Prioritization of global software requirements' engineering barriers: An analytical hierarchy process

Department of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, China Punjab Information Technology Board, Lahore, Pakistan Information and Computer Science Department, King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia School of Computer Science and Technology, Chongqing University of Posts and Telecommunication, Chongqing, China Department of Information Systems, STC's Artificial Intelligence Chair, College of Computer and Information Sciences, King Saud University, Riyadh, Saudi Arabia Information Systems Department, College of Computer and Information Sciences, Imam Muhammad Ibn‐Saud Islamic University, Saudi Arabia


| INTRODUCTION
Requirements engineering (RE) is the process of clarifying the goal by defining the needs of stakeholders and documenting these needs in such a way that can be analysed and implemented [1]. The RE process is an important phase of every software development process as it provides base for the subsequent software development activities. According to Jeremy [1], the RE process consists of five core phases, namely, requirements extraction or elicitation, requirements analysis and design, requirements specification, requirements validation, and requirements management.
Over the last two decades, software firms have increasingly adopted the phenomena of global software development (GSD) [2,3]. The GSD is a software development process in which development activities are carried out by experts who are based in different geographical locations around the globe [4,5]. The GSD is an effective development paradigm as it facilitates to hire skilled human resources at a low cost [6]. By adopting the GSD, the development activities are carried out round the clock (24/7), which is significant for in-time project delivery. For example, Mighetti and Hadad [7] underlined that the GSD paradigm assists in developing quality software due to the availability of the latest technological tools in geographically distributed sites. They further indicated that across the globe, the development provides the opportunity to GSD firms to keep update with the international market.
However, besides the advantages, GSD faces several barriers, especially in requirements elicitation and management activities [8,9]. For example, Damian, et al. [10] Lloyd, et al. [9] and Khan, et al. [11] indicate that the RE faces a number of barriers in the GSD domain due to the teams distributed across the globe. Similarly, Hsieh [12] underlined the communication and collaboration barriers associated with the RE process due to geographical distance between the RE teams of different GSD sites. Furthermore, the RE teams face various barriers while deploying RE activities such as 'establishing relationships between geographically distributed team members', 'managing time differences', and 'cultural issues' [13,14].
Significant work has been conducted to understand RE in the GSD [15][16][17]. However, no work has been carried out that identifies barriers that influence the RE process in GSD projects. Moreover, no studies have been conducted that empirically prioritize barriers associated with the RE process in GSD context. This research gap motivated us to conduct a study to better understand the barriers of the RE paradigm in the geographically distributed development environment. This study has two key goals: (i) to identify the RE barriers in the context of GSD by adopting the literature review approach and to validate them with real-world practitioners using a questionnaire survey; and (ii) to prioritize identified barriers by applying analytical hierarchy process (AHP). The prioritization of barriers will assist the practitioners to consider the high priority barriers that influence the RE process in GSD projects. The prioritization will also assist the practitioners to better understand, plan and manage RE activities in GSD context. To achieve the objectives of this study, we address the following research questions:

| BACKGROUND AND MOTIVATION
To develop software projects according to the expectations of the stakeholders, the RE process plays an important role. Various studies were carried out to address the challenges of the RE process. For example, Pandey, et al. [13] proposed a model for requirements elicitation. Pandey's model contained four core phases, namely: 'requirements elicitation and development', 'documentation of requirements', 'validation and verification of requirements', and 'requirements management and planning'. The model provides a brief description to collect and analyse the requirements according to the customers' expectations [3]. Beecham, et al. [18] introduced a requirements process improvement model, namely R-CMM. The R-CMM model contains a detailed description of the RE process. The maturity level of the R-CMM model enables software firms to assess and improve their RE activities. In the software industry prospective, the R-CMM also provides the best practices which are useful to address the faced barriers during the RE process [18]. Mellado, et al. [19] introduced a security RE process. This model is based on the common criteria (CC) (ISO/IEC 15408) [19]. This standard-based model has an agreement with the requirements security at the primary phases of software development is an iterative and systematic way by using a security resources repository; together with the combination of the CC (ISO/IEC 15408). This model is useful to address the barriers related to the RE process concerning the requirements security at the initial phase of the development life cycle [20]. Various other studies were conducted to address the challenges of the RE process [20][21][22][23][24][25][26].
A number of studies have also been conducted to better understand barriers associated with RE process in GSD projects. For example, Bhat et al. [27] indicated that RE teams working in client-vendor outsouricing projects face a different that are not addressed by traditional RE processes. The study findings indicate that shared goals, shared culture, shared process, shared responsibilyt and trust are best practices to successfully carry out RE process in client-vendor offshore outsourcing relationships.
Damian and Zowghi [28] investigated RE challenges introduced by stakeholders who are geographically distributed in a multi-site organization. They found that remote communication, knowledge sharing, cultural diversity and time zone differences negatively impact the RE process in multi-site software development organizations.
Simialarly, Alnuem et al. [29] applied structured and semistrucutred interview techniques to understand RE challenges faced by the Saudi Arabian industry during GSD projects. They found documents management and competence management as key challenges faced by the Saudi Arabian organizations during GSD projects. Jain and Suman [30] carried out a systematic literature review on the GSD life cycle and reported that temporal, socio-cultural, linguisitic, communication and coordination are challenges associated with RE process of a GSD project. However, the findings of the above discussed studies indicate that no study has been conducted to prioritize the barriers associated with RE process in the GSD context. In this study, we apply the AHP to prioritize the barriers related to the RE process in GSD projects.

| RESEARCH METHODOLOGY
To meet the objectives of the study, we applied three different research approaches; (i) literature review, (ii) empirical study (questionnaire survey), and (iii) AHP. All the approaches are briefly discussed in subsequent sections and graphically presented in Figure 1

| Literature review process
To identify the barriers of the RE process from the existing literature, we used the literature review approach.

| Data sources
To extract the research articles related to study objectives, we applied search string on the most popular digital repositories. The databases were selcted using the suggestion of Khan, et al. [4], Lloyd, et al. [9] Niazi [31] that is, 'ACM Digital Library', 'IEEE Xplore', 'Wiley Inter-Science', 'Springer Link', 'Science Direct', 'IET Digital Library' and 'Google Scholar'.

| Search string
The search string was developed using keywords and their respeoctive synonyms. The Boolean operators "AND" and "OR" were used to develop search strings for different databases by concatenating keywrods and their respective synonyms. Since, different databases have unique format for writing and executing search strings, in this article, we report a generic structure of the search string. The format of the search string executed on each selected databases are given at: https:// tinyurl.com/y944gj2a. An example of the search string is given below: ('barriers' OR 'obstacles' OR 'hurdles' OR 'difficulties' OR 'impediments' OR 'hindrance' OR 'challenges') AND ('requirements engineering' OR 'RE' OR 'requirements collection' OR 'requirements elicitation' OR 'requirements updating' OR 'requirements process assessment') AND ('global software development' OR 'GSD' OR 'distributed software development' OR 'offshore software development' OR 'outsourcing' OR 'multisite software development').
The search terms were collected based on our research experience and using the references, Shafiq et al. [2], Khan et al. [6], Khan et al. [4], Lloyd et al. [9], Niazi et al. [31]. The developed search string was carefully executed on the selected digital databases, and initially, a total of 823 research articles were collected.

| Inclusion and exclusion critera
Next, we used the inclusion and exculsion criteria to select primary publications for the study. The inclusion and execution criteria were developed by adopting the criteria published in the software engineering domain [4,32]. The inclusion and exclusion criteria for the study are as follows: Inclusion Criteria: (1) Articles that are published as a journal paper, conference article or book chapter, (2) articles written in English language, (3) findings of the articles based on empirical investigation (4) include the most complete version of an article from duplicate studies.
Exclusion Criteria: (1) The articles that do not focus on the requirement engineering in GSD context, (2) articles that do not address the research questions, (3) articles that do not explore the success factors of requirement engineering in GSD (4) articles written in other language than English.

| Final selection of studies
We have applied the steps of tollgate approach proposed by Afzal, et al. [33] to select the final primary studies. By executing the search string on the selected databases and after applying the inclusion and exclusion criteria, initially, 823 studies were extracted. By carefully applying the five steps of the tollgate approach [33], finally we have selected 29 studies ( Figure 2). The list of the final selected studies is given in appendix-D.
The selected studies were carefully reviewed to extract the data to answer the RQs of the study. The first and second authors participated in the data extraction process. First, they extracted the barriers from the finally selected studies. Later, the remaining research team carefully reviewed the extracted data. We have further conducted the reliability test with the aim to check the inter-personal biasness. As a result, F I G U R E 1 Conducted research methodology. AHP, analytical hierarchy process; GSD, global software development; RE, requirements engineering AKBAR ET AL. two independent external reviewers were requested for the reliability test. They have randomly selected seven articles and carefully performed all the phases of data extraction process. We calculated the non-parametric Kendalls coefficient of concordance (W) value to check the inter-rater agreement between the authors and external independent reviewers [34][35][36]. The value of W = 1 indicates a complete agreement and W = 0 shows a complete disagreement. The calculated results of reliability test for seven selected articles shows that W = 0.87 (p = 0.003), which indicted significant similarities between the findings of the authors and the external reviewers.
The selected studies were further analysed concerning the publication years and the adopted research methodologies ( Figure 3). According to the summarized results there is an increasing trend of publication in recent years. It is also observed that 19% 'questionnaire surveys' (QS), 21% 'case studies' (CS), 12% 'content analyses' (CA), 13%) 'grounded theory' (GT), 15 (12%) 'action research' (AR), and 23% of the studies used mixed methods (MM). Hence, it is observed that MM, CS and QS are the most widely adopted research approaches.

| Empirical study
To validate the barriers explored through the literature review, the questionnaire survey approach was applied. The phases adopted to conduct data collection form from the respondents are presented in the subsequent sections.

| Survey questionnaire development
In light of the literature review investigations, we developed an online survey instrument to explore the barriers of the RE process in GSD. The survey method can obtain information from a large targeted community [4]. Also, we can obtain data through a survey method that is hard to obtain using other observational methods [37,38]. We created a close-ended questionnaire to collect information from the RE and GSD practitioners. The survey questions are based on the 17 barriers that are investigated in the literature review. We used a five-pointed Likert scale with the following feasible reactions: 'strongly agree, agree, neutral, disagree, and strongly disagree'. According to Finstad [39], there is no critical drawback in using the 'neutral' option on a fivepointed Likert scale.
Moreover, most of the researchers verified that it is a justifiable reaction to respond as neutral to any statement or subject matter [4,40]. If there is no 'neutral' option, it might force respondents to make a negative or positive decision which makes the result biased.

| Questionnaire pilot assessment
A pilot assessment of the survey instrument was conducted with five software engineering experts to get their feedback regarding the design, measurement scale, readability, and understandability of the survey questionnaire. Three experts were from the industry (i.e. belong to DCSL SOFTWARE, ITECHART GROUP, and RADIX WEB) and the other two were from academia (i.e. 'Chongqing University of Posts and Telecommunications, Chongqing, China' and 'Nanjing University, Nanjing, China'). These experts were selected as they have direct experience in the RE and GSD paradigm. They provided suggestions to modify the structure of the questionnaire by listing all the identified barriers in tabular form. Moreover, they recommended to rephrase some of the barriers and make them more understandable. We finally updated the survey questionnaire based on the feedback of the experts and the sample of the updated questionnaire survey is given in Appendix-A. We assured the survery participants that the survey data would only be used for research purposes, and the identity of the survey respondents will not be disclosed to any organization or any individual in any circumstances.

F I G U R E 2
The phase of the adopted tollgate approach F I G U R E 3 Temporal distribution and the research methodologies used in selected studies. AR, action research; CA, content analyses; CS, Case studies; GT, grounded theory; MM, mixed methods; QS, questionnaire survey

| Data sources
The purpose of this study is to identify the barriers of the RE process in GSD context. Hence, it was important to gather information from various professionals with relevancy to RE experiences in the GSD domain. The survey participants are invited using the snowball strategy [41]. All participants were invited via different sources, for example Email, Facebook, LinkedIn, and through their professional contacts. We developed a link to the questionnaire survey and sent it to the different in-contact practitioners to participate in the survey. In addition to the questionnaire survey link, an invitation letter was also attached to the request to invite their team members to participate in the survey study. The data was collected from March 2019 to June 2019. We collected a total of 56 filled questionnaires from the experts. Although feedback was manually checked to identify incomplete entries, none was identified. We found that most of the developing firms are located either in Europe or Asia. Survey respondent's positions are ranged from software developers to software project managers. All of them have vast experiences in the RE process as well as in GSD. The detail demographics of the respondents are provided in Appendix-B.

| Survey data analysis
The survey responses are summarized using the frequency analysis method. The frequency analysis is an effective way to investigate the variable groups and for ordinal and numeric information [42]. The frequency analysis is performed, and the summarized results are presented in the tabular form. The same analysis approach has been adopted by various existing studies, for example, [4,37].

| Analytical hierarchy process
AHP is a multi-criteria decision-making process used for prioritizing the influencing factor(s) of a specific project.
Various studies have already been adopted in this method to inline the complex decision-making problems and prioritize the factors of different other software engineering domains, for example [43,44]. The same approach is adopted in this study to prioritize the investigated barriers of the GSD RE process with respect to their significance. Table 1 shows the core phases of AHP.

| Classifying the goal, categories (factors), and their corresponding barriers (sub-factors)
In this phase, we decomposed a complicated problem in interlined decision elem -making problem is divided at three levels in the hierarchy structure (as shown in Figure 4). In this hierarchy structure, the basic goal is indicated at the first level; the subcategories are indicated at the second and third levels respectively.

| Development of pair-wise comparison matrix of barriers based on the expert's opinions
At this level, we used a pair-wise comparison technique [46,47]. The purpose of adopting the pair-wise technique is to calculate the weight (W) of each reported barrier and category of barriers. In a pair-wise approach, all the barriers and their categories are compared with each other at every level based on their effect on the RE process in GSD. Based on the comparison, the pair-wise matrix was developed, and weight (W) was calculated. The weight (W) indicated the priority level of each barrier within the category and among the categories. However, experts' panel (discussed in section 3.2.1) was involved in checking the comparison score of barriers within the category and among the categories. The questionnaire used to report the responses of the experts' panel is available at the link: https://tinyurl.com/ yachfe6h. For reporting the respondents' opinions, we used a 'standardized nine-pointed comparison scale' ( Table 2). The comparison matrix [44,47] was developed for each reported barrier and their categories, as shown in matrix-A (n � n).  We further explain the pairwise comparison Matrix-A as follows: We compare the two barriers, that is, BA1 and BA2 with respect to their degree of significance for RE in GSD. If BA1 has 3 greater importance than BA2, then BA2 is one-third with respect to BA1. Using the same practice we have developed the pairwise comparison matrixes for all the barrier factors and their categories in Section 4.3.

| Calculation of priority weight for each barrier category and sub-category by using pair-wise comparisons
To calculate the priority weight (W) of the investigated barriers and their categories, all the developed matrixes of the barriers were normalized. For the normalization process, all the barriers were divided by the sum of their column and calculate the average of each row. The calculated average score of each row is the priority weigh of each barrier (Section 4.3).

| Checking the consistency of the judgements
This phase is performed to check the consistency of the pairwise comparison. To do this, the following equations are used: where, λ = [sum of each column of pairwise comparison matrix] � Weight (W) vector and n = number of criteria.

Consistency Ratio ðCRÞ ¼ CI RI
where, RI = Random consistency index with respect to matrix size (Table 3).

| FINDINGS OF THE STUDY
The results and analysis are presented in this section.

| Literature review (RQ 1)
Due to potential business gains, software firms are motivated to transform their business activities from colocated (single site) development environment to GSD [6]. However, adoption of GSD is not straightforward. There are various barriers are associated with GSD, particularly for the RE process [40]. A case study conducted by Regnell et al. [48] shows that the barriers of the RE process are one of key causes for project failure. Prikladnicki et al. [8] and Nicolás et al. [49] indicated that there is a huge research gap in the RE process in the GSD context. This encouraged us to explore the RE process barriers faced by GSD practitioners.
We have reviewed the existing literature and observed that various researchers discussed the barriers faced by RE practitioners while adopting GSD. For example, Kotonya et al. [50] and Carrillo et al. [51] conducted a survey study and indicated that there is a BA1 (lack of standard and procedure for RE at GSD sites). Ali and Lai [52] highlighted the importance of the standard procedure for RE at GSD. They further underlined that due to the lack of RE standards, practitioners lack confident during the requirements collection and managing process.
Most of the existing literature indicated that the BA2 (lack of economic maturity) is a barrier in the execution of the RE process [53]. Khan et al. [4] indicated that most of the vendor firms face budget problems, and they do not consider the RE process as an important activity in software development. Due to the lack of budget, the RE process is negatively impacted, which is one of the causes of poor and incomplete requirements collection.
BA3 (Project specific constrains in distributed sites) is also one of the barrier highlighted in the literature as there are no proper platform to specify the requirements of project in distributed sites [54]. Niazi et al. [54] indicated that sources and jobs are normally assigned among various sites without measuring the project-specific constrains, which cause problem like (unequal distribution of resources and delays in recovery status).
Akbar et al. [55] underlined that BA4 (New regulations and de-regulations across the boundaries) is a major barrier in uplifting the software development process from colocated to the GSD environment. The RE team has to work in all areas of development process to set new regulations across the boundaries, which guides the distributed team about rules and standards of tasks that is private or public provision. Similarly, Kamal et al. [56] also indicate that de facto distributed regulations are required to limit the boundaries in complex systems that will help practitioners to underline their tasks easily. Paulus et al. [57] and Khan et al. [6] underlined that BA5 (Lack of trustworthiness) is a critical barrier in the context of GSD. In global distributed environment, the development activities are dependent on geographically distributed practitioners. The main reason behind lack of trustworthiness is less physical interaction. The quality of the software development may be compromised or the practitioners may neglect some important aspects if they are not interacting or sharing information with other practitioners. For example, Gea et al. [51] underlined that for being successful in the RE process (in GSD environment), the trust between the overseas teams is an important aspect. Moreover, Khan et al. [6] stated that it is challenging to develop the confidence and trust among the geographically distributed team. Niazi et al. [54] indicated that the activities of the RE process require frequent communication and coordination. They further indicated that due to the lack of frequent and effective communication, BA11 (Lack of trust) arises. Due to lack of trust, the geographically distributed teams hesitate to share the secret and important information which leads to the poor requirements collection [54,58].
In the literature, we also found that BA6 (Political factors in GSD firms) is a critical barrier for the RE process in the GSD environment. These political barriers include government policies, political stability or instability, tax policy, labour law and trade restrictions etc. According to Gea et al. [51] firms must care about political and legal systems to globalize their businesses. The firms should be flexible enough to respond changing rules and government polices while working in the GSD environment.
BA7 (Lack of workspace awareness in GSD) is also a critical barrier indicated by [2,59] in the GSD environment. In order to satisfy our requirements in GSD, we should use elements like workspace awareness to analyse ongoing activities (as we are working in shared and distributed workspace). There is a need of awareness support model, which will reduce the BA11 (Lack of trust) and the project can expose itself about ongoing tasks and activities without coordination with other practitioners.
The literature indicates BA8 (Lack of familiarity with tools and techniques) as a critical challenge for RE process in the GSD environment. Akbar et al. [60] and Khan et al. [11] pointed out that requirement elicitation methods are required as RE partitioners are less aware of tools and scheduling activities. The main practices of requirement elicitation are interviews, questionnaires, workshops and user observations that will help practitioners to gather information about certain tools and techniques, which will help them to maintain workflow in the GSD environment. Khan et al. [61] also explored the barrier BA9 (Lack of technical maturity in distributed RE teams) which is also caused by lack of technical skills and knowledge about the ongoing process. The elicitation techniques are helpful in this case also to resolve such barriers. According to Keshta et al. [62], approaches like group sessions, workshops and traditional methods of information gathering will help practitioners in resolving issues like Lack of technical maturity in the domain of GSD.
As a result of the literature survey, we found BA10 (lack of management commitment) as a critical barrier for the RE process in the GSD environment. The physical distance in GSD sites causes the lack of good relationships among both the client and vendor organizational management, which instigates the hesitation among both managements to share the confidential material [63]. BA11 (Lack of trust) is an important factor in GSD projects due to many aspects. There can be lack of trust among GSD teams due to potential inconsistency and disparities in work practices. Lack of face-to-face meetings, poor sociocultural fit and language skills also play a role in impacting the trust level among GSD teams [64,65]. For example, Moe and Smite [66] indicated that lack of trust leads to decrease in information exchange and feedback among GSD teams and results in relationship conflicts. As a result, GSD practitioners tend to self-protect and prefer their individual goals over group goals. Moreover, the study [66], also shows that due to increase in monitoring of GSD team by the managers causes more reduction in the level of trust among practitioners.
BA12 (Cultural differences) is also a major barrier while working with the RE process in the GSD environment. This issue was highlighted by John et al. [67] based on the surveybased study on RE process management. According to Tawfeeq et al. [68], the culture influences on RE process play a major role as individual and team communicates and perform various tasks. Sethia and Pillai [35] emphasized that social and cultural aspects of RE process should not be ignored as in the distributed environment, participants participating in various activities are from different cultural background. Project managers need to bridge this culture gap by giving flexibility in terms of different cultural behaviours. Furthermore, we found BA13 (cultural differences) as a major barrier in the RE process in the GSD domain. In GSD, the development teams are geographically and culturally distinct [69]. The team member for different cultures hesitates to share the information [2]. For instance, distributed sites may not be able to communicate with each other using their native languages [59]. Moreover, misunderstandings of information might occur as a result of cultural differences that can create confusion between the overseas teams [70]. Bernbach et al. [34] highlighted the language barrier between geographically distributed teams as a significant barrier for proper requirements collection. Sethia and Pillai [35] and Kumar et al. [36] emphasized that RE is a more communication dependent process, so the language difference among the geographically distributed practitioners is a major barrier in requirements collection and management procedures.
BA14 (Lack of synchronized communication infrastructure) is also a significant barrier towards the successful execution of RE activities in GSD. Akbar et al. [14] mentioned AKBAR ET AL. that the different usage of communication channels and technological tools hinder frequent and effective communication. For example, in China the social communication mediums (e.g. Facebook, WhatsApp and IMO) are blocked. Inayat et al. [71] also recommended the same communication medium for frequent communication among geographically distributed teams.
In the literature, BA15 (Lack of geographically distributed client-vendor relationships) is reported as an important barrier for the RE process in the GSD domain. Anitha [72] and Kumari et al. [73] have shown that the relationship between clients and vendors is an ongoing, long term contractual agreement where vendors provide one or more comprehensive IT services to clients.
BA16 (Lack of mutual understanding among distributed RE teams) barrier arises due to the cultural gap between geographically distributed teams. The literature indicates that 'mutual knowledge' is a critical problem in the RE process while working in the GSD environment. According to Ali and Lai [52], to maintain communication and understanding between distributed teams, mutual understanding plays a key role, which can be maintained by continuous feedback using various agile approaches.
Literature on the RE process indicate that BA17 (Language barriers between geographically distributed teams) is an important barrier for GSD projects. For example, Kumar et al. [36] emphasized that the teams working from geographically different areas highlight barriers like 'language' and 'time zone'. To carryout the RE process in GSD context, the grouping of participants in the team plays a vital role. The interviews and virtual meetings should be conducted before assigning tasks to various teams. The project manager should note the language barriers while grouping the teams.
The identified barriers were further mapped in Ramasubbu's [74] software process improvement framework. We have adopted this framework for categorization because it is specifically developed for the improvement of the GSD process. The objective of categorization of the investigated barriers into three core categories is to develop a hierarchy structure which is required to apply the AHP approach. Ramasubbu [74] developed a framework by categorizing the software process improvement factors into their core categories. For the mapping process, we have considered the core three categories that is 'Organizational management', 'Human resources management' and 'Coordination' and classify the investigated barriers. To classify the investigated barriers, we have used the coding scheme of the GT approach [75], to identify, label, and group the identified barriers. The four core steps of the coding scheme (i.e. 'code', 'sub-categories', 'categories' and 'theory/theoretical model') were applied for mapping the barriers into three key categories. The mapping team consists of the first, second, and fourth authors of this study, and the third author validates the mapping process. In the first step, we assigned a unique code to each investigated barrier. In the second step, the investigated barriers were then categorized into two general categories. In the third step, the barriers were mapped into three categories, that is, 'Organizational management', 'Human resources management' and 'Coordination'. The categorization results are presented in Table 4.
The key aim of this categorization is to design a hierarchy framework that is used to perform the AHP analysis. Besides, this categorization will also provide the knowledge based for academic researchers and practitioners to consider the most significant area of the RE barrier of GSD context.

| Results of the empirical investigation (RQ 2)
To get the insight of the industry practitioners concerning the barriers investigated via literature review, we carried out the questionnaire survey study. The questionnaire was based on the investigated RE barriers in the context of GSD. The collected data from the survey participants were categorized as 'positive', 'negative', and 'neutral'. The positive category consists of 'strongly agree (S.A)' and 'agree (A)', negative category included 'strongly disagree (S.D)' and 'disagree (D)', and neutral (N).
The positive category presented the responses of the survey participants who considered the enlisted barriers as challenges, while negative categories presented the responses of the survey respondents who did not consider the investigated barriers as challenges for the RE process in GSD.
The neutral category presents the responses of those participates who were neutral about the impact of enlisted barriers. We have also provided open ended questions for the survey respondents to mention additional barriers that were not identified in the survey. However, no additional barriers were reported by the participants of the survey.
According to the results presented in Table 5, BA6 (political factors in GSD firms, 16%) and BA9 (lack of technical maturity in distributed RE teams, 16%) were the highest cited barriers in the neutral category. This indicated that 16% of the respondents were not sure about the effect of BA6 and BA9 on the RE process. BA11 (lack of trust, 14%), BA13 (environmental constrain at GSD sites, 14%), and BA14 (lack of synchronized communication infrastructure, 14%) were the second-highest reported barriers in the neutral category.
The survey results indicated that majority of the survey respondents agree with the core categories of the enlisted RE barriers. For example, 94% of the survey respondents agree with the C3 (Coordination) category as the mapped challenges with C3 are negatively influence the RE process in GSD. Moreover, we noted C1 (Organizational management, 91%) and C2 (Human resources management, 89%) are the second and third most important categories of RE challenging factors, respectively.

| Analytical hierarchy process (RQ 3)
To prioritize the investigated barriers of the RE process and their categories, we applied the AHP phases, presented in Section 2.3 ( Table 2). Note: The bold shows the mapping of requirements engineering barriers into key categories.
AKBAR ET AL. -9 Phase-1: (classifying the goal, categories (factors), and their corresponding barriers (sub-factors)) The investigated RE barriers were categorized into three levels (e.g. Figure 3). The aim of the study is indicated at the first level (i.e., prioritizing the barriers), the categories of the barriers are presented at the second level, and the corresponding barriers are presented at the third level. The proposed hierarchy structure of this study is indicated in Figure 5.
Phase-2: (developed pair-wise comparison matrix of barriers based on experts' opinion) Based on the opinions of the experts, the pairwise comparison was conducted. The pairwise comparison is useful to check the degree of significance of a barrier with respect to other barriers. The matrices are developed to calculate the priority weight (W) of each barrier. The comparison matrix of 'organization management', 'human resources management', and 'coordination' are presented in Table 6, Table 7, and Table 8, respectively. The pairwise comparison among the categories of the barriers is presented in Table 9.
Phase-3 (calculation of priority weight (W) of each barrier) In order to determine the priority weights of each barrier, we adopted the following two steps: � Normalization = Divide each element in every column by the sum of that column � Weight (W) = Calculate the average of each row Phase-4 (determining the consistency of the matrix) In order to determine the consistency of the matrix, we have adopted the following steps: In this phase, the local weight (LW) of each barrier was calculated. The local weight presented the priority level of a barrier within its category, for example, the local weight of BA1 (Lack of standard and procedure for RE at GSD sites, LW = 0.36) (Table 11) indicated that BA1 has the highest priority in the organizational management category. Hence, it is ranked as first and declared as having the most priority barrier in the organizational management category. BA2 (Lack of economic maturity, LW = 0.11) and BA4 (New regulations and de-regulations across the boundaries, LW = 0.11), declared as the second highest priority barriers in the organizational management category. Similarly, the LW of human resources management presented that BA7 (Lack of workspace awareness in GSD, LW = 0.15) is declared as the highest priority and BA8 (Lack of familiarity with tools and techniques, LW = 0.14) is declared as the second most prioritized barrier of the RE process in the context of GSD. Moreover, Table 11 shows that BA14 (Lack of synchronized communication infrastructure, LW = 0.17) is standout as the highest priority and BA16 (Lack of mutual understanding among distributed RE teams = 0.11) is declared as the second highest priority barrier of the RE process in GSD environment.
Phase-6: (Determining the global weight) The global weighs (GW) of all the reported barriers were calculated ( To successfully implement the activities of the RE process in the GSD environment, all the investigated barriers need to be addressed effectively. The AHP analysis provides the priority order to which the investigated barriers need to be addressed for the successful implementation of the RE process in GSD. The AHP results (Table 11) show that the C3 (coordination, 0.58) is declared as the highest priority category of the investigated barriers. This indicated that the barriers of coordination category are more significant to address on priority basis. We further noted that C2 (organizational Management, 0.31) is declared as the second and C1 (Human resources management, 0.11) is the third most imperative category of the identified RE barriers. Figure 4 demonstrated that BA1 (lack of standard and procedure for RE at GSD sites, GW = 0.112) is globally declared as the highest priority barrier for RE process in the context of GSD. BA1 is also declared as the highest priority barrier in the organizational management category.
Phase 7: (prioritizing the barriers) The final phase of AHP is the prioritization of investigated barriers. However, based on the global ranking (Table 11), we arranged the reported barriers according to the priority. The final prioritized list of barriers is presented in Table 12.

| SUMMARY OF THE STUDY
The objective of this study is to identify and prioritize the barriers of requirement engineering process in the context of GSD. To address the objective of this study, we have conducted a literature review study on a total of 17 barriers. The investigated barriers were classified into core three categories (i.e., 'organizational management', 'human resources management', and 'coordination'). The categorization results demonstrated that most of the barriers are related to the coordination category.
Besides, based on the extracted list of the barriers, we developed an online questionnaire for survey study with RE practitioners and 56 complete responses were collected from the survey participants. According to the survey results, the identified list of RE barriers form the literature are in-line with the real-world industry practitioners. Hence, investigated barriers negatively influence the RE process in GSD context.
At the final step of this study, we applied the AHP technique to prioritize the reported barriers and their categories. The results of AHP indicated that 'coordination' is the most significant barrier category for the successful execution of the RE process in the context of GSD. Furthermore, the results demonstrated that BA1 (lack of standard and procedure for RE at GSD sites), BA14 (lack of synchronized communication infrastructure), and BA16 (lack of mutual understanding among distributed RE teams) are ranked as the highest priority barriers for the successful execution of the RE process in the GSD environment. The same procedure is adopted to calculate the weights and to check the consistency of matrix in all other categories; C2 ('human resources management') category in Table 7, C3 ('coordination') category in Table 8, and in between the categories as well in Table 9. This indicated that the organizational management must take initiatives to fix this BA1 on a priority basis, which is important for the successful execution of the RE process in GSD. Similarly, BA14 (Lack of synchronized communication infrastructure, GW = 0.098) is globally declared as the second highest priority barrier for the RE process in the context of GSD. BA14 is ranked as the first in the coordination category. Moreover, the results presented in Figure 4, indicated that BA16 (Lack of mutual understanding among distributed RE teams, GW = 0.064) is globally ranked as the third highest priority barrier for the successful execution of the RE process.

| STUDY IMPLICATION
The study provides the state-of-the-art overview of the RE process barriers faced during geographically distributed development environments. The categorization of the barriers is useful for both researchers and practitioners to concentrate on the most related barriers' category. Besides, the study provides the prioritization of the investigated barriers which is helpful to examine the most significant barrier within their particular category and in overall RE process. This prioritization is helpful for the practitioners to consider the highest priority barrier that is significant for the successful execution of RE activities in the context of GSD. The study findings also assist the researchers to consider the most prioritized barriers in their future research work.

| THREATS TO VALIDITY
One of the potential limitations of the study is related to the incompleteness of the identified barriers associated with the RE process in the GSD context. We mitigated this limitation by using keywords, their synonyms, to build the search string. Furthermore, we used inclusion and exclusion criteria to select the final list of primary studies which were used for to identify the barriers associated with the RE process in a GSD project.
Another potential limitation is with reference to the small sample size. In our survey, we collected data from 56 participants which is in line with the sample size used by other researchers in their empirical studies [2,4,47].

| CONCLUSION AND FUTURE WORK
GSD is a widely used development paradigm in the software industry. The importance of RE in GSD motivated us to investigate and prioritize the barriers faced by the practitioners during the implementation of the RE process. Through the literature survey, we identified a total of 17 barriers. Based on the investigated barriers, we carried out a survey questionnaire to validate the investigated barriers. According to the responses of the survey participants, all the investigated barriers have a negative effect on the RE process. To prioritize the investigated barriers, we adopted the AHP technique. The AHP results showed that 'coordination' is the most significant barrier category for the successful execution of the RE process GSD. The findings of the study provide a framework that can help both researchers and practitioners to focus on the high priority barriers associated with the RE process in the GSD context.
As part of our future work, we plan to develop a software RE maturity model (SRE-MM) in the context of GSD. The SRE-MM will be based on the existing maturity models of other software engineering domains [18,31,37,76], as shown in Figure 6 We believe that SRE-MM will assist practitioners to manage RE process in GSD projects, better. Furthermore, we also plan to conduct a study to identify the success factors and the best practices of the RE process which are useful to address the components of the proposed SRE-MM model.