A bibliometric analysis of Agile software development publications originating from Turkey

Agile software development has reached wide adoption in various countries including Turkey, even though from which its original cultural backgrounds differ. In Turkey, many organizations have started to adopt Agile approaches more and more in their software development processes. This interest in the country's software development is parallel to what the academic literature on Agile in the country exhibits. However, despite the prevalence of Agile in Turkey, there is a lack of sufficient secondary research and comprehensive review on Agile in Turkey, which poses a significant necessity for further investigation. Considering this gap, we performed a quantitative bibliometric analysis of Agile software development publications produced by Turkish organizations in a holistic and broad approach both for scholars and practitioners. We provide a summary of relevant academic studies that emerged in Agile research in Turkey by focusing on many aspects including bibliometric properties of papers, researchers, affiliations, venues, and thematic contents that are separated into 15 sub‐research questions. After delivering results based on the questions, we discuss the results and findings of our study and present implications regarding the findings. The main contributions of our work are twofold. First, the paper may help the readers to have a quick idea, understand the subject, and gain insight from a large volume of scientific data. Second, the paper can help readers to use these analyses to form future research.


| INTRODUCTION
It seems that Agile approaches provide various advantages including faster time-to-market, competitiveness in markets, shorter development cycles, lower development costs, and abilities to move and change quickly. 1 With a remarkable rise, Agile has already penetrated various organizations. 2 Therefore, the term Agile has become representing a widely appreciated mindset, principles, practices, and methods for various domains, 2 including software development.
Due to the increasing demand from various kinds of organizations, software development has reached a wide adoption and as a complex domain, it adopts increasingly more Agile approaches 2,3 to face the inherited complexity it bears.Agile methods have spread to almost all countries including Turkey, even though from which their original cultural backgrounds differ. 4Meanwhile, as it is put forward by Šmite et al., 5 there is an interest in understanding the use of Agile methods and practices in companies outside the locations of early adopters of the Agile methods.Throughout Agile's evolution, in parallel to its spread over the globe, research on its topics has attracted significant attention. 6ile experiences in Turkey are slightly new and visible after 2010 in the Turkish software development industry. 7However, in Turkey, many organizations have started to adopt Agile approaches more and more in their growing software development processes. 8For instance, now a search on LinkedIn with "agile coach" OR "Agile koç" OR "çevik koç" brings 465 results, and "scrum master" brings 3300 results in Turkey.Among the organizations adopting Agile approaches, in addition to small and medium-sized organizations, there are large ones including the banking, aviation, and telecommunication sectors. 4is acceleration in the country can also be seen in the number of academic papers shown in Figure 1; the number and proportion of publications in Turkey compared to the globe are increasing year by year, from 0% to 2.53% at most and around 2.1% in recent years.Numerically speaking, Turkey is still behind some countries in Europe with lower populations, North America, and Oceania in terms of the number of publications in the field.On the other hand, from the same window, Turkey produces a greater number of relevant publications than some countries with higher populations including Japan, Mexico, Indonesia, the Russian Federation, and more.Despite this prevalence of Agile both in Turkey and around the globe, there is a lack of sufficient secondary research and a comprehensive review on Agile in Turkey, which poses a significant necessity for investigation.Thus, our main goal is to give ideas and insight to the researchers of Turkey who want to work on Agile and/or want to enter this field as a newcomer, to present the relevant people to work with them, and to shed light on the future by presenting trends on the axis of the past.In addition, it is aimed to contribute to the establishment of a comparable base with other countries around the world by presenting a general picture of Turkey's situation in the field.In addition, collected information can be provided to other countries that consider collaborating with Turkey on this subject.In addition to these, it can be possible to present information and insight in a case-like manner to other countries which are in a similar position to Turkey.The fact that such a study has not been carried out so far is a disadvantage when these opportunities are considered, and this study has the motivation to fill this gap.
Given the aforementioned gap and motivation, we performed a quantitative bibliometric analysis of Agile software development (ASD) publications produced by Turkish organizations, regardless of the nationality of the authors.We aimed to characterize the topic holistically and to get a view of the current situation and possible trajectory of the literature for both scholars and practitioners.On this occasion, we aimed to provide a scope and charting as a form of knowledge synthesis with a summary of all relevant academic studies and provide patterns, mapping, themes, trends, and potential and possible development avenues that emerged in the Agile research in Turkey.More specifically, we focused on certain questions to assess distinct facets of Agile publications from Turkey: (1) Who are the prominent researchers in the field of Agile?(2) Which institutions are most active?(3) What are the research topics in the field?(4) Which countries are mostly collaborating with authors from Turkey?
(5) What are the trends in the number of publications on Agile in Turkey compared to the globe?These questions then evolved into the detailed research questions (RQs) listed in Section 4.1.
F I G U R E 1 Proportion of publications by Turkey.
The rest of this paper is organized as follows: Section 2 delivers a background for bibliometric analysis techniques and related previous research.Section 3 depicts the adopted research method.Section 4 delivers results and Section 5 summarizes the outcomes of each RQ.Section 6 evaluates the results discussing them.At last, Section 7 consists of the conclusion and limitations.

| Bibliometric analysis
Bibliometric analysis is a quantitative and statistical analysis method that started to be employed in the 1950s 9 and is a utility for synthesizing research evidence and categorizing or grouping existing literature in terms of its nature, features, and volume. 10It is aimed to measure scientific research productions in the form of articles, publications, citations, and patents, which enables an evaluation of research, scientists, institutions, countries, and so forth. 11,12It is mainly used to explore emerging trends and social and intellectual structural relationships and patterns among different research constituents (e.g., authors, countries, institutions, and topics) in the extant literature. 13 has gained immense popularity in recent years, thanks to the advancement, availability, and accessibility of scientific databases such as Scopus and Web of Science (WoS) allowing to acquire large volumes of bibliometric data relatively easily. 13Bibliometric studies can advance a field in novel and useful ways, enable and empower scholars to obtain a one-stop overview, identify knowledge gaps, generate novel ideas for investigation, and offer their contributions to the field. 13,14It can focus on generating an informative summary of a specific research field. 15When knowledge develops extremely, rapidly, and massively like in software engineering, such knowledge synthesis reports using large volumes of data enable analysis promptly. 10,13, like systematic literature reviews (SLRs) and meta-analysis, provides meticulous and transparent methods to identify and analyze relevant literature to answer RQs. 10,16Unlike bibliometric analysis, SLRs require a narrow, confined, or niche research scope of study and a smaller number of papers to review. 13,17SLR poses interpretation bias resulting from massive qualitative techniques used. 18Bibliometric analysis, on the other hand, uses quantitative techniques and thus decreases the bias and is suitable when the scope of review is broad and the dataset is too large for manual review. 13Rather than an authoritative guideline like in SLR, 17 in bibliometric analysis, the data extraction process is not elaborated in-depth, RQs can be broader, inclusion/exclusion criteria can be evolved in a post-hoc manner, and it is possible to deal with broader topics, neglect quality of studies, and leverage manners and general themes to sum the results. 19,20As a similar method, meta-analysis summarizes the empirical evidence of relationships between variables while uncovering relationships that are not studied in existing studies. 13

| Related work
There are some attempts and an extensive body of literature to holistically synthesize research findings on Agile. 21Some papers focus on specific topics including Agile effort estimation, 22 measuring project and quality aspects in Agile, 23 Agile requirement management, 24 Agile software maintenance, 25 large-scale Agile transformations, 26 and quality aspects of Agile. 27 addition, there exist some bibliometric studies on analyzing the research literature related to software development including those listed by Garousi. 28In the context of Agile, there are some bibliometrics studies such as those focusing on the impact on the productivity and efficiency of ASD, 29 research themes emerged in the Scrum scientific literature, and methods, roles, artifacts, and objectives appeared in the context of the themes, 10 Lean-Agile education, 30 Agile IT governance, 31 venues, subjects and type of research methods in the field of Agile, 32 risk management in Scrum projects, 12 and so on.
Ozkan, Erdogan, and Gök 21 provide a bibliometric analysis of ASD publications by focusing on some aspects including bibliometric properties of papers, researchers, affiliations, and thematic contents.Bayram et al. 33 conducted a bibliometric analysis to find how many SLRs were published related to ASD between 2013 and 2018, which institutions and countries are most active in this regard, and the social network structure of countries and institutions in SLRs.Barud et al. 34 conducted a bibliometric study and a literature review on Lean in information technology departments or companies.However, as far as we know from the result of our narrative literature review to search for possible related bibliometrics studies focusing on Agile publications, the number of bibliometrics studies focusing on Agile is limited.
Siakas and Siakas 35 state that Agile works best in democratic organizations represented by Nordic and Anglo-Saxon cultures where Agile methods like Scrum originated from.The values in the manifesto and Scrum imply the underlying values of the authors, who are from an Anglo-Saxon culture, which include low power distance, high individualism, and low uncertainty avoidance. 36Moreover, Palokangas 36 explicitly states, "The rest of the world is more about high Power Distance and collective cultures, making adoption of these values harder for them."Turkish culture melting the West and East, being treated mostly as the Eastern, is influenced by its cultural background such as in the selection of particular Agile frameworks and implementations of Agile and Scrum originated from the West. 4lthough there are some studies using data (interviews, surveys, and so on) from Turkey, there is limited research on understanding Agile development in the Turkish software development industry. 4Among the limited current research, two of them 28,37 provide knowledge about general software engineering practices in Turkey.Garousi et al. 37 aim to understand the cross-factor correlation of various software engineering practices versus practitioner demographics including their companies and projects.They found that practitioners employed by smaller companies favor Agile more than waterfall-like development which is popular with large companies.Agile/Lean development is not popular in the military and defense sector, but it is the most popular among the participants developing software for engineering/manufacturing, IT and telecommunication, and health sectors.According to Garousi et al., 28 the most followed software development methods are waterfall life cycle, incremental development, and Agile/Lean development with adoption rates of 53%, 38%, and 34%, respectively.Pair programming, which is a highly praised popular practice within Agile methods, is not performed as frequently in Turkey.
Çetin and Durdu 38 conducted a survey to determine how Agile and traditional approaches are used in organizations in the Turkish software industry and how software developers perceive these two approaches by using data collected from 74 participants from various sectors.According to the survey results, the rate of application of Agile methods in Turkey is close to traditional methods; however, there is an increasing inclination towards the Agile approaches.Software developers think that Agile methods have many benefits; however, they think that they are more suitable for small teams and pose challenges in terms of project management.Similarly, Çalıs ¸kan et al. 39 surveyed 193 people from different companies in the software industry in Turkey.According to the results, the participants using Agile approaches suggest that Agile increases efficiency, quality, and customer satisfaction.On the other hand, participants using the classical approach are determined to be impartial on these matters.
There is also a report called Annual Agility Report, 40 dedicated to Agile in Turkey.Its content is similar to the analyses in the Annual State of Agile Report by VersionOne Inc. 41 Starting from 2012 to 2020, it is prepared annually by Agile Turkey established in 2008 as a non-profit organization.According to the last year's report, the top challenges regarding Agile in organizations are about shifting organizational culture and resistance to change.Inconsistent processes, and practices across teams, insufficient management support, and sponsorship are among other challenges.
There are some other studies on Agile gathering data from people and organizations in Turkey, yet with no further analysis specific to Turkey's context.We have encountered only one study, Gelmis ¸et al., 4 which explicitly investigated the Turkish culture concerning Agile in Turkey.In their study, Gelmis ¸et al. 4 analyzed the impact of Turkish national culture on how Agile practices are applied in Turkey by applying semi-structured interviews from three sectors.According to the study, Agile remains in many of the organizations' agendas in Turkey due to its popularity, but the Agile mindset is not fully understood.It is stated that Turkish people tend to keep up with popular trends, business fashions, and quick wins like productivity, speed, and attracting people to their organizations, even for their interests.In Turkish culture, there are some issues with open culture, feedback loops, team empowerment, process improvement, fault, and problem acknowledgment (they pretend to be successful and hide their failures), which leads to problems with building trust, transparency, and an open environment and providing psychological safety.According to the study, it may be mainly due to prevalent bureaucracy, traditional management style, organizational hierarchy, adherence to roles, obsession with titles, and authority in organizations.It is expressed that Agile transformations can be challenging for organizations in Turkey with its socioeconomic conditions and lack of social security, especially when implemented by Scrum, which comes with a radical, not evolutionary, but revolutionary approach and destroys what the Turkish nation regards important, titles and "hard-earned" positions.Then, they try to apply the existing agility to a hierarchical structure, with some workarounds to keep titles inside or around the Agile teams in Turkey.
There are already some statistical information and graphics of the WoS and Scopus databases for a given topic.However, certain fundamental aspects make our study distinct from the analyses in these databases.First, after the searches made with search criteria in these libraries, the studies that are unrelated to identified subjects should go through a manual selection process by a human.For example, when the initially returned results for our study are compared with the final dataset post-selection, it is seen that almost half of the initially returned dataset, 49% of the total, is eliminated due to irrelevance.The second main problem with using such libraries directly in this regard is that the raw data can be deficient and unstandardized.Such data are frequently seen in our study, especially with names including special characters and middle names and with the avenues of publications.The third major problem is that although the data is summarized by those libraries, there may be a need for reports to present additional and different perspectives about the data.These are the main reasons to continue with the manual operation rather than directly using ready-to-use features of the libraries.
Therefore, our study is mainly on two counts.Firstly, our study is the first to review the current bibliometrics of Agile publications from Turkey.Secondly, whereas the majority of bibliometrics studies use data obtained from databases directly, our study uses the initially obtained data after a manual human operation to refine it.The data and the results based on it affect our results positively; they could also provide insight to scholars about the possible direct impact of such a step on their results.

| RESEARCH METHOD
We followed the procedure for the bibliometric studies proposed by Donthu et al. 13 and applied by our previous study 21 to conduct this study.
The procedure includes four steps: (1) Define the aims and scope of the bibliometric study, (2) choose the techniques for bibliometric analysis, (3) collect data for bibliometric analysis, and (4) apply the bibliometric analysis and report the findings.
According to the procedure by Donthu et al., 13 the research process starts with defining the research aims and scope of the bibliometric study.The scope of the study should be large enough to warrant the use of bibliometric analysis.In the second step, bibliometric analysis techniques should be chosen to meet the aims and scope of the study.In the third step, data required for the selected bibliometric analysis and RQs are gathered from adequate databases by using relevant search terms determined to exemplify the scope of the study.In this step, the extracted data are cleaned against errors like duplicates and erroneous entries, as these databases are not exclusively designed for bibliometric analysis.At the end of this step, it is assumed that the final dataset fulfills the requirements of the bibliometric analysis techniques chosen for the study.In the final step, the bibliometric analyses are run and the findings are reported in a clear way for readers, explaining the peculiarities and implications, and aligned with the target outlet for publication.
The remainder of the section is concerned with the steps applied in our study and potential threats to validity.

| Defining aims and scope
This study aims to review and analyze studies that have been produced by authors from Turkey within the scope of ASD.Thus, we set the main goals related to our research as (1) identify the ASD studies produced in Turkey and (2) analyze and synthesize the studies' results.To specifically identify the RQs and to support our bibliometric analysis, we performed multiple iterations considering the following points: (1) We investigated the results of similar studies, (2) we examined what other analyses could be inferred from the data we obtained, and (3) we looked at the statistical information and graphics of the WoS and Scopus databases.After some iterations, based on our study's goals, we finalized the RQ as follows: RQ1. Bibliometric properties of papers: What are the bibliometric properties of the papers in terms of the trend of volume, publication types, languages, research methods, contribution types, and citation analysis?RQ 1.1.Trend of the volume of publications in Turkey compared to the globe: What is the trend of the proportion of publication numbers in Turkey compared to the globe?This is measured by using merely the Scopus data and without any manual cleaning for the data that can be similar to but is different from the main data set determined throughout our study.The ratio information was calculated by dividing those produced by Turkey by those produced globally.To identify what thematic clusters the papers pose, instead of using bibliographic coupling (through references of the papers), co-citation analysis (through citing publications to the particular papers), or automatic co-word analysis (mostly based on frequency derived from contents), we preferred to use and apply manual thematic analysis which, we believe, is more accurate than the others.

| Determining the techniques
To identify the analysis metrics to create and enrich the outcomes of the analyses, in addition to what we identified as authors, we reviewed and considered those suggested/used by multiple studies included in the related work section of this paper and those which provide a comprehensive list in this regard (e.g., previous studies 13,21,43 ).We interpreted the data both objectively (e.g., paper and citation count) and subjectively (e.g., thematic analysis).Two hundred thirty of the studies that returned from our initial search are not sufficient enough in terms of volume for automated analysis by tools, which is likely to contain a significant amount of irrelevant paper (around 50% for our study).Besides, this number is high for SLR, but it is suitable for both a manual bibliometric analysis and bibliometric analysis by tool after the manual operation to select and arrange the papers and their data.Moreover, among the 118 papers, only five papers are from WoS.This number decreased the needed effort for consolidation of the data from both libraries.
The impact of a publication is determined by the most objective and straightforward measure, which is the number of citations that the publication receives. 13

| Collecting data
To harvest the relevant publications, we conducted a manual search process of peer-reviewed studies in the digital libraries of Scopus and WoS without any filter in the year range.Scopus and WoS libraries are mainly involved in systematic mapping, SLR, and bibliometric analysis such as in Tavares et al. 12 and Donthu et al. 13 Based on the scope of this study, the search query was developed with "intervention" (agile-related terms) and "population" (software domain and Turkey-related terms/filters).The related keyword includes "agile" and the most used Agile methods in both languages, Turkish and English as shown in Table 1.The equivalent of this text was also created for WoS.We further limited our results to English and Turkish languages, manually and/or by applying relevant filters to both libraries.Regarding the search location, we anticipated and were satisfied with the effectiveness of searching in metadata instead of the full text by expecting the authors to locate the relevant terms in their papers' metadata.
After defining the keywords and libraries, a pilot search was done by the first and third authors to make sure the search process that would be applied is standard across the research.Based on the scope and context of our study, for the selection of papers, the following propositions of inclusion criteria (IC) and exclusion criteria (EC) were specified based on the guideline by Kitchenham et al. 44 and applied to the papers by the authors, by focusing on the studies from software development domain and were listed in Table 2.
During the application of inclusion/exclusion criteria, the papers were examined according to their titles and, when necessary, abstracts were examined too to identify whether they are within our scope.If the abstracts were not sufficient to decide to include or exclude the papers, then, skimming through the full texts of the papers was done to identify potentially relevant ones.After the first three authors of this study identified the selected papers, the particular paper lists of these authors were then compared to each other until reaching a consensus between them.In this step, exclusions from and inclusions into the list were made, resulting in the final agreed-upon list.The whole search process was coordinated by the second author and reviewed by the fourth author to propose improvements if needed.
The search process was done in June 2022, yielding results as listed in Table 1.In the process, a total number of 402 peer-reviewed studies were obtained from the search results as seen in Table 1.This initial list included duplicate records.After removing the duplicate records, the list included 230 distinct records.Out of 230 excluded studies, six are not accessible to full text, 33 are unrelated to Agile, 60 are unrelated to software, and 13 are secondary studies, adding up to 112 in the total.As we set the relevant filters in the libraries, we did not encounter any paper not including any authors from an organization in Turkey, papers published in non-peer-reviewed sources, or papers that are neither in English nor in Turkish.The remaining 118 papers are included in our study.For the whole list, the spreadsheet containing the search results together with inclusion and exclusion decisions is available online at https://tinyurl.com/4cft7z75.Some iterations were made to investigate what could be extracted and, accordingly, whether a possible update on the RQs was needed.Two points were taken into consideration in the first iteration: Firstly, is it possible to find the answers to the RQs?Secondly, could any other relevant RQs be added?Therefore, the RQs were preserved as they are.A data collection form holding information was designed to record the collected The collected raw data were subjected to a deduplication process.Missing data were manually completed.Citation data were obtained from Google Scholar.During the deduplication process, the author names, affiliations, and venue information had to be often fixed manually.For instance, some affiliation data includes more than one institution for a single author, which may be due to two reasons: (1) as a result of the author's multiple institutions, most of the time, a private company and a university, and (2) as a result of author's unstandardized affiliation names for a single institution.Cases in which the names of the authors (most of those including Turkish characters, a middle name, or a maiden name) were not represented in a unified way were common.
For the information that is open to a subjective evaluation, first, the relevant information was taken as it is from the relevant study and copied to the Excel file.Such information that is relatively difficult to quantify is such data from the parts about thematic classification.For this type of data, it was aimed for the authors to have a bias as little as possible and comment on the original data in this way and develop his/her comments gradually.In addition to this, two randomly selected studies were independently reviewed by the first three authors of this study and jointly evaluated until consensus was reached to ensure a common understanding and the data extraction step was applied in the agreed standard way.The rest of the remaining papers were allocated randomly to the first three authors.The authors run separate sessions in order to extract data that serve to answer the RQs by applying detailed and thorough examinations of the studies.The second author coordinated the data extraction process.

| Running bibliometric analysis and reporting findings
Once data extraction was complete, the extracted data were closely synchronized, analyzed with the first three authors, and reported by the first author.Rather than simply reporting a summary of the data, we aimed to craft insightful discussions.In line with what is suggested by Donthu et al., 13 we used bibliometric visualization in figures and tables enabling analytical discussions.We mainly used Excel visualization capabilities rather than any other software, which brings us flexibility, especially for customized analyses different from those provided by the specific tools.When Excel's capabilities were not enough to make further analysis, like in network analyses, then we used the VOSviewer application (VOSviewer is freely available at http://www.VOSviewer.com/),which can be utilized for map-based analyses. 45Additionally, when overlapped, we compared the results manually created from Excel and those results from VOSviewer and we have seen that they are consistent with each other.

| Potential threats to validity
We systematically identified and addressed potential threats to four types of validity of our research, we described the steps that we took to minimize or mitigate these threats as adopted from Wohlin et al. 46 and Erdogan and Tarhan. 47

| Internal validity
Limitations of search terms and search engines can lead to an incomplete set of primary sources.To mitigate the risk of finding irrelevant studies, a search was undertaken using defined keywords, followed by a manual search among the references in the initial pool.To minimize threats that may result from search engines, we included comprehensive academic databases.We recorded each paper that we found with its source in an Excel sheet.Therefore, we believe that an adequate and inclusive basis was established for this study and if there was any missing publication, the rate would be negligible.

| Construct validity
In this study, threats related to this type of validity concerned the suitability of RQs.The RQs with different aspects were specifically designed for the defined goal.The questions were systematically answered and finalized through several iterative improvement processes.

| Conclusion validity
To ensure the reliability of our treatments, the entire pool of primary sources was analyzed and the data were reviewed, extracted, and synthesized by the first three authors in iterations according to a research protocol, and the whole process was reviewed by the second and fourth authors.In addition, the guideline of a bibliometric study and procedure ensured the replicability of this study and that the results would not significantly deviate from those of other similar studies.

| External validity
Defining search terms in the source selection approach resulted in obtaining only primary sources written in the English and Turkish languages.
However, the main issue regards whether the selected works represent all types of literature in the area of study, and we ensure that the relevant studies collected in the study pool contained sufficient information to represent the entire related literature.

| RQ1 bibliometric properties of papers
In this RQ, we present the results of five sub-RQs.The first of the analyzes related to the bibliometrics of the publications is the trend of the volume of publications by Turkey compared to those by the globe (RQ1.1).The relevant result is depicted in Figure 1.It should be noted that the ratio information was calculated by dividing those produced by Turkey by those produced globally.Generally speaking, the observed share of Turkey in the world has almost a stable increase.Although Turkey did not produce any publications until 2006, the proportion has increased, except for 2010, 2013, and 2019.Among the years, 2011 draws attention with its high score compared to near years and 2018 does so, as the highest so far.
The distribution of papers by year according to the main data (not the one subjected to RQ1.1) is in Figure 2 (RQ1.2).With the first publication titled "A designing practice and two coding practices for extreme programming (XP)," 48 the records of the publication in the area have started to appear (the reason why this record appeared in this search and not in the one for RQ1.1 is that Agile method-based words such as XP were included in this search).The authors produced publications for 17 years, excluding the years between 2004 and 2006.Although there were few studies until 2011, the trend started to rise in 2011, reached a record number in 2018, and kept an almost constant number in the following 3 years.
In terms of the publication type (RQ1.3),83 studies (70%) are conference papers, 32 papers (27%) are journal articles, two (2%) are book chapters, and only one (1%) is a review paper, as seen in Figure 2. Regarding the language of the papers, 92 (78%) are in English, and 26 (22%) are in Turkish.
We used the contribution types that are proposed by Petersen et al. 49 to classify the papers adding new contribution types such as challenges, taxonomy, and guide.According to the figure, 38% of the papers provide empirical study only, and 22% of them come with a model/approach proposal.Almost half of the papers do not constitute any new entities such as methods, techniques, tools, and frameworks.In addition to the contribution types of the papers, in terms of the number of methods under each contribution type, while in "Challenge" and "Model/Approach" contribution types, Scrum is the prominent one; for the rest, "Agile (General)" is the top one.Regarding the items under the "Challenge" category, they include topics such as challenges of balancing and reconciling Agile approaches with traditional and regulation processes, usage of Agile methods in geographical distance contexts, scalability in Agile, and people management in Scrum.
F I G U R E 2 Number of publications by year.
Regarding RQ1.5, with 1356 total citations scattered to year as shown in Figure 4, citations accelerated after 2010.The average citation count per publication is 11.5, the median of citation counts is 4, the h-index is 27, the i10-index is 72, and the i100-index is 6.Some studies are receiving a high number of citations, as well as low numbers or none.The number of papers with at least one citation is 90 (%76.2).For the remaining 28 papers with zero citations yet, the average lifespan is 4 years (assuming all publications are published in the middle of the publication year).Although published in 2019 and before, the number of publications that have not yet been cited is 16, of which 12 papers are from the Turkish National Software Engineering Symposium (UYMS).The studies that have never been cited yet seem to have a high lifespan.
The top six papers among the most influential papers are journal articles.Journal of Information and Software Technology has the greatest number of papers on the list.The five most cited studies are given in Table 3 to demonstrate the most influential publications.The first two papers share the same authors: Mishra, D. and Mishra, A. In their first paper, 50 they analyzed the agile development and management approach used in developing a complex software project.In the second one, 51 they investigated the effect of different constituents of the physical environment on communication, coordination, and collaboration and provided a guideline for agile software developers in this regard.Tarhan and Yilmaz 52 provided an empirical method and the results from systematic analyses and comparison of development performance and product quality of incremental process as an instance of traditional methods and agile process as a combination of unified software development process, extreme programming, and Scrum adapted in two projects of a middle-size, telecommunication software development company in Turkey.The results of the comparison showed that the Agile Process performed better than the incremental process in many aspects.

| RQ2 bibliometric analysis of researchers
Figure 5 shows the top 10 authors with the highest number of papers and the average citation counts they receive from their papers (RQ2.1).
Onur Demirörs' research interest includes software measurement, process mining, maturity models, and software process improvement in conjunction with Agile.Necmettin Ozkan focuses on Agile in regulated environments, Agile project management, Agile mindset, scaling Agile, agility, and cultural perspectives of Agile.Alok Mishra deals with Agile methods adoption, teaching Agile, Agile project management, and quality aspects in ASD concerning our topic.
Figure 6 delivers the same metrics for the first authors with more than three papers.Being at the first three and different from Figure 5 above, Deepti Mishra has shared papers and topics related to Agile with Alok Mishra.Third in the list, Ozden Ozcan-Top focuses on Agile assessment and maturity models and ASD for medical devices.
Regarding RQ2.2, 108 out of 118 papers (91.5%) are co-authored publications, whereas 10 (8.5%) are sole-authored.Among the co-authored publications, there are publications with 28, 21, and 14 co-authors as the highest numbers.Even though there are hundreds of authors from Turkey and the rest of the world involved in our study, the majority of them are not connected.The largest set of connected items consists of 91 items as shown in Figure 7 obtained from VOSviewer.Because the Scopus and WoS output data format is not the same and we needed to select either of them, not allowed to select both, we selected Scopus as the data source and had to neglect the five records from WoS in the scope of this analysis.However, this constraint does not seem to have the effect to change the mainstream posed in the figure.
According to the picture, Murat Yılmaz, Onur Demirörs, and Alok Mishra occupy the central points.Among them, Murat Yılmaz has a remarkable international network.Vahid Garousi seems to be merging two network chunks.The presence of other scholars (such as Magit Y. and Scattered distribution of papers by years.
T A B L E 3 Most cited five papers.

Number of citations
Complex software project development: agile methods adoption.Mishra and Mishra 50 120 Impact of physical ambiance on communication, collaboration and coordination in agile software development: an empirical evaluation.
Mishra, Mishra, and Ostrovska. 51 Systematic analyses and comparison of development performance and product quality of incremental process and Agile Process.
Tarhan and Yilmaz 52 88 An examination of personality traits and how they impact on software development teams.Yilmaz et al. 53 85 Effective communication, collaboration, and coordination in extreme programming: human-centric perspective in a small organization.
Mishra and Mishra 54

84
G. Giray) working with the foreign connections in Murat Yılmaz's (yilmaz m.) network can demonstrate how the network has expanded to/has been shared by multiple points.
In publications with international collaboration, the proportion of papers where the total number of authors from Turkey is less than the rest of the whole is 47% and the proportion of papers where the total number of authors from Turkey is more than the rest of the whole are at 53%.Regarding the authors collaborating with the international author(s), Table 4 presents all results.Numerically speaking, Yilmaz, M., Akarsu, Z., Desharnais, J., Kocatürk, B., Tekin, N., and Demirörs, O. are separated from the rest of the group in terms of the international collaboration they established.

| RQ3 bibliometric analysis of affiliations
The following part of this paper delivers information regarding the affiliations of the authors.Figure 8 shows the affiliation category grouped into six types (RQ3.1).Government agency represents organizations such as Havelsan, Aselsan, and TÜB _ ITAK.According to the table, almost half of the papers have been produced solely by universities.In 73% of the papers, at least one author from a university is involved.16% of the papers have been produced solely by private companies and overall, 37% of the papers involve at least one author from a private company.Only 25% of the papers have collaborated through different affiliation categories such as a private company and a university together.The rest, 75% of the papers, includes only one type of affiliation; either a university, private company, or government agency.
Based on the affiliation category, Figure 9 presents the top 10 number of papers per organization overall constituting 64% of the total (RQ3.2).When it comes to the number of papers produced by universities, it exhibits the top 10 universities, overall constituting 81% of the total number of papers published.When it comes to private companies, it presents the top ones in terms of the number of papers published, overall constituting 87% of the total output in this category.The first five of them in the list follow each other with close ratios.There are five government agency organizations in total in our study.The figure shows all numbers of papers per government agency that are dominated by the first three organizations.
In our data set, in total, there are 235 distinct authors, of whom 68.9% are from Turkish affiliations.We provided the collaborating countries we obtained in our Excel file and calculated the collaboration percentages of the countries (RQ3.3).In this calculation, if a particular publication contains multiple authors from the same country, it is counted only once.According to the obtained data and the calculation, Ireland ranks first among all countries with the most co-authors abroad, with a ratio of 29%.The countries with the relevant proportions are Netherlands, 10%; T A B L E 4 Authors collaborating with the international author(s).
Author name Number of papers with international author(s) Author name Number of papers with international author(s)

| RQ4 top venues
Regarding the publication venues, Table 5 exhibits the top nine venues with three or more papers, the number of papers published at these venues, and relevant citation counts and ranking of the venues.When it comes to the citation count, these nine venues constitute 58.6% (795/1356) of the total citations.Among these places and the full list including 56 distinct places, not fully presented here, Turkish National Software Engineering Symposium is the only national and prominent venue that the authors from Turkey prefer to participate in.For the rest of the table below, it seems that they prefer to attend highly reputable venues or international conferences hosted in Turkey.

| RQ5 thematic content analysis
The following part of this paper delivers information regarding thematic content analyses conducted by the authors of this paper.We investigate thematic content analysis in terms of Agile methods, involved sectors, SWEBOK category, and themes.Among them, Figure 10 exhibits involved Agile methods information which is manually extracted from the papers (RQ5.1).To those not explicitly mentioning any Agile methods and to those mentioning "Agile" in general as the method, "Agile (General)" category is assigned.This category constitutes a considerable number of the total.For the rest, Scrum is the prominent one followed by XP.In addition to this superiority of Scrum, its integrated usages with other methods including Kanban, Nexus, SAFe, and Lean stands out.Also, we have found six studies involving an Agile scaling method.These methods cover SAFe (three times), Nexus (three times), Scrum of Scrums (two times), LeSS, Scrum at Scale, DAD, and RAGE (one time for each).
F I G U R E 9 Active organizations based on affiliation category.
Table 6 delivers the number of papers, number of total citations, and average citation counts per sector for those where the number of papers is greater than one (RQ5.2).The sector information was extracted from the contents of papers (not from the authors' sectors) when it is explicitly mentioned in which sector the particular study was conducted.Thus, a researcher from a university can conduct a study on Agile in the telecommunication sector, which adds to the telecommunication item in the table.It is also noted that each paper may not contain information about the sector in which the study is carried out.Therefore, the bottom total, 68, is less than the total number of studies, 118.• It is known by the authors that the Agile-related initiatives volume undertaken in Turkey industry is much higher than the volume of the publications.In addition, the fact that university departments in the field of IT in Turkey focus on technical aspects and the relatively low number of special department curricula and courses in IT process management (and other Agile-related fields) may have contributed to the widening of the gap between the industry and the academy.Despite this, the increasing number of research and development (R&D) centers in the sector, especially in recent years, may have contributed to the closing of this gap.

According to
• For Turkey, 2011 draws attention with its high score compared to near years; 2018 does so as the highest so far.The reason that makes these years special may be a subject of further investigation.A possible reason may be that this relatively small number of publications is severely caused by partial fluctuations.
• One of the reasons for the increase in 2015 and thereafter may be that UYMS, which hosts the most publications on Agile in Turkey, contributed to the field considerably (Even though the first record in Scopus from UYMS appears in 2013).
RQ1.3 Volume statistics for different publications types and languages: • The fact that conference publications are more than other types in terms of number indicates that the field is developing.Considering that conference paper publications are in general easier than publishing journal papers, we can infer that the majority of the Agile community in Turkey (can) create a relatively lower level of publications in the field.One more reason to have such a ratio for the conference papers can be related to having more papers from industries that may prefer to communicate their work verbally in conferences rather than in journals.The journal publication process can be longer than that of the conference in general.This factor may lead to keeping the sector participants away from such long durations as they are not usually academia driven and are not oriented to get more credibility from it.
• The effect of UYMS on the dominance of conference papers is visible.In addition, to find a place for Agile-related papers, which was a relatively new subject in its early years, there may have been a tendency for conferences that can be considered more flexible in scope compared to journals. 21In addition, the fact that there are conferences and special issues, and sections in journals dedicated to Agile 56 and the lack of (active) dedicated journals may be an implication of this tendency.
• Although some places accept publications in (both English and Turkish), the proportion of publications in English seems to be high.It is noteworthy that 23 of 26 Turkish publications are in UYMS.Even though English can be chosen as the language of the publication, publications in Turkish may seriously reduce the frequency of access to these publications.

RQ1.4 Contribution types and research method types:
• According to the included publications' results, the people in Turkey tend to use ready-made methods rather than producing their method, technique, tool, framework, process, workflow, or metric.This result also coincides with the low rate of hybrid method usage in Figure 3.
• Research method types show that a significant part of the work comes from the practices.This case confirms that Agile is a practice-lead area 57 and it is valid also in Turkey.
• Scrum is the prominent method in all contribution types, especially when a particular model, approach, method, or technique is studied.

RQ1.5 Citation analysis:
• The first five of the most cited papers are journal publications that may allow higher numbers of citations for papers.
• Among the studies that have never been cited yet, the number of those in English (15/28) is remarkable.
• It is seen that the most cited conference papers have a similar scope to the Agile topic.
• Mishra, D., Tarhan, A., Yilmaz, M., and Ozcan-Top, O. draw attention as prominent authors contributing to the most influential papers.
• It is noteworthy that all the studies published in the last 6 years (2016 and later) include Yilmaz, M., as (one of) the author.
RQ2 Bibliometric analysis of researchers: RQ2.1 Most contributing researchers: • Certain people (Demirörs, O., Mishra, A., Tarhan, A. K., and Kalıpsız, O.) are not included in the top list for the papers when they are the first author.
• There is not always a direct proportionality between the quantitative measurement of the number of publications and the number of citations received; those who have more publications do not always receive more citations.

RQ 1 . 2 .
Annual trend of publications: What is the annual trend of publications in Turkey in the field of Agile?RQ 1.3.Volume statistics for different publication types and languages.What are the percentages of publication types (conference paper vs. journal article)?What are the languages of the papers (English vs. Turkish)?RQ 1.4.Contribution types and research method types: What are the contribution types of the papers?What are the research methods of the papers (empirical versus non-empirical)?Empirical studies include case studies, which use interviews and/or surveys to gather data, experiments, experience reports, and so on.RQ 1.5.Citation analysis: What is the total citation count, average citation count per publication, the median, h-index i10-index, and i100-index for the papers?What are the most influential papers measured by citations?Additionally, this analysis investigates papers with zero citations by considering their lifespan and prominent publication venues.RQ 2. Bibliometric analysis of researchers: What are the bibliometric analyses of the researchers/authors in terms of active researchers, co-authorship, and international collaborations?RQ 2.1.Most contributing researchers: Who are the principal authors with the highest publication numbers and what is the average number of citations to their papers?In particular, who are the top first authors with the highest publication numbersand what is the total number of citations to their papers?RQ 2.2.Co-authorship: What is the proportion of co-authored papers (co-authored versus sole-authored)?Who has been collaborating?In this regard, we particularly investigated co-author networks with a special bibliometric software application, VOSviewer.RQ 2.3.International collaborations: What are the total number of authors and the proportion of authors from Turkey?These metrics investigate the total number of the determined papers and the proportion of Turkish scholars in this total number.Who are the authors from Turkey who mostly collaborate with international scholars?What is the proportion of international collaboration?We aimed to distinguish the general proportion, trend, and weight of the percentage of authors from Turkey calculated based on papers.RQ 3. Bibliometric analysis of affiliations: What are the distributions of researcher affiliations in terms of affiliation categories, contributing countries, and organizations?RQ 3.1.Most contributing affiliation categories: Which and how many of the affiliation categories contribute?RQ 3.2.Most contributing institutions: Which institutions contribute the most overall and based on affiliation categories?RQ 3.3.Most contributing countries: What are the most collaborated countries?RQ 4. Top venues: What are the top publication venues?It is measured by the number of publications published in the particular venue and yet enriched with the number of citations, citation per paper, and ranking of the venues.RQ 5. Thematic content analysis: What are the thematic content analyses of the papers?

RQ5. 1 .
Agile methods: What Agile methods are cost estimation techniques involved in the studies?RQ5.2.Sectors: What sectors are involved the most?RQ5.3.Software Engineering Body of Knowledge SWEBOK 42 category: What are the SWEBOK categories of the papers?RQ5.4.Agile themes: What are the Agile themes of the papers?In addition to manually extracted data, we also provide a network visualization of the co-occurrence of keywords to gain insight into the themes of the papers.

T A B L E 1 2 60 EC4 Papers unrelated to Agile 33 EC5
Number of sources retrieved and selected by a search query.ABS-KEY (((agile AND software) OR (çevik AND yazılım) OR scrum OR kanban OR "extreme programming" OR "uç programlama" OR (XP AND agile) OR (XP AND çevik))) AND AFFILCOUNTRY (turkey)) Inclusion criteria and exclusion criteria.least one author from a Turkish organization regardless of nationality -IC2 Primary papers not using secondary sources of data such as literature reviews and systematic mapping studies -IC3 Papers in the software development domain -IC4 The conference, workshop, journal, or book chapter papers -IC5 Papers in English or Turkish -EC1 Paper not including any authors from any organization in Turkey 0 EC2 Secondary papers such as literature reviews and systematic mapping studies 13EC3Papers in domains other than software development (for example, healthcare, construction, and supply chain) Papers published in non-peer-reviewed sources such as theses, web pages, workshop proposals, tutorials, panels, proceeding informationidentified studies.The collected information ranges from general information about each study such as Author, Title, Year, Venue, Author Affiliation, and Author Country, as well as specific information to answer the RQs.

F I G U R E 5
Top 10 researchers with high publication counts.F I G U R E 6 Most contributing first authors in publications with more than three papers.F I G U R E 7 Network visualization of co-authorship.

Table 7 ,
which provides information about the SWEBOK category identified by the authors of this paper, the majority of the Publication venues.Agile methods in the papers.Agile theme.
(2) Refactoring (2), Continuous integration (2), People management issues such as high personnel turnover and decrease in motivation (2), Usability testing (1), Functional testing (1), Acceptance tests (1), Managing technical debt (1), and Performance testing(1).Even though the SWEBOK categories provide insight into the papers' focus, they are generic, and not specific to Agile or any other domains.Therefore, we preferred to use customized Agile themes identified by the authors of this paper to give more insight into what subjects the papers deal with (RQ5.4).In this regard, Table8was created.According to the table, a considerable number of studies focus on practices, processes, and methods perspectives of Agile.There are nine papers particularly dealing with task management and effort estimation in Agile, followed by the T A B L E 5 T A B L E 6 Paper-related numbers per sector.