Sustainable Marketing Performance and Responsive Market Orientation of Enterprises in the Context of Digital Transformation: A Case Study of the Green Consumer-Goods Industry

Abstract

This study examines how digital transformation (DT) affects sustainable marketing performance (SMP) in the green consumer-goods sector, focusing on new energy vehicle (NEV) enterprises. It explores the mediating role of responsive market orientation (RMO) in this relationship. A structural path model integrating DT, RMO, and SMP is developed to analyze the impact of digital-technology adoption and market responsiveness to green marketing outcomes. This research conducts an empirical analysis using structural equation modeling based on the data of 86 Chinese A-share listed new-energy companies from 2018 to 2022. The results indicate that DT directly improves SMP and also indirectly enhances it by reinforcing RMO. RMO is found to play a significant mediating role. This study may contribute to the theoretical and methodological understanding of how digital strategies drive marketing performance in the context of green consumption and offers empirical support for advancing corporate green marketing practices.

1. Introduction

In recent years, China’s “Digital China” strategy and “Dual Carbon” goals have significantly reshaped the green consumer-goods industry. This transformation is most notable in the new energy vehicle (NEV) sector [1,2]. The “Dual Carbon” policy expands the NEV market by tightening carbon emission standards and encouraging green transportation, with digital transformation offering unprecedented opportunities for sustainable business-model innovation across consumer-goods sectors. Meanwhile, the “Digital China” initiative supports enterprise transformation through enhanced digital infrastructure, data integration, and intelligent manufacturing systems [3,4]. This policy synergy positions the NEV sector as one of the most technologically advanced areas among green industries. It provides a distinctive context for exploring corporate green marketing strategies amid digital transformation (DT), as sustainable marketing research emphasizes the need to accelerate transformations through purposeful actions that create combined value for companies, consumers, and society [5,6].

Although existing research has explored the impact of DT on corporate performance from various perspectives, three key gaps remain: first, there is limited research on how digitalization contributes to sustainable marketing performance (SMP). Current studies primarily focus on DT’s role in improving operational efficiency, innovation capacity, or financial outcomes through digital technology-business alignment mechanisms [7,8], while its influence on green brand development and consumer trust in sustainability remains underexplored. Second, empirical research on responsive market orientation (RMO) is still emerging. While RMO is often discussed conceptually or through case studies, few studies have integrated RMO into empirical models as a mediating variable between DT and SMP, despite evidence that supply-chain digital innovation policies can significantly improve sustainable-development performance [9,10,11,12]. Although some scholars hint at a link between RMO and green performance, the underlying mechanisms and pathways remain unclear—especially in complex digital environments. Third, a clear distinction between RMO and traditional market orientation is often missing. This leads to conceptual ambiguity and variable overlap. Traditional market orientation focuses on customer orientation, competitor orientation, and interdepartmental coordination, primarily emphasizing information gathering and cognitive processes. In contrast, RMO stresses a feedback loop of perception, judgment, and action, offering greater adaptability in dynamic and uncertain markets.

To address these gaps, this study proposes that RMO plays a central mediating role in the relationship between DT and SMP. A three-variable path model—DT → RMO → SMP—is constructed. This model draws on three theoretical frameworks to explain green transformation mechanisms in organizations. The Resource-Based View (RBV) views DT as a strategic resource that enhances integration efficiency [13,14]. Dynamic capabilities theory highlights the need for organizations to sense, learn, and adapt continuously in fast-changing green markets [15,16]. VBN theory, originally developed to explain individual environmental behavior, is extended here to the organizational level. It informs a five-dimensional RMO measurement system including perceived responsiveness, value orientation, belief dissemination, behavioral adjustment, and institutional enforcement, which align with green integrated marketing communication frameworks that enhance consumer awareness and purchase intentions [17,18].

Although combining these three theories introduces complexity, their roles are conceptually distinct: RBV provides the capability foundation, dynamic capabilities describe the adaptive process, and VBN represents the embedded green values. Together, they support RMO as a key mediating mechanism in the DT–SMP relationship [19].

Empirically, this study analyzes data from 86 A-share listed NEV companies between 2018 and 2022, using structural equation modeling (SEM) to test the proposed relationships. DT is measured using keyword frequencies in annual reports. RMO is evaluated using the five VBN-derived dimensions. SMP is assessed across four Environmental, Social, and Governance (ESG)-aligned indicators: economic performance, risk control, compliance, and growth sustainability. The findings confirm that DT enhances SMP both directly and indirectly via RMO. These insights offer theoretical and methodological guidance for integrating digital and green strategies in the green consumer goods sector.

2. Research Hypotheses and Variable Analysis

2.1. Proposal of Research Hypotheses

To systematically reveal the impact mechanism of DT on the SMP of NEV enterprises, this study constructs a structural path model involving three core variables based on industry-specific characteristics and relevant theoretical frameworks. The level of corporate DT is set as the independent variable, the RMO as the mediating variable, and SMP as the dependent variable. The goal is to examine how enterprises enhance organizational responsiveness through technological empowerment to achieve sustainable growth in marketing performance amid rising green consumption demands.

H1: 

The DT level has a significant positive impact on the SMP of enterprises.

In recent years, digital technologies such as artificial intelligence (AI), big data, cloud computing, and blockchain have been widely applied in business operations of NEV enterprises, including user data collection, product iteration, intelligent network control, and supply chain management [20,21]. These applications not only enhance operational efficiency but also strengthen the ability of enterprises to communicate sustainable value under green consumption trends. For example, energy optimization and carbon footprint visualization help improve green brand image [22,23]. Therefore, enterprises with stronger digital capabilities are more likely to achieve synergistic growth in both long-term performance and environmental responsibility.

H2: 

The DT level has a significant positive impact on an enterprise’s RMO.

Amid the rapid development of the NEV industry, enterprises must continuously monitor changes in green consumer preferences, policy adjustments, and public opinion trends. DT enables access to more detailed customer behavior data, faster feedback mechanisms, and more flexible strategy adjustment tools [24,25]. For instance, some enterprises utilize cloud platforms to process user charging behavior data in real time, as digital transformation enables organizations to leverage digital technologies for enhanced operational efficiency and strategic decision-making across multiple levels of analysis, thereby dynamically optimizing market layouts and marketing strategies [26,27]. Hence, the deep integration of digital technology helps activate enterprises’ perception–judgment–response mechanisms, enhancing their market responsiveness and adaptability.

H3: 

Responsive market orientation has a significant positive impact on the SMP of enterprises.

RMO reflects an enterprise’s sensitivity and responsiveness to market fluctuations. In the context of the NEV industry, consumer decisions increasingly rely on recognition of green values and transparency across the product lifecycle [28,29]. Enterprises that can identify and respond to these evolving preferences are more likely to earn consumer trust, boost brand loyalty, and improve marketing performance [30]. RMO is not only a cognitive capability but also a strategic execution mechanism driven by values, integrating belief-based orientation with institutional implementation [31,32].

H4: 

RMO partially mediates the relationship between DT and SMP.

Although DT provides a technological foundation for performance enhancement, its actual impact depends on whether enterprises have internal mechanisms to convert technological advantages into market adaptability, particularly in transforming green supply chains to achieve carbon neutrality [33,34]. RMO serves as a key channel in the transformation from “capability” to “outcome” [35]. For example, even with strong data-processing abilities, enterprises may fail to continuously enhance marketing performance if they lack adjustment mechanisms driven by consumers’ green preferences. Therefore, RMO is likely to play a partial mediating role in the effect of DT on SMP, forming an organizational response cycle of cognition–belief–behavior.

Although this study does not empirically introduce moderating variables into the model, it theoretically recognizes that enterprise scale, market maturity, and policy environment may influence the strength of the relationships among DT, RMO, and SMP. For instance, larger enterprises may possess stronger capabilities in resource allocation and technological transformation. In regions with high levels of policy subsidies, firms are more likely to activate green strategic awareness. Similarly, market maturity may affect the heterogeneity of consumer preferences regarding green products. These contextual variables offer potential directions for future research, particularly through multi-group analysis or interaction modeling, to enhance the model’s generalizability and explanatory power. This study proposes a three-variable mediating pathway—“DT → RMO → SMP”—which provides theoretical support for investigating the mechanisms linking digital strategy to marketing performance. Future research could refine this framework by decomposing RMO into a three-stage structure of “cognition–behavior–institution,” incorporating time-series data to examine the dynamic evolution of organizational responses, or adopting multilevel models to explore variations across different organizational strata. In the theoretical analysis of Hypothesis H2, this study also posits the potential existence of nonlinear relationships—such as diminishing marginal effects or stepwise improvements—in the link between DT and RMO. However, given the primary objective of identifying mechanism pathways and verifying their statistical significance, a linear SEM framework is employed to ensure model identifiability and estimation robustness [32]. These limitations are further discussed in the concluding section, along with suggestions for future studies to consider nonlinear or threshold-based modeling approaches.

To clarify the influence of DT on SMP, this study distinguishes between its direct and indirect effects. Within the model, DT exerts both a direct impact on SMP and an indirect effect mediated through RMO. Specifically, DT enhances a firm’s digital capabilities, thereby improving its market responsiveness. This heightened responsiveness enables better strategic execution and market behavior, ultimately enhancing SMP. As to why RMO serves as a partial rather than a full mediator, the study argues that while DT significantly improves RMO, RMO is not the sole channel through which DT influences SMP. Digital transformation also directly contributes to marketing performance, particularly when firms utilize digital technologies to enhance operational efficiency and strengthen brand image. Therefore, RMO is positioned as a partial mediator in the DT–SMP relationship. This assertion is grounded in a theoretical understanding of the complex interactive mechanisms between digital transformation and green marketing transformation.

2.2. Model Structure and Path Framework

To verify the theoretical pathway through which DT affects the SMP of NEV enterprises via RMO, this study constructs a SEM framework incorporating three core variables. The SEM framework is composed of two parts: a measurement model and a structural model [32,34].

Measurement model:

$$x=\mathsf{\Lambda }\xi +\delta$$

Structural model:

$$\eta =B\eta +\mathsf{\Gamma }\xi +\zeta$$

In these equations, $x$ denotes the vector of observed variables; $\xi$ and $\eta$ represent exogenous and endogenous latent variables, respectively; $\mathsf{\Lambda }$ is the factor loading matrix; $\delta$ refers to the measurement error; $B$ denotes the structural path coefficient matrix between endogenous variables; $\mathsf{\Gamma }$ represents the path coefficient matrix from exogenous to endogenous variables; and $\zeta$ is the structural residual.

In this model, DT functions as the independent variable, RMO as the mediating variable, and SMP as the dependent variable. The overall structure includes three direct paths and one mediating path. Departing from traditional unidirectional hypotheses such as “DT → SMP,” this study highlights the intermediary role of organizational response mechanisms. It emphasizes how firms convert technological strategies into actual performance through market sensing, value-driven guidance, and adaptive behavior. The model includes the following specific paths:

Path 1: DT → SMP (direct path)

This path captures the direct impact of an enterprise’s digital technology adoption on marketing performance. For instance, AI-powered personalized recommendations for green products can directly increase purchase rates and customer satisfaction.

Path 2: DT → RMO (direct path)

This path reflects how digital capabilities enhance an enterprise’s ability to perceive, interpret, and respond to market changes. For example, cloud-based user behavior analytics can improve responsiveness to evolving green consumer preferences.

Path 3: RMO → SMP (direct path)

This path demonstrates how enhanced market responsiveness leads to improved performance in green markets. RMO enables firms to convert external green value signals into internal strategic actions—such as green brand communication or supply chain adjustments—thereby fostering competitive advantage.

Path 4: DT → RMO → SMP (mediating path)

This path indicates that DT indirectly boosts SMP by enhancing RMO. Based on theoretical assumptions and preliminary empirical findings, this mediation is considered partial, suggesting that DT not only influences SMP through RMO but also has a direct effect.

To construct a multi-dimensional RMO variable system, this study—building on prior theory—attempts for the first time to extend VBN theory from the individual behavioral level to the organizational level. Originally developed to explain individual environmental behaviors, VBN theory has been shown to be applicable to the behavioral transformation processes of enterprises undergoing green transition, particularly in understanding strategic green marketing orientation drivers from an owner manager perspective [36]. Specifically, the perception of green responsibility at the organizational level parallels the value cognition of individuals; the construction of a green corporate culture reflects belief formation; and institutional mechanisms such as ESG evaluation and green Key Performance Indicators (KPIs) serve as organizational-level behavioral norms [37,38]. Based on this theoretical translation, this study conceptualizes RMO as a five-dimensional mechanism comprising (1) customer perception responsiveness, (2) green value orientation, (3) belief dissemination, (4) behavioral adjustment, and (5) institutionalized execution. These five interrelated dimensions collectively form an organizational-level “cognition–belief–norm–behavior” feedback loop, which structurally aligns with the logic of the original VBN pathway. As such, they capture the dynamic process through which enterprises transition from external green signal perception to internal strategic response within the context of green marketing transformation. From this theoretical foundation, two primary contributions emerge. First, on a conceptual level, this study extends VBN theory beyond its original scope, applying it to organizational strategic decision-making in the context of green consumption. In doing so, it proposes an innovative, multi-dimensional closed-loop response model tailored to enterprise transformation behavior, thereby enriching the theoretical interface between RMO and SMP. Second, in terms of mechanism structure, the study emphasizes the dynamic feedback process of “perception–orientation–dissemination–adjustment–institution.” In contrast to traditional sales-oriented market models, this perspective foregrounds the internalization of values and behavioral adaptation processes, offering deeper insights into how organizations embed green strategies. To improve the practical applicability and explanatory power of the model, this study also accounts for specific characteristics of the NEV industry—such as rapid technological iteration, strong policy dependence, and heterogeneous green consumer preferences. Special attention is paid to the fact that many small- and medium-sized enterprises (SMEs) in this sector face dual constraints in digital transformation and green marketing due to limited resources. Consequently, considerations of operational feasibility and adaptability across enterprise sizes are integrated into sample construction and mechanism design, supporting broader applicability and extrapolation of the model. It is important to acknowledge the boundary conditions of this study. The model is most relevant for enterprises with a foundational level of digital infrastructure and green awareness. It may not fully apply to micro-enterprises or firms lacking basic transformation capabilities. Nevertheless, by reconstructing VBN theory at the organizational level and clarifying the dimensions of the response mechanism, this study provides both theoretical innovation and empirical validation. These contributions collectively enhance the understanding of enterprise performance generation mechanisms in the dual context of digitalization and green transformation.

To construct a multi-dimensional RMO variable system, this study—building on prior theory—attempts for the first time to extend VBN theory from the individual behavioral level to the organizational level. Originally developed to explain individual environmental behaviors, VBN theory has been shown to be applicable to the behavioral transformation processes of enterprises undergoing green transition, particularly in understanding strategic green marketing orientation drivers from an owner manager perspective [36]. Specifically, the perception of green responsibility at the organizational level parallels the value cognition of individuals; the construction of a green corporate culture reflects belief formation; and institutional mechanisms such as ESG evaluation and green Key Performance Indicators (KPIs) serve as organizational-level behavioral norms [37,38]. Based on this theoretical translation, this study conceptualizes RMO as a five-dimensional mechanism comprising (1) customer perception responsiveness, (2) green value orientation, (3) belief dissemination, (4) behavioral adjustment, and (5) institutionalized execution. These five interrelated dimensions collectively form an organizational-level “cognition–belief–norm–behavior” feedback loop, which structurally aligns with the logic of the original VBN pathway. As such, they capture the dynamic process through which enterprises transition from external green signal perception to internal strategic response within the context of green marketing transformation. From this theoretical foundation, two primary contributions emerge. First, on a conceptual level, this study extends VBN theory beyond its original scope, applying it to organizational strategic decision-making in the context of green consumption. In doing so, it proposes an innovative, multi-dimensional closed-loop response model tailored to enterprise transformation behavior, thereby enriching the theoretical interface between RMO and SMP. Second, in terms of mechanism structure, the study emphasizes the dynamic feedback process of “perception–orientation–dissemination–adjustment–institution.” In contrast to traditional sales-oriented market models, this perspective foregrounds the internalization of values and behavioral adaptation processes, offering deeper insights into how organizations embed green strategies. To improve the practical applicability and explanatory power of the model, this study also accounts for specific characteristics of the NEV industry—such as rapid technological iteration, strong policy dependence, and heterogeneous green consumer preferences. Special attention is paid to the fact that many small- and medium-sized enterprises (SMEs) in this sector face dual constraints in digital transformation and green marketing due to limited resources. Consequently, considerations of operational feasibility and adaptability across enterprise sizes are integrated into sample construction and mechanism design, supporting broader applicability and extrapolation of the model. It is important to acknowledge the boundary conditions of this study. The model is most relevant for enterprises with a foundational level of digital infrastructure and green awareness. It may not fully apply to micro-enterprises or firms lacking basic transformation capabilities. Nevertheless, by reconstructing VBN theory at the organizational level and clarifying the dimensions of the response mechanism, this study provides both theoretical innovation and empirical validation. These contributions collectively enhance the understanding of enterprise performance generation mechanisms in the dual context of digitalization and green transformation.

The design of variables in this study adheres to the following specifications:

(1)

DT is measured using a text-mining approach applied to corporate annual reports. Standardized keyword frequencies—based on terms such as AI, big data, cloud computing, and blockchain—are used to reflect the degree of digital technology adoption by enterprises

(2)

RMO is constructed based on the theoretical framework of VBN theory, capturing the organizational response loop that progresses from perception to belief to behavioral execution

(3)

SMP draws on the ESG framework, and is assessed across four key dimensions: economic benefits, risk control, marketing compliance, and growth sustainability.

While these dimensions originate from broader considerations of corporate performance, they are highly applicable to the green marketing context and possess strong explanatory relevance:

• Economic benefits capture whether green marketing efforts have yielded financial gains—for example, through energy-saving cost reductions or increased sales of NEV products.
• Risk control evaluates the extent to which green strategies help enterprises mitigate potential external risks, such as regulatory scrutiny, reputational pressures, or NGO activism.
• Marketing compliance measures alignment with environmental regulations in branding, labeling, and channel practices. This is particularly critical in light of increasingly stringent oversight of green consumption.
• Growth sustainability reflects the firm’s ability to cultivate long-term customer loyalty, brand credibility, and stable revenue streams through its green marketing efforts.

In contrast to traditional marketing performance indicators (e.g., the 4A model, sales-based metrics, or brand awareness), which tend to emphasize short-term consumer response or sales outcomes, the ESG-based framework enables a more comprehensive evaluation of institutional implementation, environmental adaptability, and sustainable value creation. By integrating financial and non-financial indicators, ESG effectively addresses the core components of environmental responsibility (E), social expectations (S), and governance compliance (G), making it especially well-suited for assessing the outcomes of green marketing initiatives.

To tailor ESG to the specific context of green marketing, this study refines its internal structure—focusing on the four dimensions most relevant to marketing performance: economy, risk, compliance, and growth. This adjustment maintains the integrity of the ESG framework while enhancing its contextual relevance.

However, several limitations of the ESG framework are acknowledged. First, SMEs often disclose less ESG-related information, potentially affecting data completeness. Second, attributing specific performance outcomes directly to green initiatives remains empirically complex. These issues are further discussed in the research limitations section. To address them, future studies are encouraged to introduce complementary behavioral indicators—such as green market share growth or improvements in consumer environmental awareness—to enhance the precision and explanatory power of performance evaluation.

Importantly, the ESG framework has already been extensively validated in the fields of green finance and strategic management, and it demonstrates strong feasibility in terms of both sample coverage and theoretical alignment. Therefore, this study adopts ESG as the observational variable system for measuring green marketing performance within the SEM framework—balancing theoretical robustness, empirical operability, and green-oriented expressiveness.

Finally, it is important to note that a nonlinear relationship or threshold effect may exist between the intensity of digital technology adoption and RMO in NEV enterprises. For example, in the early stages of digital transformation, an enterprise’s market response capability may remain underdeveloped. Upon surpassing a critical threshold, however, the enterprise may experience a rapid acceleration in responsiveness. At higher levels of digital investment, diminishing marginal returns may occur. While this study primarily focuses on identifying mechanisms and verifying path significance, a linear SEM framework is employed to ensure model identifiability and estimation robustness. Future research could explore nonlinear or threshold models to capture these more complex dynamics.

2.3. Variable Description and Indicator System Construction

To ensure scientific rigor, measurability, and theoretical consistency within the SEM framework, this study designs a comprehensive observation indicator system for the three core latent variables—DT, RMO, and SMP—and incorporates control variables to enhance model robustness.

(1)

DT

DT reflects the breadth and depth of an enterprise’s application of key digital technologies, such as AI, big data, cloud computing, and blockchain, representing the strategic level of digital adoption [39]. This study employs a text-mining approach to extract standardized word frequencies from corporate annual reports. Specifically, the presence and frequency of relevant digital terms are measured, and an aggregate indicator (“DT5”) is constructed to reflect the relative intensity of digital discourse in managerial narratives.

(2)

RMO

Rooted in the VBN theory, RMO captures an enterprise’s dynamic perception, belief formation, behavioral adjustment, and institutionalization capacity in response to evolving green market demands [40,41]. The RMO construct spans five interrelated dimensions, emphasizing the organizational mechanisms that translate digital capabilities into adaptive and sustainable marketing behaviors.

(3)

SMP

Drawing on both the ESG framework and marketing performance literature, SMP is operationalized through four indicators reflecting financial and non-financial dimensions. These include the economic efficiency, stability, risk control, and compliance level of an enterprise’s green marketing efforts.

The full indicator system is presented in

Table 1. Variable and indicator system.

Variable Category	Indicator Code	Indicator Description
DT	DT1	Frequency of terms related to “AI” in annual reports (standardized).
	DT2	Frequency of terms related to “big data” in annual reports (standardized) [42].
	DT3	Frequency of terms related to “cloud computing” in annual reports (standardized).
	DT4	Frequency of terms related to “blockchain” in annual reports (standardized).
	DT5	Proportion of the combined frequency of four core technology terms relative to total text length in the management discussion section (aggregated indicator) [43,44].
RMO	RMO1	Corresponds to the “perception-awareness” stage in the VBN model. Measures whether the enterprise actively monitors green consumption trends via data analysis, market feedback, and user reports, e.g., publishing green trend reports or adjusting marketing strategy.
	RMO2	Corresponds to “belief–value orientation” in the VBN model. Assesses whether the enterprise promotes green values and visions in branding, advertising, websites, or CSR reports to shape customer value alignment [45].
	RMO3	Reflects “belief externalization.” Measures the frequency of green-related expressions (e.g., “sustainability,” “environmental protection”) in social media, corporate announcements, or media platforms [46,47].
	RMO4	Represents “norm–behavior adjustment.” Evaluates whether the enterprise incorporates green feedback into operational behaviors (e.g., eco-friendly packaging, low-carbon logistics, green certification).
	RMO5	Corresponds to the “institutionalization of norms.” Measures whether formal green management systems exist (e.g., sustainability KPIs, green procurement, carbon accounting, or dedicated green departments) [48].
SMP	SMP1	Evaluates whether the profitability of the core business has improved due to green marketing strategies. Reflects value realization through green product offerings and strategic transformation [49,50].
	SMP2	Assesses the sustainability and market stickiness of green marketing by focusing on revenue stability and long-term brand/customer growth, beyond short-term sales.
	SMP3	Examines the enterprise’s ability to manage risks, reduce costs, and avoid controversies (e.g., false green claims) during green marketing. Indicates integration of strategy into internal controls [51].
	SMP4	Measures compliance with environmental regulations and industry standards in marketing communication and channel development. Aligns with ESG rating criteria for green compliance.

.

Considering that an enterprise’s financial structure and growth capacity may significantly influence the implementation and effectiveness of green strategies, this study introduces a set of control variables to account for potential confounding effects. The selected variables are detailed in

Table 2. Control variables.

Type of Control Variables	Code	Indicator Definition
Financial structure	CV1	Debt-to-asset ratio
	CV2	Return on equity (ROE) [52]
	CV3	Cash flow ratio from operating activities
Growth ability	CV4	Revenue growth rate
	CV5	Accounts receivable ratio [53]
	CV6	Price/book value ratio

.

These control variables are theoretically grounded and widely adopted in the literature related to green strategy, corporate performance, and marketing behavior analysis. They effectively capture enterprises’ resource allocation capabilities and their potential to influence sustainable marketing behaviors.

From the financial structure perspective

• CV1 (Debt-to-Asset Ratio) reflects an enterprise’s capital structure and solvency. As a key indicator of financial stability and risk tolerance, a high leverage ratio may restrict an enterprise’s capacity or willingness to invest in digitalization and green initiatives, thereby hindering sustainable marketing performance.
• CV2 (ROE) measures the efficiency of equity utilization and reflects the enterprise’s capability to generate profit from shareholders’ investment. Prior studies have established a strong link between ROE and an enterprise’s strategic engagement in green innovation and market development.
• CV3 (Operating Cash Flow Ratio) represents the actual cash generated from core business activities. Sufficient operational cash flow is critical for sustaining long-term investments in green transformation and building flexible, responsive marketing systems.

From the growth capability perspective

• CV4 (Revenue Growth Rate) serves as a direct proxy for enterprise growth potential and agility in market response. High growth firms typically possess stronger resource input capacity and greater organizational responsiveness to green market demands;
• CV5 (Accounts Receivable Ratio) reflects liquidity risk and capital lock-in. Elevated levels may impede enterprises’ ability to respond quickly to green market signals or to invest in related promotional activities;
• CV6 (Price-to-Book Ratio) indicates market valuation and investor expectations. A higher P/B ratio generally signals positive market sentiment regarding a firm’s prospects in digitalization and sustainable development, which may translate into greater access to resources for implementing green strategies.

In summary, the control variable framework integrates considerations of both financial robustness and growth dynamics, establishing a theoretically and empirically cohesive model foundation. These variables address the question of whether enterprises possess the financial and organizational capacity to make sustainable investments and execute responsive strategies, which is essential for unbiased estimation of the relationships among DT, RMO, and SMP.

All indicators are treated as first-order reflective constructs, and will be subjected to reliability and validity assessments in subsequent empirical analyses. For text-based indicators, the Term Frequency–Inverse Document Frequency (TF-IDF) algorithm is applied for weighting, while synonym consolidation and semantic filtering are performed using Python to ensure indicator consistency, interpretability, and inter-variable correlation.

It is worth noting that variables such as research and development (R&D) investment, patent counts, and supply chain digital maturity may also influence the implementation path of corporate green strategies. However, these factors are not included in the current study’s control variable system for the following reasons. First, the core objective of this research is to investigate the marketing performance-oriented response mechanism (RMO), constructing a theoretical pathway of “DT → RMO → SMP”. Therefore, the selection of control variables primarily aims to address potential confounding effects arising from the financial capacity and growth potential corporate marketing behaviors [54,55]. Accordingly, the study prioritizes financial structure indicators such as the debt-to-asset ratio and cash flow ratio, which reflect an enterprise’s resource allocation capability. In parallel, growth metrics such as revenue growth rate and accounts receivable ratio are incorporated to control for fundamental operational factors that may influence RMO responsiveness and SMP outcomes. Second, although innovation-related variables such as R&D investment and patent output hold relevance in the context of technology strategy, their applicability in this study is limited. These indicators often exhibit high cross-industry heterogeneity, suffer from inconsistent disclosure standards, and face partial data unavailability—particularly within the NEV sector. These issues pose challenges in terms of data accessibility, comparability, and risk introducing estimation biases into the model. Third, including an excessive number of control variables—particularly those with potential collinearity—may weaken the explanatory power of key paths and reduce model fit robustness in SEM [56]. To mitigate this risk, the current study limits control variable inclusion to six indicators that successfully pass Variance Inflation Factor (VIF) tests, thereby ensuring both model parsimony and structural clarity.

Given these considerations, this study does not incorporate technological innovation indicators as control variables within the SEM framework. The external generalizability of this decision and the associated risks of omitted variable bias are explicitly acknowledged in the “research limitations” of Conclusions section.

To empirically test the proposed “DT → RMO → SMP” path model, this study develops a structured questionnaire-based measurement instrument grounded in the previously established variable system. The overall design adheres to a “theory–indicator–measurement” logic chain and adopts a five-point Likert scale for all items (1 = “completely disagree” to 5 = “completely agree”). Detailed measurement design is presented in

Table 3. Operational definitions and measurement design.

Latent Variable	Operational Definition	Observation Variable	Scale Development Basis
DT	The breadth and depth of digital technology application in fields such as AI, big data, cloud computing, and blockchain	DT1: AI or intelligent algorithms are used for precision marketing. DT2: Big data is leveraged for customer analysis and product design. DT3: Cloud platforms are used to coordinate marketing, R&D, and operations. DT4: Blockchain-based green supply chains are being explored or adopted. DT5: A strategic, integrated digitalization plan has been formulated.	Adapted from [56]; Based on the “Guidelines for Evaluating the Digitalization Level of the Manufacturing Industry” (MIIT)
RMO	The organizational ability to identify, interpret, and dynamically respond to green consumption market changes	RMO1: Regular tracking of green consumption trends. RMO2: Active communication of environmental values to customers. RMO3: Dissemination of green concepts via news and social media. RMO4: Product/service adjustments based on green market feedback. RMO5: Establishment of internal green response mechanisms and processes.	Developed from VBN theory and market-oriented response models; Referenced from existing green marketing scale items
SMP	The comprehensive financial and non-financial performance of enterprises in green marketing	SMP1: Continued growth in core business profitability. SMP2: Steady customer base growth driven by green marketing. SMP3: Effective cost and risk control in green marketing. SMP4: Brand alignment with environmental compliance requirements.	Derived from the ESG performance literature and marketing performance frameworks (e.g., Kotler)

.

To ensure theoretical consistency and empirical validity—particularly for the RMO construct—the measurement scale is developed through a rigorous, multi-step process: theoretical framework formulation → dimension extraction → item drafting → expert consultation and refinement → pilot testing → finalization of measurement instrument.

2.4. Experiment Settings

To empirically investigate the mechanism through which DT affects enterprises’ SMP via RMO, this study selects NEV enterprises in China as the research context. As a representative and dynamic sub-sector within the broader green consumer goods industry, NEV firms demonstrate notable leadership in both digital technology adoption (e.g., AI, big data, cloud computing, and blockchain) and environmental responsibility transformation. Therefore, focusing on this sector allows for a clearer examination of the interaction mechanisms between digital capabilities and green marketing performance. However, the NEV sector possesses certain unique characteristics—such as stronger policy intervention, higher technological concentration, and more integrated supply chain coordination—compared to the general green consumer goods industry. Consequently, the extrapolation of findings should be contextualized within the specificities of the NEV industry.

Sample firms were drawn from A-share listed companies on the Shanghai and Shenzhen Stock Exchanges, encompassing the Main Boards, Science and Technology Innovation Board (STAR Market), and Growth Enterprise Market. The sample includes midstream and downstream enterprises within the NEV value chain, such as complete vehicle manufacturers, core component providers, and intelligent connected system developers.

The screening criteria were as follows:

(1)

Industry classification: Firms must be classified under NEV-related subcategories of “Automobile Manufacturing” in the National Economic Industry Classification (2021);

(2)

Data completeness: Firms with missing annual reports or critical financial disclosures are excluded;

(3)

Time coverage: Only enterprises with continuous disclosure of annual reports and key indicators from 2018 to 2022 are included to ensure longitudinal data consistency;

(4)

Business focus: The primary business revenue must predominantly derive from NEV-related operations.

Following the above criteria, 86 eligible listed enterprises were identified, yielding balanced panel dataset of 430 firm-year observations.

To enhance transparency and sample interpretability, this study conduct further descriptive analysis on geographical distribution, enterprise scale, and industrial chain positioning:

(1)

Geographical distribution: Sample firms are predominantly located in economically developed eastern and central provinces, including Guangdong, Shanghai, Jiangsu, Zhejiang, and Beijing—regions that serve as strategic hubs for China’s NEV sector, demonstrating significant spatial clustering;

(2)

Enterprise scale: The majority of the sample comprises large and medium-sized firms, with an average total asset size of approximately 8.9 billion, indicating strong financial capacity and technological investment potential;

(3)

Industrial chain position: About 35% of sampled firms are complete vehicle manufacturers, 45% are core component producers (e.g., electric drives, batteries, electronic control systems), and the remaining portion includes intelligent connected systems and supporting service providers. This reflects a relatively complete and representative industrial chain structure.

It is important to acknowledge that the current sample consists exclusively of listed enterprises, enabling access to standardized, continuous, and publicly disclosed data. However, this may introduce sampling bias, as SMEs are not included. SMEs often differ significantly from listed firms in terms of resource constraints, digital transformation pathways, and policy sensitivity. These limitations are further addressed in the “Conclusions” section, where the scope of external generalizability is clarified.

Moreover, the chosen time window of 2018–2022 reflects a distinct policy and market context. On one hand, this period marks a critical transition phase for the NEV industry, shifting from government subsidy-led growth to a market-oriented development model. It is a time of significant transformation in industrial ecosystems, enterprise behavior, and market dynamics—thus offering valuable empirical insights. On the other hand, this stage is also marked by external shocks such as subsidy phase-outs, capacity restructuring, and the COVID-19 pandemic, which may introduce structural disturbances into the dataset.

To address potential confounding effects, this study controls for variables related to enterprise size, financial structure, and growth capability during model construction and SEM analysis. Furthermore, the broader policy environment and potential impact boundaries are thoroughly discussed in the later Discussion section to contextualize the findings and enhance research robustness.

The DT indicator is constructed based on the frequency and weight of digital-technology-related keywords extracted from corporate annual reports. Textual analysis is conducted using Python v3.12 in a Jupyter Notebook environment (https://jupyter.org/), following a structured pipeline to ensure accuracy and reproducibility. The steps are outlined as follows:

(1)

Text Preprocessing

All corporate annual report texts are first cleaned by standardizing formats, removing punctuation, and eliminating irrelevant content (e.g., disclaimers, appendices, or repeated headers) to ensure the quality and consistency of input data

(2)

Tokenization and Dictionary Customization

Text segmentation is performed using the Jieba Chinese tokenizer, incorporating a custom dictionary to unify and merge synonyms (e.g., “artificial intelligence” and “AI”). This ensures semantic consistency and reduces redundancy in term recognition

(3)

Keyword Selection

Based on the prior literature and industry practices [57,58], four core digital technology terms are selected: “AI,” “big data,” “cloud computing,” and “blockchain.” These terms reflect the most relevant and impactful digital capabilities within the NEV industry

(4)

Weighting via TF-IDF Algorithm

To account for differences in semantic range and prevalence across firms, the TF-IDF algorithm is applied. This weighting method corrects for term overuse or underrepresentation, adjusting raw frequency values to reflect term distinctiveness and informativeness [43,45].

The general formula for the DT index is defined as

$$D{T}_i={\displaystyle \frac{1}{n}}\sum _{j=1}^n TFID{F}_{ij}\cdot w_j$$

Here, $D{T}_i$ denotes the DT score for firm $i$; $\mathrm{T}\mathrm{F}\mathrm{I}\mathrm{D}{\mathrm{F}}_{\mathrm{i}\mathrm{j}}$ is the TF-IDF value for keyword $j$ in firm $i$’s report; $w_j$ represents the weight of keyword $j$; and $n$ is the number of selected keywords [57,58].

Specifically, the annual reports for all firms from 2018 to 2022 are used as the reference corpus. For each report, the relative frequency of each keyword is calculated by dividing the count by the total word count of the report. This ensures normalization across documents of varying lengths [59]. The document frequency (i.e., proportion of reports containing a given keyword) is then used to compute the inverse document frequency component. The final TF-IDF scores for the four keyword categories are aggregated and standardized to construct the DT index

(5)

Standardization

To facilitate comparability across firms, the calculated keyword frequencies are transformed into standard deviation units (z-scores). The final DT score is computed as follows:

$$D{T}_{\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}}={\sum }_{i=1}^4 w_i\cdot {\mathrm{StandardizedFreq}}_i$$

(1)

$w_i$ represents the TF-IDF weight for the i-th keyword category; $D{T}_{\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}}$ is the enterprise’s final DT index; and ${\mathrm{StandardizedFreq}}_i$ denotes the standardized term frequency of the i-th keyword [60,61]

(6)

Data Processing and Outlier Treatment

All continuous variables, including the DT score, are Winsorized at the 1st and 99th percentiles to mitigate the influence of extreme values. For observations with missing data, industry-year averages or linear interpolation within the same sector and year are used to ensure completeness and reduce sample bias.

3. Results

3.1. Data Robustness Test

To enhance data robustness and reduce potential bias arising from extreme observations, this study applied outlier identification and Winsorization to all continuous variables—including financial control variables and selected performance indicators—prior to modeling. Specifically, the raw data were cleaned using the Winsorization method, with the lower and upper bounds set at the 1st and 99th percentiles. This approach effectively mitigates the influence of outliers while preserving the underlying structure of the data, thereby meeting the distributional assumptions required for SEM analysis.

To verify the effectiveness of the Winsorization process, two statistical tests were conducted:

(1)

Normality test: ShapiroWilk tests were performed on both the original and treated datasets. The results indicate that the raw data significantly deviate from a normal distribution (all p-values < 0.001), whereas the Winsorized data exhibit markedly improved normality (p-values range from 0.02 to 0.10);

(2)

Distribution structure test: Kolmogorov–Smirnov (KS) tests comparing the distributions before and after Winsorization yielded small KS statistics, with all p-values exceeding 0.95. These results confirm that the Winsorization method does not introduce systematic bias and retains the stability of the original data structure.

Table 4. Results of normality and outlier tests before and after Winsorization.

Variable	Shapiro_p_Before	Shapiro_p_After	KS_Statistic	KS_p_Value
Asset_Liability_Ratio	0.0000	0.0213	0.0116	1.0000
ROE	0.0000	0.0754	0.0116	1.0000
Net_Profit_Margin	0.0000	0.1025	0.0116	1.0000

presents the normality and outlier test results before and after treatment.

To enhance transparency and facilitate visual comparison, this study also provides pre- and post-treatment values for each indicator, as shown in Figure 1.

Through the above data preprocessing procedures, this study ensures that the variables used in subsequent SEM analysis exhibit the distributional stability required for robust statistical inference. These steps provide methodological support for the reliability of path coefficient estimation and the validity of result interpretation.

3.2. Descriptive Statistics and Correlation Analysis

To verify the measurement quality of the scale, standardized reliability and validity tests were performed on its five observation indicators, including internal consistency reliability and construct validity. The results are summarized in

Table 5. Reliability and construct validity test results for the scale.

Indicator	Result
Cronbach’s α	0.812
Kaiser-Meyer-Olkin (KMO) test	0.802
Bartlett’s test of sphericity χ2 (p-value)	563.82 (p < 0.001)

.

Analysis of 430 sample groups shows that the scale achieves a Cronbach’s α value of 0.812, exceeding the recommended threshold of 0.70, indicating strong internal consistency. Construct validity tests yield a KMO value of 0.802, confirming sample adequacy for factor analysis. Bartlett’s test of sphericity produces a χ² value of 563.82 (p < 0.001), indicating sufficient inter-variable correlations for common factor extraction. Exploratory factor analysis loads all five measurement indicators into a single latent factor, with results presented in

Table 6. Factor loadings of RMO observation indicators.

Observation Indicator	Factor Loading (Factor 1)
RMO1 Perception responsiveness	0.705
RMO2 Green value guidance	0.697
RMO3 Belief dissemination capability	0.683
RMO4 Behavior adjustment mechanism	0.726
RMO5 Response institutionalization level	0.668

.

All factor loadings exceed 0.65, supporting strong convergent validity within the RMO construct.

To further enhance measurement robustness, the Multi-Trait Multi-Method (MTMM) approach was applied to cross-validate the five RMO dimensions. This method assesses convergent and discriminant validity by analyzing correlations between indicators under two distinct measurement methods:

(1)

Direct questionnaire scoring (five-point Likert scale)

(2)

Semantic scoring simulation based on ESG information disclosures, using Python TF–IDF quantification of green-related statements.

The correlation matrix for the two scoring methods shows that, for the same trait, different measurement methods produce significant correlations above 0.45 (e.g., the correlation between RMO1_questionnaire and RMO1_semantic is 0.713). The results show that, for the same trait, different measurement methods produce significant correlations above 0.45 (e.g., the correlation between RMO1_questionnaire and RMO1_semantic is 0.713). Correlations between different traits are notably lower than those within the same trait, indicating both good convergent validity and strong discriminant validity. This confirms the stability of the RMO scale across different measurement approaches, reinforcing its methodological soundness for SEM.

For the DT variable, convergent and structural validity were examined through correlation analysis and principal component analysis (PCA) using standardized keyword frequency data. The results are shown in

Table 7. Correlation matrix of DT variables.

	DT1	DT2	DT3	DT4	DT5
DT1	1.000	0.615	0.583	0.467	0.711
DT2	0.615	1.000	0.648	0.510	0.735
DT3	0.583	0.648	1.000	0.532	0.710
DT4	0.467	0.510	0.532	1.000	0.662
DT5	0.711	0.735	0.710	0.662	1.000

and

Table 8. PCA results of DT variables.

Principal Component	Characteristic Value	Explained Variance Rate (%)	Cumulative Explained Variance (%)
PC1	1.950	23.0	23.0
PC2	1.778	22.2	45.2
PC3	1.091	13.6	58.8
PC4	0.928	11.3	70.1
PC5	0.853	10.7	80.8

.

The correlation coefficients indicate strong positive relationships among DT indicators, demonstrating good convergent validity. The PCA results show that the first two components explain 45.2% of total variance (PC1: 23.0%, PC2: 22.2%), supporting the aggregation of DT indicators into a latent construct. The multidimensional nature of DT is consistent with its theoretical positioning as a strategic construct encompassing broad technology adoption.

Before conducting SEM, descriptive statistics and correlation tests were performed to assess variable distributions and preliminary relationships. To enhance measurement reliability, a dual-path quantification strategy was employed, combining (1) subjective ratings from NEV enterprise managers and (2) semantic scoring from corporate annual reports. ESG disclosures and operational data were standardized and scored according to defined indicators, generating a dataset of 430 firm-year observations from 86 listed NEV enterprises (2018–2022).

To further validate the stability and measurement validity of the latent variable scales, this study adopted a dual-path quantification strategy combining

(1)

Subjective ratings from enterprise managers, collected via five-point Likert scale surveys targeting management teams and key personnel of NEV companies.

(2)

Semantic scoring based on corporate annual reports.

All scoring data derive from firms’ annual ESG disclosures and operational records, which are standardized and scored according to predefined indicators to construct a normalized dataset. The final sample comprises 86 listed NEV-related enterprises, yielding 430 firm-year observations for 2018–2022, ensuring high completeness and representativeness.

To enhance external reliability, semantic scoring was introduced as a supplementary verification method. The detailed procedure is provided in Code.txt. Although text frequency-based indicators may be affected by differences in report-writing styles, this risk is effectively mitigated in the NEV industry for two reasons:

(1)

Annual reports—particularly ESG and strategic sections—follow standardized disclosure requirements, resulting in homogeneous content less prone to stylistic interference.

(2)

Keywords are drawn from policy-guided, high-frequency industry terms directly related to firms’ digital and green-transformation strategies, ensuring orientation and stability.

Furthermore, indicator construction employs multi-term collaborative modeling and normalization, building a composite scoring system that minimizes random noise and expression bias. This approach demonstrates strong adaptability and reliability in the NEV context, obviating the need for additional external validation.

Descriptive statistics for the main research variables (DT, RMO, SMP) and their observed indicators are presented in Figure 2.

Figure 2 shows that variable means and standard deviations fall within reasonable ranges, with no extreme outliers. For DT indicators, the mean keyword frequencies for AI, big data, and cloud computing are 0.254, 0.312, and 0.289, respectively, with standard deviations around 0.10, indicating some variability in technology-related discourse and reflecting differences in the depth of DT strategy implementation. For RMO indicators—such as perception responsiveness, green value orientation, belief dissemination capability, and behavior adjustment mechanism—means range from 3.68 to 3.95, with standard deviations below 0.65, suggesting generally strong corporate response capabilities with minor variation. SMP dimensions show even lower volatility; for example, economic performance has a mean of 0.087 (SD = 0.034), while marketing compliance and risk control capability have SD values below 0.15. These patterns indicate stable performance in executing green marketing practices across the sample.

To preliminarily assess linear relationships among variables, Pearson correlation tests were performed (

Table 9. Correlation matrix of core variables.

Variable	DT	RMO	SMP
DT	1.000	0.482 **	0.415 **
RMO	0.482 **	1.000	0.531 **
SMP	0.415 **	0.531 **	1.000

Note: ** indicates p < 0.01.

).

DT is significantly positively correlated with RMO (r = 0.482, p < 0.01), indicating that higher levels of technology adoption are associated with stronger response capabilities. DT also shows a significant positive correlation with SMP (r = 0.415, p < 0.01), providing preliminary support for hypothesis H1. The relatively strong correlation between RMO and SMP (r = 0.531, p < 0.01) suggests that RMO may mediate the relationship between DT and performance. Control variables also show significant correlations with SMP, supporting their inclusion in the model to account for confounding effects of firms’ financial conditions.

Before SEM estimation, multicollinearity diagnostics were conducted using the variance inflation factor (za) method (

Table 10. Multicollinearity diagnosis.

Variable	VIF Value
RMO3	1.026
RMO5	1.023
DT4	1.023
CV2	1.015
RMO4	1.015
RMO1	1.014
DT3	1.013
RMO2	1.013
CV1	1.011
DT2	1.011
DT1	1.003

).

All VIF values are well below the conventional thresholds of 5 or 10, indicating no serious multicollinearity among independent variables.

To assess structural stability, a one-by-one variable elimination sensitivity analysis was performed, sequentially removing each DT and RMO indicator and recalculating VIF statistics (

Table 11. Sensitivity analysis of VIF.

Removed Variable	Average VIF	Maximum VIF
DT1	1.0163	1.0256
DT2	1.0146	1.0263
DT3	1.0141	1.0236
DT4	1.0123	1.0234
RMO1	1.0143	1.0245

).

The fluctuations in average and maximum VIF values are minimal (within ±0.02), confirming the model’s stability under variable elimination and the absence of structural multicollinearity risks.

3.3. Model-Fitting Results

The study sample comprises 86 listed enterprises in the NEV industry, with five years of data coverage (2018–2022), forming a balanced panel dataset of 430 observations. This sample size meets the requirements for SEM with a medium-sized dataset.

According to effect size guidelines, an $f^2$ = 0.15 is considered a medium effect size for path estimation in structural models. For testing single-path mediating effects, the required sample size generally falls within N ≥ 138–200. The effect size $f^2$ is calculated using the following equation to assess the influence of a specific path in the structural equation

$$f^2={\displaystyle \frac{R^2-R_{excluded}^2}{1-R^2}}$$

Here, $R^2$ represents the model’s goodness-of-fit (GIF) when the explanatory variable is included, and $R_{excluded}^2$ denotes the GIF after excluding the explanatory variable.

Given that the proposed model contains latent variables (DT, RMO, SMP), 17 observed indicators, and control variables—resulting in 23 estimated paths—the minimum sample size for SEM estimation should satisfy N ≥ 10 × the number of parameters (approximately N ≥ 230).

To further validate adequacy, a statistical power simulation was conducted using G*Power v3.1 with the following parameters:

(1)

Test type: One-tailed z-test for fixed regression paths;

(2)

Expected effect size: f² = 0.15 (medium);

(3)

Significance level: α = 0.05;

(4)

Desired power level: 0.90;

(5)

Number of independent variables: 6–8 (based on actual paths);

The results indicate that, at a desired power of 0.90, the minimum required sample size is approximately 172–185. The actual sample size of 430 observations therefore exceeds this lower bound, meeting the estimation requirements for multi-path, multi-indicator SEM. This ensures high statistical power for identifying causal relationships among DT, RMO, and SMP. All latent variables were modeled using a first-order reflective structure, with clearly defined relationships between observed and latent variables and adequate degrees of freedom, satisfying SEM identifiability conditions.

Using AMOS 26.0, the structural model was estimated to examine the direct and mediating effects of DT on SMP through RMO. The model includes three latent variables (DT, RMO, SMP), six control variables, and four hypothesized core paths (H1–H4). The GIF results are shown in

Table 12. Model GIF indices.

Indicator	Fitted Value	Recommended Threshold	Status
Chi-square/degree of freedom ratio	2.681	<3.0	Good
Root Mean Square Error of Approximation (RMSEA)	0.064	<0.08	Good
Standardized Root Mean Square Residual (SRMR)	0.045	<0.08	Good
Comparative Fit Index (CFI)	0.938	>0.90	Excellent
Tucker–Lewis Index (TLI)	0.921	>0.90	Excellent
Goodness-of-Fit Index (GFI)	0.902	>0.90	Better
Adjusted Goodness-of-Fit Index (AGFI)	0.875	>0.85	Better

.

The reported indices indicate an overall good model fit. The chi-square/degrees-of-freedom ratio, RMSEA, and SRMR meet common acceptability criteria, while incremental fit indices (CFI, TLI) exceed the 0.90 benchmark. Absolute fit indices (GFI, AGFI) are within acceptable ranges. Auxiliary indices such as normed fit index, incremental fit index, and parsimony goodness-of-fit index were examined but are not emphasized because primary indices already meet strong standards and the model is conceptually parsimonious. Introducing numerous additional fit statistics risks redundancy and may complicate interpretation without materially changing conclusions. The model therefore relies on primary fit metrics to demonstrate adequate explanatory power and structural stability.

The standardized path coefficients were estimated using AMOS 26.0, and the results are presented in

Table 13. Path coefficient and hypothesis testing results.

Path	Standardized Coefficient	Standard Error	p-Value	Hypothesis Test Result
DT → SMP(H1)	0.267 ***	0.041	<0.001	Support
DT → RMO(H2)	0.482 ***	0.052	<0.001	Support
RMO → SMP(H3)	0.396 ***	0.044	<0.001	Support

Notes: DT = Digital Transformation; SMP = Sustainable Marketing Performance; RMO = Relational Market Orientation; *** p < 0.001.

.

The results indicate that

• H1 is supported—Enterprises with higher levels of DT achieve significantly better SMP outcomes.
• H2 and H3 both have high coefficients, underscoring the critical mediating role of RMO. Strengthening RMO enables enterprises to better align strategic initiatives with evolving market demands.

From a model explanatory perspective:

• The structural model explains 46.8% of the variance in SMP, suggesting that DT and RMO jointly account for a substantial share of performance variation.
• The model also explains 36.1% of the variance in RMO, confirming that DT substantially enhances enterprises’ relational and response capabilities.

Regarding control variables:

• ROE and growth rate have significant positive effects on SMP.
• The debt-to-asset ratio exerts a significant negative effect, reinforcing that financial stability is an essential foundation for achieving strong green marketing performance.

3.4. Mediating Effect Analysis

Following the verification of the main path significance in the SEM framework, the study further employed the Bootstrap method to assess the mediating role of RMO in the DT → SMP relationship.

The decomposition of mediation effects follows the equation

$$c=c^{\prime }+ab$$

Here, $c$ refers to the total effect (overall influence of DT on SMP); $c^{\prime }$ stands for the direct effect (DT on SMP after controlling for RMO); $a$ represents the effect of DT on RMO; and $b$ denotes the effect of RMO on SMP.

This mediation analysis aims to uncover how digital strategies indirectly enhance corporate marketing performance through organizational mechanism innovation, thereby providing deeper theoretical insight and practical guidance.

In line with Hypothesis H4, RMO is expected to partially mediate the relationship between DT and SMP. To identify the indirect effect and evaluate its statistical significance, a non-parametric Bootstrap resampling approach (5000 resamples; 95% confidence interval [CI]) was applied. This method is preferred over the Sobel test when sample sizes are small or data distributions are non-normal and is widely adopted in contemporary SEM mediation research.

Table 14. Bootstrap test results.

Path Type	Effect Size	95% CI	Significance	The Type of Path
Direct effect (DT → SMP)	0.267	[0.192, 0.342]	Significant	Direct effect (DT → SMP)
Indirect effect (DT → RMO → SMP)	0.191	[0.139, 0.250]	Significant	Indirect effect (DT → RMO → SMP)
Total effect	0.458	[0.381, 0.527]	Significant	Total effect

reports the results.

Both the direct and indirect effect CIs exclude zero, indicating significant mediation. The indirect path (DT → RMO → SMP) confirms that digital capabilities improve marketing performance partly by strengthening market-oriented relational mechanisms. Although the indirect effect constitutes 41.7% of the total effect, this value should not be simplistically interpreted as “mediation accounts for over 40%.” Instead, it reflects a strong statistical association within the current model framework.

To test the robustness of this mediation, a sensitivity comparison was performed. Partial perturbation modeling removed one RMO dimension at a time, while control variables remained constant. The results are shown in

Table 15. Sensitivity comparison of the mediating path.

Situational Model	Mediating Effect	95% CI Lower	95% CI Upper	Significance
Original model	0.196	0.150	0.251	Significant
Remove RMO1	0.190	0.136	0.243	Significant
Remove RMO2	0.197	0.146	0.245	Significant
Remove RMO3	0.206	0.154	0.253	Significant
Remove RMO4	0.189	0.143	0.235	Significant

. All variants remain significant, indicating structural stability of the mediating path.

Further analysis reveals that DT enhances SMP not through a single channel, but via a series of nested and mutually reinforcing mechanisms embedded in RMO’s five dimensions:

(1)

Perception Capability—this strengthen real-time capture of green consumer preferences, market dynamics, and policy shifts via tools such as big data analytics and AI recognition. Improves the alignment of product positioning and communication strategies;

(2)

Green Value Orientation—Internalizes external sustainability concepts into corporate strategy, fostering strategic unity and motivating organizational behavioral shifts toward sustainability;

(3)

Belief Dissemination Capability—Embeds green values into organizational culture through training, internal campaigns, and cultural programs, enhancing the consistent execution of sustainability strategies;

(4)

Behavioral Adjustment Mechanism—Enables rapid adaptation in product design, supply-chain optimization, and market launches (e.g., green procurement, low-carbon innovation), enhancing operational execution in SMP;

(5)

Institutionalized Feedback—Integrates green principles into governance structures, performance indicators, and process standardization, ensuring long-term strategy adherence and continuous improvement.

The sensitivity analysis further confirms that the mediation chain remains operational even if one dimension is removed (e.g., removing Belief Dissemination Capability still yields a mediating effect of 0.206, CI: [0.154, 0.253]). This indicates a redundant yet interdependent mechanism structure, where no single element alone drives the effect, but each reinforces the others.

In sum, DT influences SMP via a closed-loop “perception → cognition → dissemination → behavior → institutionalization” chain. Compared with traditional linear mediation models, the proposed mechanism is dynamic, multi-dimensional, and feedback-oriented, offering both theoretical advancement and actionable insight into how digital strategies drive sustainable marketing outcomes.

3.5. Causal Identification and Robustness Analysis

To address potential bidirectional causality and endogeneity among the model variables (DT, RMO, SMP), this study applied causal identification strategies, including the lagged variable method, instrumental variable regression, and fixed-effect controls. These approaches enhance the credibility of empirical inferences and the robustness of the conclusions.

The two-stage least squares (2SLS) approach was used to mitigate endogeneity concerns. n the first stage, the density of enterprise-level digitalization policies and regional green investment intensity were employed s instrumental variables (IVs) for DT. The second stage regressed SMP on the fitted values of RMO. he results are summarized in

Table 16. Causal identification results.

Variable/Model	Explained Variable	Key Explanatory Variable	Regression Coefficient	Standard Error	T-Value	p-Value	R2/Adj. R2	N
Instrumental variable method (2SLS)—Phase I	RMO	DT	0.527	0.067	7.865	<0.001	0.443	430
Instrumental variable method (2SLS)—Phase II	SMP	RMO (IV)	0.361	0.052	6.942	<0.001	0.398	430
Fixed-effect model	SMP	RMO	0.319	0.049	6.51	<0.001	0.376	430
Lagged variable models (Lag period I I)	SMP	RMO_t-1	0.284	0.047	6.043	<0.001	0.362	430
Lagged variable models (Lag period I II)	SMP	RMO_t-2	0.243	0.051	4.765	<0.001	0.329	430

, showing that both stages yield significant coefficients, confirming that the IVs are strongly correlated with the endogenous variable and exogenous to the error term.

By leveraging time-series characteristics, this study incorporates one-period and two-period lagged values of key variables into the regression analysis. The results show that the main path coefficients remain significant, indicating that the pathway from DT to RMO, and subsequently to SMP, exhibits both temporal logic and stability. To address potential endogeneity concerns, the 2SLS method was applied, using the density of enterprise-level digitalization policies and regional green investment intensity as IVs. The first-stage results confirm that the instruments are strongly correlated with the endogenous variables, while the second-stage coefficients remain robust, suggesting that the model’s explanatory power is driven by exogenous variation. Additionally, industry, region, and year fixed effects were introduced to control for unobserved heterogeneity. After accounting for these factors, the model’s overall fit improved and the core path relationships remained significant, further confirming the robustness and broad applicability of the proposed research model.

To address potential endogeneity, reverse causality, and unobserved heterogeneity in the “DT → RMO → SMP” pathway, this study applied a series of robustness checks: (i) lagged variable regression, (ii) instrumental variable estimation, and (iii) fixed-effect controls. These methods leverage temporal characteristics, exogenous variation, and structural constraints to validate the stability and causal interpretation of the model relationships.

Using time-series characteristics, the one-period and two-period lagged values of key variables are incorporated into the regression. Specifically, the main variables DT and RMO are replaced with their one-period lags to form the path Lag_DT → Lag_RMO → SMP. This specification tests whether the mediation relationship remains when contemporaneous correlations are eliminated.

Table 17. The path estimation results of robustness tests of the lagged variable method.

Path	Original Model	Lagged Variable
DT → RMO	0.336 **	0.321 **
RMO → SMP	0.372 **	0.379 **
Model fitting indicators (CFI/TLI/RMSEA)	0.919/0.905/0.041	0.911/0.898/0.039

Note: ** indicates p < 0.01.

presents the comparison between the original and lagged-variable models. The coefficient of Lag_DT on Lag_RMO is 0.321 (p < 0.01), and the coefficient of Lag_RMO on SMP is 0.379 (p < 0.01), both highly significant. Model fit indices (CFI = 0.911, TLI = 0.898, RMSEA = 0.039) remain within accepted thresholds. Although coefficients are slightly smaller than in the baseline model, their direction and significance are preserved, indicating clear temporal logic and reduced risk of reverse causality.

Given that enterprise DT adoption may be jointly influenced by unobserved factors such as managerial cognition or regional policy priorities, this study employed an exogenous instrumental variable for DT. The provincial digital infrastructure level (IV_DT)—computed as the average of the internet penetration rate and industrial digitalization index—was used as the instrument. Data were sourced from the China Digital Economy Development White Paper (2018–2022) and provincial statistical yearbooks.

First-stage regression results confirm strong relevance (F-statistic = 24.38 > 10). The Hansen over-identification test (p = 0.413) does not reject the validity of the instrument. As shown in

Table 18. Regression path estimation results of IVs.

Path	OLS Coefficient	2SLS Coefficient
DT→ RMO	0.336 **	0.318 **
RMO → SMP	0.372 **	0.361 **
F-value of the first stage	—	24.38
Hansen over-identification test (p-value)	—	0.413
Model-fitting index (Adj-R2)	0.366	0.351

Note: ** indicates p < 0.01.

, the 2SLS coefficients closely match the OLS estimates, with stable significance levels, indicating that the key pathway is driven by exogenous variation in DT.

To eliminate unobservable heterogeneity related to industry, region, and year, three sets of fixed effects were introduced. dummy variable matrix was constructed based on enterprise industry classification (e.g., vehicle manufacturing, battery, and electric control system), registration province, and year (2018–2022). These were added to the SEM to control for systematic variation from industrial characteristics, geographic location, nd macro-policy changes

Results in

Table 19. The path regression estimation results after controlling for fixed effects.

Path Relationship	Standardized Coefficient	Standard Error	z-Value	p-Value	Significance
DT → RMO	0.448	0.038	11.79	<0.001	***
RMO → SMP	0.398	0.044	9.05	<0.001	***
DT → SMP (direct)	0.265	0.049	5.41	<0.001	***
Indirect effect	0.178	0.021	8.48	<0.001	***

Note: *** indicates p < 0.001.

show that DT still exerts significant positive effect on MO (coefficient decreases slightly from 0.461 to 0.448, p < 0.001), and MO continues to significantly enhance SMP. The indirect effect DT → MO → SMP passes both Sobel and Bootstrap tests, confirming robust mediation.

Across all robustness strategies—lagged variables, IV estimation, and fixed-effect models—the “DT → RMO → SMP” pathway remains significant in both magnitude and direction. This consistency supports the temporal validity, exogeneity, and broad applicability of the model, reducing concerns over reverse causality and omitted variable bias.

3.6. Time-Series Analysis

To further explore the dynamic evolution of relationship among DT, RMO, and SMP, this study supplemented the original cross-sectional data with time-series analysis. Using consecutive annual reports from 86 listed NEV enterprises in China spanning 2018 to 2022, a five-period panel dataset was constructed (n = 86 × t = 5, totaling 430 observations). A fixed-effects panel model was applied to identify the causal and lagged effects among the three core variables over time. The results are summarized n

Table 20. Fixed-effects model results.

Variable	Coefficient Estimate	t-Value	Significance
DT	0.231	4.23	***
RMO	0.388	5.11	***

Note: Industry and year fixed effects are controlled. *** indicates p < 0.001.

:

From a dynamic perspective, investment in DT improves firms’ market sensing and responsiveness in the short term, which subsequently influences SMP through strategic actions—demonstrating a time-lagged “capability mediation effect.” Additionally, sustained enhancements in RMO exert a long-term positive impact on SMP, underscoring its strategic significance in the context of green consumption. This analysis strengthens the temporal foundation of the proposed mechanism model and provides a data-driven basis and methodological framework for future research addressing nonlinear evolution and policy shock responses in time-series contexts.

4. Discussion

This study uses the NEV industry as a case to explore how enterprises enhance SMP through RMO during the process of DT. The SEM analysis confirms the proposed path hypotheses. The results reveal that the effect of DT on RMO (standardized coefficient = 0.482) is significantly stronger than its direct effect on SMP (0.267), while RMO exerts a significant positive influence on SMP (β = 0.396). This indicates that RMO plays a crucial mediating role between DT and performance outcomes. Further quantitative decomposition shows that 41.7% of the total effect of DT on SMP operates through this indirect pathway, emphasizing the pivotal role of RMO in transforming technological investments into market performance.

From a managerial perspective, these findings provide strategic insights. Rather than treating technology investments in isolation, enterprises should embed them within an integrated “technology–mechanism–performance” framework. Firms can leverage intelligent systems to collect user data and optimize product feedback via Over-The-Air updates. They can also enhance green traceability and compliance through digital supply chains. This approach establishes a closed-loop mechanism to upgrade the green value chain. Senior management must acknowledge the importance of mechanism-building to avoid the misconception of “technological omnipotence.”

Subgroup regression analysis reveals structural heterogeneity across enterprise types. Main board firms, characterized by higher technological maturity, tend to convert technology investments directly within existing governance frameworks. Growth-oriented enterprises rely more heavily on RMO mechanisms to enhance organizational responsiveness. This divergence suggests that firms should consider their resource endowments and developmental stages when formulating transformation strategies. While the current study models the relationships using a linear framework, theoretical insights and practical observations imply a possible nonlinear relationship between DT and RMO. For instance, in the initial stages of DT, mechanism activation effects may be limited, whereas at higher DT levels, diminishing marginal returns or mechanism saturation might emerge. Although quadratic terms or threshold variables were not introduced for empirical testing here, recognizing this potential nonlinearity offers a valuable direction for future research.

Regarding causal inference, despite employing IVs and robustness tests to strengthen the model’s explanatory power, residual endogeneity among DT, RMO, and SMP cannot be entirely ruled out. For example, firms’ existing responsiveness or performance levels may influence DT investment decisions, and the maturity of RMO mechanisms may affect DT implementation effectiveness. Future studies could enhance causal identification precision by incorporating dynamic panel data models or difference-in-differences approaches. Additionally, some variables in this study are constructed based on semantic frequency analysis from text mining. Although standardization and manual coding procedures were applied to enhance consistency, measurement bias due to writing style variations may persist, especially across different industries where textual conventions vary. Thus, for cross-industry or cross-lingual applications, integrating external labels or deploying deep semantic classification techniques is recommended to improve variable validity. Finally, the sample focuses on Chinese A-share listed NEV enterprises. While this sample is highly representative in its context, limitations exist regarding international generalizability and applicability to SMEs. Differences in institutional environments, market orientations, and digital infrastructures across countries and industries may affect the transferability of the DT–RMO–SMP framework. Future research could explore this mechanism’s applicability by incorporating manufacturing and service industries in Europe, North America, or other regions through multi-case studies or qualitative interviews.

In summary, this study contributes theoretically by extending the structural pathway linking sustainable marketing and digital transformation, while offering practical guidance for organizational mechanism design in corporate green strategies. However, limitations related to model assumptions, sample scope, and variable construction provide fruitful avenues for further theoretical refinement and cross-context validation.

5. Conclusions

Driven by the dual forces of the “dual carbon” strategy and DT, certain green consumer goods industries are experiencing a fundamental restructuring of their marketing logic and performance evaluation systems. This study focuses on A-share listed companies in the NEV industry and constructs a three-variable path model of “DT—RMO –SMP.” Utilizing annual report texts and financial data, SEM analysis reveals that DT exerts a significant positive effect on SMP. Through the integrated application of digital technologies such as AI, big data, and cloud computing, enterprises not only enhance internal management efficiency but also improve their capability to respond to green consumer preferences. RMO plays a pivotal mediating role between DT and SMP, indicating that firms can amplify the performance-enhancing effects of digital technology by establishing rapid, green value–oriented response mechanisms, thereby forming a systematic pathway: “technology-driven → mechanism transformation → performance improvement.”

Several limitations constrain the generalizability of these findings. First, the study does not systematically incorporate external moderating variables such as firm size, market maturity, or policy support strength. These factors may cause context-dependent variations in the observed relationships. Second, the cross-sectional data structure limits consideration of bidirectional causality and dynamic lag effects. Third, text-mining variables may contain measurement bias due to writing style variations across industries. Fourth, the sample focuses exclusively on Chinese A-share listed NEV enterprises, limiting international generalizability and applicability to SMEs.

In conclusion, despite limitations regarding model applicability and variable measurement, this study provides a theoretical foundation and empirical evidence for understanding the relationship between enterprise digital transformation and green market responsiveness within the green consumer goods industry. The “capability–mechanism–performance” framework constructed herein offers a practical measurement pathway and a basis for mechanism expansion in future empirical research.