A Structured Framework for Circular Supplier Selection: A Hybrid Multi-Criteria Decision-Making Approach

Abstract

Background: Circular supply chains (CSC) have emerged as a strategic response to sustainability challenges, while adoption remains uneven. Supplier selection is a key driver of effectiveness, shaped by organizational capabilities, institutional support, and leadership. This study develops a structured framework for circular supplier selection (CSS) using a hybrid multi-criteria decision-making approach, addressing fragmented research and strengthening the link between methodological innovation and practice. Methods: The proposed framework integrates fuzzy DEMATEL, the Best-Worst Method (BWM), and the Analytic Hierarchy Process (AHP) within MCDM. Fuzzy DEMATEL identifies cause-and-effect relationships among criteria, distinguishing net causes from net effects. The most influential and dependent criteria serve as anchors for the BWM weighting, followed by AHP to evaluate sub-criteria and alternatives. Results: Environmental governance emerged as the most influential driver in the causal analysis, while circular performance received the highest weight in BWM. The final AHP evaluation ranked Alternative 5 as the most suitable, followed by A9 and A3, confirming the framework’s ability to deliver consistent, actionable insights for circular supplier selection. Conclusions: This integration enables a more granular and robust evaluation of supplier strategies within CSC, reinforcing their role in accelerating sustainability transitions. It establishes a structured framework for CSS, highlighting CSS performance and upstream supply chain decision-making.

1. Introduction

Liang et al. [1] proved the increasing recognition of circular supply chains (CSC) as a strategic response to sustainability challenges, although their adoption varies across contexts. Prior studies highlight different drivers: digital dynamic capabilities fostering environmental sustainability, organizational enablers supporting effective CSC implementation [2], and leadership orientations combined with institutional governance shaping green supply chain adoption and circular practices [3]. Together, these perspectives suggest that CSC effectiveness relies on both internal capabilities and external institutional support. These elements shape how firms establish criteria and make strategic decisions in supplier selection within CSC.

Building on this foundation, circular supplier selection (CSS) has become a central concern in advancing CSC, with recent studies proposing diverse methodological approaches. For instance, Nayeri and Sazvar [4] developed stochastic models to integrate the circular economy (CE) and Industry 5.0 dimensions into raw-material-provider evaluation. Similarly, scenario-based stochastic Multi-Criteria Decision-Making (MCDM) frameworks address supplier assessment under uncertainty in project-driven supply chains [5]. Other contributions include multi-phase decision-support systems that combine fuzzy and optimization techniques to enhance supplier ranking and order allocation in manufacturing contexts [6], as well as practical approaches to embed circularity into procurement decisions [7]. Collectively, these studies demonstrate that CSS requires balancing methodological rigor with practical applicability and incorporating sustainability and circularity criteria in supplier evaluation.

While these contributions provide valuable insights, the existing literature remains fragmented. Stochastic MCDM approaches [5] have not yet explored alternative fuzzy models, broader sectoral applications, or long-term supplier performance evaluation. Approaches embedding circularity into procurement [7] do not sufficiently address collaboration mechanisms or sector-specific tool development. Similarly, raw-material-provider evaluation [4] stops short of offering mathematical models for order allocation after supplier selection. These limitations highlight a clear research gap: existing studies tend to emphasize isolated methodological advances without integrating causal analysis, weighting consistency, and hierarchical evaluation into a single structured framework for CSS. Our study addresses this gap by combining Fuzzy Decision-Making Trial and Evaluation Laboratory (Fuzzy DEMATEL), the Best-Worst Method (BWM), and the Analytic Hierarchy Process (AHP) into a structured hybrid approach. This integration advances both methodological rigor and practical applicability in circular supplier selection.

Beyond the scope of this study, further opportunities remain. There is a need to explore sector-specific adaptations of tools, investigate long-term supplier performance under circular criteria, and enhance collaboration mechanisms that integrate circular practices across the entire supply chain [5,7]. Addressing these gaps will strengthen the alignment between methodological innovation and practical implementation. Consequently, this opens the way for new frameworks that advance the role of CSS in accelerating sustainability transitions.

To bridge these gaps, the objective of this study is to develop a structured hybrid MCDM framework for CSS by integrating Fuzzy DEMATEL, BWM, and AHP. This structured framework for CSS enhances both methodological rigor and practical applicability. The structured framework for CSS applies these methods to support the strategic evaluation of CSS, incorporating criteria, sub-criteria, and alternatives for a more granular assessment. Methodologically, the study advances existing decision-support approaches by integrating Fuzzy DEMATEL to identify cause–and–effect relationships among the analyzed criteria, distinguishing those that function as net causes and those that function as net effects. This approach allows us to identify the most influential and the most dependent criteria. These critical dimensions then serve as anchors for the subsequent application of BWM to weigh the criteria, before the embedding of AHP for the evaluation of sub-criteria and alternatives.

This study advances literature through two main contributions. First, it proposes a structured decision-making framework tailored for CSS. Second, it integrates Fuzzy DEMATEL, BWM, and AHP into a transparent and replicable approach for ranking strategic alternatives. This integration establishes a structured framework for CSS that emphasizes performance outcomes and upstream supply chain decision-making, thereby enabling effective governance of CSC.

In addition to the Introduction, this paper has four more sections: Section 2 examines the foundations of CSS; Section 3 outlines the research methodology; Section 4 reports the results of the deployment of the CSS model; Section 5 provides the discussion; and Section 6 concludes the study.

2. Literature Review

2.1. Foundations of CSS

CSS, within the context of sustainability and the CE, has evolved beyond traditional cost-based approaches. It now incorporates broader dimensions such as social responsibility, resilience, and organizational purpose. For instance, Castellani et al. [8] extended the Triple Bottom Line framework to advance the Sustainable Development Goals (SDGs). In contrast, a multidimensional framework integrates sustainable, resilient, and circular criteria to formalize CSS knowledge [9]. These contributions emphasize that supplier selection must align with sustainability transitions rather than remain confined to economic efficiency. They also reveal a tension between normative aspirations and operational rigor, demonstrating that CSS research oscillates between conceptual expansion and methodological consolidation.

Sector-specific applications further illustrate the diversity of approaches within the literature. Particularly, dedicated CSS frameworks tailor evaluation criteria to the cosmetics industry [10] and the fresh fruit and vegetable supply chain [11]. Moreover, optimizing production in the energy sector enhances sustainability while reducing carbon footprints [12]. These examples demonstrate how supplier-related decisions directly influence product performance and environmental outcomes, proving that CSS must adapt to industry-specific challenges. However, this contextualization generates fragmentation: criteria tailored to sectoral needs but lacking comparability across industries, making it difficult to establish generalizable CSS principles.

The integration of resilience and digitalization into CSS frameworks has also gained significant prominence. A robust multidimensional framework aligns CSS with Industry 4.0 principles [13], while data-driven frameworks address supplier selection and order allocation by combining resiliency, CE, and customer-based dimensions [14]. Furthermore, integrating fuzzy logic and AI-based capabilities helps manage uncertainty during supplier scrutiny [15]. Together, these studies highlight the importance of embedding technological innovation and resilience into CSS to ensure adaptability in dynamic environments. Nevertheless, methodological emphases differ; some studies rely on deterministic optimization, while others use fuzzy or AI-based tools to manage uncertainty. This divergence underscores a theoretical tension between robustness and flexibility in CSS design.

Large-scale and complex supply chains present additional challenges for supplier evaluation. To reduce complexity and enhance consistency in weight determination, recent methods consider the critical interactions between circular and non-circular criteria [16]. At the same time, remanufacturing research identifies resource optimization and landfill reduction as key drivers, reinforcing the importance of context-specific sustainability factors [17]. Finally, linking supplier selection to routing efficiency and inventory management demonstrates how to manage waste material collection and processing under uncertainty [18]. These contributions share a concern with complexity management but differ in scope: some advance abstract methodological refinement, while others prioritize concrete operational problem-solving. This divergence shows that CSS must balance methodological rigor with practical applicability in large-scale and uncertain environments.

Research has also explored supplier selection in specialized supply chains, where energy efficiency and transparency are paramount. In construction, integrated frameworks incorporate blockchain-enabled smart contracts to align supplier governance with CE objectives [19]. Similarly, in the semiconductor industry, resilient and sustainable supply chain networks optimize performance across economic, environmental, social, and resilience metrics [20]. In summary, these studies demonstrate how to extend structured frameworks for CSS to complex industrial contexts by integrating advanced optimization techniques and digital technologies. At the same time, they raise questions about scalability and accessibility, particularly for SMEs or less digitized sectors.

Overall, these contributions highlight the evolution of CSS across diverse sectors and contexts and reinforce the growing importance of criteria selection. The identification, weighting, and interaction of sustainability, resilience, and circularity criteria remain central to advancing the structured framework for CSS, setting the stage for the next section.

2.2. CSS Criteria

CSS has emerged as a critical factor for overcoming CE barriers and advancing sustainable practices. Recent studies show that effective adoption requires not only compliance with waste regulations but also collaborative eco-partnerships between buyers and suppliers [21]. Systematic literature mapping underscores the role of regenerative business models and digital technologies, such as blockchain, in enabling zero-waste strategies while strengthening supplier evaluation [22]. These insights highlight the need for robust criteria that integrate economic, environmental, social, and technological dimensions. Consequently,

Table 1. The main criteria in the structured framework for CSS.

Criterion	Description	References
C1 Circular Performance	The capability of the supplier to enable closed-loop systems, resource recovery, product life extension, and circular value creation.	[23,24,25,26]
C2 Environmental Governance	Degree of formalization of environmental management practices, emission control, compliance, and sustainability reporting.	[23,27,28]
C3 Economic & Operational Performance	Traditional dimensions of supplier performance include cost efficiency, quality, delivery reliability, flexibility, and financial stability.	[6,23,29]
C4 Technological Enablement	Technological capability to support circular supply chains through digitalization, traceability, system integration, and Industry 4.0 readiness.	[6,24,29]
C5 Social & Collaborative Sustainability	Supplier commitment to social responsibility, ethical standards, stakeholder collaboration, transparency, and long-term partnership alignment.	[23,24,25,27]

outlines the main evaluation criteria, their definitions, and the foundational literature supporting the structured framework for CSS.

Circular performance reflects a supplier’s ability to enable closed-loop systems, resource recovery, product life extension, and circular value creation. A foundational framework [23] assesses circular maturity using key indicators such as eco-friendly packaging, waste management, reduced virgin material use, and product recovery. Similarly, the framework in [24] incorporates circular and environmental sub-criteria, including green sourcing, pollution control, and environmental management systems, to align smart supply chains with zero-waste objectives. Furthermore, research indicates that energy efficiency, recyclable materials, clean technologies, and reverse logistics often outweigh purely economic considerations [25].

In food supply chains, the strategic role of sustainable packaging design is also critical. Specifically, reverse logistics serves as a core mechanism for conserving resources and reducing food waste [26]. Multi-criteria decision-making allows the assessment of collaborative packaging sustainability to effectively balance ecological impact, food safety, and end-of-life options.

Complementing this, environmental governance captures the formalization of environmental management, emission control, compliance, and sustainability reporting. Evidence indicates that the environmental dimension strongly influences supplier selection, proving that compliance and transparent governance are decisive for supplier credibility [23]. When evaluating these practices, a clear distinction exists between one-time assessments and continuous monitoring [27]. Effective governance requires suppliers to actively reduce consumption, emissions, and pollution through advanced environmental management systems. Accordingly, low-carbon models emphasize carbon governance practices, establishing emission targets and risk assessments as decisive evaluation factors [28].

Economic and operational performance encompasses cost efficiency, quality, delivery reliability, flexibility, and financial stability. Although traditional economic dimensions remain essential within CSS, competitiveness and operational efficiency must align with overarching sustainability goals [23]. Sub-criteria such as cost, quality, timely delivery, productivity, and financial capacity frequently take a secondary role compared to circular priorities [25]. To address this balance, hybrid decision-making models integrate competitive cost, production quality, efficiency, and financial availability [29]. This research highlights that financial readiness and IT infrastructure are essential for supporting supplier performance in circular and Industry 4.0 contexts.

From this perspective, comprehensive metric sets at macro, meso, and micro levels support supplier selection and monitoring [30]. By incorporating managerial psychology and data uncertainty, this approach helps resolve conflicting judgments among decision-makers. Consequently, integrating economic competitiveness and material reuse into evaluations effectively bridges the gap between CE theory and practical management.

Technological enablement further enhances this process by reflecting the supplier’s capacity to support CSC through digitalization, traceability, system integration, and Industry 4.0 readiness. Industry 4.0 criteria—including connectivity and intelligent automation—function as key drivers of efficiency and reduced environmental impact [24]. Digital tools and investment in environmental R&D serve as critical enablers for green supplier selection, balancing financial viability with innovation [29]. For example, integrating sustainability with cyber-physical manufacturing and IT infrastructure optimizes supplier selection in sectors like the beverage industry, enhancing competitiveness while conserving resources [6].

Social and collaborative criteria refer to a supplier’s commitment to social responsibility, ethical standards, stakeholder collaboration, transparency, and long-term partnership alignment. Collaboration and shared awareness strengthen circular maturity through product return incentives, cross-sector cooperation, and targeted training programs [23].

Building on these perspectives, social criteria must integrate deeply with circular and technological dimensions [24]. Key sub-criteria include occupational health, labor rights, information transparency, and the prevention of forced labor [24,25]. Transparency and stakeholder engagement ensure a balanced, resilient supplier evaluation, encouraging partners to extend beyond mere regulatory compliance [27]. Based on these dimensions,

Table 2. Sub-criteria in the structured framework for CSS.

Criterion	Sub-Criterion	Description	References
	SC1 Circular design capability	Design for reuse, modularity, disassembly, and recyclability.	[23,26]
	SC2 Circular innovation capability	Development of innovative circular business and production models.	[24,25]
C1	SC3 Reverse logistics capability	Ability to collect, return, and reintegrate products/materials into closed-loop systems.	[23,24]
	SC4 Recycling & recovery rate	Extent of material recovery, reuse, and remanufacturing practices.	[23,27]
	SC5 Use of secondary materials	Adoption of recycled or regenerated inputs in production.	[25,28]
	SC6 Carbon Management	Monitoring and reduction in greenhouse gas emissions.	[25,28]
	SC7 Environmental Management Systems	Structured environmental management practices and certifications.	[27,28]
C2	SC8 Environmental Transparency	Disclosure of environmental data and sustainability reporting.	[24,29]
	SC9 Regulatory Compliance	Compliance with environmental and CE regulations.	[23]
	SC10 Waste Management Systems	Structured waste minimization, treatment, and safe disposal practices.	[23,27]
	SC11 Cost Efficiency	Total acquisition cost and economic competitiveness.	[23,24]
	SC12 Delivery Reliability	On-time delivery and supply continuity.	[30]
C3	SC13 Financial Stability	Financial robustness and investment capacity.	[25,29]
	SC14 Operational Flexibility	Ability to adapt to demand or specification changes.	[25,30]
	SC15 Quality Reliability	Defect rates, certifications, and historical quality performance.	[23]
	SC16 Analytics for Circular Optimization	Use of digital analytics to improve resource efficiency.	[24]
	SC17 Collaborative Digital Platforms	Digital tools enabling coordination and information sharing.	[6,29]
C4	SC18 Data Integration Capability	Integration of ERP, IoT, and information systems.	[6]
	SC19 Digital Traceability	Digital tracking of materials across the supply chain.	[29]
	SC20 Industry 4.0 Readiness	Adoption of smart manufacturing and cyber-physical systems.	[6,24]
	SC21 Corporate Social Responsibility	Commitment to social responsibility and ethical standards.	[25,27]
	SC22 Ethical & Sustainable Sourcing	Adherence to ethical sourcing and sustainable procurement standards.	[25,28]
C5	SC23 Health & Safety Practices	Occupational health and safety compliance and performance.	[24]
	SC24 Stakeholder Collaboration	Level of cooperation and partnership alignment within the supply chain.	[23,24]
	SC25 Transparency & Trust	Information transparency and long-term trust relationships.	[23]

presents the corresponding sub-criteria alongside concise operational descriptions and key references.

Established supply chain reference models, such as the SCOR Digital Standard (SCOR-DS), also incorporate sustainability dimensions across operational, economic, social, and environmental measures. Recent studies show that SCOR-DS integrates resilience and ESG indicators into performance assessments. This reinforces the relevance of aligning CSS criteria with recognized industry frameworks [31].

Following Saaty’s principles for structuring decision problems in AHP, three requirements guide the selection of criteria (Table 1) and sub-criteria (Table 2): they must be mutually exclusive, conceptually comprehensive, and hierarchically coherent [32,33]. Then, conceptual similarity across environmental, economic, social, technological, and circular dimensions allows for the grouping of the criteria and sub-criteria identified in the literature. Single operational constructs consolidate overlapping elements after semantic comparison. For example, “environmental transparency” consolidates concepts related to disclosure, reporting, and environmental visibility. Similarly, relational aspects associated with stakeholder confidence and long-term cooperation fall under “transparency and trust” within the social dimension.

To refine the structured framework for CSS, sub-criteria were retained only when they satisfied three conditions: (1) recurrent appearance across prior studies on CSS and CE; (2) conceptual distinctiveness from other sub-criteria; and (3) operational relevance for within the structured framework for CSS in CSC contexts. The resulting structure comprises five main criteria and twenty-five sub-criteria, ensuring logical consistency between strategic dimensions and operational indicators. Finally, prior to expert evaluation, the structured framework for CSS underwent conceptual validation to ensure it accurately captures the essential dimensions of CSS without redundancy or omission.

2.3. Decision-Making in CSS

Decision-making in CSS incorporates a wide range of approaches, including MCDM frameworks, hybrid simulations, and risk-oriented frameworks. These methods reflect the growing need to integrate sustainability, technological readiness, and resilience into supplier evaluation.

Integrated MCDM frameworks clearly demonstrate how sector-specific characteristics influence these choices. For example, in the paint industry, frameworks combining sustainability, risk management, and circularity have identified efficiency and safety as dominant criteria [34]. This emphasizes the importance of tailoring decision tools to specific industrial contexts. Similarly, manufacturing research shows that while delivery and quality remain central, environmental and social criteria are gaining prominence. These findings highlight a gradual transition toward sustainability-oriented decisions in energy-intensive industries [35].

Digital technologies have also enhanced modern decision-making frameworks. For instance, surveys and readiness assessments in Small and Medium-sized Enterprises (SMEs) reveal that IoT and cyber-physical systems are critical enablers of CSC. However, organizational orientation heavily influences their adoption [36]. In the textile sector, combining MCDM with expert consultation highlights that technological infrastructure and an innovation-oriented culture are decisive for successful circular transitions [37].

Furthermore, risk-based frameworks refine CSS by integrating performance evaluation with sustainability objectives. These structured frameworks combine risk analysis with circular criteria. This approach strengthens alignment with the SDGs, particularly regarding waste reduction and resource efficiency [38].

Subsequent studies highlight methodological innovations that reconcile economic and environmental goals. In the electronics sector, closed-loop supply chain models using MCDM and industrial symbiosis demonstrate superior performance compared to linear systems. Specifically, eco-industrial parks reduce carbon emissions, while tailored incentive schemes improve reverse logistics flows [39].

Additionally, hybrid simulations within agro-industries show that dynamic supplier selection can simultaneously increase profitability and reduce emissions. This illustrates the dual benefits of integrated decision-making [40]. Finally, strategic modeling in the pharmaceutical industry identifies critical barriers, such as financial constraints and coordination failures, while positioning blockchain and reverse logistics as key facilitators for circular adoption [41].

Taken together, these studies demonstrate that CSS decision-making is evolving from descriptive evaluations toward integrative frameworks. Modern frameworks effectively combine traditional performance, sustainability, technology, and risk considerations. While delivery, quality, and cost remain fundamental, multidimensional frameworks that reflect the complexity of CSC increase their incorporation. The diversity of methods, from MCDM to hybrid simulations, confirms that methodological pluralism is essential. Ultimately, refining these processes requires bridging conventional supplier performance with emerging sustainability imperatives. This integration ensures that selection frameworks are both operationally robust and strategically aligned with CE principles.

2.4. Identified CSS Gap

Although developing decision-making frameworks for CSS has achieved considerable progress, the literature reveals several persistent gaps that limit their applicability and robustness. These limitations highlight the need for more comprehensive frameworks that integrate methodological rigor with practical relevance.

One recurring limitation concerns the applicability of existing frameworks across different firm sizes and economic contexts. Studies emphasize the necessity of expanding current frameworks to Small and Medium-sized Enterprises (SMEs) and developing countries, where sustainability and resource security are critical challenges [34,37]. Furthermore, existing frameworks require validation across diverse industrial sectors, such as the automotive, glass, and paper industries. Scholars also recommend extending these methodologies to other multi-criteria problems beyond supplier selection [42].

Another gap relates to methodological refinement. Several studies call for integrating quantitative data with expert opinions to reduce subjectivity. Additionally, adopting fuzzy and hybrid MCDM approaches can better capture uncertainty and imprecision in complex decision environments [36,38]. Implementing sensitivity analysis and prescriptive analytical methods is also crucial to strengthen result robustness and mitigate potential biases in expert-driven evaluations [36].

Economic and operational trade-offs represent an additional challenge. Research points to the need for models that account for dynamic demand, investment costs, and the long-term benefits of circular facilities. Concurrently, frameworks should incorporate penalties and incentives for environmentally responsible practices. Robust approaches to managing uncertainties in cost, recovery, and recycling rates are particularly vital [39].

Sector-specific studies further highlight the importance of adapting criteria to unique resource contexts. For instance, agro-industrial applications require models that evaluate the suitability of agricultural residues while adjusting weights and criteria to specific resource typologies [38]. More broadly, scholars stress the need to develop standardized metrics and multi-level models for circular procurement, thereby contributing to a stronger scientific consensus on CE principles [41].

Taken together, these gaps reveal the necessity of a structured framework for CSS. Integrating approaches such as Fuzzy DEMATEL, BWM, and AHP can effectively address uncertainty, balance trade-offs, and ensure applicability across industries. By combining methodological rigor with practical adaptability, such structured frameworks for CSS can overcome current limitations, providing transparent and replicable tools for ranking strategic CSS alternatives. Rather than claiming to fully resolve these gaps, this study positions its proposed structured framework for CSS as a conceptual contribution that advances theoretical development. It offers a structured basis for practical implementation, while the empirical validation of its applicability, particularly in SMEs and across diverse sectors, remains an important direction for future research.

3. Methodology

3.1. Research Design

This section outlines the methodological foundation of the structured framework for CSS. This study adopts an integrated MCDM approach to evaluate suppliers under multiple, interrelated, and potentially conflicting criteria. The proposed framework combines Fuzzy DEMATEL, BWM, and AHP in a sequential, complementary structure.

Conceptual complementarity within CSC supports the integration of these methods. Fuzzy DEMATEL enables the identification of causal relationships among criteria under conditions of uncertainty [43,44]. BWM, in turn, ensures consistent weights with reduced cognitive burden, overcoming the limitations of traditional pairwise comparison methods [45]. Compared to ranking-oriented methods such as TOPSIS, BWM is particularly advantageous because it derives reliable and consistent weights under uncertainty. While TOPSIS remains effective for ranking alternatives, it depends on predefined weights; conversely, BWM provides a structured mechanism to obtain them with greater consistency [46]. Finally, AHP structures the hierarchies of criteria and alternatives, a feature widely applied in the structured framework for CSS [47].

Unlike other hybrid frameworks, such as fuzzy AHP–VIKOR [48], which primarily emphasize alternative classification, the proposed integration combines causality, consistency, and hierarchy. This combination offers a more robust and transparent solution for circular supplier selection. Such a structured framework for CSS is particularly suited to CSC, where interdependencies, uncertainty, and the need for transparent evaluation demand a structured framework for CSS that goes beyond isolated techniques.

Prior multi-criteria decision-making literature supports the selection of these methods. Fuzzy DEMATEL effectively captures causal interdependencies among evaluation criteria [44,49]. Meanwhile, BWM generates reliable weight while maintaining high consistency in expert judgments [50]. Lastly, AHP remains a robust tool for structuring hierarchical decision problems and aggregating local priorities into global rankings [39]. Fernández [51], Tramarico [52], and Tramarico et al. [53] recognize this specific sequential integration as effective in addressing complex, multidimensional evaluation contexts.

To visualize the process, the research design has five analytical layers, as illustrated in Figure 1. This structure enables a smooth transition from criteria definition to final CSS ranking. Ultimately, integrating these methods simultaneously captures causal interdependencies, ensures weighting consistency, and supports the hierarchical evaluation of alternatives.

The integration of these methods follows logical progression. The causal insights obtained from Fuzzy DEMATEL guide the identification of the most influential and dependent criteria, which then serve as anchors for the weighting process in BWM. This ensures the weight assignment is based on concepts rather than being substantively arbitrary. Subsequently, the AHP method processes the resulting weights, enabling the evaluation of sub-criteria and supplier alternatives with consistent priorities. This sequential order derives formally from the methodological complementarities among the techniques. Specifically, Fuzzy DEMATEL must precede BWM to provide causal anchors for the Best and Worst criteria, while BWM must precede AHP to ensure that the hierarchical evaluation is based on consistent, validated weights. In this way, each stage builds directly on the outcomes of the previous one, creating a coherent flow that reflects the complexity of CSC.

As illustrated in Figure 1, the input layer represents the first step of the proposed structured framework for CSS, involving the identification and structuring of the evaluation criteria and sub-criteria for CSS. The comprehensive literature review presented in Section 2.2 established the relevant criteria (C1–C5) and their associated sub-criteria (SC1–SC25), as presented in Table 1 and Table 2. This provides the conceptual foundation for the subsequent analytical procedures.

In the second layer, causal analysis examines the interrelationships among the main criteria. There, the Fuzzy DEMATEL method results in the total influence matrix and computes the prominence (R + C) and relation (R − C) scores. These results enable the classification of criteria into cause-and-effect groups.

In the third layer, criteria weighting establishes the relative importance of the main criteria using the BWM application. Unlike conventional implementations that select the best and worst criteria purely based on subjective judgment, this study utilizes the prior causal analysis to guide their identification. Specifically, the criterion exhibiting the strongest net causal influence is the Best criterion, while the most influenced criterion is Worst. Based on these reference points, the BWM model determines the final weights by minimizing the maximum absolute deviation, thereby ensuring high consistency in expert judgments.

The fourth layer operationalizes the assessment of supplier alternatives. At this stage, AHP is employed to structure and evaluate the decision problem. A decision hierarchy (Figure 2) links the overall objective, the previously weighted criteria, the respective sub-criteria (Table 2), and the supplier alternatives (A1–A10). It is important to note that AHP does not determine the weights of the main criteria. Instead, the AHP process uses the weights of criteria obtained from BWM. Then, AHP results in the local priorities of sub-criteria and alternatives.

Pairwise comparisons at the sub-criteria level result in the local priorities of the alternatives. Overall priorities for alternatives were derived from the sum of local priorities weighted by criteria.

In the final layer, the output consolidates these evaluations to present the definitive supplier ranking. Radar charts visualize performance across criteria, supporting sensitivity analysis and highlighting trade-offs among alternatives. By integrating causal analysis, consistent criteria weighting, and hierarchical evaluation, the proposed framework functions as a robust, structured framework for CSS, enabling transparent and informed supplier selection. This integration ensures that the framework goes beyond a simple evaluation, operating as a structured tool to guide managerial decision-making in CSC.

To contextualize a structured framework for CSS application, the study incorporates a case analysis involving a leading multinational chemical company operating across North, Central, and South America. This organization features a complex, geographically dispersed supply chain and relies on a diversified supplier base, including raw material providers, packaging suppliers, logistics operators, and industrial service partners with varying circular maturity levels. The chemical sector presents significant sustainability challenges related to hazardous waste management, emission control, regulatory compliance, resource recovery, and reverse logistics. Consequently, a structured framework for CSS becomes critical. In addition, increasing pressure from ESG reporting, environmental transparency, and traceability requirements intensifies the need for procurement tools capable of integrating circular, technological, operational, and governance dimensions.

This study interviewed fifteen supply chain and sustainability experts. These professionals possess over 15 years of industry experience across the Americas, with academic and practical backgrounds in business, environmental studies, and production engineering. This sample size is well-suited for expert-driven decision-making methods, which prioritize domain knowledge, representativeness, and the quality of judgment over statistical generalization. Foundational work on expert-based modeling highlights that such techniques elicit structured insights from knowledgeable individuals [54] and that no universally prescribed sample size exists since suitability depends on the specific research context [55,56].

To mitigate conformity bias, moderate discussion rounds elicited and consolidated the expert judgments. Professional experience and CSC practices based open debates and justification of divergent views during these sessions, allowing a transparent consensus to emerge [57]. Supply chain consultants and managers located across North, Central, and South America participated, ensuring functional and geographic diversity without any conflicts of interest. Simple majority voting resolved discussions that did not result in immediate convergence. Judgment intervals were used to represent disagreements that were not yet resolved, in accordance with established protocols for group decision-making [58]. This procedure enhances both the transparency and replicability of the evaluation.

3.2. Fuzzy DEMATEL

The structured framework for CSS integrates fuzzy logic to manage uncertainty in expert judgments and to enhance multi-criteria assessment robustness. Chen-Yi et al. [59] summarize the Fuzzy DEMATEL in the following steps:

A.

Define the criteria and introduce a fuzzy scale. Express performance levels through linguistic terms mapped to triangular fuzzy values. The scale ranges from no influence (values close to 0), through low and medium influence (intermediate values between 0.25 and 0.75), up to high and strong influence (values approaching 1.0). This linguistic-to-numeric mapping provides a structured way to capture subjective judgments [59,60].

B.

Build the fuzzy decision matrix. In this mathematical step, the triangular fuzzy parameters, $l_{ij}, m_{ij}, \mathrm{a}\mathrm{n}\mathrm{d} u_{ij}$ represent the lower, middle, and upper bounds of the influence of the criterion $C_i$ on criterion $C_j$, where $C_i$ function as the influencing criterion and $C_j$ as the influencing criterion.

C.

Derive the area center numbers $\left(l,m,u\right)$ using (1). Equation (1) computes the area center number (AC), summarizing the fuzzy judgments into a single representative value.

$$AC={\displaystyle \frac{(u-l)+(m-l)}{3}}+l$$

(1)

This step condenses the uncertainty of expert inputs into a usable score for weighting.

D.

Adjust the weights of the criteria by defining the coefficients ${AC}_j$ for each criterion j. Equation (2) normalizes these (${AC}_j)$ values to calculate the relative weight $w_j$ of each criterion.

$$w_j={\displaystyle \frac{{AC}_j}{{\sum }_{j=1}^n{AC}_j}}$$

(2)

where the variables $w_j$ denote the normalized weight of the criterion $j, {AC}_j$ represents the adjusted coefficient value associated with criterion j, and i and j denote the influencing and influenced criteria, respectively.

DEMATEL uses these weights to build the causal relation diagram.

E.

Apply DEMATEL to create a causal relation diagram. Equations (3) and (4) present the sum of rows (R) and columns (C). These equations compute the total influence exerted by each criterion (R) and the degree to which the other criteria influence each criterion (C).

$$r_i={\sum }_{j=1}^n{t}_{ij}$$

(3)

$$c_i={\sum }_{i=1}^n{t}_{ij}$$

(4)

where the components $r_i$ represent the total influence exerted by the criterion $i,c_j$ the total influence received by the criterion $j,t_{ij}$ and the influence of the criterion $i$ on criterion $j$, respectively.

The resulting R + C and R − C scores classify criteria into cause-and-effect groups, guiding the identification of the Best and Worst criteria in BWM. Specifically, R + C reflects the overall significance of the factor within the decision context, while R − C highlights its net effect. While alternative methods for decision-making in CSS include Data Envelopment Analysis (DEA) and the BOCR framework [44,49,61]. With the Fuzzy DEMATEL, it is possible to capture these causal interdependencies.

Expert judgments compose the fuzzy influence matrix, and group discussion validated the causal relations. This causal structure not only classifies criteria into cause-and-effect groups but also provides the conceptual basis for identifying the Best and Worst criteria in the subsequent BWM stage.

3.3. Best-Worst Method

In this study, BWM determines the relative importance of the CSS criteria. The choice to adopt this method lies in its analytical versatility and its capacity to address complex MCDM issues within CSC environments [50,62]. Rezaei [45] describes the process of BWM application:

A.

Identify the criteria as the decision set $\left\{c_1,c_2,\dots ,c_n\right\}$.

B.

Identify the criterion considered most important (Best) and the one deemed least relevant (Worst).

C.

Conduct pairwise judgments among the criteria, applying a 1–9 rating scale. The best-to-other representation is as a vector $A_B=(a_{B1},a_{B2},\dots ,a_{Bn})$, where the components $a_{Bj}$ denotes the preference of the Best criterion over each criterion j.

D.

Express the worst-to-others representation as a vector $A_w={\left(a_{1w},a_{2w},\dots ,a_{nw}\right)}^T$, where T indicates the transpose of the vector.

E.

Determine the optimal set of weights $\left(w_1^{\mathrm{*}},w_2^{\mathrm{*}},\dots ,w_n^{\mathrm{*}}\right)$.

F.

Execute an integrity check. The consistency ratio (CR) has ${\xi }^{\mathrm{*}}$ in the numerator representing the optimal consistency value obtained from the linear optimization model. Equation (5) presents the calculation of the CR, which measures the reliability of the weights obtained in BWM. The parameter measures represent the maximum deviation in the pairwise comparisons, while CI denotes the consistency index.

$$CR={\displaystyle \frac{{\xi }^{\mathrm{*}}}{CI}}$$

(5)

The CI in the BWM is defined on a 1–9 preference scale, with values ranging approximately from 0 to 5.53, as reported in the methodological literature [45,63]. This ensures that the derived weights maintain logical coherence and that the decision-making process remains reliable before proceeding to the AHP stage.

In practice, the participating experts identified the Best and Worst criteria based on the prior causal insights, and consistency was checked using the CR index. The weights obtained through BWM are subsequently embedded into the AHP hierarchy. This step ensures that the evaluation of sub-criteria and supplier alternatives reflects consistent and conceptually justified priorities. Once BWM establishes these primary criterion weights, the next stage introduces AHP to assess the final supplier alternatives for CSS.

3.4. Analytic Hierarchy Process

AHP is introduced to extend the analysis by ranking the sub-criteria and alternatives. The method has long been regarded as a benchmark in MCDM research [64,65] and is widely implemented in CE applications [66].

The foundations of AHP rely on the Saaty scale [67]. A pairwise comparison matrix A is developed, and from it, the eigenvector $w$ and eigenvalue ${\lambda }_{max}$ are derived. These components support the decision-making process by allowing the calculation of relative priorities, as shown in (6). Specifically, Equation (6) computes the eigenvector of priorities from the pairwise comparison matrix.

$$Aw={\lambda }_{max}w$$

(6)

This procedure ensures that the relative importance of sub-criteria and alternatives is mathematically consistent. When judgments are perfectly consistent, the maximum eigenvalue equals the matrix order ${\lambda }_{max}=n$. Conversely, in the presence of inconsistency, ${\lambda }_{max}$ becomes greater than n. The Consistency Index (CI) in (7) reflects this level of inconsistency. Equations (7) and (8) collectively verify the reliability of the judgments:

$$CI=({\lambda }_{max}-n)/(n-1)$$

(7)

Equation (8) defines the Consistency Ratio (CR) by integrating the Random Index (RI) for a given matrix size n. Values of CR exceeding 0.10 imply that the judgments should be reviewed [66].

$$CR=CI/RI$$

(8)

The RI values are derived from reference tables established by Saaty [66] through simulations of random pairwise comparison matrices. Because RI depends on the matrix size (n), for example, $RI=0.58 \mathrm{w}\mathrm{h}\mathrm{e}\mathrm{n} n=3$, $RI=0.90 \mathrm{w}\mathrm{h}\mathrm{e}\mathrm{n} n=4$, and $RI=1.12 \mathrm{w}\mathrm{h}\mathrm{e}\mathrm{n} n=5.$ Representative parameters are applied accordingly. Mathematically, a CR value below 0.10 confirms judgment reliability, whereas higher values require a collaborative review of the inputs. These pairwise comparisons, executed based on expert input, remain under consistency thresholds to preserve hierarchical reliability.

The decision to apply AHP in combination with Fuzzy DEMATEL and BWM is justified by its capacity to structure complex problems into a transparent hierarchy, complementing both causal analysis and criteria weighting [68,69,70]. This integration provides methodological coherence and establishes a robust, structured framework for CSS. In addition, a sensitivity analysis was conducted to assess supplier ranking stability under varying weight scenarios, thereby reinforcing evaluation credibility. By incorporating the weights derived from BWM, the AHP stage extends the analysis to supplier alternatives. This ensures that the final hierarchical evaluation aligns completely with the causal relationships and consistent weights established in the preceding stages.

4. CSS Structured Framework for Deployment Results

4.1. Fuzzy DEMATEL Results

In the application of Fuzzy DEMATEL, the five main CSS criteria (C1 to C5) detailed in Table 1 were evaluated. Following the sequential methodological steps, expert linguistic judgments were converted into fuzzy decision matrices, defuzzified, and normalized. The resulting row (R) and column (C) sums, along with the prominence (R + C) and relation (R − C) scores, are detailed in

Table 3. Fuzzy DEMATEL results for CSCS criteria.

Criterion	R (Row Score)	C (Column Score)	R + C (Importance/Relation)	R − C (Cause/Effect)	Classification
C1	11.830	11.987	23.817	−0.157	Neutral
C2	12.501	10.921	23.422	1.580	Strong Cause
C3	10.825	12.100	22.925	−1.275	Effect
C4	11.580	11.334	22.914	0.247	Slight Cause
C5	11.700	12.096	23.797	−0.396	Effect

.

Table 3 summarizes the Fuzzy DEMATEL scores (R + C and R − C) and the resulting classifications. Within this matrix, the prominence values range from 22.914 to 23.817, while the directional relation values range between −1.275 and +1.580. These analytical outcomes are synthesized in the classification column and visually mapped in the distribution diagram in Figure 3.

Figure 3 highlights the causal–effect distribution across the system. Environmental governance (C2) stands out as the most influential factor, with the highest positive R − C value, positioning it as a strong cause. Technological enablement (C4) shows moderate influence, classified as a slight cause. Conversely, economic and operational performance (C3) and social and collaborative sustainability (C5) exhibit negative R − C scores, confirming their roles as net effects. Circular performance (C1), with a near-zero R − C score, occupies a neutral position, indicating a balanced interplay between systemic influence and dependence.

These results demonstrate robust logical consistency. Governance and technology act as primary systemic drivers, while economic and social outcomes emerge as consequences. Circularity, positioned neutrally, reflects its integrative role across the operational pipeline. This structural coherence strengthens the methodological transition, as the Fuzzy DEMATEL outcomes establish the formal criteria for the subsequent BWM phase, designating environmental governance (C2) as the Best criterion and economic and operational performance (C3) as the Worst criterion.

The emergence of environmental governance (C2) as the dominant causal factor indicates that institutional and regulatory mechanisms are structural drivers rather than mere supportive elements in CSC. Consequently, governance establishes the foundational conditions under which technological innovation, economic performance, and social collaboration thrive. The classification of technological enablement (C4) as a secondary cause reinforces the argument that digitalization and innovation act as amplifiers of governance, accelerating the diffusion of circular practices.

In contrast, identifying the economic and social dimensions as effects highlights their dependency on upstream drivers. This dependency implies that performance improvements in these areas rely heavily on prior investments in governance and technology. This causal configuration reveals a systemic hierarchy: governance and technology function as strategic levers, while economic and social outcomes materialize as consequences. Such a pattern underscores that circularity requires institutional alignment and technological transformation rather than isolated operational gains. This finding aligns directly with decision science literature regarding the primacy of structural drivers in complex networks.

These findings resonate with broader discussions in supply chain management and CE literature. Prior studies [71,72,73] emphasize that governance mechanisms establish the institutional foundation for sustainable practices, while technological innovation accelerates their diffusion. Furthermore, classifying performance dimensions as effects supports decision science perspectives that emphasize outcome dependence over isolated actions. Accordingly, the Fuzzy DEMATEL results transcend numerical classification by substantiating theoretical debates on systemic levers in CSC. This integration provides valuable academic insight and managerial guidance, proving that robust governance and technological readiness are mandatory prerequisites for achieving long-term economic and social sustainability.

4.2. BWM Results

Based on the Fuzzy DEMATEL outcomes, environmental governance ($C2$) was designated as the Best criterion, reflecting its dominant systemic role. Conversely, economic and operational performance (C3) was identified as the Worst criterion, reinforcing its interpretation as an outcome rather than a driver. This alignment between causal influence and preference anchors represents a structured methodological assumption. While Fuzzy DEMATEL identifies structural drivers and dependencies, BWM captures decision-maker preferences. Consequently, the equivalence adopted here serves as a pragmatic design choice to ensure framework coherence, though its limitations require cautious interpretation.

The remaining criteria—circular performance (C1), technological enablement (C4), and social and collaborative sustainability (C5)—occupied intermediate positions. Based on these preferences, BWM was applied to determine the optimal weight allocation among the five dimensions, where the variables $w_{C1},w_{C2},w_{C3}, w_{C4}, \mathrm{a}\mathrm{n}\mathrm{d} w_{C5}$ denote the respective weights for each criterion (

Table 4. Pairwise judgments.

Judgment (BO/OW)	C1	C2	C3	C4	C5
BO (Best = C2, O = Others)	2	1	7	3	4
OW (O = Others, Worst = C3)	2	1	1	3	3

).

To derive optimal criteria weights, the BWM mathematical framework minimizes the maximum deviation between the pairwise comparison ratios and the estimated weights. The objective function and constraints are constructed based on the Best-to-Others (BO) and Others-to-Worst (OW) preference vectors elicited during the expert evaluation (Table 4). This optimization maximizes comparison consistency while computing the final weights, as detailed in Equations (9)–(13):

$$\min \xi$$

(9)

Based on the evaluations comparing (C2 to others) (10):

$$\left\{\begin{array}{c}\left|{\displaystyle \frac{w_{C2}}{w_{C1}}}-2\right|\le \xi \\ \left|{\displaystyle \frac{w_{C2}}{w_{C3}}}-7\right|\le \xi \\ \left|{\displaystyle \frac{w_{C2}}{w_{C4}}}-3\right|\le \xi \\ \left|{\displaystyle \frac{w_{C2}}{w_{C5}}}-4\right|\le \xi \end{array}\right.$$

(10)

Based on the evaluations comparing (others to C3) (11):

$$\left\{\begin{array}{c}\left|{\displaystyle \frac{w_{C1}}{w_{C3}}}-2\right|\le \xi \\ \left|{\displaystyle \frac{w_{C2}}{w_{C3}}}-1\right|\le \xi \\ \left|{\displaystyle \frac{w_{C4}}{w_{C3}}}-3\right|\le \xi \\ \left|{\displaystyle \frac{w_{C5}}{w_{C3}}}-3\right|\le \xi \end{array}\right.$$

(11)

Normalization and non-negativity (12) and (13):

$$w_{C1}+w_{C2}+w_{C3}+w_{C4}+w_5=1$$

(12)

$$w_j\ge 0 \forall j$$

(13)

The optimal value (9) obtained from this optimization problem is denoted by ξ*, which is subsequently used to compute the consistency ratio defined in Equation (5). The solver was applied, producing the criteria weights. The solver generated a CR of 0.143, remaining safely below the admissible limit of 0.282 for a five-criterion structured framework for CSS. Although this value indicates a minor level of inconsistency, such variations are typical in complex multi-criteria environments and fall well within the thresholds established in the BWM literature, thereby confirming methodological validity.

The resulting distribution reveals that circular performance (C1) received the highest value, 0.325, followed by technological enablement (C4) at 0.217. Interestingly, although C2 was designated as the initial anchor for preferences, the final optimization prioritized C1, highlighting the operational emphasis placed on active circular practices.

This strong score for technological enablement (C4) aligns with operational reality, as circular suppliers depend on technological enablement for traceability and innovation. Social and collaborative sustainability (C5) registered 0.162, reinforcing that collaboration and engagement are important but not dominant. Finally, economic and operational performance (C3) obtained the lowest weight, 0.093. This outcome follows the logic that traditional cost matrices are secondary to sustainability dimensions in circular supply networks. Collectively, these BWM results provide the foundational weights for the subsequent evaluation within the structured framework for CSS using the AHP method.

4.3. AHP Results

At this stage, the weights derived from the BWM serve as inputs for the AHP evaluation, enabling systematic prioritization of criteria and sub-criteria. The resulting local and overall priorities for each dimension are presented in

Table 5. Priorities, criteria, and sub-criteria.

Criteria and Sub-Criteria	Local	Overall
C1 Circular Performance	0.325	0.325
SC1 Circular design capability	0.477	0.155
SC2 Circular innovation capability	0.234	0.076
SC3 Reverse logistics capability	0.122	0.040
SC4 Recycling & recovery rate	0.093	0.030
SC5 Use of secondary materials	0.074	0.024
C2 Environmental Governance	0.203	0.203
SC6 Carbon Management	0.247	0.050
SC7 Environmental Management Systems	0.203	0.041
SC8 Environmental Transparency	0.326	0.066
SC9 Regulatory Compliance	0.131	0.027
SC10 Waste Management Systems	0.094	0.019
C3 Economic & Operational Performance	0.093	0.093
SC11 Cost Efficiency	0.042	0.004
SC12 Delivery Reliability	0.350	0.033
SC13 Financial Stability	0.265	0.025
SC14 Operational Flexibility	0.228	0.021
SC15 Quality Reliability	0.115	0.011
C4 Technological Enablement	0.217	0.217
SC16 Analytics for Circular Optimization	0.250	0.054
SC17 Collaborative Digital Platforms	0.161	0.035
SC18 Data Integration Capability	0.379	0.082
SC19 Digital Traceability	0.122	0.027
SC20 Industry 4.0 Readiness	0.087	0.019
C5 Social & Collaborative Sustainability	0.162	0.162
SC21 Corporate Social Responsibility	0.452	0.073
SC22 Ethical & Sustainable Sourcing	0.221	0.036
SC23 Health & Safety Practices	0.146	0.024
SC24 Stakeholder Collaboration	0.117	0.019
SC25 Transparency & Trust	0.064	0.010

.

The consistency ratio (CR) was calculated for each of the five sub-criterion matrices. The resulting values (0.064, 0.076, 0.065, 0.095, and 0.022) remain strictly below the admissible threshold of 0.10, confirming acceptable consistency and validating the derived weights.

In Table 5, the results reveal a coherent structure across the five criteria. In C1 circular performance, circular design capability emerged as the most influential sub-criterion, with a score of 0.155. This outcome is consistent with the idea that design is the foundation of circularity. Without products designed for reuse, recycling, and reverse logistics, other circular practices cannot be fully implemented. In C2 environmental governance, environmental transparency scored 0.066. This reflects the growing importance of disclosure and traceability in CSC, where suppliers are increasingly required to provide clear and verifiable environmental data. Carbon management and environmental management systems remain relevant, but transparency is the key differentiator.

For C3 economic and operational performance, the delivery reliability obtained 0.033. This is coherent with CSS, as reliability in delivery directly impacts operational efficiency and customer satisfaction. Financial stability and operational flexibility also contribute substantially, ensuring resilience and adaptability. In C4 technological enablement, data integration capability registered a score of 0.082. This highlights the role of digitalization in enabling circular practices, as integrated data systems allow for the traceability, monitoring, and optimization of circular flows. Analytics for circular optimization also contributes significantly, reinforcing the technological backbone of CSC.

Finally, in C5 social responsibility and ethics, corporate social responsibility (SC21) reached 0.073. Although SC21 is not strictly a circularity measure, its prominence reflects the broader expectation that CSS must operate ethically and responsibly. Importantly, the adjustment reduced its dominance, compared to earlier versions, allowing ethical sourcing and health and safety practices to carry more weight, which creates a more balanced evaluation.

This research employed performance levels to evaluate the alternatives, with each option assessed individually against a defined rating. The scale included the following categories: excellent 1.000, very good 0.830, between good and very good 0.670, good 0.500, between weak and good 0.250, and weak 0.10 [36]. The overall ranking of the alternatives is shown in

Table 6. Rank of alternatives.

Alternative	Overall	Rank
A1	0.754	4th
A2	0.603	8th
A3	0.787	3rd
A4	0.459	9th
A5	0.884	1st
A6	0.612	7th
A7	0.745	5th
A8	0.412	10th
A9	0.855	2nd
A10	0.635	6th

. For confidentiality reasons, the ten supplier alternatives (A1–A10) are presented anonymously; however, they represent real suppliers with varying sizes, locations, and levels of circularity maturity, ensuring that the ranking remains meaningful while preserving anonymity.

In Table 6, the results indicate a clear differentiation among the CSS alternatives, with performance scores ranging from 0.412 to 0.884. The ranking reveals that A5 achieved the highest overall score, 0.884, followed by A9 with 0.855 and A3 with 0.787. These scores indicate superior performance across the evaluated criteria, positioning these alternatives as the most suitable options within the CSS context.

A second group of alternatives, including A1 at 0.754 and A7 at 0.745, also demonstrates strong performance, although slightly lower than the top-ranked suppliers. This suggests that these alternatives present competitive characteristics but with some limitations in specific criteria. Mid-ranked alternatives such as A10 (0.635), A6 (0.612), and A2 (0.603) show moderate performance, indicating a balanced but less competitive profile compared to the leading suppliers.

Finally, A4 with 0.459 and A8 with 0.412 occupy the lowest positions in the ranking, reflecting comparatively weaker performance across the evaluated dimensions. These outcomes suggest that such alternatives may require significant improvements to meet the desired circular and sustainability standards. The comparison between alternatives and criteria is illustrated in Figure 4.

The radar chart (Figure 4) illustrates the priorities of alternatives (A1–A10) across criteria (C1–C5), confirming the structured framework for CSS’s ability to differentiate suppliers and support structured decision-making in CSS. The sensitivity analysis for the circular performance criterion (C1) is introduced in Figure 5.

It should be noted that this evaluation represents an initial robustness check, limited to criterion C1 and a single variation. A more comprehensive sensitivity analysis, systematically varying all criteria and assessing the stability of the entire ranking, is recognized as necessary and will be addressed in future work.

Specifically, Figure 5 displays the results for C1, where the black vertical line represents the original weight of 0.325, while the red line indicates the doubled weight of 0.650. Even under this adjustment, the ranking of alternatives remains unchanged, demonstrating robustness. This stability reinforces that the framework delivers consistent outcomes and serves as a reliable, structured framework for CSS.

To complement the sensitivity analysis illustrated in Figure 5,

Table 7. Ranking stability of alternatives under scenario variations.

Scenario Description	Top-Ranked Alternatives	Mid-Ranked Alternatives	Low-Ranked Alternatives	Rank Changes Observed
Base weights (original)	A5, A9, A3	A1, A7, A10, A6, A2	A4, A8	None
C1 weight doubled (Figure 5)	A5, A9, A3	A1, A7, A10, A6, A2	A4, A8	None
C1 weight reduced by half	A5, A9, A3	A1, A7, A10, A6, A2	A4, A8	None

summarizes the stability of alternative rankings under different scenario variations. As shown, the ranking remains unchanged across base, doubled, and reduced weights for circular performance (C1), confirming robustness and facilitating managerial interpretation.

Importantly, the observed stability does not suggest that the alternatives are homogeneous in performance. On the contrary, the results reveal a persistent separation between high-, medium-, and low-performing suppliers across the entire variation range of C1. Top-ranked suppliers such as A5 and A9 maintained superior positions even when the weight of circular performance was substantially increased. This trend indicates that these suppliers demonstrate balanced capabilities across multiple dimensions rather than isolated strengths in a single criterion. Similarly, lower-ranked alternatives remained comparatively weak despite variation, suggesting the presence of structural gaps in sustainability and circularity performance.

From a managerial perspective, this robustness indicates that procurement decisions supported by the proposed structured framework for CSS are not excessively sensitive to moderate or even substantial changes in strategic priorities related to circular performance. In practical terms, firms may adjust the relative importance of circularity objectives without causing disruptive ranking instability or inconsistent supplier selection outcomes. This stability strengthens a structured framework for CSS reliability within CSC environments characterized by evolving ESG pressures, regulatory requirements, and sustainability priorities.

Furthermore, the absence of significant rank reversals suggests that the structured framework for CSS maintains internal coherence across environmental, technological, operational, and social dimensions. Rather than favoring suppliers based on isolated excellence, the structured framework for CSS appears to reward consistently balanced performance. This characteristic is particularly relevant for long-term supplier relationship management in CSC.

The sequential application of Fuzzy DEMATEL, BWM, and AHP does not exclusively represent procedural integration but also creates analytical complementarity. The causal hierarchy identified by Fuzzy DEMATEL guided the designation of Best and Worst criteria in BWM, ensuring that weight allocation reflected systemic drivers. These weights were then embedded into the AHP hierarchy, allowing the prioritization of sub-criteria and alternatives to be directly linked to the causal logic. This interaction demonstrates that the integration of methods provides significant added value by linking structural drivers to practical supplier rankings, thereby reinforcing both theoretical debates and managerial decision-making within CSC.

5. Discussion

These findings are consistent with prior studies applying Fuzzy DEMATEL to sustainable supply chains. Environmental governance mechanisms have been identified as strong causal drivers shaping systemic sustainability outcomes [74]. Technological enablers support systemic outcomes by enhancing transparency and traceability, although they do not act as primary determinants [75]. In contrast, economic and social criteria tend to emerge as dependent effects, reflecting a reactive nature within the system. This coherence reinforces the methodological transition to the BWM, where environmental governance (C2) serves as the best criterion and economic and operational performance (C3) as the worst.

The BWM analysis highlights a shift in emphasis compared to the causal structure identified by Fuzzy DEMATEL. While environmental governance (C2) emerged as the main driver in the causal model, the prioritization indicates that circular performance (C1) is perceived as the most decisive factor in CSS. This variation highlights the complementary nature of the two methods: Fuzzy DEMATEL explains systemic influence, whereas BWM captures practical decision-making priorities.

The divergence between Fuzzy DEMATEL and BWM results provides a theoretically enriching insight. While Fuzzy DEMATEL identified environmental governance (C2) as a systemic driver, BWM highlighted circular performance (C1) as the most decisive priority. This deviation illustrates that structural influence and managerial preference may diverge systematically. Specifically, governance mechanisms shape the system, but decision-makers emphasize operational circularity as more actionable. Such tension contributes to the theoretical debate on multi-criteria decision-making in sustainability, showing that integrated methods can reveal not only complementarities but also analytically and managerially significant misalignments.

While our findings align with prior studies emphasizing governance and circularity as key drivers in sustainable supply chains, critical literature also highlights tensions that complicate this picture. Ref. [72] argues that assessment frameworks for CSC remain underdeveloped, making prioritization highly context-dependent and potentially unstable. Similarly, recent reviews note that governance mechanisms, though structurally influential, often face implementation barriers that limit their practical prioritization [71].

These perspectives suggest that the divergence we observed between systemic influence (C2) and managerial priority (C1) may reflect broader challenges in translating structural sustainability drivers into operational decision-making. Engaging with such critical viewpoints enriches the theoretical contribution of our study. Ultimately, it shows that methodological integration not only reveals complementarities but also exposes systemic tensions that require further exploration.

The prominence of circular performance (C1) reflects a paradigm shift in CSS evaluation, where sustainability outcomes outweigh traditional economic considerations. Technology (C4) also appears as a strong enabler, suggesting that digital tools and innovations are essential to operationalize circular practices. Social and collaborative (C5) aspects provide supportive value by reinforcing the importance of stakeholder engagement. The economic criterion (C3), although present, remains secondary, consistent with the principles of CSC. In sum, this study demonstrates that supplier selection in circular contexts is guided more by the ability to deliver sustainable and technologically enabled solutions than by cost efficiency, aligning with the broader transition toward CE models.

AHP results highlight that the distribution of priorities across the five criteria and sub-criteria demonstrates a coherent and realistic structured framework for CSS. Circular design and data integration anchor the evaluation in circularity principles, while environmental transparency aligns with global ESG reporting requirements. Operational reliability and financial stability provide the necessary economic foundation, whereas social responsibility ensures that suppliers meet ethical and societal expectations.

These results suggest that CSS cannot rely solely on environmental or social criteria; instead, it requires a balanced approach that integrates design, technology, governance, operations, and ethics. The fact that transparency and corporate social responsibility emerged as dominant in their respective dimensions reflects current market pressures, but the overall model still places circularity-related criteria at the core. This balance strengthens the coherence of the framework and makes it suitable for practical application in CSS processes.

The AHP outcomes highlight the structured framework for CSS’s ability to clearly distinguish supplier performance within CSC. The presence of top-performing alternatives demonstrates that excellence in circular practices is achievable, while mid-ranked suppliers reveal partial alignment that could be strengthened through targeted improvements. Lower-ranked suppliers expose structural or strategic gaps, underscoring the need for investment and adaptation to meet sustainability requirements.

Methodologically, the consistency between Fuzzy DEMATEL, BWM-derived weights, and AHP rankings reinforces the robustness of the integrated approach. From a managerial perspective, the results provide actionable insights for procurement professionals operating within circular supply networks:

• Institutional ESG and Audits: Since environmental governance (C2) emerged as a key systemic driver and environmental transparency (SC8) achieved the highest priority within this dimension, firms should incorporate standardized ESG disclosure requirements, environmental reporting protocols, and traceability-based audits into their supplier qualification and contract renewal processes. For instance, procurement managers can apply these findings by requiring suppliers to submit sustainability disclosures aligned with global frameworks, such as the Global Reporting Initiative (GRI) or Sustainability Accounting Standards Board (SASB). Additionally, mandating annual audits—such as Standard (ISO 14001) certification [76]—can effectively verify environmental compliance. Rather than limiting audits to regulatory compliance, managers should adopt recurring circularity assessments covering circular design capability (SC1), recycling and recovery practices (SC4), and the use of secondary materials (SC5).
• Digital Systems Infrastructure: The strong performance of technological enablement (C4), particularly data integration capability (SC18), indicates that firms should prioritize investments in interoperable digital systems capable of monitoring supplier environmental and operational data in real time. To achieve this, managers should allocate capital to Enterprise Resource Planning (ERP) modules equipped with sustainability dashboards. They should also pilot blockchain-enabled traceability tools within high-risk sourcing categories and implement collaborative digital platforms to support reverse logistics coordination, circular-flow monitoring, and supplier transparency across the supply chain.
• Data-Driven Supplier Evaluation: The results further suggest that procurement managers should adopt data-driven systems within the structured framework for CSS that integrate circularity indicators, ESG metrics, delivery reliability, and digital readiness into supplier scorecards. By embedding the exact priority weights derived from this hybrid Fuzzy DEMATEL-BWM-AHP framework directly into corporate procurement scorecards, firms can ensure that day-to-day purchasing decisions systematically prioritize suppliers with stronger long-term circular capabilities over short-term cost efficiencies.
• Targeted Supplier Action Plans: The findings also indicate differentiated managerial actions according to supplier performance profiles:
  • Top-ranked suppliers (such as A5 and A9) should be integrated into long-term strategic partnerships focused on circular innovation and co-development initiatives.
  • Mid-ranked suppliers should receive targeted capability-building support related to digital integration, environmental reporting, and operational resilience. For example, procurement teams can provide technical workshops on digital traceability tools or supply standardized environmental reporting templates to help these partners drive continuous improvement.
  • Low-ranked suppliers should be subject to corrective improvement plans involving Corporate Social Responsibility (CSR) compliance, sustainability audits, and minimum circularity performance requirements linked to future sourcing eligibility. This operational turnaround can be managed by setting explicit, time-bound compliance thresholds as mandatory conditions for remaining in the sourcing pool.

Therefore, the proposed framework contributes not only as a supplier ranking mechanism but also as a governance-oriented managerial tool capable of supporting procurement policies, supplier development strategies, and digital traceability investments within CSC environments.

6. Conclusions

This study presents a structured framework for CSS by integrating Fuzzy DEMATEL, BWM, and AHP. The contribution lies in demonstrating how the complementary use of these methods can capture causal relationships, prioritize criteria, and rank alternatives in a coherent manner. Rather than introducing a fundamentally new methodology, the study advances the integration of established MCDM techniques within the CSC context, providing actionable insights for managers seeking alignment with CE principles.

The expert evaluations carried out within a particular CSC context resulted in the structured framework for CSS. Then, the framework is adaptable to different industries. However, the outcomes may vary according to sector-specific characteristics and expert perspectives.

While the study advances the integration of established MCDM techniques and demonstrates their complementarity, the results remain context-specific and should be interpreted with caution. The reliance on expert consensus, drawn from a single industry setting, may introduce bias. Furthermore, this approach does not capture how supplier capabilities evolve, nor does it allow a direct comparison with actual performance data. The framework is adaptable to other industries and regions, but empirical validation across SMEs and diverse sectors is still required to ensure robustness and generalizability. In this sense, the contribution is conceptual and methodological rather than definitive, providing a foundation that future research can build upon to strengthen both theoretical development and practical implementation in CSS.

Future research could analyze the application of ANP within the MCDM method. This approach is particularly suited to capture complex interdependencies among criteria that Fuzzy DEMATEL cannot fully address, thereby strengthening the robustness of the structured framework for CSS.

In addition, testing the structured framework for CSS in diverse industrial contexts would help validate its adaptability and scalability, while exploring hybrid approaches that integrate digital traceability and ESG reporting not only as complementary indicators but as methodological components of the evaluation process. This integration could be achieved, for example, by embedding traceability data into supplier performance metrics and using ESG disclosures as structured inputs for multi-criteria analysis.

As suggested by an anonymous reviewer, future studies could also investigate the alignment of the proposed structured framework for CSS with established supply chain reference models, such as SCOR digital, to strengthen comparability with resilience, economic, social, and environmental performance measures. In addition, the anonymous reviewer highlighted the need for empirical validation in SMEs and across different sectors, which remains a crucial direction for future research to ensure robustness and generalizability. Future research should also extend the sensitivity analysis by systematically varying all criteria and examining the stability of the entire ranking, thereby providing a more comprehensive robustness assessment of the structured framework for CSS.

Overall, the framework offers a structured framework for CSS that supports governance in CSC and contributes to aligning supplier evaluation with CSS performance outcomes, while acknowledging its contextual limitations.