You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320027777

What Questions do Requirements Engineers Ask?

Conference Paper · September 2017


DOI: 10.1109/RE.2017.76

CITATIONS READS
12 4,974

4 authors, including:

Michael Vierhauser Jane Cleland-Huang


Johannes Kepler University Linz DePaul University
66 PUBLICATIONS   617 CITATIONS    247 PUBLICATIONS   6,015 CITATIONS   

SEE PROFILE SEE PROFILE

Smita Ghaisas
Tata Research Development and Design Centre (TCS Research)
71 PUBLICATIONS   439 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Data Driven Requirements Engineering View project

Architectually Significant Requirements View project

All content following this page was uploaded by Smita Ghaisas on 10 July 2018.

The user has requested enhancement of the downloaded file.


What Questions do Requirements Engineers Ask?
Sugandha Malviya Michael Vierhauser and Jane Cleland-Huang Smita Ghaisas
School of Computing University of Notre Dame Tata Consultancy Services
DePaul University, Chicago, USA. South Bend, IN, USA Pune, India
slohar@depaul.edu mvierhau@nd.edu, JaneClelandHuang@nd.edu smita.ghaisas@tcs.com

Abstract—Requirements Engineering (RE) is comprised of var- data is sometimes incomplete, inconsistent and important trace
ious tasks related to discovering, documenting, and maintaining links can be missing [22]. Accessing these data sources and
different kinds of requirements. To accomplish these tasks, a combining them to produce meaningful and desired results
Requirements Engineer or Business Analyst needs to retrieve
and combine information from multiple sources such as use case can therefore be considered a cumbersome and error-prone en-
models, interview scripts or business rules. However, collecting deavor [19]. Such problems can best be alleviated by strategic
and analyzing all the required data can be tedious and error- upfront project planning and with appropriate instrumenting
prone and often results in untraceable or missing requirements of the environment.
or inadequate requirements verification. To understand, and Several user studies [12, 6, 25] have investigated and
moreover support requirements engineering activities in practice,
we conducted an online survey with requirements professionals collected important questions asked by software practitioners
in the IT industry. Our main goal was to discover the kinds in various project roles, and then analyzed the questions to
of queries and artifacts they are interested in. Our analysis discover the information snippets needed to support their
included 29 surveys responses and a total of 159 natural language tasks. These studies have focused primarily on the developer
queries. Using open coding and grounded theory, we analyzed and and activities related to programming and testing and have
grouped these queries into 9 different query purposes and 53 sub-
purposes, and also identified frequently used artifacts. Analyzing not focused on activities such as requirements elicitation,
real-world queries can shed light on the questions requirements stakeholder analysis, and change impact analysis. To shed light
professionals would like to ask and the artifacts needed to on the tasks performed by requirements engineers and business
support such questions. This is potentially useful for project-level analysts in practice – and to provide better support for strate-
planners, and could help them to identify important questions, gic project planning – we conducted an online survey with
proactively instrument their environments with supporting tools,
and strategically collect data that is needed to answer the queries requirements professionals in the IT industry. In contrast to
of interest to their project. previous studies we focused on discovering questions pertinent
Index Terms—Requirements Engineering Tasks, Requirements to requirements engineers and business analysts. As a result
Engineering Questions, Requirements Traceability of this survey we collected 159 questions from 29 participants
and extracted 9 main purposes and 83 artifact types using open
I. Introduction coding and grounded theory [11, 1].
Requirements Engineering is a vital process for creating The goal of the study is to discover different kinds of
high quality software systems [9, 8]. It involves diverse questions that business analysts and requirements engineers
tasks including requirements elicitation, analysis, validation, are interested in asking to support requirements-related tasks
verification, risk management, and change control [18]. A such as quality assessments, business rule analysis, and change
plethora of different tools, methods, and techniques are needed impact analysis. More specifically, we investigate two research
to successfully perform these tasks; however, even with proper questions:
tool-support in place, such tasks can be time-consuming and • RQ1 : What kinds of queries are useful for requirements
difficult to perform. Researchers have identified numerous engineers and business analysts?
problems associated with performing tasks as diverse as • RQ2 : What artifact types are required to address their
stakeholder identification, requirements elicitation and analy- questions?
sis, requirements specification, change management. However, We then show how this knowledge can be used to help
providing support for such requirements engineering tech- project stakeholders strategically instrument their environ-
niques involves collecting and managing information from ments with the project data and tools needed to service their
many different data sources. For example, performing stake- questions.
holder analysis requires retrieving, collating, and analyzing The remainder of this paper is structured as follows.
information from interview notes, scenarios, goals and other Section II discusses related work and empirical research in
artifacts [30]. Typically, such artifacts are neither managed by software engineering. Section III describes the survey design
a single person, nor stored in a single location but distributed and the procedures for data collection and analysis. Sections
across multiple repositories (e.g., document management sys- IV- VI report results, describe a usage scenario, and provide a
tems, source-code repositories, issue trackers, or application detailed discussion of our findings. Finally, Sections VII and
lifecycle management (ALM) services). Furthermore, artifact VIII discuss threats to validity, conclusions and future work.
II. Related Work specific roles and responsibilities. Such knowledge drives
strategic decisions about which data should be collected, when
Several studies in the past decade have explored the kinds of project stakeholders should invest time and effort constructing
questions software practitioners are interested in asking. Many trace links, which scripts should be created to retrieve data
of these studies focused on questions asked by software devel- from individual tools, and which tools are needed to col-
opers. In addition, numerous techniques exist for supporting late and present data to developers, requirements engineers,
users in navigating through complex artifacts and information. project managers, and other stakeholders. Furthermore, such
Various studies have identified questions that support soft- studies provide guidance for tool developers and researchers
ware practitioners’ tasks. For instance, Fritz et al. [6] inter- interested in addressing informational needs. To the best of
viewed professional developers and identified 78 questions that our knowledge however, there has been no systematic effort
were of interest to them. They found that answering these to understand the kind of questions Requirement Engineers
questions required integration of information from multiple or Business Analysts working in the Software Engineering
sources including source code, change sets, team information, industry are interested in asking.
and work items. To support co-located software development
teams, Ko et al. [12] collected 21 questions and grouped them III. Research Context & Survey Design
into seven different categories including submitting a change, We now outline the procedures and guidelines followed in
and reproducing a failure. Other related studies focused on the design and execution of the survey.
specific development tasks, questions for detecting security
vulnerabilities [27], maintaining source code [25] and analyz- A. Identifying Requirements Tasks
ing bug reports [5]. Begel et al. [3] created a catalog of 145 As a preparatory step for the survey, we constructed a list of
software analytical questions of interest to Data Scientists. common requirements engineering tasks. 15 tasks were iden-
The questions were gathered through two different surveys tified primarily through analyzing requirements engineering
and included participants from three disciplines: development, text books [13, 23, 14, 29]. Following extensive discussion
testing, and program management. Based on the participants’ we eliminated similar tasks and reduced the list to 11 distinct
roles and interests, the questions were grouped into 12 differ- tasks that would form the focus of our survey. An overview of
ent categories. For example, a category of Testing Practices these tasks, with a brief description, can be found in Table I.
covered questions related to unit-testing, test coverage, and
test automation. B. Subjects
These studies all focused on software development activi- We targeted professional Requirements Engineers and Busi-
ties, while our interest [15, 16] is in requirements engineering. ness Analysts working in the IT industry. To recruit partici-
Our survey therefore deliberately targets requirements engi- pants we created several different mailing lists. The first was
neers and business analysts and focuses on their associated comprised of authors who had published papers in indus-
activities. We collect questions they perceive as important try tracks of requirements engineering related conferences.
and identify data artifacts which are needed to answer such We included papers from the past five years (2012-2016)
questions and to support requirements engineering activities. of the Requirements Engineering conference (RE), and the
In similar work, Bouillon et al. [4] conducted a survey Requirements Engineering for Software Quality conference
involving practitioners where they identified 29 traceability (REFSQ). Only authors affiliated with a concrete IT company
usage scenarios such as refining and detailing requirements were included. Authors from academia were not included.
and test coverage analysis of specification and code. They As a result, we created a list of 101 potential candidates.
grouped these scenarios into 6 different groups, each pertaining The second list, containing contact information for 23 people,
to a phase of the software development process. was comprised from our personal networks of professional
In addition to surveys that were designed to directly elicit Business Analysts and Requirements Engineers.
queries of interest from practitioners, other researchers have Moreover, to reach additional participants, we
sought to observe and understand requirements engineering used the professional social network site LinkedIn
activities. For example, Zowghi and Coulin [30] studied tech- (http://www.linkedin.com) to advertise the study in virtual
niques, approaches, and tools used to perform requirements practitioner groups. We posted our survey information
elicitation. Ramesh and Jarke [21] conducted an extensive and link to three different groups: the IREB Certified
empirical study of RE practices across different organizations. Professional for Requirements Engineering (2,523 members),
They first identified specific information needs for software the Requirements Engineering Specialist Group (2,093
engineering tasks, and then proposed traceability models that members), and the Requirements Engineering Forum of
would provide the necessary support for answering a wide Asia-Pacific (737 members). For this part we used a separate
range of queries. They documented their findings as a set survey link for collecting responses.
of three reference level traceability meta-models describing
artifact types and traceability links. C. Survey Design & Process
All of these studies provide invaluable insights into the Interviews and surveys are both popular techniques for
information needs of project stakeholders according to their performing empirical research [26]. We decided to conduct an
TABLE I: RE Tasks included in our study. The number in familiarize participants with the kinds of queries a software
parenthesis reflects the number of queries collected for each practitioner could ask about their task such as “Have any
of the selected task. components repeatedly failed acceptance tests?” However, we
found that users tended to closely mimic the examples, rather
Task Description than creatively thinking of their own queries. We therefore
Elicitation & This task involves working with stake- replaced the examples with more descriptive scenarios (cf.
Specification (43) holders such as customers to understand Table II) and associated queries and then asked the participants
the business domain, discover, and specify to think of their own scenarios and associated queries.
functional requirements and give adequate We also reduced the number of tasks from three to one,
inputs to architects so that they can design a in order to increase the chance that participants would com-
technical solution that meets requirements. plete the survey [26, 20]. We removed a question that asked
participants how they currently performed each task as this
Business Rules This task involves capturing and document-
was overly time-consuming to answer while providing limited
Analysis (15) ing business rules such as regulations and
company policies. It also involves under-
benefit to the study. Finally, we limited the maximum number
standing the implicatioons of the rules. of queries per task for each user to five and reduced the widget
space for entering the query text to five lines so as to avoid
Quality Assessing the quality of individual require- verbose queries.
Assessment (38) ments and the overall specification. The final questionnaire comprised the following parts:
Safety In a safety-critical system, failure or mal- • An introductory information sheet including an overview
Assessment (5) function may result in death, serious injury, of tasks to be performed and a request for consent to
severe damage, or environmental harm. In participate.
the safety-assessment task, an analyst as- • Questions regarding the participant’s experience in the IT
sesses whether identified hazards have been
industry and their role as a Business Analysts or a Re-
fully mitigated [17].
quirements Engineer. Participants without this experience
Threat Identifying security threats and specifying were excluded and directed to the end of the survey.
Modeling (0) security requirements. • A list of commonly performed requirements tasks as
Performance Dealing with requirements related to scala- shown in Table I. Participants were asked to select a task
Requirements (6) bility, availability, performance etc. they were familiar with.
• A series of questions about artifacts and tools used to
Change Impact Determining those parts of a software sys- perform the selected task, the difficulty level for accom-
Analysis (12) tem that might affected when a certain
plishing the task, and frequency at which the task was
change is performed [2].
performed.
Coverage Evaluating the extent to which requirements • Two sample project scenarios with example questions.
Analysis (2) are covered in the design. These are shown in Table II.
Compliance Ensuring that external regulatory codes have • A prompt for up to five useful queries associated with a
Analysis (4) been fully addressed by the requirements. scenario relevant to the participant’s work.
• An option for the participant to select another task and
User Acceptance This task involves testing against speci- to provide additional queries.
Testing (16) fied requirements and proving that the cus-
• A request for additional comments about the study and
tomers’ needs are met by the developed
to optionally provide contact information for follow-up
software.
purposes.
Project Managing requirements including priorities The online questionnaire was created using Qualtrics Sur-
Management (31) and schedules.
vey. The survey link1 was shared through a recruitment email.

D. Analysis
online survey as this minimizes the effort for collecting data, The goal of the survey was to collect requirements related
increases the number of potential participants, and allows sub- questions from practitioners, to categorize them based on the
jects to answer the questions according to their own schedule purpose of the query, and to identify artifact types needed
and availability [20]. As an incentive for participating in the to support the query. From the completed surveys, the first
study, we offered to share study results. researcher filtered out surveys containing no queries and those
As recommended by Yin [28] we performed a pilot received from participants with no industry and requirements
study with a couple of experienced Requirements Engi- engineering experience. The remaining surveys were consid-
neers/Business Analysts. Based on the responses received, ered valid responses.
we made small updates to our survey. For example, in our
initial survey we included several sample project queries to 1 http://depaul.qualtrics.com/SE/?SID=SV_7aMFE3v3CC8gqk5
Step 1: Step 2: Step 3: Step 4: Step 5:
Requirements Aggregated
Purposes and
Surveys Queries Purposes and
Tasks Artifacts
Artifacts

... refinement

Purposes 54 Purposes 9
11 29 159 Artifacts 83 Sub-Purposes 53
Artifacts 83

Senior Senior
Researcher 2 Researcher 3
cross-check
Researcher 1 and validation

Fig. 1: Research method followed: after creating a list of 11 requirements-related tasks, survey responses and queries were
collected. The anticipated purposes of the queries and the extracted artifacts were then used to iteratively derive codes and
theories.

The first researcher then extracted quantitative data for the the end of the survey.
queries provided by the participants into a spreadsheet (cf. We followed the same approach for queries that were
Fig. 1 - Step 3). We then used open coding [24, 11] to analyze deemed out of scope, not useful, or provided too little in-
the queries and extract relevant information. In particular, we formation on either the involved artifacts or the purpose.
were interested in two main aspects of the query: (1) the After thoroughly discussing those queries we either removed
purpose of the query, i.e. the intention the user had when them from the set of used queries – if all three researchers
issuing the query and (2) the artifacts referenced by the query. agreed – or discussed their purpose and artifact references
until agreement was reached.
TABLE II: Example Queries presented to Survey Participants In the next step (cf. Fig. 1 - Step 5) we followed the
grounded theory [1, 7] approach to further group and cate-
Scenario Example Queries
gorize the queries according to their purpose. In a step-wise
Project Management - Ray wants to -Is there any schedule refinement all three researchers discussed the coded purposes
know if there are any delayed slippage?
requirements before announcing the -How many requirements and attempted to group them into high-level purposes with
release date of their new phone. What have their status as subsequent sub-groups. The final set of purposes, sub-purposes
will be his query? pending? and artifacts provides guidance for instrumenting real-world
Threat Modeling - Susie and her team -List all the identified projects.
are specifying security requirements for threats for the
the drone remote control system. She authentication application.
thinks she may have missed some of the -Have all the threats been IV. Survey Results
threats related to operator authentication. identified and prioritized
She issues a query to confirm this. What for the drone remote Using the first mailing list comprised of authors of con-
is her query? control system? ference papers in industry tracks, we were able to collect 26
survey responses. From our network of contacts we collected
We selected these two elements as our main focus of in- 13 additional responses for an overall response rate of 33%.
terest with the expectation of identifying frequently addressed There was only a single response from the LinkedIn group. We
purposes and/or artifacts across the different tasks. As a pilot thus were able to collect 40 responses in total. Out of these
for the coding procedure, all three researchers involved in the 40 responses, 11 (6 due to the lack of industry experience
process coded 30 randomly selected queries and discussed the and 5 responses based on our exclusion criteria discussed in
outcome until mutual agreement for both purpose and artifacts Section III-D) were removed, resulting in a final set of 29
was reached. The first researcher then coded the remaining surveys.
queries and the other two researchers, on a frequent basis, Based on the quantitative data collected in the first part of
checked random samples. In cases where the first research our survey we can conclude that we were able to reach a broad
was unsure about either the purpose or the artifact references, spectrum of practitioners and experts with several years of
the query was tagged and discussed by all three researchers. industrial requirements engineering experience (cf. Table III).
In the case of eight queries, additional clarification was sought From our valid 29 surveys we calculated that more than 62%
directly from the participants. This was only possible for 11 of the survey participants can be considered very experienced
cases where participants had provided their email address in with more than 10 years of general IT experience and 31%
TABLE III: Survey participants’ background and experience. • Project Management: Queries related to activities such
From the total number of 40 participants 6 were removed in as monitoring the project schedule, estimating project
the due to no practical experience and 5 due to missing or costs or planning releases were added to this group. For
incomplete data. example, “List all requirements that have their status as
Experience IT Experience BA Experience
pending.”
• Quality: Queries addressing an artifact’s (or its
None 4 2
Less than one year 1 1 attributes) quality such as queries looking for undefined
1-5 years 5 13 terms or queries addressing quality analysis (cf. [10]).
6-10 years 8 10 For example, “Find all ambiguous words in
more than 10 years 22 10
requirements.”
Total 40 36 • Risk Management: Queries related to compliance or
hazard analysis were assigned to this category. For
with more than 10 years of business analysis and requirements example, “In what frequency can a hazard happen?”.
engineering. • Search: Queries that require searching through a set of
We were able to collect queries for 10 out of the 11 origi- artifacts in order to retrieve the desired information
nally targeted RE tasks (cf. Table I). None of the participants were assigned to this purpose. The search could either
selected the Threat Modeling task. Figure 2(a) presents an employ a specific matching rule or perform a text-based
overview of the selection rate of the different tasks. Further- search, e.g., “What previous projects had requirements
more, participants were asked to rate how frequently they describing synchronization between multiple
performed the selected tasks on a three-step scale (i.e., Rarely, instruments?”
Sometimes, and Regularly). Figure 2(b) provides an overview • Stakeholder Analysis: Queries primarily targeting an
of the responses and confirms that most selected tasks were individual (or a group of) customer, supplier, user etc.
performed on a regular basis. such as “Show me the stakeholder(s) with an interest in
As a result of the survey we collected 172 queries, with a given business rule.” or “For a given requirement,
each participant providing an average of four queries with a who are the stakeholders of interest?” were assigned to
minimum of 2 and a maximum of 10. Following an initial this purpose.
screening we found that several queries were either incomplete To preserve the diverse intentions behind the high-leve
or inconclusive. We thus removed 13 queries resulting in a categorization of queries we introduced sub-purposes. These
final set of 159 which we used in the coding step. Our initial are depicted in Table IV alongside several example queries
effort to code the intentions of the users issuing the different from our survey. A more comprehensive list of gathered
queries resulted in 59 different purposes. questions and their encodings can be found here: https://
In a joint effort we then tried to find and group related tinyurl.com/ j6flyts
purposes which led to a final aggregate set of 9 high-level We therefore address our first research question RQ1 by
purposes. We categorized queries solely on the query’s identifying and reporting questions for Requirements Engi-
primary aim, even if a participant had initially selected a neers and Business Analysts.
different ‘umbrella’ task in the survey. We define query As a second contribution of our survey we extracted the re-
purposes as follows: lated artifacts for each query. In total, we collected 83 different
• Business Rule Analysis: Queries where the user was artifact types. Furthermore, we identified related artifacts for
primarily interested in business related aspects. This, for each purpose. Table V provides a comprehensive overview
example, includes business objectives, business rules, of artifacts and associated purposes. We then calculated the
and business processes. One example query assigned to occurrence of each artifact type across all 9 query purposes.
this group was “What are the business rule Figure 3 presents the artifacts that were used to address more
requirements [for this project]?”. than one query responses. The size of each artifact bubble
• Change Impact Analysis: Queries that referred to any is proportional to the number of purposes it addresses. We
kind of change such as adding new features, change also evaluated the occurrence of each artifact type across the
requests or issues arising due to a change. For example, complete query set. Not surprisingly, Requirements appeared
“How to minimize system impact with historical data on in 83 queries, Business Rules in 17 queries, Stakeholder in 17
such changes?”. queries, Requirements Type in 15 queries, Test Cases in 15
• Metrics: Queries that employed any sort of queries and Business Processes in 10 queries Other artifacts
measurement as their primary purpose such as counting, such as Change Requests, Release, and Processes appeared in
calculating maxima or averages were assigned to this at least 6 queries. However, 45 artifact types appeared in only
purpose. For instance, “How many requirements are in one query.This suggests that a core set of artifacts are common
this specification?” would require to perform a count on across many queries; however, project planners must also
requirements. consider less frequently used artifacts, that support questions
• Process: Queries related to a process step. For example, important to their specific project goals and environments.
“Is there any defect to test case mapping?”. We thus, address RQ2 by reporting all the associated artifact
TABLE IV: The nine extracted purposes with their respective Sub-Purposes, and Example Queries taken from the survey
Purpose # Queries Sub-Purposes Examples
Business Planning and Strategy -At what step in a business process X does a business rule A
Business Rationale apply?
Business Rule
7 Search Business Rules -What type of constraints are embedded in this rule?(
Analysis
Search Business Rules Constraints geographic constraints?, deadlines?)
Reverse Engineering Business Rules -List all business objectives.
Business Impact Analysis -Do we know the impact is minor or major and will it affect
Managing Sub-Systems upstream downstream applications?
Managing Release -Does the new enhancement creating any issues in the testing of
Change Impact Product Evolution existing functionalities?
Analysis 12
Rationale -What are the changes? and what are the business modules that
Requirements History are going to be impacted for the change?
Requirements Management -How many systems are impacted by these requirements and
Testing for Finding Issues which business/system process are they associated to?
Document Analysis -How many non-functional requirements are there?
Issue Tracking -Do all requirements have a direct traceability to documented
Project Management business processes?
Metrics 16
Requirements Coverage -How many features exist?
Security Analysis -Are there test cases for all requirements?
User Acceptance Testing
-Is there any defect to test case mapping?
Process Status -How many stories are there not done in the Release?
Process 6
Metrics -How many bugs are there for the Epic called "Support"?

Cost Estimation -List all requirements that have their status as pending.
Critical Path Analysis -Is there any schedule slippage?
Managing tasks -What is the current utilization of resources based on the RAG
Project Planning status of the project?
Project
29 Progress -How many security requirements have their status as pending?
Management
Resource Estimation -Have we estimated the task?
Release Planning -Do we have task prioritized?Critical tasks needs to be
Scheduling performed in a timely manner.
Tracking Project Performance
-Find all ambiguous words in requirements.
Analysis
-Are there any duplicate requirements?
Backward Tracing (origins)
-Are all requirements traced to a stakeholder need?
Completeness
-How many processes are identified as a GAP for this release
Meeting Project Objectives
and do all of them have a technical response identified?
Quality 33 Metrics
-Are there any duplicate requirements?
Quality Control
-Do all of the process steps have requirements defined for them?
Quality Set
-Are functional requirements traced to the relevant function?
Reviewing Requirements
-Did I missed any requirements from stakeholders?
Uncovering Errors or Inconsistencies

-What are regulations to comply with?


-What are the main hazards that can happen to the system?
Risk Compliance Analysis -What can cause a hazard?
Management 8
Hazard Analysis -What are the consequences of a hazard?
In what frequency can a hazard happen?

-Which requirements are related to the requirement "xyz"?


Attributes -How many concurrent users would be there for the application?
Business Rule -Please list the business rules relevant to Travel Insurance
Data Provenance applicable for Multicountry travel.
Identifying Requirements -Does requirement match to a business objective?
Search 36
Finding Matching Requirements -Do previous projects have requirements related to measuring
Project Glossary Extraction total harmonic distortion?
Rationale -Need to know the regulatory compliance requirements pertinent
Uncovering Inconsistencies to Payer process in Healthcare

-For a given requirement,who are the stakeholders of interest?


Stakeholder Identification -What kind of users are going to use the system?
Stakeholder
14 Use Case Analysis -How many bugs have been found during acceptance testing?
Analysis
User Acceptance Testing -For a given business rule, identify the originator of the business
rule
Project  Management    
Threat  Modeling     User  Acceptance  Tes.ng    
Compliance   0%  
Analysis     Project   Compliance  Analysis    
3%   Management    
12%   Coverage  Analysis   Rarely  
Elicita'on  and  
Specifica'on   Safety     Change  Impact  Analysis   Some.mes    
22%   Assessment   Performance  Requirements    
Business  Rules   2%   Regularly  
Analysis     Threat  Modeling    
15%   Quality   Safety  Assessment  
Assessment     Quality  Assessment    
Coverage  Analysis   17%  
5%   Business  Rules  Analysis    
User  Acceptance   Performance   Elicita.on  and  Specifica.on  
Change  Impact   Tes'ng    
Analysis   Requirements  
12%   3%   0   2   4   6   8   10   12   14   16  
9%  
(a) Distribution of selected tasks. (b) Frequency of performing the task as reported by the partipants.

Fig. 2: Analysis of covered RE Tasks regarding their relative occurrence (a) and frequency (b).

that address the collected questions and by presenting the implementing the change. He could also search for any related
artifacts that are important in the context of different query requirements, and if necessary find stakeholders who might
purposes. be knowledgeable about the topic of the change request.
The fact that the TIM was established in advance, necessary
V. A Usage Scenario data collected and preserved, and the project environment
Collecting and analyzing queries and uncovering frequently instrumented appropriately allows Michael to answer questions
used artifacts can provide foundations for successfully im- that support his current review to enhance an existing feature.
plementing query support in real-world projects. With the
following scenario we illustrate how this information could VI. Discussion
be used. Imagine that a software firm called JMS Corp. is in A. Task Coverage
the process of developing a product.The company uses diverse Our initial set of RE tasks contained 11 different activ-
artifacts and respective tools for managing them. Their require- ities of which ten were covered by survey responses. We
ments are managed in the JIRA Issue Tracking system, impact collected only 2 queries for the least frequently selected task
analysis reports of previous projects are stored as an Excel of Coverage Analysis ranging to 43 queries for the Elicitation
spreadsheets and test cases stored in their own proprietary & Specification task. No queries were collected for Threat
tools. Let us further assume that JMS utilizes the artifacts Modeling. This is perhaps indicative of the lack of threat
identified in Table V to strategically construct the Traceability modeling activities performed in general, or could also reflect
Information Model (TIM) shown in Fig. 4. This TIM in- lack of diversity in our participant pool. The same holds true
cludes the artifacts used by our survey respondents to support for Coverage Analysis (2 queries) and Compliance Analysis
“Change Impact Analysis” queries and now guides JMS’ (4 queries). Although we did include Business Analysts in our
decisions for creating infrastructure, for example, deploying list of potential participants these tasks are only very sparsely
tools such as TaskTop or Mylyn to integrate heterogeneous populated. We believe this may be because other types of
data sources, or to write customized data retrieval scripts to project stakeholders such as risk management officers or even
make such data accessible for querying. developers are more likely to perform these tasks.
Now consider what happens if a customer submits an
enhancement change request and that the request has been B. Purpose Overlaps
approved and is opened for implementation. Michael is work- While extracting the query purposes and during the aggre-
ing as a Business Analyst and requires information about the gation phase we noticed an overlap of query purpose and sub-
change request. purpose. This means that a single query may serve multiple
The TIM displays the possible artifact types that can be used purposes. For instance, we defined Metrics as a purpose but
as well as the associations established between artifacts. This also as a sub-purpose under Process. Consider the query,
helps Michael determine what kind of queries are currently How many bugs are there for the Epic called “Support”?
supported in his project environment. For example, he could whose immediate objective is Process, but achieving this
consider searching for current features related to the en- objective requirements counting bugs. Therefore Metrics was
hancement request and retrieve their historical impact reports recognized as a sub-purpose. We deliberately refrained from
to better understand how much effort would be involved in further aggregating or merging such cases since our intention
Change Change
Impact Requests Metrics
TBD Items
Analysis

Process
Test Status

Business
Rules

Stakeholders Release
Analysis
Test Cases
System
Specification
System
Design
Business Rule
Analysis
Bugs

Requirements
Search Project
Business
Objective Management
Model/Goals

Reqs. Type

Constraints

Stakeholder Processes

Process Reqs.
Status
Steps
Reqs.
Quality
Attributes
Business
Processes

Quality Risk
Trace Data Management

Fig. 3: Overview of frequently used artifact types and their occurrence in the different purposes.

was to preserve the original motivation when issuing the Likewise, some sub-purposes dominated their respective
query. We furthermore noticed that some of the sub-purposes categories. For example,13 of the 29 project management
were across multiple purposes. For instance, the sub-purpose queries were related to progress. Similarly Finding matching
Rationale is defined under both Change Impact Analysis and requirements and Finding Business Rules covered 42% of
Search. search queries.

C. Related Sub-Purposes E. Artifacts Type Coverage


Although we identified, a large number of artifact types
Some of the extracted sub-purposes are closely related.
from our query set, only a small subset of them are used across
For example, Finding Matching Requirements and Identifying
the majority of query purposes. Most artifacts addressed only
Requirements both belong to Search. However, the first type
one or two query purpose categories. This is partially due
of query finds matches based on artifact content, for example
to the fact that we did not aggregate similar artifact types.
for queries such as “Which requirements are related to the
However, such aggregation would be possible. For exam-
requirement xyz?”, while the second uses trace links or at-
ple, Hazard Cause, Hazard Consequence, Hazard Occurrence
tributes to identify artifacts based on constraints. An example
Rate, Hazard Type could be consolidated into a composite
is “Which requirements have pending status?” Each type of
artifact such as Hazard or Fault Tree. Here we decided to
query requires completely different project tooling.
report raw artifact references and leave artifact analysis for
D. Purpose and Sub-Purpose Coverage future work.

Another observation was that 61% of queries were rec- VII. Threats To Validity
ognized under the query purposes of Search, Quality and Construct validity: The queries collected during the survey
Project Management, and additional 26% of queries were were produced outside the context of real software engineering
tagged under the purposes of Change Impact Analysis, Metrics tasks and thus need to be evaluated in practice. However
and Stakeholder Analysis, while only 13% were associated all queries were provided by practitioners with several years
with the query purposes of Risk Management, Business Rule of experience and therefore serve as a good starting point
Analysis and Process. This can partially be attributed to for further study. Another threat to validity stems from the
overlap between categories (especially ones such as metrics interpretation and analysis of the queries. To mitigate possible
or search); while results suggest that some tasks are more misinterpretations three researchers were involved in the pro-
common than others, further studies are needed to determine cess. The first researcher conducted the surveys and collected
if this observation is generalizable. survey and query data. The other researchers participated in the
TABLE V: Artifact Types Referenced by each Query Group
Business   Business   Business  
Business Business Objectives Model/Business Goals, Business Rules   Processes  
Modules  
Rule Processes, Business Rule Constraints, Business Rules, supported  by  
Analysis Reqs., Comply   derived  
to   from  
Change Business Modules, Business Processes, Business Rules,
Business  
Impact Change Requests, Functionalities, Impact, Issues, Items, Historical  Data   Func,onali,es   Requirements  
Analysis Reqs., Stakeholders, Sub-System, System
scope  
Specifications, Test Cases, implements   change  
Metrics Business Processes, Change Requests, Design, Features,
Feature Tree, Reqs., Reqs. Type, Test Cases, Test Goals   Requirements   Change  
describes  
Scenarios, Test Status, Trace Links, Requests  
affects  
Process Bugs, Defects, Epic, Release, Reqs., Reqs. History, submits    
Solution Reqs.,Test Cases, Test Plan, User Stories, System   Impact  
Project Budget, Business Objectives Model/Business Goals, Specifica,ons   Stakeholders  
tests   affects  
Manage- Business Processes, Development Status, Documents,
ment Estimation Metrics, Process Status, Processes, Project
Test  Cases   Issues  
Goals, Project Health, QA Team, Release, Release
Schedule, Reqs., Reqs. Priority, Reqs. Status, Reqs.
Type, Resource Usage, ROI, Schedule, Signature, Fig. 4: A Traceability Information Model constructed to
Signature Status, Task Priority, Task Type, Tasks, Test support “Change Impact Analysis” queries using artifacts
Cases, Version referenced by survey respondents.
Quality Bugs, Business Processes, Business Teams, Functional
Areas, Functions, GAP Analysis, Metrics(Defect
Density, High Impact, System Sensitiveness),
Performance Measures, Process Steps, Processes, Reqs.,
personal network of contacts.
Reqs. Model, Reqs. Quality Attributes, Reqs. Status,
Reqs. Type, Stakeholders, Stakeholder’s Reqs., Regarding external validity and the generalizability of the
Technical response/solution, Test Cases, Trace Data, results we reached out for practitioners in different domains
Transactions, User Profiles, User Stories, with different kinds of skills and experience. In total 29
Risk Hazard Cause, Hazard Consequence, Hazard practitioners participated in our survey working on different
Manage- Occurrence(frequency), Hazard Type, Hazards,
ment Products, Reqs. Type, Stakeholders, System
tasks such as requirements elicitation, business analysis and
Specifications, project management. We are thus confident that our survey
Search Business Objectives Model/Business Goals, Business covers a sufficiently broad area of practice. As part of our
Processes, Business Rules, Components, Constraints, future work we plan to extend the initial survey to reach
Design, Disaster Recovery, Functional Scope additional participants from different domains to cover those
Boundaries, Interface, Items, Metrics(output resistance,
tasks that are currently only sparsely populated with queries.
output impedance, or path resistance), Model, Previous
Projects, Process Steps, Processes, Reqs., Reqs. Quality
Attributes, Resolver Reqs., Reqs. Type, Schedule, VIII. Conclusion and Future Work
Stakeholders, System Specifications, Test Cases,
Stake- Activity Diagram, Bugs, Business Rules, Checklists, In this paper we presented an empirical study to elicit and
holder Data(Critical transactions pertaining to business analyze queries of interest to Requirements Engineers and
Analysis Processes), Reqs., Stakeholders, Systems Model, Business Analysts. The data was collected through an online
System Specifications, Test Cases, Test Status,
survey including 29 requirements professionals with the aim
of understanding the kinds of questions they would like to ask
to accomplish those tasks. In total we collected 172 questions
coding and cross-checked the results in all subsequent steps. and used 159 question for further analysis. From analyzing
In case an e-mail address was provided (on an optional basis) these queries we were able to extract 9 different purposes
we contacted the participant to clarify any uncertainties about and 83 different types of artifacts that our participants were
their queries. interested in. Furthermore, we identified the artifacts that were
Internal Validity: One of the challenges with surveys is used frequently across queries of different purposes; however,
recruiting a target group of respondents. To increase the prob- we did not aggregate the identified list of artifacts. We reserve
ability of receiving more responses, we conducted an online this as a part of the future work.
survey. However, as all the respondents were IT engineers, Furthermore, we aim to collect additional queries for tasks
taking an online survey was not an issue in our case [20]. that were not sufficiently covered and create a comprehensive
We distributed a separate anonymous survey link on LinkedIn set of query purposes with their associated artifacts. We are
Requirements engineering groups. This introduces a risk of confident that providing query support for requirements-related
inviting participants outside the focus group. We however re- artifacts can ease the task of combining and linking project-
ceived a single survey response. All the other survey responses specific information from different sources and in the long run
were received from the requirements engineers recognized simplify these tasks hence improving the quality of the whole
from the requirements related conferences and through our project.
Acknowledgements collections of domain documents. In Proc. of the 24th
The work described in this paper was partially funded by US Int'l Requirements Engineering Conf., pages 156–165.
National Science Foundation grants CCF-0959924 and CCF- IEEE, 2016.
1265178. [16] S. Lohar, J. Cleland-Huang, and A. Rasin. Evaluating
the interpretation of natural language trace queries. In
References Proc. of the Int'l Working Conf. on Requirements Engi-
[1] S. Adolph, W. Hall, and P. Kruchten. Using grounded neering: Foundation for Software Quality, pages 85–101.
theory to study the experience of software development. Springer, 2016.
Empirical Software Engineering, 16(4):487–513, 2011. [17] NASA. NASA Software Safety Guidebook. 2004.
[2] R. S. Arnold and S. A. Bohner. Impact analysis-towards [18] B. Nuseibeh and S. Easterbrook. Requirements engineer-
a framework for comparison. In Proc. of the 1993 ing: A roadmap. In Proc. of the Conf. on the Future of
Conference on Software Maintenance, pages 292–301, Software Engineering, pages 35–46. ACM, 2000.
1993. [19] P. Pruski, S. Lohar, R. Aquanette, G. Ott, S. Amorn-
[3] A. Begel and T. Zimmermann. Analyze this! 145 borvornwong, A. Rasin, and J. Cleland-Huang. Tiqi:
questions for data scientists in software engineering. In Towards natural language trace queries. In Proc. of the
Proc. of the 36th Int'l Conf. on Software Engineering, IEEE 22nd Int'l Requirements Engineering Conf., pages
pages 12–23. ACM, 2014. 123–132, 2014.
[4] E. Bouillon, P. Mäder, and I. Philippow. A survey on [20] T. Punter, M. Ciolkowski, B. Freimut, and I. John.
usage scenarios for requirements traceability in practice. Conducting on-line surveys in software engineering. In
In Proc. of the Int'l Working Conf. on Requirements Proc. of the 2003 Int'l Symp. on Empirical Software
Engineering: Foundation for Software Quality, pages Engineering, pages 80–88. IEEE, 2003.
158–173. Springer, 2013. [21] B. Ramesh and M. Jarke. Toward reference models for
[5] S. Breu, R. Premraj, J. Sillito, and T. Zimmermann. requirements traceability. IEEE Transactions on Software
Information needs in bug reports: improving cooperation Engineering, 27(1):58–93, 2001.
between developers and users. In Proc. of the 2010 ACM [22] P. Rempel, P. Mäder, T. Kuschke, and J. Cleland-Huang.
Conf. on Computer supported cooperative work, pages Mind the gap: Assessing the conformance of software
301–310. ACM, 2010. traceability to relevant guidelines. In Proc. of the 36th
[6] T. Fritz and G. C. Murphy. Using information fragments Int'l Conf. on Software Engineering, pages 943–954,
to answer the questions developers ask. In Proc. of the 2014.
32nd ACM/IEEE Int'l Conf. on Software Engineering- [23] S. Robertson and J. Robertson. Mastering the require-
Volume 1, pages 175–184. ACM, 2010. ments process: Getting requirements right. Addison-
[7] B. G. Glaser and A. L. Strauss. The discovery of wesley, 2012.
grounded theory: Strategies for qualitative research. [24] C. Robson. Real World Research. John Wiley & Sons,
Transaction Books, 2009. 2011.
[8] H. F. Hofmann and F. Lehner. Requirements engineering [25] J. Sillito, G. C. Murphy, and K. De Volder. Questions
as a success factor in software projects. IEEE Software, programmers ask during software evolution tasks. In
18(4), 2001. Proc. of the 14th ACM SIGSOFT Int'l Symp. on Founda-
[9] E. Hull, K. Jackson, and J. Dick. Requirements engineer- tions of Software Engineering, pages 23–34. ACM, 2006.
ing. Springer Science & Business Media, 2010. [26] J. Singer, S. E. Sim, and T. C. Lethbridge. Software
[10] IEEE Computer Society. Software Engineering Standards engineering data collection for field studies. In Proc. of
Committee and IEEE-SA Standards Board. IEEE Rec- the Guide to Advanced Empirical Software Engineering,
ommended Practice for Software Requirements Specifi- pages 9–34. Springer, 2008.
cations. Institute of Electrical and Electronics Engineers, [27] J. Smith, B. Johnson, E. Murphy-Hill, B. Chu, and H. R.
1998. Lipford. Questions developers ask while diagnosing
[11] S. H. Khandkar. Open coding, 2009. potential security vulnerabilities with static analysis. In
[12] A. J. Ko, R. DeLine, and G. Venolia. Information needs Proc. of the 10th Joint Meeting on Foundations of
in collocated software development teams. In Proc. of Software Engineering, pages 248–259. ACM, 2015.
the 29th Int'l Conf. on Software Engineering, pages 344– [28] R. K. Yin. Case study research: Design and methods.
353. IEEE Computer Society, 2007. Sage, 2009.
[13] G. Kotonya and I. Sommerville. Requirements engineer- [29] R. R. Young. Effective requirements practices. Addison-
ing: processes and techniques. Wiley Publishing, 1998. Wesley Longman Publishing Co., Inc, 2001.
[14] P. A. Laplante. Requirements engineering for software [30] D. Zowghi and C. Coulin. Requirements elicitation: A
and systems. CRC Press, 2013. survey of techniques, approaches, and tools. In Proc. of
[15] X. Lian, M. Rahimi, J. Cleland-Huang, L. Zhang, R. Fer- the Engineering and Mmanaging Software Requirements,
rai, and M. Smith. Mining requirements knowledge from pages 19–46. Springer, 2005.

View publication stats

You might also like