Agile practices such as continuous integration and delivery promise shorter time to market and improved quality. For this reason, such practices have been adopted in many software companies. In the context of systems development, additional constraints apply, e.g., as a result of scale or parallel development of hardware and software. Traditionally, stage-gate processes with a focus on up-front requirements analysis are common in large-scale systems engineering. However, long upfront analysis is considered anti agile and there is some friction between RE (which is often considered as a waterfall activity or phase) and agile practices. This paper describes a research agenda that plans to address this friction. The work proposes a series of empirical studies to discover the information needs and related knowledge, pertinent to product development. We expect to make a contribution by establishing guidelines and frameworks that can be used to make requirements engineering a foundation for agile systems development.
Large-scale systems development companies are operating in increasingly
competitive and dynamic markets. These new business challenges call for a product
development process which is able to respond faster to changing situations. For
this reason, large systems development companies have adopted agile methods
[
Agile methods, which were originally designed for small teams, have been
applied successfully but with challenges in large organizations [
At the same time Requirements Engineering (RE), which is necessary for
project success [
Based on that, we start by exploring the area of RE with two large-scale software intensive companies that are using agile methods in their development process. We then derive, based on actual industry needs and related work, the speci c research goals we have and derive a research agenda to address the friction between RE and agile methods. 2
As a starting point, using semi-structured interviews and workshops, we have explored the problem in one telecommunications company and one automotive company. We found that companies are using agile methods and want to adapt the new agile strategy of continuous delivery and deployment. Thus increasing the focus on value creation. In this situation, there is a pattern, where teams receive the high level requirements which they break down into user stories to be implemented in each sprint. However, there are some discrepancies that prevent the delivery of value. Through thematic analysis of the transcribed interviews, we identi ed several discrepancies. The following are the key results that relate to this thesis.
Gap between customer and developer: The agile methods advocate face-to-face communication between the customers and developers. However, that is practically impossible in the large-scale context as the customer base for large-scale companies is quite wide and distributed in most, if not all, cases. Developers complain that the levels between them and the customers are so many that there is so much value lost by the time the requirements get to them. Also developers complain that they are not given access to the requirements early enough, thus increasing the probability of misinterpreting the requirements. The earlier the programmers participate in requirements engineering process, the less the risk of developing the wrong product.
Information ow supporting agility: Since the direct communication between developer and customer is very complex, there is concern and therefore a need to address the way the information should ow to (and) or from and within the organization. To receive requirements, the levels within the organizations use varying tools or artifacts that cause a loss of value during conversion. Developers for instance use user stories as a way of communication among themselves which the product owners may not understand. After development, the customers are unable to try out the products in a short time for timely feedback. They still nd a need to go through their long veri cation cycles. This could be because of the lack of avenues to give them an easy way to try out features early enough or the lack of documentation to guarantee quality of the release. Managing documentation requirements: For software-intensive systems, the regulatory requirements are one of the primary factors. These regulations and policies, which guide the development of certain software components, require some sorts of documentation to back them up. Agile methods, however, do not provide for handling such requirements.
These challenges are all addressing the con icts that agile ways of working have with the traditional RE practices. The far away customer as opposed to on-site customer for agile methods, the lack of emphasis on documentation as opposed to speci c documents for knowledge sharing in the traditional ways all disturb the ow of information between stakeholders. So the general question for this thesis is: In what way can RE be modi ed to complement the agile process of development?
The remainder of the paper discusses related work, research aim, methodology and gives a conclusion. 3
The applicability of agile methods in large-scale development of software
intensive systems is challenging [
RE is a process by which the services that the customer requires from a
system and the constraints under which it operates are established. Traditionally,
during the RE process, a requirements document was created for knowledge
sharing [
To answer the question of scaling agile methods, Ronkainen and
Abrahamsson [
To further try to understand agile methods in large-scale, more studies [
Despite the fact that works have come out to identify these challenges, not
much has been done to address them. Some works [
The aim of this research is to design a framework that consists of RE techniques that are suitable for requirements engineers to enhance the value provided through handling documentation and regulations in software-intensive systems development while also supporting agile development, continuous integration, and continuous deployment. 4.1
As a starting point, we focus on the practices and processes in use by asking the following questions: RQ A What practices can be used to manage documentation without compromising the agile structure? Since documentation serves both internal for (e ective information ow between stakeholders) and external purposes (quality assurance) to capture both we further break it down into:
RQ 1 In what way can information ow be managed so that a project organization consistently maintains value and keeps a competitive edge in product?
This can be broken down into other research questions: RQ 1A What processes ensure that all stakeholders get the same understanding of the requirements in order to deliver value? RQ 1B What practices for managing user stories ensure that the stories contain strategic value for a given sprint? RQ 2 How can agile methods be extended to cater for the quality (regulatory) requirements in software intensive systems? 4.2
The research expects to contribute to the body of knowledge of tailoring RE and agile methods to work in the systems engineering context. The contribution will be through frameworks addressing how RE should be structured and executed to cope with the possibly unique challenges in embedded systems development. Focus will be speci cally in regards to: 1. Requirements ow in the organization from customer to release | to help rms manage a uni ed understanding of value in the organization. 2. Documentation requirements handling | to help rms deal with the regulatory requirements in an agile setting. 5 5.1
The research will follow the research approach proposed by Sein et al. [
This research will build on the foundation of related work, which have been
gathered and synthesized in relation to RE [
As Inayat et al. [
Triangulation will be needed in order to establish and generalize the challenges from the di erent domains. Qualitative data will be derived from interviews with identi ed stakeholders in the di erent rms. A combination of in depth interviews with open ended questions, focus groups and workshops will be used when confronting the rm representatives. Document analysis will be another possible data source. 5.2
This research is part of a bigger project with many experienced researchers with varying backgrounds. We plan to have atleast two interviewers per session to reduce the risk of bias and misinterpretation. To ensure shared understanding of interview questions, we plan to start each interview with an introduction part in which we share our understanding of the subject with the interviewee. Anonymity will be guaranteed to the interviewees to ensure practitioners are not threatened to express their opinions. The results of the study may be limited to the case companies. However, we aim to expand our research to other companies dealing in di erent types of products. Triangulation will be used Each company's characteristics will be presented to the readers 6
It is important for RE to be done rightly for the projects to be successful. In the advent of agile practices like continuous integration and delivery, rms should be able to fully reap their bene ts without completely getting rid of their old ways but to tailor their methods to fully support the agile ways of working. The approaches or frameworks to achieve that purpose are still lacking and need to be explored.
Currently the researcher is on the verge of completing her initial problem analysis and identi cation of challenges relating to agile development in the software-intensive system context and is planning for an empirical study to analyze the problem yet deeper through engagement with the companies. Once it has been done, the researcher will analyze the data collected and triangulate to devise a solution.
Acknowledgments. This research is funded by SIDA Bright project and partly supported by the Software Center (Large-Scale RE project) of Chalmers University of Technology, Sweden.