Application of FAHP–QFD–NLP Hybrid Decision Model in Assessment of Priority Strategies of Corporate Sustainability

Abstract

This study presents a structured decision-making framework to evaluate and prioritize corporate sustainability actions under the triple bottom line, encompassing economic performance, social value, and environmental responsibility. First, we use fuzzy analytic hierarchy process to determine the relative importance of sub-criteria across TBL dimensions based on expert judgments and selection. Second, we use quality function deployment to translate those weighted priorities into ranked indicators by the house of quality. Finally, using natural language processing, we employ BERT semantic similarity models and assess the alignment between the model’s recommended actions and actual corporate sustainability disclosures across 10 publicly listed companies. The integrated framework enables a systematic comparison between ideal strategic priorities and practical implementation. By aggregating expert-derived priorities and disclosure-based alignment, the model provides a quantitative basis for benchmarking firms and identifying gaps between intention and action. This approach offers a replicable, data-driven method for evaluating corporate sustainability strategies within the TBL structure.

Introduction

In the face of growing climate risks, social inequality, and regulatory pressure, firms are increasingly challenged to reconcile profit maximization with long-term sustainability. Traditional management models centered solely on shareholder value have proven inadequate for addressing these multidimensional goals. Corporate decision-making has gradually shifted from a shareholder-centric model toward the Triple Bottom Line (TBL) framework, including economic, social, and environmental performance [1, 2–3]. With increasing supervision from regulatory agencies and heightened societal expectations, Corporate Social Responsibility (CSR) and Environmental, Social, and Governance (ESG) criteria have placed greater emphasis on decision-making frameworks that synergize financial performance with social value [4]. For instance, the United Nations Sustainable Development Goals (SDGs) have become global benchmarks [5], the European Union’s Carbon Border Adjustment Mechanism influences supply chain decisions [6], and ESG rating agencies such as MSCI and Sustainalytics significantly impact investment attractiveness [7]. A recent study by PricewaterhouseCoopers [8] shows that over 70% of business leaders acknowledge ESG’s influence on competitiveness, yet most still lack robust, quantitative decision-making tools. Recent research also underscores the inadequacy of traditional approaches, which often overemphasize financial metrics while neglecting social and environmental outcomes [9, 10, 11–12]. As a result, sustainability strategies are frequently implemented in an unsystematic manner, limiting their long-term effectiveness [9, 10, 11–12].

Although sustainable development is widely accepted in principle, converting such goals into viable, operational strategies remains difficult for several reasons. First, ESG reporting serves as a disclosure tool, but does not guide specific resource allocation [11, 12]. Companies may be aware of their carbon footprint but are often uncertain whether to prioritize investments in renewable energy or optimize production efficiency [13]. ESG ratings tend to be backward-looking and lack predictive or prescriptive power, making strategic adjustments difficult. Second, corporate strategy must balance the conflicting interests of multiple stakeholders, including investors, regulators, consumers, and employees. For example, investors may focus on short-term profits, while regulators emphasize long-term environmental objectives [14]. Consequently, a persistent ‘intention–action gap’ remains in corporate sustainability management.

This intention–action gap underscores a critical research gap: existing frameworks often fail to move beyond disclosure and compliance, leaving firms uncertain about how to allocate resources across competing sustainability priorities. ESG ratings provide backward-looking evaluations but rarely prescribe concrete strategic actions. At the same time, stakeholder conflicts—such as reconciling short-term shareholder returns with long-term environmental commitments—create further ambiguity for managers [13, 14–15]. These limitations highlight the need for a hybrid decision-making approach that not only prioritizes sustainability objectives under uncertainty but also ensures their operationalization and external validation.

This study addresses the research question: How can firms systematically prioritize and operationalize sustainability strategies within the TBL framework to bridge the gap between intention and action? Unlike classical tools such as AHP, which provide structured prioritization but lack translation into actionable plans, or QFD, which links goals to actions but lacks external verification, our framework integrates FAHP, QFD, and NLP to address these shortcomings directly [16, 17–18].

To address these challenges, this study integrates FAHP, QFD, and NLP, where FAHP handles uncertainty in prioritizing criteria, QFD translates priorities into actionable indicators, and NLP validates alignment with corporate disclosures [16, 19, 20]. This layered integration enables a closed-loop mechanism: expert-driven prioritization (FAHP), structured action planning (QFD), and empirical validation (NLP). In doing so, the framework bridges the gap between sustainability intention and action, a limitation consistently noted in prior studies but rarely addressed holistically.

Building upon the FAHP and QFD foundation, this study further incorporates NLP as a third analytical layer. By applying BERT-based sentence embeddings with cosine similarity models, we evaluate the alignment between the 27 prioritized sustainability action indicators and actual disclosures in sustainability reports from the target corporations. This NLP-driven benchmarking mechanism enables empirical validation and comparison of corporate actions across the TBL dimensions, enhancing the model’s credibility and real-world applicability.

This study contributes theoretically by integrating FAHP, QFD, and NLP into a novel framework for sustainability decision-making, and practically by offering actionable guidelines for ESG-driven strategic planning [10, 21, 22, 23–24]. The objectives are:

1. To develop the FAHP–QFD–NLP model for prioritizing action indicators.

2. To validate the model’s predictive power through empirical analysis.

3. To provide resource allocation guidelines for enhanced ESG compliance.

Literature Review

Evolution of Sustainability Disclosures: From CSR to ESG and TBL

Sustainable development and its associated reporting frameworks have undergone significant evolution over the past few decades. The United Nations Brundtland Commission [24] introduced the concept of sustainable development as meeting present needs without compromising future generations’ ability to meet theirs. Initially, corporate sustainability efforts were disclosed as CSR reports, emphasizing philanthropy and ethical practices. Over time, these evolved into ESG reports, which introduced standardized evaluation criteria for external stakeholders and investors [25]. Rooted in stakeholder theory, CSR emphasized accountability to a broad range of parties, including shareholders, customers, suppliers, and communities [3]. ESG expanded this focus through quantitative indicators to guide investment decisions and sustainability performance evaluation [26, 27]. More recently, the term "sustainability report" has emerged, reflecting an integration of CSR and ESG concepts. While ESG primarily offers an investor-oriented external evaluation, TBL functions distinctly as a core management philosophy that emphasizes internal strategy alignment and long-term value creation. Despite these frameworks, businesses continue to face challenges in balancing economic performance with environmental and social objectives under resource constraints. This tension highlights the need for a structured internal decision-making model that can bridge the gap between qualitative sustainability aspirations and quantitative strategic implementation.

TBL as a Decision Framework

TBL theory [28, 29–30] has emerged as a robust conceptual model for guiding corporate sustainability decisions across three interdependent dimensions: economic, social, and environmental. TBL is increasingly applied in internal decision-making contexts such as supply chain optimization, resource planning, and corporate strategy formulation [28, 31].

In comparison with ESG, which serves as an investor-oriented evaluation framework [32], TBL functions as a management philosophy that emphasizes internal strategic alignment and long-term value creation. ESG frameworks are supported by standardized—though often inconsistent—metrics issued by agencies such as MSCI and Sustainalytics, serving external stakeholders including investors, analysts, and regulators [33, 34]. By contrast, TBL lacks strict standardization but offers a proactive, values-driven foundation for embedding sustainability into corporate strategy, organizational culture, and daily operations. ESG is frequently compliance-driven and reactive, whereas TBL promotes a forward-looking orientation that integrates sustainability into the firm’s core mission. While ESG evaluations can indirectly shape internal practices through investor pressure and regulatory oversight, their principal function remains financial risk and opportunity assessment. TBL, in contrast, is intrinsically linked to decision execution, stakeholder trust-building, and long-term resilience.

The distinct roles of TBL and ESG provide the rationale for our hybrid framework. Given this study's primary objective—to facilitate structured internal decision-making and resource prioritization—TBL is favored as the conceptual backbone for the FAHP and QFD stages. TBL’s focus on internal alignment and strategic action perfectly fits the needs of defining and prioritizing corporate action indicators. Conversely, ESG’s standardized metrics and external focus make it the essential benchmark for the NLP component, enabling us to empirically validate the model’s output against publicly disclosed corporate performance. This integrated approach leverages the internal strength of TBL and the external credibility of ESG, thereby ensuring the coherence of our proposed FAHP–QFD–NLP methodology (Table 1).

Table 1. Comparison of TBL and ESG

Aspect	TBL	ESG	Role in the FAHP–QFD–NLP framework
Main content	Balanced integration of economic, social, and environmental objectives	Adds the Governance dimension and quantifies sustainability dimensions	TBL's scope guides criteria definition; ESG provides the external performance dimensions for analysis
Purpose and audience	Internally oriented; guides managers and employees toward long-term sustainability goals	Externally oriented; used by investors, analysts, and regulators for financial evaluation	TBL's internal focus forms the foundation for FAHP–QFD prioritization (Action Planning)
Standardization and metrics	No universal standardization; qualitative and context-dependent	Standardized metrics exist but vary across rating agencies (e.g., MSCI, Sustainalytics)	ESG's structured metrics and public data are essential for the NLP external validation and benchmarking
Use cases	Internal strategic planning, supply chain optimization, organizational culture development	Investment screening, regulatory compliance, corporate risk management	TBL ensures the model focuses on actionable strategy; ESG aligns results with market/investor expectations
Limitations	Hard to benchmark across firms; less suitable for direct financial evaluation	Risk of inconsistent evaluations; often compliance-driven rather than transformative	The hybrid model integrates both to mitigate TBL’s lack of external validation and ESG’s lack of internal prescription

FAHP–QFD–NLP for Structured TBL Decision-Making

Decision-making in corporate sustainability is characterized by multiple conflicting criteria and high degrees of uncertainty. To address this, fuzzy set theory has been combined with the AHP, resulting in the FAHP, which improves the consistency and robustness of expert pairwise comparisons in multi-criteria environments [35, 36]. FAHP has been widely applied in evaluating sustainability initiatives, risk management, and resource allocation [37].

While FAHP effectively handles the inherent uncertainty in TBL criteria prioritization, it remains purely a prioritization tool that lacks a systematic mechanism to translate abstract priorities into concrete operational actions. This crucial gap necessitates the linkage to an action-oriented framework like QFD.

QFD, originally developed to translate customer needs into technical product specifications, has evolved into a versatile tool for decision-making across industries. As the central component of QFD, HOQ translates strategic priorities into actionable indicators and improves implementation efficiency [38, 39]. QFD has been applied in service innovation, healthcare, strategic planning, and sustainability strategy formulation [40, 41].

QFD typically relies on expert judgment to determine the relationships between goals and actions, which can introduce subjectivity. By integrating FAHP with QFD, strategic priorities derived from fuzzy logic can be systematically translated into executable plans, enhancing decision reliability and feasibility [42, 43]. For example, Haber et al. [44] combined FAHP, QFD, and the Kano model to improve hemodialysis device development, while Akbaş and Bilgen [45] used FAHP–QFD–TOPSIS to select optimal fuels for wastewater treatment. Ocampo et al. [46] incorporated FAHP, DEMATEL, and QFD to design sustainable edible oil products, and Fetanat & Tayebi [47] applied FAHP–QFD with linear programming for optimizing water purification systems. Liu et al. [48] combined FAHP and fuzzy QFD with Grey Relational Analysis to enhance user experience in vehicle design.

Although the FAHP–QFD integration provides a robust path from strategic prioritization to action planning, its results are limited to internal documentation and lack external empirical validation against actual market performance or disclosure reality. This limitation, which is critical for modern ESG accountability, justifies the final integration of a quantitative validation tool like NLP.

With the increasing digitization of corporate disclosures, NLP has emerged as a promising tool for analyzing sustainability reports and assessing disclosure alignment—gaining traction in both academic research and corporate practice. Recent work has also combined NLP with AHP. For instance, Darzi et al. [49] applied a social media–based AHP model to rank cellphone life-cycle issues for strategic CSR marketing, demonstrating the feasibility of integrating textual analytics with multi-criteria decision frameworks. Following these initial efforts, studies have increasingly adopted NLP techniques for evaluating the semantic alignment of sustainability disclosures. For example, Gutierrez-Bustamante and Espinosa-Leal [50] applied NLP methods to analyze the sustainability reporting quality of Nordic listed companies, demonstrating the method’s potential to systematically score and benchmark corporate sustainability narratives.

Although this integrated approach has seen extensive application in product development and operations management, its application in corporate governance and sustainability strategy remains scarce. To address this gap, the present study extends FAHP–QFD by incorporating NLP, enabling firms to align internal strategic planning with external sustainability disclosures through semantic similarity analysis. This hybrid framework provides a structured, data-driven approach to prioritize and benchmark sustainability actions under the TBL framework, thereby addressing both methodological robustness and practical relevance.

To strengthen comprehensiveness, this study extends the literature review by including recent contributions on FAHP in sustainability evaluation, QFD in ESG strategy development, and NLP in sustainability disclosure analysis. These additions allow for a more nuanced understanding of methodological diversity and thematic applications in the field. To provide a structured overview, prior studies are categorized in Table 2 by thematic focus (e.g., sustainability strategy, ESG compliance, CSR marketing) and methodological approach (e.g., FAHP, QFD, NLP).

Table 2. Structured categorization of recent studies on sustainability evaluation

Theme	Method	Representative studies	Connection to the proposed framework (Contribution/Limitation)
Sustainability strategy and disclosure evolution	CSR / ESG frameworks, bibliometric analysis	Mishra & Pandey (2025) [30]; Chipimo et al. (2025) [32]	Contribution: Defines the problem space (ESG/TBL distinction)
ESG compliance and reporting	QFD for ESG integration, survey-based ESG analysis	Fernandes & Pegino (2024) [19]; Khatib (2024) [21]	Contribution: Addresses action translation (QFD). Limitation: Often lacks rigorous upstream prioritization (FAHP) and empirical validation (NLP)
CSR marketing and stakeholder perspectives	NLP–AHP integration	Darzi et al. (2025) [49]	Contribution: Demonstrates the feasibility of NLP–AHP hybrid. Limitation: Limited to qualitative assessment and a specific scope (marketing/social media)
Sustainability performance evaluation	FAHP applications in risk and resource allocation	Govindan et al. (2021) [37]; Kamvysi et al. (2023) [22]	Contribution: Establishes rigorous priority setting (FAHP). Limitation: Focuses on abstract criteria; lacks action translation (QFD) and external validation (NLP)
Product and process design	FAHP–QFD hybrids with DEMATEL, TOPSIS, or LINMAP	Haber et al. (2020) [44]; Akbaş & Bilgen (2017) [45]; Fetanat & Tayebi (2021) [47]; Lin et al. (2024) [62]	Contribution: Solves the prioritization–action gap (FAHP–QFD). Limitation: Focus is on product design; lacks external ESG validation (NLP) required for corporate strategy
Corporate disclosure analysis	NLP for sustainability reports, semantic scoring	Gutierrez-Bustamante & Espinosa-Leal (2022) [50]; Schimanski et al. (2024) [63]	Contribution: Validates performance via disclosure (NLP). Limitation: Descriptive only; lacks upstream strategic planning (FAHP) and action planning (QFD)
Advanced fuzzy extensions in sustainability	Circular Intuitionistic Fuzzy; SWARA–CoCoSo; Cubic Picture Fuzzy	Athar Farid et al. (2025) [64]; Riaz et al. (2024) [65]; Ghoushchi et al. (2022) [66]	Limitation: Highly specialized and complex; fragmented (sector-specific) and not applicable as an integrated corporate strategic framework

Beyond the above categories, recent studies have further advanced fuzzy set extensions to tackle sustainability challenges. For example, circular intuitionistic fuzzy frameworks have been applied in sustainable logistics and robotics evaluation, entropy-based SWARA combined with CoCoSo has supported renewable energy prioritization, and cubic picture fuzzy approaches integrated with blockchain and the metaverse have addressed uncertainty in supply chain management.

While previous contributions have demonstrated methodological innovation across various contexts like logistics and supply chain management, they remain largely confined to sector-specific applications or fragmented approaches. What is still lacking is an integrated framework that can simultaneously address strategic prioritization (via FAHP), action translation (via QFD), and disclosure validation (via NLP). Our FAHP–QFD–NLP framework fills this gap by offering a multi-layer approach that directly addresses the shortcomings of these otherwise fragmented studies, thus bridging the distance between abstract sustainability intentions and verifiable corporate implementation.

Research Gap and Positioning

Although fuzzy MCDM approaches have enhanced decision-making under uncertainty, they remain largely confined to priority ranking without providing mechanisms for translating strategies into actionable plans. Similarly, hybrid frameworks such as FAHP–TOPSIS or SWARA–CoCoSo typically combine only two methodological layers but lack a validation mechanism to ensure alignment with actual corporate practices. Consequently, the literature still lacks an integrated framework that systematically links strategic prioritization (via FAHP), actionable deployment (via QFD), and disclosure validation (via NLP). By establishing this three-step chain, the present study addresses a critical research gap and contributes a novel approach that bridges abstract sustainability intentions with verifiable corporate implementation.

Research Methodology

This study develops an integrated FAHP–QFD–NLP decision-making framework to evaluate and prioritize corporate sustainability actions across the TBL dimensions: economic, social, and environmental. The methodology consists of three sequential phases: (1) determining the strategic importance of sustainability criteria using FAHP, (2) translating those criteria into actionable indicators through QFD, and (3) validating the alignment of these indicators with actual corporate practices via NLP.

Research Framework

To initiate the evaluation process, a hierarchical structure based on the TBL is developed. Figure 1 illustrates the overall flow of the integrated FAHP–QFD–NLP framework, beginning with strategic weight determination, progressing through the translation into actionable indicators, and culminating in real-world benchmarking. Figure 2 depicts the hierarchical structure that forms the foundation for FAHP, consisting of three main criteria—Economic Performance (C1), Social Value (C2), and Sustainable Environment (C3).

[See PDF for image]

Fig. 1

Research framework flowchart

[See PDF for image]

Fig. 2

Hierarchical Structure of the Decision Model

For clarity of presentation, the complete specification of criteria, sub-criteria, and indicators (H1–H27), together with their supporting references, is consolidated in Table 3. This table consolidates the full hierarchical structure into a single integrated format, providing a transparent foundation for the subsequent FAHP–QFD–NLP analysis.

Table 3. Indicators of the hierarchical structure

Goal	Criteria	Sub-criteria	Supporting references	Indicator 1	Indicator 2	Indicator 3
	Economic Performance C1 [51, 52]	Financial Returns C11	[51]	Increase market share of high-profit products (H1)	Optimize resource allocation, cut inefficient spending (H2)	Expand investment scale in high-return businesses (H3)
		Market Competitiveness C12	[52]	Increase advertising investment, enhance market brand awareness (H4)	Cooperate with major clients to expand market share (H5)	Launch differentiated product strategies (H6)
		Resource Efficiency C13	[53]	Strengthen internal management, improve production efficiency (H7)	Implement digital systems, optimize resource allocation (H8)	Streamline the supply chain, reduce costs (H9)
Achieve Synergistic Development of Economic Performance and Social Responsibility	Social Value C2 [1, 3]	Employee Satisfaction C21	[1, 3]	Improve compensation and benefits, establish incentive mechanisms (H10)	Implement flexible work models, enhance job satisfaction (H11)	Provide training opportunities, enhance employee skills (H12)
		Customer Loyalty C22	[38, 54]	Improve after-sales service levels, enhance customer experience (H13)	Launch loyalty programs (member points system), enhance customer stickiness (H14)	Establish customer service management systems (H15)
		Community Contribution C23	[1, 3]	Cooperate with local NPOs/charities, participate in community building (H16)	Conduct community activities, enhance brand reputation (H17)	Provide local employment opportunities, promote local economy (H18)
	Sustainable Environment C3 [2, 55]	Carbon Reduction C31	[2, 55]	Invest in low-carbon technologies, reduce carbon emissions in production (H19)	Implement clean energy plans, reduce carbon footprint (H20)	Use green logistics, optimize transport efficiency (H21)
		Resource Conservation C32	[2, 56]	Improve production equipment efficiency, reduce resource waste (H22)	Adopt circular economy models, achieve resource recycling and reuse (H23)	Optimize product packaging, reduce material usage (H24)
		Green Innovation C33	[2, 57]	Develop environmentally friendly products, enter green markets (H25)	Invest in R&D, promote sustainable innovation (H26)	Obtain green certifications, enhance market competitiveness (H27)

Expert judgments are collected through pairwise comparison questionnaires, where sub-criteria within the same main category are compared using linguistic scales represented as fuzzy triangular numbers. This method captures uncertainty and reduces the ambiguity in expert preferences. From these comparisons, local weights for each sub-criterion are derived and then aggregated according to the TBL hierarchy to compute global weights. These global weights reflect the strategic importance of each sub-criterion and are subsequently applied as row weights in the HOQ matrix of the QFD phase.

In this matrix, experts evaluate the strength of the relationship between each sub-criterion and the 27 sustainability action indicators, forming the basis for determining action priorities. To ensure practical alignment with real-world disclosures, the framework incorporates a mirrored benchmarking mechanism in the NLP phase, as illustrated in Fig. 1. Sustainability reports from listed companies are semantically matched against the 27 action indicators using BERT-based similarity analysis. This reflective comparison provides a practical validation of whether the prioritized actions identified through FAHP and QFD are being implemented in disclosed corporate strategies.

Questionnaire Survey

The questionnaire design was based on the Pairwise Comparison Method of AHP [58], utilizing a 1–9 ratio scale to quantify the relative importance of factors. To address uncertainty in decision-making, fuzzy linguistics were incorporated, and the questionnaire was adapted for FAHP to enhance evaluation accuracy and consistency (see Table 4 for the fuzzy linguistic scale). Practicing accountants from accounting firms were invited to complete the questionnaire via email, telephone, and face-to-face interviews to ensure its completeness and accuracy. Respondents possessed professional expertise in finance and sustainable development, enabling them to provide expert assessments of the criteria and sub-criteria in line with the TBL framework. Accountants were selected as respondents because the assurance level of sustainability reports affects their credibility, and as assurance providers, accountants play a crucial role in enhancing the transparency and reliability of CSR reporting [59]. A total of 20 questionnaires were distributed, with 19 valid responses collected. The collected fuzzy judgments were then aggregated and processed using the Buckley FAHP method described below to determine the weights of criteria and sub-criteria, which subsequently served as input for the QFD analysis. Attention was given not only to the accuracy of weight calculation but also to ensuring the questionnaire design genuinely reflected corporate decision-making needs.

Table 4. Fuzzy linguistic scale

Triangular Fuzzy Number (TFN)	Linguistic Value	Ratio Scale
(0.8,0.9,1.0)	Absolutely Important	9:1
(0.7,0.8,0.9)	Very Strongly Important	7:1
(0.6,0.7,0.8)	Strongly Important	5:1
(0.5,0.6,0.7)	Moderately Important	3:1
(0.4,0.5,0.6)	Equally Important	1:1
(0.3,0.4,0.5)	Moderately Unimportant	1:3
(0.2,0.3,0.4)	Strongly Unimportant	1:5
(0.1,0.2,0.3)	Very Strongly Unimportant	1:7
(0.1,0.1,0.1)	Absolutely Unimportant	1:9

FAHP Weight Calculation

While traditional AHP effectively handles multi-criteria decision problems, human judgment inherent in expert opinion collection often involves subjectivity and fuzziness. The crisp scale of traditional AHP struggles to fully capture this uncertainty, potentially affecting decision flexibility and result robustness. To better simulate decision-makers' fuzzy judgments, this research incorporated fuzzy set theory, adopting the geometric mean method for FAHP [35]. This method calculates fuzzy weights directly based on fuzzy pairwise comparisons. Additionally, a consistency check was performed on the aggregated judgments before proceeding with weight calculation. Additionally, a consistency check was performed on the aggregated judgments before proceeding with weight calculation.

The membership function $u_{\tilde{A}}(x)$ of a Triangular Fuzzy Number (TFN) Ã = (a, b, c) can be defined as follows:

1

$$u_{\tilde{A}}(x)=\left\{\begin{array}{cc}(x-a)/(b-a),& a\le x\le b\\ (c-x)/(c-b),& b\le x\le c\\ 0,& o.w.\end{array}\right)$$where a, b, and c represent the lower bound, middle value, and upper bound of the fuzzy number, respectively.

The detailed steps employed are as follows:

Step 1: Aggregate Expert Judgments and Construct Fuzzy Matrix

Pairwise comparison results using the fuzzy linguistic scale (see Table 3) were collected from k experts and converted into Triangular Fuzzy Numbers (TFNs) ${\tilde{a}}_{\mathit{ij}}^e=\left(l_{\mathit{ij}}^e,,,m_{\mathit{ij}}^e,,,u_{\mathit{ij}}^e\right),\forall i,j>1,e=1,2,\cdots ,t$. For each comparison (i, j), the TFN judgments from the k experts were aggregated using the geometric mean method to obtain the aggregated fuzzy judgment value ${\tilde{a}}_{\mathit{ij}}=\left(l_{ij,},m_{ij,},u_{\mathit{ij}}\right)$ as follows:

2

$$l_{\mathit{ij}}={\left(\prod _{e=1}^{\text{k}},l_{\mathit{ij}}^e\right)}^{\frac{1}{\text{t}}}$$

3

$$m_{\mathit{ij}}={\left(\prod _{e=1}^{\text{k}},m_{\mathit{ij}}^e\right)}^{\frac{1}{\text{t}}}$$

4

$$u_{\mathit{ij}}={\left(\prod _{e=1}^{\text{k}},u_{\mathit{ij}}^e\right)}^{\frac{1}{\text{t}}}$$where Π denotes the product across experts e = 1 to t. The aggregated ${\tilde{a}}_{\mathit{ij}}$ values were then used to construct the final fuzzy pairwise comparison matrix Ã = [${\tilde{a}}_{\mathit{ij}}$] shown in Eq. (5).

5

$$A=\left[{\tilde{a}}_{\mathit{ij}}\right]=\left[\begin{array}{cccc}\left(1,1,1\right)\hfill & \left(l_{12},m_{12},u_{12}\right)& \dots & \left(l_{1n},m_{1n},u_{1n}\right)\\ \left(\frac{1}{u_{12}},\frac{1}{m_{12}},\frac{1}{l_{12}}\right)\hfill & \left(1,1,1\right)& \dots & \left(l_{2n},m_{2n},u_{2n}\right)\\ ⋮\hfill & ⋮& \dots & ⋮\\ ⋮\hfill & \ddots & \dots & ⋮\\ ⋮\hfill & ⋮& \ddots & ⋮\\ \left(\frac{1}{u_{1n}},\frac{1}{m_{1n}},\frac{1}{l_{1n}}\right)\hfill & \left(\frac{1}{u_{2n}},\frac{1}{m_{2n}},\frac{1}{l_{2n}}\right)& \dots & \left(1,1,1\right)\end{array}\right]$$

Step 2: Consistency Test for Aggregated Matrix

To ensure the aggregated expert judgments possess an acceptable level of consistency, a consistency test was performed on the fuzzy pairwise comparison matrix Ã. As Buckley's FAHP method does not directly compute Saaty's consistency index, a pragmatic approach was adopted: First, the aggregated fuzzy matrix à obtained in Step 1 was defuzzified using the Centroid Method (see Eq. (6) below) to yield a crisp pairwise comparison matrix, $A_{crisp\_agg}=\left[{\mathit{DF}}_{\mathit{ij}}\right]$.

6

$${\mathit{DF}}_{\mathit{ij}}=\frac{l_{\mathit{ij}}+m_{\mathit{ij}}+u_{\mathit{ij}}}{3}$$where ${\mathit{DF}}_{\mathit{ij}}$ is the defuzzified crisp value representing the aggregated comparison of element i to element j. Next, based on this crisp matrix $A_{crisp\_agg}$, its maximum eigenvalue (${\lambda }_{\mathit{max}}$) was calculated [56]. Then, the Consistency Index (CI) was calculated as follows:

7

$$CI=\frac{{\lambda }_{\mathit{max}}-n}{n-1}$$where n is the order of the matrix. Finally, using the appropriate Random Index (RI) value based on the matrix order n (RI = 0.58 for n = 3), the Consistency Ratio (CR) was calculated as follows:

8

$$CR=\frac{\mathit{CI}}{\mathit{RI}}$$

If CR ≤ 0.1, the consistency of the aggregated expert judgments was considered acceptable, allowing continuation to the subsequent Buckley FAHP weight calculations. This test must be performed separately for the aggregated matrices at the main criteria level and each sub-criteria level (all CR values were confirmed to be ≤ 0.10 in this study).

Step 3: Calculate Fuzzy Geometric Mean Per Row

For each row (i) of the aggregated fuzzy matrix à (verified for consistency in Step 2), its fuzzy geometric mean ${\tilde{r}}_i=\left(r_i^L,,,r_i^M,,,r_i^U\right)$ was calculated as follows:

9

$$r_i^L={\left(\prod _1^n,l_{\mathit{ij}}\right)}^{\frac{1}{n}}$$

10

$$r_i^M={\left(\prod _1^n,m_{\mathit{ij}}\right)}^{\frac{1}{n}}$$

11

$$r_i^U={\left(\prod _1^n,u_{\mathit{ij}}\right)}^{\frac{1}{n}}$$where n is the number of elements being compared at that level.

Step 4: Calculate Vector Sum of Fuzzy Geometric Means

The vector sum of all fuzzy geometric means ${\tilde{r}}_i$ calculated in Step 3 was computed, denoted as ${\tilde{R}}_{\_sum}=\left(R_{\_sum}^L,,,R_{\_sum}^M,,,R_{\_sum}^U\right)\text{as follows}$:

12

$$R_{\_sum}^L=\sum _{k=1}^n{r}_k^L$$

13

$$R_{\_sum}^M=\sum _{k=1}^n{r}_k^M$$

14

$$R_{\_sum}^U=\sum _{k=1}^n{r}_k^U$$

Steps 5: Calculate Final Fuzzy Weights

Buckley's normalization method was used to calculate the final fuzzy weight ${\tilde{W}}_i=\left(w_i^L,,,w_i^M,,,w_i^U\right)$ for each element i as follows:

15

$$w_i^L=\frac{r_i^L}{R_{\_sum}^U}$$

16

$$w_i^M=\frac{r_i^M}{R_{\_sum}^M}$$

17

$$w_i^U=\frac{r_i^U}{R_{\_sum}^L}$$

(Note the divisors involve the L, M, U sums in the order U, M, L).

Step 6: Defuzzification for Crisp Weights

To facilitate ranking and provide input for the subsequent QFD analysis, the final fuzzy weights ${\tilde{W}}_i$ obtained in the previous step were defuzzified into single crisp values $w_{i\_crisp}$ using the Centroid Method as follows:

18

$$w_{i\_crisp}=\left(w_i^L,+,w_i^M,+,w_i^U\right)/3$$

To ensure the weights sum to 1, these defuzzified weights were normalized as follows:

19

$$w_{i\_final}=\frac{w_{i\_crisp}}{{\sum }_{k=1}^n{w}_{i\_crisp}}$$

This resulted in the final crisp weight $w_{i\_final}$ for each element i, and Σ denotes the sum of the crisp weights for all elements k = 1 to n.

Step 7: Hierarchy Aggregation for Global Weights

The final crisp weights ($w_{i\_final}$) calculated for each level of the hierarchy were synthesized to obtain the overall global weights for the lowest-level sub-criteria relative to the main goal. The calculation involves multiplying the final crisp local weight of a sub-criterion k (denoted $w_{k\_local}$) by the final crisp weight of the parent criterion i (denoted $w_{i\_parent}$) as follows:

20

$$W_{k\_global}=w_{k\_local}\times w_{i\_parent}$$

• $W_{k\_global}$: The global weight of sub-criterion k.

• $w_{k\_local}$: The local weight of sub-criterion k to its parent criterion i.

• $w_{i\_parent}$: The weight of the parent criterion i.

These global weights $W_{k\_global}$ reflect the final importance ranking of the sub-criteria and served as inputs for the QFD analysis.

QFD Analysis and Validation via NLP

The core objective of QFD is to translate customer requirements into executable technical specifications, ensuring that a company's products or services closely align with market demands [37]. This research combines QFD with the final crisp weights derived from the preceding FAHP analysis (specifically, the global weights $W_{k\_global}$ from Step 6 of 3.3 FAHP Weight Calculation), employing a data-driven approach to enhance the accuracy and feasibility of corporate decision-making regarding economic performance, social value, and sustainable environment. Specifically, QFD is applied to map the strategic priorities derived from FAHP to actionable indicators, which are then compared against actual implementation practices. This comparison utilizes semantic similarity analysis based on BERT and cosine similarity, covering sustainability reports from the sampled companies (2021–2023). This structured QFD model enables firms to align strategy formulation with the TBL, ensuring comprehensive alignment. The specific steps involved are detailed below.

Step 1: Define the Scope of QFD Analysis

The analysis concentrates on decision optimization within the three core dimensions: Economic Performance, Social Value, and Sustainable Environment. Specific elements selected for analysis include the WHATs, representing the 9 key sub-criteria identified under the three TBL dimensions; the HOWs, referring to the 27 proposed executable action indicators corresponding to the sub-criteria; and a benchmarking layer, consisting of 27 matched sentences extracted from the sustainability reports of the target corporations over three years (2021–2023). This benchmarking layer is used to evaluate the alignment between the proposed action indicators and real-world corporate disclosures.

Step 2: Construct the HOQ

Based on the final crisp weights derived from FAHP and the QFD methodology, the HOQ is constructed to evaluate the relevance and priority of each implementation strategy. The specific structure of this HOQ includes the following components:

(A) Requirements Layer (Left Wall): This section represents the 9 sub-criteria derived from the TBL framework. Their relative importance is captured by the final crisp global weights ($W_{k\_global}$) from the FAHP analysis (Sect. 3.3, Step 6), serving as the basis for decision evaluation.

(B) Implementation Strategy Layer (Ceiling): This layer includes 27 proposed action indicators (HOWs), each mapped to one or more sub-criteria. These indicators serve as candidate strategies designed to address the corporate sustainability goals.

(C) Relationship Matrix (Main Body): evaluates the degree of contribution of the Implementation Indicators (indexed by j = 1 ~ 27) towards achieving the Sub-criteria (indexed by i = 1 ~ 9). Nineteen experts were invited to score each (i, j) combination using a Likert 1–9 scale, representing the relationship strength or contribution degree (denoted $R_{\mathit{ij}}^{(e)}$, where e = 1, 2, …, 19 represents the expert). To synthesize the expert opinions, the Average Relationship Score (${\overline{R}}_{\mathit{ij}}$) for each (i, j) pair was first calculated as follows:

21

$${\overline{R}}_{\mathit{ij}}=\frac{{\sum }_{e=1}^{19}{R}_{\mathit{ij}}^{(e)}}{19}$$where Σ denotes the sum over experts e = 1 to 19.

(D) Benchmarking Layer (Ceiling Extension): to evaluate the alignment between proposed implementation indicators and actual corporate disclosures, a benchmarking layer is added to HOQ structure. Sustainability reports from the evaluated organizations in Taiwan across three years (2021–2023) were analyzed using BERT sentence embeddings and cosine similarity techniques. The selected companies listed in Table 5, including semiconductors, telecommunications, finance, aviation, petrochemicals, and manufacturing for ensuring the cross-sector applicability and generalizability of the proposed model. The semantic similarity between the j-th action indicator ${\overrightarrow{H}}_j$ and the k-th company's report sentence $\overrightarrow{S_{\mathit{jk}}}$ is computed as:

22

$${\text{Cosine Similarity}}_{\mathit{jk}}=\frac{{\overrightarrow{H}}_j\bullet \overrightarrow{S_{\mathit{jk}}}}{\Vert {\overrightarrow{H}}_j\Vert \bullet \Vert \overrightarrow{S_{\mathit{jk}}}\Vert }$$where: ${\overrightarrow{H}}_j$ is the BERT embedding of the j-th implementation indicator; $\overrightarrow{S_{\mathit{jk}}}$ is the embedding of the matched sentence from the k-th company's report.

Table 5. Overview of benchmarking companies (2021–2023)

No	Company name	Abbreviation	Industry sector
1	Taiwan Semiconductor Manufacturing Co	TSMC	Semiconductors
2	Advanced Semiconductor Engineering Inc	ASE	Semiconductor Packaging
3	Cathay Financial Holding Co	Cathay	Financial Services
4	EVA Airways Corp	EVA	Aviation
5	Far Eastern New Century Corp	Far Eastern	Textiles and Petrochemicals
6	Formosa Plastics Corp	FPC	Petrochemicals
7	Fubon Financial Holding Co	Fubon	Financial Services
8	MediaTek Inc	MediaTek	IC Design
9	Taiwan Cement Corp	TCC	Cement and Building Materials
10	Taiwan Power Company	Taipower	Energy and Utilities

To evaluate the alignment between each proposed action indicator and real-world corporate disclosures, we employed BERT sentence embeddings and cosine similarity to quantify semantic closeness. A high-precision threshold of 0.85 was adopted to identify strong semantic matches—stricter than the commonly used 0.80 in prior studies [60] and more conservative than the 0.75 benchmark suggested in official SBERT utilities [61]. If the cosine similarity is higher than 0.85, the sentence is considered semantically aligned with the target indicator, and the similarity score is retained. This dual-layer integration of expert-derived priorities and text-based disclosure alignment establishes a closed-loop evaluation mechanism, enhancing the model’s validity and practical relevance.

(E) Prioritization Layer (Bottom Row): Based on the final crisp global weights of the sub-criteria ($W_i$) derived from FAHP (Sect. 3.3, Step 6) and the average relationship scores (${\overline{R}}_{\mathit{ij}}$) from the relationship matrix (Eqs. (12)–(14)), the final Priority Score for each implementation indicator j was calculated as follows:

23

$$Priority{\mathit{Score}}_j=\sum _{i=1}^9\left(W_i,\times ,{\overline{R}}_{\mathit{ij}}\right)$$where $W_i$ is the final global weight of the i-th sub-criterion from FAHP, ${\overline{R}}_{\mathit{ij}}$ is the average relationship score between indicator j and sub-criterion i, and Σ denotes the sum across all 9 sub-criteria (i = 1 to 9). A higher calculated $Priority{\mathit{Score}}_j$ indicates a higher overall priority for implementation indicator j. Finally, the 27 implementation indicators were ranked based on these scores to guide strategic resource allocation. These rankings also served as the basis for further benchmarking against actual sustainability practices disclosed in the reports of the evaluated organizations over a three-year period.

Results and Discussion

FAHP Weight Calculation Results

Professional evaluations regarding corporate decisions were collected from 19 experts. Through the FAHP methodology, the relative weights for the three main dimensions—Economic Performance, Social Value, and Sustainable Environment—and their sub-criteria were calculated to assess the importance of different factors. Before presenting the final weights, the consistency of the aggregated expert judgments was verified. A consistency test was performed on the defuzzified aggregated pairwise comparison matrix for each level of the hierarchy. As shown in Table 6, the CR values were found to be less than or equal to 0.1, indicating the consistency.

Table 6. Consistency test results for FAHP matrices

Hierarchy level	Main criterion		Sub-Criteria
Matrix	[C1,C2,C3]	[C11,C12,C13]	[C21,C22,C23]	[C31,C32,C33]
${\lambda }_{\mathit{max}}$	3.0418	3.0441	3.0443	3.0485
Consistency Index (CI)	0.0209	0.0221	0.0221	0.0242
Random Index (RI)	0.58	0.58	0.58	0.58
Consistency Ratio (CR)	0.0361	0.0381	0.0382	0.0418
Result(CR ≤ 0.1)	Consistent	Consistent	Consistent	Consistent

The resulting final crisp weights for the three main criteria are shown in Table 7, while the detailed final global weights and ranking for the sub-criteria after defuzzification are presented in Table 8 and visualized in the radar chart in Fig. 3. First, Fig. 3 indicates that Green Innovation (C33) emerged as highly significant (Weight = 0.2226), indicating that companies should focus on developing environmental technologies, innovating low-carbon products, and implementing green manufacturing processes to enhance market competitiveness. Accessing additional resources through policy subsidies and green financing can make this a key component of long-term competitive advantage; Resource Conservation (C32) was identified as crucial (Weight = 0.1499). Enterprises should prioritize improving energy and water resource utilization efficiency, reducing waste, and developing circular economy models. Amid global carbon neutrality trends, resource conservation contributes to cost reduction, improved regulatory compliance, and enhanced brand competitiveness; Community Contribution (C23) is vital (Weight = 0.1320) for strengthening corporate image through community development initiatives, philanthropic activities, and local job creation. Such efforts help build social trust, foster customer loyalty, and promote long-term competitive advantage. Other sub-criteria, ranked 4th through 9th, include Carbon Reduction (C31), Resource Efficiency (C13), Customer Loyalty (C22), Market Competitiveness (C12), Employee Satisfaction (C21), and Financial Returns (C11). These factors influence internal operations, customer relations, and market competition strategies. It is recommended that while pursuing sustainable development and CSR objectives, companies simultaneously ensure robust Economic Performance and maintain market competitiveness to sustain long-term growth.

Table 7. Weights of main criteria

Main criterion	Global weight	Rank
C1	0.2241	3
C2	0.2969	2
C3	0.4790	1

Table 8. Overall weights and ranking of sub-criteria

Sub-criterion	Global weight	Rank
C11	0.0480	9
C12	0.0709	7
C13	0.1052	5
C21	0.0671	8
C22	0.0978	6
C23	0.1320	3
C31	0.1065	4
C32	0.1499	2
C33	0.2226	1

[See PDF for image]

Fig. 3

Visualization of sub-criteria global weights

The final crisp weight results obtained from the FAHP analysis (presented in Table 8) serve as crucial inputs for the QFD relationship matrix. This integration allows for a subsequent evaluation of the relevance of various strategies to corporate requirements and provides a basis for prioritizing action plans. Consequently, corporations can formulate more competitive and sustainable decision strategies that consider both expert judgments and market or stakeholder demands. Furthermore, when developing sustainability and CSR goals, ensuring internal Economic Performance and maintaining market advantages are essential for achieving stable long-term growth. Through this integrated approach, businesses can balance expert recommendations with practical requirements during the decision-making process to formulate strategies that are both more competitive and sustainable.

QFD Results Analysis

By constructing the HOQ and integrating the final crisp weights derived from the FAHP analysis, the impact of the 27 implementation strategies (HOWs) on the 9 key sub-criteria (WHATs) was explored to identify the strategies with the highest priority for execution. Based on the calculated priority score rankings, the most critical strategies for corporate implementation were identified. The final priority scores and ranks for all 27 action indicators are presented in Table 9, while the top five implementation strategies with the highest scores are visually summarized in Fig. 4. Second, in Fig. 4, the top three strategies in the priority ranking fall under the Sustainable Environment dimension (C3), particularly within the categories of Resource Conservation and Green Innovation. This finding suggests that companies should place strategic emphasis on promoting circular economy practices, enhancing resource reuse, and improving equipment efficiency to reduce waste and strengthen competitiveness through technological innovation. Additionally, optimizing green logistics and transportation emerged as another high-impact area, reinforcing the critical role of corporations in achieving sustainable development objectives. The benchmarking analysis based on cosine similarity further substantiates the practicality and robustness of the proposed FAHP–QFD–NLP model. As shown in Table 10, all 27 action indicators achieved average cosine similarity scores above 0.85 across the 10 sampled companies, with the lowest observed value being 0.8393—still indicative of high semantic similarity. This consistently strong alignment suggests that the proposed indicators are robust and widely applicable across industries, validating the generalizability of the model. The results reflect a broader shift in stakeholder and corporate focus—from traditional financial performance to broader sustainability-oriented concerns—demonstrating the evolving priorities of both corporate management and society.

Table 9. HOQ priority scores and ranking of action indicators

No	Action indicators	Priority score	Rank
H1	Increase market share of high-profit products	4.377	14
H2	Optimize resource allocation, cut inefficient spending	4.788	12
H3	Expand investment scale in high-return businesses	3.979	17
H4	Increase advertising investment, enhance market brand awareness	3.619	24
H5	Cooperate with major clients to expand market share	4.097	16
H6	Launch differentiated product strategies	4.202	15
H7	Strengthen internal management, improve production efficiency	4.808	11
H8	Implement digital systems, optimize resource allocation	4.917	10
H9	Streamline the supply chain, reduce costs	4.605	13
H10	Improve compensation and benefits, establish incentive mechanisms	3.676	21
H11	Implement flexible work models, enhance job satisfaction	3.503	27
H12	Provide training opportunities, enhance employee skills	3.669	22
H13	Improve after-sales service levels, enhance customer experience	3.613	25
H14	Launch loyalty programs, enhance customer stickiness	3.608	26
H15	Establish customer service management systems	3.782	18
H16	Cooperate with local NPOs/charities, participate in community building	3.626	23
H17	Conduct community activities, enhance brand reputation	3.710	20
H18	Provide local employment opportunities, promote local economy	3.741	19
H19	Invest in low-carbon tech, reduce production carbon emissions	5.779	9
H20	Implement clean energy plans, reduce carbon footprint	5.852	8
H21	Use green logistics, optimize transport efficiency	6.099	4
H22	Improve production equipment efficiency, reduce resource waste	6.157	2
H23	Adopt circular economy models, achieve resource recycling/reuse	6.355	1
H24	Optimize product packaging, reduce material usage	6.077	5
H25	Develop eco-friendly products, enter green markets	6.076	6
H26	Invest in R&D, promote sustainable innovation	6.142	3
H27	Obtain green certifications, enhance market competitiveness	6.046	7

[See PDF for image]

Fig. 4

Top 5 Action Indicators

Table 10. Similarity analysis by reports of 10 companies (2021–2023)

Indicator	TSMC	ASE	Cathay	EVA	Far Eastern	FPC	Fubon	MediaTek	TCC	Taipower	Average	Rank
H1	0.9260	0.9162	0.9256	0.9083	0.9170	0.9129	0.9119	0.9160	0.9136	0.9147	0.9162	6
H2	0.9192	0.9170	0.9113	0.9121	0.9048	0.9141	0.9148	0.9187	0.9159	0.9175	0.9145	9
H3	0.9077	0.8855	0.8894	0.8871	0.8860	0.8839	0.8863	0.8892	0.8850	0.8858	0.8886	16
H4	0.8517	0.8518	0.8556	0.8460	0.8502	0.8501	0.8494	0.8529	0.8511	0.8514	0.8510	26
H5	0.9072	0.8971	0.9016	0.9022	0.8980	0.8994	0.8979	0.8977	0.8982	0.8962	0.8995	11
H6	0.8817	0.8785	0.8882	0.8827	0.8811	0.8802	0.8794	0.8806	0.8781	0.8805	0.8811	19
H7	0.9018	0.8783	0.8480	0.8836	0.8781	0.8787	0.8789	0.8807	0.8776	0.8774	0.8783	20
H8	0.9180	0.9147	0.9151	0.9115	0.9164	0.9145	0.9135	0.9171	0.9156	0.9169	0.9153	8
H9	0.8901	0.8799	0.8848	0.8819	0.8775	0.8809	0.8829	0.8800	0.8805	0.8800	0.8818	18
H10	0.9236	0.8858	0.8881	0.8850	0.8875	0.8847	0.8857	0.8888	0.8847	0.8860	0.8900	15
H11	0.8621	0.8513	0.8562	0.8499	0.8501	0.8531	0.8526	0.8537	0.8514	0.8532	0.8534	25
H12	0.9411	0.9282	0.9327	0.9247	0.9279	0.9257	0.9261	0.9285	0.9264	0.9275	0.9289	4
H13	0.8824	0.8696	0.8741	0.8677	0.8702	0.8690	0.8704	0.8713	0.8696	0.8702	0.8714	22
H14	0.8518	0.8393	0.8469	0.8442	0.8398	0.8418	0.8397	0.8433	0.8406	0.8418	0.8429	27
H15	0.8721	0.8633	0.8660	0.8642	0.8608	0.8632	0.8639	0.8668	0.8636	0.8657	0.8650	24
H16	0.8835	0.8978	0.8979	0.8988	0.8950	0.8969	0.8960	0.9003	0.8962	0.8975	0.8960	12
H17	0.8545	0.8675	0.8698	0.8688	0.8678	0.8660	0.8651	0.8678	0.8668	0.8673	0.8661	23
H18	0.8992	0.8917	0.8951	0.8930	0.8915	0.8893	0.8892	0.8931	0.8918	0.8931	0.8927	14
H19	0.9646	0.9470	0.9461	0.9449	0.9451	0.9465	0.9461	0.9473	0.9463	0.9478	0.9482	1
H20	0.9442	0.9329	0.9324	0.9295	0.9316	0.9336	0.9319	0.9327	0.9315	0.9331	0.9333	3
H21	0.9039	0.8867	0.8890	0.8860	0.8858	0.8873	0.8868	0.8856	0.8863	0.8871	0.8885	17
H22	0.9250	0.8918	0.8903	0.8958	0.8912	0.8906	0.8910	0.8926	0.8924	0.8913	0.8952	13
H23	0.9260	0.9139	0.9142	0.9133	0.9149	0.9157	0.9144	0.9141	0.9145	0.9152	0.9156	7
H24	0.9001	0.8722	0.8711	0.8691	0.8736	0.8738	0.8727	0.8736	0.8730	0.8720	0.8751	21
H25	0.9068	0.9070	0.9041	0.9044	0.9053	0.9071	0.9059	0.9069	0.9065	0.9068	0.9061	10
H26	0.9339	0.9347	0.9341	0.9299	0.9335	0.9328	0.9334	0.9345	0.9340	0.9339	0.9335	2
H27	0.9217	0.9218	0.9170	0.9221	0.9234	0.9215	0.9217	0.9234	0.9215	0.9228	0.9217	5
Average	0.9037	0.8934	0.8943	0.8928	0.8927	0.8931	0.8929	0.8947	0.8931	0.8938
Rank	1	5	3	9	10	6	8	2	7	4

Third, Fig. 5 illustrates the top five action indicators with the highest cosine similarity over three years. These indicators exhibit not only semantic stability but also conceptual consistency, reinforcing their strategic importance across industries and time. Notably, H19 (low-carbon technology), H26 (R&D investment), and H20 (clean energy usage) consistently ranked highest, underscoring a cross-sectoral emphasis on innovation and environmental responsibility. Finally, Fig. 6 presents the average cosine similarity scores for each of the 10 participating companies. TSMC retained the top rank, with consistently high alignment scores, followed closely by MediaTek and Cathay Financial. These results highlight sectoral differences in sustainability reporting maturity and reveal that leading firms in the semiconductor and financial industries demonstrate stronger alignment with structured sustainability action models. The cross-industry sample affirms the adaptability and universality of the proposed framework, supporting its application across diverse corporate contexts.

[See PDF for image]

Fig. 5

Top 5 Action Indicators with Highest Cosine Similarity

[See PDF for image]

Fig. 6

Listed Companies by Average Cosine Similarity Action Indicators

Comparative Discussion with Advanced Fuzzy Models

While prior studies have developed advanced fuzzy aggregation models—such as the T-spherical fuzzy Dubois–Prade approach and Cubic Picture Fuzzy Fair Aggregation Operators—to improve information representation and robustness under uncertainty, these frameworks remain focused on enhancing aggregation operators. By contrast, the proposed FAHP–QFD–NLP framework extends beyond aggregation, offering a three-layer integration that links strategic prioritization (FAHP), action translation (QFD), and disclosure validation (NLP). In this way, the present study complements existing fuzzy extensions by addressing gaps in strategy operationalization and empirical benchmarking, thereby ensuring that sustainability intentions are translated into verifiable corporate practices. This comparative positioning further demonstrates the novelty and added value of our contribution.

Conclusion

This study emphasizes both theoretical and practical implications by integrating FAHP, QFD, and NLP into a comprehensive sustainability evaluation framework. By situating sustainability as a multidimensional objective—encompassing economic, social, and environmental performance—the model demonstrates how strategic priorities can be systematically translated into actionable indicators and validated against real-world corporate disclosures. This dual focus highlights the framework’s value as both an academic contribution to decision science and a practical tool for corporate governance, enabling organizations to align internal strategies with external expectations in a transparent and evidence-based manner.

The empirical findings reveal that expert evaluations placed the highest importance on Green Innovation (C33), Resource Conservation (C32), and Community Contribution (C23), underscoring the prominence of innovation, efficiency, and stakeholder engagement in sustainable value creation. The QFD phase identified five key implementation indicators with the greatest execution urgency, including adopting circular economy models (H23) and improving production equipment efficiency (H22). Benchmarking through BERT-based similarity further confirmed the framework’s robustness, as all 27 indicators exhibited high semantic alignment with corporate sustainability disclosures, with cosine scores consistently above 0.85. These results validate that the framework not only prioritizes relevant strategies but also reflects practices currently emphasized by leading firms.

In addition to these findings, this study makes distinct methodological and practical contributions. Methodologically, it advances decision science by integrating FAHP, QFD, and NLP into a unified three-layer framework that links strategic prioritization, action translation, and disclosure validation—an integration not addressed in prior fuzzy or hybrid MCDM studies. Practically, it provides corporations with a transparent and actionable tool to prioritize sustainability strategies, allocate resources, and align ESG reporting with strategic execution. This dual contribution clarifies the study’s novelty by extending fuzzy–QFD hybrids beyond product development toward governance-level sustainability planning.

Despite its contributions, this study has several limitations. First, the FAHP weighting process relies on expert judgment, which may introduce subjectivity despite the use of fuzzy logic to mitigate inconsistencies. Second, the case analysis is limited to ten publicly listed Taiwanese firms, restricting the generalizability of the findings to broader international or SME contexts. Third, the constructed HOQ is based on static inputs and does not capture temporal changes in sustainability practices. Finally, the FAHP phase assumes independence among criteria, whereas real-world sustainability dimensions often interact in complex and dynamic ways.

Future research can address these limitations by applying the FAHP–QFD–NLP framework to different industries and cross-national settings to enhance generalizability. Methodologically, integrating emerging approaches such as circular intuitionistic fuzzy sets or machine learning–driven weighting schemes could further improve robustness and reduce subjectivity. In addition, extending the framework to dynamic sustainability evaluation—through longitudinal data or system-based modeling—would allow continuous adaptation to evolving corporate and regulatory priorities. These directions would ensure that the proposed framework remains flexible, scalable, and relevant to future sustainability challenges.

Author Contributions

Yin-Yin Huang: conceptualization, validation, resources, data curation, writing—original draft, writing—review & editing. Tsai-Sung Lin: conceptualization, methodology, software, formal analysis, writing—original draft. Jingchao Pan: software, validation, formal analysis, resources, data curation. Minh T.N. Nguyen: methodology, software, validation, formal analysis, investigation. Ruey-Chyn Tsaur: conceptualization, validation, resources, data curation, writing—original draft, writing—review & editing.

Funding

The authors declare that no funding was received from any organization or agency in support of this research.

Data availability

Data is provided within the manuscript.

Declarations

Conflict of interest

The authors declare that there is no conflict of interest.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Carroll, AB. The pyramid of corporate social responsibility: toward the moral management of organizational stakeholders. Bus. Horiz.; 1991; 34, 4 pp. 39-48. [DOI: https://dx.doi.org/10.1016/0007-6813(91)90005-G]

2. Elkington, J. Cannibals with forks: the triple bottom line of 21st century business. Alternat J; 1998; 25, 4 42.

3. Freeman, R. E.: Strategic management: a stakeholder approach. Pitman. (1984)

4. Freeman, RE; Harrison, JS; Wicks, AC. Managing for stakeholders: Survival, reputation, and success; 2004; Yale University Press:

5. United Nations: Transforming our world: The 2030 agenda for sustainable development. United Nations. (2015). https://sdgs.un.org/2030agenda

6. European Commission: The carbon border adjustment mechanism: a tool to prevent carbon leakage. European Commission. (2021)

7. Dai, R; Wang, W. The impact of ESG rating disagreement on corporate risk-taking: evidence from China. Dig. Econ. Sustain. Develop.; 2024; [DOI: https://dx.doi.org/10.1007/s44201-024-00060-0]

8. PricewaterhouseCoopers: 25th Annual Global CEO Survey: ESG and the path to long-term value. PwC. (2022). https://www.pwc.com/gx/en/ceo-survey/2022/main-report.html

9. Dyllick, T; Muff, K. Clarifying the meaning of sustainable business: introducing a typology from business-as-usual to true business sustainability. Organ. Environ.; 2016; 29, 2 pp. 156-174. [DOI: https://dx.doi.org/10.1177/1086026615575179]

10. Gillan, SL; Koch, A; Starks, LT. Firms and social responsibility: a review of ESG and CSR research in corporate finance. J. Corp. Finan.; 2021; 66, [DOI: https://dx.doi.org/10.1016/j.jcorpfin.2020.101889] 101889.

11. Harun, MH; Bakar, AB; Abdullah, N. Semiparametric option-implied information and median-variance approach: a game-changer in integrating sustainability into financial decision-making. Financ. Res. Lett.; 2024; 60, [DOI: https://dx.doi.org/10.1016/j.frl.2023.104852] 104852.

12. Purcarea, IM; Cociuba, C-V. ESG reporting and corporate governance: trends and challenges. Corporate Govern. Rev.; 2024; 22, 1 pp. 50-64.

13. Bálint, J; Várallyai, L; Botos, B. Evaluation of data-driven sustainability potential at SMEs using an altered Ecocanvas model. Sustainability; 2024; 16, 3 1245. [DOI: https://dx.doi.org/10.3390/su16031245]

14. Mi, J. Esg performance and enterprise competitive advantage construction: a case study of Walmart. J. Bus. Ethics; 2024; 189, 2 pp. 345-368. [DOI: https://dx.doi.org/10.1007/s10551-022-05315-5]

15. Abramova, E. Environmental, social, and governance (ESG) principles in the banking sector development: integration and challenges. Bank. Fin. Rev.; 2024; 39, 1 pp. 89-106.

16. Zadeh, LA. Fuzzy sets. Inf. Control.; 1965; 8, 3 pp. 338-353. [DOI: https://dx.doi.org/10.1016/S0019-9958(65)90241-X]

17. Alodat, T; Hao, C; Nobanee, H. How sustainability committees moderate the link between ESG performance and environmental innovation. J. Bus. Strat.; 2024; 45, 1 pp. 204-220. [DOI: https://dx.doi.org/10.1108/JBS-05-2023-0096]

18. Ragazou, M; Lemonakis, C; Passas, I. ESG-driven ecopreneur selection in European financial institutions: entropy and TOPSIS analysis. Fin. Sustain. J.; 2024; 12, 4 pp. 765-789.

19. Fernandes, R; Pegino, J. Quality function deployment in sustainable supply chain management: a systematic review. Sustain. Supply Chain J.; 2024; 9, 1 pp. 45-67.

20. Giacotto, F; Marques, S; Martinetti, A. Prescriptive maintenance: a comprehensive review of current research and future directions. J. Manuf. Syst.; 2023; 71, pp. 789-812. [DOI: https://dx.doi.org/10.1016/j.jmsy.2023.10.009]

21. Khatib, F. Integrating QFD and sustainability for corporate strategy development. Sustain. Res. J.; 2024; 18, 1 pp. 145-160.

22. Kamvysi, K; Tsironis, L; Gotzamani, K. Prioritizing sustainable strategies in corporate decision-making using FAHP-QFD. Eur. J. Oper. Res.; 2023; 308, 1 pp. 108-122. [DOI: https://dx.doi.org/10.1016/j.ejor.2022.12.015]

23. Freeman, RE; Harrison, JS; Zyglidopoulos, SC. Stakeholder theory: concepts and strategies; 2018; Cambridge, Cambridge University Press: [DOI: https://dx.doi.org/10.1017/9781108539500]

24. World Commission on Environment and Development. Our common future; 1987; Oxford, Oxford University Press:

25. Becchetti, L; Bobbio, E; Prizia, F; Semplici, L. Going deeper into the S of ESG: a relational approach to the definition of social responsibility. Sustainability; 2022; 14, 19 12668. [DOI: https://dx.doi.org/10.3390/su141912668]

26. Baratta, A; Cimino, A; Longo, F; Solina, V; Verteramo, S. The impact of ESG practices in industry with a focus on carbon emissions: insights and future perspectives. Sustainability; 2023; 15, 8 6685. [DOI: https://dx.doi.org/10.3390/su15086685]

27. Park, SR; Jang, JY. The impact of ESG management on investment decision: institutional investors' perceptions of country-specific ESG criteria. Int. J. Financ. Stud.; 2021; 9, 3 48. [DOI: https://dx.doi.org/10.3390/ijfs9030048]

28. Elkington, J: 25 years ago I coined the phrase ‘triple bottom line.’ Here’s why it’s time to rethink it. Harvard Business Review. (2018). https://hbr.org/2018/06/25-years-ago-i-coined-the-phrase-triple-bottom-line-heres-why-im-rethinking-it

29. Slaper, T. F., & Hall, T. J.: The triple bottom line: What is it and how does it work? Indiana Business Review, 86(1), 4–8. (2011). https://www.ibrc.indiana.edu/ibr/2011/spring/article2.html

30. Mishra, P; Pandey, VK. Triple bottom line and environmental sustainability: evolution of global ESG research—a bibliometric analysis. Environ. Sci. Eur.; 2025; 37, 136. [DOI: https://dx.doi.org/10.1186/s12302-025-01184-9]

31. IBM Institute for Business Value: ESG and the triple bottom line: aligning business strategies with sustainability goals. IBM. (2023). https://www.ibm.com/thought-leadership/institute-business-value/en-us/report/esg-triple-bottom-line

32. Chipimo, ML; Bwalya, J; Kanyanga, JK. The role of ESG in driving firm profitability: implications for stakeholder, resource-based view, and triple bottom line theories in emerging markets. SEISENSE J. Manag.; 2025; 8, pp. 119-131. [DOI: https://dx.doi.org/10.33215/m1c78498]

33. Amel-Zadeh, A; Serafeim, G. Why and how investors use ESG information: evidence from a global survey. Financ. Anal. J.; 2018; 74, 3 pp. 87-103. [DOI: https://dx.doi.org/10.2469/faj.v74.n3.2]

34. Montiel, I. Corporate social responsibility and corporate sustainability. Organ. Environ.; 2008; 21, 3 pp. 245-269. [DOI: https://dx.doi.org/10.1177/1086026608321821]

35. Buckley, JJ. Fuzzy hierarchical analysis. Fuzzy Sets Syst.; 1985; 17, 3 pp. 233-247.819361 [DOI: https://dx.doi.org/10.1016/0165-0114(85)90040-X]

36. Chang, DY. Applications of the extent analysis method on fuzzy AHP. Eur. J. Oper. Res.; 1996; 95, 3 pp. 649-655. [DOI: https://dx.doi.org/10.1016/0377-2217(95)00300-2]

37. Govindan, K; Shaw, M; Majumdar, A. Social sustainability tensions in multi-tier supply chain: a fuzzy AHP approach. Int. J. Prod. Econ.; 2021; 239, [DOI: https://dx.doi.org/10.1016/j.ijpe.2021.108192] 108192.

38. Akao, Y. Quality function deployment: Integrating customer requirements into product design; 1990; New York, Productivity Press:

39. Chan, LK; Wu, ML. Quality function deployment: a literature review. Eur. J. Oper. Res.; 2002; 143, 3 pp. 463-497. [DOI: https://dx.doi.org/10.1016/S0377-2217(02)00168-9]

40. Fargnoli, M; Costantino, F; Di Gravio, G; Tronci, M. Product service-systems implementation: a customized framework to enhance sustainability and customer satisfaction. J. Clean. Prod.; 2020; 257, [DOI: https://dx.doi.org/10.1016/j.jclepro.2020.120466] 120466.

41. Jagannath, AR; Bhola, SS. Quality function deployment (QFD) for development and innovative business practices: a systematic review. Seybold Rep.; 2020; 15, 8 pp. 1-19.

42. Bottani, E; Rizzi, A. An adapted multi-criteria approach to supplier selection based on quality function deployment. Int. J. Prod. Econ.; 2008; 111, 2 pp. 763-781. [DOI: https://dx.doi.org/10.1016/j.ijpe.2007.02.038]

43. Kahraman, C; Cebeci, U; Ruan, D. Multi-criteria decision-making using fuzzy AHP. Fuzzy Syst. Knowl. Discov.; 2014; 8, 4 pp. 456-469.

44. Haber, N; Fargnoli, M; Sakao, T. Integrating QFD for product-service systems with the Kano model and fuzzy AHP. Total Qual. Manag. Bus. Excell.; 2020; 31, 9–10 pp. 929-954. [DOI: https://dx.doi.org/10.1080/14783363.2018.1462925]

45. Akbaş, H; Bilgen, B. An integrated fuzzy QFD and TOPSIS methodology for choosing the ideal gas fuel at WWTPs. Energy; 2017; 125, pp. 484-497. [DOI: https://dx.doi.org/10.1016/j.energy.2017.03.010]

46. Ocampo, LA; Jumao-as, AMB; Labrador, JJT; Rama, AMO. On a generic framework for sustainable product design: an application to a food product. J. Agric. Technol. Manag.; 2016; 19, 1 pp. 9-21.

47. Fetanat, A; Tayebi, M. Sustainable design of the household water treatment systems using a novel integrated fuzzy QFD and LINMAP approach: a case study of Iran. Environ. Dev. Sustain.; 2021; 23, 10 pp. 15031-15061. [DOI: https://dx.doi.org/10.1007/s10668-021-01286-8]

48. Liu, Z; Chen, X; Liang, X. Jiyu FAHP he FQFD de chengyongche cheneikongjian ke kuozhan sheji [Extensible design of passenger vehicle interior space based on FAHP and FQFD]. PLoS ONE; 2024; 19, 1 [DOI: https://dx.doi.org/10.1371/journal.pone.0296309] e0296309.

49. Darzi, MA; Ahmadi, S; Shokouhyar, S. Cross-cultural ranking of cellphone life cycle issues for strategic CSR marketing with social media-based AHP. Environ. Dev. Sustain.; 2025; 27, pp. 3201-3232. [DOI: https://dx.doi.org/10.1007/s10668-023-04009-y]

50. Gutierrez-Bustamante, M; Espinosa-Leal, L. Natural language processing methods for scoring sustainability reports—a study of Nordic listed companies. Sustainability; 2022; [DOI: https://dx.doi.org/10.3390/su14159165]

51. Jensen, MC. Value maximization, stakeholder theory, and the corporate objective function. J. Appl. Corp. Financ.; 2010; 22, 1 pp. 32-42. [DOI: https://dx.doi.org/10.1111/j.1745-6622.2010.00259.x]

52. Porter, ME. Competitive strategy: techniques for analyzing industries and competitors: with a new introduction; 1998; Free Press:

53. Heskett, JL; Sasser, WE; Schlesinger, LA. The service profit chain: How leading companies link profit and growth to loyalty, satisfaction, and value; 1997; Free Press:

54. Oliver, RL. Whence consumer loyalty?. J. Mark.; 1999; 63, 4_suppl1 pp. 33-44. [DOI: https://dx.doi.org/10.1177/00222429990634s105]

55. Hart, SL. A natural-resource-based view of the firm. Acad. Manage. Rev.; 1995; 20, 4 pp. 986-1014. [DOI: https://dx.doi.org/10.2307/258963]

56. Ellen MacArthur Foundation: Circular economy and climate change. (2022). Retrieved from https://ellenmacarthurfoundation.org

57. Kahraman, C; Cebeci, U; Ulukan, Z. Multi-criteria supplier selection using fuzzy AHP. Logist. Inf. Manag.; 2003; 16, 6 pp. 382-394. [DOI: https://dx.doi.org/10.1108/09576050310503367]

58. Saaty, TL. The analytic hierarchy process: planning, priority setting, resource allocation; 1980; McGraw-Hill:

59. Tu, R.-R., Hsu, W.-H., & Deng, Y.-S: Qiye shehui zeren baogaoshu zhi quexin fuwu tigongzhe, quexin chengdu ji chanye zhuanjia yu kexindu [Assurance providers, assurance levels, and industry experts in the credibility of corporate social responsibility reports]. Zhonghua Kuaiji Xuekan [Chinese Accounting Review], 4(4), 503–536. (2017). https://www.airitilibrary.com/Publication/alDetailedMesh?DocID=P20140110001-201704-201705170020-201705170020-503-536

60. Reimers, N., & Gurevych, I: Sentence-BERT: sentence embeddings using siamese BERT-networks (No. arXiv:1908.10084). arXiv. (2019). https://doi.org/10.48550/arXiv.1908.10084

61. SBERT.net: SBERT Utility Functions. Retrieved June 17, 2025, (2023). from https://www.sbert.net/docs/package_reference/util.html

62. Lin, C-L; Hsu, C-Y; Ting, C-H. Research on the key influencing goals for visual design sustainability: a dual perspective. Sustainability; 2024; 16, 1885. [DOI: https://dx.doi.org/10.3390/su16051885]

63. Schimanski, T; Reding, A; Reding, N et al. Bridging the gap in ESG measurement: using NLP to quantify environmental, social, and governance communication. Financ. Res. Lett.; 2024; 61, [DOI: https://dx.doi.org/10.1016/j.frl.2024.104979] 104979.

64. Athar Farid, HM; Dabic-Miletic, S; Jameel, T et al. Promoting sustainable logistics in the electronics industry: circular intuitionistic fuzzy framework for evaluating smart robotics technologies. Expert Syst. Appl.; 2025; 287, [DOI: https://dx.doi.org/10.1016/j.eswa.2025.128031] 128031.

65. Riaz, M; Kausar, R; Jameel, T; Pamucar, D. Cubic picture fuzzy topological data analysis with integrating blockchain and the metaverse for uncertain supply chain management. Eng. Appl. Artif. Intell.; 2024; 131, [DOI: https://dx.doi.org/10.1016/j.engappai.2023.107827] 107827.

66. Ghoushchi, SJ; Jalalat, SM; Bonab, SR et al. Evaluation of wind turbine failure modes using the developed SWARA-CoCoSo methods based on the spherical fuzzy environment. IEEE Access; 2022; 10, pp. 86750-86764. [DOI: https://dx.doi.org/10.1109/ACCESS.2022.3199359]