Influence Mechanism of Data-Driven Dynamic Capability of Foreign Trade SMEs Based on the Perspective of Digital Intelligence Immunity

Abstract

Against the backdrop of digital transformation, this study constructs an analytical framework for the influence mechanism of the data-driven dynamic capabilities of foreign trade SMEs from the perspective of digital intelligence immunity, aiming to clarify the complex relationships among influencing factors and multi-combination paths for capability improvement. The research employs the fuzzy AHP-DEMATEL method to quantify the complex influence relationships among factors and uses fsQCA to analyze the configuration paths of high-level data-driven dynamic capabilities. Results show that digital intelligence management and analysis, digital intelligence supervision and early warning, and digital intelligence ecosystem are key drivers of data-driven dynamic capabilities, with digital intelligence talents serving as a guarantee and digital foundation as a foundation. The study identifies the following two core paths for forming high-level capabilities: “management–talent–ecology collaboration” and “early warning–technology–mechanism enhancement.” It concludes that foreign trade SMEs should strengthen digital intelligence management and ecological construction, improve early warning mechanisms, and adopt multi-pronged approaches to build data-driven dynamic capabilities, providing a theoretical basis for their digital transformation and capability upgrading.

1. Introduction

Small- and medium-sized enterprises (SMEs) in foreign trade serve as a crucial cornerstone for a country’s economic development and foreign trade [1]. In 2023, China’s foreign trade industry achieved a historic breakthrough, with the number of foreign trade operators having import–export records exceeding 600,000 for the first time. Among them, private enterprises, with a huge number of 556,000, became the main force, with a total import–export volume of 22.36 trillion CNY throughout the year, accounting for 53.5% of the country’s total import–export value, demonstrating strong market vitality. As a core component of private enterprises, SMEs account for as much as 90%. It can be seen that SMEs occupy a pivotal position among China’s foreign trade operators. In the current wave of digital economy sweeping the globe, these foreign trade SMEs have keenly captured opportunities, broken through the time–space limitations of traditional trade, and obtained a large number of trade opportunities by building online exhibition and sales platforms, using big data analysis to accurately match customers, and expanding overseas markets through cross-border e-commerce platforms [2].

However, the development of foreign trade by foreign trade SMEs is not smooth sailing. In this dynamic environment, the sustainable development of foreign trade SMEs faces severe challenges. From the perspective of the external environment, the uncertainty of the international market has significantly increased. The “2025 China Export Risk Index Report for Small, Medium, and Micro Foreign Trade Enterprises” [3] shows that due to the complex changes in the international trade environment and the rise of payment risks of overseas enterprises, the overall credit risk faced by China’s small, medium, and micro foreign trade enterprises in export has continued to rise in the past three years. From the perspective of internal conditions, foreign trade SMEs are easily restricted by the limitations of resource endowment, the low maturity of digital transformation, the structural shortage of professional talents, and the asymmetry of market information. The “2024 China SME Digital Transformation Report” [4] points out that nearly 40% of SMEs face tight funds (36.8%), more than half of SMEs lack professional talents (50.8%), and more than 60% of SMEs are still in the early stage of transformation (62.6%). Although the use of digital technology helps to enhance the sustainable development capacity of enterprises, SMEs often become victims in the digital wave due to the lack of sufficient technical and financial investment [5].

Based on the dynamic capability theory, enterprise digital transformation can enhance perception, decision making, and reconstruction capabilities, thereby promoting enterprise sustainable development [6]. Foreign trade SMEs can not only track the search popularity of keywords in emerging markets and public opinion on social media through AI big data analysis tools (such as Google Trends and SEMrush) to identify unmet segmented needs [7] but also improve production efficiency and innovation capabilities by introducing an MES (manufacturing execution system) to achieve digital management of production lines; they can not only open stores on platforms such as Alibaba International Station, Amazon, and Shopee and reach consumers in emerging markets by using platform traffic support policies [8] but also build independent websites (such as through Shopify) and accurately attract traffic by combining SEO and social media advertising (Facebook, TikTok) [9]. It is worth noting that the digital transformation of enterprise elements drives the reform of the constituent elements of dynamic capabilities. Due to the lack of systematic deconstruction in the interaction mechanism between parameters such as talents, culture, and technology in the digital scenario and enterprise dynamic capabilities, especially the failure to reveal the collaborative effect path of multi-dimensional parameters, there is an obvious explanatory gap in theoretical research when guiding the digital sustainable development of SMEs. In the above real-world scenarios, the construction of dynamic capabilities of foreign trade SMEs faces new requirements, and the current construction of enterprise dynamic capabilities urgently needs to focus on the systematic construction of crisis resistance.

2. Theoretical Background

2.1. New Dynamic Capability Driven by Data

Unlike general capabilities for explicit tasks, dynamic capabilities enable organizations to sense changes, deliberate strategies, and reconfigure resources for adaptation in turbulence [10]. It helps enterprises gain sustainable competitive advantages in volatile and complex environments [11]. McAfee and Brynjolfsson (2012) proposed the concept of “data-driven enterprises” [12,13]. Driven by the development of big data storage, analysis, and processing technologies, enterprises are gradually shifting to mobile data-based data-driven decision making. Specifically, it refers to enterprises’ automatic decision making based on data analysis [13]. The relationships among digitization, dynamic capabilities, and data-driven decision making are shown in Figure 1. Dynamic capability theory suggests digital technologies enhance enterprises’ perception, decision making, and reconstruction capabilities and boost data efficiency.

A comprehensive literature analysis identifies gaps in data-driven capability research from the dynamic capability perspective, as follows: (1) Though dynamic capabilities have a three-dimensional framework, scale development is scarce, lacking a unified indicator system [14]. (2) Digital integration transforms organizational capabilities and structures, requiring new digital variables in dynamic capability frameworks. (3) Data-driven decision-making mechanisms need further analysis. Existing studies rarely clarify data-driven capability formation paths, despite strategic process analyses following the sensing, decision making, and reconstructing principles of dynamic capabilities. (4) High-level data-driven capabilities result from multiple factors. Yet, current research relies on theoretical or traditional statistical analyses, hardly revealing conditional combinations and causal complexity. (5) Existing studies show limited research on SMEs’ core competitiveness in the digital era [15,16]. Research on data-driven dynamic capabilities of foreign trade SMEs also needs strengthening.

2.2. New Dynamic Capability of Foreign Trade SMEs

Digitization enhances enterprises’ dynamic capabilities to perceive opportunities, support data-driven decision making, and restructure resources. Unlike pure data analysis, data-driven decision making covers activities based on automated data analysis [13]. Compared with large enterprises, SMEs have disadvantages, such as weak innovation, low professionalism, small scale, and scarce resources. In uncertain international environments with fierce digital competition, data-driven effects matter greatly for foreign trade SMEs. Data alone is insufficient; clarifying data-driven mechanisms is key to building high-level capabilities, with dynamic capabilities offering theoretical support [12].

Based on dynamic capability-based research, this paper follows dynamic capabilities’ logic, including perception, decision, and refactoring. Combined with prior studies, it summarizes micro-factors of data-driven decision making in foreign trade SMEs (

Table 1. Research on new dynamic capability of foreign trade enterprises.

Stage	Action	Representative Document
Perception/search	Digital intelligence transformation enables enterprises to perceive the dynamic and complex international competitive environment with higher precision	Koch [17]
	Use the “Internet + government Service” platform to improve the ability to obtain and interpret enterprise policy information	Chen [18]
	The use of digital platform improves the availability of information such as local market, transaction history, and behavior of transaction objects, and reduces the uncertainty of cross-border transactions	De [19]
	The use of digital products is beneficial for enterprises to interact with suppliers and customers, and helps enterprises track, record, and master consumers’ behaviors and preferences in time	Nambisan [20]
Decisions/choices	Using cloud computing and big data to process foreign consumer behavior data and analyze their behavior characteristics	Yoon [21]
	Using digital marketing to reduce overseas sales costs and export risks	Coviello [22]
Refactoring/configuration	The construction of cross-border digital platform breaks the limitation of geographical and cultural distance	Kim [23]
	Adapt to the host country environment better	Antra S [24]
	Promote the transformation of digital, intelligent, and precise digital corporate culture	Shah [25]
	Enhancing the position of enterprises in the value chain	Liu [26]

). In perception/search, digital technologies help foreign trade SMEs acquire diverse information and accumulate knowledge for international business [17,18,19,20]. In decisions/choices, big data, cloud computing, and AI help overcome limited experience, boosting decision efficiency and scientificity [21,22]. In restructuring/configuration, digital technologies significantly affect foreign trade enterprises’ structure, processes, operations, and culture [23,24,25,26]. Digital technologies also ease “low-end lock-in” from scale constraints, aiding global value chain position reshaping [26]. Compared with large foreign trade enterprises, SMEs are smaller with flatter structures, boasting stronger reshaping and transformation capabilities. This enables rapid internal restructuring, resource integration, and business innovation via digital means.

Enterprise sustainable development includes the following three dimensions: economic, social, and environmental. In digital transformation, foreign trade SMEs’ new dynamic capabilities link operations and sustainable development via data-driven mechanisms. Economically, new dynamic capabilities boost resource efficiency and competitiveness via data-driven decisions and digital technologies. Socially, new dynamic capabilities help break resource constraints, create social value, and optimize employment structures. Environmentally, new dynamic capabilities are used to perceive markets and promote ecological sustainability via resource-intensive models. Thus, building data-driven dynamic capabilities relates closely to enterprises’ sustainable competitive development. Thus, based on existing dynamic capability definitions and foreign trade enterprises’ digital transformation, this paper defines foreign trade SMEs’ data-driven new dynamic capabilities as the ability to gain sustainable competitiveness in dynamic/uncertain environments via digital technologies and flowing data.

2.3. Tissue Resilience Theory

Organizational resilience, a special dynamic capability formed under sustained stress and disruption of adverse events, serves as one guiding philosophy for enterprises to overcome complex dynamic environments [27]. Current analyses of enterprise resilience mainly adopt the following two perspectives: embeddedness and process. (1) From the embeddedness perspective, existing studies show that influencing factors of enterprise resilience can be roughly categorized into soft factors (cognition, emotion, social relations) and hard factors (resources, architecture, strategy). Soft parameters of organizational resilience from this perspective cover cognition, emotion, and social relations, such as employees’ cognitive and emotional capabilities, organizational social network, and leadership level; hard parameters involve resources, architecture, and strategy, including financial slack, backup equipment, resilient resources, decentralized organizational structure, and enterprise strategic planning. (2) From the process perspective, research on enterprise resilience factors can be divided into the following three stages: pre-event, mid-event, and post-event. In the pre-event stage, organizational resilience is manifested as the accumulation of enterprise resources and capabilities, including technology, information, interpersonal relationships, and enterprise structure. In the mid-event stage, it is reflected in the organization’s response and adjustment capabilities, namely the countermeasures taken under new scenarios based on pre-event preparations. In the post-event stage, organizational resilience is demonstrated as enterprise improvement and overtaking. That is, enterprises can learn from adversity through learning and memory, absorbing knowledge and experience to achieve upgrading.

Organizational resilience theory offers unique advantages for constructing data-driven dynamic capabilities. It organically integrates soft elements (cognition, emotion) and hard elements (resources, architecture) from the embeddedness perspective to form a capability system combining rigidity and flexibility. Meanwhile, the three-stage evolutionary logic of the process perspective provides a time-sequenced path for the full-cycle cultivation of dynamic capabilities, enabling enterprises to achieve capability leaps from passive response to active evolution in complex environments. However, although organizational resilience can achieve recovery and improvement after crises, it lacks the systematic construction of pre-event defense mechanisms in dynamic capabilities, failing to establish hierarchical response mechanisms, and deeply embed digital intelligent technologies in the full process of defense, monitoring, memory, etc.

2.4. Tissue Immunity

2.4.1. Tissue Immune Theory

Organizational immunity theory originates from inspiration by biological immunity theory [28]. Drawing on biological immunity concepts, organizational immunity enables enterprises to dynamically identify internal/external disturbances, eliminate threats, and retain adaptive memory for sustained health. Related research on organizational immunity is shown in

Table 2. Tissue immunity-related research.

Research Perspective	Tissue Immunity	Representative Document
System structure	Full-time immunization organizations and part-time immunization organizations	Wan Yihua [28]
	Central immune system, professional immune system, and peripheral immune system	Zhao Jianbo [29]
	Practical artifacts, organizational operating practices, and enterprise strategic dynamic capability system	Jiang Tao [30]
Response mechanism	There are two lines of defense of non-specific immunity and specific immunity, and four stages are recognition, feedback, variation, selection, and memory	Wang Yihua [28]
	It includes three dimensions: organizational cognition, organizational defense, and organizational memory	Lv Ping [31]
	There are two dimensions of specific immunity and non-specific immunity. Non-atopic immunity includes organizational structure, institutional rules, and organizational culture, while specific immunity includes organizational surveillance, organizational defense, and organizational memory	Lv Ping [32]
	Routine–cyclic interaction mode of immune response	Jiang Tao [30]
	Automatic implementation of non-specific, extensive-specific, and unique-specific responses at corresponding levels according to dissident intensity	Jiang Tao [33]
	It makes up for the vacancy that the traditional immune system is defaulted to be a system that can automatically transmit information and analyze causality and emphasizes embedding safety construction content directly in design and production	Ni Qing [27]
Systematic elements	It has the dual characteristics of psychological immunity and behavioral immunity	Ni Qing [34]

. With the continuous deepening of theoretical studies, the structural dimensions, constituent elements, and response mechanisms of organizational immunity theory are constantly being improved and developed [27,28,29,30,31,32,33,34]. Scholars generally agree that the organizational immune system can be divided into the following three subsystems: the central immune system, the full-time immune system, and the peripheral immune system. There are also the following two lines of defense: specific immunity and non-specific immunity. Specific immunity generally includes monitoring, defense, and memory functions, while non-specific immunity comprises organizational structure, institutional rules, and organizational culture. In the digital economy era, organizational immunity has ushered in a leap of digital intelligence, emphasizing that enterprises directly use new-generation digital technologies (such as artificial intelligence, big data, cloud computing, and digital twin) to embed security mechanisms into design and production processes [27].

Both organizational resilience and organizational immunity center on addressing external shocks and uncertainties, aiming to help organizations survive and develop in complex environments. They are complementary theoretical perspectives; resilience is the manifestation of results, while immunity is the support of capabilities. The former focuses on how to survive in crises, and the latter focuses on how to build a crisis-resistant system. Together, they constitute the core framework for organizations to handle uncertainties. The former emphasizes post-event recovery, while the latter highlights full-cycle capability building of pre-event defense + mid-event response + post-event repair. Their differences are specifically manifested, as follows: (1) Different conceptual connotations. Organizational resilience emphasizes an organization’s ability to return to normal after being impacted or to adapt and maintain development in adverse environments, focusing on resilience and recovery speed at the result level. Organizational immunity uses the system capability framework of the human immune system. By identifying and responding to outsiders (threats or opportunities), it achieves organizational self-repair and dynamic balance with a stronger emphasis on defense mechanisms and proactive responses during this process. Organizations need to have the ability to make continuous responses to ongoing changes. (2) Different theoretical perspectives. Research on organizational resilience often focuses on state changes before and after crises, paying attention to crisis losses and recovery time, featuring a combination of static and dynamic characteristics. Starting from the positioning of system capabilities, organizational immunity regards the organization as an organic whole composed of a central system, specialized systems, and peripheral systems. It emphasizes the collaborative role of each subsystem in identifying and eliminating dissenters, as well as the dynamic balance between immunity and environmental changes. (3) Different goals. The goal of organizational resilience is to reduce crisis damage and recover quickly, such as restoring production efficiency after a pandemic. The goal of organizational immunity is to establish long-term defense mechanisms. By enhancing immunity (such as corporate culture, leadership, employee quality, etc.), it proactively prevents and responds to various dissents to achieve the sustainability of organizational health. The perspective of digital intelligence immunity can precisely make up for the gap in organizational resilience, providing a more forward-looking “immunity–response–evolution” closed-loop framework for the construction of data-driven dynamic capabilities and promoting the upgrading of enterprise dynamic capabilities from adaptive adjustment to preventive immunity.

In order to maximize the applicability of tissue immune theory, this paper extracts the best parts of existing research to form the latest tissue immune system, and the results are shown in

Table 3. Tissue immune system.

Immune Subsystem	Immune Pathway	Functions	Content
Immune central system	Unique-specific immunity	Strategic planning, learning, and memory	Supervision, defense, memory
Full-time immunization system	Extended-specific immunity	Directly incorporate safety compliance construction into the digital system
Immune peripheral system	Non-specific immunity	Complete self-control and feedback	Structure, system, culture

. The tissue immune system consists of three systems and two lines of defense. On the basis of the two immune paths, this paper adopts the research of Jiang Tao et al. (2017) [30], categorizing immune pathways into strategic-level specific immunity, operational-level extended specific immunity, and non-specific immunity based on an enterprise’s conventional autonomous mechanisms. These three immune pathways correspond to different levels of immune subsystems. The immune system can automatically carry out immune response at the corresponding level according to the intensity of internal and external dissidents, thereby forming a cyclical interaction between organizational conventions and immune responses.

2.4.2. Characteristics and System Construction of Digital Intelligence Immune

(1)

Digital Intelligence Immune Characteristics

Currently, digital intelligence immunity possesses the following four characteristics: intelligence, ecologization, integrity, and interactivity.

In terms of intelligence, enterprises can establish a digital intelligent twin system penetrating all business links, drive business processes through digital foundation, and use new technologies like big data, cloud computing, and AI to provide information and decision support for management activities, such as supervision warning and prediction planning. This makes up for the gap in traditional immunity theory where the enterprise immune system is assumed to automatically conduct causal analysis and store/manage massive data experience, enhancing business efficiency and promoting system self-control feedback [35].

At the ecologization level, given the close connections among stakeholders (consumers, producers, etc.) and prominent eco-economic features in the digital economy era, organizational immunity draws on interspecific cooperation and symbiosis in biology to improve the environment, proposing the concept of digital intelligent collaborative immunity. Enterprises reduce the impact of internal/external dissenters by establishing partnerships with relevant stakeholders [35].

In terms of integrity, digitization strengthens internal enterprise connections. Digital information technology enables more timely and accurate information transmission/feedback, promoting employee workflow coordination. Enterprises can also implement internal control through new-generation digital technologies—AI for monitoring employee behavior and big data for acquiring dynamic risk models—thus proposing digital intelligent communication and tracking to supplement the organizational immunity theoretical system [36,37].

Concerning interactivity, digital intelligent interconnection technologies enhance enterprise–external links and open up influence dissemination channels. Aiming at the deficiency that traditional organizational immunity theory does not consider external enterprise interactions, it proposes digital intelligent output to supplement the theoretical system. Enterprises can use digital technologies for intelligent push, enhancing resistance by conveying positive values and establishing a good public image [38].

(2)

Digital Intelligence Immune System

Based on the above summaries and expansions, this paper further improves the enterprise digital intelligent immune system in the context of the digital economy, as shown in Figure 2. The digital intelligent immune system consists of three subsystems, each playing a distinct role.

The digital intelligent immune central system is responsible for strategic planning, learning, and memory, serving as the leadership core of the organizational immune system. It includes the following three components: digital intelligent management and analysis, digital intelligent learning and memory, and digital intelligent talent. Enterprises obtain decision support for the entire process of management activities—prediction and planning, execution and control, assessment and evaluation—through digital intelligent management and analysis. They accumulate experience and knowledge through digital intelligent learning and memory. The central system is the primary site for providing unique and specific responses.

The digital intelligent immune full-time system is a complex relational network with security construction anchors distributed throughout. It comprises the following four components: digital intelligent ecosystem, digital intelligent supervision warning, digital intelligent communication and tracking, and digital intelligent output. Using this network, enterprises can directly integrate relevant security compliance and audit tasks into the digital intelligent capability construction system through new-generation digital technologies (such as digital twin, code vaccine, cloud security testing, etc.), while conducting real-time monitoring of equipment, data, and employees. The full-time system is the main site for providing extensive and specific responses.

The digital intelligent immune peripheral system includes the following four conventional components: digital foundation, employee quality, rules and regulations, and digital intelligent culture. According to established conventions, departments such as product R&D, marketing, procurement supply and production, and processing operate autonomously to achieve self-control and feedback. The peripheral system is the primary site for providing non-specific responses.

The digital intelligent immune system activates corresponding levels of immune responses based on the severity of intruders, forming a cycle of interaction between conventions and immune responses. Once changes in the internal or external environment are detected, enterprises first respond through non-specific immune reactions of conventional components. At this stage, enterprises only use existing traditional components (i.e., “trial and error”) to resolve dissent without modification. When existing non-specific immune efforts resist external attacks, the system immediately activates specific immune response activities. The organization first assesses the impact level of the intruder. If the change’s impact is limited to the organizational operation level, enterprises will respond only through extensive specific immune reactions at the operational level. If it also involves the organizational strategic level, enterprises will simultaneously activate unique specific immune reactions at the strategic level. In the specific analysis and implementation process, the organization will witness the “birth” of new practices or the “mutation” of old ones. Finally, through learning and memory, new conventions are internalized into conventional components, preparing for the re-invasion of the same intruders and triggering non-specific immune reactions.

2.5. New Dynamic Ability from the Perspective of Digital Intelligence Immunity

Organizational immunity is a type of dynamic capability [27]. Monitoring, defense, learning, and memory in organizational immunity theory align with the sensing, decision making, and reconfiguring logic of dynamic capabilities. Both theories focus on adapting to environmental changes for competitive advantage. Previous research has confirmed their compatibility and developed a multi-level organizational immune response model [30]. With the in-depth development of the digital economy, enterprise dynamic capabilities have gradually evolved into digital-driven dynamic capabilities. However, through the literature review, it is found that current research on new dynamic capabilities of enterprises still has obvious deficiencies, as follows: (1) Although there is a three-dimensional framework, scale development is insufficient, and there is a lack of a unified indicator system. (2) Digital integration has promoted the need to incorporate new digital variables into dynamic capability frameworks. (3) Data-driven decision-making mechanisms need further analysis, and existing studies rarely clarify the formation paths of data-driven capabilities. (4) High-level data-driven capabilities result from multiple factors, but current research hardly reveals conditional combinations and causal complexity. Digital intelligence immunity provides a highly appropriate research perspective for the study of new data-driven dynamic capabilities of enterprises, which can effectively make up for the above research gaps.

First, the immune subsystem supports the implementation of dynamic capabilities, as follows: (1) The digital intelligence immune system is conducive to expanding the scope of surveillance. The intelligent monitoring system scans the external environment in real time, such as technological breakthroughs and policy changes of competitors, to enhance the timeliness and accuracy of perception. The peripheral system of the immune subsystem cultivates employees’ digital sensitivity, enabling front-line teams to proactively identify minor changes in the market. (2) The digital and intelligent immune system is conducive to optimizing the efficiency of resource integration. Digital tools, such as data middle platforms and AI algorithms, accelerate resource integration and shorten the response cycle from perception to action. (3) It is conducive to establishing an immune-driven repair mechanism. For instance, we can code the immunization rules and embed them into digital business processes to achieve process-oriented immunization capabilities. Or through the memory function of the immune subsystem, the experience of responding to a single crisis can be transformed into organizational-level capabilities.

Second, the digital intelligence immunity perspective can provide theoretical support for constructing an indicator system of data-driven dynamic capabilities. The digital intelligent immune system includes the following three subsystems: the central immune system, the full-time immune system, and the peripheral immune system. According to the hierarchical logic of this system, this paper divides the data-driven dynamic capability operation system into the following three subsystems: the digital intelligent peripheral system, the digital intelligent full-time system, and the digital intelligent central system. Each subsystem has different functions and can automatically perform non-specific processing, extensive specific processing, or unique specific processing at different levels according to the severity of opportunities or crises.

In the response mechanism model, the Talents element at the strategic level is the core kinetic energy driving the process. When entering response execution, if conventional products fail and flow to the Scan–Analysis stage, talents’ cross-domain collaboration and problem-solving capabilities become crucial. (1) Operational talents collaborate with technology and market teams to quickly sort out the Ecosystem resource adaptation path and promote adjustments in the Output link. (2) Management talents ensure the smooth transmission of Transmission and Attacking instructions through efficient communication, making counterattack strategies accurately implemented and avoiding process stuck. (3) In the Learning and Memory link at the backend of the capability evolution, talents undertake the responsibilities of knowledge extraction, coding, and inheritance, refining tacit experience from immune practices (such as decision-making logic for special threat responses) into explicit knowledge storable in the Database. Precipitating immune capabilities as the strategic memory of the entire organization through training and case review supports rapid adaptation in subsequent responding again.

Third, the immune consensus logic of digital intelligence immunity and the convention–immune response cycle interaction model help clarify the internal mechanisms of enterprise data-driven capabilities from the dynamic capability perspective. The mechanism of enterprise data-driven dynamic capabilities is illustrated in Figure 3. When abnormal data is detected, enterprises first respond through conventional components. At this stage, enterprises address anomalies solely via “trial and error” without modifications. When existing conventional components fail to achieve effective responses, the system immediately initiates digital intelligence analysis to assess the impact level. If the environmental change affects only the organizational operation level, the enterprise will activate only extensive specific responses at the operational level; if the impact extends to the strategic level, unique specific responses at the strategic level will be simultaneously activated. Through experience summarization, learning, and memory, the system stores decision information and corresponding scenarios in the database, forming new conventions to improve the efficiency of responding to the same scenarios in the future.

2.6. Literature Evaluation

Digitization can coordinate the tension among an organization’s triple performance (economic, social, and environmental), facilitating corporate sustainable development. In the context of digital transformation, the data-driven dynamic capabilities of foreign trade SMEs, enabled by digital technologies and data-driven mechanisms, serve as the core link connecting enterprise operations with sustainable development goals. Current research on enterprises’ data-driven dynamic capabilities has the following deficiencies: (1) The indicator system for data-driven dynamic capabilities has inherent flaws, necessitating the reintroduction and in-depth development of digital variables. (2) The sensing, decision making, and reconstructing logic from the strategic perspective of dynamic capabilities fails to reveal the full picture of data-driven dynamics, with internal mechanisms of data-driven dynamic capabilities requiring further exploration. (3) Data-driven dynamic capabilities result from multiple factors, demanding conditional configuration analysis. (4) Research on the core competitiveness of SMEs remains extremely limited, urgently calling for strengthened research on the data-driven new dynamic capabilities of foreign trade SMEs.

Studies have found that the digital intelligence immunity perspective provides a highly suitable framework for researching the data-driven new dynamic capabilities of foreign trade SMEs: (1) Organizational immunity theory and dynamic capability theory are highly compatible. The digital intelligent development of organizational immunity can address existing research gaps in data-driven new dynamic capabilities, helping to strengthen crisis scenario control in dynamic capability construction. (2) It offers theoretical support for constructing an indicator system of data-driven dynamic capabilities. (3) Its immune consensus logic and convention–immune response cycle interaction model clarify the internal mechanisms of enterprises’ data-driven capabilities from the dynamic capability perspective.

Therefore, this study will make the following contributions: (1) Construct an indicator system for the data-driven dynamic capabilities of foreign trade SMEs from the digital intelligence immunity perspective. (2) Evaluate the influencing factors of new dynamic capabilities in foreign trade SMEs using the Analytic Hierarchy Process (AHP) and fuzzy Decision-Making Trial and Evaluation Laboratory (fuzzy-DEMATEL). (3) Adopt fuzzy-set Qualitative Comparative Analysis (fsQCA) to analyze the multiple formation paths of high-level data-driven new dynamic capabilities.

The main contributions of this paper are as follows: (1) Constructing an indicator system for the new dynamic capabilities of foreign trade SMEs, deepening the understanding of influencing factors and internal mechanisms and promoting the sustainable development of dynamic capability theory. (2) Expanding organizational immunity theory and further improving digital intelligence immunity research by integrating the development trend of the digital economy. (3) Providing hierarchical, systematic, and multi-path insights for foreign trade SMEs to build high-level data-driven dynamic capabilities, addressing the deficiency in research on the core competitiveness of foreign trade SMEs in the digital economy.

3. Indicator Construction

Based on the above analysis and expert interviews and referring to the “SME Digital Level Evaluation Indicators (2024 Edition)” [39], this study constructs an indicator system for data-driven dynamic capabilities of foreign trade SMEs, including 3 first-level indicators and 11 second-level indicators, as shown in

Table 4. New dynamic capability index system of foreign trade SMEs.

First-Class Index	Secondary Index	Source of Literature
A1 Digital Intelligence Central System	B1 Digital Intelligence Management and Analysis	Yoon [21]
	B2 Digital Intelligence Memory	Shah [25]
	B3 Digital Intelligence Talents	Pisar [35]
A2 Digital Intelligent Full-time System	B4 Digital Intelligence Communication and Tracking	Wamba [36]
	B5 Digital Intelligence Supervision and Early Warning	Banker [37]
	B6 Digital Intelligence Ecosystem	Nambisan [20]
	B7 Digital Intelligence Output	Sajko [38]
A3 Digital Intelligence Autonomous System	B8 Digital foundation	Pisar [36]
	B9 Staff Quality	Wang [28]
	B10 Rules and Regulations	Wang [28]
	B11 Digital Intelligence Culture	Wang [28]

.

Digital Intelligent Central System: Digital Intelligent Management and Analysis covers planning management, element guarantee, and information security. It involves using big data, AI, and other technologies to conduct full-process management and analysis of enterprise operation data, providing decision support for management activities, such as strategic planning, execution control, and evaluation. This serves as a key driving factor for data-driven dynamic capabilities. Digital Intelligent Memory focuses on the digital storage and iteration of an enterprise’s historical experience and knowledge. By accumulating data assets, it forms organizational learning capabilities to support the continuity of dynamic decision making. Digital Intelligent Talent refers to talent groups with digital technology application, data analysis, and intelligent decision-making capabilities. They are an important force to ensure the implementation of data-driven dynamic capabilities, promoting the deep integration of technology and business.

Digital Intelligent Full-Time System: Digital Intelligent Supervision and Early Warning leverages technologies, like digital twin and AI monitoring to real time, scan market risks, internal vulnerabilities, and other situations in the internal and external environments, achieving dynamic early warning and response. This is a key defense mechanism in the capability system. Digital Intelligent Communication and Tracking optimizes internal information transmission and process coordination through digital tools while tracking the dynamics of external stakeholders, thereby improving system response efficiency and collaboration capabilities. Digital Intelligent Ecosystem is a digital collaborative network constructed by enterprises with stakeholders such as consumers, suppliers, and platforms. It reduces external risks and enhances enterprises’ dynamic adaptability through data sharing and cooperative symbiosis. Digital Intelligent Output relies on digital platforms (such as independent websites and social media) to output brand values and data insights, enhancing enterprise influence. This is an important link in outward-oriented immune responses.

Digital Intelligent Peripheral System: Digital Foundation includes network construction, equipment digitization, equipment networking, data collection, information systems, network security, and data security. These items evaluate the enterprise’s digital foundation level, providing basic support for data-driven dynamic capabilities. Employee Quality includes employees’ digital literacy, professional skills, and innovative awareness. Organizational application capabilities of digital technologies are improved through training and cultural infiltration. Rules and Regulations refer to management systems and process specifications related to digital transformation, such as data security systems and intelligent decision-making processes, to ensure the standardization and stability of organizational operations. Digital Intelligent Culture cultivates a data-driven decision-making culture and innovative atmosphere, promoting the deep integration of organizational routines and digital technologies to form sustainable driving forces for digital transformation. The index system comprises 3 first-level indicators and 11 second-level indicators, as presented in Table 4.

4. New Dynamic Capability Evaluation Method for Foreign Trade SMEs

4.1. Research Method Framework

Based on its high compatibility with the analysis of complex causal relationships in dynamic capability theory, this paper selects the following three methods for the research: Analytic Hierarchy Process (AHP), Fuzzy Decision-Making Trial and Evaluation Laboratory (fuzzy-DEMATEL), and Fuzzy-set Qualitative Comparative Analysis (fsQCA).

Firstly, this paper proposes an integrated approach to evaluate the comprehensive weights of each factors in new dynamic capability evaluation of foreign trade SMEs [40]. We use AHP to transform the subjective judgments of decision makers into quantifiable weights. This is a method that breaks down complex decision-making problems into multiple levels and factors and conducts decision analysis through a combination of qualitative and quantitative approaches. As a structural model analysis method, fuzzy-DEMATEL overcomes the limitations of traditional DEMATEL by introducing triangular fuzzy numbers. Through the reconstruction of the analysis process with triangular fuzzy numbers, after fuzzy semantic quantification, matrix operation, and defuzzification, it realizes the accurate description of the fuzzy influence relationship between indicators. This method abandons the assumption of indicator independence in AHP, can prove the causal relationship between indicators to derive weight values, and thus forms a methodological complement to AHP.

Subsequently, fsQCA is used to mine multiple paths. QCA based on the ideas of set theory and configurational thinking, effectively connects qualitative and quantitative analyses. With the help of architecture theory and Boolean algebra operations, it examines the relationship between antecedent conditions and condition combinations and results from the perspective of sets, so as to explain the complex causal relationships behind phenomena. The specific operation process of fsQCA selected in this study is divided into the following seven steps: variable selection, typical case selection, data calibration, necessary analysis of single conditions, truth table construction, combinatorial analysis, and result interpretation [41].

4.2. AHP Steps

Analytic Hierarchy Process (AHP) was proposed by Satty in the 1980s [42]. It is a method that decomposes complex decision-making problems into multiple levels and multiple factors and conducts decision-making analysis through a combination of qualitative and quantitative approaches. The core lies in transforming the subjective judgments of decision makers into quantifiable weights, thereby achieving scientific decision making. The judgment matrix is the key to AHP quantitative analysis, used to represent the importance of each factor (or scheme) at the same level relative to a certain factor at the previous level. In this paper, the 1–9 scaling method proposed by Saaty is adopted as the assignment criterion (

Table 5. Influencing factors of new dynamic capability of foreign trade SMEs.

Scale	Meaning
1	Equally important
2	Between equally important and slightly important
3	Slightly important
4	Between slightly important and more important
5	Moderate important
6	Between more importance and strong importance
7	Strong important
8	Between strong importance and extreme importance
9	Extremely important

).

Invite experts to assess the mutual importance of the factors by using Table 5. Assuming that there are N influencing factors in an index layer for K experts, the specific steps are as follows:

(1)

Construct the judgment matrix A^k = ${\left[{\mathrm{a}}_{\mathrm{i}\mathrm{j}}\right]}_{\mathrm{n}\times \mathrm{n}}^{\mathrm{k}}$, each a_ij represents the kth experts’ preference of the factor i over the factor j; i, j ∈ N. The fundamental relationship between the elements of the reciprocal matrix is (a_ij). (a_ji) = 1

(2)

Product the average judgment matrix B, [a_ij] with elements the average values of experts’ preferences, as follows:

$${\mathrm{b}}_{\mathrm{ij}}={\displaystyle \frac{\sum {{\mathrm{a}}_{\mathrm{ij}}}^1+{{\mathrm{a}}_{\mathrm{ij}}}^2+\dots +{{\mathrm{a}}_{\mathrm{ij}}}^{\mathrm{k}}}{\mathrm{k}}}$$

(1)

$$\mathrm{B}={\left[\begin{array}{c}{\mathrm{b}}_{\mathrm{ij}}\end{array}\right]}_{\mathrm{n}\times \mathrm{n}}=\left[\begin{array}{cccc}1& {\mathrm{b}}_{12}& \dots & {\mathrm{b}}_{1\mathrm{n}}\\ {\mathrm{b}}_{21}& 1& \dots & {\mathrm{b}}_{2\mathrm{n}}\\ ⋮& ⋮& \ddots & ⋮\\ {\mathrm{b}}_{\mathrm{n}1}& {\mathrm{b}}_{\mathrm{n}2}& \dots & 1\end{array}\right]$$

(2)

(3)

Normalize each column element of the judgment matrix (the sum of the column elements is 1):

$$\begin{array}{c}{\mathrm{c}}_{\mathrm{ij}}\end{array}={\displaystyle \frac{\begin{array}{c}{\mathrm{b}}_{\mathrm{ij}}\end{array}}{{{\displaystyle \sum }}_{\mathrm{k}=1}^{\mathrm{n}}\begin{array}{c}{\mathrm{b}}_{\mathrm{kj}}\end{array}}}$$

(3)

(4)

Sum the normalized matrix row by row:

$${{\mathrm{w}}_{\mathrm{i}}}^{\prime }={\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{n}}\begin{array}{c}{\mathrm{c}}_{\mathrm{ij}}\end{array}$$

(4)

(5)

Normalize w_i′ to obtain the weight vector W:

$${\mathrm{w}}_{\mathrm{i}}={\displaystyle \frac{{{\mathrm{w}}_{\mathrm{i}}}^{\prime }}{{{\displaystyle \sum }}_{\mathrm{k}=1}^{\mathrm{n}}\begin{array}{c}{{\mathrm{w}}_{\mathrm{i}}}^{\prime }\end{array}}}$$

(5)

$$\mathrm{W}={\left({\mathrm{w}}_1,{\mathrm{w}}_2,\dots ,{\mathrm{w}}_{\mathrm{n}}\right)}^{\mathrm{T}}$$

(6)

(6)

Calculate the weight vector. By calculating the judgment matrix B, the eigenvector W corresponding to the maximum eigenroot λmax of this judgment matrix can be obtained:

$${\mathrm{\lambda }}_{\max }={\displaystyle \frac{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}\frac{{\left(\mathrm{BW}\right)}_{\mathrm{i}}}{{\mathrm{w}}_{\mathrm{i}}}}{\mathrm{n}}}$$

(7)

(7)

Calculate the Consistency Index (CI):

$$\mathrm{CI}={\displaystyle \frac{\mathrm{\lambda }\max -\mathrm{n}}{\mathrm{n}-1}}$$

(8)

(8)

Calculate the Consistency Ratio (CR): Generally, if CR < 0.1, the consistency of the judgment matrix is considered acceptable, meaning that although the judgment matrix may have slight inconsistency, it is insufficient to affect the rationality of the decision-making process. However, if CI > 0.1 or CR exceeds 0.1, and there is inconsistency among judgment matrices, the judgment matrix needs to be adjusted.

Table 6. Random consistency RI value.

N	3	4	5	6	7	8	9	10
RI	0.58	0.89	1.12	1.26	1.36	1.41	1.46	1.49

provides specific values of RI:

$$\mathrm{CR}={\displaystyle \frac{\mathrm{CI}}{\mathrm{RI}}}$$

(9)

(9)

Calculate the total weight of the scheme layer relative to the target layer, that is, the total importance degree of each scheme to the ultimate goal.

Let the criterion layer have m factors, and their weights relative to the target layer are as follows:

$$\mathrm{W}={\left({\mathrm{w}}_1,{\mathrm{w}}_{2\mathrm{i}},\dots ,{\mathrm{w}}_{\mathrm{m}}\right)}^{\mathrm{T}}$$

(10)

There are n factors in the scheme layer, and the weight vector for the J-th factor in the criterion layer is as follows:

$${\mathrm{W}}_{\mathrm{j}}={\left({\mathrm{w}}_{1\mathrm{j}},{\mathrm{w}}_{2\mathrm{j}},\dots ,{\mathrm{w}}_{\mathrm{nj}}\right)}^{\mathrm{T}}$$

(11)

The total weight of the i-th scheme in the scheme layer is as follows:

$${\mathrm{h}}_{\mathrm{i}}={{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{m}}{\mathrm{w}}_{\mathrm{j}}\times {\mathrm{w}}_{\mathrm{ij}}$$

(12)

(10)

Check the overall consistency of the total ranking results at each level to ensure the logical unity of judgments at all levels. Calculate the total ranking consistency index CI₂, where CI_j is the consistency index of the judgment matrix corresponding to the J-th factor of the criterion layer:

$${\mathrm{CI}}_2={\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{m}}{\mathrm{w}}_{\mathrm{j}}\times {\mathrm{CI}}_{\mathrm{j}}$$

(13)

(11)

Calculate the total ranking average random consistency index RI₂, where RI_j is the random consistency index of the judgment matrix corresponding to the J-th factor of the criterion layer:

$${\mathrm{RI}}_2={\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{m}}{\mathrm{w}}_{\mathrm{j}}\times {\mathrm{RI}}_{\mathrm{j}}$$

(14)

(12)

Calculate the total sorting consistency ratio. If CR₂ < 0.1, the total sorting result is acceptable; otherwise, the relevant judgment matrix needs to be backtraced and corrected.

$${\mathrm{CR}}_2={\displaystyle \frac{{\mathrm{CI}}_2}{{\mathrm{RI}}_2}}$$

(15)

(13)

Sort according to the total weight of the scheme layer and select the scheme with the highest weight as the optimal decision result.

4.3. Fuzzy-DEMATEL Steps

Decision Making Trail and Evaluation Laboratory (DEMATE) was proposed by Battelle in the 1970s [43]. By using this method to invite experts to judge the mutual influence relationship among system indicators and transform qualitative research into quantitative research, it can help decision makers filter out the core indicators that affect the realization of system utility. Considering the subjective judgment of experts and the uncertainty caused by language fuzziness, the fuzzy-DEMATEL method will be used to judge the mutual influence between indicators [44]. The mapping relationship between linguistic variables and fuzzy numbers is established, as shown in

Table 7. Conversion relationship between judgment terms and fuzzy numbers.

Scale	Meaning	Semantic Transformation (${\mathbf{l}}_{\mathbf{i}\mathbf{j}},{\mathbf{m}}_{\mathbf{i}\mathbf{j}},{\mathbf{u}}_{\mathbf{i}\mathbf{j}}$)
0	No impact	(0.0, 0.1, 0.3)
1	The impact is minimal	(0.1, 0.3, 0.5)
2	Have little impact	(0.3, 0.5, 0.7)
3	Have a great influence	(0.5, 0.7, 0.9)
4	Have a great impact	(0.7, 0.9, 1.0)

.

Using Table 7 to collect the judgment results of experts on the mutual influence degree among various factors, the initial direct influence matrix is constructed. The initial value of expert score is defuzzified, and the triangular fuzzy number is transformed into specific clear value. Assuming that there are n influencing factors, the specific steps are as follows:

(1)

If the k-th expert judges that the influence degree of factor I on factor J is as follows, the expression result of expert K can be expressed as:

$$B^k={\left[{b_{ij}}^k\right]}_{\mathrm{n}\times \mathrm{n}}=\left[\begin{array}{cccc}0, 0.1, 0.3& {l_{12}}^k, {m_{12}}^k, {u_{12}}^k& \dots & {l_{1n}}^k, {m_{1n}}^k, {u_{1n}}^k\\ {l_{21}}^k, {m_{21}}^k, {u_{21}}^k& 0, 0.1, 0.3& \dots & {l_{2n}}^k, {m_{2n}}^k, {u_{2n}}^k\\ ⋮& ⋮& \ddots & ⋮\\ {l_{n1}}^k, {m_{n1}}^k, {u_{n1}}^k& {l_{n2}}^k, {m_{n2}}^k, {u_{n2}}^k& \dots & 0, 0.1, 0.3\end{array}\right]$$

(16)

(2)

Standardize the judgment matrix of the k-th expert:

$${{\mathrm{a}}_{\mathrm{ij}}}^{\mathrm{k}}={{\mathrm{l}}_{\mathrm{ij}}}^{\mathrm{k}}-{{\mathrm{minl}}_{\mathrm{ij}}}^{\mathrm{k}}/{\Delta }_{\min }^{\max }$$

(17)

$${{\mathrm{b}}_{\mathrm{ij}}}^{\mathrm{k}}=\left({{\mathrm{m}}_{\mathrm{ij}}}^{\mathrm{k}}-{{\mathrm{minl}}_{\mathrm{ij}}}^{\mathrm{k}}\right)/{\Delta }_{\min }^{\max }$$

(18)

$${{\mathrm{c}}_{\mathrm{ij}}}^{\mathrm{k}}={{\mathrm{u}}_{\mathrm{ij}}}^{\mathrm{k}}-{{\mathrm{minl}}_{\mathrm{ij}}}^{\mathrm{k}}/{\Delta }_{\min }^{\max }$$

(19)

$${\Delta }_{\min }^{\max }={{\mathrm{maxu}}_{\mathrm{ij}}}^{\mathrm{k}}-{{\mathrm{minl}}_{\mathrm{ij}}}^{\mathrm{k}}$$

(20)

(3)

Calculate the left and right standard values:

$${{\mathrm{p}}_{\mathrm{ij}}}^{\mathrm{k}}={{\mathrm{b}}_{\mathrm{ij}}}^{\mathrm{k}}/\left(1+{{\mathrm{b}}_{\mathrm{ij}}}^{\mathrm{k}}-{{\mathrm{a}}_{\mathrm{ij}}}^{\mathrm{k}}\right)$$

(21)

$${{\mathrm{q}}_{\mathrm{ij}}}^{\mathrm{k}}={{\mathrm{c}}_{\mathrm{ij}}}^{\mathrm{k}}/\left(1+{{\mathrm{c}}_{\mathrm{ij}}}^{\mathrm{k}}-{{\mathrm{b}}_{\mathrm{ij}}}^{\mathrm{k}}\right)$$

(22)

(4)

Calculate the comprehensive standardized value:

$${{\mathrm{W}}_{ij}}^k={\displaystyle \frac{{{\mathrm{p}}_{ij}}^k\left(1-{{\mathrm{p}}_{ij}}^k\right)+{\left({{\mathrm{q}}_{ij}}^k\right)}^2}{1-{{\mathrm{p}}_{ij}}^k+{{\mathrm{q}}_{ij}}^k}}$$

(23)

(5)

Calculate the clarity value:

$${{\mathrm{e}}_{ij}}^k=min{{\mathrm{l}}_{ij}}^k+{{\mathrm{w}}_{ij}}^k·{\Delta }_{min}^{max}$$

(24)

(6)

Record the direct influence matrix E of the k-th expert and find the average value of the influence degree of I factor on J factor judged by k-th expert:

$${\mathrm{E}}^k=\left(\begin{array}{ccc}{{\mathrm{e}}_{12}}^k& \dots & {{\mathrm{e}}_{1n}}^k\\ ⋮& \ddots & ⋮\\ {{\mathrm{e}}_{n1}}^k& \dots & {{\mathrm{e}}_{nn}}^k\end{array}\right)$$

(25)

$${{\mathrm{e}}_{ij}}^k=\left({\mathrm{e}}_{ij}^1+{\mathrm{e}}_{ij}^2+\dots +{\mathrm{e}}_{ij}^n\right)/\mathrm{K}$$

(26)

(7)

Construct the fuzzy direct influence matrix M according to e_ij^k.

(8)

Calculate the normalization direct influence matrix G.

$$\mathrm{G}={\displaystyle \frac{1}{{\max }_{1\le \mathrm{i}\le \mathrm{n}}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{e}}_{\mathrm{ij}}}} M$$

(27)

(9)

Determine the comprehensive influence matrix T.

$$\mathrm{T}=\mathrm{G}{\left(1-G\right)}^{-1}$$

(28)

(10)

The elements in matrix T are added row by row to form the influence degree f_i, which represents the comprehensive influence value of the factor in that row on all other factors. The elements in matrix T are added by column to form the degree of influence g_i, which represents the comprehensive influence value of the factors in this column on all other factors:

$${\mathrm{f}}_{\mathrm{i}}={\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{t}}_{\mathrm{ij}},\mathrm{i}=1,2,\dots \mathrm{n}$$

(29)

$${\mathrm{g}}_{\mathrm{i}}={\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{t}}_{\mathrm{ij}},\mathrm{i}=1,2,\dots \mathrm{n}$$

(30)

(11)

The sum of the degree of influence and the degree of being influenced is called the centrality n_i, which indicates the position of the factor in the system and the extent of its effect. The difference between the degree of influence and the degree of being influenced is called the cause degree mi, which reflects the causal relationship among various influencing factors. If the cause degree is greater than 0, it indicates that the factor has a greater effect on other factors and is called a cause factor. If the cause degree is less than 0, it indicates that the factor is more affected by other factors and is called a result factor:

$${\mathrm{m}}_{\mathrm{i}}={\mathrm{f}}_{\mathrm{i}}-{\mathrm{g}}_{\mathrm{i}},\mathrm{i}=1,2,\dots \mathrm{n}$$

(31)

$${\mathrm{n}}_{\mathrm{i}}={\mathrm{f}}_{\mathrm{i}}+{\mathrm{g}}_{\mathrm{i}},\mathrm{i}=1,2,\dots \mathrm{n}$$

(32)

4.4. Calculation of Comprehensive Influence Degree Based on Fuzzy AHP-DEMATEL

Using the total weight (h_Bi) obtained by AHP and the centrality (n_Bi) obtained by fuzzy-DEMATEL, the comprehensive influence degree is calculated. Through the collaborative calculation of the two methods, we finally obtain the comprehensive impact degree of each indicator. This integration strategy effectively makes up for the limitations of using the two methods alone—it not only reduces the problem of semantic ambiguity in expert scoring caused by the rigid hierarchical judgment scales in AHP but also weakens the defect in fuzzy-DEMATEL, where indicator weight allocation is prone to being affected by subjective experience biases. Thus, it more comprehensively and objectively captures the actual importance of each influencing factor in the system, providing a more accurate quantitative basis for subsequent decision making.

$${\mathrm{w}}_{\mathrm{Bi}}={\displaystyle \frac{{\mathrm{h}}_{\mathrm{Bi}}{n}_{Bi}}{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{h}}_{\mathrm{Bi}}{n}_{Bi}}}$$

(33)

4.5. fsQCA Steps

fsQCA (Fuzzy-set Qualitative Comparative Analysis) was presented by Charles C. Ragin in 1987. It was proposed in that this method is used to explore the multiple concurrent causal relationships that cause the occurrence of result variables in complex systems, which is more in line with practical needs. The fsQCA method breaks the research bottleneck of efficiency orientation in traditional methods, is effect-oriented, and opens up a new research perspective for scholars. It has unique advantages for exploring the causal mechanism of problems in complex systems [45]. fsQCA is usually divided into the following five steps: case analysis and variable selection, data calibration, necessity analysis of single condition, truth table construction, and combination analysis. The specific operation is as follows:

(1)

Case analysis and variable selection: fsQCA is an analysis method with both qualitative and quantitative requirements, which can obtain the complex causal mechanism of the combination of different factors to the results. Research usually uses literature analysis, field interviews with experts, and other methods to comprehensively sort out and define the outcome variables and conditional variables and collect data.

(2)

Data calibration, including direct calibration and indirect calibration. The fsQCA method is based on Boolean algebra, which requires the data to be distributed in the interval of [0, 1], but usually the collected original data does not meet this condition, so the original data corresponding to each antecedent condition and result condition is transformed into fuzzy membership degree. For this reason, the existing research often adopts direct calibration method, that is, setting three anchor points (complete membership, intersection, and complete non-membership) to generate the membership degree of the research object (i.e., Case) in each factor (i.e., Condition set) [46].

(3)

Necessity analysis (test necessary conditions): Before qualitative comparative analysis of fuzzy sets, it is necessary to test the necessity of antecedent variables. Consistency is usually used to measure the necessity of antecedent variables to results, that is, to what extent the results originate from the influence of antecedent variables. When the consistency is greater than or equal to 0.9, it can be judged that the antecedent variables are necessary conditions for results [47].

(4)

Construct truth tables: Truth tables are suitable tools for testing set relationships, and they shift the focus from empirical cases to conditional combinations. The truth table shows all possible combinations of causal conditions. There are usually three criteria to check whether a combination is relevant to the result: frequency, original consistency, and reduction rate of inconsistency. Among them, according to the total samples and their distribution among configurations, the critical value of configuration case frequency (1 is the minimum critical value) should be selected to keep at least 75% of observed cases, and the original consistency should be at least 0.8 [48].

(5)

Combinatorial analysis: The antecedent variables appearing in both the intermediate solution and the reduced solution are defined as “core conditions”, and the antecedent variables appearing only in the intermediate solution but not in the reduced solution are defined as “edge conditions”. They have different contributions to the results, and the core conditions have an important impact on the results, while the edge conditions only play an auxiliary role [49].

4.6. Comments on Theoritical Framework

This paper adopts the method of integrating AHP, fuzzy-DEMATEL, and fsQCA. First of all, the AHP method is used to solve the relativity problem of index weights. Again, the fuzzy-DEMATEL method is utilized to reveal the causal influence of the variables. Secondly, the fuzzy AHP-DEMATEL is adopted to calculate the comprehensive influence degree of factors, which is convenient for scholars to judge the importance of a certain variable relative to the outcome variable accordingly and conduct more in-depth research in sequence. Finally, the fsQCA method is utilized to explore the configuration path for generating high-level digital and intelligent immunity capabilities of foreign trade SMEs. This method covers the complete analysis dimensions of “importance → causality → configurability” and meets the essential requirements of complex system research.

5. Analysis of New Dynamic Capability Evaluation Mechanism of Foreign Trade SMEs

5.1. Determination of Weight Coefficient Based on AHP

In order to enhance the reliability of the data, 25 experts in related fields were invited to carry out a consultation and questionnaire survey. The expert information is shown in

Table 8. Information table of expert group members.

Project		Frequency	Percentage
Gender	Male	14	56%
	Female	11	44%
Professionalism	Graduate student	2	8%
	Doctoral student	7	28%
	Professor	11	44%
	Industry expert	5	20%

, and the investigation time is from September 2023 to March 2024. The 25 experts participating in the scoring include two master’s students, seven doctoral students, 11 professors, and five senior executives of foreign trade enterprises, with professional coverage of international trade, management science and engineering, data science, and other fields.

According to the evaluation criteria in Table 5, experts are invited to judge the importance of the influencing factors of the new dynamic capability of foreign trade SMEs. Each expert needs to give four judgment matrices, including one judgment matrix of criterion indicators and three judgment matrices of scheme layer indicators. Then, Formulas (1)–(15) are used to calculate the total weight of the factors. Through our efforts, we have met all the consistency conditions. The results are shown in

Table 9. The results of index weights and consistency checks.

Criterion Layer			Scheme Layer			$\mathbf{Total} \mathbf{Weight}: {\mathit{h}}_{\mathit{B}\mathit{i}}$ CR = 0.0749
Name	Weight	CR	Name	Weight	CR
A1	0.2923	0.0327	hB1	0.6115	0.0327	0.1787
			hB2	0.0963		0.0282
			hB3	0.2922		0.0854
A2	0.6120		hB4	0.5245	0.0900	0.3210
			hB5	0.1135		0.0695
			hB6	0.3208		0.1963
			hB7	0.0413		0.0253
A3	0.0957		hB8	0.5513	0.0611	0.0528
			hB9	0.2243		0.0215
			hB10	0.0923		0.0088
			hB11	0.1320		0.0126

.

5.2. Analysis of Influencing Factors Based on Fuzzy-DEMATEL

According to the judgment standard in Table 7, 25 experts in Table 8 are invited again to give judgment opinions on the mutual influence degree of each index. According to Formulas (11)–(19), the fuzzy influence of Expert 1 is first fuzzified to obtain matrix E. Then, the calculation results of all experts are averaged, and the direct influence matrix M is calculated. The results are shown in

Table 10. Direct impact matrix M.

Serial Number	B1	B2	B3	B4	B5	B6	B7	B8	B9	B10	B11
B1	0.1217	0.7015	0.5149	0.3282	0.5299	0.3344	0.8783	0.3132	0.5349	0.7086	0.7027
B2	0.1875	0.1217	0.3132	0.1809	0.3046	0.1217	0.4851	0.2832	0.3055	0.3055	0.3129
B3	0.3420	0.5075	0.1217	0.3064	0.2976	0.1875	0.7095	0.3570	0.3129	0.5224	0.5149
B4	0.5075	0.7374	0.4925	0.1217	0.7089	0.3120	0.8783	0.5149	0.7077	0.8059	0.7024
B5	0.1217	0.2902	0.3123	0.1217	0.1217	0.1217	0.7162	0.3055	0.3129	0.5149	0.4851
B6	0.2979	0.6810	0.4851	0.2902	0.4851	0.1217	0.7021	0.5075	0.7086	0.7021	0.6951
B7	0.1217	0.2911	0.1217	0.1217	0.1217	0.1344	0.1217	0.1217	0.3052	0.2979	0.3420
B8	0.3200	0.5149	0.3052	0.3123	0.3417	0.3190	0.7024	0.1217	0.4925	0.7021	0.4376
B9	0.2994	0.3135	0.2902	0.1477	0.3078	0.1742	0.5149	0.1217	0.1217	0.4925	0.3126
B10	0.1217	0.3273	0.1217	0.1742	0.1217	0.1293	0.2767	0.1217	0.3420	0.1217	0.3052
B11	0.1217	0.2982	0.1217	0.1217	0.3049	0.1477	0.2832	0.1217	0.3126	0.2976	0.1217

. Using the traditional fuzzy-DEMATEL method to standardize the direct influence matrix, the standardized influence matrix G is obtained, and the influence degree, affected degree, centrality degree, and cause degree of each factor are further calculated. The results are shown in

Table 11. Influence degree, affected degree, centrality degree, and cause degree of each factor.

Serial Number	${\mathit{f}}_{\mathit{i}}$	${\mathit{f}}_{\mathit{i}}$ Sort	${\mathit{g}}_{\mathit{i}}$	${\mathit{g}}_{\mathit{i}}$ Sort	${\mathit{n}}_{\mathit{i}}$	${\mathit{n}}_{\mathit{i}}$ Sort	${\mathit{m}}_{\mathit{i}}$	${\mathit{m}}_{\mathit{i}}$ Sort	Factor Attribute
B1	1.8099	3	0.8331	9	2.6430	3	0.9768	3	Cause factor
B2	0.9607	8	1.5400	4	2.5007	5	−0.5793	8	Outcome factor
B3	1.3605	5	1.0253	7	2.3858	9	0.3352	5	Cause factor
B4	2.0850	1	0.7354	10	2.8203	1	1.3496	1	Cause factor
B5	1.0683	6	1.1661	6	2.2344	11	−0.0978	6	Outcome factor
B6	1.8229	2	0.6930	11	2.5159	4	1.1299	2	Cause factor
B7	0.6824	11	2.0063	1	2.6887	2	−1.3238	11	Outcome factor
B8	1.4843	4	0.9216	8	2.4059	8	0.5628	4	Cause factor
B9	1.0052	7	1.4505	5	2.4557	7	−0.4453	7	Outcome factor
B10	0.7114	10	1.7561	2	2.4675	6	−1.0447	10	Outcome factor
B11	0.7339	9	1.5972	3	2.3311	10	−0.8633	9	Outcome factor

.

Observing Table 11, we can find that the cause degree of indexes B1, B3, B4, B6, and B8 is greater than 0, which indicates that these factors are easy to affect the output results by affecting other factors in the system. The influence degree of these indicators is ranked 1–5, while the affected degree is ranked 8–11, which shows that these factors have strong influence on other factors but are difficult to be affected. Among them, B1 and B4 have high degree of cause and centrality, which shows that they play an important role in the system.

Indicators B2, B5, B7, B9, and B11 have negative causes, which indicates that these indicators are easily affected by other factors in the system. The influence degree of these indicators is in all factors in 6–11, and the affected degree is in 1–6, indicating that these factors are easily affected by other factors and cannot easily play an influential role.

5.3. Comprehensive Influence Degree Calculation on Fuzzy AHP-DEMATEL

The total weight (h_Bi) obtained by AHP and the centrality (n_Bi) obtained by fuzzy-DEMATEL are combined to calculate the comprehensive influence degree, and the results are shown in

Table 12. Comprehensive influence degree of each factor.

Factors	$\mathbf{Total} \mathbf{Weigh} \mathbf{t}{\mathbf{h}}_{\mathbf{B}\mathbf{i}}$	$\mathbf{Centrality} {\mathbf{n}}_{\mathbf{B}\mathbf{i}}$	$\mathbf{Comprehensive} \mathbf{Influence} \mathbf{Degree} {\mathbf{w}}_{\mathbf{B}\mathbf{i}}$	Rank
B1	0.1787	2.6430	0.1817	3
B2	0.0282	2.5007	0.0271	7
B3	0.0854	2.3858	0.0784	4
B4	0.3210	2.8203	0.3482	1
B5	0.0695	2.2344	0.0597	5
B6	0.1963	2.5159	0.1900	2
B7	0.0253	2.6887	0.0262	8
B8	0.0528	2.4059	0.0489	6
B9	0.0215	2.4557	0.0203	9
B10	0.0088	2.4675	0.0084	11
B11	0.0126	2.3311	0.0113	10

. The results show that B4 (Digital Intelligence Supervision and Early Warning), B6 (Digital Intelligence Ecosystem), and B1 (Digital Intelligence Management and Analysis) are the core of the new dynamic capability index system. They are both key factors and play important roles in the system.

5.4. Conditional Configuration Analysis Based on fsQCA

5.4.1. Data Collection

However, to date, industrial businesses are struggling with digital transformation [50]. It is difficult for small- and medium-sized foreign trade enterprises to have a perfect data-driven system. Therefore, this paper takes the form of an expert questionnaire and uses fsQCA to study the new dynamic capability driven by the data of foreign trade SMEs. The result variables are “significant improvement of new dynamic ability” and “significant reduction in new dynamic ability”, and the condition variables are 11 related variables designed in the above study. In this paper, 25 experts in Table 8 are invited to make adjustments for sample enterprises on the basis of five points. The 25 experts participating in the scoring include two master’s students, seven doctoral students, 11 professors, and five senior executives of foreign trade enterprises, with professional coverage of international trade, management science and engineering, data science, and other fields. The score threshold is 1–9 points, taking the “significant improvement” and “significant reduction” in the new dynamic ability as the result orientation.

The research sample covers 70 foreign trade SMES. In terms of geographical distribution, the Yangtze River Delta accounts for 26, the Pearl River Delta accounts for 25, and the remaining 19 are distributed in other regions. The industries focus on main foreign trade categories such as textiles (30), 3C products (22), and machinery (18), ensuring regional and industrial representativeness. Enterprise samples are matched based on the expert knowledge background, and each enterprise is independently scored by 3–5 experts, forming an expert evaluation panel for that enterprise. The process followed these steps: First, we aggregated individual expert ratings and compiled collective feedback from the panel. The results were then fed back to the panel for continuous refinement until consensus was reached. After screening, 9 invalid questionnaires were excluded, and finally, 61 valid samples were retained, with an effective rate of 87.1%. The judgment results of Expert 1 of foreign SME1 are shown in

Table 13. Judgment Results of Expert 1.

	B1	B2	B3	B4	B5	B6	B7	B8	B9	B10	B11
Basic score	5	5	5	5	5	5	5	5	5	5	5
Significant improvement	7	4	4	9	4	8	3	6	4	3	4
Significantly reduced	4	5	3	4	5	4	4	3	4	4	3

.

5.4.2. Data Calibration

Using direct calibration method [49] for the selection of the following three anchor points: “complete membership”, “intersection point”, and “complete non-membership”. Referring to the existing research practices, the average value and standard deviation are calculated first, then the average value is taken as the intersection point, and the average value plus or minus standard deviation is taken as the complete membership point and the complete non-membership point, respectively. The descriptive statistics of each variable and anchor points are shown in

Table 14. Anchor points and descriptive statistics of conditional variables.

Conditional Variable	Ambiguity Correction			Descriptive Statistics
	Completely Subordinate to	Crossing Point	Completely Unaffiliated	Mean Value	Standard Deviation
B1	7.173	5.016	2.860	5.016	2.156
B2	4.645	3.852	3.060	3.852	0.792
B3	4.416	3.262	2.109	3.262	1.153
B4	8.492	4.902	1.311	4.902	3.590
B5	5.233	4.623	4.013	4.623	0.610
B6	7.846	4.951	2.056	4.951	2.895
B7	3.755	2.787	1.819	2.787	0.968
B8	6.165	4.557	2.950	4.557	1.608
B9	5.345	4.803	4.261	4.803	0.542
B10	4.225	3.148	2.070	3.148	1.078
B11	4.394	3.590	2.786	3.590	0.804

.

5.4.3. Necessity Test

Before the qualitative comparative analysis of fuzzy sets, it is necessary to test the necessity of antecedent variables, and the results are shown in

Table 15. Necessity test results.

Conditional Variable	Consistency of Result Variables
	Significant Improvement	Significantly Reduced
B1	0.925	0.090
~B1	0.075	0.910
B2	0.497	0.532
~B2	0.503	0.468
B3	0.810	0.184
~B3	0.190	0.816
B4	0.942	0.071
~B4	0.058	0.929
B5	0.596	0.514
~B5	0.404	0.486
B6	0.936	0.073
~B6	0.064	0.927
B7	0.235	0.730
~B7	0.765	0.270
B8	0.910	0.112
~B8	0.090	0.888
B9	0.537	0.607
~B9	0.463	0.393
B10	0.174	0.815
~B10	0.826	0.185
B11	0.362	0.679
~B11	0.638	0.321

Note: “~” indicates the negation of the corresponding condition.

. When the result variable is “significantly improved digital dynamic capability”, the consistency of B1 (digital intelligence management and analysis), B4 (digital intelligence supervision and early warning), B6 (digital intelligence ecosystem), and B8 (digital foundation) is all greater than 0.9, indicating that they can independently constitute the necessary conditions for “significantly improved new dynamic capability”. When the result variable is “significant reduction in new dynamic capability”, the consistency of B1 (lack of digital intelligence management and analysis), B4 (lack of digital intelligence supervision and early warning), and B6 (lack of digital intelligence ecosystem) is greater than 0.9, indicating that they can independently constitute the necessary conditions for “significant reduction in digital dynamic capability”.

5.4.4. Constructing Truth Table

After determining the necessary conditions, the truth table is constructed, and the data is analyzed in a standardized way. The setting frequency is 1, and the original consistency threshold is 0.80 [48]. The 61 samples studied are coded and summarized.

5.4.5. Conditional Configuration Analysis

Based on the critical values of coverage (>0.20) and consistency (>0.75) defined by Ragin [45], the frequency threshold is 1 and consistency threshold is 0.8, and the coverage and consistency of the model solution are 0.546 and 0.960, which are satisfactory enough to support the occurrence of result conditions. According to the research of Pappas [49], the core conditions are analyzed by using intermediate solution and simple solution, and the results are shown in

Table 16. High-level data driven dynamic capability configuration sufficiency analysis of foreign trade SMEs.

Conditional Variable	Configuration 1	Configuration 2	Configuration 3	Configuration 4	Configuration 5	Comprehensive Weight	Rank
B1 Data Management and Analysis	◎	◎	●	●	●	0.1916	3
B2 Digital Intelligence Memory		◎	⊗		◎	0.0307	7
B3 Digital Talents	●	●	●	●	●	0.0978	4
B4 Digital Intelligence Supervision and Early Warning	◎	◎	●	●	●	0.2790	2
B5 Digital Intelligence Communication and Tracking	◎		⊗	⨯	◎	0.0768	5
B6 Digital Intelligence Ecosystem	◎	◎	●	●	●	0.2050	1
B7 Digital Intelligence Output	⨯		⨯	●	●	0.0223	9
B8 Digital Foundation	◎	◎	●	●	●	0.0478	6
B9 Staff Quality		◎	⊗	◎	◎	0.0238	8
B10 Rules and Regulations	⨯	⨯	⨯	⊗	⊗	0.0115	10
B11 Digital Intelligence Culture	⊗	⊗		⊗		0.0136	11

Note: “●” indicates that the core condition exists, “⊗” indicates that the core condition does not exist, “◎” indicates that the edge condition exists, “⨯” indicates that the edge condition does not exist, and blank indicates that the existence of condition variables is irrelevant to the result.

.

As shown in the results of fsQCA condition combination analysis in Table 15, the formation of new data-driven dynamic capabilities of foreign trade SMEs is based on the simultaneous existence of B1 (digital intelligence management and analysis), B4 (digital intelligence supervision and early warning), B6 (digital intelligence ecosystem), and B8 (digital foundation). Combined with the analysis of fuzzy AHP-DEMATEL comprehensive weight results, it can be seen that B1 (digital intelligence management and analysis), B4 (digital intelligence supervision and early warning), and B6 (digital intelligence ecosystem) have significant effects on the high-new dynamic capability of foreign trade SMEs. In addition, by observing the formation path of high-level new dynamic ability, it can be found that B3 (digital intelligence talents) can make up for the lack of B2 (digital intelligence memory), B5 (digital intelligence communication and tracking), B9 (employee quality), and B11 (digital intelligence culture).

6. Conclusions

Against the backdrop of the intertwined digital economy and complex international landscape, this study enhances enterprise dynamic capability research and deepens organizational immunity theory from both practical scenarios and theoretical innovation dimensions, while providing targeted insights for the sustainable development of foreign trade SMEs. On the one hand, by integrating digital practice scenarios of foreign trade SMEs—such as tracking emerging market public opinion with AI tools like Google Trends, achieving digital pipeline management through MES systems, opening stores on platforms like Taobao, building independent websites, and implementing precise traffic acquisition via SEO and social media advertising—this research reveals the deep integration path of digital technologies and business scenarios. Meanwhile, based on the escalating overseas credit risks indicated in the “2025 China Export Risk Index Report for Small, Medium, and Micro Foreign Trade Enterprises” and the internal resource constraints (36.8% tight funds, 50.8% professional talent shortages) reflected in the “2024 China SME Digital Transformation Report”, it clarifies that the construction of dynamic capabilities for foreign trade SMEs must break through traditional frameworks and shift toward crisis-resistant constitution building centered on digital intelligence immunity.

On the other hand, this study comprehensively uses the Analytic Hierarchy Process (AHP) and Fuzzy Decision-Making Trial and Evaluation Laboratory (fuzzy-DEMATEL) to determine indicator weights. AHP method is used to solve the relativity problem of index weights. Fuzzy-DEMATEL introduces triangular fuzzy numbers, breaking through the limitations of traditional methods to accurately depict fuzzy relationships between indicators—complementing each other’s strengths. Additionally, it employs fuzzy-set Qualitative Comparative Analysis (fsQCA) to mine multiple paths. Based on set theory, fsQCA integrates qualitative and quantitative analyses, using Boolean algebra operations to dissect the relationship between antecedent condition combinations and outcomes, thereby revealing complex causal mechanisms. This not only provides micro-level empirical support for the expansion of dynamic capability theory in the digital economy context but also offers a solution with both theoretical innovation and practical guidance for foreign trade SMEs to achieve sustainable development amid high cross-border risks and resource constraints.

6.1. The Significance for Theories

(1)

This paper improves the research on the dynamic capabilities of enterprises in the context of the digital economy. Digital intelligence immunity can provide a suitable research perspective for the research on the new dynamic capabilities driven by enterprise data and can effectively make up for the deficiencies in the current research on the new dynamic capabilities driven by enterprise data. Firstly, the digital intelligence immune subsystem supports the implementation of dynamic capabilities. First, the digital and intelligent immune system expands the scope of immune surveillance and enhances the dynamic perception ability. Second, the digital intelligence immune system is conducive to optimizing the efficiency of resource integration and enhancing the ability of dynamic decision making. Thirdly, it is conducive to establishing an immune-driven “tissue repair” mechanism and enhancing the dynamic configuration capability. The perspective of digital intelligence immunity can provide theoretical support for the construction of a new dynamic capability index system driven by enterprise data. Secondly, the digital intelligence immune system is conducive to supporting the construction of the dynamic capability theoretical system. The three subsystems, namely the digital intelligence peripheral system, the digital intelligence dedicated system, and the digital intelligence central system, each perform their own duties and can automatically carry out non-specific processing, broad specific processing, or unique specific processing at different levels based on the degree of opportunity and crisis handling, providing framework support for the construction of new dynamic capabilities. Finally, the immune response logic of digital intelligence immunity and the circular interaction mode of routine immune responses are helpful to clarify the internal mechanism of enterprise data driven from the perspective of dynamic capabilities and improve the re-response efficiency of the system to the same situation by constructing the trial-and-error–analysis–learning–memory path.

(2)

This article deepens the theoretical research on tissue immunity and has significant academic and practical significance. Theoretically, four major characteristics of digital intelligence immunity can be summarized as follows: intelligence, ecology, integrity, and interactivity. Intelligence achieves data-driven defense through technologies such as AI. Ecology breaks through organizational boundaries to build symbiotic networks. Integrity strengthens internal collaboration with the help of digital technology. Interactivity actively outputs influence through digital intelligence interconnection. In practice, the traditional abstract concept of tissue immunity based on biological metaphors has been transformed into a system framework that is technologically implementable, ecologically scalable, and capable of evolving. A digital and intelligent immune system including central, specialized, and peripheral systems has been constructed, and the hierarchical response and circular evolution mechanisms have been clearly defined. This provides more operational theoretical tools for organizations to deal with uncertainties in the digital economy era, promotes the research on organizational immunity from theoretical exploration to engineering construction, and helps organizations achieve resilient development in a dynamic environment.

6.2. The Significance for Foreign Trade SMEs

(1)

B1 (Digital Intelligence Management and Analysis), B4 (Digital Intelligence Supervision and Early Warning), and B6 (Digital Intelligence Ecosystem) have an important impact on the new dynamic capability driven by data of foreign trade SMEs. Fuzzy AHP-DEMATEL analysis shows that B1 (digital intelligence management and analysis), B4 (digital intelligence supervision and early warning), and B6 (digital intelligence ecosystem) have a significant impact on the formation of new data-driven dynamic capabilities of foreign trade SMEs. FsQCA analysis also shows that B1 (Digital Intelligence Management and Analysis), B4 (Digital Intelligence Supervision and Early Warning), and B6 (Digital Intelligence Ecosystem) are the necessary conditions for the formation of digital dynamic capabilities of foreign trade SMEs and the core conditions in multiple paths. Therefore, in order to improve their new data-driven dynamic capability level, foreign trade SMEs must grasp the construction of these three core links.

(2)

B3 (digital talents) has a strong influence on the new dynamic capability driven by data of foreign trade SMEs and is an important guarantee for the formation of high-level new dynamic capability of foreign trade SMEs. Fuzzy AHP-DEMATEL analysis shows that B3 (digital intelligence talents) has a strong influence on new dynamic ability. FsQCA analysis shows that B3 (digital intelligence talents) has a high coverage rate, and in path analysis, it can be found that B3 (digital intelligence talents) can play a certain role in the formation of high dynamic ability, such as B2 (digital intelligence memory), B5 (digital intelligence communication and tracking), B9 (employee quality), B10 (rules and regulations), and B11 (digital intelligence culture). Therefore, in order to improve their new dynamic ability level, foreign trade SMEs should attach importance to the training of digital talents, actively improve the talent incentive system, and speed up the introduction of digital talents.

(3)

B8 (Digital foundation) is an important foundation for the formation of new dynamic capabilities driven by high-level data of foreign trade SMEs. Fuzzy AHP-DEMATEL analysis shows that B8 (digital foundation) has a medium influence on the new dynamic capability of enterprises, but fsQCA analysis shows that B8 (digital foundation) is not only the necessary result of the formation of high-level new dynamic capability but also the core condition of multiple paths. Therefore, in order to strengthen the new dynamic capacity building, foreign trade SMEs should actively apply the digital foundation.

(4)

The formation of new dynamic capability driven by high data of foreign trade SMEs is the result of multi-factor linkage. Fuzzy AHP-DEMATEL analysis obtains the net influence effect of each variable on the new dynamic capability of foreign trade SMEs in the model composed of specific factors, which provides an intuitive understanding for understanding the influence effect of each antecedent variable on the new dynamic capability. The fsQCA analysis results show that the antecedent conditions for the formation of high-level new dynamic capabilities of foreign trade SMEs are all composed of two or more antecedent conditions. Through the necessity test (consistency ≥ 0.9), it is found that a single condition (such as only B1 or B6) cannot independently trigger high-level capabilities. When any one of B1, B4, and B6 is missing, the configurational consistency drops sharply to below 0.5. This finding breaks through the traditional linear causal hypothesis, confirming that dynamic capabilities in the digital economy are the result of the networked interlocking of soft and hard elements. Hard conditions (such as B6 ecological resources) provide the material foundation, while soft conditions (such as B3 talent cognition) activate factor synergy. Both are precipitated into reusable organizational routines through the regularization of the digital foundation (B8). This method effectively makes up for the deficiencies of the fuzzy AHP-DEMATEL analysis method, indicating that to build high-level data-driven new dynamic capabilities, enterprises should not be limited to optimizing single elements but should focus on the linkage and matching of multiple forces.