An Exploration of the Influence Mechanism of Resource Bricolage and Ambidextrous Learning on Micro-Innovation in New Ventures: The Moderating Roles of Customer Participation and Government Support

Abstract

Resource bricolage, centered on breaking through resource constraints, emphasizes providing innovation momentum for new ventures by creatively reorganizing existing resources at hand. Existing studies have confirmed that resource bricolage exerts an impact on corporate innovation, yet explorations into this approach’s mechanism of action remain insufficient. Based on resource bricolage theory and organizational ambidexterity theory, this study constructs a theoretical framework of “resource bricolage–ambidextrous learning–micro-innovation”. Using 319 new ventures as samples, hierarchical regression analysis is adopted to empirically test the mediating effect of ambidextrous learning and the moderating effects of government support and customer participation. The results show that resource bricolage has a significant positive impact on micro-innovation, with ambidextrous learning playing a partial mediating role. The research conclusions indicate that under resource-constrained contexts, new ventures can activate idle resources and improve resource efficiency through resource bricolage, thereby reducing reliance on new resource inputs. Under this method, new ventures can also rely on ambidextrous learning to accumulate knowledge and capabilities, laying the foundation for the continuous improvement of micro-innovation and further leverage government support to stabilize resource supply and absorb customer participation to align with market demand. This mechanism ultimately enables the achievement of micro-innovation while enhancing its sustainability. This study enriches resource bricolage theory by unpacking the “resource–learning–innovation” mechanism and provides practical guidance for new ventures to leverage resource bricolage and external support for micro-innovation under resource constraints, which is of reference value for sustainable entrepreneurial practices.

1. Introduction

As the core carriers of innovation vitality, new ventures are crucial for maintaining dynamic and sustainable entrepreneurial ecosystems. According to data from the State Market Regulatory Administration of China, the number of newly registered enterprises in China reached 8.305 million in 2020, representing an increase of 8.71% compared with that in 2019, demonstrating strong vitality in the entrepreneurial ecosystem. However, new ventures generally face the inherent disadvantages of being “new, small, and weak” in their initial stage. These issues are specifically manifested in multiple constraints such as a scarcity of resource endowments, insufficient embedding in social networks, and limited accumulation of legitimacy [1]. These constraints significantly inhibit the sustained development of the innovative activities among such firms. In this context, micro-innovation, as a market-demand-oriented innovation paradigm that builds competitive advantages through incremental improvements, has become a key path to break through resource bottlenecks due to its high compatibility with the resource characteristics of new ventures [2]. Thus, determining how to effectively activate micro-innovation practices has become a core issue for new ventures wishing to achieve survival and growth.

Resource bricolage, as a strategic response to resource constraints, embodies the logic of “sustainable resource utilization” by creatively reorganizing idle or underused resources (e.g., repurposing idle equipment for product testing) to generate innovation value, thereby reducing waste and enhancing resource efficiency. These functions align with the core definition of resource bricolage as “the process of using existing resources at hand to solve new problems or seize new opportunities” [3] and supports Senyard et al.’s [4] finding that resource bricolage expands the boundaries of resource utilization. Ambidextrous learning, by balancing exploratory learning (expanding resource application scenarios) and exploitative learning (deepening resource utilization efficiency) [5], enables new ventures to accumulate knowledge continuously. As Cao et al. [6] confirmed, this dual learning mechanism avoids short-term resource waste caused by blind trial-and-error, thus supporting the sustainability of micro-innovation. However, existing studies have not systematically explained how resource bricolage affects micro-innovation in new ventures through ambidextrous learning, and the transmission mechanism between these two concepts remains theoretically unknown.

From a contingency perspective, the innovative activities of new ventures are embedded in complex external contexts, and government support and customer participation may constitute important boundary conditions. As a core element of the formal institutional environment, government support provides resource supplements and legitimacy endorsements for enterprises through financial support and policy inclinations [7]. Thus, the moderating effect of government support in the transformation process from resource bricolage to ambidextrous learning is worthy of exploration. Customer participation, as an interface mechanism linking enterprises and markets, optimizes learning directions and innovation focuses through demand feedback [8]; the contingency role of this factor in the relationship between ambidextrous learning and micro-innovation also needs to be clarified. However, existing studies have paid insufficient attention to the above contextual variables and failed to fully reveal the boundary conditions under which resource bricolage affects micro-innovation.

The theoretical advancements of this study are reflected in the following aspects. First, Senyard et al. [4] explored the direct relationship between resource bricolage and innovation performance, while Wu et al. [9] focused on the association between ambidextrous innovation and performance. However, this study takes “micro-innovation”—a unique form of innovation under resource-constrained contexts—as the core dependent variable for the first time, systematically analyzing the action path of resource bricolage on non-disruptive innovation. Second, this study constructs a mediating framework of “resource bricolage–ambidextrous learning–micro-innovation”, revealing the differential roles of exploratory learning (expanding resource uses) and exploitative learning (deepening resource efficiency) within this approach, thus compensating for the deficiencies of existing studies in the dynamic analysis of the “resource–innovation–transformation mechanism”. Third, this study integrates the moderating effects of government support and customer participation, clarifying the differences in the roles of institutional factors and market factors at different transformation stages, thereby providing a segmented perspective for the contextualized research of resource bricolage theory.

Overall, based on resource bricolage theory and ambidextrous learning theory, this study takes resource bricolage as the research focus, constructs a theoretical framework of “resource bricolage–ambidextrous learning–micro-innovation”, and adopts hierarchical regression analysis to conduct an empirical analysis on sample data from 319 new ventures. This work aims to clarify the influence mechanism of resource bricolage on micro-innovation in new ventures and further explore the mediating effect of ambidextrous learning and the moderating effects of government support and customer participation.

2. Theoretical Background

The resource-based view holds that the competitive advantage of enterprises stems from their ability to integrate and restructure heterogeneous resources [10,11]. For new ventures, the “liability of newness” and resource constraints pose dual challenges. As a core strategy to break through this dilemma, resource bricolage achieves value creation through the creative reorganization of existing resources at hand [12,13,14]. Baker and Nelson [3] defined resource bricolage as “the process of using existing resources at hand to solve new problems or seize new opportunities,” whose essence is to break through the limitations of the inherent attributes of resources and construct new means–ends relationships (for example, converting idle office space into a product testing room to align resource utilization with innovation goals). Zhu and Li [1] further noted that resource bricolage is not only an expedient measure to passively cope with resource shortages but also a strategic behavior to actively stimulate innovation potential, especially with unique advantages in the field of micro-innovation.

As an incremental improvement strategy, micro-innovation builds competitive advantages through continuous small-scale optimizations, characterized by low risk, short cycles, and proximity to the market [2,15,16]. Zhou et al. [16] noted that micro-innovation focuses on subtle optimizations of products, services, or processes, rather than disruptive changes, which is highly compatible with the resource endowments of new ventures. Based on the research of Zhou et al. [2,16] and Ye [17], this study defines “micro-innovation” as an innovation paradigm through which new ventures build competitive advantages via incremental improvements in products, services, or processes under resource constraints. The core characteristics of micro-innovation include the following: (1) incrementality (local optimization based on existing foundations, rather than disruptive changes), (2) market orientation (taking user demand feedback as the core basis), (3) low risk (low input costs and short trial-and-error cycles). Compared with incremental innovation, micro-innovation places greater emphasis on “resource frugality” (prioritizing the utilization of existing resources over new investments). Compared with frugal innovation, micro-innovation focuses more on the “market demand response,” rather than mere technical simplification. There are clear boundaries between the two in terms of “resource constraint adaptability” and “innovation goal orientation”. Through empirical analysis, Ye [17] found that micro-innovation enhances corporate competitive advantages through the “differentiation–specialization” path, while Zhang et al. [18] revealed the mechanism by which micro-innovation can be achieved from the perspective of ambidextrous capabilities. However, existing studies have mostly focused on the external driving factors of micro-innovation [19,20], with insufficient exploration of the mechanism by which resource bricolage affects micro-innovation through internal capability transformation [21,22].

The “specificity of resource constraints” in new ventures makes them inherently dependent on the mechanism of resource bricolage. Unlike mature enterprises that can break through constraints via external resource purchases, new ventures face issues such as financing barriers and insufficient legitimacy due to the “liability of newness” [1] and thus must rely on the creative reorganization of existing resources to achieve innovation. Moreover, the characteristics of micro-innovation—”low input, rapid iteration, and market proximity”—are highly compatible with ambidextrous learning. Exploratory learning helps new ventures identify unmet market demands when resources are limited (e.g., capturing potential pain points through user interactions), while exploitative learning reduces innovation costs by deepening existing knowledge (e.g., optimizing current technologies to achieve partial improvements). This dual function of “opportunity identification–efficiency optimization” is a core feature that distinguishes new ventures from mature enterprises [2].

Ambidextrous learning theory provides a key perspective for this transformation mechanism. March et al. [5] proposed that ambidextrous learning includes exploratory learning (acquiring new knowledge) and exploitative learning (deepening existing knowledge), noting that the dynamic balance between the two represents the core capability for the continuous innovation of organizations. Cao et al. [6] confirmed that in the process of resource bricolage, exploratory learning explores new uses of resources, and exploitative learning improves the efficiency of resource utilization. Here, the synergistic effect of ambidextrous learning significantly promotes innovation performance. Wu and Liu [22] further found that the essence of resource bricolage is the “utilization and exploration” process of knowledge resources and that ambidextrous learning can transform resource reorganization into innovative achievements. Zhao et al. [23] also verified the mediating role of ambidextrous learning between resources and innovation.

Both institutional theory and stakeholder theory emphasize the influence of the external environment as a key moderating mechanism [24,25]. As an important institutional variable, government support provides resource supplements and legitimacy endorsements for enterprises through financial support and policy inclinations, which may strengthen the promoting effect of resource bricolage on ambidextrous learning. Cui et al. [26] showed that government funding enhances enterprises’ innovation motivation by alleviating resource constraints. Wei and Wan [27] additionally found that government support can improve the efficiency of resource integration and accelerate learning transformation, while Xiong et al. [28] noted that moderate government subsidies can incentivize enterprises’ R&D investment and enhance the feasibility of ambidextrous learning. In addition, China’s unique institutional environment provides contextual support for the core mechanism. First, “guanxi” (interpersonal relationships), as an informal institution, helps new ventures acquire idle resources or information support through social networks, indirectly enhancing the feasibility of resource bricolage [29]. Second, the national “mass entrepreneurship and innovation policy” provides legitimacy endorsement for new ventures through forms such as subsidies for science and technology projects and tax incentives, making government support not only a resource supplement but also a signal of market recognition, which may strengthen this factor’s moderating role in the resource transformation process. These institutional factors collectively constitute the unique context for innovation in new Chinese ventures.

As a key stakeholder interaction mechanism, customer participation affects the transformation of ambidextrous learning into micro-innovation through feedback effects. Bai et al. [30] proposed that customer participation promotes knowledge transfer through information sharing, thereby affecting innovation effectiveness. Fan et al. [31] noted that customer participation is a “double-edged sword,” whose in-depth interactions can guide the direction of exploratory learning. Tian et al. [25] verified that customer participation improves ambidextrous innovation through knowledge transfer. Zhang et al. [32] took Xiaomi mobile phones as an example and confirmed that customer participation can strengthen the correlation between ambidextrous learning and micro-innovation. Behind this Chinese local practice lies the in-depth shaping of the logic of customer participation by cultural contexts: in China’s high-collectivism cultural context [33], customer participation is not merely an individual behavior based on “self-efficacy” (as is the case in individualistic cultural contexts), but rather more closely embedded in group interactions. Willingness to participate stems not only from the functional benefits of product/service optimization, but also from needs such as “maintaining community harmony” and “integrating into user groups”. This tendency synergizes with China’s unique “Guanxi” culture, further strengthening the connection between participatory behavior and interaction quality, and ultimately providing a cultural foundation for enterprises like Xiaomi to achieve micro-innovation through user participation [34].

In summary, existing studies have explored the roles of resource bricolage, micro-innovation, ambidextrous learning, and the moderating variables of government support and customer participation, but there remain research gaps. First, the internal logic of how resource bricolage affects micro-innovation through ambidextrous learning has not been clarified. Second, the pre-moderating effect of government support on the relationship between resource bricolage and ambidextrous learning and the post-moderating effect of customer participation on the relationship between ambidextrous learning and micro-innovation lack systematic discussion. Therefore, this study constructs a mediating model of “resource bricolage–ambidextrous learning–micro-innovation” and introduces government support and customer participation as moderating variables, aiming to improve the theoretical framework.

3. Hypotheses Development

3.1. The Relationship Between Resource Bricolage and Micro-Innovation

The resource-based view holds that resource input is the foundation of corporate innovation, while new ventures face resource acquisition dilemmas due to the “liability of newness” and their “smallness and weakness” [1]. Resource bricolage, centered on breaking through resource constraints, provides possibilities for micro-innovation in new ventures by activating and creatively reorganizing existing resources at hand. The mechanism of this approach can be explored from three perspectives. First, resource bricolage emphasizes in-depth exploration of existing resources. New ventures rely on accessible resources in the market, including easily overlooked user needs or entrepreneurs’ own knowledge and capabilities [1]. In this way, new ventures can explore their potential needs through continuous interaction with users to develop novel resource value. For example, Zhang and Zhang’s [35] study on small and micro technology enterprises in Guangdong showed that Lansheng Pharmaceutical explored user needs based on existing drugs and ultimately achieved micro-innovation in pharmaceutical products. Second, the underlying logic of “making do” and “improvisation” in resource bricolage promotes the rapid iteration of micro-innovation. Under resource constraints, new ventures do not overly focus on the effectiveness of existing resources but promote innovation through proactive improvisational actions [1,36]. This micro-innovation process often disassembles projects into short-term small goals, collects market feedback through repeated testing, and generates a large number of ideas [15,19], which then become an effective path for micro-innovation. Finally, resource reconstruction achieves new purposes by integrating internal and external resources. Fang and Huang [37] noted that resource reconstruction not only serves existing goals but also points to new goals through re-integration. Zhou et al. [2] argued that micro-innovation in small- and medium-sized enterprises relies on cross-border integration mechanisms and that the progress of information technology enables enterprises to interact with external entities such as consumers and peers. The collision of different information generates new ideas, which become a key driving force for micro-innovation. Under this background, the following hypothesis is proposed:

H1. 

Resource bricolage positively affects micro-innovation in new ventures.

3.2. The Relationship Between Resource Bricolage and Ambidextrous Learning

Resource bricolage can promote ambidextrous learning in new ventures, namely exploratory learning and exploitative learning. First, from the perspective of organizational ambidexterity, although new ventures can flexibly carry out exploratory learning, it is difficult to ensure stability, and excessive or long-term exploration may lead to failure. Exploitative learning has strong stability but may produce organizational inertia due to long-term focus [5,22]. As a management activity, resource bricolage can improve both organizational stability and adaptability [38], creating conditions for ambidextrous learning [22,39,40]. Specifically, the “heterogeneous reorganization” (for example, combining technical knowledge, user feedback, and unused equipment to develop a new product) mechanism of resource bricolage enhances organizational adaptability [41,42,43]. By reorganizing waste or idle resources to generate new ideas, this approach helps new ventures adapt to market changes and promotes exploratory learning. The “making-do utilization” mechanism of heterogeneous reorganization, moreover, enhances organizational stability [3,4]. By developing new service attributes for old resources to alleviate resource constraints, this approach also supports exploitative learning. From a knowledge perspective, resource bricolage requires new ventures to deconstruct and reorganize existing resources and knowledge. On the one hand, resource bricolage strengthens application by examining existing resources and relying on past knowledge accumulation, promoting exploitative learning [6,44]. On the other hand, this approach breaks through inherent resource analysis methods to capture new opportunities, promoting exploratory learning [1,42,43]. Senyard [4] also noted that the process of integrating existing resources to reduce constraints in resource bricolage reflects the promotion of applied learning, while sorting heterogeneous resources and re-integrating them to establish competitive advantages requires exploring knowledge in unknown fields, directly promoting exploratory learning. Under this background, the following hypotheses are proposed:

H2a. 

Resource bricolage positively affects exploratory learning in new ventures.

H2b. 

Resource bricolage positively affects exploitative learning in new ventures.

3.3. The Relationship Between Ambidextrous Learning and Micro-Innovation in New Ventures

The promotion of ambidextrous learning on micro-innovation in new ventures stems from this approach’s core role in knowledge accumulation and innovation implementation. Studies by Bontis et al. [45] and Zhou and Feng [46] showed that organizations can accumulate knowledge, achieve breakthrough innovations, and optimize decisions through ambidextrous learning, thereby supporting the development of new ventures. Exploitative learning helps new ventures acquire, utilize, and create knowledge to cope with external challenges, while exploratory learning helps such ventures find solutions to new problems, quickly capture potential market demands, and achieve creative strategies. The organizational learning mechanism itself is an important component of micro-innovation in new ventures [2,15,16,17,47,48]. From the perspective of customer participation, the interaction between enterprises and users involves the transmission of product or service information, and feedback promotes internal trial-and-error learning, advancing micro-innovation in a dynamic and efficient rhythm [2,48]. From the perspective of cross-border integration, new ventures jointly drive micro-innovation by exploring external knowledge, opportunities, and resources while utilizing internal experience [2]. Li’s [47] study on new ventures in Zhejiang Province also showed that the “prudent learning” and “knowledge filtering” links in achieving micro-innovation are highly consistent with the concept of organizational learning—new ventures clarify the direction of micro-innovation through exploratory learning and then use exploitative learning to compare the differences between market demands and themselves, ultimately integrating the results of the two types of learning to achieve micro-innovation. Specifically, for different types of micro-innovation, exploratory learning promotes new product development, product line extension, or market development through multi-channel research and experimentation, expanding the knowledge base and surpassing the venture’s original innovation capabilities [2,46]. Exploitative learning, based on existing resources, knowledge, and capabilities, improves technical capabilities through screening and reorganization, efficiently leverages the value of existing knowledge, and lays the foundation for the development of current technologies, products, and processes [49]. Ye’s [17] empirical research also confirmed that independent micro-innovation relies on internal knowledge accumulation and breakthroughs and that exploitative learning can integrate internal knowledge to achieve this goal, while exploratory learning identifies new market demands by acquiring external knowledge, supporting independent and imitative micro-innovation. Against this background, the following hypotheses are proposed:

H3a. 

Exploratory learning positively affects micro-innovation in new ventures.

H3b. 

Exploitative learning positively affects micro-innovation in new ventures.

3.4. The Mediating Role of Organizational Ambidexterity

Resource bricolage impacts micro-innovation in new ventures through the mediation of ambidextrous learning. This transmission path can be elaborated based on the logic of resource constraints and innovation transformation. Resource bricolage carried out by new ventures under resource constraints is essentially a process of exploring and developing resources. Organizational exploratory learning acquires knowledge through continuous trial and error, expanding the scope of resources [49]. In the face of resource redundancy, resource bricolage can study existing technologies and knowledge through exploitative learning, enrich and improve knowledge, improve the efficiency of information sharing, and control innovation risks [6], providing guarantees for micro-innovation. At the same time, resource bricolage can provide guidance for ambidextrous learning, improving innovation efficiency and quality [6,39]. Senyard et al. [4] and Wu et al. [22] also noted that resource bricolage can control the innovation risks of new ventures, increase the possibility of obtaining results, and create a safe environment for micro-innovation. In addition, external knowledge obtained through exploratory learning can enrich innovation inspiration, and exploitative learning reduces risks through efficient resource bricolage [22]. Both approaches jointly improve the possibility of micro-innovation. Under this background, the following hypotheses are proposed:

H4a. 

Exploratory learning mediates the impact of resource bricolage on micro-innovation in new ventures.

H4b. 

Exploitative learning mediates the impact of resource bricolage on micro-innovation in new ventures.

3.5. The Moderating Role of Government Support

The government plays an important role in ensuring enterprise development in social governance, and the positive impact of government support on innovation in new ventures has been fully verified. In the context of resource scarcity, the government maintains the legitimate rights and competitive status of new ventures by establishing a sound protection system, ensuring that the results of their knowledge exploration and utilization can be effectively transformed into innovative practices; at the same time, as a key resource provider, the government can compensate for the insufficient R&D capabilities of new ventures through R&D support [50] and alleviate the inhibition of resource constraints on innovation through tax subsidies and talent support [51]. In addition, when environmental uncertainty is high, the government can play a safety net role, providing support for enterprises’ knowledge acquisition and integration [52]. Specifically, government support can reduce the trial-and-error costs in the process of resource bricolage, making new ventures more motivated to explore new knowledge through exploratory learning. At the same time, the stable resource supply such support provides can strengthen the in-depth application of existing knowledge through exploitative learning, thereby amplifying the driving effect of resource bricolage on ambidextrous learning.

However, excessive government support may also lead to potential risks. As noted by Cui et al. [26], over-reliance on government subsidies could reduce new ventures’ motivation for independent resource bricolage, forming a “resource-dependence trap” that weakens the initiative of ambidextrous learning. Thus, the moderating effect of government support is more pronounced under moderate support intensity, while excessive support may diminish the positive impact of this support on the relationship between resource bricolage and ambidextrous learning. This pattern suggests that the moderating role of government support may not follow a simple linear trend but could exhibit an inverted U-shaped relationship. Specifically, moderate government support alleviates resource constraints and enhances the effectiveness of resource bricolage in driving ambidextrous learning; yet beyond a certain threshold, excessive support may induce reliance on external resources, reducing enterprises’ willingness to engage in independent resource recombination and thus weakening the positive linkage between resource bricolage and ambidextrous learning. Future research could further validate this non-linear relationship by introducing quadratic terms into the analytical model. Under this background, the following hypotheses are proposed:

H5a. 

Government support positively moderates the impact of resource bricolage on exploratory learning in new ventures.

H5b. 

Government support positively moderates the impact of resource bricolage on exploitative learning in new ventures.

3.6. The Moderating Role of Customer Participation

Customer participation strengthens the promotion effect of ambidextrous learning on micro-innovation in new ventures through knowledge sharing and demand coupling. Tian et al. [25] noted that customer participation builds a bridge for knowledge sharing between enterprises, enabling enterprises to effectively transform and supplement heterogeneous knowledge obtained from customers. By integrating and utilizing this knowledge, the promotion of ambidextrous learning on micro-innovation is further strengthened. Micro-innovation itself has a distinct customer participation orientation [16], and the core of its precise optimization lies in deeply catering to customer needs. By fully capturing customers’ existing or potential demand information, new ventures use internal rapid iterative learning mechanisms to continuously couple product or service prototypes with customer needs, ultimately achieving micro-innovation. Zhang et al.’s [32] research further confirmed that different intensities of customer participation can bring explicit and tacit knowledge to enterprises. Enterprises determine needs by iteratively learning these two types of knowledge and promoting micro-innovation via circular interaction with customers. Consequently, the higher the degree of customer participation, the more accurately the new market knowledge obtained through exploratory learning and the existing knowledge deepened through exploitative learning can match customer needs, thereby enhancing the positive impact of ambidextrous learning on micro-innovation. Nevertheless, customer participation may act as a “double-edged sword” [30]. When customer demands are highly heterogeneous or dispersed, excessive participation could increase coordination costs and lead to conflicting feedback, which may confuse the direction of ambidextrous learning and weaken its positive impact on micro-innovation. This suggests that the moderating effect of customer participation may not follow a simple linear pattern, but rather a potential inverted U-shaped relationship: moderate customer participation (with relatively concentrated and consistent demands) can maximize its role in aligning ambidextrous learning with market needs, thereby strengthening the promotion of micro-innovation; however, once participation exceeds a certain threshold (e.g., overly fragmented demands), its positive moderating effect may diminish or even reverse. Unfortunately, this study only focuses on testing the linear positive moderating role of customer participation under the condition of relatively concentrated demands, without empirically verifying such non-linear relationships. Future research could incorporate quadratic terms of customer participation into the analytical model to further examine its potential inverted U-shaped moderating effect. Under this background, the following hypotheses are proposed:

H6a. 

Customer participation positively moderates the impact of exploratory learning on micro-innovation in new ventures.

H6b. 

Customer participation positively moderates the impact of exploitative learning on micro-innovation in new ventures.

In summary, this study takes micro-innovation as the dependent variable, resource bricolage as the independent variable, ambidextrous learning as the mediating variable, and government support and customer participation as moderating variables, constructing a theoretical model as shown in Figure 1.

4. Research Methodology

4.1. Sample Selection and Data Collection

This study collected data through questionnaire surveys, targeting new ventures established for no more than 8 years [53]. Based on the division of regional entrepreneurial activity in the China Regional Innovation and Entrepreneurship Development Index Report (2020), regions with high entrepreneurial activity such as Shanghai, Beijing, Guangdong, Jiangsu, Zhejiang, and Fujian were selected as survey sites. To ensure the accuracy and objectivity of the questionnaire data, respondents were limited to founders and middle-to-senior managers of the new ventures, who were familiar with their companies’ operations. During the formal survey, all questionnaires were distributed in electronic form, conducted from October 2022 to February 2023 (i.e., lasting 4 months), which fulfilled the expected research needs. Questionnaire samples came from three channels: first, MBA students from Hainan University were invited to fill them out; second, responses were collected through the authors’ social networks; third, professional survey companies visited the target enterprises to collect responses. To control questionnaire quality, attention check questions and response timers were set in the electronic questionnaires. After the survey, 25 samples were randomly selected, and respondents were followed up with via phone numbers provided in the questionnaires for verification. A total of 514 questionnaires were distributed. Invalid samples were excluded, including those with incomplete responses, incorrect answers to attention-checking questions, logical inconsistencies, and excessively short response times. Ultimately, 319 valid questionnaires were recovered, with 195 invalid samples, resulting in a sample recovery rate of 62.06%.

Regarding the sample characteristics, new ventures established for 1–3 years accounted for 13.1%, those established for >3–5 years accounted for 55.2%, and those established for >5–8 years accounted for 31.7%. In terms of firm size, enterprises with 1–20 employees accounted for 4.1%, those with >20–100 employees accounted for 35.1%, those with >100–200 employees accounted for 35.7%, and those with >200 employees accounted for 25.1%. For industry distribution, manufacturing enterprises accounted for 45.1%; information transmission, computer services, and software industries accounted for 17.2%; scientific research and technical services accounted for 12.9%; and other industries accounted for 24.8%.

4.2. Variable Measurement

Mature scales from domestic and international studies were adopted to measure resource bricolage, ambidextrous learning, government support, customer participation, and micro-innovation. These scales were appropriately modified based on a pre-survey to better align with the research objectives and context. All scales used a 5-point Likert scale, where 1 represents “strongly disagree,” and 5 represents “strongly agree”.

Specifically, resource bricolage was measured using the 8-item scale developed by Senyard et al. [4]. Ambidextrous learning was assessed with the 10-item scale developed by Ge et al. [54], which includes 5 items for exploratory learning and 5 items for exploitative learning. Micro-innovation was measured using the 4-item scale developed by Ye et al. [17]. Government support was measured using an 8-item scale integrating the scales developed by Li et al. [55] and Niu et al. [56], and customer participation was assessed using a 4-item scale combining the scales developed by Bai [30] and Carbonell et al. [57] (

Table 1. Research scale.

Construct	Item	Content
Resource bricolage (RB)	RB1	We are confident of our ability to find workable solutions to new challenges by using our existing resources.
	RB2	We gladly take on a broader range of challenges than others with our resources would be able to.
	RB3	We use any existing resource that seems useful to respond to a new problem or opportunity.
	RB4	We deal with new challenges by applying a combination of our existing resources and other resources inexpensively available to us.
	RB5	When dealing with new problems or opportunities, we take action by assuming that we will find a workable solution.
	RB6	By combining our existing resources, we take on a surprising variety of new challenges.
	RB7	When we face new challenges, we put together workable solutions from our existing resources.
	RB8	We combine resources to accomplish new challenges for which the resources were not originally intended.
Exploratory learning (ERL)	ERL1	Our company values acquiring strategic knowledge related to experiments and high market risks in products/services.
	ERL2	Our company tends to collect information without specific market strategic targeting to ensure the smooth progress of new product/service development experiments.
	ERL3	Our company learns new knowledge and experiences to develop products/services.
	ERL4	Our company collects new information beyond existing market and technical experiences.
	ERL5	Our company aims to collect new information that enables us to acquire new knowledge in the development of new products/services.
Exploitative learning (EIL)	EIL1	Our company collects information to improve or solve problems in new product/service development.
	EIL2	Our company aims to search for ideas and information that can ensure production capacity.
	EIL3	Our company seeks commonly accepted methods and solutions to problems in new product/service development.
	EIL4	Our company uses information collection methods to help us understand and update current products/services.
	EIL5	Our company emphasizes the utilization of knowledge related to existing product/service experiences.
Government support (GS)	GS1	We have obtained corporate income tax incentives.
	GS2	We have obtained talent policy subsidies from the government.
	GS3	We have obtained intellectual property rewards.
	GS4	We have obtained support for science and technology projects.
	GS5	We have obtained subsidies for entrepreneurial financing loans.
	GS6	The government has helped the company introduce talents.
	GS7	The government has provided entrepreneurial guidance and counseling.
	GS8	The government has broken unnecessary entry barriers to create a fair market environment.
Customer participation (CP)	CP1	Our company communicates with customers frequently.
	CP2	Our company widely solicits customers’ opinions.
	CP3	Our company invites some customers to join the new service development team.
	CP4	Our company communicates with customers through multiple channels.
Micro-innovation (MI)	MI1	Our company has innovated existing products/services by learning and imitating the ideas and behaviors of pioneering innovators.
	MI2	Our company has improved production/operation processes through independent design and R&D.
	MI3	Our company can independently set goals in innovation activities.
	MI4	Our company has gradually improved products or services through its own efforts.

).

Ye’s [17] study indicated that firm age, firm size, and industry type may affect micro-innovation. To avoid unnecessary errors, firm age, firm size, and industry type were included as control variables. Firm age was coded as 1–3 corresponding to 1–3 years, >3–5 years, and >5–8 years, respectively. Firm size was coded as 1–4 corresponding to 1–20 employees, >20–100 employees, >100–200 employees, and >200 employees, respectively. Industry type was coded as 1–4 corresponding to manufacturing, information transmission, computer services and software, scientific research and technical services, and other industries respectively.

4.3. Robustness Tests

To further address concerns about sampling bias and external environmental interference, supplementary tests were conducted. Independent sample t-tests confirmed no significant differences in the mean values of core variables (resource bricolage, ambidextrous learning, and micro-innovation) across data collection channels or regions (all p > 0.05). One-way ANOVA and subgroup regression analyses showed that the core path relationships remained stable across firm age subgroups (1–3 years, 3–5 years, and 5–8 years) and industry subgroups (manufacturing vs. non-manufacturing), with no statistically significant differences in effect sizes (all p > 0.05).

To rule out potential impacts of the zero-COVID-19 policy, data were divided into two periods: October–December 2022 (strict policy period, N = 174) and January–February 2023 (relaxed policy period, N = 145). Independent sample t-tests revealed no significant differences in core variable means between periods (resource bricolage: 3.68 vs. 3.57, t = 1.153, p = 0.250; micro-innovation: 3.45 vs. 3.33, t = 1.643, p = 0.096), indicating minimal policy interference on the core mechanism. These tests confirm that sampling characteristics and policy changes do not significantly affect the reliability and generalizability of the conclusions.

4.4. Testing for Common Method Bias

This study examined common method bias using Harman’s single-factor test. The results revealed 6 factors with eigenvalues greater than 1. Here, the maximum variance explained by a single factor is 32.22%, which is lower than the 40% threshold. Therefore, there is no serious common method bias in this study.

To further verify the robustness of the data, in addition to the existing Harman single-factor test, this study supplemented the Unmeasured Latent Method Construct (ULMC) test to more strictly control the risk of common method bias. The specific steps were as follows.

A confirmatory factor analysis (CFA) model incorporating a latent method factor was constructed in AMOS 25.0, with the factor loadings of all items loaded simultaneously onto their theoretically assigned constructs and a common method factor. The results showed that after adding the method factor, there was no significant improvement in the model fit indices (ΔRMSEA = 0.003, ΔCFI = 0.003). This result indicates that method variance interfered little with the data and that common method bias did not pose a substantial threat to the conclusions of this study.

4.5. Reliability and Validity Tests

This study used SPSS 25.0 and AMOS 25.0 to test the reliability and validity of the scales. As shown in

Table 2. Reliability test.

Construct	Item	Scale
		Factor Loading	Cronbach’s Alpha	AVE	CR
Micro-innovation (MI)	MI1	0.654	0.807	0.517	0.810
	MI2	0.742
	MI3	0.756
	MI4	0.720
Resource bricolage (RB)	RB1	0.761	0.924	0.602	0.924
	RB2	0.767
	RB3	0.768
	RB4	0.791
	RB5	0.779
	RB6	0.791
	RB7	0.789
	RB8	0.762
Government support (GS)	GS1	0.672	0.882	0.485	0.883
	GS2	0.662
	GS3	0.702
	GS4	0.734
	GS5	0.683
	GS6	0.711
	GS7	0.709
	GS8	0.697
Customer participation (CP)	CP1	0.747	0.803	0.509	0.805
	CP2	0.744
	CP3	0.637
	CP4	0.719
Exploratory learning (ERL)	ERL1	0.825	0.898	0.641	0.899
	ERL2	0.788
	ERL3	0.758
	ERL4	0.822
	ERL5	0.809
Exploitative (EIL)	EIL1	0.831	0.898	0.639	0.898
	EIL2	0.713
	EIL3	0.791
	EIL4	0.829
	EIL5	0.827

, Cronbach’s Alpha values of micro-innovation, resource bricolage, government support, customer participation, exploratory learning, and exploitative learning are all greater than 0.7, indicating good scale reliability. The validity of each variable was measured from four perspectives: content validity, convergent validity, discriminant validity, and construct validity. In terms of content validity, all variables adopted mature foreign scales, which were translated and appropriately modified in combination with the research of domestic scholars to ensure that each item conformed to the current research context. Thus, each scale offered good content validity. Regarding construct validity, the results of the confirmatory factor analysis showed that the six-factor model offered a good fit (χ²/df = 1.562, RMSEA = 0.042, CFI = 0.952, TLI = 0.947, IFI = 0.952). Here, the square root of the average variance extracted (AVE) for micro-innovation, resource bricolage, government support, customer participation, exploratory learning, and exploitative learning is greater than the correlation coefficient between each variable (as shown in

Table 3. Correlation coefficients and discriminant validity.

Pearson Correlation and AVE
Variables	1	2	3	4	5	6	7	8	9
Company size	1
Company age	0.302 **	1
Company industry	−0.162 **	−0.047	1
Exploitative learning (EIL)	−0.005	0.006	0.049	0.799
Exploratory learning (ERL)	−0.049	−0.039	−0.002	0.485 **	0.801
Customer participation (CP)	0.011	0.021	0.057	0.436 **	0.276 **	0.713
Government support (GS)	0.056	0.09	0.092	0.423 **	0.213 **	0.299 **	0.697
Resource bricolage (RB)	−0.084	−0.029	−0.051	0.424 **	0.472 **	0.236 **	0.210 **	0.776
Micro-innovation (MI)	−0.018	0.017	0.053	0.593 **	0.644 **	0.435 **	0.359 **	0.499 **	0.719
Mean	2.818	5.408	2.172	3.527	3.690	3.476	3.676	3.627	3.497
Standard deviation	0.857	1.785	1.243	0.966	0.844	0.909	0.754	0.827	0.894

Note: (1) ** indicates p < 0.01; (2) the bold values on the diagonal are the square root of the AVE values.

), so each scale has good discriminant validity and construct validity. In terms of convergent validity, the composite reliability (CR) values of all variables are greater than 0.7, indicating that each scale possesses good convergent reliability. Consistent with the arguments of Fornell and Larcker (1981) [58] and Hair et al. (2010) [59], a slightly lower AVE (>0.45) remains acceptable when CR is high and the model fit is good. Additionally, the actual measurement performance of the scales, such as all factor loadings exceeding 0.6 (see Table 2), further validated their applicability in this study.

5. Results

5.1. Correlation Analysis of Variables

As shown in Table 3, the six main variables—micro-innovation, resource bricolage, government support, customer participation, exploratory learning, and exploitative learning—exhibit positive correlations with each other. Specifically, resource bricolage is significantly positively correlated with exploratory learning, exploitative learning, government support, customer participation, and micro-innovation; exploratory learning and exploitative learning are significantly positively correlated with government support, customer participation, and micro-innovation; and government support and customer participation are significantly positively correlated with micro-innovation. These results provide preliminary validation for the hypotheses proposed in this study.

5.2. Result Analysis

This study used SPSS 25.0 to perform hierarchical regression analysis to verify whether the research hypotheses are supported. Before regression analysis, a multicollinearity diagnosis was conducted, and the variance inflation factor (VIF) of all variables was less than 2, indicating no multicollinearity in the sample data. The results of the main effect test are shown in

Table 4. Main effect test.

Variable	Dependent Variable: MI
	Adding Three Control Variables	Adding Three Control Variables and RB
Company size	−0.018	0.032
	(−0.290)	(0.586)
Company age	0.012	0.013
	(0.419)	(0.518)
Company industry	0.037	0.061
	(0.898)	(1.718)
RB		0.547 **
		(10.350)
R2	0.003	0.257
Adjusted R2	−0.006	0.247
F value	0.362	27.142 **
ΔR2	0.003	0.253

Note: (1) N = 319; (2) ** indicates p < 0.01 The data listed are standardized β coefficients, with t-values in parentheses. The same is true below.

, which indicates that resource bricolage has a positive impact on micro-innovation. Thus, Hypothesis H1 is supported, allowing further analysis.

This study adopted the method of Wen et al. [60] to test the mediating effect, with the following steps. The total effect of resource bricolage on micro-innovation was verified in the main effect test. The mediating effect was tested in three stages, as shown in

Table 5. Mediating effect test.

Variable	Dependent Variable: ERL		Dependent Variable: EIL		Dependent Variable: MI
	Adding 3 Control Variables	Adding 3 Control Variables and RB	Adding 3 Control Variables	Adding 3 Control Variables and RB	Adding 3 Control Variables	Adding 3 Control Variables and ERL	Adding 3 Control Variables and EIL	Adding 3 Control Variables, RB and ERL	Adding 3 Control Variables, RB and EIL
Company size	−0.043	0.001	0.000	0.046	−0.018	0.009	−0.018	0.031	0.009
	(−0.724)	(0.023)	(0.002)	(0.751)	(−0.290)	(0.175)	(−0.380)	(0.650)	(0.208)
Company age	−0.012	−0.012	0.004	0.005	0.012	0.020	0.010	0.019	0.011
	(−0.443)	(−0.471)	(0.139)	(0.182)	(0.419)	(0.848)	(0.429)	(0.836)	(0.499)
Company industry	−0.007	0.014	0.038	0.06	0.037	0.041	0.014	0.054	0.032
	(−0.185)	(0.413)	(0.861)	(1.505)	(0.898)	(1.253)	(0.448)	(1.723)	(1.062)
RB		0.482 **		0.503 **				0.312 **	0.303 **
		(9.441)		(8.424)				(5.900)	(6.182)
ERL						0.631 **		0.487 **
						(13.140)		(9.430)
EIL							0.596 **		0.484 **
							(14.901)		(11.578)
R2	0.003	0.224	0.002	0.186	0.003	0.357	0.416	0.421	0.480
Adjusted R2	−0.006	0.214	−0.006	0.214	−0.006	0.349	0.409	0.412	0.471
F value	0.334	22.603 **	0.257	17.977 **	0.362	43.582 **	55.971 **	45.581 **	57.727 **
△R2	0.003	0.220	−0.007	0.176	0.003	0.354	0.413	0.064	0.064

** indicates p < 0.01.

. After adding the three control variables, resource bricolage had a positive impact on both exploratory learning and exploitative learning, supporting Hypotheses H2a and H2b, while exploratory learning and exploitative learning had a positive impact on micro-innovation, supporting Hypotheses H3a and H3b. Finally, the independent variable and mediating variables were added to test the mediating effect. The results showed that resource bricolage still had a positive impact on micro-innovation, indicating that exploratory learning and exploitative learning play a mediating role between resource bricolage and micro-innovation, supporting Hypotheses H4a and H4b.

To further verify the robustness of the mediating effect of ambidextrous learning, this study adopts the bias-corrected bootstrap method with 5000 resamples [61] to calculate the bias-corrected 95% confidence interval (CI) of the indirect effect, and the results are shown in

Table 6. Test results of indirect, direct, and total effects using bias-corrected bootstrap method (5000 resamples).

Parameter	Effect	SE	Bias-Corrected 95% CI
			Lower	Upper	p
RB-ERL-MI (Indirect Effect)	0.193	0.041	0.122	0.287	0.000
RB-EIL-MI (Indirect Effect)	0.254	0.046	0.172	0.354	0.000
RB-MI (Direct Effect)	0.161	0.067	0.028	0.293	0.021
RB-MI (Total Effect)	0.608	0.053	0.493	0.702	0.000

. It can be seen that (1) the indirect effect of resource bricolage (RB) on micro-innovation (MI) through exploratory learning (ERL) is 0.193, with a bias-corrected 95% CI of [0.122, 0.287] (does not contain 0); (2) the indirect effect through exploitative learning (EIL) is 0.254, with a bias-corrected 95% CI of [0.172, 0.354] (does not contain 0). Both results confirm that ERL and EIL play partial mediating roles between RB and MI, which is consistent with the stepwise test results in Table 5, further verifying the reliability of Hypotheses H4a and H4b.

Figure 2 presents the structural equation model (SEM) results for verifying the mediating effect of ambidextrous learning, which further visualizes the core path relationships. The model fit indices meet the ideal criteria (χ²/df = 2.261, RMSEA = 0.063, CFI = 0.941, TLI = 0.933), indicating good model fit. From the path coefficients, the following can be seen: (1) resource bricolage has a significant positive impact on both ERL (β = 0.483 **, p < 0.01) and EIL (β = 0.527 **, p < 0.01); (2) ERL (β = 0.526 **, p < 0.01) and EIL (β = 0.366 **, p < 0.01) both have significant positive impacts on micro-innovation; (3) the direct effect of RB on MI remains significant (β = 0.161 *, p < 0.05). These results are consistent with Table 5 and Table 6, further confirming the partial mediating role of ambidextrous learning through SEM.

To test whether the mediating effect varies across industry subgroups (information transmission, computer services, and software industry vs. other industries), Process Model 58 and the percentile bootstrap method (5000 resamples) [62] were used for analysis. The results are shown in

Table 7. Conditional indirect effect of ambidextrous learning on micro-innovation.

Mediating Variable	Level	Level Value	Effect	BootSE	BootLLCI	BootULCI
EIL	−1SD	−0.206	0.175	0.044	0.089	0.262
	Mean Value	0.172	0.153	0.031	0.094	0.219
	+1SD	0.551	0.133	0.046	0.056	0.234
ERL	−1SD	−0.206	0.210	0.050	0.118	0.315
	Mean Value	0.172	0.189	0.035	0.126	0.262
	+1SD	0.551	0.169	0.055	0.077	0.289

Note: Calculated via Process Model 58; Level Value is the standardized score of “information transmission industry” (dummy variable: 1 = information transmission, computer services, and software industry, 0 = others); BootSE = bootstrap standard error; BootLLCI/BootULCI = percentile method 95% confidence interval (5000 resamples).

: (1) for the EIL-mediated pathway, the indirect effect values remained stable across different industry subgroup levels (0.133–0.175), and the 95% CIs all excluded 0; (2) for the ERL-mediated pathway, the indirect effect values were also stable (0.169–0.210), with the 95% CIs all excluding 0. The above results indicate that the mediating effect of ambidextrous learning does not change with industry subgroup. In addition, the heterogeneity tests for firm age and firm size follow the same logic (indirect effects are stable and significant), which are not elaborated here due to space constraints.

Before conducting the formal moderating effect analysis, the variables included in the interaction terms were centered to reduce multicollinearity between variables. as shown in

Table 8. Moderating effect test.

Variable	Dependent Variable: ERL		Dependent Variable: EIL		Dependent Variable: MI
Company size	−0.007	0.004	0.019	0.031	0.001	0.002	−0.021	−0.019
	(−0.139)	(0.078)	(0.328)	(0.552)	(0.016)	(0.043)	(−0.450)	(−0.418)
Company age	−0.016	−0.021	−0.01	−0.015	0.016	0.017	0.008	0.006
	(−0.664)	(−0.861)	(−0.368)	(−0.568)	(0.726)	(0.781)	(0.369)	(0.271)
Company industry	0.004	0.015	0.029	0.041	0.028	0.026	0.009	0.003
	(0.126)	(0.446)	(0.761)	(1.088)	(0.905)	(0.844)	(0.285)	(0.103)
RB	0.454 **	0.453 **	0.413 **	0.412 **
	(8.726)	(8.789)	(7.231)	(7.285)
GS	0.138 *	0.180 **	0.444 **	0.492 **
	(0.401)	(3.031)	(7.057)	(7.545)
RB × GS		0.157 *		0.177 *
		(2.47)		(2.529)
ERL					0.545 **	0.567 **
					(11.614)	(12.099)
EIL							0.519 **	0.528 **
							(11.989)	(12.209)
CP					0.285 **	0.317 **	0.186 **	0.215 **
					(6.538)	(7.162)	(4.064)	(4.521)
ERL × CP						0.140 **
						(3.079)
EIL × CP								0.088 *
								(2.216)
R2	0.238	0.252	0.298	0.312	0.434	0.451	0.445	0.454
Adjusted R2	0.225	0.238	0.287	0.299	0.425	0.440	0.436	0.443
F value	19.510 **	17.541 **	26.577 **	23.596 **	48.051 **	42.708 **	50.243 **	43.212 **
△R2	0.238	0.015	0.298	0.014	0.431	0.017	0.442	0.009

* indicates p < 0.05, ** indicates p < 0.01.

, government support had a positive moderating effect on the relationship between resource bricolage and both exploratory learning and exploitative learning, indicating that government support can strengthen the positive relationship between resource bricolage and exploratory learning/exploitative learning, thus supporting Hypothesis H5a and H5b. Customer participation had a positive moderating effect on the relationship between exploratory learning/exploitative learning and micro-innovation, indicating that customer participation can strengthen the positive relationship between exploratory learning/exploitative learning and micro-innovation, thus supporting Hypotheses H6a and H6b. In the analysis of moderating effects, we conducted multicollinearity tests for models with interaction terms. As shown in

Table 9. Multicollinearity test for models with interaction terms.

Variable	Dependent Variable: ERL			Dependent Variable: EIL
	95% CI	Collinearity Diagnosis		95% CI	Collinearity Diagnosis
		VIF	Tolerance		VIF	Tolerance
Company size	−0.098~0.106	1.15	0.869	−0.080~0.143	1.15	0.869
Company age	−0.069~0.027	1.114	0.898	−0.068~0.037	1.114	0.898
Company industry	−0.052~0.083	1.064	0.939	−0.033~0.115	1.064	0.939
RB	0.352~0.555	1.066	0.938	0.301~0.523	1.066	0.938
GS	0.063~0.297	1.172	0.853	0.364~0.619	1.172	0.853
RB × GS	0.032~0.282	1.129	0.885	0.040~0.314	1.129	0.885

and

Table 10. Multicollinearity test for models with interaction terms.

Dependent Variable: MI	95% CI	Collinearity Diagnosis		95% CI	Collinearity Diagnosis
		VIF	Tolerance		VIF	Tolerance
Company size	−0.090~0.094	1.130	0.885	−0.111~0.072	1.128	0.887
Company age	−0.026~0.061	1.102	0.908	−0.037~0.049	1.103	0.907
Company industry	−0.034~0.086	1.032	0.969	−0.057~0.063	1.038	0.963
ERL	0.475~0.659	1.113	0.899
EIL				0.443~0.613	1.244	0.804
CP	0.230~0.404	1.151	0.869	0.121~0.309	1.336	0.748
ERL × CP	0.050~0.229	1.110	0.901
EIL × CP				0.010~0.167	1.138	0.878

, the variance inflation factor (VIF) values of all interaction terms are less than 2, and the tolerance values are greater than 0.5. Additionally, the 95% confidence intervals of all interaction terms do not contain 0, indicating that there is no serious multicollinearity issue in the models and that the results are reliable.

To more intuitively reflect the moderating effects of government support on the relationships between resource bricolage and micro-innovation, exploratory learning, and exploitative learning, as well as the moderating effects of customer participation on the relationships between exploratory learning, exploitative learning, and micro-innovation, this study used the standard deviation of government support and customer participation to plot the impact of resource bricolage on exploratory learning and exploitative learning under different levels of government support, as well as the impact of exploratory learning and exploitative learning on micro-innovation under different levels of customer participation. As shown in Figure 3 and Figure 4, when government support is high, the positive impact of resource bricolage on exploratory learning and exploitative learning is stronger than when government support is low, i.e., higher government support strengthens the positive relationships between resource bricolage and exploratory learning/exploitative learning. As shown in Figure 5 and Figure 6, when customer participation is high, the positive impact of exploratory learning and exploitative learning on micro-innovation is stronger than that when customer participation is low; i.e., higher customer participation strengthens the positive relationships between exploratory learning/exploitative learning and micro-innovation.

To quantify the practical impact intensity of each key path, this study calculates Cohen’s f² (effect size) based on the Adjusted R² of hierarchical regression models (raw R² used if Adjusted R² is negative), with the following criteria: f² = 0.02 (small), 0.15 (medium), 0.35 (large) [63].

Table 11. Summary of Cohen’s f² for key pathways.

Core Pathways	Cohen’s f2	Effect Size Category	Data Source
RB→MI (Direct Effect)	0.324	Medium-to-Large	Table 4
RB→ERL	0.268	Medium	Table 5
RB→EIL	0.211	Medium	Table 5
ERL→MI	0.531	Large	Table 5
EIL→MI	0.687	Large	Table 5
GS moderates RB→ERL	0.017	<Small	Table 8
GS moderates RB→EIL	0.017	<Small	Table 8
CP moderates ERL→MI	0.027	Small	Table 8
CP moderates EIL→MI	0.013	<Small	Table 8

Note: Calculated via $f^2={\displaystyle \frac{R_{full}^2-R_{reduced}^2}{1-R_{full}^2}}$.

shows the following: (1) core mechanism pathways (RB→MI, RB→ERL/EIL, ERL/EIL→MI) have medium-to-large effects, especially ERL/EIL→MI (large effects), confirming their strong explanatory power for micro-innovation; (2) three moderating pathways have f² < 0.02, but all are significant (Table 8: β = 0.157 */0.177 */0.088 *), indicating reliable moderating roles despite small effect sizes.

6. Conclusions and Discussion

Based on the perspective of ambidextrous learning, this study empirically examined the relationships and mechanisms among resource bricolage, ambidextrous learning, micro-innovation in new ventures, government support, and customer participation using hierarchical regression analysis with questionnaire data from 319 new ventures. In the specific Chinese context, this study enriches the application of relevant theories in the field of micro-innovation among new ventures, providing new perspectives and directions for subsequent theoretical research. The core values of this study are as follows: (1) this work reveals the mediating role of ambidextrous learning between resource bricolage and micro-innovation, clarifies the differentiated functions of exploratory learning and exploitative learning, and offers detailed explanations for the resource transformation mechanism; (2) it verifies the moderating effects of government support and customer participation, clarifies the role of external contextual factors in different transformation stages, and enriches the contextualized research on resource bricolage theory; (3) it provides practical implications for new ventures under resource constraints, namely activating internal potential through resource bricolage and achieving micro-innovation by integrating ambidextrous learning with external support, thus offering a reference path for similar enterprises.

6.1. Research External Validity Boundaries

The applicable context of this study’s conclusions should be defined in conjunction with sample characteristics. The research samples were derived from regions with high entrepreneurial activity in China, so its conclusions are more applicable to the context of “emerging markets with strong government intervention and significant resource constraints”. In resource-abundant developed economies, new ventures may rely more on external resource acquisition, rather than bricolage, and the intensity of resource bricolage’s role may vary. In markets with low government intervention, the moderating role of market factors such as customer participation may be more prominent. Future research could further verify the universality of the core mechanisms through cross-regional or cross-national sample comparisons.

6.2. Research Implications

(1) Managerial implications. New ventures should formulate differentiated strategies based on industry type and resource constraint levels. From the perspective of segmented industries, new technology-oriented ventures (such as those in the information transmission, computer services, and software industries, accounting for 17.2% of the sample) should focus on aligning exploratory learning with government technological support. For example, such ventures could obtain R&D funds by applying for technological project support, which could be used to test new technology or explore the functions of new products, and leveraging exploratory learning to break through technological bottlenecks. New manufacturing ventures (accounting for 45.1% of the sample) should strengthen the driving role of customer participation in exploitative learning, e.g., establishing quarterly feedback mechanisms with core customers to collect suggestions for optimizing production processes or improving product details and refining existing production technologies and service models through exploitative learning.

From the perspective of segmented resource constraint levels, highly resource-constrained enterprises (with weak resource bricolage capabilities) should prioritize reducing the trial-and-error costs of exploitative learning through customer participation. For example, they could rely on high-frequency customer feedback to quickly adjust product parameters and thus avoid wasting scarce resources on low-value experiments. Medium resource-constrained enterprises could balance exploratory and exploitative learning by allocating 50–60% of government subsidies to exploratory learning (such as testing new market demands) and 40–50% to exploitative learning (such as upgrading existing products based on user data) to improve resource utilization efficiency. Low resource-constrained enterprises (with strong resource bricolage capabilities) could expand the scope of exploratory learning, e.g., by collaborating with universities to acquire cutting-edge knowledge, while consolidating existing market advantages through exploitative learning.

(2) Policy implications. Governments should provide targeted support for different types of new ventures and varying levels of resource constraints. New technology-oriented ventures should focus on the supply of technological resources. Such ventures could expand the coverage of subsidies for entrepreneurial financing loans, increase incentives for intellectual property rights (especially invention patents for core technologies), and incentivize exploratory learning. New manufacturing ventures should strengthen their demand-side guidance. For instance, they could establish enterprise–customer docking platforms to reduce customer participation costs and provide 30% subsidies for equipment upgrades to process improvement projects based on customer feedback to support exploitative learning. Enterprises with high resource constraints could implement “resource relief policies”. Such policies could alleviate pressure on basic resources through corporate income tax reductions and exemptions, as well as provide support for talent introduction, creating conditions for effective learning and innovation.

6.3. Research Limitations and Prospects

This study has four main limitations. First, this study used cross-sectional data collected from October 2022 to February 2023, making it difficult to capture the dynamic evolution of resource bricolage, ambidextrous learning, and micro-innovation. Thus, the existing results cannot fully reflect the long-term mechanisms between variables. Moreover, cross-sectional data make it hard to completely rule out reverse causality between variables (e.g., successful micro-innovation may help enterprises accumulate more resources, which in turn promotes resource bricolage behavior), leading to potential endogeneity issues that affect the robustness of causal inferences. Second, the micro-innovation scale was adapted from the existing literature, and its generalizability across new ventures in different industries remains to be verified with additional samples. Third, contextual variables were limited to government support and customer participation, excluding other potential moderating variables such as industry competition intensity and entrepreneurial team characteristics, which may restrict the comprehensiveness of the theoretical framework. Fourth, social desirability bias in data collection was not addressed. Although no significant differences in core variables were found across data collection channels or regions, respondents to self-reported questionnaires may overstate positive outcomes and understate problems, and no statistical methods were used to control this bias.

Future research could advance in six directions. First, a longitudinal tracking design could be adopted to conduct a 3–5-year follow-up survey of the 319 new ventures, collect annual panel data, dynamically observe the temporal changes in resource bricolage, ambidextrous learning, and micro-innovation, and clarify the long-term associations among them. Meanwhile, endogeneity issues can be further controlled through the instrumental variable method (e.g., using regional differences in resource endowments as instrumental variables) to enhance the robustness of causal inference. Second, research on contextual variables can be expanded by examining the moderating effects of variables such as industry competition (e.g., the impact of market competition intensity on the effectiveness of resource bricolage) and entrepreneurial team heterogeneity (e.g., the role of team knowledge structure in balancing ambidextrous learning), so as to improve the contextual adaptability of the theoretical framework. Third, cross-context validation can be conducted by comparing realities across different industries (e.g., tech startups vs. new manufacturing ventures) to test industry-specific differences in the core mechanisms; the research can also be extended to cross-country/cross-cultural comparisons to explore the generalizability boundaries of the theory. Fourth, subsequent studies can incorporate quadratic terms of government support and customer participation to test their non-linear moderating effects. Existing studies mainly focus on their linear moderating roles, but in practice, government support may have a “too much of a good thing” effect (e.g., excessive subsidies leading to inertia from resource dependence), and customer participation may also cause problems such as surging coordination costs due to overly heterogeneous demands. Testing via quadratic terms can more accurately reveal their moderating patterns. Fifth, subsequent studies can combine multiple case studies (e.g., selecting 3–5 high-micro-innovation enterprises) to reveal the practical paths of resource bricolage (e.g., “idle resource identification—cross-departmental reorganization—market feedback iteration”) through in-depth interviews, and enhance the ecological validity of conclusions through triangulation (e.g., cross-validation with internal enterprise documents, industry reports, and interview data), thereby providing more operable practical references for new ventures. Sixth, in future data collection, a social desirability scale can be included as a control variable, or the marker variable method can be used to adjust parameters via structural equation modeling to reduce the impact of this bias.