Agile Software Development (ASD) is commonly used in environments where people have to solve complex adaptive problems. Agile methodologies (e.g. Scrum [1] , Kanban [2] or Extreme Programming (XP) [3] ) are often combined with Human-Centered Design (HCD) activities [4] when applying them in industry. One objective of this integration of methodologies is to evolve to a value-driven organization where systems can be developed to fulfil user needs by offering a positive User Experience (UX) [5] .

Focusing on value delivery becomes an emergent aspect in system development due to the increasing competition in industry. Therefore, established Requirements Engineering (RE) approaches need to be adapted to the requirements of ASD. Analyzing existing literature reviews related to agile RE [5] [6] [7] [8] , we can observe that a couple of authors propose different process models for agile RE. Such process models have in common the agile values on which they rely [9] :

• Individuals and interactions over processes and tools;

• Working software over comprehensive documentation;

• Customer collaboration over contract negotiation;

• Responding to change over following a plan.

Nevertheless, these process models for agile RE differ in how the way of working is described. For instance, which methods or artifacts are recommended to use for discovering, refining, prioritizing, reviewing or documenting requirements in an agile environment. We can build an abstract view on this field by means of a metamodel so as to gain a better understanding of agile RE and reduce the complexity of existing process models. Moreover, the analysis of related work shows that the description of agile RE management in the literature varies a lot from each other and that there is no common language. The metamodel for agile RE helps us to describe key concepts of agile RE management. This gives us a common definition for this research area and supports us in the development of a unified language. As the environment where agile is used is complex, due to the number of diverse people working together in different roles to create one system, we need an extremely adaptive metamodel for describing agile RE on an abstract level.

This paper presents a metamodel for agile RE (agile RE metamodel), aiming to build an abstraction layer about the variety of existing process models in the field. The agile RE metamodel allows building a common understanding of the complex research field of agile RE. In addition, we provide two instantiations to domain specific models for agile RE by means of two cases run in industry: i) case a discusses on a Scrum-based process model for developing several systems in the E-government area; and ii) case b presents a Kanban-based process model used to relaunch an internet-based newspaper portal.

The paper is structured as follows: Section 2 summarizes the related work and points out the research gap. Then, Section 3 describes the applied method for creating and instantiating the metamodel. Section 4 presents the metamodel for agile RE. Later, Section 5 discusses on the instantiation into process models. Section 6 analyzes the results and limitations of this research and finally, Section 7 states conclusions and provides an outlook on future research.

We conducted a systematic literature review [5] in order to analyze the state-of-the-art of agile RE with strong focus on stakeholder and user involvement. In particular, we investigated what approaches exist to involve stakeholder in the process, which methodologies are commonly used to present the user perspective and how requirements management is carried out. In sum, 27 papers were included in the study [5] and analyzed according to our predefined research protocol.

The results of the state-of-the-art analysis revealed that the research field of agile RE is very close to current work practices in companies, since most of the included studies report results from case studies (19/27 = 70%). In addition, we observed that agile RE is a complex research field with a lot of different cross-functional influences, for instance, from the fields of HCD, ASD or RE.

However, we also identified some gaps in present literature. We learned that building a shared understanding concerning the user perspective is not very well established in agile environments. Thus, we were not able to find a common process model for stakeholder and user involvement. Nevertheless, organizations deal with those shortcomings by integrating additional methodologies like HCD, Design Thinking, Contextual Inquiry or Participatory Design.

Furthermore, we found studies presenting process models for agile RE [10] [11] [12] [13] [14] . All these process models share the continuous management of requirements by involving stakeholder and user. The most related ones are highlighted below.

Memmel et al. [10] develop a XP-based process model called Cross-discipline User Interface and Software Engineering (CRUISER). The process starts with an Initial Requirements Up-front Phase (IRUP) whose results are agile models that describe user needs by means of agile techniques (e.g. essential use cases, scenarios and prototypes). The gathered information is elaborated and processed during the different phases of CRUISER.

Kautz [11] carried out a case study with the aim to investigate user and customer involvement in ASD. He does not claim to propose a process model but the results of the study present an implicitly applied process model. Kautz integrates Participatory Design activities in XP. The agile team can detect problems concerning misunderstanding of requirements early before they can grow into larger problems by the application of an onsite customer as well as iteratively reviews with users and customers.

Maguire [12] extends the HCD [4] framework for ASD. The HCD process consists of the following steps: plan the HCD process, understand and specify the context of use, describe user requirements, produce design solutions to meet such requirements and evaluate designs against them. He suggests a set of agile techniques that can be used in each step. Besides, he proposes a couple of artifacts that are generated while using agile techniques.

Rivero et al. [13] propose a process model known as Mockup-Driven Development (MockupDD). Their approach supports Model-Driven Web Engineering (MDWE) and is integrated into Scrum. In the beginning of MockupDD, a quick requirements gathering stage is performed resulting in a set of user stories. Based on this, costumers and users create mockups to represent these user stories graphically. These mockups lay the foundation for the following modeling process.

Olsson et al. [14] design a process model based on a conceptual model named Qualitative/Quantitative Customer-Driven Development. It stresses the need for combining qualitative customer feedback in early stages of development with quantitative observations in later stages. Olsson et al. treat requirements as hypotheses that are validated with customers before development. Hypotheses are derived from business strategies, innovation initiatives, customer feedback and on-going validation cycles.

Analyzing the commonalities of the proposed approaches, we can conclude that all these process models have in common that they utilize different types of artifacts, meetings, methods and roles (referred to as agile techniques) for the management of requirements in an agile environment. In this context, the related work lacks in: 1) providing guidelines that help to choose an appropriate set of agile techniques for solving a specific agile RE problem [15] ; and 2) providing generic concepts on an abstraction layer. These generic concepts are important in today’s business world since companies utilize different types of process models for diverse teams.

This leads to increasing complexity within an organization (e.g. scaled organization or other teams working with sequential approaches, like waterfall model). This paper proposes a metamodel for agile RE to handle this complexity (see Section 4). It is evaluated by an instantiation into concrete process models (see Section 5). To the best of our knowledge this is the first metamodel for agile RE.

The objectives of this paper are contributing a metamodel for agile RE, which allows building a common understanding of the complex research field of agile RE; and showing how the metamodel can be used in industry by means of tailoring domain specific models.

In the following subsections, we will explain how we developed the agile RE metamodel iteratively. Subsequently, we will demonstrate how the metamodel can be applied to industry by means of instantiations.

The agile RE metamodel (see Figure 3) visualizes the big picture of agile RE without providing a concrete domain specific model. We can use it to derive a domain specific model tailored according to the environment where the system is developed. Moreover, the agile RE metamodel represents a set of generic concepts without describing a domain specific model itself. Therefore, it builds a framework for a process. Hence, the domain specific model is an instantiation of the metamodel (see Figure 2).

The agile RE metamodel looks at the cross-functional area of agile RE that consists of HCD [4] (pink, metaclasses Context of Use and User), ASD (green, metaclasses Stakeholder, Agile Team, Methodology and Impact), Domain Driven

Design [20] (DDD, orange, metaclasses System and Domain) and RE [21] (grey and metaclass Agile RE Activity). The blue metaclasses (Organizational Environment, Agile RE Problem, Agile RE Pattern and Requirements) are part of our contribution to the body of knowledge of agile RE. The initial version of the agile RE metamodel is presented in [16] , whereas Figure 3 represents the optimized version.

The agile RE metamodel can be tailored to domain specific models in order to be applicable to industry. In the following subsections, we will present two cases in which we have built instances of the metamodel. The first process model (case a) is carried out in an environment that uses Scrum for the development of a web application for E-government, with the aim to solve the problem of discovering services (E-government services). In contrast, the second process model (case b) describes an environment that uses Kanban for the relaunch of an internet-based newspaper portal. We have decided to show these two cases, because they demonstrate the adaptability of the agile RE metamodel in terms of tailoring domain specific models for RE based on different agile methodologies.

One of the most important benefits of the proposed metamodel for agile RE is cross-functionality. Agile methodologies are based on cross-functional collaboration of people from different disciplines (e.g. programming, test or UX design). It is essential to consider all the disciplines while creating a process model for agile RE, for becoming a value-driven organization. The proposed metamodel for agile RE supports the understanding of this complexity and enables companies to become clear about the influencing parameters. Furthermore, researchers as well as practitioners can use the metamodel for evaluating their existing process models for agile RE aiming to improve them.

The application of the agile RE metamodel allows us to reflect on the agile RE problems that happened in the presented cases (see Section 5.1.2 and Section 5.2.2) as well as to find out appropriate solutions for handling them.

Once these two cases are analyzed, we can figure out some differences in terms of the elaboration of information related to specific stereotypes. Basically, the context of use is deeper explored and more elaborated in case b (see Section 5.2.2). This occurs because of the integration of HCD, as in HCD the context of use is meaningful in terms of user research. Taking this as an example, we can discuss whether it makes sense to put some effort in the analysis of the context of use in case a, and discuss what impact this analysis might have on the value delivery. Besides, they also differ in the manner of how people are related among them. In case a, for instance, the user is part of the agile team, since some members of the agile team use the system, such as citizens or employers of the Ministry of Culture. Finally, another difference can be observed in the relation between the methodology and the agile RE pattern. In case b, the agile RE pattern is part of the methodology, since the agile technique Kanban board is rooted in the Kanban methodology, whereas the agile technique in case a (DS4aRE process) is a new technique independent of a methodology.

Nonetheless, the analyzed cases in this work present some shortcomings in terms of the analysis of the organizational environment. We have to state that we were not able to gather all the data the metamodel would have the capacity to include due to time spent when the case studies were carried out. The final version of the metamodel was created in July 2017, whereas the case studies were conducted before that time.

As opposed to this, both cases have confirmed that domain specific models allow us to receive a detailed view on complex organizational environments. This detailed view enables us to identify problems and impediments that have to do with people, processes or tools. Furthermore, it helps us to compare specific excerpts of different organizational environments with the aim to reflect on all the aspects that run properly and should be shared.

The use of patterns is a reasonable approach since agile RE is a problem-solving approach to ASD. The usage of patterns is widely spread in the field of software engineering [30] . Most of the proposals rely on the approach by Alexander et al. [31] . The concept of agile RE patterns presented in the agile RE metamodel (see Figure 3) is also based on Alexander et al.’s [31] understanding for the use of patterns.

In light of this, the combination of agile RE problems and their corresponding solutions can be inferred to new agile RE patterns that can make easy sharing the experience with the community. The gathered knowledge can encourage other people to solve similar individualized agile RE problems to those presented in this work.

Nevertheless, there might be some limitations to our research. We have identified the agile RE problems and agile RE patterns by empirical evaluation. It could be argued that there are some limitations in terms of the general application of the results due to the usage of an expert validation. However, the experts work for different companies and therefore they have different levels of experience in the agile transformation of an organization. Additionally, the identified key problems (Table 3) only give guidance for the optimization of existing process models for agile RE in industry and do not claim to be fully complete.

Deriving a process model for agile RE based on the proposed agile RE metamodel might still be a challenge in some organizations due to the complexity of the environment (e.g. scaled organization or other teams working with sequential approaches like waterfall model). Nonetheless, the metamodel reduces this complexity by visualizing the influential parameters that can be changed according to the explicit view.

This paper presents the agile RE metamodel that builds an abstraction layer about the variety of existing agile RE process models used in industry. The metamodel for agile RE helps us to describe key concepts of agile RE management. This gives us a common definition for this research area and supports us in the development of a unified language. Besides, we have demonstrated how we can build instances of the metamodel in order to provide process models for industry, concretely in Scrum and Kanban environments. We have presented two cases that analyze an instantiation of the metamodel: in case a, an agile RE process model for NDT-Agile (based on Scrum and Earned Value Management) is used for creating a web application that enables the Ministry of Culture to comply with Law 39/2015; in case b, a Kaban-based agile RE process model is utilized for developing an internet-based newspaper portal.

We have contributed to the software development body of knowledge by providing a metamodel for agile RE. The metamodel has several implications for both, researchers and practitioners. On the one hand, the metamodel can be used to model new process models for agile RE that are strongly focused on value delivery and therefore can be classified as value-driven process models. On the other hand, existing process models can be evaluated as well as improved by the metamodel.

Future research may specifically deal with providing practitioners and researches with tool support so that they can improve their agile RE process models based on the proposed agile RE metamodel. We would like to encourage them to improve their agile RE process models. To this end, we are currently working on a web application that provides guidance for solving agile RE problems by choosing the proper agile technique. The objective is to share knowledge regarding agile RE by means of agile RE patterns. The web application is a knowledge base concerning the most common agile RE problems in industry and appropriate solutions.

We would like to thank Dr. Nora Koch for her support in the improvement of the agile RE metamodel and Dominique Winter for his support in carrying out the empirical evaluation related to agile RE problems. We are also very grateful to Dr. Francisco José Domínguez Mayo for the valuable discussion concerning the agile RE metamodel.

This research has been supported by the Pololas project (TIN2016-76956-C3-2-R) of the Spanish Ministry of Economy and Competitiveness.

The authors declare no conflicts of interest regarding the publication of this paper.

Schön, E.-M., Sedeño, J., Mejías, M., Thomaschewski, J. and Escalona, M.J. (2019) A Metamodel for Agile Requirements Engineering. Journal of Computer and Communications, 7, 1-22. https://doi.org/10.4236/jcc.2019.72001