Sustainable Supply Chains: Bridging Theory and Practice Through Hybrid Analysis

Abstract

Sustainable Supply Chain Management (SSCM) integrates economic, environmental, and social considerations across global supply networks. Despite extensive research, the field remains fragmented across strategic, tactical, and operational levels, with limited theoretical integration and inconsistent alignment of sustainability dimensions. To address this gap, this study develops a data-driven perspective on SSCM research using a hybrid analytical framework. A PRISMA-guided systematic review is combined with bibliometric science mapping with VOSviewer 1.6.16 and transformer-based topic modeling (BERTopic) with Python 3.10 to analyze literature published between 2011 and 2025. The analysis integrates two complementary datasets: highly cited studies to capture established research structures and recent publications to identify emerging trends. The findings reveal the growing prominence of digital technologies, including artificial intelligence, blockchain, and big data analytics, alongside the central role of collaboration and governance mechanisms in enabling sustainability. The topic modeling identifies eleven coherent research themes, highlighting both well-established areas, such as circular economy, and emerging directions, such as risk-oriented decision-making and digital traceability. By introducing a cross-method semantic correspondence approach that integrates citation-based and embedding-based analyses, this study advances a more coherent and multi-layered understanding of SSCM research. This integrated perspective reveals the field’s evolution, core thematic structures, and emerging gaps, while providing a robust foundation for future theoretical and practice-oriented developments.

1. Introduction

Supply Chain Management (SCM) represents a comprehensive approach to the strategic coordination of processes the entire supply chain (SC), from upstream suppliers to downstream consumers. Its main purpose is to enhance overall performance while reducing operational costs [1,2,3]. In recent decades, growing environmental, economic, and social pressures have increased the need to integrate sustainability considerations into SC practices [4,5,6]. Accordingly, Sustainable Supply Chain Management (SSCM) refers to the integration of environmental, social, and economic sustainability objectives into supply chain planning, coordination, and decision-making across firms and stakeholders [7,8]. SSCM is now widely recognized as an important source of competitive advantage across various sectors. It requires coordinated actions at strategic, tactical, and operational levels, both within organizations and across SC networks [9,10,11,12,13]. As SSCM research has expanded over the past two decades, the literature has become increasingly diverse and fragmented. This fragmentation is evident across sustainability dimensions, industrial contexts, methodological approaches, and conceptual definitions [14,15,16]. Although previous review studies have contributed to understanding specific aspects of SSCM, relatively few have systematically mapped the thematic growth and intellectual structure of the field in a comprehensive manner [17,18,19]. Therefore, review approaches are needed that can capture both the structural development of the literature and the emerging semantic patterns within recent research. Therefore, before advancing new theoretical and managerial insights in SSCM, a comprehensive and methodologically robust review of the existing literature is essential [20].

SSCM operates at the intersection of business strategy, environmental stewardship, and social responsibility. Despite more than two decades of research, the field continues to face persistent tensions and fragmentation that constrain both theoretical development and practical application. Although SSCM spans strategic, tactical, and operational levels, these layers remain weakly integrated. Moreover, findings on the relationships among economic, environmental, and social performance are often inconsistent. As sustainability pressures intensify across global SCs, there is a growing need to systematically examine the evolution of SSCM, its dominant themes, underlying contradictions, and the role of emerging technologies. In this context, supply chain digitalization refers to the use of digital technologies, such as artificial intelligence (AI), machine learning (ML), big data analytics, blockchain, Internet of Things (IoT), and digital platforms. These technologies enable SC actors to collect, process, verify, and share information, thereby improving visibility, traceability, coordination, and sustainability-related decision-making [9]. A key challenge lies in the fragmented nature of the literature across different decision-making levels, which limits the development of a coherent understanding. Much of the existing research emphasizes macro-level strategies, whereas comparatively less attention is given to operational practices such as logistics and consumer behavior [21,22]. This imbalance creates a disconnect between high-level strategic intentions and their practical implementation. Similarly, the tactical level-particularly in areas such as network design and supplier collaboration-remains underexplored. This leaves important gaps in understanding how strategic objectives are translated into operational outcomes [21,23].

Beyond structural fragmentation, SSCM research is further challenged by a lack of theory-driven synthesis. Existing reviews suggest that the field remains conceptually dispersed and under-theorized. Many studies offer descriptive insights without clearly articulating causal mechanisms [24,25]. In particular, prior work highlights the limited integration of theoretical perspectives, noting that key constructs often fail to converge into testable frameworks, especially in explaining the link between sustainability practices and economic performance [24]. At this point, the circular economy (CE) represents an emerging sustainability-oriented paradigm aimed at minimizing resource consumption and waste. It achieves this by prolonging the use of materials, products, and resources through closed-loop resource flows, reuse, remanufacturing, recycling, reverse logistics, and resource recovery. By transcending traditional linear SC models of “take-make-dispose,” CE fosters a systemic understanding of the interdependencies among materials, processes, firms, and sustainability outcomes [23]. Although this concept promotes a shift toward systemic thinking, scholars increasingly emphasize the relevance of paradox theory and systems thinking for capturing the inherent tensions within SSCM [8,24,26]. Without a stronger theoretical grounding, the field risks accumulating empirical findings that remain fragmented and lack conceptual coherence.

Several recent studies have attempted to map and structure the SSCM literature using bibliometric and systematic review approaches. For instance, Nimsai et al. (2020) [27] examine the intellectual structure of SSCM through co-citation and science mapping techniques. Amofa et al. (2023) [28] provide a comprehensive overview of global SSCM research trends, with particular emphasis on CE and contextual differences across economies. More recently, Qu and Kim (2024) [29] focus on the role of AI-integrated technologies in SSCM by combining bibliometric analysis with topic modeling. These studies have made important contributions by identifying influential knowledge structures, dominant research clusters, CE-related developments, contextual differences, and technology-oriented research directions in SSCM. However, they offer valuable insights; they generally rely on predefined scopes or domain-specific filtering strategies. As a result, they often focus on particular themes or subsets of the literature. In addition, the validation of emerging themes through complementary analytical methods remains limited in prior research. These limitations are closely linked to the broader challenges observed in SSCM research. The field continues to be characterized by theoretical fragmentation and inconsistent findings, which constrain its ability to provide clear guidance for practice and policy. While existing empirical studies offer valuable insights, these contributions remain dispersed. Therefore, a more integrative perspective is needed. Such a perspective should go beyond descriptive reviews to clarify the intellectual structure of the field and uncover its underlying thematic patterns. In response to these challenges, this study addresses the lack of an integrated analysis that concurrently links the intellectual framework of SSCM, emerging semantic themes, and implications for theory and practice. More specifically, existing SSCM reviews and bibliometric studies have largely examined consolidated knowledge structures, selected thematic domains, or recent developments separately. However, the relationship between conceptually mature research and citation-latency-affected emerging themes remains insufficiently explained. The present study addresses this limitation through an unsupervised and data-driven hybrid design. This design combines VOSviewer-based science mapping with BERTopic-based semantic topic modeling. It enables the analysis to retain the strengths of citation-based bibliometrics while also capturing emerging themes that may not yet be visible in citation networks. This comprehensive perspective is essential for elucidating the evolution of the field, the emergence of new research trajectories, and how accumulated knowledge can more effectively inform both academic advancement and managerial decision-making.

This study adopts a hybrid analytical approach that combines bibliometric science mapping with ML-based topic modeling to examine SSCM research published between 2011 and 2025. Unlike prior studies that mainly focus on systematic synthesis, bibliometric structures, or specific thematic domains, this approach employs an unsupervised, data-driven methodology. It analyzes a broad corpus of SSCM literature without imposing predefined thematic boundaries. This enables the identification of latent structures and emerging themes in a more flexible, less assumption-driven manner. To ensure a reliable representation of the field’s intellectual structure, the analysis focuses on conceptually mature studies while complementing them with recent publications to capture emerging research trends. The novelty and primary contribution of this hybrid framework lie in its dual-dataset design. The framework connects consolidated citation-based knowledge structures with emerging semantic themes and interprets these patterns through established theoretical lenses. Overall, this study aims to establish a comprehensive and data-driven perspective on SSCM research by combining bibliometric science mapping and semantic topic modeling. First, it investigates the evolution of SSCM research through an analysis of citation dynamics, publication trends, and network-based frameworks based on conceptually mature studies. Second, it uncovers the thematic structure of the literature via keyword co-occurrence and a detailed examination of titles and abstracts. It also captures emerging themes through a broader text-based corpus. Finally, by integrating bibliometric and semantic insights, the study identifies consolidated knowledge structures, emerging research themes, and implications for both theoretical advancement and practical application. This structure clarifies the study’s gap-method-insight logic. The gap concerns the limited integration of mature and emerging SSCM knowledge. The method combines citation-based and text-based analytical layers. The insight lies in explaining how stable intellectual structures, emerging semantic developments, and theory–practice implications are connected. The research objectives are outlined accordingly.

RO-1. To map and analyze the evolution of SSCM research between 2011 and 2025 through citation dynamics, publication trends, and network-based structures.

RO-2. To uncover the thematic structure of the literature using keyword co-occurrence and title–abstract analysis, identifying both dominant and emerging research themes in shaping sustainability discourse.

RO-3. To integrate insights from bibliometric science mapping and transformer-based topic modeling to develop a structured and comprehensive understanding of SSCM research, highlighting key research gaps and emerging trajectories that can inform future scholarly inquiry.

The remainder of the paper is structured as follows. Section 2 outlines the study design and methodology. Section 3 presents descriptive and bibliometric analyses of science mapping. In this section, co-occurrence-based thematic patterns and theoretical interpretations are examined using VOSviewer. It also provides a BERTopic analysis and introduces a cross-method semantic correspondence of the findings. Section 4 discusses the results, including research and managerial implications, limitations, and future research directions. Finally, Section 5 concludes the paper.

2. Study Design and Methodology

To ensure methodological rigor, transparency, and reproducibility, this study adopts the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework. PRISMA provides a structured and widely accepted protocol for systematic literature reviews. It enables transparent study selection and explicitly defined inclusion and exclusion criteria [30]. The research design follows a multi-stage methodological approach that integrates systematic screening, bibliometric science mapping, semantic topic modeling, cross-method semantic correspondence analysis, and framework development. The process begins with the formulation of a search strategy to retrieve SSCM-related records from selected academic databases. The final search was conducted on 17 December 2025. The metadata of the retrieved studies were downloaded as separate Excel files from the selected databases. These files included information such as title, authors, abstract, keywords, source title, publication year, DOI, and citation data. Separate datasets were prepared for the highly cited studies used in bibliometric science mapping and for the broader corpus used in semantic topic modeling, which included recent and lower-cited studies. Subsequently, selection and deduplication procedures were applied to eliminate irrelevant and duplicate records. Duplicate records and studies containing irrelevant terms or not meeting the predefined scope were manually identified and removed in Excel for the highly cited studies dataset, based on the available metadata. Eligibility screening was then conducted according to predefined inclusion and exclusion criteria. The PRISMA checklist was used to enhance transparency and methodological rigor (see Supplementary Materials, Table S1). Based on the cleaned Excel dataset, the remaining studies were searched and organized in Mendeley. A dedicated folder was created for the selected records, and the RIS file exported from this folder was used as the input for the VOSviewer analysis. In contrast, the broader Excel dataset was converted into CSV format and used in Python (Version 3.10) for the BERTopic analysis. Because this dataset was substantially larger and was not restricted by citation thresholds, duplicate records were removed in the Python environment. Thus, the final analytical datasets were constructed for both citation-based and text-based analyses. VOSviewer Text Mining and Network Analysis Software (Version 1.6.16) was employed to develop co-occurrence networks based on keywords and abstracts. This analysis revealed the intellectual structure and thematic evolution of SSCM research [31]. Following this step, BERTopic-based topic modeling was applied to recent publications to identify emerging semantic themes that were not yet apparent in citation-based analyses. The results of the VOSviewer and BERTopic analyses were then compared through cross-method semantic correspondence analysis to identify convergent and complementary patterns. Finally, the integrated findings are synthesized into a conceptual framework that connects established intellectual structures, emerging themes, and theory–practice implications in SSCM research.

According to the PRISMA protocol, the primary stages of the literature selection process are outlined in the following subsections.

2.1. Information Sources

The selection of databases was informed by previous studies, including those by Abrizah et al. (2013) [32], Aghaei Chadegani et al. (2013) [33], Bar-Ilan (2010) [34], Mongeon & Paul-Hus (2016) [35], and Vieira & Gomes (2009) [36]. These studies indicate that the Web of Science (WoS) and Scopus are the most widely used databases for literature searches. They also show that most bibliometric analyses rely on data from these two databases. As demonstrated by De Oliveira et al. (2017) [37], WoS and Scopus covered 95% of the articles included in their study. These databases are widely recognized and trusted tools within the research community. They provide access to diverse research topics and recent advances across academic fields, making them valuable sources for comprehensive literature reviews. Based on the findings from previous studies, WoS and Scopus were used as engines and the main search databases in this study. Mendeley 1.19.8 was used as a reference management tool.

2.2. Eligibility Criteria and Screening Rationale

Bibliometric science mapping generally does not use detailed methodological appraisal tools, such as CASP or MMAT, because its focus is not on assessing the methodological rigor of individual studies [19,38]. Instead, it relies on structured inclusion and exclusion criteria, often described as a “coarse sieve,” together with conceptual relevance. Articles that meet these criteria are generally more influential and conceptually developed contributions to the SSCM literature. This supports the construction of stable and interpretable bibliometric networks. The inclusion criteria for this review focused on studies related to SSCM and Green Supply Chain Management (GSCM) published in Social Science Citation Index (SSCI), Science Citation Index (SCI), and SCI-Expanded journals between 2011 and 2025. This period was selected using a predefined search protocol and citation criteria to capture both the consolidation and recent expansion of SSCM research. After the PRISMA-guided screening process, a citation-based calibration step was implemented exclusively to construct the VOSviewer science mapping dataset. Period-specific citation thresholds were applied to select influential and conceptually mature publications, thereby improving the stability and interpretability of the bibliometric structures [39,40]. Highly cited studies tend to exhibit more robust conceptual linkages, whereas recent low-citation publications may introduce fragmentation and noise into network visualizations [41,42]. However, although citation thresholds improve bibliometric stability and interpretability, they may also introduce selection bias. Therefore, the VOSviewer science mapping dataset should be viewed as a citation-filtered core of significant SSCM knowledge structures rather than as a neutral representation of the entire field.

To ensure balanced coverage across publication periods, citation thresholds were defined according to publication year for the construction of the VOSviewer-based science mapping dataset. Articles published between 2011 and 2015 required at least 250 citations, those published between 2016 and 2019 required at least 100 citations, and those published between 2020 and 2023 required more than 30 citations. These thresholds were designed to account for differences in citation accumulation over time and to avoid disadvantaging more recent publications. Given the rapid expansion of SSCM research, applying a single uniform threshold would likely have excluded a substantial portion of newer studies. To improve methodological transparency, the distributional role of each threshold was examined within its respective publication window, and the corresponding retention ratios are reported in

Table 1. Period-Specific Citation Thresholds and Retention Ratios for the Science Mapping Dataset.

Publication Window	Share in the Final Dataset	Citation Threshold	Mean Citations	Median Citations	Distributional Position	Methodological Role
2011–2015	16%	≥250	352.8	229	Slightly above the median; retained approx. 47% of the dataset	Citation-mature core literature
2016–2019	34.9%	≥100	197.3	134	Below median; retained approx. 77% of the dataset	Citation-visible intermediate literature
2020–2023	49.1%	>30	106.3	78.5	Inclusive threshold; retained approx. 94% of the dataset	Recent citation-visible literature under citation latency

. The threshold of 250 citations for 2011–2015 was slightly above the median and retained approximately 47% of the records, representing a citation-mature core subset of earlier studies. The threshold of 100 citations for 2016–2019 retained approximately 77% of the records, reflecting a more inclusive calibration for the intermediate period. Similarly, the threshold of more than 30 citations for 2020–2023 retained approximately 94% of the records, accounting for the stronger citation-latency effect in recent publications. The share of studies in the final dataset was not expected to be equal across publication windows because SSCM research has expanded substantially in recent years. The period-specific thresholds were therefore intended to balance citation maturity with temporal coverage. Earlier publications had longer citation accumulation periods and were filtered more selectively to retain a mature core subset. In contrast, recent publications were included more inclusively to avoid missing relevant studies due to citation latency.

2.3. Search Strategy

The search strategy was designed to capture the core and most established terminology of the SSCM field while maintaining conceptual focus and avoiding excessive scope expansion. The search terms were applied to the title, abstract, and keyword fields in Scopus using TITLE-ABS-KEY and to the topic field in Web of Science using TS. A structured Boolean search strategy was implemented to ensure both breadth and precision. In Scopus, the following search string was used: TITLE-ABS-KEY (“Sustainable Supply Chain Management” OR “Sustainable Supply Chain” OR (“Sustainability” AND “Supply Chain Management”) OR (“Green” AND “Supply Chain Management”)). In Web of Science, the equivalent search string was applied as follows: TS = (“Sustainable Supply Chain Management” OR “Sustainable Supply Chain” OR (“Sustainability” AND “Supply Chain Management”) OR (“Green” AND “Supply Chain Management”)). The OR operators linked synonymous or related terms, allowing the inclusion of studies that used different terminology. The AND operator ensured that both sustainability and SCM were explicitly addressed in the studies. Quotation marks were placed around multi-word phrases to improve accurate results.

The review focused on dominant and widely indexed SSCM terminology to ensure consistency in the construction of the bibliometric and semantic datasets. This approach is also consistent with prior bibliometric reviews in the field. Expanding the search string to include all adjacent terms could have substantially broadened the corpus. For example, it could have included studies on corporate sustainability, ESG reporting, or responsibility issues without a clear SCM focus. GSCM was included in the search strategy because it represents a foundational and historically important stream within the broader development of SSCM. While SSCM encompasses the environmental, social, and economic dimensions of sustainability, GSCM primarily focuses on the environmental aspect of SCM. It has contributed substantially to SSCM through concepts such as green purchasing, eco-design, cleaner production, reverse logistics, recycling, waste reduction, and environmental collaboration with suppliers [11]. Excluding GSCM would therefore risk omitting early and influential studies that shaped the intellectual structure of SSCM. Therefore, the search was intentionally scoped to SSCM and GSCM terminology. To reduce the risk of excessive broadening, GSCM-related studies were included only when their titles, abstracts, or keywords explicitly referenced SCM alongside a sustainability-related concept. This assessment was conducted during the title and abstract screening stage.

2.4. Selection Process

A two-step screening process was used to determine study eligibility. First, titles and abstracts were reviewed to identify potentially relevant studies based on the search strategy. Second, full-text articles were assessed against the predefined inclusion and exclusion criteria. No automation tools were used; therefore, all screening decisions were manually verified.

2.5. Data Analysis and Synthesis Methods

The synthesis of the studies followed a qualitative approach to summarize key trends, themes, and gaps in the SSCM literature. The studies were first categorized by article type to identify the eligible set and then further analyzed using citation thresholds. Descriptive statistics were visualized through bar charts, pie charts, and distribution plots in MS Excel. In addition, keyword and abstract co-occurrence maps were created using VOSviewer to illustrate relationships among common themes. The qualitative synthesis focused on interpreting these patterns and identifying gaps in the literature.

2.6. Methodological Considerations and Bias Mitigation

Although a formal assessment of reporting bias was not conducted, this review follows established bibliometric practices to enhance the robustness and interpretability of the analysis. Citation thresholds were applied to identify influential and conceptually mature publications. This supported the construction of stable and meaningful co-occurrence structures. It is important to emphasize that this approach does not imply that studies with lower citation counts are of lower quality. Rather, it reflects the methodological need for sufficient citation density in reliable science mapping analyses. During the selection process, many recent publications, particularly those from 2023, did not meet the citation thresholds because of citation latency. Including these studies in the core science mapping dataset could have compromised the stability of co-occurrence structures and the reliability of thematic interpretations. To address this limitation, publications from 2021 to 2025 were excluded from the core bibliometric analysis but incorporated into a complementary BERTopic analysis. This complementary approach allowed recent low-citation studies to contribute to the identification of emerging themes while preserving the methodological rigor of the citation-based analysis. By integrating bibliometric science mapping with ML-based topic modeling, the study provides a balanced perspective that captures both established research structures and evolving thematic developments.

2.7. Data Extraction and Variables

The review aimed to analyze trends in SSCM studies, with a focus on publication patterns, citation trends, and emerging themes. It also identified key components and gaps in the literature, including research objectives and contributions. Data on publication year, citation counts, study types, and research methods were extracted from the full-text articles of the selected studies. Following the study selection phase, a co-word cluster bibliometric network analysis was conducted using VOSviewer. This analysis was used to examine the intellectual structure and thematic patterns of SSCM research. It enabled the identification of research trends, conceptual relationships, and dominant clusters within the literature. In this context, bibliometric analysis provides a quantitative and systematic means of mapping the research landscape [43,44]. To further enhance the robustness of the findings and address the limitations of citation-based filtering for recent studies, a cross-method semantic correspondence analysis was performed using BERTopic. This machine learning-based approach enabled the identification of emerging themes from recent low-citation studies. It therefore offered a more balanced and forward-looking perspective on SSCM research.

3. Results

3.1. Study Selection Process

The initial search followed the PRISMA protocol, as outlined in Figure 1, to systematically delineate the study selection process. This process distinguished between the broad search pool, the qualified review corpus, and the two analytical datasets used in the study. The search initially generated 799,696 records. After applying database restrictions, the review was limited to peer-reviewed, English-language articles indexed in SSCI, SCI, and SCI-Expanded between 2011 and 2025. Duplicate records, irrelevant entries, and non-article document types, such as conference papers and editorials, were then removed. As a result, 899 articles were identified as the PRISMA-eligible SSCM corpus, which established the core boundary of the systematic review. From this corpus, a citation-filtered subset of 212 highly cited articles was selected for VOSviewer-based bibliometric science mapping. This dataset was designed to capture established intellectual structures, as citation thresholds support the formation of coherent co-occurrence networks. In parallel, a broader text-based dataset comprising 3799 articles was created for BERTopic analysis (see Section 3.3). Both analytical datasets were constructed using the database-specific search strings reported in Section 2.3, where the search terms were applied in Scopus through TITLE-ABS-KEY and in Web of Science through TS. The BERTopic analysis was intended to reveal emerging semantic themes, particularly from recent and less-cited studies that may not yet appear in citation-based structures. Accordingly, the three datasets were constructed for different but complementary analytical purposes. The 899 articles defined the PRISMA-eligible review boundary. From this set, 212 highly cited articles formed the VOSviewer-based science mapping dataset, which was used to capture the established intellectual structure of SSCM research. By contrast, the 3799-article BERTopic dataset was broader and was not restricted by citation thresholds. It therefore enabled the identification of recent, lower-cited, and emerging semantic themes. Thus, the VOSviewer and BERTopic datasets should be understood as complementary rather than identical analytical datasets.

Table 2. Overview of the Data Collection and Study Design (Source: Authors).

Research Databases	WoS and Scopus
Publication Time Frame	2011–2025
Publication Type	Only peer-reviewed journals.
Article Type	Review, Original research, Case study
Article Language	English
Unit of Analysis	Document-level metadata (authors, journals, publication years, citation counts, DOIs), and text (keywords, titles, abstracts).
Inclusion/Exclusion Criteria	Inclusion: Articles published in SCI, SCI-Expanded, and SSCI journals were included. Exclusion: Conference papers, short notes, book chapters, and editorial notes were excluded.
Eligibility Criteria	- Articles published between 2011 and 2015 needed at least 250 citations. - From 2015 to 2019, a minimum of 100 citations were required. - From 2019 to 2023, more than 30 citations were needed. - From 2021 to 2025, included only in ML-based analysis due to citation latency.
PRISMA Screening Rationale	Two-stage selection following PRISMA guidelines: Coarse sieve: title–abstract screening for thematic relevance. Fine sieve: full-text screening against inclusion/exclusion criteria.
Search Terms	(“Sustainable Supply Chain Management” OR “Sustainable Supply Chain” OR (“Sustainability” AND “Supply Chain Management”) OR (“Green” AND “Supply Chain Management”))
Search Strategy	Boolean search; synonyms combined using OR; conceptual linkages defined using AND; multi-word concepts enclosed in quotation marks.
Core Bibliometric Analysis	Keyword co-occurrence analysis; title–abstract co-occurrence analysis; cluster detection via VOSviewer; overlay visualization for temporal interpretation.
Cross-Method Semantic Correspondence Analysis	Although citation counts are low, recent studies (2021–2025) were analyzed using BERTopic to validate thematic evolution and capture emerging research trends.
Softwares and Tools	VOSviewer 1.6.16 (science mapping), MS Excel (data extraction, descriptive analyses), Python 3.10 (BERTopic).

presents a structured overview of the study design and data collection process adopted in this study.

The selected studies by the PRISMA, comprising literature reviews, original research, and case studies, were analyzed using VOSviewer. A complete list of these papers is provided in Appendix A,

Table A1. List of The Studies Selected Based on The PRISMA Framework.

Reference Number	Author(s)	Year	Methodological Category	DOI
[7]	Koberg & Longoni	2019	Review-based study	10.1016/j.jclepro.2018.10.033
[11]	Brandenburg et al.	2019	Conceptual study	10.3390/su11247239
[15]	Cloutier et al.	2020	Review-based study	10.1080/00207543.2019.1660821
[17]	Ansari & Kant	2017	Review-based study	10.1016/j.jclepro.2016.11.023
[20]	Patel & Desai	2019	Review-based study	10.1080/13675567.2018.1534946
[26]	Sarkis et al.	2021	Conceptual study	10.1108/IMDS-08-2020-0450
[46]	Karmaker et al.	2023	Empirical study	10.1016/j.jclepro.2023.138249
[47]	Abbasi et al.	2023	Optimization study	10.1016/j.jclepro.2023.137935
[48]	Da Silva et al.	2023	Case-based study	10.1016/j.resconrec.2023.106969
[49]	Khan et al.	2023	Empirical study	10.1016/j.jclepro.2023.136609
[50]	Liu et al.	2023	Conceptual study	10.1016/j.cie.2023.109113
[51]	Mridha et al.	2023	Optimization study	10.1016/j.jclepro.2022.135629
[52]	Debnath et al.	2023	Optimization study	10.1016/j.jclepro.2022.135477
[53]	Michael Rodriguez-Gonzalez et al.	2022	Empirical study	10.1016/j.jclepro.2022.134670
[54]	Khan et al.	2023	Empirical study	10.1002/bse.3207
[55]	Adams et al.	2023	Case-based study	10.1002/bse.3198
[56]	Shekarian et al.	2022	Review-based study	10.3390/su14137892
[57]	Kazancoglu et al.	2022	Empirical study	10.1016/j.jclepro.2022.132431
[58]	Agrawal et al.	2023	Decision-support study	10.1007/s10668-022-02396-2
[59]	Bai et al.	2022	Decision-support study	10.1016/j.jclepro.2022.131896
[60]	Kayikci et al.	2022	Decision-support study	10.1002/bse.3087
[61]	Mogale et al.	2022	Optimization study	10.1016/j.cie.2022.108105
[62]	Kunkel et al.	2022	Case-based study	10.1016/j.resconrec.2022.106274
[63]	Mangla et al.	2022	Decision-support study	10.1002/bse.3027
[64]	Faramarzi-Oghani et al.	2023	Review-based study	10.1080/00207543.2022.2045377
[65]	Xu et al.	2022	Empirical study	10.1108/JEIM-06-2021-0260
[66]	Sheng et al.	2023	Review-based study	10.1007/s10668-022-02109-9
[67]	Michael Rodriguez-Gonzalez et al.	2022	Empirical study	10.1002/csr.2233
[68]	Rajaeifar et al.	2022	Review-based study	10.1016/j.resconrec.2021.106144
[69]	Guo et al.	2022	Optimization study	10.1016/j.tre.2021.102593
[70]	Kim et al.	2022	Empirical study	10.1080/00207543.2021.1992032
[71]	Raian et al.	2022	Decision-support study	10.1016/j.resconrec.2021.105975
[72]	Paul et al.	2023	Empirical study	10.1108/IJLM-04-2021-0238
[73]	Wang & Rani	2022	Decision-support study	10.1108/JEIM-05-2021-0222
[74]	Yang et al.	2022	Decision-support study	10.1108/JEIM-06-2021-0261
[75]	Centobelli et al.	2021	Empirical study	10.1016/j.ijpe.2021.108297
[76]	Kazancoglu et al.	2023	Decision-support study	10.1080/09537287.2021.1980910
[77]	Kumar et al.	2021	Decision-support study	10.1016/j.resconrec.2021.105879
[78]	Bechtsis et al.	2022	Review-based study	10.1080/00207543.2021.1957506
[79]	Acero et al.	2022	Conceptual study	10.1111/jscm.12269
[80]	Nayal et al.	2021	Conceptual study	10.1108/JEIM-09-2020-0381
[81]	Paul et al.	2021	Review-based study	10.3390/su13137104
[82]	Kusi-Sarpong et al.	2021	Decision-support study	10.1016/j.omega.2021.102502
[83]	Jraisat et al.	2023	Case-based study	10.1080/00207543.2021.1936263
[84]	Gong et al.	2023	Case-based study	10.1080/00207543.2021.1930238
[85]	Cui et al.	2023	Decision-support study	10.1080/00207543.2021.1924412
[86]	Siems & Seuring	2021	Case-based study	10.1002/bse.2792
[87]	Liu et al.	2021	Optimization study	10.1016/j.tre.2021.102319
[88]	Daddi et al.	2021	Empirical study	10.1002/csr.2144
[89]	Olan et al.	2022	Empirical study	10.1080/00207543.2021.1915510
[90]	Ma et al.	2021	Optimization study	10.1016/j.tre.2021.102290
[91]	Edwin Cheng et al.	2022	Empirical study	10.1080/00207543.2021.1906971
[92]	Dominguez et al.	2021	Optimization study	10.1016/j.omega.2020.102268
[93]	Feng et al.	2023	Decision-support study	10.1080/00207543.2021.1890260
[94]	Sun et al.	2021	Empirical study	10.1002/csr.2128
[95]	Koot et al.	2021	Review-based study	10.1016/j.cie.2020.107076
[96]	Mc Loughlin et al.	2023	Conceptual study	10.1080/09537287.2021.1884764
[97]	Liu et al.	2021	Optimization study	10.1016/j.tre.2021.102237
[98]	Tsai et al.	2021	Decision-support study	10.1016/j.resconrec.2021.105421
[99]	Hu et al.	2021	Conceptual study	10.1016/j.cie.2020.107079
[100]	Khan et al.	2021	Review-based study	10.1016/j.jclepro.2020.123357
[101]	Raut et al.	2021	Empirical study	10.1016/j.tre.2020.102170
[102]	Zhou et al.	2021	Review-based study	10.1016/j.omega.2020.102295
[103]	Esmaeilian et al.	2020	Review-based study	10.1016/j.resconrec.2020.105064
[104]	Banik et al.	2022	Decision-support study	10.1080/13675567.2020.1839029
[105]	Chirra et al.	2021	Decision-support study	10.1080/00207543.2020.1832272
[106]	Khan et al.	2021	Empirical study	10.1002/csr.2057
[107]	Nayeri et al.	2020	Optimization study	10.1016/j.cie.2020.106716
[108]	Singh et al.	2020	Empirical study	10.1108/IJLSS-06-2017-0068
[109]	Sundarakani et al.	2021	Optimization study	10.1108/IJLM-12-2019-0333
[110]	Sanchez-Flores et al.	2020	Review-based study	10.3390/su12176972
[111]	Zhang et al.	2021	Review-based study	10.1108/JEIM-12-2019-0381
[112]	Agyabeng-Mensah et al.	2020	Empirical study	10.1108/IJLM-10-2019-0275
[113]	Tseng et al.	2022	Decision-support study	10.1080/13675567.2020.1800608
[114]	Singh et al.	2021	Review-based study	10.1080/00207543.2020.1792000
[115]	Hong et al.	2022	Empirical study	10.1080/13675567.2020.1795094
[116]	Sharma et al.	2020	Review-based study	10.1016/j.cor.2020.104926
[117]	Nazam et al.	2020	Decision-support study	10.1108/JEIM-09-2019-0271
[118]	Yu et al.	2021	Empirical study	10.1080/09537287.2020.1774675
[119]	Bird & Soundararajan	2020	Conceptual study	10.1007/s10551-018-4067-z
[120]	Hussain & Malik	2020	Decision-support study	10.1016/j.jclepro.2020.120375
[121]	Alghababsheh & Gallear	2021	Empirical study	10.1007/s10551-020-04525-1
[122]	Lis et al.	2020	Review-based study	10.3390/su12103987
[123]	Fattahi et al.	2021	Optimization study	10.1080/00207543.2020.1746427
[124]	Jiang et al.	2020	Empirical study	10.1108/IMDS-07-2019-0394
[125]	Tseng et al.	2022	Empirical study	10.1080/13675567.2020.1749577
[126]	Yadav & Singh	2021	Decision-support study	10.1108/JEIM-09-2019-0301
[127]	Bag et al.	2021	Empirical study	10.1108/JEIM-10-2019-0324
[128]	Inamdar et al.	2021	Review-based study	10.1108/JEIM-09-2019-0267
[129]	Mardani et al.	2020	Review-based study	10.1016/j.jclepro.2019.119383
[130]	Bux et al.	2020	Decision-support study	10.1002/csr.1920
[131]	Van Engeland et al.	2020	Review-based study	10.1016/j.omega.2018.12.001
[132]	Dewi et al.	2023	Empirical study	10.3390/su15021148
[133]	He et al.	2021	Decision-support study	10.1080/00207543.2020.1724343
[134]	Frei et al.	2020	Optimization study	10.1002/bse.2479
[135]	Digalwar et al.	2020	Decision-support study	10.1002/bse.2455
[136]	Han & Huo	2020	Empirical study	10.1108/IMDS-07-2019-0373
[137]	Jazairy & von Haartman	2020	Empirical study	10.1080/13675567.2019.1584163
[138]	Vivas et al.	2020	Optimization study	10.1016/j.cie.2019.01.044
[139]	Shashi et al.	2020	Review-based study	10.1002/bse.2428
[140]	Shareef et al.	2020	Case-based study	10.1080/09537287.2019.1695917
[141]	Laosirihongthong et al.	2020	Decision-support study	10.1080/09537287.2019.1701233
[142]	Pan et al.	2020	Empirical study	10.1080/09537287.2019.1631457
[143]	Zimon et al.	2019	Decision-support study	10.3390/su11247227
[144]	Meherishi et al.	2019	Review-based study	10.1016/j.jclepro.2019.07.057
[145]	Carter et al.	2019	Review-based study	10.1108/IJPDLM-02-2019-0056
[146]	Gong et al.	2019	Empirical study	10.1016/j.ijpe.2019.01.033
[147]	Jia et al.	2019	Case-based study	10.1016/j.ijpe.2018.07.022
[148]	Manupati et al.	2020	Decision-support study	10.1080/00207543.2019.1683248
[149]	Narayanan et al.	2019	Decision-support study	10.1108/JMTM-06-2017-0114
[150]	Dahlmann & Rohrich	2019	Empirical study	10.1002/bse.2392
[151]	Govindan et al.	2019	Optimization study	10.1016/j.cor.2018.11.013
[152]	Hu et al.	2019	Empirical study	10.1016/j.resconrec.2019.05.042
[153]	Bai et al.	2020	Empirical study	10.1080/00207543.2019.1661532
[154]	Mehdikhani & Valmohammadi	2019	Empirical study	10.1108/JEIM-07-2018-0166
[155]	Sajjad et al.	2020	Empirical study	10.1002/bse.2389
[156]	Tseng et al.	2019	Decision-support study	10.1016/j.jclepro.2019.04.201
[157]	Seuring et al.	2019	Case-based study	10.1016/j.jclepro.2018.12.102
[158]	Yun et al.	2019	Review-based study	10.1108/IJLM-05-2017-0112
[159]	Saeed et al.	2019	Review-based study	10.3390/su11041137
[160]	Rebs et al.	2019	Review-based study	10.1016/j.jclepro.2018.10.100
[161]	Ardakani & Soltanmohammadi	2019	Empirical study	10.1002/csr.1671
[162]	Gardas et al.	2019	Case-based study	10.1016/j.spc.2018.11.005
[163]	Xiao et al.	2019	Case-based study	10.1111/jscm.12170
[164]	Carter & Washispack	2018	Review-based study	10.1111/jbl.12196
[165]	Chacon Vargas et al.	2018	Empirical study	10.1016/j.resconrec.2018.08.018
[166]	Das	2018	Empirical study	10.1016/j.jclepro.2018.08.250
[167]	Luthra & Mangla	2018	Decision-support study	10.1016/j.resconrec.2018.07.005
[168]	Badiezadeh et al.	2018	Decision-support study	10.1016/j.cor.2017.06.003
[169]	Kaur & Singh	2018	Optimization study	10.1016/j.cor.2017.05.008
[170]	Silvestre et al.	2018	Empirical study	10.1016/j.jclepro.2018.05.127
[171]	Jia et al.	2018	Review-based study	10.1016/j.jclepro.2018.03.248
[172]	Qorri et al.	2018	Review-based study	10.1016/j.jclepro.2018.04.073
[173]	Luthra & Mangla	2018	Decision-support study	10.1016/j.psep.2018.04.018
[174]	Ahmed & Sarkar	2018	Optimization study	10.1016/j.jclepro.2018.02.289
[175]	Bastas & Liyanage	2018	Review-based study	10.1016/j.jclepro.2018.01.110
[176]	Moktadir et al.	2018	Decision-support study	10.1016/j.jclepro.2018.01.245
[177]	Kot	2018	Review-based study	10.3390/su10041143
[178]	Castillo et al.	2018	Conceptual study	10.1111/jbl.12176
[179]	Hong et al.	2018	Empirical study	10.1016/j.jclepro.2017.06.093
[180]	Allaoui et al.	2018	Optimization study	10.1016/j.cor.2016.10.012
[181]	Hosseinifard & Abbasi	2018	Optimization study	10.1016/j.cor.2016.08.014
[182]	Kazancoglu et al.	2018	Decision-support study	10.1108/IMDS-03-2017-0121
[183]	Mathivathanan et al.	2018	Empirical study	10.1016/j.resconrec.2017.01.003
[184]	Roy et al.	2018	Review-based study	10.1108/IJOPM-05-2017-0260
[185]	Wang, Jing; Dai, Jun	2018	Empirical study	10.1108/IMDS-12-2016-0540
[186]	Zhang et al.	2018	Empirical study	10.1016/j.resconrec.2016.06.015
[187]	Chen et al.	2017	Review-based study	10.1016/j.ijpe.2017.04.005
[188]	Ahmadi et al.	2017	Decision-support study	10.1016/j.resconrec.2017.07.020
[189]	Ansari & Kant	2017	Review-based study	10.1002/bse.1945
[190]	Lim et al.	2017	Decision-support study	10.1016/j.jclepro.2017.06.056
[191]	Rajeev et al.	2017	Review-based study	10.1016/j.jclepro.2017.05.026
[192]	Oelze	2017	Case-based study	10.3390/su9081435
[193]	Stindt	2017	Optimization study	10.1016/j.jclepro.2017.03.126
[194]	Madani & Rasti-Barzoki	2017	Optimization study	10.1016/j.cie.2017.01.017
[195]	Bechtsis et al.	2017	Review-based study	10.1016/j.jclepro.2016.10.057
[196]	Dubey et al.	2017	Review-based study	10.1016/j.jclepro.2016.03.117
[197]	Busse et al.	2017	Empirical study	10.1111/jscm.12129
[198]	Dubey et al.	2017	Review-based study	10.1108/IJLM-07-2015-0112
[199]	Genovese et al.	2017	Optimization study	10.1016/j.omega.2015.05.015
[200]	Reefke & Sundaram	2017	Review-based study	10.1016/j.omega.2016.02.003
[201]	Varsei & Polyakovskiy	2017	Optimization study	10.1016/j.omega.2015.11.009
[202]	Su et al.	2016	Decision-support study	10.1016/j.jclepro.2015.05.080
[203]	Schoeggl et al.	2016	Decision-support study	10.1016/j.jclepro.2016.04.035
[204]	Fahimnia & Jabbarzadeh	2016	Optimization study	10.1016/j.tre.2016.02.007
[205]	Luthra et al.	2016	Decision-support study	10.1016/j.jclepro.2016.01.095
[206]	Oelze et al.	2016	Empirical study	10.1002/bse.1869
[207]	Busse	2016	Empirical study	10.1111/jscm.12096
[208]	Markman & Krause	2016	Conceptual study	10.1111/jscm.12105
[209]	Hussain et al.	2016	Empirical study	10.1016/j.rser.2015.07.097
[210]	Chiappetta et al.	2016	Conceptual study	10.1016/j.jclepro.2015.01.052
[211]	Formentini & Taticchi	2016	Case-based study	10.1016/j.jclepro.2014.12.072
[212]	Govindan et al.	2016	Conceptual study	10.1016/j.jclepro.2015.11.084
[213]	Lin & Tseng	2016	Decision-support study	10.1016/j.jclepro.2014.07.012
[214]	Giannakis & Papadopoulos	2016	Empirical study	10.1016/j.ijpe.2015.06.032
[215]	Matthews et al.	2016	Conceptual study	10.1111/jscm.12097
[216]	Luthra et al.	2015	Decision-support study	10.1016/j.resourpol.2014.12.006
[217]	Taticchi et al.	2015	Review-based study	10.1080/00207543.2014.939239
[218]	Sajjad et al.	2015	Empirical study	10.1002/bse.1898
[219]	Mota et al.	2015	Optimization study	10.1016/j.jclepro.2014.07.052
[220]	Govindan et al.	2015	Optimization study	10.1016/j.cor.2014.12.014
[221]	Marshall et al.	2015	Empirical study	10.1080/09537287.2014.963726
[222]	Brandenburg & Rebs	2015	Review-based study	10.1007/s10479-015-1853-1
[223]	Luthra et al.	2015	Conceptual study	10.1080/09537287.2014.904532
[224]	Azadi et al.	2015	Decision-support study	10.1016/j.cor.2014.03.002
[225]	Boukherroub et al.	2015	Optimization study	10.1016/j.cor.2014.09.002
[226]	Validi et al.	2015	Optimization study	10.1016/j.cor.2014.06.015
[227]	Xie, Gang	2015	Optimization study	10.1016/j.cor.2013.11.020
[228]	Meixell & Luoma	2015	Conceptual study	10.1108/IJPDLM-05-2013-0155
[229]	Tseng et al.	2015	Decision-support study	10.1108/IMDS-10-2014-0319
[230]	Diabat et al.	2014	Case-based study	10.1016/j.jclepro.2014.06.081
[231]	Turke & Altuntas	2014	Case-based study	10.1016/j.emj.2014.02.001
[232]	Beske et al.	2014	Review-based study	10.1016/j.ijpe.2013.12.026
[233]	Brandenburg et al.	2014	Review-based study	10.1016/j.ejor.2013.09.032
[234]	Pagell & Shevchenko	2014	Conceptual study	10.1111/jscm.12037
[235]	Gold et al.	2013	Case-based study	10.1016/j.ibusrev.2012.12.006
[236]	Morali & Searcy	2013	Case-based study	10.1007/s10551-012-1539-4
[237]	Al Zaabi et al.	2013	Decision-support study	10.1007/s00170-013-4951-8
[238]	Ahi & Searcy	2013	Review-based study	10.1016/j.jclepro.2013.02.018
[239]	Harms et al.	2013	Empirical study	10.1002/csr.1293
[240]	Golicic & Smith	2013	Review-based study	10.1111/jscm.12006
[241]	Winter & Knemeyer	2013	Review-based study	10.1108/09600031311293237
[242]	Ageron et al.	2012	Empirical study	10.1016/j.ijpe.2011.04.007
[243]	Gopalakrishnan et al.	2012	Case-based study	10.1016/j.ijpe.2012.01.003
[244]	Zailani et al.	2012	Empirical study	10.1016/j.ijpe.2012.02.008
[245]	Wittstruck & Teuteberg	2012	Empirical study	10.1002/csr.261
[246]	Beske	2012	Conceptual study	10.1108/09600031211231344
[247]	Seuring	2011	Review-based study	10.1002/bse.702
[248]	Wolf	2011	Empirical study	10.1007/s10551-011-0806-0
[249]	Carter et al.	2011	Review-based study	10.1108/09600031111101420
[250]	Zhang et al.	2022	Case-based study	10.1016/j.resconrec.2022.106536
[251]	Cai & Choi	2020	Conceptual study	10.1016/j.tre.2020.102010

[7,11,15,17,20,26,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,225,226,227,228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251].

3.2. Bibliometric and Science Mapping Analysis of Highly Cited Studies

3.2.1. Descriptive Analysis by Journal, Document Type, and Citation Metrics

A significant number of publications in the SSCM domain address diverse topics across multiple disciplines, including business, engineering, management, computer science, decision-making, and operations [20,252]. In accordance with the inclusion and exclusion criteria outlined in the methodology, 212 highly cited SSCM articles were analyzed and classified according to journal and document type. Figure 2 presents the distribution of publications across leading international journals, based on their share within the selected dataset. The results indicate a clear concentration of publications in a limited number of journals. The Journal of Cleaner Production accounts for 19% of the selected studies, followed by the International Journal of Production Research (8%), Business Strategy and the Environment (7%), Resources, Conservation & Recycling (6%), and Sustainability (5%). In contrast, journals ranked in the lower tier of the top 20 contribute only marginal shares, with individual contributions of around 1%. This finding indicates that SSCM research is concentrated within a core group of journals, while a long tail of journals contributes smaller but diverse perspectives to the field.

The concentration of journals in the SSCM field significantly shapes its structure and indicates a degree of intellectual consolidation. A limited number of journals serve as central hubs for discussions on sustainability performance, green practices, circular economy, and supply chain coordination. However, the journal distribution also reveals an imbalance in thematic focus. Environmentally oriented journals, such as the Journal of Cleaner Production, Business Strategy and the Environment, and Resources, Conservation & Recycling, are particularly prominent. This pattern highlights the historical emphasis on environmental issues over social and economic dimensions. At the same time, a disciplinary divide persists between technically oriented journals, such as the International Journal of Production Research and Computers & Operations Research, and managerially focused outlets, such as the Journal of Supply Chain Management and Corporate Social Responsibility and Environmental Management. This divide has led to overlapping but fragmented dialogues. The combination of consolidation and fragmentation points to a critical challenge in SSCM. Although the field has an identifiable intellectual core, its theoretical insights and practical applications remain dispersed across different publication communities. This segmented knowledge landscape emphasizes the need for a cohesive analytical framework that integrates intellectual structures, emerging themes, and theory–practice implications.

The articles were classified into three document types, with original research constituting the dominant category in the SSCM literature. A proportional analysis was conducted to examine the distribution of article types across journals while accounting for differences in publication volume. The results indicate that Production Planning & Control and the Journal of Business Ethics have the highest proportion of case study articles (75%), whereas Computers & Operations Research has the highest share of original research articles (73%). In contrast, Sustainability publishes the largest proportion of review papers (64%). Citation analysis further reveals differences in impact across document types. As shown in Figure 3, original research articles account for the highest total number of citations (15.933), followed closely by review papers (15.770). Case studies receive comparatively fewer citations (10.542).

To ensure comparability across journals, citation counts were normalized by the number of publications in each category. This allowed for a more balanced assessment of citation impact. Building on this normalized perspective, Figure 4 presents the distribution of citation impact across journals. The International Journal of Physical Distribution & Logistics Management records the highest citation count (873), followed by the International Journal of Production Economics (739). Although the Journal of Cleaner Production is the most prolific publisher in the SSCM field, its citation impact per article appears comparatively lower (412). This difference may be partly explained by variations in the types of articles published. Journals with a higher share of review and conceptual studies often show stronger citation performance than those publishing more case-based research.

3.2.2. Temporal Analysis of Publications and Citation Trends

An annual analysis of SSCM publications was conducted to examine temporal trends in the field. Figure 5 presents the yearly distribution of studies from 2011 to 2023. The results reveal a generally increasing trajectory, with moderate fluctuations over time. A notable rise in publication volume is observed around 2020–2021. This increase may be associated with heightened academic interest during the COVID-19 period and the growing relevance of resilient and SSCs. A decline in the number of publications is observed in 2022 and becomes more pronounced in 2023. However, this decrease should be interpreted with caution, as it primarily reflects methodological constraints rather than an actual reduction in research activity. Specifically, the application of citation thresholds limits the inclusion of recently published studies, as these publications have not yet accumulated sufficient citations to meet the selection criteria. The year 2023 can be considered partially citation-mature, whereas publications from 2024 and 2025 remain in the citation-latency phase. Therefore, different analytical treatments are methodologically justified. Publications from 2024 and 2025 were excluded from the core analysis because of insufficient citation accumulation. Despite these limitations, the overall trend indicates sustained growth in SSCM research. This suggests that the field has evolved from an emerging topic into a more established and expanding area of study.

Figure 6 illustrates the annual distribution of SSCM publications by paper type. The results show a clear increase in the number of case study publications in recent years. This indicates a growing emphasis on empirical and application-oriented research within the field. The trend also suggests that, as SSCM gains wider recognition, journals increasingly value context-specific and practice-driven contributions. Original research articles represent the second-most prevalent paper type and closely follow case studies in publication volume. Their steady growth reflects the expanding research capacity and methodological diversity of the SSCM domain. In contrast, review papers remain relatively limited in number, particularly in recent years. This pattern indicates a shift in publication priorities toward original and application-focused studies.

The final analysis in this section examines unit citations by paper type over time (see Figure 7). Unit citations are defined as the ratio of total citations to the number of publications in each year. The results indicate that review papers achieved the highest unit citation values in the early stage of the period, particularly in 2011. However, citation patterns show considerable variability in subsequent years. From 2014 onward, unit citation values for original research articles, case studies, and review papers converge to relatively similar levels. This convergence suggests a gradual shift in the distribution of scholarly impact. In this pattern, the citation advantages traditionally associated with review papers become less pronounced over time. One possible explanation is the increasing maturity of the SSCM field. As the field develops, high-quality original and application-oriented studies may attract comparable academic attention. In addition, the expansion of publication volume may have contributed to a more balanced citation landscape across different paper types.

3.2.3. Co-Occurrence-Based Thematic Analysis Using VOSviewer

Text mining enables the extraction of meaningful patterns from large volumes of unstructured textual data. It facilitates the identification of recurring concepts, semantic relationships, and thematic structures in scientific literature [253]. In bibliometric studies, text mining provides essential input by transforming textual information, such as titles, abstracts, and keywords, into structured formats. These structured data can then be analyzed through co-occurrence and network-based approaches. Building on this foundation, the present study employs bibliometric science mapping to explore the conceptual and intellectual structure of SSCM research. Unlike descriptive bibliometric analysis, which focuses mainly on publication statistics, science mapping examines relationships between scholarly elements by constructing networks based on co-occurrence patterns [254].

To achieve this, the study employed the VOS mapping technique developed by Van Eck and Waltman. This technique is a widely used distance-based visualization algorithm in bibliometric research and is known for producing stable and interpretable maps [255,256,257,258,259,260,261,262]. It enables the identification of thematic clusters, conceptual linkages, and emerging research trends within the SSCM literature. While author keywords reflect the intentional representation of core concepts, terms derived from titles and abstracts capture broader thematic and methodological relationships.

In this study, VOSviewer was employed to conduct a co-occurrence-based thematic analysis of both author keywords and title–abstract terms. By examining the frequency with which terms co-occur across publications, bibliometric networks were constructed. In these networks, nodes represent terms, and links reflect co-occurrence relationships. Author keywords provide a focused representation of core concepts, whereas title–abstract terms capture broader thematic and methodological patterns [258,260]. Analyzing these datasets separately allows for a more comprehensive understanding of SSCM’s intellectual structure. It also enables the identification of dominant themes, emerging topics, and their interrelationships.

Keyword Co-Occurrence Analysis

A keyword co-occurrence analysis was performed using VOSviewer, based on bibliographic data exported in RIS format from Mendeley. The full counting method was applied, with co-occurrence specified as the type of analysis and keywords as the unit of analysis. Among the 629 unique keywords identified in the dataset, 124 met the minimum occurrence threshold of two. Because all 124 keywords satisfied this threshold, the full set was retained in accordance with VOSviewer’s default recommendations. Thesaurus and stopword files were used to consolidate terminological variants before network construction, and the outputs of these files are provided in Appendix A,

Table A2. VOSviewer Keyword Co-Occurrence Analysis Thesaurus Terms and Stopwords.

Label	Replace by	Thesaurus Outputs	Stopwords
ahp	analytical hierarchy process	analytical hierarchy process	adoption
artificial intelligence	N/A	N/A	base of the pyramid
automotive industry	N/A	N/A	flexibility
best worst method	N/A	N/A	practices
best-worst method	best worst method	best worst method	drivers
big data	N/A	N/A	barriers
big data analytics	N/A	N/A	sustainable
blockchain	N/A	N/A	environment
carbon emissions	N/A	N/A	review
case study	N/A	N/A	risk
case study research	case study	case study	performance
china	N/A	N/A	framework
circular economy	N/A	N/A
closed-loop supply chain	N/A	N/A
collaboration	N/A	N/A
conceptual framework	N/A	N/A
corporate theory building	N/A	N/A
corporate social responsibility	N/A	N/A
corporate sustainability	N/A	N/A
critical success factors	N/A	N/A
critical success factors (csf)	critical success factors	critical success factors
decision-making	N/A	N/A
decision making	decision-making	decision-making
dematel	N/A	N/A
developing countries	N/A	N/A
dynamic capabilities	N/A	N/A
economic performance	N/A	N/A
emerging economies	N/A	N/A
emerging economy	emerging economies	emerging economies
environmental management	N/A	N/A
environmental performance	N/A	N/A
environmental sustainability	N/A	N/A
factor analysis	N/A	N/A
financial performance	N/A	N/A
firm performance	N/A	N/A
food industry	N/A	N/A
food supply chain	N/A	N/A
fuzzy ahp	N/A	N/A
fuzzy delphi method	N/A	N/A
fuzzy set theory	N/A	N/A
global supply chains	N/A	N/A
goal programming	N/A	N/A
governance	N/A	N/A
governance mechanism	N/A	N/A
green logistics	N/A	N/A
green practices	N/A	N/A
green supply chain management	N/A	N/A
green supply chain management (gscm)	green supply chain management	green supply chain management
grey theory	N/A	N/A
knowledge management	N/A	N/A
literature review	N/A	N/A
manufacturing industry	N/A	N/A
multi-criteria decision-making	N/A	N/A
meta-analysis	N/A	N/A
mcdm	multi-criteria decision-making	multi-criteria decision-making
multi-tier supply chain	N/A	N/A
multi-tier supply chains	multi-tier supply chain	multi-tier supply chain
multi-objective optimization	N/A	N/A
new zealand	N/A	N/A
operations management	N/A	N/A
organizational theories	N/A	N/A
performance measurement	N/A	N/A
recycling	N/A	N/A
remanufacturing	N/A	N/A
resilience	N/A	N/A
resource efficiency	N/A	N/A
reverse logistics	N/A	N/A
risk management	N/A	N/A
scale development	N/A	N/A
social performance	N/A	N/A
social responsibility	N/A	N/A
social sustainability	N/A	N/A
stakeholder engagement	N/A	N/A
stakeholder theory	N/A	N/A
structural equation modeling	N/A	N/A
structural equation modelling (sem)	structural equation modeling	structural equation modeling
supply chain	N/A	N/A
supply chain design	N/A	N/A
supply chain dynamic capabilities	N/A	N/A
supply chain flexibility	N/A	N/A
supply chain integration	N/A	N/A
supply chain management	N/A	N/A
supply chain management (scm)	supply chain management	supply chain management
supply chain network design	N/A	N/A
supply chain performance	N/A	N/A
supply chain planning	N/A	N/A
supply chain resilience	N/A	N/A
supply chain sustainability	N/A	N/A
supply management	N/A	N/A
sustainability	N/A	N/A
sustainable development	N/A	N/A
sustainable operations	N/A	N/A
sustainable performance	N/A	N/A
sustainable supplier selection	N/A	N/A
sustainable supply chain	N/A	N/A
sustainable supply chain management	N/A	N/A
sustainable supply chain management (sscm)	sustainable supply chain management	sustainable supply chain management
sustainable supply chains	sustainable supply chain	sustainable supply chain
systematic literature review	N/A	N/A
tea supply chain	N/A	N/A
textile industry	N/A	N/A
traceability	N/A	N/A
transportation	N/A	N/A
triple bottom line	N/A	N/A
uncertainty	N/A	N/A
waste management	N/A	N/A
ındustry 4.0	industry 4.0	industry 4.0
ınterpretive structural modeling	interpretive structural modeling	interpretive structural modeling

. The limited number of retained keywords, together with thesaurus-based cleaning, was considered adequate to ensure terminological clarity at this stage of the analysis. Specifically, of the 614 cleaned keywords, 110 met the threshold. In addition, custom stopwords were applied to remove generic and low-informative terms that do not contribute to meaningful thematic interpretation, such as “adoption,” “framework,” “practices,” “performance,” and “drivers.” However, methodologically oriented keywords such as “literature review” and “meta-analysis” were deliberately retained, because they reflect the methodological diversity of SSCM research and contribute to a more complete representation of the field’s intellectual structure. The results of this analysis are demonstrated in both network and density visualizations in Figure 8a,b.

Figure 8a presents the network visualization generated by VOSviewer and illustrates the structural relationships among keywords. Node size represents keyword frequency, whereas the distance between nodes reflects similarity based on co-occurrence patterns. The map reveals a highly interconnected conceptual structure. Central terms such as “sustainability,” “SSCM,” and “SCM” occupy prominent positions because of their strong connections across multiple thematic areas. By contrast, peripheral and intermediate nodes indicate emerging and co-evolving themes. This pattern reflects the multidimensional nature of SSCM research and points to several coherent thematic clusters discussed in the following subsection. Figure 8b provides the corresponding density visualization and offers a complementary perspective on the distribution of research intensity. High-density regions, depicted in yellow, represent frequently occurring and influential keywords, whereas green areas indicate transitional or developing themes. Lower-density regions, shown in blue to purple, correspond to less-explored or niche topics that contribute to the diversification of the field. Overall, the density distribution highlights a balanced structure between well-established core concepts and emerging research directions, underscoring the dynamic and evolving nature of SSCM scholarship.

Title–Abstract Term Co-Occurrence Analysis

A title–abstract co-occurrence analysis was performed using VOSviewer to identify broader thematic and contextual patterns in SSCM research. The analysis covered 212 articles with title and abstract data exported in RIS format from Mendeley. Titles and abstracts were used as the text source, while structured abstract labels and copyright statements were excluded. The full counting method was applied, with co-occurrence specified as the analysis type and terms as the unit of analysis. Initially, the dataset comprised 4949 terms. A minimum occurrence threshold of five was then applied to remove low-frequency terms, resulting in 418 eligible terms. Before network construction, a custom thesaurus was used to merge terminological variants, including abbreviations and their expanded forms. A stopword file was also applied to remove generic terms that did not contribute to thematic interpretation. The outputs of the thesaurus and stopword files are provided in Appendix A,

Table A3. VOSviewer Text-Based Co-Occurrence Analysis Thesaurus Terms and Stopwords.

Label	Replace by	Thesaurus Outputs	Stopwords
agriculture supply chain			academician
ahp	analytical hierarchy process	analytical hierarchy process	actor
analytical hierarchy process			adoption
automotive industry			association
bangladesh			attempt
bda	big data analytics	big data analytics	attribute
best worst method			barrier
bibliometric analysis			base
big data			basis
big data analytic	big data analytics	big data analytics	capability
big data analytics			category
biofuel			classification
blockchain			community
blockchain technology	blockchain	blockchain	competitiveness
bwm	best worst method	best worst method	consumer
carbon emission			coordination
causal relationship			corruption
ce practice	circular economy practice	circular economy practice	cost
china			current study
climate change			decision maker
competitive advantage			decision making trial
competitive priority			definition
conceptual model			demand
corporate social responsibility			determinant
COVID	COVID-19	COVID-19	effect
critical success factor			effectiveness
csf	critical success factor	critical success factor	efficiency
csfs	critical success factors	critical success factors	end
csr	corporate social responsibility	corporate social responsibility	era
customer pressure			ethic
dcs	distribution control system	distribution control system	evolution
dematel			evolution laboratory
dependence power			extant literature
digital technology			expert
digitalisation			field
disruption			flexibility
dynamic capability			flow
economic performance			further research direction
electric vehicle battery			future research
electronics industry			hand
empirical investigation			help
empirical study			hierarchical structure
enterprise performance			hypothesis
environmental dimension			identification
environmental performance			important role
financial performance			improvement
firm performance			impure biofuel
food industry			influence
global supply chain			initiative
global warming			interest
governance			interrelationship
governance mechanism			journal
government			key challenge
government policy			key role
gpla	green procurement and logistics acceptance	green procurement and logistics acceptance	key supply chain strategy
green supply chain			lack
green supply chain management			life
green supply chain management practice			linear
green warehousing			link
gscm	green supply chain management	green supply chain management	list
gscm practice	green supply chain management practice	green supply chain management practice	managerial implication
knowledge management			mean
logistics optimization			mediating role
manufacturer			methodology
manufacturing company	manufacturing firm	manufacturing firm	network
manufacturing firm			observation
mathematical model			opr
mcdm method	multi criteria decision-making method	multi criteria decision-making method	optimization
meta analysis	meta-analysis	meta-analysis	policy
moderating effect			positive effect
multi criteria decision	multi-criteria decision	multi-criteria decision	positive impact
multi tier supply chain	multi-tier supply chain	multi-tier supply chain	post
oasc	organic agriculture supply chain	organic agriculture supply chain	power
operational performance			present research
organizational learning			present study
organizational sustainability			pressure
pandemic	COVID-19	COVID-19	prioritization
paper industry			production
resilience			product
retailer			profit
rubber product			publication
sc dynamic capability	supply chain dynamic capability	supply chain dynamic capability	quality
scc	supply chain capability	supply chain capability	question
sdgs	sustainable development goals	sustainable development goals	relational capital
sem	structural equation modeling	structural equation modeling	relative importance
sensitivity analysis			research field
shipper			research gap
slr	systematic literature review	systematic literature review	research limitations implication
social issue			response
social performance			responsiveness
social sustainability			return
social value			review
sscm adoption	sustainable supply chain management adoption	sustainable supply chain management adoption	scenario
sscm implementation	sustainable supply chain management implementation	sustainable supply chain management implementation	scope
sscm literature	sustainable supply chain management literature	sustainable supply chain management literature	scı
sscm practice	sustainable supply chain management practice	sustainable supply chain management practice	selection
sscm research	sustainable supply chain management research	sustainable supply chain management research	service
sscp	sustainable supply chain practice	sustainable supply chain practice	set
stakeholder management			significance
stakeholder theory			solution
structural equation modeling			solution method
supplier			state
supply chain manager			structure
supply chain network			sub criterium
supply chain performance			successful implementation
supply chain sustainability			supply
sustainability initiative			tension
sustainability issue			theme
sustainability practice			thing
sustainability risk			tier
sustainability risk factor			total cost
sustainable performance			transition
sustainable practice			trend
sustainable solution			trade off
sustainable supply chain flexibility			triad
sustainable supply chain management practice			value chain
sustainable supply chain performance			variety
sustainable supply chain practice			ımpact
synchromodality
systematic review
tbl	triple bottom line	triple bottom line
technology
theoretical framework
top management commitment
traceability
transportation
tscm	total supply chain management	total supply chain management
waste
wcsscm	world class sustainable supply chain management	world class sustainable supply chain management
ıks	integrated kanban system	integrated kanban system
ındia	india	india
ındian automobile industry	indian automobile industry	indian automobile industry
ındustry	industry	industry
ınternet	internet	internet
ıot	iot	iot
ısm	interpretive structural modeling	interpretive structural modeling

. The limited number of retained terms, together with thesaurus-based cleaning, was considered adequate to ensure terminological clarity at this stage of the analysis. Specifically, of the 4925 cleaned terms, 401 met the threshold. Following VOSviewer’s default recommendation, 241 of the most relevant terms, representing 60% of the cleaned term set, were retained for building the co-occurrence network and the corresponding density visualization presented in Figure 9a,b.

The network visualization in Figure 9a reveals a thematically rich and operationally grounded structure of SSCM research. Frequently occurring terms, such as barrier, implementation, technology, supplier, and SC sustainability, indicate a strong emphasis on the practical challenges and enabling factors of sustainability adoption. The proximity of terms related to collaboration, partnership, and supplier engagement highlights the critical role of inter-organizational coordination. In parallel, the presence of digital technology-related terms, such as blockchain, big data, traceability, and digitalization, reflects a growing research focus on technology-driven solutions for improving transparency, resilience, and decision-making. The co-occurrence of environmental and performance-related terms further suggests that sustainability is increasingly evaluated through multidimensional performance outcomes. In addition, methodological constructs, including Multi-Criteria Decision-Making (MCDM), Structural Equation Modeling (SEM), and dynamic capabilities, point to a diverse and evolving methodological landscape. The density visualization in Figure 9b provides further insight into the maturity and distribution of research themes. High-density areas correspond to well-established topics such as implementation challenges, GSCM practices, and performance outcomes. Medium-density regions reflect themes that have gained prominence in response to recent disruptions, particularly those related to technology and SC coordination. In contrast, low-density areas highlight underexplored topics, including uncertainty, ethics, and organizational learning, indicating potential avenues for future research. The dispersed distribution of sector-specific terms suggests that industry-focused SSCM research remains fragmented rather than consolidated around dominant themes.

Overall, the title–abstract analysis complements the keyword-based findings by offering a more context-sensitive and practice-oriented perspective. While the keyword analysis captures the conceptual core of SSCM, the title–abstract analysis shows how sustainability challenges are operationalized across different contexts. In particular, it highlights implementation dynamics, technological transformation, and emerging research gaps.

Temporal Overlay Analysis

The thematic evolution of SSCM research was examined using the temporal overlay and co-occurrence structures generated through VOSviewer. The overlay visualizations for both author keywords and title–abstract terms represent the average publication year of co-occurring terms rather than specific publication years. VOSviewer calculates the mean occurrence year for each term without normalization, producing a color gradient that reflects the temporal concentration of related concepts [257,263]. Although the dataset covers the period from 2011 to 2023, the overlay maps predominantly display values between 2012 and 2020. This occurs because terms outside this range tend to appear less frequently. Such a pattern is commonly observed in bibliometric analyses and results from the aggregation of temporal information at the term level. Therefore, the color distribution in Figure 10 should be interpreted as an indicator of relative thematic evolution rather than exact temporal boundaries.

The theoretical interpretations presented in this section are not directly derived from the bibliometric algorithm. Instead, they are used to contextualize the observed thematic patterns within established theoretical frameworks. Accordingly, the thematic structure is interpreted through three temporally evolving phases. These phases were identified based on the relative positioning and color distribution of co-occurring terms in the overlay visualization. The phases should be understood as researcher-constructed theoretical readings of the observed keyword co-occurrence patterns rather than algorithmically defined groupings. To avoid post hoc theoretical labeling, the discussion focuses on the mechanisms through which the observed terms relate to SSCM theory. Foundational sustainability terms are interpreted through the Resource-Based View (RBV) and Natural Resource-Based View (NRBV). Operational terms, such as GSCM, reverse logistics, circular economy, and performance, reflect the operationalization of sustainability capabilities. More recent digital and uncertainty-related terms are interpreted through Dynamic Capabilities Theory (DCT) and Information Processing Theory (IPT)/Organizational Information Processing Theory (OIPT). These terms indicate information-processing capacity, resilience, and dynamic adaptation in multi-tier SCs. These theoretical mechanisms, together with the related tensions and boundary conditions, are discussed in detail in the following subsections. These subsections present the temporal thematic patterns and their theory-informed interpretation.

3.2.4. Glossary of Theoretical Background

• Resource-Based View (RBV): RBV explains how firms achieve competitive advantage through resources and capabilities that are valuable, rare, inimitable, and non-substitutable [264]. In the context of SSCM, RBV highlights how sustainability-related capabilities, such as green innovation, supplier collaboration, environmental management systems, and sustainability knowledge, can enhance firm performance and support competitive advantage [265,266].
• Natural Resource-Based View (NRBV): NRBV extends RBV by explicitly linking competitive advantage to a firm’s ability to address environmental challenges. It emphasizes capabilities related to pollution prevention, product stewardship, and sustainable development [267,268,269]. In this study, NRBV is used to interpret themes such as environmental strategy, green practices, circular economy, and sustainability-oriented value creation [270].
• Dynamic Capabilities Theory (DCT): DCT focuses on how organizations adjust, renew, and reconfigure their resources and capabilities in response to changing environmental conditions. Within SSCM, DCT explains how firms respond to sustainability pressures, regulatory shifts, technological advancements, and evolving stakeholder expectations. This theory is particularly relevant for analyzing themes of resilience, adaptation, circular transformation, and strategic change [271,272,273,274].
• Information Processing Theory (IPT): IPT suggests that organizations must enhance their information-processing capabilities to address uncertainty, complexity, and interdependence [275]. In SSCM, firms face significant information-processing challenges because sustainability performance depends on data from multiple suppliers, products, processes, and stakeholders. Therefore, IPT is useful for analyzing information visibility, monitoring, decision-making, and uncertainty reduction.
• Organizational Information Processing Theory (OIPT): OIPT extends information-processing logic to the organizational level. It explains how firms design structures, systems, and technologies to process information effectively. In SSCM research, OIPT helps explain the role of digital technologies, such as blockchain, big data analytics, AI, ML, and IoT, in improving transparency, traceability, coordination, and compliance across complex and multi-tier SCs [276,277,278,279,280].
• Triple Bottom Line (TBL): The TBL framework conceptualizes sustainability through three interconnected dimensions: economic, environmental, and social performance. In supply chains, the economic dimension includes cost efficiency, profitability, productivity, and competitiveness. The environmental dimension covers emissions, energy consumption, waste management, resource efficiency, and pollution mitigation. The social dimension focuses on labor conditions, stakeholder welfare, ethical sourcing, safety, and social responsibility. TBL therefore serves as a foundational sustainability rationale for much of the SSCM literature [141,235].

3.2.5. Theoretical Interpretation of Temporal Thematic Patterns

Theoretic Interpretive Phase 1: Foundational Sustainability Concepts (Earlier Themes)

Earlier themes, represented by darker blue-green tones and corresponding approximately to 2011–2015, reflect the conceptual foundation of SSCM research. Frequently occurring terms such as sustainability, SSCM, sustainable development, and triple bottom line indicate a period of conceptual consolidation. During this period, the core dimensions of sustainability were defined and integrated into SC contexts. These patterns can be interpreted through the lens of RBV and NRBV. These theories conceptualize sustainability as a strategic capability linked to competitive advantage and environmental responsibility [264,265]. In this phase, research primarily emphasized the development of conceptual frameworks and the articulation of sustainability principles. These efforts formed the basis for subsequent operational and empirical studies [266,267,268].

The use of broad foundational terms indicates that early SSCM research focused on establishing the theoretical and managerial legitimacy of sustainability within the field. From the perspectives of RBV and NRBV, sustainability-related practices can be understood as firm-specific capabilities. These capabilities allow organizations to reduce environmental impacts, enhance operational performance, and meet stakeholder expectations. Therefore, the earlier themes in this research do not simply serve as general labels for sustainability. Instead, they show how the field originally framed sustainability as a value-creating capability for organizations. This conceptual foundation later facilitated the emergence of more operational, technology-driven, and sector-specific research streams.

Theoretic Interpretative Phase 2: Operationalization of Sustainability Capabilities (Mid-Phase Themes)

Mid-phase themes, represented by green tones and corresponding approximately to 2015–2019, indicate a transition toward the operationalization of sustainability within SCs. Terms such as GSCM, reverse logistics, circular economy, environmental management, and SC performance became increasingly prominent during this period. This shift suggests a growing emphasis on translating sustainability concepts into measurable practices and performance outcomes [269,270]. These developments are consistent with RBV- and NRBV-informed perspectives, in which firms use operational capabilities and resource management strategies to enhance environmental and economic performance.

In this context, GSCM integrates environmental criteria into procurement, production, logistics, and supplier management. Reverse logistics and closed-loop supply chains support resource recovery, waste reduction, and product returns. Terms associated with the circular economy indicate a transition from linear resource use to reuse, recycling, and renewal. These concepts align closely with the NRBV mechanisms of pollution prevention, product stewardship, and sustainable development. They illustrate how firms convert environmental responsibility into tangible supply chain capabilities. This evolution from foundational sustainability to practice-oriented and performance-driven capabilities signifies a shift from sustainability as a strategic orientation to sustainability as an operational capability. Moreover, sustainability capabilities gain strategic value when embedded in measurable processes linked to supply chain performance outcomes. This provides stronger empirical support for RBV and NRBV by showing that resource optimization and environmental management can serve as sources of competitive advantage within SSCM.

Theoretic Interpretative Phase 3: Digital, Resilient, and Information-Intensive SSCM (Recent Themes)

More recent themes, represented by lighter green–yellow tones and corresponding approximately to 2019–2023, reflect the emergence of digitally enabled and resilience-oriented perspectives in SSCM research. Terms such as technology, blockchain, big data, traceability, capability, flexibility, and uncertainty appear more frequently in the most recent layers of the overlay visualization. These developments can be interpreted in relation to DCT, which explains how firms sense, integrate, and reconfigure resources to adapt to turbulence and support sustainability-oriented innovation [271,272,273]. Socially and institutionally oriented concepts, such as stakeholder theory and collaboration, have also intensified in recent years. This indicates an expanding interest in governance and relational sustainability [274]. It also suggests that digital and dynamic capabilities in SSCM are not purely internal firm-level capabilities. Rather, they depend on inter-organizational coordination, stakeholder alignment, and collaborative governance.

The rising prominence of uncertainty-intensive and information-heavy concepts also corresponds with IPT. This theory argues that organizations facing environmental complexity must enhance their information-processing capacity to maintain performance [275]. In SC contexts, this view is operationalized through OIPT, which emphasizes the need for advanced technologies to reduce information gaps, improve visibility, and manage disruptions. These technologies include big data analytics, blockchain, and traceability systems [276,277,278,279]. Technologies such as blockchain, traceability systems, big data analytics, AI, and digital platforms do not merely represent the adoption of new tools. They also function as mechanisms through which firms collect, verify, share, and process sustainability-related information across multi-tier supply chains [280]. Blockchain and traceability systems enhance transparency by reducing information asymmetry regarding product origin, supplier practices, and sustainability compliance. Big data analytics and AI support the processing of diverse information on environmental performance, supplier risk, and disruption signals. The VOSviewer network structure further supports this theoretical interpretation. Digital information-processing terms, such as blockchain, big data, and traceability, appear in technology-oriented regions. In contrast, relational and institutional terms, such as collaboration and stakeholder engagement, are positioned in adjacent or different thematic areas, together with supplier participation and trust-related constructs discussed in the related literature. This pattern suggests that the effectiveness of digital mechanisms under OIPT depends not only on technological capability but also on relational governance, supplier participation, and institutional trust. Similarly, the visual separation between risk-oriented terms, such as risk, fuzzy methods, DEMATEL, and decision-making, and adaptability-oriented terms, such as flexibility, adaptability, and capability, reflects a control-adaptability gap in SSCM research. This supports the interpretation that structured risk routines may enhance robustness. However, they may also conflict with the agility required by DCT under turbulent market, regulatory, or technological conditions. Consequently, the effectiveness of digital and information-processing mechanisms is contingent on data quality, digital infrastructure maturity, supplier participation, inter-organizational trust, regulatory pressure, and the complexity of multi-tier supply networks.

These three thematic clusters demonstrate a progression from foundational conceptualization to the operationalization of sustainability practices. This progression culminates in a focus on digitally enabled, information-intensive, and resilience-oriented SSCM. It also underscores the growing complexity and maturity of SSCM research. In this evolving field, sustainability is increasingly viewed as a dynamic, information-processing, and capability-driven process shaped by technological, organizational, relational, and environmental factors.

3.3. Analysis of Recent Studies Using BERTopic

To establish a stable bibliometric science mapping structure, citation thresholds were initially applied to identify highly cited SSCM studies. However, publication activity increased sharply after 2020 and peaked in 2021 (see Figure 5). As a result, many recent high-quality studies were excluded because of citation lag. To address this limitation, these studies were analyzed separately using a transformer-based topic modeling approach based on titles and abstracts.

BERTopic was selected because transformer-based topic modeling requires richer linguistic context than author keywords can provide. By contrast, VOSviewer is more suitable for co-occurrence-based keyword mapping [281]. Accordingly, author keywords were analyzed using VOSviewer to construct a citation-based conceptual structure. BERTopic was then employed to complement and extend the thematic patterns identified in recent publications, independently of citation counts. This hybrid approach captures both the established structure of the field and emerging research themes, thereby enhancing the study’s analytical depth and relevance.

To address citation-latency effects and capture emerging research themes, a second dataset of 3799 articles was compiled from the Web of Science and Scopus databases without applying citation thresholds. This dataset was analyzed using BERTopic, based on titles and abstracts. The analysis incorporated both document-level metadata, such as authors, journals, publication years, and citation counts, and textual content, including titles, abstracts, and keywords. The DOI numbers of these papers are provided in Supplementary Materials, Table S2.

3.3.1. Topic Extraction Process

To explore thematic structures in recent SSCM studies affected by citation lag, BERTopic, a transformer-based topic modeling method, was employed. Unlike traditional frequency-based approaches, such as Latent Dirichlet Allocation (LDA) or Non-negative Matrix Factorization (NMF), BERTopic does not rely on bag-of-words representations. Instead, it uses contextual embeddings from pre-trained BERT models to generate semantically rich topic clusters [282,283]. This capability is particularly suitable for interdisciplinary fields such as SSCM, where conceptual nuances extend beyond simple term co-occurrence.

Before applying BERTopic, a preprocessing pipeline was established to ensure consistency and quality in the textual corpus. Records with missing titles or abstracts were removed. The article title and abstract were then combined into a single text field for each document. The text corpus was processed using a CountVectorizer with English stop words, bigram extraction through ngram_range = (1, 2), and a minimum document frequency threshold of min_df = 0.01. In addition to standard English stop words, domain-generic terms such as “supply,” “chain,” “study,” “studies,” “sustainable,” “sustainability,” and “research” were treated as non-informative expressions. These terms appeared frequently across the corpus but did not help distinguish topics. Sentence embeddings were generated using the all-MiniLM-L6-v2 Sentence Transformer model. This produced high-dimensional vectors that capture contextual meaning and inter-sentence semantic relationships [284]. The model was selected because it offers a practical balance between semantic representation quality and computational efficiency. This made it suitable for a medium-sized collection of short-to-medium-length texts, such as titles and abstracts.

To improve computational efficiency and visualization clarity, Uniform Manifold Approximation and Projection (UMAP) was used for dimensionality reduction. The UMAP parameters were set as follows: min_dist = 0.0, n_components = 6, n_neighbors = 15, and metric = cosine. These settings were selected to preserve local semantic relationships among similar documents while enabling clearer separation among dense thematic clusters. The parameter settings were selected based on a sensitivity-focused interpretability assessment rather than purely metric-based optimization. This assessment considered topic separability, keyword clarity, thematic coherence, and alignment with the VOSviewer-based science mapping results. Specifically, n_neighbors was set to 15, after the values of 10, 15, and 20 were examined. This setting helped maintain local relationships among semantically similar documents while allowing broader thematic structures to emerge. A min_dist of 0.0 was selected to promote clearer separation among dense semantic clusters. The number of components was set to 6 to retain sufficient information from the embedding space while reducing dimensional complexity for subsequent clustering. Alternative min_dist values of 0.3 and 0.5 were also examined to assess the sensitivity of topic separation. The final value of 0.0 was retained because it produced more clearly separated and interpretable thematic clusters. After dimensionality reduction, document embeddings were clustered using HDBSCAN, a density-based algorithm in BERTopic that preserves local semantic structure. HDBSCAN was selected because it can identify density-based topic structures and detect outliers or low-confidence documents without forcing all documents into clusters. In this study, HDBSCAN was configured with min_cluster_size of 15, metric = Euclidean, and cluster_selection_method = eom as recommended for UMAP-transformed embeddings [285]. Alternative min_cluster_size values of 10 and 20 were examined to evaluate topic granularity. The value of 15 was retained because it provided a suitable balance between avoiding excessive fragmentation into very small topics and preserving meaningful thematic clusters.

To further support the parameter selection, a scenario-based robustness check was conducted by comparing alternative BERTopic configurations with the selected model. The evaluation focused on outlier ratio, topic diversity, and largest topic share as complementary indicators of topic quality and interpretability. Outlier ratio indicates the proportion of documents assigned to low-confidence or noise categories. Topic diversity measures the proportion of unique terms among the top representative keywords. Largest topic share assesses whether the solution is dominated by a single topic. The results of this robustness check are presented in

Table 3. Robustness Check of BERTopic Parameter Settings.

Model Settings	Outlier Ratio	Topic Diversity	Largest Topic Share
n_neighbors = 15; min_cluster_size = 10; min_dist = 0.0	0.3393	0.6545	0.3235
n_neighbors = 10; min_cluster_size = 15; min_dist = 0.0	0.2693	0.6727	0.3833
n_neighbors = 15; min_cluster_size = 15; min_dist = 0.0	0.2674	0.6636	0.2429
n_neighbors = 20; min_cluster_size = 15; min_dist = 0.0	0.2885	0.6334	0.3322
n_neighbors = 15; min_cluster_size = 20; min_dist = 0.0	0.2538	0.6545	0.3707
n_neighbors = 15; min_cluster_size = 15; min_dist = 0.3	0.4793	0.7091	0.3660
n_neighbors = 15; min_cluster_size = 15; min_dist = 0.5	0.5609	0.7091	0.2344

.

As shown in Table 3, the selected configuration (n_neighbors = 15, min_cluster_size = 15, min_dist = 0.0) provided a balanced solution. It produced an outlier ratio of 0.2674, a topic diversity score of 0.6636, and a largest topic share of 0.2429. Although higher min_dist values produced higher topic diversity, they also substantially increased the outlier ratio. This indicates a larger proportion of low-confidence topic assignments. Therefore, the selected configuration was retained because it offered a suitable balance among topic coverage, thematic distinctiveness, and topic-size balance.

The final number of topics was not predefined. Instead, it was determined by the data-driven clustering structure generated by HDBSCAN and then assessed through interpretability checks. Because BERTopic results can be sensitive to the parameter settings of UMAP and HDBSCAN, the extracted topics were evaluated together with their representative keywords, document-level coherence, and consistency with the science mapping results obtained from VOSviewer. Documents identified by HDBSCAN as outliers or low-confidence assignments were treated cautiously. They were not used as the primary basis for defining the substantive meaning of the topics. Thus, the BERTopic analysis served as a complementary semantic layer rather than a definitive classification of the SSCM literature.

Topic extraction in BERTopic uses class-based TF-IDF (c–TF–IDF). This weighting scheme identifies terms that are both frequent within a specific cluster and distinctive across the larger corpus. It therefore helps ensure that the extracted topics are semantically coherent and contextually discriminative [284,286]. The model then assigns each document a probability distribution over the detected topics. This allows for the assessment of thematic dominance, overlap, and internal coherence. This probabilistic interpretation enhances the interpretability of the results and helps identify subtle emerging themes that may not be captured through citation- or keyword-based analyses alone. The topic labels were not treated as automatically generated theoretical categories. Instead, they were manually interpreted by the authors based on c–TF–IDF representative terms, representative documents, titles, abstracts, and the conceptual coherence of each topic. The structured implementation of BERTopic is illustrated in Figure 11. The pipeline outlines the sequential steps of preprocessing, embedding, dimensionality reduction, clustering, and c–TF–IDF-based topic extraction for identifying emerging themes in recent SSCM studies.

In line with the study’s hybrid methodological design, the BERTopic results were combined with the citation-based conceptual structure generated by VOSviewer to assess cross-method semantic correspondence. While VOSviewer maps established conceptual structures based on keyword co-occurrence, BERTopic complements this framework by identifying emerging and semantically rich research topics that are independent of citation counts. In this study, cross-method semantic correspondence refers to an interpretive comparison of convergence, divergence, and complementarity between VOSviewer-based bibliometric structures and BERTopic-based semantic topics. Therefore, the semantic correspondence score should be interpreted as an indicator of cross-method alignment rather than as a statistical validation metric. This complementary approach strengthens the analysis by capturing both the stable intellectual foundations of SSCM and its evolving thematic frontiers. It therefore increases the depth and relevance of the study.

3.3.2. BERTopic Results

Table 4. Table of Extracted Topics and Their Most Representative Keywords Generated by BERTopic Analysis.

Topic	Top Keywords
0	supply (0.0245) + chain (0.0219) + supply chain (0.0214) + performance (0.0195) + sustainability (0.0183) + study (0.0181) + research (0.0158) + sustainable (0.0155) + green (0.0148) + management (0.014)
1	supply (0.0243) + chain (0.0222) + supply chain (0.0219) + carbon (0.0196) + model (0.0186) + green (0.0169) + manufacturer (0.0151) + cost (0.0124) + products (0.0114) + optimal (0.0113)
2	food (0.0482) + waste (0.0185) + supply (0.0163) + sustainability (0.0145) + chain (0.014) + food waste (0.0132) + study (0.013) + sustainable (0.0126) + environmental (0.0119) + food supply (0.0118)
3	circular (0.0527) + circular economy (0.0336) + economy (0.0332) + ce (0.0317) + waste (0.0187) + study (0.0147) + circularity (0.0142) + research (0.014) + supply (0.0134) + business (0.0127)
4	blockchain (0.0854) + technology (0.0354) + blockchain technology (0.0354) + supply (0.0261) + chain (0.0248) + supply chain (0.0237) + adoption (0.0208) + traceability (0.0166) + study (0.015) + bct (0.0136)
5	risk (0.0289) + fuzzy (0.0271) + criteria (0.0266) + decision (0.0198) + supply (0.0182) + supplier (0.0181) + selection (0.0173) + sustainable (0.0171) + management (0.0164) + method (0.0163)
6	water (0.0279) + china (0.0275) + trade (0.0237) + consumption (0.0209) + energy (0.0207) + carbon (0.0201) + emissions (0.018) + sectors (0.0159) + economic (0.0157) + development (0.015)
7	learning (0.03) + machine (0.026) + machine learning (0.0243) + data (0.0207) + ai (0.0203) + supply (0.0191) + supply chain (0.0171) + chain (0.0167) + model (0.0166) + using (0.0129)
8	fashion (0.0523) + textile (0.0338) + clothing (0.0321) + sustainable (0.0275) + industry (0.0273) + sustainability (0.0224) + circular (0.021) + study (0.0171) + fashion industry (0.0153) + value (0.0148)
9	construction (0.0769) + projects (0.0361) + project (0.0345) + procurement (0.0305) + sustainable (0.0211) + research (0.0209) + management (0.0203) + sustainability (0.02) + construction projects (0.0168) + construction industry (0.016)
10	tourism (0.0454) + consumers (0.0249) + sustainable (0.0229) + green (0.0223) + study (0.0211) + media (0.0183) + social media (0.0181) + social (0.018) + twitter (0.0148) + behavior (0.0142)

summarizes the topics extracted by BERTopic and their most representative keywords. The numerical values in parentheses next to each keyword represent the c–TF–IDF (class-based term frequency-inverse document frequency) scores from the BERTopic model. These scores indicate the importance of each term within a topic by considering both its frequency in the topic cluster and its rarity in the overall corpus. Higher c–TF–IDF values indicate that a term is representative of a given topic and less common in other topics. Therefore, these terms serve as key descriptors of topic semantics rather than as raw frequencies or probabilities.

The findings indicate that SSCM research is structured around several prominent and recurring themes. These include SC sustainability and performance (Topics 0 and 1), circular economy and circularity (Topic 3), blockchain and traceability (Topic 4), and risk-aware decision-making using multi-criteria methods (Topic 5). These topics show high document frequency and well-defined keyword structures, underscoring their significance in the literature. Within this representation, semantically related themes tend to co-occur within the same topic clusters. This reflects shared contextual meaning captured through embedding-based modeling. Topic numbering reflects the clustering structure generated by the BERTopic model and does not imply ranking.

Beyond the core themes, BERTopic reveals several application- and context-specific topics. These include sustainable agri-food SCs and food waste (Topic 2), water–energy–carbon interactions with a focus on China (Topic 6), fashion and textile sustainability (Topic 8), and construction procurement and sustainability management (Topic 9). These topics reflect the application of SSCM concepts in specific industry contexts. Topic 7 highlights the increasing integration of AI, ML, and data-driven modeling. This suggests that intelligent decision-support systems are increasingly positioned as integral components of SSCM research rather than solely as methodological tools. Topic 10 focuses on sustainable consumption, tourism, and social media-driven behavioral analysis. Although relatively small, its distinct emphasis suggests the growing visibility of consumer-centric sustainability research, which remains relatively peripheral to the dominant SSCM research agenda. These context-specific themes are less visible in the keyword-based co-occurrence analysis. This highlights the added value of BERTopic in capturing emerging and application-oriented research directions.

The intertopic distance map in Figure 12 illustrates the semantic relationships among the identified topics. Similar topics are positioned closer together, whereas more distinct topics are located farther apart. The visualization suggests a separation between central topics, such as Topics 0, 1, 3, and 4, and more specialized or peripheral themes, such as Topics 6, 8, and 10. Topics 0 and 1 appear as the most centrally positioned and densely connected clusters, reflecting their central role within the field. In contrast, Topic 10, which focuses on tourism and consumer sustainability, appears relatively isolated. This indicates a more peripheral and emerging research direction.

The hierarchical clustering analysis presented in Figure 13 provides additional insight into the relationships among the identified topics. The dendrogram shows that topics related to SC performance, sustainability management, risk, and digital intelligence cluster at shorter linkage distances. This suggests a relatively strong semantic connection among these themes. In contrast, sector-specific topics, such as fashion, construction, and tourism, merge with the core clusters only at higher linkage distances. This indicates a lower degree of conceptual integration. The hierarchical structure therefore highlights the distinction between central thematic areas in SSCM and more context-dependent application domains. It can also be interpreted as reflecting the field’s expansion from established conceptual foundations toward increasingly specialized and application-oriented research areas.

The document-topic density map shown in Figure 14 reveals the distribution of individual publications across the extracted topics. Dense concentrations around Topics 0 and 1 indicate a large number of studies associated with general SSCM concepts. In contrast, smaller and more localized clusters corresponding to Topics 6, 8, and 10 reflect more specialized research areas. The relatively limited overlap among these clusters further suggests that some emerging themes remain less integrated with the dominant SSCM research agenda.

The BERTopic findings complement the VOSviewer results by showing that contemporary SSCM research includes both central and sector-specific themes. Central themes include sustainability performance, circular economy, blockchain, AI/ML, and risk-oriented decision-making. Sector-specific topics include agri-food, fashion, construction, tourism, and resource interdependence. While the central themes indicate the increasing integration of operational and digital practices, the peripheral topics suggest that application-oriented research remains only partially connected to the dominant SSCM agenda. Accordingly, the following cross-method semantic correspondence analysis examines the alignment between BERTopic-derived themes and VOSviewer-based bibliometric structures. This analysis highlights areas of convergence, extension, and divergence.

3.3.3. Cross-Method Semantic Correspondence Analysis of the VOSviewer and BERTopic Results

To enhance the robustness and interpretability of the results, a cross-method semantic correspondence analysis was conducted between the VOSviewer and BERTopic outputs. In this study, cross-method semantic correspondence refers to the comparative interpretation of findings across different analytical approaches. VOSviewer maps dominant research structures through keyword co-occurrence and citation patterns, whereas BERTopic uncovers latent semantic patterns in titles and abstracts using contextual embeddings. Comparing these complementary approaches enables the identification of thematic consistency, convergence, and divergence within the SSCM literature.

Table 5. Cross-Method Semantic Correspondence Table of The VOSviewer and BERTopic Results.

BERTopic ID(s)	BERTopic Thematic Label	Related BERTopic Keywords	Total c–TF–IDF Weight of Each Topic	Sum of c–TF–IDF Weights for Matched Keywords	BERTopic Weighted Overlap Score	Corresponding VOSviewer Clusters	VOSviewer Representation Score	Semantic Correspondence Score
T4	Blockchain and traceability	Blockchain; traceability; technology; adoption	0.297	0.158	0.530	Blockchain, Industry 4.0, big data analytics	2/2 = 1	0.53
T1	Carbon and green manufacturing	Carbon; green manufacturing; manufacturer; cost; products	0.174	0.075	0.430	Carbon emissions, manufacturing industry, GSCM	2/2 = 1	0.43
T0	SSCM performance and sustainability	Supply chain; performance; sustainability; green practices	0.184	0.059	0.320	SSCM, sustainability, and GSCM	2/2 = 1	0.32
T7	AI and ML in SSCM	AI; ML; data; model	0.204	0.063	0.310	Big data analytics, Industry 4.0, digital technologies	2/2 = 1	0.31
T5	Risk and MCDM methods	Risk; fuzzy methods; criteria; decision-making; supplier	0.206	0.121	0.590	DEMATEL, fuzzy AHP, fuzzy set theory, decision-making	1/2 = 0.5	0.30
T3	Circular economy and circularity	Circular economy; circularity; waste	0.239	0.067	0.280	Circular economy, closed-loop supply chain, reverse logistics	2/2 = 1	0.28
T2	Food waste and agri-supply chains	Food; waste; sustainability; food supply	0.174	0.093	0.530	Food supply chain, carbon emissions, supply chain sustainability	1/2 = 0.5	0.27
T8	Fashion and textile sustainability	Fashion; textile; clothing; circularity	0.264	0.139	0.530	Textile industry, sustainability	1/2 = 0.5	0.27
T9	Sustainable construction	Construction; procurement	0.293	0.107	0.370	Conceptual framework, supply chain design, procurement-related terms	1/2 = 0.5	0.19
T6	Consumption-driven emissions and international trade	China; water; energy; carbon; emissions; trade	0.205	0.138	0.670	Carbon emissions, developing countries, global supply chains, resource-related terms	0/2 = 0	0
T10	Sustainable consumption and tourism	Consumers; tourism; sustainability	0.220	0.093	0.420	Weak/limited representation	0/2 = 0	0

Note: The “Related BERTopic keywords” column was harmonized using short descriptive keyword labels derived from the representative BERTopic terms. c–TF–IDF weights and semantic correspondence scores were rounded to three decimal places for readability.

presents a systematic mapping of BERTopic-derived topics (see Table 4) to the major thematic clusters identified by VOSviewer. Several BERTopic topics show clear semantic alignment with established VOSviewer clusters. This indicates consistency between the two methods despite their different analytical foundations. To improve the transparency of the cross-method semantic correspondence between the BERTopic and VOSviewer results, a normalized c–TF–IDF-based keyword overlap score was calculated for each BERTopic topic. First, the total c–TF–IDF weight for each topic was calculated by summing the c–TF–IDF scores of the representative keywords listed in Table 5. The keywords under each BERTopic thematic label in Table 5 were then compared with the representative keywords and c–TF–IDF scores in Table 4. For each topic, the c–TF–IDF scores of keywords that matched or were semantically related to the corresponding VOSviewer cluster terms were summed. This produced the total c–TF–IDF weight for the matched keywords. To prevent double-counting, phrase-level matches were prioritized. For example, the phrase “supply chain” received its c–TF–IDF weight, while the individual terms “supply” and “chain” were not counted again (e.g., Table 5, Topic 7). If a multi-word phrase was absent from the BERTopic output but its components appeared separately and collectively represented the same VOSviewer concept, the average c–TF–IDF weight of the individual terms was used as a phrase-equivalent overlap weight. Finally, the normalized BERTopic weighted overlap score for each topic was calculated using Equation (1).

$${\mathrm{B}\mathrm{E}\mathrm{R}\mathrm{T}\mathrm{o}\mathrm{p}\mathrm{i}\mathrm{c}\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{O}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{l}\mathrm{a}\mathrm{p}\mathrm{S}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}}_i={\displaystyle \frac{\sum {\mathrm{c}–\mathrm{T}\mathrm{F}–\mathrm{I}\mathrm{D}\mathrm{F}\mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}\mathrm{s}\mathrm{o}\mathrm{f}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{c}\mathrm{h}\mathrm{e}\mathrm{d}\mathrm{k}\mathrm{e}\mathrm{y}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{s}}_{\mathrm{i}}}{\sum {\mathrm{c}–\mathrm{T}\mathrm{F}–\mathrm{I}\mathrm{D}\mathrm{F}\mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}\mathrm{s}\mathrm{o}\mathrm{f}\mathrm{a}\mathrm{l}\mathrm{l}\mathrm{r}\mathrm{e}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{k}\mathrm{e}\mathrm{y}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{s}}_{\mathrm{i}}}}$$

(1)

where $i$ denotes a BERTopic topic. The numerator represents the sum of the c–TF–IDF weights of keywords that either match or are semantically equivalent to terms in the corresponding VOSviewer cluster. The denominator represents the total c–TF–IDF weight of all representative keywords for the same BERTopic topic. The final semantic correspondence score integrates the BERTopic weighted overlap score and the VOSviewer representation score. These two indicators provide complementary insights into semantic alignment. The BERTopic weighted overlap score assesses the degree of overlap between the representative keywords of a BERTopic topic and the corresponding VOSviewer theme, based on the c–TF–IDF weights of matched terms. However, relying solely on keyword overlap may lead to overestimation, particularly when general terms such as “sustainability” or “emissions” span multiple topics. The VOSviewer representation score was assigned using a level-based scale: 2 for dense and clearly identifiable clusters, 1 for themes that were partially represented or visible but less central than the core clusters, and 0 for themes that were peripheral, dispersed, or practically absent in the VOSviewer network and density maps. In this scheme, topics without meaningful VOSviewer representation were assigned a semantic correspondence score of 0. This prevented very small scores from being misinterpreted as substantive semantic correspondence. The final semantic correspondence score was therefore derived by combining the normalized BERTopic weighted overlap score with the normalized VOSviewer representation score, as specified in Equation (2).

$${\mathrm{S}\mathrm{e}\mathrm{m}\mathrm{a}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{C}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{S}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}}_i={BERTopicOverlapScore}_i·{\displaystyle \frac{{VOSviewerRepresentationScore}_i}{2}}$$

(2)

where i denotes a BERTopic topic. Since the BERTopic weighted overlap score ranges from 0 to 1, the VOSviewer representation score was normalized by dividing it by its maximum possible value of 2. This transformation converted the three-level VOSviewer representation scale into a 0–1 scale, making it compatible with the BERTopic overlap score in the final semantic correspondence calculation. Higher semantic correspondence scores indicate stronger alignment between the semantic structure detected by BERTopic and the bibliometric structure observed in VOSviewer.

In particular, the cross-method semantic correspondence results presented in Table 5 indicate that several topics derived from BERTopic closely align with identifiable clusters in VOSviewer. Topic 4, which addresses blockchain and traceability, achieves the highest semantic correspondence score (0.53) and corresponds directly to the related VOSviewer cluster. Topic 1, which focuses on carbon and green manufacturing, aligns with the carbon emissions and manufacturing cluster, with a score of 0.43. Topic 0, centered on SSCM performance and sustainability, corresponds to the SSCM and GSCM cluster. Similarly, Topic 7, which discusses AI and ML in SSCM, aligns with the big data, AI, and ML cluster. These correspondences suggest that the core themes identified by BERTopic are also prominently represented in the VOSviewer network. In addition, Table 5 identifies several topics as thematic extensions of the VOSviewer structure. Topic 5, which examines risk and MCDM methods, has a high BERTopic weighted overlap score (0.59) but a lower semantic correspondence score (0.30). This suggests partial representation in VOSviewer. Topic 3, focused on circular economy and circularity, is similarly interpreted as a thematic extension. Its score of 0.28 indicates a more nuanced representation than that shown in the VOSviewer map. Topics 2, 8, and 9 also represent sectoral extensions in areas such as food waste, fashion sustainability, and sustainable construction. In contrast, Topics 6 and 10 show limited correspondence with VOSviewer, with a semantic correspondence score of 0. This indicates that these topics are better captured by BERTopic’s semantic modeling.

To improve the presentation of the BERTopic results, the 11 extracted topics were organized into five higher-order thematic dimensions based on thematic labels, keywords, and cross-method semantic correspondence patterns (see Table 5). This synthesis illustrates the broader semantic structure of SSCM research and the connections among individual topics within broader research streams. As shown in

Table 6. Higher-Order Thematic Structure of BERTopic Results and Cross-Method Semantic Correspondence Interpretation.

Higher-Order Thematic Dimension	Related BERTopic Topics	Representative Focus	Cross-Method Semantic Correspondence Interpretation
Core SSCM performance and environmental sustainability	T0, T1	SSCM performance, sustainability, GSCM, carbon emissions, green manufacturing	Direct correspondence with established VOSviewer clusters
Digital and data-driven SSCM transformation	T4, T7	Blockchain, traceability, AI, ML, big data, data-driven SSCM	Direct correspondence with digitalization-related VOSviewer clusters
Circularity and resource-oriented sustainability transitions	T3, T6	Circular economy, circularity, resource use, carbon emissions, trade, consumption-driven environmental impacts	Thematic extension of established circularity and emissions-related clusters
Decision-support and risk-based SSCM methods	T5	Risk, fuzzy methods, MCDM, supplier selection, decision-making	Thematic extension of risk and decision-making clusters
Sector-specific sustainability applications	T2, T8, T9, T10	Food waste, agri-supply chains, fashion and textile sustainability, sustainable construction, sustainable consumption, tourism	Sectoral extension with partial or limited VOSviewer correspondence

, the five dimensions are: (1) core SSCM performance and environmental sustainability, (2) digital and data-driven SSCM transformation, (3) circularity and resource-oriented sustainability transitions, (4) decision-support and risk-based SSCM methods, and (5) sector-specific sustainability applications. The interpretation indicates whether each dimension aligns with established VOSviewer clusters or represents a thematic extension of the bibliometric structure.

Overall, the revised cross-method semantic correspondence approach shows that BERTopic and VOSviewer provide complementary perspectives. VOSviewer captures more established and bibliometrically visible structures, whereas BERTopic helps reveal semantically coherent thematic extensions and less consolidated research directions within the SSCM literature.

4. Discussion

4.1. Research Implications

Building on insights from the VOSviewer-based co-occurrence analysis and the BERTopic modeling results, the data-driven conceptual framework presented in Figure 15 provides a structured representation of both established and emerging components of SSCM research. Rather than being purely conceptual, the framework is grounded in empirical patterns derived from the integration of citation-based and embedding-based analyses.

At its foundation, the framework reflects the core components of SSCM, including sustainability integration, SC performance, CE practices, and collaborative governance mechanisms. These themes were identified through VOSviewer and represent well-established research streams that form the conceptual basis of the field. The framework then extends to an operational perspective. It captures how sustainability principles are translated into practice through approaches such as lean management, reverse logistics, environmental management, and risk-based decision-making. It also emphasizes the growing role of digital and data-driven approaches. Technologies such as AI, ML, big data analytics, and blockchain emerge prominently in the BERTopic analysis, indicating a shift toward more data-intensive and technology-enabled SC practices. The dashed structure in this part of the framework reflects the bidirectional and iterative nature of the interactions among these components. This highlights that these relationships evolve dynamically rather than following a strictly linear progression.

Another important aspect of the framework is the inclusion of context-specific and emerging themes that are less visible in citation-based analyses. These include sectoral applications, such as agri-food systems and the textile and construction industries, as well as topics related to resource interdependence and consumer-oriented sustainability. Their presence suggests that SSCM research is increasingly extending beyond traditional boundaries toward more interdisciplinary and application-driven perspectives. In addition to these thematic dimensions, the framework incorporates a methodological perspective. It highlights the role of analytical approaches such as MCDM, SEM, optimization techniques, and ML. These methods act as cross-cutting enablers that support multiple layers of the framework.

The framework further underscores the importance of coordination among policymakers, firms, and supply chain partners. It highlights governance mechanisms and information-sharing practices as critical for implementing sustainability initiatives [15,287,288]. In parallel, the increasing use of digital technologies, such as big data analytics and blockchain, demonstrates how transparency, traceability, and operational efficiency can be improved through data-driven approaches [289,290,291]. At the organizational level, strategies such as financial management and green supplier management contribute to aligning economic and environmental objectives by supporting both cost optimization and emission reduction [292,293]. Overall, the framework provides a coherent and empirically grounded synthesis of SSCM research by linking established knowledge structures with emerging directions. It offers a structured perspective for researchers and supports practitioners and policymakers in integrating sustainability into SC decision-making. In this way, it contributes to improved long-term performance and resilience.

4.2. Key Managerial Implications

The findings of this study suggest that sustainability initiatives in SCs are unlikely to be effective when implemented as isolated actions. Instead, sustainability should be approached as an integrated and multi-layered process. This process connects conceptual priorities with operational practices, technological capabilities, and context-specific applications. The framework presented in Figure 15 and the structured roadmap in

Table 7. Practical Managerial Roadmap for SSCM Implementation.

Managerial Priority	Example Indicators	Supporting Tools	Decisions Supported
Define sustainability baseline	Logistics cost and delivery reliability; carbon emissions and energy use; waste generation and recycling rate; supplier compliance score; GRI/CDP alignment score; CE recovery rate (%)	KPI dashboards; data analytics; GRI/CDP reporting frameworks	Identify sustainability gaps and priority areas; align SC performance with sustainability objectives; embed CE principles into business processes
Improve transparency and traceability	Product origin and certification status; audit results and supplier compliance; material flow visibility; traceability coverage (%); Scope 3 emissions by supplier tier	Blockchain; IoT; traceability platforms	Verify suppliers and ethical sourcing; monitor compliance; document product provenance; certify multi-tier suppliers
Predict risks and disruptions	Supplier risk score; delay probability and disruption frequency; demand variability; inventory shortages; disruption detection lead time	AI; ML; big data analytics	Predict supplier risks; forecast demand; detect disruptions; plan inventory; manage sustainability risks proactively and reactively
Optimize logistics and operations	Transport cost and lead time; emissions per shipment; vehicle utilization and service level; waste reduction rate (%); resource efficiency ratio	AI; ML; optimization models; reverse logistics tools	Select routes and transportation modes; reduce emissions; allocate inventory; align lean management with sustainability targets; design reverse logistics networks
Balance sustainability trade-offs	Cost vs. environmental impact; social responsibility score; resilience and technological readiness; supplier weighted score; investment payback with sustainability criteria	MCDM methods	Select suppliers and technologies; prioritize circular strategies; prioritize investments; document auditable trade-off decisions across cost, quality, and sustainability
Adapt to sector-specific contexts	Sectoral compliance requirements; lifecycle emissions and water use; waste rate and labor compliance; consumer-oriented sustainability index; system-level resource interdependency score	Blockchain; AI/ML; MCDM; sector-specific KPI systems	Tailor SSCM strategies to sectoral needs; extend SC boundaries to include consumer-oriented and system-level sustainability dynamics

provide a basis for translating these dimensions into actionable managerial decisions. These decisions are organized around six managerial priorities.

At the foundational level, managers should recognize that sustainability in supply chains extends beyond environmental considerations. It requires the integration of economic, social, and governance dimensions. This involves aligning sustainability objectives with overall SC performance, strengthening collaboration with partners, and embedding CE principles into core business processes. In practice, this requires defining a measurable sustainability baseline through indicators such as carbon emissions, energy use, waste generation, recycling rates, supplier compliance scores, and CE recovery rates. These indicators can be monitored through Key Performance Indicator (KPI) dashboards and aligned with established reporting frameworks such as the Global Reporting Initiative (GRI) and the Carbon Disclosure Project (CDP). At the operational and capability level, practitioners should focus on implementing practices such as lean management, reverse logistics, environmental management, and risk-based decision-making. These practices enable organizations to translate sustainability goals into measurable outcomes while improving resource efficiency and maintaining operational performance. In this context, structured decision-support tools, such as MCDM methods, can support more balanced and transparent evaluations across competing sustainability criteria. These criteria include cost, environmental impact, social responsibility, and technological readiness. Such tools can also enable documented and auditable trade-off decisions in supplier selection, technology adoption, and investment prioritization.

The increasing importance of digital and data-driven enablers highlights the need for organizations to invest in technologies such as AI, ML, big data analytics, and blockchain. These technologies serve distinct but complementary functions. Blockchain enhances traceability and supplier certification, particularly in high-risk SC segments such as agri-food provenance and textile sourcing. AI and ML enable predictive capabilities, including supplier risk scoring, demand forecasting, and disruption detection. Big data analytics support the real-time monitoring of sustainability KPIs, such as Scope 3 emissions, water usage intensity, and transport emissions per shipment across multiple SC tiers. Scope 3 emissions refer to indirect greenhouse gas emissions occurring across the value chain beyond direct organizational operations.

In addition, managers should consider the role of context-specific and emerging applications. Sectoral differences, such as those observed in agri-food systems, construction, and textile SCs, require tailored sustainability strategies. Similarly, the growing relevance of consumer-oriented sustainability and resource interdependencies suggests that managers need to extend their focus beyond traditional SC boundaries. This requires engagement with broader system-level dynamics. From a strategic perspective, these elements should be integrated through coordinated governance mechanisms, performance measurement systems, and collaborative decision-making processes. This includes aligning internal sustainability initiatives with external regulations, strengthening partnerships with suppliers and stakeholders, and adopting data-driven approaches to support long-term planning and adaptation. From a theoretical standpoint, these managerial implications are consistent with the sustainability-oriented management perspectives discussed in Section 3. The emphasis on capability development and efficiency aligns with RBV, while environmental stewardship and CE practices reflect NRBV. The increasing role of digital technologies and adaptive decision-making corresponds to IPT/OIPT and DCT. This highlights the importance of flexibility and responsiveness in complex and uncertain environments.

4.3. Limitations and Future Research Directions

This study has several limitations despite its contributions. First, the bibliometric science mapping analysis relies on citation-based criteria to construct stable co-occurrence structures. This is a widely accepted approach in bibliometric research [40,41]. However, it may overlook recent but potentially influential studies because of citation latency. Citation counts are influenced by factors beyond conceptual maturity, including publication age, journal visibility, language, topic popularity, disciplinary focus, and regional publishing dynamics. Therefore, the core map of highly cited studies represents a citation-filtered view of influential and stabilized knowledge structures in SSCM. It should not be interpreted as a neutral or exhaustive depiction of the entire field. Although this limitation was partially addressed through the ML-based topic modeling analysis of recent publications from 2021 to 2025, future research could refine this approach by incorporating alternative weighting schemes or dynamic updating mechanisms. This would help balance bibliometric stability better with emerging research trends. Second, the search strategy may not capture all studies that use alternative terminology, such as ESG-oriented or responsibility-oriented supply chain research, without explicitly referring to SSCM or GSCM. However, expanding the search string to include all adjacent terms could substantially broaden the corpus. It could also introduce studies on corporate sustainability, ESG reporting, or responsibility issues that fall outside the core SSCM focus of this review. This limitation is acknowledged, and future reviews may further expand the search strategy to examine these adjacent streams in greater detail. Third, the bibliometric and text-mining techniques employed in this study are based on document-level metadata, such as titles and abstracts, rather than full-text analysis. While this enables systematic and large-scale mapping of SSCM literature, it may limit the capture of deeper contextual insights and theoretical nuances. Future studies could enhance analytical depth by integrating full-text analysis, qualitative content analysis, or mixed-method approaches. Fourth, the scope of the study is limited to peer-reviewed English-language publications indexed in the Web of Science and Scopus databases. Although these sources provide extensive coverage of high-quality research, they may exclude relevant insights from non-English studies and grey literature. Expanding data sources in future reviews could provide a more comprehensive and inclusive understanding of SSCM research.

Beyond these methodological considerations, the findings point to several promising directions for future research. More focused investigation is needed into governance and risk management in supply chains, especially in digitally enabled environments. In these settings, issues such as data privacy, cybersecurity, and algorithmic transparency increasingly influence sustainability outcomes. Adopting a life-cycle and system-wide perspective also represents an important avenue for future work. Examining sustainability impacts across the entire SC, from raw material sourcing to end-of-life management, would support a more comprehensive understanding of trade-offs between operational performance and environmental and social outcomes. Further empirical research in sector-specific contexts, such as agri-food, construction, fashion, and energy-intensive SCs, would help refine and validate the proposed framework under different conditions. Developing context-sensitive risk management strategies and sustainability standards is particularly important in emerging markets characterized by diverse regulatory and operational challenges. In addition, longitudinal and mixed-method research designs could provide valuable insights into how sustainability enablers evolve and interact over time across different decision contexts. Such approaches would enhance both theoretical development and practical relevance.

Finally, future studies could further explore the role of governance mechanisms in shaping sustainable procurement and logistics systems. Advanced analytical approaches, including MCDM, optimization models, and predictive techniques such as PLS-SEM, offer strong potential for examining trade-offs among cost, emissions, and SC resilience. They can also support empirical testing of the relationships among sustainability enablers across different levels of decision-making.

5. Conclusions

This study provides a comprehensive and structured overview of SSCM research by examining its evolution, core themes, and emerging directions. By integrating bibliometric science mapping with transformer-based topic modeling, it captures both established knowledge structures and newer research trajectories shaping contemporary sustainability practices. The findings indicate a clear shift from early conceptual discussions toward more operational, data-driven, and technology-enabled approaches. This shift highlights the growing strategic importance of SSCM. Key thematic areas include sustainability performance, CE practices, governance mechanisms, and digital technologies. The findings also show the increasing prominence of context-specific and interdisciplinary research domains. The convergence of citation-based and embedding-based analyses enhances the robustness of these insights. It also offers a deeper understanding of how foundational and emerging themes are interconnected within the SSCM knowledge structure. Building on these findings, the study develops a data-driven conceptual framework that synthesizes the structural and evolving dimensions of SSCM research. Grounded in empirical evidence derived from the integration of VOSviewer and BERTopic analyses, the framework captures the progression of the field through interconnected layers. It links core conceptual foundations with operational practices, digital enablers, and context-specific applications. In doing so, it illustrates how sustainability objectives can be translated into practical SC strategies in complex and dynamic environments.

The divergence between citation-based clusters and BERTopic-derived themes further clarifies the developmental structure of SSCM research. Citation-based clusters reveal consolidated and central themes such as sustainability performance, GSCM, circular economy, governance, and digital technologies. However, bibliometric centrality does not necessarily imply complete theoretical maturity. For instance, digitalization-related themes are increasingly visible in SSCM research. Yet, their theoretical contribution depends on explaining how technologies such as blockchain, big data analytics, AI, and traceability function as information-processing, coordination, and governance mechanisms in multi-tier SCs. In contrast, BERTopic highlights semantically coherent but less citation-integrated themes, including consumption-driven emissions, sustainable consumption, tourism, and sector-specific sustainability applications. These themes suggest that SSCM research is expanding toward demand-side, interdisciplinary, and context-sensitive directions that have not yet been fully integrated into the core citation structure. Therefore, future research should not only extend empirical applications in these areas but also connect them more explicitly to established SSCM theories. It should also examine how these themes become embedded in supply chain governance, decision-making, and performance systems.

The theoretical contribution of this study is conceptualized as an integrative interpretive framework rather than the establishment of a new theory. It suggests that SSCM research can be explained through the interplay of complementary theoretical mechanisms. From the perspectives of RBV and NRBV, sustainability-oriented resources and environmental capabilities are expected to enhance firms’ capacity to convert SSCM practices into economic, environmental, and social performance outcomes. The co-evolution of these capabilities, including environmental management, governance mechanisms, and CE practices, creates resource complementarities that standalone initiatives cannot achieve. This indicates that cross-domain integration is essential for sustained competitive advantage. From a DCT perspective, the implementation of SSCM practices is most effective when firms develop dynamic capabilities. These capabilities enable firms to recognize sustainability pressures, capitalize on emerging opportunities, and adapt supply chain processes under uncertainty, regulatory demands, and technological change. Conversely, adopting sustainability practices without corresponding adaptive governance mechanisms may lead to diminishing returns, especially in volatile and multi-tier supply chain settings. From IPT and OIPT perspectives, digital technologies such as artificial intelligence, machine learning, big data analytics, blockchain, and the IoT can strengthen SSCM by improving information processing and reducing information asymmetry across supply chain tiers. These mechanisms support transparency, traceability, and inter-organizational coordination in complex networks. The findings suggest that SSCM has not yet converged around a single theoretical core. Rather, it remains a pluralistic and multi-theoretical domain in which RBV/NRBV, DCT, IPT/OIPT, systems thinking, paradox theory, and governance perspectives explain different but complementary aspects of sustainability-oriented SC transformation. Collectively, these theoretical interpretations suggest that SSCM development cannot be fully captured through isolated perspectives. Sustainability outcomes are contingent on the synergistic effects of resources and capabilities, environmental strategies, adaptive capacity, and information-processing mechanisms. Sector-specific contextual factors, including regulatory landscapes, consumer pressures, and resource interdependencies, serve as additional moderating dimensions. These factors influence the extent to which such mechanisms affect sustainability performance, particularly in high-complexity sectors such as agri-food, construction, and textiles.

This framework provides a foundation for future empirical research. Such research can investigate how sustainability-oriented capabilities, dynamic capabilities, and digital information-processing mechanisms collectively influence SSCM implementation and performance across different sectoral contexts. The framework also highlights that the theory–practice contribution of SSCM research is strengthened when sustainability strategies are aligned with measurable indicators, decision-support tools, and context-specific implementation methods. Practically, this framework serves as a structured and adaptable guide for decision-makers. It illustrates how SSCM concepts can inform managerial priorities, performance metrics, digital tools, and strategic decisions tailored to specific sectors. By facilitating more informed and coordinated decision-making, the framework may enhance performance, resilience, and long-term sustainability in global SCs.