Variable science mapping as literature review method

Tomczyk, Przemyslaw; Brüggemann, Philipp; Paul, Justin Article — Published Version Variable science mapping as literature review method Journal of Marketing Analytics Provided in Cooperation with: Springer Nature Suggested Citation: Tomczyk, Przemyslaw; Brüggemann, Philipp; Paul, Justin (2024) : Variable science mapping as literature review method, Journal of Marketing Analytics, ISSN 2050-3326, Palgrave Macmillan, London, Vol. 12, Iss. 4, pp. 829-841, https://doi.org/10.1057/s41270-024-00336-9 This Version is available at: https://hdl.handle.net/10419/316664 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Journal of Marketing Analytics (2024) 12:829–841 https://doi.org/10.1057/s41270-024-00336-9 ORIGINAL ARTICLE Variable science mapping asliterature review method PrzemyslawTomczyk1· PhilippBrüggemann2 · JustinPaul3 Revised: 9 April 2024 / Accepted: 19 June 2024 / Published online: 2 July 2024 © The Author(s) 2024 Abstract This study investigates a novel mapping approach for the systematic analysis of empirical research, termed Variable Science Mapping (VSM). This approach enhances the current capabilities of Systematic Literature Reviews (SLRs) by incorporating variables and their interrelationships, surpassing traditional methods, such as Science Mapping (SM), which primarily analyze keywords, citations, and authorship. We present a step-by-step conceptual protocol for implementing the VSM approach. Subsequently, the strengths and limitations of VSM compared to SM are examined across 12 SLR stages. To this end, we assess the actual usage of SM for each stage based on an analysis of 63 papers employing the SM approach. Additionally, expert interviews are conducted to evaluate the utility of both SM and VSM across identical analytical stages. Notably, a distinct alignment emerged between the outcomes of the SLR and expert assessments pertaining to SM. The findings reveal VSM’s favorable ratings in eight out of 12 stages. Equivalence in expert ratings between SM and VSM surfaced in one stage, while SM was deemed more beneficial in three stages. This nuanced evaluation underscores the contextual strengths and limitations of both approaches. The implications extend to both scientific and managerial domains, offering valuable insights into the prospective advancements in SLRs. In conclusion, this analysis not only sheds light on the potential advantages of VSM but also serves as a foundation for guiding future research methodologies to widen capabilities among different SLR stages. Keywords Literature analysis· Empirical mapping· Systematic literature review· Science mapping· Variable science mapping Introduction Systematic literature review (SLR) is a research method that enables the identification, selection, critical evaluation, and synthesis of existing literature in a rigorous, transparent, and repeatable manner, leading to robust conclusions about what is known and what is not known in peer-reviewed research areas (Christofi etal. 2021). According to Scopus, in 2022, 667 scientific articles in the field of management, accounting, and finance were published, in which a SLR was the basic one. Compared to 487 in 2021 and 379 in 2021, an upward trend close to exponential can be seen. The procedure of SLRs can be complex. The use of this method requires a review question (Christofi etal. 2021; Leonidou etal. 2018; Mcquade etal. 2021; Vrontis and Christofi 2019), data collection (Christofi etal. 2021; Mcquade etal. 2021), inclusion or exclusion criteria, selection of relevant studies, final database preparation (Leonidou etal. 2018; Vrontis and Christofi 2019), bibliometric analysis (Mcquade etal. 2021; Siemieniako etal. 2022; Vrontis and Christofi 2019), research results presentation in the form of thematic analysis (Leonidou etal. 2018; Mcquade etal. 2021; Siemieniako etal. 2022; Vrontis and Christofi 2019) or synthesis (Christofi etal. 2021; Leonidou etal. 2018; Vrontis and Christofi 2019), contribution presentation (Christofi etal. 2021), and developing a section on future research agenda (Paul and Menzies 2023; Christofi etal. 2021). One frequently employed technique in SLRs is Science Mapping (SM). SM is used in bibliometric analysis, understood as a part of a SLR or a separate analysis to achieve visual data presentation (Chen 2017; Ghorbani etal. 2021; ElKattan etal. 2023). The current approach that dominates today consists of mapping areas, keywords, terms, authors, or citations (LeónCastro etal. 2021). Numerous different software solutions * Philipp Brüggemann [email protected] 1 Department ofMarketing, Kozminski University, Jagiellonska St. 59, Warsaw, Poland 2 University ofHagen, Hagen, Germany 3 Plaza Universitaria, Río Piedras, SanJuan, PuertoRico, USA 830 P.Tomczyk et al. exist for the generation of Science Maps (e.g., VOSviewer, CiteSpace, or SciMAT). In recent publications, SM has been used to perform extensive literature reviews, especially in areas with a relatively high number of publications, such as integrated marketing communication (Christian etal. 2021; Wu etal. 2022) or digital marketing (Aksoy etal. 2021; León-Castro etal. 2021; Gao etal. 2021). Despite the distinct advantages offered by SM, such as the comprehensive presentation of the research field from multiple perspectives and its ease of use, this approach is not without limitations. While this approach has strengths in analyzing papers on a descriptive basis, e.g., by considering keywords and citations, it might be impossible to investigate key variables, key theories, or antecedents and consequences of the studied literature. While this approach is limited to summarizing metadata, the degree of information extraction and the closeness to an appropriate picture of the underlying literature are severely limited (Tomczyk 2022). Furthermore, there has been little research on the boundaries of the SM approach, especially as limitations in research papers (e.g., Mavric etal. 2021; Zupic and Čater 2015). While there is merit in using the approach of SM, it is constrained to the descriptive content analysis phase (Najaf etal. 2022). As such, this approach can hardly provide support at the first stage of the literature review, i.e., formulating a research problem. Therefore, researchers have criticized bibliometric reviews based on SM (Paul etal. 2021; Paul etal. 2023) because the theoretical contributions can be limited. In general, the development of a domain area is based on the development of knowledge about variables as an expression of social phenomena. Progress is made by creating new variables, investigating new relationships, and analyzing these connections in different contexts. Surprisingly, there is no approach to systematically investigate such evolving variables and relationships. Another widely recognized and commonly employed method in SLRs is Meta-analysis (Brüggemann and Rajguru 2022). This technique is particularly effective for examining selected relationships across multiple studies (Glass 1976) using specialized software. Despite its merit and relevance, Meta-analysis is constrained to offering an aggregated overview of selected relationships. Differences within a given body of research, which encompass positive, negative, and non-significant results across several papers, cannot be fully addressed by SM or Meta-analysis. To address this gap, we propose a novel approach termed Variable Science Mapping (VSM). This technique aims to conduct a comprehensive synthesis of relationships between variables in empirical research. Additionally, it provides a graphical visualization encompassing all variables and their interrelationships. Consequently, this new approach facilitates the presentation of the current state of knowledge for any selected variable within any domain of the social sciences. This mapping technique is expected to be valuable both in formulating research problems and in analyzing content to address these problems. In this study, our objective is not to determine whether VSM is superior to existing methods but to explore potential enhancements in SLRs through the application of VSM. To ascertain the appropriate contexts for using SM and VSM in SLRs, we first identify 12 relevant stages in SLRs based on the existing literature. We then conduct a SLR of 63 top-tier articles, examining the utilization of SM across these predefined 12 stages. Subsequently, through a series of in-depth interviews with experienced researchers in management, we investigate where and how VSM and SM can be beneficial in SLRs. Synthesizing knowledge inSLRs SM andMeta‑analysis To synthesize knowledge from scientific articles, researchers can utilize SM and Meta-analysis methodologies. SM usually employs citation, co-citation, and co-occurrence analyses to depict the knowledge bases within a field (Sanguankaew and Ractham 2019). Using this approach, researchers can understand the relationships between scientific works and the evolution of knowledge within a domain. SM tools, such as bibliometric software, help visualize knowledge domains, literature, and citation networks (Chen 2017). These tools facilitate the identification of trends, influential works, and emerging topics within a field, aiding in the interpretation of extensive scholarly information, as well as mapping research landscapes and visualizing connections between studies (Cobo etal. 2011). Meta-analysis, on the other hand, provides a quantitative synthesis of coefficients from multiple studies, allowing researchers to draw conclusions by analyzing empirical findings from various sources (Harlos etal. 2016). This statistical method is widely used across disciplines to aggregate and analyze results reported in different studies (Huang and Hu 2017). Through Meta-analysis, researchers can calculate effect sizes and assess trends or patterns present in the literature (Suciana and Sausan 2023). Metaanalysis offers a systematic approach to aggregating and analyzing data from multiple studies to draw overarching conclusions or identify patterns across research findings (Huang and Hu 2017). By applying meta-analytical techniques, researchers can quantitatively assess the consistency and magnitude of effects observed in different studies, providing a comprehensive understanding of a research question (Dolapçıoğlu and Subaşi 2022). Metaanalysis is valuable in synthesizing evidence from diverse 831Variable science mapping asliterature review method sources and resolving discrepancies present in individual studies (Anjani 2023). By incorporating Meta-analysis, researchers can quantitatively analyze and synthesize data from identified studies, offering a rigorous and evidencebased approach to knowledge synthesis (Huang and Hu 2017). The integration of SM tools and Meta-analysis techniques enhances the understanding of complex research landscapes, providing a robust framework for synthesizing knowledge from scientific articles. However, these approaches also have certain limitations. SM cannot identify variables, relationships, or research models across different papers; it can only compare elements such as keywords and authorship between papers in a map. While Meta-analysis focuses on empirical studies and can analyze relationships, it does so only from an aggregated perspective across all examined papers, providing insight into specific relationships. Meta-analysis cannot create a depiction of the research implemented in the papers, nor can it analyze contradictory results, such as those with positive or negative coefficients, for further research. To address these limitations of current approaches, we introduce VSM in the next section. The concept ofVSM This section introduces the VSM technique to extend the capability of SLRs. The rationale behind this approach is not to supplant SM or Meta-analysis but to enhance SLRs for empirical papers. We contend that this approach is particularly valuable in research areas with numerous empirical studies on similar topics. While SM can identify keywords, co-authorship, and other relevant elements for a specific variable, VSM aims to map all variables across the papers under investigation. This comprehensive mapping can offer researchers a more holistic overview of a predefined research field. Tailored for systematic analysis of empirical research, VSM also complements the well-established method of Meta-analysis by capturing variables and relationships, examining commonalities and disparities across diverse papers. This approach, e.g., unveils potential research gaps or inconsistent findings. Figure1 delineates the procedural steps involved in employing VSM. We detail these steps below following from a VSM protocol we suggest to follow when using VSM. Figure1 presents a conceptual protocol for applying the VSM approach. This protocol can be utilized Fig. 1 Conceptual protocol for applying the VSM approach 832 P.Tomczyk et al. by researchers in future to implement the VSM approach step by step, thereby ensuring an adequate execution of this methodology. Before incrementally implementing VSM, it is imperative to precisely delineate the research domain under investigation in step 1. This initial step holds significant importance as it profoundly influences all subsequent procedures. Opting for a broad research scope (e.g., technology acceptance in electronic commerce) will result in a correspondingly extensive and intricate Variable Science Map. While this breadth can offer a comprehensive overview of entire research streams, it also entails increased effort and complexity in analysis. For emerging and narrower research domains, VSM can serve to generate an overview of past research endeavors and extract potential future research directions. Once the research field under scrutiny has been identified, the subsequent step 2 involves seeking pertinent literature. This phase should adhere to a structured approach akin to a SLR method, as advocated by Vrontis and Christofi (2019) and outlined by Paul and Criado (2020). Upon completion of the preparatory stages, the application of VSM ensues in step 3. This involves the initial identification of variables within each of the articles under investigation. These variables might inherently represent antecedents, consequences, mediators, or moderators. It is imperative to note that a variable could serve diverse roles across multiple articles. Following the identification of variables within each of the examined papers, step 4 involves extracting the necessary information for the Variable Science Map. This encompasses the identified variables (antecedents, consequences, mediation, or moderation), the analyzed directions of relationships and details regarding the significance of empirical findings. To facilitate information extraction, it might be beneficial to consider the figures or formulas portraying the empirical analyses within the papers, provided such information is available. In step 5, all variables are incorporated into a model to depict the discovered relationships. This process results in a comprehensive research model based on the previously examined articles. Figure1 illustrates, on the right side, an exemplary analysis of five articles. These five articles encompass five distinct research models, which are subsequently integrated into an aggregated research model. Following the creation of this aggregated model, details regarding significance and relationship direction (± or nonsignificant) can be documented. In the subsequent step 6, the augmentation of evaluation metrics becomes pivotal for a more comprehensive contextualization of the analysis outcomes. These metrics encompass diverse dimensions, including the number of included papers, variables, identified relationships, the theoretical maximum of potential relationships, and instances of conflicting results. Leveraging these descriptive specifics, various metrics can be calculated, such as the ratio of relationships derived from articles to the theoretical maximum, the proportion of conflicting results, or the prevalence of non-significant findings. Figure3 serves as an illustrative instance, providing a portrayal of a Variable Science Map coupled with evaluation metrics. In step 7, presenting the research results is recommended, involving the display of both the Variable Science Map and accompanying metrics. This inclusive approach offers a comprehensive view of empirical research within a specific field of study. Simultaneous consideration of the Variable Science Map and metrics aids in conveying and assessing the analysis’s quality and extent. Moving to step 8, the interpretation of findings relies on the Variable Science Map and evaluation metrics. By consolidating multiple empirical works, this approach enables result verification, identification of contradictory findings, and the revelation of further research needs. While deriving intricate empirical models is plausible, it is vital to stress caution in utilizing the resulting research model as a template for subsequent empirical research. The aggregation of multiple models may introduce statistical concerns, such as multicollinearity or endogeneity. In the next section, we will investigate the usefulness of SM and VSM related to 12 SLR stages. We have deliberately chosen SM as the comparative method because it is also a mapping approach, making it particularly suitable for comparison with VSM. It is important to note that VSM is not intended to replace SM, but rather to complement this established mapping approach. Analyzing SM andVSM alongSLR stages Derivation ofSLR stages In this investigation, our focus centers on a SLR that incorporates the SM technique. Our primary aim is to contrast the approach outlined in this article with the established methodologies prevalent in the scientific community. Employing the SLR method recommended by experts (Vrontis and Christofi 2019; Paul and Criado 2020), we initiated our exploration using Scopus, the largest repository of peer-reviewed academic publications, employing widely accepted search algorithms (Glińska and Siemieniako 2018). Our search criteria comprised a specific string of three keywords related to SM along with the names of the two most widely software: VOSviewer and Citespace. Within the domains of management, finance, and accounting, our search yielded 590 occurrences in VOSviewer and 163 in Citespace from Scopus using title–abstract–keywords criteria. Subsequently, another SM application, Biblioshiny presented only 66 occurrences. The search string utilized is provided below for reference. 833Variable science mapping asliterature review method TITLE-ABS-KEY(”science mapping” OR “bibliometric mapping” OR “variable mapping” OR “vosviewer” OR “citespace”) AND (LIMIT-TO(SRCTYPE, “j”)) AND (LIMIT-TO(DOCTYPE, “ar”)) AND (LIMITTO(LANGUAGE, “English”)) AND (LIMIT-TO(SUBJAREA, “BUSI”)). On December 27, 2022, we executed the string search, initially casting a wide net across the entire database. Given the extensive yield of several thousand results, we refined our search parameters to encompass only the categories of management, finance, and accounting. This focused approach yielded 532 articles. Subsequent analysis of the abstracts led to the retention of 443 articles. Among these, 12 lacked accessible PDFs, resulting in a final count of 431 articles for qualitative analysis. Applying the Academic Journal Guide (AJG) criterion, we sieved through these articles, ultimately selecting those falling within the 3, 4, and 4* categories. This meticulous selection process culminated in the final sample comprising 63 articles of high quality. We analyze each of the 63 articles, gathering general information, such as software type, industry, mapping type, and research field. Furthermore, each article underwent an examination based on a SLR approach consisting of 12 distinct stages derived from Vrontis and Christofi (Christofi etal. 2021; Vrontis and Christofi 2019). Within these stages, spanning from review question identification to recommendations identification, we systematically checked whether the outcomes from SM were utilized. The 12 stages involved in this assessment encompass: 1. Review question identification, 2. Data collection analysis, 3. Bibliometric analysis, 4. Key variable analysis, 5. Key theories analysis, 6. Thematic clustering, 7. Antecedents’ identification, 8. Consequences identification, 9. Research gap identification, 10. Trend identification, 11. Conclusions, 12. Recommendations identification. In the subsequent section, in addition to presenting the outcomes of the SLR, we introduce findings derived from in-depth interviews. These interviews aimed to ascertain experts’ perspectives on the usefulness of both SM and VSM concerning the 12 stages integral to SLRs, as previously delineated. Following this, we consolidate the SLR results regarding the use of SM in scientific research with the experts’ evaluations and discuss the findings. In‑depth expert interviews In the subsequent phase of our research, we conducted a series of nine in-depth interviews, adhering to the Kvale and Brinkmann (2018) methodology. Our participants comprised professionally active researchers in management, holding a PhD degree, and hailing from Poland and Germany. The experts were selected from our professional network. We deliberately chose experts who are not familiar with the topic being analyzed through SM and VSM methodologies to avoid potential bias arising from prior engagement with the subject matter. For the interviews, we curated a set of five articles within the domain of customer ideas: Barasa etal., (2021), Burnham etal. (2020), Casaló and Romero (2019), Chan etal. (2015), Chan etal. (2021). This selection was deliberate, as these articles were deemed sufficient to reveal discernible differences between the methods under scrutiny. We reasoned that if disparities were clearly discernible within this limited set, their presence would be even more pronounced in a broader selection. Moreover, the field of customer ideas within management was chosen for its relevance and unfamiliarity to the participants, mitigating any foreknowledge bias. We generated two distinct maps: a conventional map using VOSviewer (aligned with the SM approach) and a variable map manually crafted (by the VSM approach). To create the Variable Science Map, we followed steps 3 to 6 as outlined in Fig.1. Figures2 and 3 within this context depict the visual representations presented to the interviewed experts. The expert interviews followed a consistent protocol to avoid bias arising from different procedures. Initially, the experts were shown the conventional Science Map, which was briefly explained. Subsequently, for each of the 12 SLR stages, the experts were asked to assess whether SM could be effectively utilized at that stage. Afterward, the Variable Science Map was presented and briefly explained, and the same questions were posed. To facilitate comprehension during the interviews, succinct explanations were provided for both maps. Figure2 represents a visualization of keyword occurrences occurring at least twice, generated through the VOSviewer software. The visualization displays two distinct clusters, indicated by red and green colors. Specifically, the three keywords highlighted in red (i.e., idea generation, perceptions, word of mouth) were frequently interlinked within the analysis of the five articles. Similarly, the three keywords highlighted in green (i.e., participation, customer feedback, innovation) exhibited recurrent co-occurrences within this set of articles. Moreover, Fig.2 demonstrates instances where articles utilized keywords from both clusters (e.g., participation and idea generation). Notably, within the five articles under 834 P.Tomczyk et al. analysis, there was no instance of an article using both innovation and perception as keywords. Figure3 presents a manually created Variable Science Map developed by the authors in accordance with the conceptual protocol for applying the VSM approach, as outlined in Fig.1. This map integrates information derived from the same set of five articles utilized in the earlier SM approach utilizing VOSviewer. Figure3 includes variable names, illustrating all investigated relationships, and enumerates the number of positive, negative, and non-significant empirical findings for each relationship within the subset of five papers utilized in our SLR. For instance, examining the relationship between Benefits (SYNT) and Customer ideation, our analysis revealed one positive, one negative, and two not significant outcomes among the five articles scrutinized. This finding underscores the ambiguity inherent in the current empirical results uncovered by the VSM approach. Furthermore, within this limited subset, only one of the five articles explored the association between Customer ideation and Innovation, detecting a significant positive relationship. However, the VSM approach identifies potential avenues for further investigation to validate these outcomes. It is important to note that these conclusions can only be tentatively drawn when comprehensive literature pertinent to the research question is considered. In our instance, the analysis was limited to a Fig. 2 Conventional Science Map made with VOSviewer (according to the SM approach) idea generation perceptions word-of-mouth participation c c u u s s t t o o m m e e r r f f e e e e d d b b a a c c k k innovation Fig. 3 Variable Science Map made manually by the authors (according to the VSM approach) 835Variable science mapping asliterature review method small set of five articles. Nonetheless, this illustrative example underscores the significant utility of the VSM approach in facilitating multifaceted insights. Figure3 is accompanied by descriptive information and precision metrics. It is important to note that the research models under scrutiny have not yet undergone evaluation by expert judges. However, their presentation here serves as an exemplar of the sixth point within the VSM protocol, as outlined in Fig.1. Exactly as with the presentation of the conventional Science Map, evaluation of these consolidated research models was conducted by experts. This evaluation relied on a succinct description of the Variable Science Map provided by the interviewers. Drawing inspiration from the application of goodness-offit metrics in statistical methodologies, such as regression or structural equation models, we propose methodologies aimed at offering a comprehensive evaluation of information generation and the precision inherent in VSM outcomes. Initially, we present the count of articles (5) and variables (5) identified in our analysis. Following this, we enumerate the relationships revealed through the VSM approach (7). Subsequently, utilizing this data, we calculate the theoretical maximum number of relationships attainable within this analytical framework (5 × 7 = 35). However, it is crucial to note that this theoretical maximum is seldom achieved in VSM analyses due to continual refinements in research models and relationships, often aimed at generating novel insights in response to evolving circumstances. Despite this, using the Theoretical Maximum of Relationships (35), we derive the Model Accuracy (MA) as a ratio of observed relationships to the theoretical maximum (7/35 = 20%). This metric functions as an indicator of the comprehensiveness concerning the analysis of relationships within the papers under consideration. An additional vital metric pertains to conflicting results, where the occurrence of both positive and negative significant outcomes in a relationship denotes a conflicting result. From these instances, we compute the Conflict Rate (CR) to gauge the prevalence of conflicting results relative to all relationships identified using the VSM approach. For instance, within our analysis (Fig.3), with 1 conflicting result among 7 relationships, the indicative CR stands at 28.57% (2/7). This metric aids in identifying ambiguous or contradictory findings, potentially uncovering lacunae in existing research and illuminating unexplored research avenues. Further aiding our evaluation is the Non-Significance Rate (NSR), calculated by determining the ratio of non-significant results (3) to all identified relationships (7), resulting in an NSR of 42.86% (3/7). This illustrates that 42.86% of the relationships studied in the five articles lacked statistical significance. These metrics offer valuable insights for researchers and practitioners, enabling the assessment of result value, identification of disparities among articles, and the identification of research gaps. While long-term utilization necessitates defining thresholds akin to quality measures in regressions or structural equation models, immediate establishment of substantial thresholds for valuation remains elusive due to limited practical experience. Rigorous testing and comparison of numerous VSM results are imperative for establishing credible thresholds. Results In Fig.4, the temporal distribution of the 63 articles utilized for the SLR is depicted. Notably, all these articles utilize the SM approach, as VSM does not currently appear in the prevailing literature. Until 2017, the utilization of the SM approach was sparse, evident only in isolated articles. However, a noteworthy surge in publications commenced in 2018, with a substantial spike observed in 2021. Remarkably, the dynamics illustrate a significant and consistent increase in the number of high-quality articles employing the SM approach, extending until the conclusion of 2022. Despite relatively modest absolute figures, the discernible trend in Fig.4 underscores the growing popularity of the SM approach as a preferred research methodology within top-quality articles. Following the acknowledgment of the escalating importance of the VSM approach, our focus now shifts to presenting, comparing, and interpreting the outcomes derived from distinct investigations—specifically, the findings obtained from our SLR and the in-depth expert interviews pertaining to SM and VSM. Summarizing these results, Fig.5 presents values transformed into percentages. For instance, the uppermost black bar depicted in Fig.5 indicates an 83% value, signifying that within the 63 articles, the utilization of SM for thematic clustering amounted to 52 instances. Observing the black bars within Fig.5 reveals that in six out of the 12 stages analyzed, the utilization of the SM approach exceeded 20% in our literature review. Notably, researchers predominantly employed SM (83% occurrence) for thematic clustering. Furthermore, we identified two stages where a mere 3% of the articles (i.e., 2 out of 63 articles) incorporated SM, while SM was absent in four stages. The absence of SM usage in identifying key variables, antecedents, and consequences can be attributed to a fundamental limitation: Science Maps do not facilitate their identification. Operating primarily on keywords that do not always represent variables, these maps lack the capacity to decipher and report the strengths of relationships among variables. This limitation stands as one substantial constraint within the SM technique. Interestingly, our investigation indicate a disparity between the perceptions of experts we interviewed and the 836 P.Tomczyk et al. Fig. 4 SM usage trend (quality criteria: minimum AJG 3; n = 63) Fig. 5 Presentation of the results from three different investigations on SM and VSM