lowing (more general) problems in requirements engi-
neering apply to your projects?
Likert
Q29 Considering your personally experienced problems
(stated in the previous question), which ones would you
classify as the five most critical ones (ordered by their
relevance).
Closed
Q 30 Considering your personally experienced most critical
problems (selected in the previous question), which
causes do they have?
Open
Q 31 Considering your personally experienced most critical
problems (selected in the previous question), which im-
plications do they have?
Open
Q 32 Considering your personally experienced most critical
problems (selected in the previous question), which
mitigations do you define (if at all)?
Open
Q 33 Considering your personally experienced most critical
problems (selected in the previous question), which
would you classify as a major cause for project failure
(if at all)?
Closed(MC)
the answers given to selected questions on the status quo to answer RQ 2. Table 3
summarises an excerpt of our questionnaire demonstrating the scope of our study.
The full questionnaire can be taken from the publication section of our project
website.
In this part, we use a mix of open questions and closed ones. The type of
question is denoted in the table (last column). In case of closed questions, the
answers can be mutually exclusive single choice answers (SC) or multiple choice
ones (MC). Most of the closed multiple choice questions include a free text option,
e.g., “other” so that the respondents can express company-specific deviations from
standards we ask for. We furthermore use Likert scales on an ordinal scale of 5 and
define for each a maximum value (e.g., “agree”, or “very important”), a minimum
value (e.g., “disagree”, or “very unimportant”), and the middle (“neutral”). The
latter allows the respondents to make a selection when they have, for example, no
opinion on the given answer options. Finally, we define selected questions as condi-
tional ones to guide through the survey by filtering subsequent question selection.
Naming the Pain in Requirements Engineering 9
For instance, if respondents state that they have not defined any company-specific
RE standard (Q 16), the remaining questions on the standards are omitted.
For the analysis of the problems (Q 28 to Q 33), we first present a list of prob-
lems practitioners are meant to typically encounter in practice. This list emerged
from previously conducted external studies (see also our related work section 2)
and has been already used in our first survey round (see also [28] for a richer
discussion on the elaboration of the list). For this survey round, we use the same
list which includes a set of 21 pre-compiled problems shown next in no particular
order:
Communication flaws within the project team
Communication flaws between project team and the customer
Terminological problems
Unclear responsibilities
Incomplete and / or hidden requirements
Insucient support by project lead
Insucient support by customer
Stakeholders with diculties in separating requirements from previously known
solution designs
Inconsistent requirements
Missing traceability
Moving targets (changing goals, business processes and / or requirements)
Gold plating (implementation of features without corresponding requirements)
Weak access to customer needs and / or (internal) business information
Weak knowledge of customer’s application domain
Weak relationship to customer
Time boxing / Not enough time in general
Discrepancy between high degree of innovation and need for formal acceptance
of (potentially wrong / incomplete / unknown) requirements
Technically unfeasible requirements
Underspecified requirements that are too abstract and allow for various inter-
pretations
Unclear / unmeasurable non-functional requirements
Volatile customer’s business domain regarding, e.g., changing points of contact,
business processes or requirements
The respondents are then asked to report the relevance of the presented prob-
lems for their project setting, before being asked to select the 5 most critical
ones (Q 29). The subsequent questions on the causes, the eects, and potential
mitigation strategies consider then those 5 problems.
3.3 Data Collection
The survey is conducted by invitation only to have a better control over the dis-
tribution of the survey among specific companies and also to control the response
rate. The responses where, however, anonymous to allow our respondents to freely
share their experiences made within their respective company. For each company,
we invited one respondent as a representative of the company. In case of large
companies involving several autonomous business units working each in a dier-
ent industrial sector and application domain, we selected a representative of each
10 D. M´endez Fern´andez et al.
unit. For the data collection, each country representative defined an invitation list
including contacts from dierent companies and initiated the data collection inde-
pendently as an own survey project. All surveys relied on the same survey tool2
hosted and administrated by the representatives for Germany. Information on the
overall data collection procedure can be also taken from our project website.
For the study at hands, we conducted the survey in the countries summarised
in Table 4. The data collection phase varied among the countries and some of
the countries also collected the data in multiple tranches potentially resulting in
longer inactivity phases.
Tabl e 4 Data collection phase (overview).
Area Country Data Collection Phase
Central Europe (CE)
Austria (AT) 2014-05-07 to 2014-09-15
Germany (DE) 2014-05-07 to 2014-08-18
Ireland (IE) 2014-05-07 to 2014-12-31
North America (NA)
Canada (CA) 2014-05-07 to 2015-08-15
United States of America (US) 2014-05-07 to 2015-05-01
Northern Europe (NE)
Estonia (EE) 2014-05-07 to 2014-10-31
Finland (FI) 2015-06-01 to 2015-08-28
Norway (NO) 2014-05-07 to 2014-09-15
Sweden (SE) 2014-05-07 to 2014-09-15
South America (SA)
Brasil (BR) 2014-12-09 to 2015-03-31
3.4 Data Analysis
To answer our research questions, we first need to quantify the answers given for
the selection of the predefined problems the participants shall rank as they have
experienced them in their projects. As part of this quantification, we also sum up
to which extent the given problems have led to project failures in the experience
of the participants.
For analysing the answers given to the open question on what causes and
eects the RE problems have (Q 30 to Q 32), we apply qualitative data analysis
techniques as recommended in context of Grounded Theory [15, 1]. In particular,
we considered the following basic coding steps:
1. Open coding to analyse the data by adding codes (representing key character-
istics) to small coherent units in the answers, and categorising the developed
concepts in a hierarchy of categories as an abstraction of a set of codes – all
repeatedly performed until reaching a state of saturation. We define the (theo-
retical) saturation as the point where no new codes (or categories) are identified
and the results are convincing to all participating researchers [4].
2. Axial coding to define relationships between the concepts, e.g., “causal condi-
tions” or “consequences”.
2We im p l e m e nted the s u r v e y as a Web appli c a t i o n u s i n g the Enterprise Feedback Suite.
Naming the Pain in Requirements Engineering 11
3. Internal Validation as a form of internal quality assurance of the obtained
results.
Please note that we deviate from the approach to Grounded Theory as intro-
duced by Glaser and Strauss [15] in two ways. First, given that we analyse data
from an anonymously conducted survey after the fact, we are not able to follow a
constant comparison approach where we iterate between the data collection and
the analysis. This also means that we are not able to validate our results with
the participants, but have to rely in internal validation procedures to reduce the
resulting threats to the validity (see also Sect. 3.5 discussing our validity proce-
dures). Second, we do not inductively build a theory from bottom up, as we start
with a predefined conceptual model (i.e. the problems) whereby we do not apply
selective coding to infer a central category.
In our instrument, we have already a predefined set of RE problem codes (given
RE problems, see Q 28 and 29) for which we want to know how the participants
see their causes and eects. For this reason, we rely on a mix of bottom-up and
top-down approach. That is, we start with our pre-defined core category, namely
RE problems and a set of codes each representing one key RE problem, and three
sub-categories: Causes,Eects,andMitigation Strategies, which then group the
codes emerging from the free text answers given by the participants. Within the
causes and eects, we pre-define again the sub-categories. These sub-categories are
as follows:
For the causes, we use the sub-categories Input,Method,Organization,People,
Tools suggested in our previous work on defect causal analysis [20] as we want
to know from where in the socio-economic context the problems stem.
For the implications, we use the sub-categories Customer,Design or Imple-
mentation,Product,Project or Organization,andVerification or Validation
as done in our previous work [31] as we want to know where in the software
project ecosystem the problems manifest themselves.
For each answer given by the participant, we then apply open coding and
axial coding until reaching a saturation in the codes and relationships, and we
allocate the codes to the previously defined sub-categories. A rich discussion on
the principles of analysing textual data and how we generally apply it in context of
the NaPiRE initiative can also be found in our previously published material [41].
For coding our results, we first coded in a team of two coders the first 250
statements to get a first impression of the resulting codes, the way of formulat-
ing them and the level of abstraction for capturing the codes. After having this
overview, we organised a team of five coders within Germany and Brazil. Each
of the coders then coded approx. 200 statements for causes and additional 200
statements for eects, while getting the initial codes from the pilot phase as orien-
tation. In case the coder was not sure how to code given statements, she marked
the code accordingly for the validation phase. During that validation phase, we
formed an additional team of three independent coders who then reviewed those
codes marked as “needs validation” as well as an additional sample, comprising
20% of the statements assigned to each coder, selected on their own. After the
validation phase, we initiated a call where we discussed last open issues regarding
codes which still needed further validation, before closing the coding phase. The
overall coding process took in total three months.
To interpret the resulting codes, in particular the answers to research question
2 where we want to know whether there exist patterns of problems and context
12 D. M´endez Fern´andez et al.
characteristics, we need to put the answers to the problems in relation to the
answers given to previous questions in the questionnaire. To this end, we apply
manual blocking to our results, i.e. we select subsets of results which include spe-
cific variable selections; for instance all results for which a specific process model
has been selected. We then discuss in the group of researchers whether there are
specific dierences in the problems visible, e.g. compared to the problems when
other process models have been selected. However, blocking the codes according to
all possible permutations of the variables in the questionnaire is not feasible. For
this reason, we intentionally block the codes according to a combination of the two
variables company s ize (Q 1) and the type of software process models used (Q 8)
(agile or plan-driven), which we believe to be suitable for an initial observation of
relevant patterns. Of course, further blocking variables from the status quo section
of the questionnaire could be used. However, the relation of the whole underlying
NaPiRE theory to the manifestation of the problems also involves significant eort
and is left for future work.
3.5 Validity
The overall NaPiRE endeavour includes several procedures for checking validity,
i.e., concerning the data collection and analysis phases, as described in detail in
our previously published material [28]. For the analysis of qualitative data, which
is in the scope of this article, we defined additional procedures as described next.
The major threat to validity arises from the actual coding process as coding
is essentially a creative task. Subjective facets of the coders, such as experiences,
expertise, and expectations, strongly influence the way we code free text state-
ments. A further threat arises from the fact that we cannot validate our results
with the respondents given the anonymous nature of our survey. Finally, coding
over a distributed team of researchers can additionally lead to a possibly limited
degree of saturation in the emerging codes.
To decrease the threats, we first conducted a pilot phase in the analysis. After
agreeing on the first resulting codes within the group of coders, i.e. after get-
ting a common understanding on the wording and the level of abstraction in the
codes, we then applied researcher triangulation for the actual coding process. An
independent group of researchers coded all the statements given by the respon-
dents, before we subsequently conducted a validation phase within the group of
researchers. The validation phase during coding should then minimise the amount
of incorrect codes. This validation focussed on codes declared as “needs validation”,
but also on further codes presumably being coded correctly. There, we focused on
the occurrences of the codes rather than on the choice of the labels for the codes
(e.g., CRs and change requests are seen as the same code).
4 Study Results
In the following, we present the survey results concerning the RE problems (RQ 1),
observable patterns (RQ 2) and their common causes and eects (RQ 3) as re-
ported by our respondents. We first summarise the information about the study
Naming the Pain in Requirements Engineering 13
population, characterising the responding organisations, as this information is cru-
cial to enable a suitable interpretation of the results.
4.1 Study Population
In total, 354 organisations spread across 10 dierent countries agreed to answer
the survey. Out of these, 228 (63%) completed the survey by going through all
of its questions and successfully reaching its end. Table 5 shows the number of
completed datasets and the completion rate per country.
Tabl e 5 Study population including response rates, total datasets obtained, completed
datasets, and completion rates. The explanation of the country codes can be take from Table 4.
Response Total Completed Completion
Area Country Rate Datasets Datasets Rate
CE
AT 72 % 18 14 78 %
DE 36.8% 50 41 82%
IE 39.7% 25 18 72%
NA
CA 75% 15 13 87%
US 36.2% 25 15 60%
NE
EE 25.7% 9 8 89%
FI 37.5% 18 15 83%
NO 70.8% 17 10 59%
SE 51.8% 59 20 34%
SA
BR 35.3% 118 74 63%
Total: 354 228 64%
The results reported in this article consider the completed datasets only. These
228 organisations were active in a variety of dierent business domains. The do-
mains were provided by the respondents in free text format (see Table 3, question
Q2) and coded by the researchers. The tag cloud for the coded business domains
can be seen in Figure 1.
This figure shows the frequency of each domain code and highlights the most
frequent ones. At all, 215 of the 228 organisations provided answers for their busi-
ness domain. There is a huge variety in the business domains ranging from embed-
ded software systems (described by the respondents as “Automotive, Embedded
Software” or “Software for medical devices”) over business systems (“business in-
telligence for data centres” or “Software ERP”) to consulting (“IT Consulting” or
“Consulting for secure systems and software”). The most frequent code assigned
was cross-cutting which means that the organisation is actively working with prod-
ucts and/or services that can be applied to several dierent business domains (e.g.,
cloud computing and web applications, custom software development, enterprise
resource planning products, IT consulting services). Additionally, we identified a
very large amount of dierent domains with few data points in each one. Therefore,
considering the amount of organisations active in several domains and the large
variety of dierent domains reported, we decided to characterise the responding
14 D. M´endez Fern´andez et al.
crosscutting (60)
public (28)
sector (28)
enterprise (20)
finance (26)
planning (20)
resource (23)
automotive (15)
healthcare (14)
management (16)
business (9)
energy (11)
insurance (11)
logistics (8)
telecom (9) transportation (8)
aerospace (5)
education (5)
gas (5) geoprocessing (4)
informed (5)
intelligence (4) oil (5)
process (5)
aviation (3)
communication (3) construction (3)
defence (3) e-commerce (3)
entertainment (3) games (3)
human (3)
shipping (3)
workforce (3)
agriculture (2)
chemicals (2)
customer (2)
electronic (2)
railway (2) relationship (2)
scientific (2) software (2)
automation (1)
e-
government (1)
funerary (1)
goods (1) industrial (1)
networking (1)
pulp (1)
steel (1)
Fig. 1 Tag cloud of the business domains of the responding organisations.
organisations independently of their business domain, in terms of domain cross-
cutting characteristics of size and process model used (see also Sect. 3.4).
Concerning size, we grouped organisations as small, medium, and large-sized.
For this grouping, as in [31], we used the number of employees (software and
other areas). Organisations with up to 50 employees were considered small-sized
organisations, with 51 to 250 were considered medium-sized organisations, and
with more than 250 were considered large-sized. Out of the 228 organisations that
completed the questionnaire, 216 provided their number of employees. The sizes
of these organisations are shown in Table 6.
Tabl e 6 Sizes of responding organisations.
Central North Northern South
Size Europe America Europe America Total
Small 12 11 20 26 69
Medium 4 0 12 17 33
Large 36 16 34 28 114
Total 52 27 66 71 216
We can observe that the datasets include relatively large samples of both,
small and large-sized organisations. Considering the distributions of size per region,
except for SA, the responding large-sized organisations slightly outweigh the small
and medium-sized organisations.
Regarding the process models used, respondents answered a multiple choice
question with the following options: RUP, Scrum, V-Model XT, Waterfall, XP, and
Other (in this case informing textually which process model they use). We grouped
these process models into agile (Scrum and XP), plan-driven (RUP, Waterfall and
V-Model XT), and mixed (for those organisations that informed to use agile and
plan-driven process models or variations therein). Out of the 228 organisations
that completed the questionnaire, 196 selected one of the five predefined options
for their process model. The process model of these organisations is shown in
Table 7.
Again, the datasets include relatively large samples of both, agile and plan-
driven organisations. Considering the distribution per region, except for CE, the
Naming the Pain in Requirements Engineering 15
Tabl e 7 Software process models used in responding organisations.
Central North Northern South
Model Europe America Europ e America Total
Agile 12 13 35 32 92
Plan-driven 15 4 8 19 46
Mixed 17 8 19 14 58
Total 45 24 62 65 196
responding organisations following agile process models in the respondents envi-
ronment outweigh the plan-driven ones. The amount of organisations using mixed
process models (or more than one) is large. However, we decided to exclude the
mixed ones from our corresponding analyses to remove a potential confounding
factor, as in these cases we had no information on the extent to which each pro-
cess model is applied in the organisation.
As we believe that agile and plan-driven process models may have dierent
eects on small, medium, and large-sized organisations, we also crossed these two
characterisation dimensions aiming at further exploring potential RE problem pat-
terns (cf. Section 4.3). The result of this crossing is shown in Table 8.
Tabl e 8 Responding organisations by size and process models used.
Process Central North Northern South
Model Size Europe America Europe America Total
Agile Small 2 4 10 14 30
Agile Medium 2 0 10 10 22
Agile Large 9 8 14 8 39
Plan-driven Small 3 2 1 5 11
Plan-driven Medium 1 0 1 2 4
Plan-driven Large 10 2 6 12 20
Total 27 16 42 51 136
Excluding the 58 organisations with mixed process models, 136 organisations
that completed the questionnaire informed the number of employees and prede-
fined process model options. While agile process models are being applied by small
and large-sized representatives of the responding organisations, plan-driven pro-
cess models are mainly applied by the large-sized ones (although we still have some
samples of small sized plan-driven organisations).
We therefore could obtain a balanced characterisation of small, medium and
large-sized organisations of dierent regions enrolled in both, plan-driven and agile
development methods.
4.2 Problems in RE (RQ 1)
Based on the set of 21 pre-defined general RE problems listed in the NaPiRE
questionnaire, the respondents were asked to rank the five most critical ones.
The top 10 most critical RE problems, as informed by the 228 respondents, are
visualised together with the frequency in which they are meant to lead to project
failure in a simplified manner in Figure 2.
16 D. M´endez Fern´andez et al.
Fig. 2 Overall frequency of top 10 RE problems and their relation to project failure.
Naming the Pain in Requirements Engineering 17
Table 9 further summarises the 10 most cited RE problems providing more
details. There, we report on how many of the respondents cited particular prob-
lems, how many considered them as a major cause for project failure, and how
often each problem was ranked in each of the five potential ranking positions, thus,
showing how the bars in Figure 2 are composed.
Tabl e 9 Most cited top 10 RE problems.
RE Problem Total
Cause for project failure
Ranked as #1
Ranked as #2
Ranked as #3
Ranked as #4
Ranked as #5
Incomplete and / or hidden requirements 109 (48%) 43 34 25 23 17 10
Communication flaws between project team
and customer
93 (41%) 45 36 22 15 9 11
Moving targets (changing goals, business pro-
cesses and / or requirements)
76 (33%) 39 23 16 13 12 12
Underspecified requirements that are too ab-
stract
76 (33%) 28 10 17 18 19 12
Time boxing / Not enough time in general 72 (32%) 24 16 11 14 17 14
Communication flaws within the project
team
62 (27%) 25 19 13 11 9 10
Stakeholders with diculties in separating
requirements from known solution designs
56 (25%) 10 13 13 12 9 9
Insucient support by customer 45 (20%) 24 6 13 12 6 8
Inconsistent requirements 44 (19%) 15 8 9 6 9 12
Weak access to customer needs and / or busi-
ness information
42 (18%) 16 7 10 8 8 9
Out of these critical RE problems we highlight the first five, which were cited
by more than 30% of the respondents. Noteworthy is also, however, that some
problems even if they seem not to occur as often as others, seem to be still more
critical as they are meant to lead more often to project failure; for instance, In-
compl ete a nd / or h idd en requ irements being the most frequently cited problem
is, from a relative point of view, not meant to lead as often to project failure
as Communication flaws between project team and the customer even if it occurs
more often. Furthermore, Moving targets do lead more often to project failures
than Underspecified requirements that are too abstract even if they show the same
total frequency of occurrence.
The analysis of the problem patterns described next, concentrates on the top
five problems.
4.3 Problem Patterns (RQ 2)
Given the diversity of the responding organisations, in particular concerning their
sizes and process models, we block the results according to those context factors
to further investigate how the problems manifest within such clusters, aiming at
18 D. M´endez Fern´andez et al.
identifying potential RE problem patterns (see also Sect. 3.4). Table 10 shows
the five most critical RE problems per process model and organisational size.
We can see tha t b esides Gold Plating, which appears for plan-driven medium-
sized organisations, all other problems are also listed in the list of the overall 10
most cited RE problems (Table 9). Nevertheless, it is noteworthy that this specific
cluster had only 4 organisations and that this fact might not represent a relevant
dierence. Also the textual statements of the corresponding respondents did not
show much specifics of plan-driven, medium-sized companies in that respect. Only
the statement “The team believes to be very qualified to understand the business“
hints in the direction that in a plan-driven development process, the developers
are not exposed to as much customer feedback as necessary and think they already
know the customer’s business.
Concerning the occurrence of the problems within the clusters, the only prob-
lem being consistently in the top 3 is Incomplete and/or hidden requirements.Itis
also the most cited problem overall. We will discuss its causes and eects in detail
in Sec. 4.4.2.
Very common is also Communication flaws between project team and the cus-
tomer. It appears in the three most cited problems in all clusters except for plan-
driven and small-sized organisations. We can see one reason in the free-text an-
swers especially for large companies. They tend to split the work in several teams
of which some work directly with customers, while others don’t. One respondent
describes that their “sales or account teams, product managers [. .. ] act as proxies
for the end user”. Small and agile companies seem to suer especially from cus-
tomers not willing to participate with a considerable amount of time (“Not enough
customers willing to help out and also time constraints”, “Customer is busy and
skips meetings.” or “Customers have no time to explain what they actually need”).
The plan-driven, small companies who rated this problem as important, did not
show a consistent pattern of reasons in their free-text answers.
Another dierence concerns the Moving targets problem. This problem is faced
by all plan-driven but also large agile organisations. That plan-driven companies
cite this problem often in comparison to agile companies supports the basic premise
of agile software development that it helps to quickly adapt to changing needs. The
respondents from plan-driven organisations mention as reasons the “Lack of change
management on the customer side”, the “Unclear business vision and understand-
ing by stakeholders” and overall “badly written requirements”. The negative eects
on their projects are manifold including “pro ject delays; extended engagement of
resources beyond original plan; customer dissatisfaction” and “expensive projects,
time consuming implementation, bad quality”.
But why do also large, agile companies often experience the Moving targets
problem? We do not see a clear answer from the free-text responses. Some of the
answers could be explained such that large companies in general have larger, more
complex projects which also might run for a longer time. Then the mentioned
problems are more significant. For example, a change in the management of the
customers was mentioned and seems to have large eects: “senior management
confusion/churn”. But also the chance that over time other people bring in new
ideas and constraints seems to be more likely: “There are always some stakeholders
involved in later part of the project who would come up with new things”. Even
agile development processes cannot compensate this.
Naming the Pain in Requirements Engineering 19
Tabl e 1 0 5 most critical problems per process model used and company size.
Process Citation
Model Size Total Top 5 Problems Count
Agile Small 30 1. Incomplete and / or hidden requirements 18 (60%)
2. Communication flaws between project team and
the customer
14 (47%)
3. Underspecified requirements that are too abstract 13 (43%)
4. Communication flaws within the project team 10 (33%)
5. Time boxing / Not enough time in general 13 (43%)
Agile Medium 20 1. Communication flaws between project team and
the customer
12 (55%)
2. Incomplete and / or hidden requirements 9 (41%)
3. Communication flaws within the project team 8 (36%)
4. Stakeholders with diculties in separating re-
quirements from known solution designs
8 (36%)
5. Weak access to customer needs and / or business
information
7 (32%)
Agile Large 39 1. Incomplete and / or hidden requirements 17 (44%)
2. Moving targets (changing goals, business pro-
cesses and / or requirements)
17 (44%)
3. Communication flaws between project team and
the customer
15 (38%)
4. Time boxing / Not enough time in general 14 (36%)
5. Underspecified requirements that are too abstract 11 (28%)
Plan-
driven
Small 11 1. Incomplete and / or hidden requirements 7 (64%)
2. Communication flaws within the project team 6 (55%)
3. Moving targets (changing goals, business pro-
cesses and / or requirements)
6 (55%)
4. Time boxing / Not enough time in general 5 (45%)
5. Underspecified requirements that are too abstract 5 (45%)
Plan-
driven
Medium 4 1. Communication flaws between project team and
the customer
2 (50%)
2. Gold plating (implementation of features without
corresponding requirements)
2 (50%)
3. Incomplete and / or hidden requirements 2 (50%)
4. Moving targets (changing goals, business pro-
cesses and / or requirements)
2 (50%)
5. Underspecified requirements that are too abstract 2 (50%)
5. Weak access to customer needs and / or business
information
2 (50%)
Plan-
driven
Large 30 1. Incomplete and / or hidden requirements 14 (47%)
2. Communication flaws between project team and
the customer
13 (43%)
3.Underspecified requirements that are too abstract 10 (33%)
4. Communication flaws within the project team 9 (30%)
5. Moving targets (changing goals, business pro-
cesses and / or requirements)
8 (27%)
5. Stakeholders with diculties in separating re-
quirements from known solution designs
8 (27%)
Another dierence between the clusters concerns the Time boxing problem,
which appears mainly in agile and in small organisations. In both agile and plan-
driven, small companies, we found three (related) reasons for this prevalence of
time boxing problems: bad estimations, unrealistic release dates and scope changes.
Our respondents often mentioned that estimations were not accurate: “A combi-
nation of bad planning and bad estimation of time for development” or “Bad
estimates, unrealistic expectations”. Especially sales and marketing is blamed for
promising unrealistic dates: “Sales shouldn’t give wishful promises” or “Release
dates are sometimes arbitrary and often released early to customers creating a
20 D. M´endez Fern´andez et al.
hard deadline”. At last, frequent scope changes seem to contribute to this prob-
lem: “Last minute changes; change of priority; Business urgency”.
Finally, we noticed that Weak access to customer needs and / or business
information only appeared in the medium-sized clusters. For all medium-sized
organisations, including the ones with mixed process models, it was the third most
cited problem, with 13 citations, while it did not appear in the top 5 RE problems
for small- and large-sized organisations. We could not find any consistent patterns
in the free-text answers of the medium-sized companies. We can only speculate
that small-sized organisations might adapt themselves to fit the availability of their
customers, while large organisations might have more influence on their customers
to achieve the required access.
4.4 Cause-Eect Analysis (RQ 3)
In the following, we will summarise causes and eects as reported by our respon-
dents for the discussed RE problems. After selecting the five most critical RE
problems, we asked our respondents to provide what they believe to be the main
causes and eects for each of the problems. They provided the causes and eects
in an open question format, with one open question for the cause and another
for the eect for each the previously selected RE problems. Details on the data
analysis procedure can be taken from Section 3.4. In the following, we first discuss
the main causes for the RE problems, before discussing the causes and eects for
the top problems in detail.
4.4.1 Main Causes for RE Problems
In total, 177 of the 228 organisations that completed the survey provided textual
information for at least one cause and we received in total 820 textual answers
for causes and eects of RE problems. The coding process yielded 92 dierent
codes for causes of RE problems and 49 dierent codes for their eects. While
it does not make sense to analyse eects out of the context of the RE problems
that provoke them, causes are at the beginning of the causal system [7]. Thus,
an isolated view on causes (without consideration of their specific RE problem
context) may provide valuable information, for example, on how to prevent RE
problems in general. We therefore first provide a descriptive view on the most cited
causes of RE problems.
The ten most reported causes and how often they have been reported within
each of the analysed clusters are shown in Table 11. For the percentages, we
considered the total amount of organisations that completely answered the survey,
given that empty answers could mean that they did not consider any specific cause.
We can observe that the main reported causes of RE problems are Lack of time,
Lack of e xperience of RE team m embers,andWeak qualification of RE team mem-
bers. While none of the causes was cited by more than 20% of the organisations,
this figure changes within the specific clusters, where some causes were commonly
reported. The cause frequencies above 20% within each cluster are highlighted in
bold. What can also be seen, even if implicitly, are cycles in the causes and the
Naming the Pain in Requirements Engineering 21
Tabl e 1 1 Most cited causes of RE problems. Clusters yielding in more than 20% frequency
of the causes are highlighted.
Agile Agile Agile Plan- Plan- Plan-
driven driven driven
All Small Medium Large Small Medium Large
Cause 228 30 22 39 11 4 30
Lack of time 42 (18%) 2 (7%) 6 (27%) 3 (8%) 3 (27%) 0 (0%) 10 (33%)
Lack of experience of
RE team members
41 (18%) 5 (17%) 6 (27%) 4 (10%) 4 (36%) 0 (0%) 8 (27%)
Weak qualification of
RE team members
31 (14%) 1 (3%) 8 (37%) 1 (3%) 2 (18%) 2 (50%) 4 (13%)
Communication flaws
between project team
and the customer
30 (13%) 2 (7%) 2 (9%) 6 (15%) 3 (27%) 0 (0%) 7 (23%)
Requirements remain
too abstract
29 (13%) 4 (13%) 2 (9%) 5 (13%) 0 (0%) 0 (0%) 6 (20%)
Changing business
needs
21 (9%) 1 (3%) 2 (9%) 3 (8%) 1 (9%) 0 (0%) 0 (0%)
Customer does not
know what he wants
20 (9%) 3 (10%) 0 (0%) 8 (21%) 0 (0%) 1 (25%) 1 (3%)
Missing direct com-
munic ation to cu s-
tomer
18 (8%) 1 (3%) 4 (18%) 3 (8%) 1 (9%) 0 (0%) 0 (0%)
Language barriers 17 (7%) 0 (0%) 1 (5%) 3 (8%) 1 (9%) 0 (0%) 4 (13%)
Strict time schedule
by customer
16 (7%) 2 (7%) 0 (0%) 4 (10%) 0 (0%) 0 (0%) 2 (7%)
problems, i.e. some of the causes are, in fact, problems; for instance, Communi-
catio ns bet ween p roject team and the customer is given as one problem, but also
named by our respondents as a cause.
To analyse the influence of the most cited causes on the most cited problems
and, in turn, of those problems to project failure (as reported by the survey re-
spondents), we visualise the relationships via an alluvial diagram. This diagram is
shown in Figure 3. The decision to relate only the most cited causes to the most
cited RE problems was taken to enhance the visualisation.
Communication flaws between project team and the customer
Customer does not know what he wants
Lack of a well-defined RE process
Lack of experience of RE team members
Lack of time
Missing direct communication to customer
Requirements remain too abstract
Too high team distribution
Unclear roles and responsonsibilities at customer side
Weak qualification of RE team members
Communication flaws between project team and the customer
Communication flaws within the project team
Incomplete and / or hidden requirements
Inconsistent requirements
Insufficient support by customer
Moving targets (changing goals, business processes and / or requirements)
Stakeholders with difficulties in separating requirements from previously known solution designs
Time boxing / Not enough time in general
Underspecified requirements that are too abstract and allow for various interpretations
Weak access to customer needs and / or (internal) business information
Project Completed
Project Failed
Fig. 3 Relation of top 10 causes, top 10 problems, and the project impact.
22 D. M´endez Fern´andez et al.
As it appears, some of the ten most cited causes are more related to some spe-
cific problems than to others. Typical examples that can be seen are Lac k of time
leading mainly to the Time boxing problem,Lack of experience of R E team members
leading mainly to Incomplete and / or hidden requirements and Underspecified re-
quirements,orToo high team distribution leading mainly to Communication flaws
within the project team.
Concerning project failure, the occurrence of some RE problems seems to lead
very often to project failure. Out of these, we highlight Communication flaws be-
tween project team and the customer,Incomplete and / or hidden requirements,
Underspecified requirements,Communication flaws within the project team, and
Insucient support from the customer. However, in particular relating to project
failure, it is noteworthy that this diagram is based on a reduced dataset, con-
sidering only instances which contain one of the main causes, one of the main
problems, and a project impact. The complete data on how often each of the most
cited problems was related by the respondents to project failure is provided in
Table 9.
4.4.2 Causes and Eects of Top RE Problems
To provide a more complete view on the causes and eects reported for some of
the most critical RE problems, in particular, the three most cited ones (which are
also the three most cited ones for project failure), Incomplete and / or hidden
requirements,Communication flaws between us and the customer,andMoving
Targets, we built probabilistic cause-eect diagrams. Those diagrams have already
been applied in previous eorts in the NaPiRE context, based on data from Brazil
(see also Section 2.2). It is noteworthy to mention that, despite of the name of
those diagrams, within this paper we use them to represent relative frequencies,
i.e. how often each cause or eect was cited out of the total citations, and not
probabilities.
Figures 4 and 5 respectively show such cause-eect diagrams for the causes and
the eects of the Incomplete and / or hidden requirements problem. For instance,
in Figure 4, we can see that the most frequently cited causes were related to the
categories Input (34%, i.e. 31 out of 91 reported causes were from that category),
Method (33%), and People (29%). The five most frequent reported causes
for this problem are the Weak qualification (9%) and the La ck of experience
(9%) of the RE team members, Time pressure (5%), Stakeholders lacking
business vision and understanding (4%), the use of Poor requirements elicitation
techniques (4%), Specifying the requirements in an too abstract way (4%), and
Missing completeness checks (4%).
Concerning the eects of this problem, it can be seen in Figure 5 that the main
aected categories were Project or Organization (43%, i.e. 37 out of 87 reported
eects were from that category), Design or Implementation (23%), and Product
(21%). The most frequently cited causes were Time overrun (10%), Post
implementation rework (9%) and Poor product quality (9%).
The causes and eects of the Communication flaws between project team and
the customer problem are depicted in Figures 6 and 7. The prevailing cause cate-
gories for this RE problem are Method (38%, i.e. 30 out of 78 cited causes were
from that category) and Input (33%). The five most frequently reported causes
Naming the Pain in Requirements Engineering 23
Fig. 4 Probabilistic cause-eect diagram for Incomplete and / or hidden requirements fo-
cussing on the causes.
Fig. 5 Probabilistic cause-eect diagram for Incomplete and / or hidden requirements fo-
cussing on the eects.
for this problem are Inherent communication flaws (12%), Missing direct com-
munication (10%), Langua ge barriers (9%), Time pressure (6%), Missing
engagement by the customer (6%), and a Too high team distribution (6%).
In this case (Figure 7), the main aected categories were Project or Organiza-
tion (47%, i.e. 32 out of 68 eects were from that category), Product (22%),
and Customer (19%). The main cited eects for this problem were Customer dis-
satisfaction (16%), Time overrun (13%), and Poor product quality (10%).
24 D. M´endez Fern´andez et al.
Fig. 6 Probabilistic cause-eect diagram for Communication flaws between project team and
customer focussing on the causes.
Fig. 7 Probabilistic cause-eect diagram for Communication flaws between project team and
customer focussing on the eects.
Finally, Figures 8 and 9 show the probabilistic cause-eect diagram for the
causes and the eects of Moving Targets. As shown in Figure 8, the causes of this
problem are heavily concentrated on the Input category (66%, i.e. 40 out of 61
cited causes were from that category). Its main causes were coded to Changing
business needs (15%), Customers who do not know what they want (13%),
Vol a t i l e i n d u s t r y s egment s t h a t l ead to changes (10%), Poor project management
(7%), and Weak management at the customer side (5%).
Again, as expected for one of the most cited RE problems, the eects are severe
(Figure 9). Most of the eects are concentrated in the Project or Organization
(68%, i.e. 38 out of 56 cited eects were from that category) category, which
Naming the Pain in Requirements Engineering 25
Fig. 8 Probabilistic cause-eect diagram for Moving targets focussing on the causes.
might explain why this problem has such a strong relation to project failure. In
fact, 51% of the organisations that cited Moving Targets as a problem stated that
it led to project failure. In this case, the eect Time overrun is clearly the most
cited one (27%), followed by Budget overrun (13%), and Overall demotivation
(7%).
Fig. 9 Probabilistic cause-eect diagram for Moving targets focussing on the eects.
26 D. M´endez Fern´andez et al.
5 Conclusions
In this article, we contributed the analysis of contemporary problems practitioners
experience in their industrial project setting. To this end, we relied on the NaPiRE
initiative (http://www.re-survey.org), a global family of bi-yearly replicated sur-
veys where the aim is to overcome the problem of by now isolated investigations in
RE that are not yet representative. Our analysis of contemporary problems uses
data provided by 228 companies spread over 10 countries and included an investi-
gation of problems practitioners experience, what the causes of those problems are,
and how the problems manifest themselves in the process going beyond simplified
views on project failure.
5.1 Discussion of Results
In this section, we discuss our research questions based on the obtained survey
results and their potential implications.
Problems in RE (RQ 1) Our first research question concerned the contemporary
problems in RE. We identified and ranked the problems cited as the most critical
ones by the 228 organisations that completed the survey. This result reflects the
contemporary opinion of organisations spread throughout ten dierent countries,
of dierent sizes and using dierent process models. We believe that this result
provides further insights into industrial RE problem trends and that it helps to
lay the foundation to steer academic and industrial research in a problem-driven
manner where scientific contributions to RE can be put in tune with practically
relevant problems. Out of the identified problems, we highlighted Incomplete and
/orhiddenrequirements,Communication flaws between project team and the cus-
tomer,andMoving targets, which were the most cited ones and also the ones
mostly related to project failure.
Problem Patterns (RQ 2) The second research question relates to identifying pat-
terns between problems and context characteristics. Of course, there are several
ways of blocking our results that could have been performed for this analysis. In
this initial eort, we focused on the type of process model (agile or plan-driven)
and on the organisational size, blocking those clusters of survey responses. Within
these blocks, it was already possible to observe some relevant behavioural dier-
ences. For instance, considering the three most cited problems, Incomplete and/or
hidden requirements appears in all clusters, while Communication flaws between
project team and the customer does not seem to be a major problem for small
plan-driven organisations and Moving targets occurs mainly in plan-driven and in
large organisations. Future work includes investing additional eort relating the
problems to other relevant constructs of the underlying NaPiRE theory. Based
on our initial observations, we believe that specific advice to organisations with
dierent characteristics on how to prevent critical RE problems could be valuable
beyond the type of advice available in current guidelines and maturity models.
Naming the Pain in Requirements Engineering 27
Cause-Eect Analysis (RQ 3) The third research question concerned the causes
and the eects as they have been perceived by our respondents. We identified
the most reported causes and analysed their influence on the most critical RE
problems. We could observe that the causes tend to dier within the selected
blocks and that some of the ten most cited causes have more influence on some
specific problems than on others.
Additionally, and still in the context of RQ 3, we analysed the causes and eects
related to the three most critical RE problems. We believe that the identification
of the causes can already help organisation to focus their prevention eorts. For
instance, we identified that, in general, the main reported causes for Incomplete
and/or hidden requirements are the Weak qualification and the Lack of experience
of the RE team members, Time pressure,Stakeholders without business vision,
Poor elicitation techniques,Too abstract specifications,andMissing completeness
checks. Based on this information, an organisation facing this or similar problems
could take first counter measures, such as:
1. Checking on the qualification and experience of its team members, providing
training if needed, in particular, on avoiding abstract specifications. This could
also be supported by including and training RE standards that put emphasis
on the way requirements should be elicited and specified.
2. Adjusting its portfolio management to avoid accepting projects under extreme
time pressure or involving stakeholders that lack business vision.
3. Assessing and improving the eciency of their elicitation techniques.
4. Improving their completeness check, within the philosophy of their develop-
ment paradigm. Plan-driven organisations, for instance, could institutionalise
requirements inspections based on RE standards for the artefacts, while agile
organisations could introduce the Definition of Ready (DoR) practice, which
is commonly used in agile projects to avoid the beginning of work on features
that do not comply with clearly defined completion criteria.
Future work in this direction includes setting up a knowledge base on typical
causes of RE problems and on actions that could be taken to mitigate or prevent
them (success factors). However, we reinforce that these are informal suggestions of
the authors based on the identified causes and that our analyses need to be backed
up by complementary investigations, ideally also by applying dierent empirical
research methods on project data (e.g., case studies and experiments). Moreover,
organisations should perform in depth causal analysis in their projects to assure
addressing the right causes, the ones that are really happening in their concrete
context.
5.2 Relation to Existing Evidence
In this section, we relate the results of this article to evidence from previous
NaPiRE studies and other related RE surveys presented in Section 2.
Previous NaPiRE Evidence The first NaPiRE run with data from 58 respondents
from Germany [28] provided a very similar picture of the top problems as we found
here. The by far most cited problem in both was surveys incomplete / hidden
requirements.Moving targets,time boxing and underspecified requirements are too
abstract being in the top 5 in both surveys. Diculty of separating requirements
28 D. M´endez Fern´andez et al.
from known solutions,inconsistent requirements and co mmu nica tion flaws in the
team occurred in both top tens with a slightly dierent ranking. Most interesting
is commu nica tio ns flaws between team and customer . It moved from rank 6 in
the first run to rank 2 in second run. Furthermore, the first run included missing
traceability and gold plating (ranks 9 and 10) which were replaced in the second
run by insucient support by customer and weak access to customer needs (ranks
8 and 10). Hence, the extension to other countries and a larger sample emphasised
mainly customer-related problems. This suggests that these problems are not so
prevalent in Germany. Analysing the data from the second run in this respect
somewhat supports this: The problems insucient support by customer and weak
access to customer needs are mentioned less often than in most other countries. A
reason could be that more often than in other countries, the team and the customer
are geographically close and speak the same language. Yet, communication fla ws
betwee n team and customer are also very prevalent in the German data of the
second run.
In [18], the the Brazilian data set was analysed w.r.t. the cited problems, causes
and eects. The set of top problems in Brazil matches exactly the top problems
in the global data set. The ranking of the problems diers only slightly. There are
dierences in the causes and eects, but the general categories are consistent with
some variation in the weights.
The data sets from Brazil and Germany were also compared separately [31].
The top five problems from both countries are in the top 8 of the global data set.
Also all other results are very similar to the results from the full data set.
Finally, the data sets from Brazil and Austria was used to investigate in detail
the incomplete / hidden requirements problem [19]. The causes found there are
also included in the results of this article. The distribution over the categories of
causes changed slightly from a strong focus on people (40%) to input (34%) and
method (33%). Yet, the results of this article contain far more causes than we
found in [19].
Furth e r Existing Evidence Our results corroborate and extend existing findings
reported by other researchers. For example, the German Success study [6] investi-
gated factors that influence project success in general, not limited to requirements
engineering. Most of the factors are consequently more abstract than ours. How-
ever, the study found that project success is independent of the degree of manage-
ment support. We can support this to a certain extent in the sense that missing
management support was not often perceived as a problem.
Kamata et al. [23] investigate relations between requirements quality and
project success or failure, which aimed at a more limited view than we took as the
requirements specification was in scope of their investigation rather then the whole
RE process. Therefore, the results are not directly comparable. Nevertheless, their
finding that a relatively small set of requirements has strong impact on project
success or failure is relevant for our top RE problem of incomplete and / or hidden
requirements, because it further substantiates the problem of single missing and /
or hidden requirements.
Nikula et al. [33] find in their state of practice survey on requirements engineer-
ing in small- and medium-sized companies that completeness, change management
and descriptions are the three most needed techniques to further develop RE in
the participating companies. These three top needs are directly related to our
Naming the Pain in Requirements Engineering 29
three top RE problems and, thus, further underpin their high rating in our sur-
vey and the relevance of the provided cause-eect analysis for these problems in
Section 4.4.2.
Solemon et al. [37] report requirements engineering problems and practices in
software companies. Diering from our study, the focus of the researchers was to
identify RE problems rather than relating them to possible causes. They classified
RE problems into organisational and RE process related ones. The reported main
organisational problems are lack of customer and user communication problem,
lack of developer communication, as well as poor time and resources allocation.
The main problems in the RE process are related to changing requirements, incom-
plete requirements, ambiguous requirements and poor user understanding. With
regard to the mentioned communication, resource, change, completeness and un-
derstandability issues, our survey supports their findings, and additionally provide
a more fine-grained distinction and relationships between causes and eects.
Liu et al. [25] present results of a survey on why requirements engineering fails.
According to that survey, major failure reasons are an unclear understanding of
the system by the customer, constant change of user needs and understanding,
missing access to domain knowledge for software engineers, reuse of existing design
in wrong context and environment, lack of domain and technical expertise for RE
decision makers, tight project schedule, broken communication links, as well as
lack of standardised data and interface definitions. The authors do not clearly
distinguish between problems and their causes, but all listed major failure reasons
are related to our main RE problems and their causes.
Verner et al. [40] further ran a survey in Australia and the USA. They con-
centrated on success factors in RE and found good requirements, customer/user
involvement, and eective requirements management to be the best predictors of
project success. Disregarding the diculties to precisely capture what “good re-
quirements” are, our results still can be considered in tune with their observations
as we identified problems and their causes which (if negated) can be used to refine
the abstract success factors identified by Verner et al. For instance, we identified
incomplete and/or hidden requirements as a main problem and weak qualification
as well as lack of experience as its main causes.
Al-Rawas and Easterbrook [2] finally present a field study on communication
problems in requirements engineering. The results show that organisational issues
have great influence on the eectiveness of communication, and furthermore that
in general users find the notations used by software practitioners to model their
requirements dicult to understand and validate. The topic of the study shows
that communication is an important RE problem, which is also reflected by our
results. Also the presented results are in tune with the causes for communication
flaws that we present.
5.3 Impact/Implications
Our findings complement existing evidence on problems in RE in various ways.
First, we could distill a detailed picture of problems practitioners experience in
their project setting including a rich analysis of eects going beyond project failure.
Second, the analysis of the causes and eects in dependency to context factors
allows us to steer a first empirically founded discussion on phenomena that hold for
30 D. M´endez Fern´andez et al.
particular contexts. Third, and most importantly, revealing not only the problems,
but also their causes, allows us to get a first picture of requirements engineering
success factors which, if met, should mitigate the problems.
Based on this analysis, we can already steer further problem-driven research,
for example on agile requirements engineering. It also allows us to further explore
RE success factors suitable to establish maturity models and RE improvement
endeavours that are grounded on empirical data.
Overall, our contribution does not only support researchers to steer their re-
search, but also practitioners to evaluate their own current RE situation against
overall industrial trends presented here.
5.4 Limitations
We have analysed the results from the survey conducted in ten countries yielding
a broad population, while applying the dierent procedures to control the validity
as described in section 3.5. Still, we are aware that our study has limitations. Most
importantly, our results are based on a reasonable but still limited number of re-
spondents. Also, the responses are not equally distributed over business domains
and families of systems (such as embedded systems). This probably has an influ-
ence on the rankings of the problems if specific domains have specific problems.
We can, for example, not say reliably how the picture would change if considering
safety-critical systems only, let alone as they appear less frequently in our data
than information systems. We can only assume that the picture would change for
practices that tend to be seen as more important for the development of highly
dependable, safety-critical systems than for business information systems (such as
traceability). Furthermore, we cannot make concrete statements about how gen-
eralisable the results eventually are, because we still are not able to estimate the
representativeness of our population (given the unavailability of empirical data
characterising all industry segments in every considered country). This means in
consequence that we cannot generalise going beyond the contexts described and
that we might even expect partially dierent results in dierent countries. There-
fore, we need to follow our design of a family of surveys and further steer the
continuous replications and syntheses of the results while capturing precisely the
context to establish a more reliable and empirically solid theory.
Furthermore, inherent to survey research is that surveys can only reveal stake-
holders’ perceptions on current practices rather than empirically backed-up knowl-
edge about those practices. To some extent, we aim at revealing exactly those
perceptions. However, the answers given by our respondents might still be biased.
We mitigated this threat by conducting the survey anonymously, but need to back
up (and further explore) the insights revealed via this survey by adding more
investigations using other empirical methods, e.g., interviews and case studies.
A further limitation arises from the qualitative data and its analysis. Manual
coding of qualitative data is by nature a creative process where experiences, ex-
pectations and the expertise of the coders influence the results. We mitigated this
threat to a certain extent by applying researcher triangulation, yet the picture
might still change if involving other researchers.
Finally, a major threat arises from the way we designed our instrument as it
still includes too many (to a certain extent closed) questions which might enforce
Naming the Pain in Requirements Engineering 31
too simplified views by the respondents on the particularities of their project en-
vironments. We asked our respondents to categorise their software process model
used based on a pre-defined list of options. This means that although the respon-
dents might have selected, for instance, Scrum, we still have no guarantee that the
process model followed was indeed Scrum or even agile. It could have also been a
plan-driven, iterative model baptised as “Scrum” in that organisational context.
That is, respondents might have misinterpreted the options or they might have
relied on a dierent understanding on the process model used (or even on how
the process model should be rather than how it actually is), let alone because of
the fuzzy notion of “agile” that allows for too many variations. So far, we can-
not mitigate the threats arising from this circumstance, but use the lessons we
learnt in this replication to further foster our learning curve during the follow-up
re-redesign of our instrument. Similarly, the part of our instrument used to reveal
problems relies, from a conceptual point of view, on a simplified understanding
of causes and eects. We realised that such a simplified view on phenomena and
their sequential cause-eect dependencies is not sucient to capture the complex
dependencies of the various factors in a project; we need, in fact, to extend our
instrument with means that allow to capture complex systems of phenomena as
recommended in context of system theory (and system thinking respectively). The
re-design of our instrument therefore includes, inter alia, a more indirect approach
that relies on triangulation from multiple questions to support richer and more
robust results (see also our next section).
5.5 Future Work
We are currently analysing the overall data obtained from the current NaPiRE
survey round with respect to the responses on the status quo in RE. This current
analysis complements the investigation presented here with a first holistic theory
on the current status quo in RE. Another work already in progress is the usage
of results from the cause-eect analysis to establish a maturity model for RE
including practical recommendations on how to improve RE in response to certain
context-specific situations.
Besides current analyses of the survey data and subsequent work based on the
data, we are planning the next survey round to be started in 2017. To this end, we
will first revise our instrument to better mitigate the limitations discussed above.
This re-design includes:
A richer set of questions to describe the context of our respondents including
a more specific focus on teams and pro ject environments rather the overall
organisation.
A more dierentiated view on the particularities of the projects and the process
models used including a more indirect approach that relies on triangulation
from multiple, more objective questions on the practices actually used.
A more dierentiated view relying on more open questions following an ap-
proach as recommended by the 5-why methodology to capture the complex
system of dependent phenomena rather than relying of single causes and ef-
fects of a set of pre-compiled problems.
32 D. M´endez Fern´andez et al.
Given that NaPiRE is a community eort aimed at bringing forward the em-
pirical RE community, we cordially invite researchers and practitioners to join us
in our follow-up activities to steer future replications of our family of surveys.
5.6 Acknowledgements
We want to thank all members of ISERN who supported the design of this research
initiative. Furthermore, we want to thank all our contacts from industry for their
participation in the survey. Our gratitude also goes to the reviewers whose con-
structive feedback supported us improving our manuscript and in particular the
further re-design of our instrument for the follow-up replications. Finally, Dietmar
Pfahl was supported by the Estonian Research Council.
References
1. Adolph S, Hall W, Kruchten P (2011) Using Grounded Theory to study the
Experience of Software Development. Journal of Empirical Software Engineer-
ing 16(4):487–513
2. Al-Rawas A, Easterbrook S (1996) Communication problems in requirements
engineering: a field study. National Aeronautics and Space Administration
3. Beecham S, Hall T, Rainer A (2003) Software process improvement problems
in twelve software companies: An empirical analysis. Empirical software engi-
neering 8(1):7–42
4. Birks, M and Mills, J (2011) Grounded Theory - A Practical Guide. Sage
Publications, Inc.
5. Brand A, Allen L, Altman M, Hlava M, Scott J (2015) Beyond Author-
ship: Attribution, Contribution, Collaboration, and Credit. Learned Publish-
ing 28:151–155
6. Buscherm¨ohle R, EekhoH, Josko B (2006) Success – Erfolgs- und Misser-
folgsfaktoren bei der Durchf¨uhrung von Hard- und Softwareentwicklungspro-
jekten in Deutschland. ISBN-13 978-3-8142-2035-2, BIS-Verlag der Carl von
Ossietzky Universit¨at Oldenburg
7. Card, DN (2005) Defect Analysis: Basic Techniques for Management and
Learning. In: Advances in Computers, vol 65, pp 259–295
8. Cheng, BHC and Atlee, JM (2007) Research Directions in Requirements En-
gineering. In: Future of Software Engineering (FOSE’07), IEEE Computer
Society, pp 285–303
9. Condori-Fernandez N, Daneva M, Sikkel K, Wieringa R, Dieste O, Pastor
O (2009) A systematic mapping study on empirical evaluation of software
requirements specifications techniques. In: Proceedings of the 2009 3rd Inter-
national Symposium on Empirical Software Engineering and Measurement,
IEEE Computer Society, pp 502–505
10. Condori-Fern´andez N, Daneva M, Wieringa R (2012) Preliminary Survey on
Empirical Research Practices in Requirements Engineering. Tech. Rep. TR-
CTIT-12-10, University of Twente, Centre for Telematics and Information
Tech n o logy (CTI T )
Naming the Pain in Requirements Engineering 33
11. Cox K, Niazi M, Verner J (2009) Empirical Study of Sommerville and Sawyer’s
Requirements Engineering Practices. IET Software 3(5):339–355
12. Damian D, Chisan J (2006) An Empirical Study of the Complex Relationships
between Requirements Engineering Processes and other Processes that lead to
Payos in Productivity, Quality, and Risk Management. IEEE Transactions
on Software Engineering 32(7):433–453
13. Davis A, Dieste O, Hickey A, Juristo N, Moreno AM (2006) Eectiveness of re-
quirements elicitation techniques: Empirical results derived from a systematic
review. In: Requirements Engineering, 14th IEEE International Conference,
IEEE, pp 179–188
14. Eveleens J, Verhoef T (2010) The Rise and Fall of the Chaos Report Figures.
IEEE Software 27(1):30–36
15. Glaser B, Strauss A (1967) The Discovery of Grounded Theory: Strategies for
Qualitative Research. Aldine Transaction
16. Hsia P, Davis A, Kung D (1993) Status report: Requirements engineering.
IEEE Softw 10(6):75–79
17. IEEE (1998) IEEE Recommended Practice for Software Requirements Spec-
ifications – IEEE Std 830-1998. Technical Standard IEEE Std 830-1998, The
Institute of Electrical and Electronics Engineers, Inc.
18. Kalinowski M, Spinola R, Conte T, Prikladnicki R, Mendez Fernandez D, Wag-
ner S (2015) Towards building knowledge on causes of critical requirements
engineering problems. In: Proc. 27th International Conference on Software
Engineering and Knowledge Engineering (SEKE)
19. Kalinowski M, Felderer M, Conte T, Spinola R, Prikladnicki R, Winkler D,
Mendez Fernandez D, Wagner S (2016) Preventing incomplete/hidden require-
ments: Reflections on survey data from austria and brazil. In: Proc. Software
Quality Days 2016
20. Kalinowski, M and Card, DN and Travassos, GH (2012) Evidence-Based
Guidelines to Defect Causal Analysis. IEEE software 29(4):16–18
21. Kalinowski, M and Mendes, E and Travassos, GH (2011) Automating and
Evaluating the Use of Probabilistic Cause-Eect Diagrams to Improve De-
fect Causal Analysis. In: Proc. International Conference on Product Focused
Software Process Improvement
22. Kalinowski, M and Travassos, GH and Card, DN (2008) Towards a Defect
Prevention Based Process Improvement Approach. In: Proc. Euromicro Con-
ference on Software Engineering Advanced Applications
23. Kamata M, Tamai T (2007) How Does Requirements Quality Relate to Project
Success or Failure? In: Proceedings of the 15th International Conference on
Requirements Engineering, IEEE Computer Society, pp 69–78
24. Lemos J, Alves C, Duboc L, Rodrigues GN (2012) A systematic mapping
study on creativity in requirements engineering. In: Proceedings of the 27th
Annual ACM Symposium on Applied Computing, ACM, pp 1083–1088
25. Liu L, Li T, Peng F (2010) Why requirements engineering fails: A survey
report from china. In: Requirements Engineering Conference (RE), 2010 18th
IEEE International, IEEE, pp 317–322
26. endez Fern´andez D, Wagner S (2013) Naming the Pain in Requirements
Engineering – NaPiRE Report 2013. Tech. Rep. TUM-I1326, Technische Uni-
versit¨at M¨unchen
34 D. M´endez Fern´andez et al.
27. endez Fern´andez D, Wagner S (2013) Naming the Pain in Requirements
Engineering: Design of a global Family of Surveys and first Results from Ger-
many. In: Proc. 17th International Conference on Evaluation and Assessment
in Software Engineering (EASE 2013), ACM
28. endez Fern´andez D, Wagner S (2014) Naming the Pain in Requirments En-
ginering: A Design for a global Family of Surveys and First Results from Ger-
many. Information and Software Technology - DOI: 101016/jinfsof201405008
57:616–643
29. endez Fern´andez D, Wagner S, Lochmann K, Baumann A, de Carne H (2012)
Field Study on Requirements Engineering: Investigation of Artefacts, Project
Parameters, and Execution Strategies. Information and Software Technology
54(2):162–178
30. endez Fern´andez D, Mund J, Femmer H, Vetr`o A (2014) In Quest for Re-
quirements Engineering Oracles: Dependent Variables and Measurements for
(good) RE. In: Proc. 18th International Conference on Evaluation and Assess-
ment in Software Engineering (EASE 2014), ACM
31. Mendez Fernandez D, Wagner S, Kalinowski M, Schekelmann A, Tuzcu A,
Conte T, Spinola R, Prikladnicki R (2015) Naming the Pain in Requirements
Engineering: Comparing Practices in Brazil and Germany. IEEE Software
32(5):16–23
32. Napier N, Mathiassen L, Johnson R (2009) Combining Perceptions and Pre-
scriptions in Requirements Engineering Process Assessment: An Industrial
Case Study. IEEE Transactions on Software Engineering 35(5):593–606
33. Nikula U, Sajaniemi J, K¨alvi¨ainen H (2000) A State-of-the-practice Survey on
Requirements Engineering in Small-and Medium-sized Enterprises. Research
Report 951-764-431-0, Telecom Business Research Center Lappeenranta
34. Nuseibeh B, Easterbrook S (2000) Requirements Engineering: A Roadmap. In:
Proceedings of the Conference on the Future of Software Engineering, ACM,
New York, NY, USA, pp 35–46
35. Pekar V, Felderer M, Breu R (2014) Improvement methods for software re-
quirement specifications: A mapping study. In: Quality of Information and
Communications Technology (QUATIC), 2014 9th International Conference
on the, IEEE, pp 242–245
36. Pettersson F, Ivarsson M, Gorschek T, ¨
Ohman P (2008) A practitioner’s guide
to light weight software process assessment and improvement planning. Jour-
nal of Systems and Software 81(6):972–995
37. Solemon B, Sahibuddin S, Ghani AAA (2009) Requirements engineering prob-
lems and practices in software companies: An industrial survey. In: Advances
in Software Engineering, Springer, pp 70–77
38. Sommerville I, Sawyer P (1997) Requirements Engineering - A Good Practice
Guide, 1st edn. ISBN-13: 978-0471974444, John Wiley and Sons, Inc.
39. Staples M, Niazi M, Jeery R, Abrahams A, Byatt P, Murphy R (2007) An ex-
ploratory study of why organizations do not adopt CMMI. Journal of Systems
and Software 80(6):883–895
40. Verner J, Cox K, Bleistein S, Cerpa N (2007) Requirements engineering and
software project success: an industrial survey in australia and the us. Aus-
tralasian Journal of information systems 13(1):225–238
41. Wagner, S and Mendez Fernandez, D (2015) Analysing Text in Software
Projects. In: Bird, C and Menzies, T and Zimmermann, T (ed) The Art and
Naming the Pain in Requirements Engineering 35
Science of Analyzing Software Data, Morgan Kaufmann
AListofContributorsandRoles
Table 12 introduces the details of the authorships with respect to the roles taken in the NaPiRE
project and in context of this particular manuscript. To this end, we introduce a role concept
based on the classification scheme as discussed by Brand et al. in [5]. In our context, we
distinguish the following roles3:
Conceptualisation: Ideas; formulation or evolution of overarching research goals and aims.
Project Administration: Management and coordination responsibility for the research
activity planning and execution.
Methodology: Development or design of methodology; creation of models.
Instrument Design: Development / re-design of the instrument used in this replication.
Data Collection: Data collection as national representative in the respective country using
the provided infrastructure.
Data Analysis: Application of textual analysis techniques to study and interpret the data.
Data Curation: Management activities to annotate (produce metadata), scrub data and
maintain research data (including software code, where it is necessary for interpreting the
data itself) for initial use and later reuse.
Data Visualisation: Preparation, creation and/or presentation of the published work,
specifically visualisation/ data presentation.
Wri t i n g - O r i g i nal Draft : Preparation, creation and/or presentation of the published work,
specifically writing the initial draft (including substantive translation).
Data - Review & Editing: Preparation, creation and/or presentation of the published
work by those from the original research group, specifically critical review, commentary or
revision including pre- or post-publication stages
The first two authors are the initiators and coordinators of the overall initiative. Further,
we formed a group of lead authors for this particular manuscript (the first six authors) to do
the data analysis and / or the writing process.
3Those roles marked with an are the exact same as introduced in [5].
36 D. M´endez Fern´andez et al.
Tabl e 1 2 Authorship details: Roles and contribution.
Author
Conceptualisation
Project Administration
Methodology
Instrument Design
Data Collection
Data Analysis
Data Curation
Data Visualisation
Writ i n g - D r a f t
Writ i n g - R e v i e w & E d i t i n g
D. endez Fern´andez X X X X X X X X X X
S. Wagner X X X X X X X X
M. Kalinowski X X X X X X X
M. Felderer X X X X X X
P. Ma f r a X X
A. Vetr`o XX
T. Conte X X X
M.-T. Christiansson X X
D. Greer X X X
C. Lassenius X X X
T. M¨annist¨o X X X
M. Nayabi X X
M. Oivo X X X
B. Penzenstadler X X
D. Pfahl X X X
R. Prikladnicki X X X
G. Ruhe X X
A. Schekelmann X X
S. Sen X X
R. Spinola X X X
A. Tuzcu X X
J. L. de la Vara X X
R. Wieringa X X X
... In real-world programming contexts, requirements are often conveyed in natural language, which is prone to ambiguity, misinterpretation, and the need for clarification [13,36]. Failure to clarify ambiguities early can be costly in software development [10]. Thus, identifying gaps and clarifying missing details in specifications are essential practices for developers. ...
Preprint
Introductory programming courses often rely on small code-writing exercises that have clearly specified problem statements. This limits opportunities for students to practice how to clarify ambiguous requirements -- a critical skill in real-world programming. In addition, the emerging capabilities of large language models (LLMs) to produce code from well-defined specifications may harm student engagement with traditional programming exercises. This study explores the use of ``Probeable Problems'', automatically gradable tasks that have deliberately vague or incomplete specifications. Such problems require students to submit test inputs, or `probes', to clarify requirements before implementation. Through analysis of over 40,000 probes in an introductory course, we identify patterns linking probing behaviors to task success. Systematic strategies, such as thoroughly exploring expected behavior before coding, resulted in fewer incorrect code submissions and correlated with course success. Feedback from nearly 1,000 participants highlighted the challenges and real-world relevance of these tasks, as well as benefits to critical thinking and metacognitive skills. Probeable Problems are easy to set up and deploy at scale, and help students recognize and resolve uncertainties in programming problems.
... Natural language is commonly used to express requirements, because it can be understood by nearly all stakeholders [31]. The use of unstructured natural language phrasing has been shown, however, to have negative impacts on the quality and utility of resulting requirements, e.g., vagueness, inconsistency, ambiguity, duplication, errors of omission, and a lack of testability [43,44]. To address this, researchers have developed and evaluated a variety of structured natural language (SNL) methods for expressing requirements [26,29,31], and applied such approaches to domainspecific expression of functional requirements [27,28]. ...
Preprint
Full-text available
Deep neural network (DNN) testing is crucial for the reliability and safety of critical systems, where failures can have severe consequences. Although various techniques have been developed to create robustness test suites, requirements-based testing for DNNs remains largely unexplored -- yet such tests are recognized as an essential component of software validation of critical systems. In this work, we propose a requirements-based test suite generation method that uses structured natural language requirements formulated in a semantic feature space to create test suites by prompting text-conditional latent diffusion models with the requirement precondition and then using the associated postcondition to define a test oracle to judge outputs of the DNN under test. We investigate the approach using fine-tuned variants of pre-trained generative models. Our experiments on the MNIST, CelebA-HQ, ImageNet, and autonomous car driving datasets demonstrate that the generated test suites are realistic, diverse, consistent with preconditions, and capable of revealing faults.
... Balancing these and ensuring that all stakeholders' needs are adequately addressed is important. • Compatibility with current systems -Ensuring compatibility and coherence between new emerging and existing requirements can be complex and a careful analysis and classification of requirements is required [20]. • Insufficient customer requirements -Customers may provide incomplete requirements, making it difficult to classify them accurately [21]. ...
Chapter
The construction of a computer-based system (CBS) begins ideally with the elicitation of its requirements and the writing of a specification of its requirements, usually in the form of a natural-language (NL) requirements specifications (RS). This chapter begins by describing (1) three kinds of NL RSs: software requirements specification, a.k.a. SRSs, user manuals, and user stories, and (2) the qualities, such as correctness and lack of ambiguity, that they as NL RSs must have to serve their purpose of adequately specifying the requirements of a CBS. The chapter then defines the kinds of defects that an RS, being written in an NL, may suffer, with an emphasis on ambiguity as the kind of defect arising specifically from the use of NL to write the RS. The chapter describes how some defect-detection tools work. It observes that whether an NL RS has any of these kind of defects is fundamentally algorithmically undecidable. However, searching for instances of a finite number of indicators of some kinds of defects, such ambiguity or vagueness, is feasible with 100% recall. While an RS may not have such a defect without the presence of one its indicators, not every occurrence of an indicator of a defect is part of an instance of the defect. The chapter reviews efforts to empirically evaluate a defect-detection tool’s (1) effectiveness: how well a defect-detection tool detects the defects in its scope, often expressed in terms of recall and precision, and (2) usefulness: whether detecting these defects in an RS positively affects the quality of the CBS that is constructed from the RS. The chapter observes that the construction of many a defect-detection tool is accompanied by an empirical demonstration of its effectiveness, but not by any demonstration of its usefulness. Moreover, each of three attempts to empirically demonstrate usefulness of abstraction finding failed; neither did any of the found ambiguities cause serious problems nor were any of the serious problems caused by ambiguities. Instead, in each case, as a result of a robust requirements analysis process, all the stakeholders happened to agree on the meaning of every considered ambiguity. The chapter concludes by wondering whether any defect-detection tool is useful.
Article
Full-text available
Ambiguity in software requirements is a significant challenge as it often leads to misunderstandings, implementation errors, and costly project delays. This research proposes a hybrid framework that combines rule-based techniques with machine learning to identify ambiguity in software requirements with precision and efficiency. The framework begins with a rule-based model that systematically detects ambiguities using a carefully prepared list of ambiguous phrases. The analysis utilizes a dataset of 1,553 software requirements drawn from diverse project domains. To capture more intricate ambiguities that traditional rule-based systems might miss, the framework integrates TF-IDF vectorization and a Random Forest classifier, enhancing the precision and coverage of classification. In addition, clustering analysis identifies patterns to provide deeper insights into ambiguous requirements, while sentiment analysis explores the relationship between ambiguity and the emotional tone of requirements. Together, these analyses offer a broader understanding of ambiguity trends and stakeholder perceptions. The framework’s performance is validated using standard evaluation metrics, achieving an accuracy of 97%, precision of 97%, recall of 89%, and an F1-score of 92%, significantly surpassing traditional rule-based methodologies. This research advances automatic ambiguity detection by delivering a flexible and interpretable solution. The proposed approach enhances the clarity and quality of software requirements, strengthening requirements engineering practices and supporting more effective software development.
Article
Context: The Internet of Things (IoT) involves heterogeneous devices that interact and process data via the Internet. In the development of IoT systems, requirement elicitation is crucial. However, challenges such as heterogeneity, interoperability, scalability, and requirements volatility necessitate new approaches or adapting traditional techniques. Objective: In this context, this work proposes the Sensorina Map and IoT Mind as techniques adapted from the Empathy Map and Mind Map, respectively, to support requirement elicitation in IoT systems. Method: Two empirical studies were conducted in an academic environment to assess the feasibility of the techniques, then, a case study in industry environment. Results: The first study analyzed the ease of use and evaluated if it assisted software engineers in remembering the system requirements. The participants’ perceptions were collected through a Focus Group, refining the techniques. Subsequently, an observational study evaluated the techniques’ usefulness and ease of use. The results of the study demonstrated that the participants considered the methods feasible. The case study results revealed that the Sensorina Map is more suitable for advanced stages. At the same time, the Mind IoT suits better the initial phases, emphasizing the need for practical examples and adaptations to suit diverse user profiles. Conclusion: This work is expected to advance research in IoT systems and benefit professionals and researchers in this area.
Chapter
In this chapter, we provide advice on how to effectively teach survey research based on lessons learned from several international teaching experiences on the topic and from conducting large-scale surveys published at various scientific conferences and journals. First, we provide teachers with a potential syllabus for teaching survey research, including learning objectives, lectures, and examples of practical assignments. Thereafter, we provide actionable advice on how to teach the topics related to each learning objective, including survey design, sampling, data collection, statistical and qualitative analysis, threats to validity and reliability, and ethical considerations. The chapter is complemented by online teaching resources, including slides covering an entire course.
Article
Full-text available
As part of the Naming the Pain in Requirements Engineering (NaPiRE) initiative, researchers compared problems that companies in Brazil and Germany encountered during requirements engineering (RE). The key takeaway was that in RE, human interaction is necessary for eliciting and specifying high-quality requirements, regardless of country, project type, or company size.
Article
Full-text available
Software Requirement Specifications (SRS) are a key result of the Requirement Engineering (RE) process and an important basis for every large industrial software development project. Nevertheless, accurate, complete and consistent SRS are still a challenge in practice. Many publications related to SRS address problems and provide solutions for solving these. The most frequently researched SRS problems and improvement methods are listed and referenced in this mapping study. As final result of the search process, publication are analyzed and mapped to each other. This mapping is important for a further evaluations of SRS improvement methods that helps practitioners and researchers to compare those approaches.
Article
Full-text available
As part of the Naming the Pain in Requirements Engineering (NaPiRE) initiative, researchers compared problems that companies in Brazil and Germany encountered during requirements engineering (RE). The key takeaway was that in RE, human interaction is necessary for eliciting and specifying high-quality requirements, regardless of country, project type, or company size.
Conference Paper
Full-text available
[Context] Many software projects fail due to problems in Requirements Engineering (RE). [Objective] The goal of this paper is to gather information on relevant RE problems and to represent knowledge on their most common causes. [Method] We replicated a global family of RE surveys in Brazil and used the data to identify critical RE problems and to build probabilistic cause-effect diagrams to represent knowledge on their common causes. [Results] The survey was answered by 74 different organizations, including small, medium and very large sized companies, conducting both, plan-driven and agile development. The most critical RE problems, according to those organizations, are related to communication and to incomplete or underspecified requirements. We provide the full probabilistic cause-effect diagrams with knowledge on common causes of the most critical identified RE problems online. [Conclusion] We believe that the knowledge presented in the diagrams can be helpful to support organizations in conducting causal analysis sessions by providing an initial understanding on what usually causes critical RE problems.
Article
Full-text available
• As the number of authors on scientific publications increases, ordered lists of author names are proving inadequate for the purposes of attribution and credit. • A multi-stakeholder group has produced a contributor role taxonomy for use in scientific publications. • Identifying specific contributions to published research will lead to appropriate credit, fewer author disputes, and fewer disincentives to collaboration and the sharing of data and code.
Conference Paper
Full-text available
Context: For many years, researchers and practitioners have been proposing various methods and approaches to Requirements Engineering (RE). Those contributions remain, however, too often on the level of apodictic discussions without having proper knowledge about the practical problems they propagate to address, or how to measure the success of the contributions when applying them in practical contexts. While the scientific impact of research might not be threatened, the practical impact of the contributions is. Aim: We aim at better understanding practically relevant variables in RE, how those variables relate to each other, and to what extent we can measure those variables. This allows for the establishment of generalisable improvement goals, and the measurement of success of solution proposals. Method: We establish a first empirical basis of dependent variables in RE and means for their measurement. We classify the variables according to their dimension (e.g. RE, company, SW project), their measurability, and their actionability. Results: We reveal 93 variables with 167 dependencies of which a large subset is measurable directly in RE while further variables remain unmeasurable or have too complex dependencies for reliable measurements. We critically reflect on the results and show direct implications for research in the field of RE. Conclusion: We discuss a variety of conclusions we can draw from our results. For example, we show a set of first improvement goals directly usable for evidence-based RE research such as "increase flexibility in the RE process", we discuss suitable study types, and, finally, we can underpin the importance of replication studies to obtain generalisability.
Article
Context Requirements Engineering (RE) is a critical discipline mostly driven by uncertainty, since it is influenced by the customer domain or by the development process model used. Volatile project environments restrict the choice of methods and the decision about which artefacts to produce in RE. Objective We aim to investigate RE processes in successful project environments to discover characteristics and strategies that allow us to elaborate RE tailoring approaches in the future. Method We perform a field study on a set of projects at one company. First, we investigate by content analysis which RE artefacts were produced in each project and to what extent they were produced. Second, we perform qualitative analysis of semi-structured interviews to discover project parameters that relate to the produced artefacts. Third, we use cluster analysis to infer artefact patterns and probable RE execution strategies, which are the responses to specific project parameters. Fourth, we investigate by statistical tests the effort spent in each strategy in relation to the effort spent in change requests to evaluate the efficiency of execution strategies. Results We identified three artefact patterns and corresponding execution strategies. Each strategy covers different project parameters that impact the creation of certain artefacts. The effort analysis shows that the strategies have no significant differences in their effort and efficiency. Conclusions In contrast to our initial assumption that an increased effort in requirements engineering lowers the probability of change requests or project failures in general, our results show no statistically significant difference between the efficiency of the strategies. In addition, it turned out that many parameters considered as the main causes for project failures can be successfully handled. Hence, practitioners can apply the artefact patterns and related project parameters to tailor the RE process according to individual project characteristics.
Conference Paper
[Context] Many software projects fail due to problems in requirements engineering (RE). [Goal] The goal of this paper is analyzing a specific and relevant RE problem in detail: incomplete/hidden requirements. [Method] We replicated a global family of RE surveys with representatives of software organizations in Austria and Brazil. We used the data to (a) characterize the criticality of the selected RE problem, and to (b) analyze the reported main causes and mitigation actions. Based on the analysis, we discuss how to prevent the problem. [Results] The survey includes 14 different organizations in Austria and 74 in Brazil, including small, medium and large sized companies, conducting both, plan-driven and agile development processes. Respondents from both countries cited the incomplete/hidden requirements problem as one of the most critical RE problems. We identified and graphically represented the main causes and documented solution options to address these causes. Further, we compiled a list of reported mitigation actions. [Conclusions] From a practical point of view, this paper provides further insights into common causes of incomplete/hidden requirements and on how to prevent this problem.
Chapter
Most of the data produced in software projects is of textual nature: source code, specifications, or documentations. The advances in quantitative analysis methods drove a lot of data analytics in software engineering. This has overshadowed to some degree the importance of texts and their qualitative analysis. Such analysis has, however, merits for researchers and practitioners as well. In this chapter, we describe the basics of analysing text in software projects. We first describe how to manually analyse and code textual data. Next, we give an overview of mixed methods to automatic text analysis including N-Grams and clone detection as well as more sophisticated natural language processing identifying syntax and contexts of words. Those methods and tools are of critical importance to aid in the challenges in today's huge amounts of textual data. We illustrate the introduced methods via a running example and conclude by presenting two industrial studies.