Big Data Capability and Sustainable Competitive Advantage: The Mediating Role of Ambidextrous Innovation Strategy

Abstract

After the rapid expansion of data variety, velocity and volume, human civilization experienced rapid changes in the period between the “IT” and “Data” ages. Researchers view big data and data capability as new sustainable competitive advantages that enhance the sustainability of organizational development. This paper aims to develop and empirically test a framework that will investigate how big data capability is achieved through exploitative, explorative and ambidextrous modes of innovation strategies, and will also explore how they can, in turn, build firms’ sustainable competitive advantage. Using data from surveys of 229 respondents working in Chinese manufacturing firms, we test the framework using regression and bootstrapped mediation analyses. It also shows how big data capability will make firms more inclined to implement exploitative innovation strategy and construct sustainable competitive advantage, as opposed to explorative innovation strategy; when viewed from an ambidextrous perspective, combined dimension of ambidextrous innovation strategy is found to partially mediate between big data capability and sustainable competitive advantage while balanced dimension of ambidextrous innovation strategy does not. The conclusions are of great significance because they will help firms to deal with challenges that arise in big data applications and digital transformation. The findings offer new insight into the strategic choices of organizational innovations.

1. Introduction

The explosive growth of information and tremendous advancements in science and technology have resulted in human civilization experiencing rapid change in the period between the “IT” and “Data” ages. New data is being produced from all industries at an unprecedented rate. International Business Machines Corporation (IBM) observed that in excess of 2.5 trillion bytes of data are generated globally every day, with unstructured data accounting for 90 percent of this amount [1]. However, the phenomenon of the “big data productivity paradox” in management practices has often been observed.

The growing application of big data notwithstanding, it is hugely important for a firm to understand how to utilize big data to transform its outdated value creation, as this will provide competitive advantages in the new economy [2]. Of the organizational capabilities, big data capability (BDC) has become key to accumulating the power of data and capitalizing on its value in order to create competitive advantages for businesses (see Appendix A for a check list of abbreviations). This is the ability to identify sources where large volumes of various kinds of data flow out at high speed, and to collect, store and analyze big data with the aim of accomplishing the firm’s strategic and operational goals [3]. Most studies have primarily focused on the definition, measurements and dimensions of big data-related capabilities, and have also addressed their effects on organizational performance [4,5,6], supply chain performance [7,8], organizational innovation [9,10,11], strategic decision-making [12,13] and competitive advantage [14,15,16]. Some researchers have dynamically and systematically taken BDC into account with the Dynamic Capability Theory [3,9] framework to further clarify the mechanism for transforming BDC into organizational actions and operational value. In providing a way to enhance the performance expression of organizational ability, sustainable competitive advantage (SCA) has a crucial contribution to make to firm’s core competitiveness, survival and sustainability [17,18]. As noted in the reference, BDC is an important factor that influences the development of SCA for companies in the digital economy context [15].

Research into the subject has been mostly restricted to limited perspectives of organizational behavior [19], organizational capacity and employees’ activity [20]. Innovation Strategy Theory provides appropriate strategic management techniques and measures that can be used to augment the impact of the firm’s innovation activities on firm growth and performance [21], and it is well-suited to provide theoretical insight into how BDC can be transformed into SCA. The literature has focused in particular on the mechanisms that may enable big data analytics capabilities to enhance the strategic flexibility of organizations through ambidextrous innovation [22] and has also confirmed the impact that big data-related capabilities can have on organizational capabilities by providing a competitive advantage [23]. From this, we deduce that organizations can build BDC into the resource and capability guarantee for innovation strategy implementation that will ultimately drive innovation strategy selection and implementation. In addition, we also propose that this will promote an ambidextrous development that helps to achieve SCA and superior performance.

Innovation strategy is the process of planning for the implementation of innovative practices and is critical for businesses to remain competitive [24]. From a single perspective, exploitation and exploration represent two opposite modes of innovation strategy based on differences in the magnitude of innovation [24]. Exploitative innovation strategy (EIIS) is market-oriented and adapts innovation practice on the basis of existing knowledge and learning about how to respond to user needs [25,26]; exploratory innovation strategy (ERIS), meanwhile, is technology-oriented and breaks down old technological barriers to gain a competitive advantage by experimenting with new knowledge and fields [25,27]. Although there are logical differences between the two, they complement each other. From an ambidextrous perspective, innovation strategy can be classified into balanced dimension of ambidextrous innovation strategy (BDAIS) and combined dimension of ambidextrous innovation strategy (CDAIS) [28,29]. The ”balanced dimension of ambidextrous” perspective seeks to reduce the risk of innovation imbalance and resource allocation imbalance by narrowing the gap between the two, thereby avoiding polarization and over-skewing, and realizing innovation by “avoiding shortcomings”; meanwhile, research based on the “combined dimension of ambidextrous” perspective centers on the theory of synergy, which clarifies how strengthening the complementarity between the two can enable the “width” of technology to be broadened and the “depth” of technology to be deepened, in a way that helps to achieve the “growth” of innovation [28].

Some research into innovation strategies has engaged from a single perspective [30,31], which indicate that innovation strategy enables the business organization to have a successful formulation and implementation of innovation related activities [30]. Chen et al. [31] state that in the group of organizations exhibiting either exploratory or exploitative innovation strategy, the more similar the organizational culture configurations are to those of the top performers, the higher their innovation speed and innovation quality are. Other studies have begun to focus on how the relatively balanced or combined relationship between EIIS and ERIS can impact on firm performance. Luo et al.’s [29] research indicates that balanced dimension and combined dimension of ambidextrous strategy have different correlations with firm performance in growth and maturity stage. As suggested by previous studies, the effect of big data analytics capability on organizational performance could be mediated by exploitative innovation, explorative innovation or ambidextrous innovation [1,19]. However, the question of why some firms prefer one innovation strategy mode over another, and the related question of the specific insights that different levels of BDC can offer in this regard, have been largely overlooked. There is still an incomplete understanding about the mediation mechanism of innovation strategy between data capability and sustainable competitive advantage from single or ambidextrous perspective. Those studies that have engaged with these questions have rarely done so in sufficient theoretical detail or depth. Research into how EIIS and ERIS work independently and collaboratively in the wider context of the relationship between BDC and SCA therefore needs to be further clarified, and this is the essential contribution that this paper seeks to make. It achieves this by drawing on the perspective of both singularity and ambidexterity to explore how innovation strategy mediates between BDC and SCA.

We seek to answer the following research questions:

RQ1: What effect does BDCs have on firm’s SCA?

RQ2: From a single perspective, does the firm’s EIIS and ERIS influence the relationship between BDC and SCA? If so, through what mechanism?

RQ3: From an ambidextrous perspective, does the firm’s BDAIS and CDAIS influence the relationship between BDC and SCA? If so, through what mechanism?

According to Dynamic Capability, Innovation Strategy and Organizational Ambidexterity Theory, this study aims to analyze the role that BDC plays in influencing SCA under the mediations of innovation strategy from both single and ambidextrous perspectives. The research expands basic research related to BDC and provides suggestions for how organizations can seek to achieve sustainable development by drawing on BDC. Overall, the results of this study offer useful guidelines to help firms comprehend the important role of BDC on SCA.

2. Theoretical Background and Research Hypotheses

2.1. Big Data Capability and Sustainable Competitive Advantage

SCA is a capability (or set of capabilities) or resource (or set of resources) that current or potential competitors will find it difficult to duplicate. It reflects a quality that outperforms competitors within a specific timeframe and a position at the forefront of the competitive market [32]. The theory of Dynamic Capability establishes that businesses can gain SCA by integrating and allocating resources that enable them to quickly respond to changing customer demands and technical opportunities [33]. BDC is the dynamic capability that enables firms to identify, collect, store and analyze large amounts of internal and external data of different types that flow at high speeds. It enables them to extract potential business value that supports the accomplishment of strategic and operational goals [3] and is an important factor that affects the construction of SCA in the digital economy [15].

Firms with BDC are better positioned to gain and maintain a competitive advantage because they can meet a greater variety of market demands more frequently. In general, BDCs benefit SCA in four ways.

First, an organization’s ability to acquire, prepare and analyze large amounts of data could make a difference, especially if it is able to make these procedures difficult to imitate. As described in the existing literatures, the acquisition and availability of data resources makes a critical contribution by helping companies make decisions and build competitive advantages [10,15]. BDCs help enterprises identify, collect, store and rapidly analyze large amounts of data of various types that flow at high speeds, which lays a foundation for the construction of a data-driven competitive advantage [3].

Second, while data is itself a core resource, it is also important for firms to possess an infrastructure that is capable of storing, sharing and analyzing it [34]. In this perspective, the technical infrastructure, which is related to enterprise competitiveness [35], provides a basis for enterprises to achieve daily operations and improve business management capabilities. Big data analytics infrastructures, which are the ensemble of information systems capable of collecting, storing, processing and analyzing large amounts of data, should be able to adapt to different data types [19]. Ferraris et al. [6] observe that if an enterprise has a sufficient number of software applications and hardware devices that support big data and its analysis, it will be able to efficiently transfer and share massive amounts of unstructured data that are scattered across different organizational information systems. This greatly reduces the time and capital costs of processing large amounts of scattered data and improves management efficiency. Conversely, it is difficult to internalize data and turn it into a competitive advantage [36].

Third, from a technological perspective, the construction of competitive advantage and the improvement of sustainable performance in the big data era depends on the mastery of big data and the use of predictive analytics technology to drive innovation [37]. Big data analytics could overcome the limitations of traditional market analytics by providing the real-time speed demanded by rapid technological change. Xu et al. [38] show how enterprises use BDCs to achieve data mining, text analytics, web analytics, mobile analytics and other data analytics technologies and methods that support efficient operations management. By making it possible to code, compute and analyze massive amounts of structured and unstructured data, these help to: (1) identify the factors that hinder the enterprise’s internal operational efficiency; (2) more accurately distinguish customer and non-customer needs; (3) grasp the subjective initiative; and (4) establish forward-looking innovation in operation management that could build a competitive advantage.

Finally, from a strategic perspective, BDCs provide enterprises with digital and intelligent “minds” [38,39]; improve strategic flexibility through knowledge management and innovation [22]; and encourage enterprises to continuously optimize and improve their development strategies in order to avoid the “inertia trap” caused by the low timeliness of resources. On this basis, we hypothesize that:

Hypothesis 1 (H1).

Big data capability is positively related to sustainable competitive advantage.

2.2. The Mediating Effect of Pure Exploitative and Explorative Innovation Strategy

The theory of innovation strategy suggests that the strategic allocation of innovation resources and innovation activities will be key to building competitive advantages in organizations [21]. Referring to the difference in the magnitude of innovation, we classify innovation strategies as EIIS and ERIS from a single perspective. On the basis of different theories, we analyze the mediation mechanism effect of EIIS and ERIS between BDC and SCA.

Market-oriented theory suggests that companies use EIIS to integrate and utilize existing knowledge and resources; improve organizational processes and structures in order to adapt and improve innovation practice; optimize and upgrade existing products and services; and respond to the needs of existing customers and markets in a timely manner [25,26,40]. Rialti et al.’s [41] study, which is based on the big data context, shows that firms can drive exploitative innovation and improve organizational performance by building BDC. On the one hand, big data analysis infrastructures are highly flexible, interoperable and scalable, and facilitate the flow and sharing of heterogeneous data across different contexts within an organization [42]. They allow timelier detection of production line failures; enable the improvement of production processes; make it possible to address established services’ defects and operational barriers; and improve decision-making efficiency and enable incremental innovation and organizational improvement [43,44].

On the other hand, companies cultivate BDC with the aim of identifying, collecting and storing large amounts of higher quality data that come in various types and flow at high speeds [3,45]. Big data-based market analysis deepens organizations’ understanding of the market’s and customer’s behavioral patterns and preferences, and this in turn drives improvements in products and services that meet the needs of existing markets or customers, enhance customer stickiness and, in turn, increase organizational competitiveness [46]. On this basis, we hypothesize that:

Hypothesis 2 (H2).

When the level of big data capabilities is higher, a firm will be more inclined to build a sustainable competitive advantage by implementing exploitative innovation strategy.

Technology-oriented theory establishes that ERIS helps enterprises break the original technical barriers and gain a competitive advantage by engaging with new knowledge and fields [25,30]. The implementation of ERIS can help companies target a unique market niche and build a competitive advantage based on new products or services [26]. BDCs can provide enterprises with the heterogeneous resources needed to implement exploratory innovation, stimulate forward-thinking patterns and create unique innovation knowledge sets [39]. These can help companies mine business value through descriptive analysis, make trend judgments through predictive analysis, reduce uncertainty (in consumer demand, product supply and among competitors) and enhance companies’ awareness of the outside world [47]. The alertness and adaptability to changes in the external environment will enable enterprises to explore market trends and potential customer needs in a timely manner, enhance technological advancement and tap into potential customer needs.

In other words, once a high level of BDCs is attained, enterprises could be able to mine, test and analyze massive amounts of data, and this will in turn greatly improve the reliability and accuracy with which they identify customer needs, position market segments and provide enterprises with more opportunities and space to implement exploratory innovation [20]. On this basis, we propose that BDCs improve firms’ ability to integrate diverse external information and knowledge that enables them to cope with exploratory innovation uncertainty; create value that competitors struggle to imitate; and, consequently, drive firms to build SCA. We accordingly propose a hypothesis that competes with H2.

Hypothesis 3 (H3).

When the level of big data capabilities is higher, a firm will be more inclined to build a sustainable competitive advantage by implementing explorative innovation strategy.

2.3. The Mediating Effect of Ambidexterity Innovation Strategy

There are logical differences and complementarities between EIIS and ERIS. From an ambidextrous perspective, innovation strategy can be classified into BDAIS and CDAIS. Based on resource-based theory and strategic synergy, this study investigates the mediation mechanism effect of BDAIS and CDAIS in the causal chain of BDC that acts on SCA. Resource-based theory refers to BDAIS and emphasizes that EIIS and ERIS co-exist in an organization, jointly compete for its scarce resources and are self-reinforcing in character [24]. We draw on Cao et al.’s [28] research to argue that companies can achieve lasting competitiveness by referring to resource allocation and risk control, accurately grasping the balance between EIIS and ERIS and enhancing the organization’s adaptability to the external environment. BDCs are dynamic and enable enterprises to identify, acquire, store and analyze large amounts of diverse types of internal and external data that flow at a high speed. They can also be used to extract potential business value that supports the achievement of strategic and operational goals [3]; contribute rich heterogeneous data, technology, tools and skill resources to enterprises; and also shape a data-driven culture in the capability building process.

The process of building a data-driven culture, which is based on data analysis, updating technology and learning about tools and skills, enables processes to be improved and the potential value of resources to be explored. It also improves resource scheduling capabilities to make full use of redundant resources and enhances the technical ability of enterprises to develop new products and services [34,48]—this alleviates resource competition between EIIS and ERIS by better aligning the two, and this in turn facilitates performance improvement. We propose a hypothesis on this basis.

Hypothesis 4 (H4).

When the level of big data capabilities is higher, a firm will be more inclined to build a sustainable competitive advantage by implementing the balanced dimension of ambidextrous innovation strategy.

CDAIS, which is based on the theory of strategic synergy, emphasizes that when a strategy is matched with other related organizational strategies, it will be possible to achieve strategic goals with half the effort. The coordination of EIIS and ERIS will realize complementary advantages and mutually promote them, and this will in turn contribute to improving performance [24,26,28]. Exploitative innovation is, in the first instance, the foundation of exploration innovation. Enterprises will only be able to achieve quantitative and qualitative change by progressively innovating, whether by creating new knowledge or technology, developing new products or services, achieving breakthrough development by gaining a fuller understanding of existing business fields and stock knowledge or by deeply exploring existing resources and capabilities [49]. In the second instance, a high level of exploratory innovation can stimulate further improvement of existing products and marketing tactics, and can also, by developing new products or technologies, improve the competitive position of the firm’s other items in the market.

In a rapidly changing environment, big data analytics capabilities will help companies more quickly and accurately identify areas where they can gain a competitive advantage [9], and this will in turn enable them to allocate and share resources from a systematic and global perspective in a way that helps to ensure synergy in innovation strategies. Alternatively, BDC can be conceived and understood as an adaptive mechanism that provides companies with a clearer understanding and judgment of innovation strategy and path selection; by helping to break cognitive inertia and limitations, it enables them to free themselves from the constraints of experience-based decision-making and develop a comprehensive and systematic understanding of their environment. On this basis, it can be deduced that CDAIS serves as a vital link between BDC and SCA. We therefore propose a hypothesis that competes with H4.

Hypothesis 5 (H5).

When the level of big data capabilities is higher, firms will be more inclined to build a sustainable competitive advantage by implementing the combined dimension of ambidextrous innovation strategy.

The theoretical model of this research is shown in Figure 1.

3. Methodology and Data Sources

3.1. Sample and Data Collection

This study commissioned a third-party professional research organization to collect questionnaires in China, which ensured data availability and validity by limiting the sample collection conditions to enterprises with experience of big data applications. The data-collection process lasted for approximately 3 months (October–December 2020). The subjects were working employees—primarily senior/middle/junior managers and technicians who were given certain bonus rewards and various other enticements. After data cleaning and sample screening, 346 sample data were collected, and 229 valid questionnaires were retrieved, with an effective response rate of 66.18 percent. The samples came primarily from Beijing, Fujian, Guangdong, Hebei, Hunan, Shandong, Shanxi, Sichuan and Zhejiang, along with other provinces and cities with a consistent geographical distribution. The other specific descriptive statistical results of the sample are shown in

Table 1. Descriptive statistics of interviewed companies and personal characteristics.

Characteristics
Category	Classification	Samples	Percent (%)
Gender/Sex	Male	110	48.0
	Female	119	52.0
Position	Senior managers	19	8.3
	Middle managers	69	30.1
	Junior managers	70	30.6
	Others	71	31.0
Enterprise Age	8 years or less	81	35.4
	9–19 years	86	37.6
	20–30 years	42	18.3
	More than 30 years	20	8.7
Number of employees	Less than 20	29	12.7
	20–299	119	52.0
	300–999	39	17.0
	1000 or more	42	18.3
Enterprise Ownership	State-owned enterprises	41	17.9
	Private enterprise	137	59.8
	Others	51	22.3
Years of big data application experience	Less than 3 years	107	46.7
	3–6 years	89	38.9
	More than 6 years	33	14.4
Industry Type	Computer, communications and other electronic equipment manufacturing	104	45.4
	General and special equipment manufacturing	26	11.4
	Electric Equipment and Machinery	25	10.9
	Furniture Manufacturing	14	6.1
	Others	60	26.2
Total respondents = 229

.

3.2. Measures

We developed the instruments based on previously validated measures. All the items in the questionnaire were measured by using a 7-point Likert scale, from “strongly disagree” to “strongly agree”.

Table 2. Coefficients of reliability and validity.

	Items	Cronbach’α	Factor Loading	CR	AVE
BDC	16	0.959	0.665–0.861	0.958	0.586
EIIS	4	0.796	0.608–0.750	0.773	0.461
ERIS	4	0.846	0.721–0.795	0.849	0.585
SCA	6	0.886	0.665–0.816	0.892	0.580

presents the information about the measurement scale in the current study of the measurement of BDC, EIIS, ERIS and SCA.

3.2.1. Big Data Capability

According to Lin and Kunnathur [3] measurement indicators, BDC can be measured by 16 items. (see Appendix A for a check list of variables items).

3.2.2. Innovation Strategy

The innovation strategy was measured by eight items adopted from He and Wong’s study [24]. This measurement approach had two dimensions, specifically EIIS and ERIS. Furthermore, BDAIS was based on the organic equilibrium view [50], in which x represents the EIIS, y represents ERIS and the Equation (1) is used to measure the relative equilibrium of the two. The equilibrium is 1 when the EIIS and ERIS are equal, and the equilibrium value is close to 1 when they are not.

$$BDAIS=1-\left|x-y\right|∕\left(x+y\right)$$

(1)

We referred to Cao et al. [30], who calculate CDAIS by multiplying EIIS and ERIS, which is:

$$CDAIS=xy$$

(2)

3.2.3. Sustainable Competitive Advantage

We drew on Chang’s six-item scale to measure SCA. Its six dimensions (corporate image, difficulty of competitor imitation, management capability, products and services, profitability and R&D capability) include both products and services [51].

3.2.4. Control Variables

The realization of the value of a firm’s BDCs is also influenced by factors that include firm age, size and type and industry type [19,34]. These factors were then included as control variables in this study. “1” was assigned to firms that had existed for less than 8 years; “2” to those that had existed for 9–19 years; “3” to those that had existed for 20–30 years; and “4” to those that had existed more than 30 years. The firm size was the number of employees; “1” was assigned to firms with less than 20 employees; “2” to firms with between 20–299 employees; “3” to firms with between 300–999 employees; and “4” to firms with 1000 or more employees. “1” was assigned to state-owned enterprises, and “0” to all other types were assigned to “0”. With regard to industry type, “1” was assigned to computer, communication and other electronic equipment manufacturing and “0” to all other manufacturing industry types.

4. Empirical Results and Analysis

4.1. Common Method Bias Test

This study used process and statistical control to circumvent the influence of the common method bias and guarantee the objectivity of the survey results. It was made clear to respondents that the contents would be kept strictly confidential and would only be used for academic research. They were collected anonymously to reduce the psychological pressure on respondents and to use process control to obtain more realistic information. Statistical control that used Harman’s one-way test indicated that the first principal component explained 49.459 percent of the total variance in the unrotated case, which was less than the 50 percent threshold proposed by Hair et al. [52]. A test for common method bias that used a latent variable that is uncorrelated with the other factors was used to include a latent variable in the four-factor model through AMOS; the results showed that the model fit metrics performed better when the latent method factor was included (χ²/df = 1.376, RMSEA = 0.041, CFI = 0.973, TLI = 0.966, IFI = 0.973), and the change values of TFI and CFI (∆TLI = 0.012, ∆CFI = 0.014) did not exceed the 0.05 threshold suggested by Bagozzi and Yi [53]. The common methodological bias in the study’s data was therefore not serious.

4.2. Validity Test

First, in terms of reliability, as shown in Table 2, the Cronbach’s alpha coefficient and the combined reliability CR of each variable in the study model were higher than 0.7, which indicates good reliability. Second, in terms of validity, as shown in Table 2, the factor loadings of each variable were greater than the recommended value of 0.5, and the lowest AVE value (average extracted variance) was 0.461 within an acceptable range; the study built a competitive model for validated factor analysis, and, as shown in

Table 3. CFA results.

	MODEL	χ2	df	χ2/df	RMSEA	IFI	TLI	CFI	Model Comparison	Δχ2	Δdf
1	Four-factor model	581.279	383	1.518	0.048	0.960	0.954	0.959
2	Three-factor model	1163.464	402	2.894	0.091	0.845	0.831	0.844	2 vs. 1	582.185	19
3	Two-factor model	1316.745	404	3.259	0.100	0.814	0.798	0.813	3 vs. 1	735.466	21
4	One-factor model	1493.296	405	3.687	0.109	0.778	0.760	0.777	4 vs. 1	912.017	22

Note: Four-factor model (BDC; ERIS; EIIS; SCA), Three-factor model (BDC + ERIS; EIIS; SCA), Two-factor model (BDC + ERIS + EIIS; SCA), One-factor model (BDC + ERIS + EIIS + SCA).

, the four-factor model had the best fit (χ²/df = 1.518 < 3; RMSEA = 0.048 < 0.05; IFI = 0.960 > 0.9; TLI = 0.954 > 0.9; CFI = 0.959 > 0.9), which was significantly better than the other three models, indicating a high degree of discrimination among the variables and better convergent and discriminant validity of the overall scale.

4.3. Descriptive Statistical Analysis

The results presented in

Table 4. Correlation coefficients.

Constructs	1	2	3	4	5	6	7	8
1 Firm age	1
2 Firm size	0.527 **	1
3 Firm type	0.216 **	0.321 **	1
4 Industry type	−0.014	0.107	0.009	1
5 BDC	0.152 *	0.129 +	0.032	0.117 +	1
6 EIIS	0.073	0.099	0.066	0.087	0.562 **	1
7 ERIS	0.048	0.078	0.046	0.055	0.592 **	0.800 **	1
8 SCA	0.081	0.085	0.005	0.108	0.718 **	0.647 **	0.647 **	1
Mean	2.004	2.410	0.179	0.454	5.153	5.407	5.401	5.221
SD	0.944	0.931	0.384	0.499	1.050	0.924	1.012	1.026

Note: SD (standard deviation), ** p < 0.01, * p < 0.05, + p < 0.1.

show that there was a significant correlation between the main variables, providing preliminary support for further regression analysis to obtain more robust analytical results.

4.4. Hypothesis Testing

This study used SPSS 22.0 and its macro program PROCESS to perform hypothesis testing. In order to guarantee the stability and reliability of the regression analysis results, it calculated the variance inflation factor (VIF) of the involved models, and the obtained results show that the VIF values all fall between 0 and 10, which indicates there was no multicollinearity problem between the variables. The following section presents a detailed analysis of the results.

4.4.1. Direct Effect

The direct effect of BDC on SCA is presented in

Table 5. Results of hierarchical regressions.

	EIIS	ERIS	BDAIS	CDAIS	SCA
	M1	M2	M3	M4	M5	M6	M7	M8	M9
Firm age	−0.037	−0.065	−0.017	−0.042	−0.030	0.046	−0.010	0.036	−0.014
Firm size	0.030	0.027	−0.016	0.029	0.011	0.007	−0.001	0.008	−0.001
Firm type	0.046	0.032	0.095	0.047	−0.016	−0.047	−0.032	−0.055	−0.032
Industry type	0.017	−0.019	0.075	−0.018	0.023	0.058	0.022	0.065	0.031
BDC	0.560 ***	0.599 ***	0.293 ***	0.625 ***	0.719 ***		0.482 ***		0.483 ***
EIIS						0.351 ***	0.242 **
ERIS						0.363 ***	0.169 *
BDAIS								0.044	−0.017
CDAIS								0.663 ***	0.386 ***
R2	0.320	0.354	0.104	0.390	0.517	0.472	0.614	0.472	0.607
ΔR2	0.302	0.345	0.082	0.375	0.497	0.452	0.096	0.451	0.090
F	20.965 ***	24.455 ***	5.155 ***	28.514 ***	47.824 ***	33.110 ***	50.155 ***	33.015 ***	48.788 ***
VIF	1.009 ≤ VIF ≤ 3.048

Note: *** p < 0.001, ** p < 0.01, * p < 0.05.

. Model 5 from Table 5 shows the regression model of BDC when there was SCA with a significantly positive coefficient (β = 0.719, p < 0.001). It indicates a positive relationship between BDC and SCA. Therefore, hypothesis 1 was confirmed.

4.4.2. Mediating Effect of Pure Exploitative and Explorative Innovation Strategy

We performed the mediation tests using both regression and bootstrapped mediation analyses. Models 1 and 2 from Table 5 were used to examine the impact of BDCs on EIIS and ERIS, and the coefficients were both found to be significantly positive (β = 0.560, p < 0.001; β = 0.599, p < 0.001), which indicates that BDCs positively impact EIIS and ERIS. Model 6 was used to test the effect of EIIS and ERIS on SCA, and the coefficients were found to be significantly positive (β = 0.351, p < 0.001; β = 0.363, p < 0.001), which indicates that both EIIS and ERIS positively affect SCA; Model 7 is a regression model that adds variables (EIIS and ERIS) on the basis of Model 5, which it was then compared with. The coefficient of BDC decreased by 0.237 and was found to be significantly positive; R² increased by 0.097; and the F-value increased from 47.824 to 50.155. The coefficient of EIIS was found to be significantly positive (β = 0.242, p < 0.01) and the coefficient of ERIS was found to be significantly positive but weaker than EIIS (β = 0.169, p < 0.05).

This study used the macro program PROCESS’s Model 4 and inserted mediating variables (EIIS and ERIS) to test their parallel mediating effects in the relationship between BDC and SCA (the results are shown in

Table 6. Results of mediation analysis.

Path		Index	BootSE	95% Confidence Interval
				LLCI	ULCI
Total effect		0.7031	0.0464	0.6116	0.7945
Direct effect	BDC → SCA	0.4715	0.0524	0.3682	0.5748
Indirect effect	BDC → EIIS → SCA	0.1323	0.0525	0.0362	0.2460
	BDC → ERIS → SCA	0.0992	0.0612	−0.0160	0.2279

). The 95 percent confidence interval of the direct effect of BDC on SCA did not contain 0, which applies to the counterpart confidence interval for the indirect effect of the EIIS’s mediation path. Meanwhile, the 95 percent confidence interval of the ERIS’s mediation path was found to contain 0, which indicates the parallel mediation effects of EIIS and ERIS in the BDC–SCA relationship. The mediating effect of the EIIS partially mediated, and the mediating effect of ERIS was not found to be significant. This finding is consistent with Hypothesis 2 and inconsistent with Hypothesis 3. The former was supported empirically and the latter (the competing hypothesis) was not tested.

4.4.3. Mediating Effect of Ambidextrous Innovation Strategy

Models 3 and 4 from Table 5 were used to examine the impact of BDCs on BDAIS and CDAIS, and the coefficients were significantly positive (β = 0.293, p < 0.001; β = 0.625, p < 0.001), which indicates that BDCs positively impact BDAIS and CDAIS. Model 8 was used to test the effect of BDAIS and CDAIS on SCA, and the coefficient of BDAIS was not significant (β = 0.044, p > 0.1), and the coefficient of CDAIS was significantly positive (β = 0.663, p < 0.001). Model 9 is a regression model that adds variables (BDAIS and CDAIS) on the basis of Model 5, which it was then compared with. The coefficient of BDC decreased by 0.236 and was found to be significantly positive; R² increased by 0.090; and F-value increased from 47.824 to 48.788. The coefficient of BDAIS was found to be not significantly positive (β = −0.017, p > 0.1) and the coefficient of CDAIS was found to be significantly positive and much stronger than that of the balanced dimension (β = 0.386, p < 0.001).

Further, this study used the macro program PROCESS to test the parallel mediating effects of BDAIS and CDAIS between BDC and SCA (the results are shown in

Table 7. Results of mediation analysis.

Path		Index	BootSE	95% Confidence Interval
				LLCI	ULCI
Total effect		0.7031	0.0464	0.6116	0.7945
Direct effect	BDC → SCA	0.4723	0.0541	0.3657	0.5788
Indirect effect	BDC → BDAIS → SCA	−0.0048	0.0199	−0.0492	0.0314
	BDC → CDAIS → SCA	0.2356	0.0606	0.1340	0.3710

). The 95 percent confidence interval of the direct effect of BDC on SCA did not contain 0, which applies to the counterpart confidence interval for the indirect effect of the CDAIS’s mediation path. Meanwhile, the 95 percent confidence interval of the BDAIS’s mediation path was found to contain 0, which indicates the parallel mediation effects of BDAIS and CDAIS in the BDC–SCA relationship. This finding is inconsistent with Hypothesis 4 and consistent with Hypothesis 5. The latter (the competing hypothesis) was supported empirically and the former was not tested.

5. Discussion and Conclusions

There is a widespread consensus within the academic and practical fields that it is important and necessary to use big data management and innovation to improve SCA in a dynamic competitive environment. This paper draws on theories of Dynamic Capability, Innovation Strategy and Organizational Ambidexterity to explore how BDCs impact on SCA. In addition, it also assesses how innovation strategy mediates by reviewing the existing literature.

5.1. Discussion

First, this study’s empirical analysis shows that BDC is a significant antecedent of firms’ SCA. Following the Dynamic Capability Theory, this study used the initial work conducted by Lin and Kunnathur [3] and Ramadan et al. [15] to explore the effects of BDC in determining SCA. The results support Bag et al.’s [54] and Hao et al.’s [10] view (that “enterprises cultivate a high level of BDCs for management support in order to shape sustainable performance”), which broadens the meaning ascribed to the internalization of BDCs into organizations. Additionally, in contrast to previous research that focused on the impact of BDC on organizational performance [4,6], supply chain performance [7,8], organizational innovation [9,10,11], strategic decision-making [12,13] and competitive advantage [14,15,16], this study asserts that SCA is linked to an enterprise’s survival and sustainability, and proposes that it is a critical representational element that will enable businesses to shape their core competences, improve organizational performance and achieve long-term development. The research also adds further exploration of how decision-makers take advantage of data and analytical capabilities to activate and generate competitive advantage [14].

Second, this study examines how innovation strategy mediates the relationship between BDC and SCA from both single and ambidextrous perspectives. In this framework, BDC is theorized to strengthen SCA, directly and indirectly. When perceived from a single perspective, it is apparent that increases in the level of BDCs will make firms more inclined to implement EIIS and construct SCA, as opposed to ERIS. This mediated research builds on Rialti et al. [19] and Shamim et al. [20], which indicate that big data-related capability is an important driver of exploitative and exploratory innovation in firms. It is also in accordance with previous research, which shows that EIIS and ERIS encourage superior innovation speed, innovation quality and performance [31,55]. Contrary to previous studies, this interesting finding shows that the mediation effect of ERIS was not found to be significant. This means ERIS does not only need to rely on the “opportunity insights” brought by BDC to create new innovative changes, but also need to consider other factors, including the rational allocation of innovation resources, the path dependence of the organization’s sustainable development and the importance of balancing the present and long-term.

When viewed from an ambidextrous perspective, CDAIS is found to partially mediate between BDC and SCA; however, the mediating effect of BDAIS between BDC and SCA was not found to be significant. The findings support Luo et al.’s [29] and Soetanto and Jack’s [55] conclusion that “integrated ambidextrous innovation techniques are advantageous for firm performance enhancement”. Contrary to the hypotheses of previous studies, this paper does not find BDAIS significantly mediate the relation between BDC and SCA, which leads us to ask if potential influences affect how firms balance EIIS and ERIS in a big data context when seeking to build a core competitive advantage. According to Sirén et al. [56], companies may end up in an exploitation trap, where high levels of exploitation consume the firm’s limited strategic learning resources, weakening the possibility of explorative innovations arising by neglecting the role of exploration strategy. Firms need to overcome more uncertainties and influence factors to establish a good balance of ambidextrous innovation, which means there may be a longtime lag in the enhancement of competitive advantage. On balance, this strategy may not significantly affect firms’ SCA within a certain time period. It adds further exploration of Revilla et al.’s study [57], which indicates that the “unbalanced” innovation strategy may result in superior outcomes compared with a more “balanced” strategy. In addition, the process mechanism through which BDAIS affects SCA and its role in the relationship between BDC and SCA both need to be further explored.

5.2. Theoretical Contribution

This paper’s contribution mostly relates to two main aspects. The first is enriching the research of BDCs for businesses and using Dynamic Capability Theory to increase awareness of the mechanism through which BDCs influence SCA. Researchers typically regard BDC as a resource and investigate its impact on organizational performance by using Resource-Based Theory [1,4], which means the dynamic and systematic consideration of organizational capability has not yet been investigated [3]. This finding will help academics and practitioners to conceive and understand BDC as a contextual factor and will also provide insights into the endogenous factors of enterprises’ competitive advantage.

Second, this article incorporates both single and ambidextrous viewpoints to investigate how the mechanism of innovation strategy mediates between BDC and SCA. As suggested by previous studies, the effect of big data-related capability on sustainable competitive advantage could be mediated by ambidextrous innovation [19,20]. However, there is still an incomplete understanding about the relationships among BDC, EIIS, ERIS and SCA. This is the question we explored in this study. The finding (that “as the level of BDCs increases, firms are more inclined to implement EIIS to build SCA”) provides two important strategic perspectives: (1) “In the new context of big data, the advantage of big data-based capabilities enhances firms’ adaptive improvement of internal innovation practices and insight into the external dynamic competitive environment”; (2) “As the level of BDCs increases, firms are more inclined to implement EIIS to build SCA.”

Following this, consider the conclusion that “as the level of BDCs increases, companies tend to implement CDAIS to build SCA “. In contrast to previous research that focused on a single aspect of singularity or ambidexterity, this paper investigates the mediating role of EIIS and ERIS, with specific reference to their ambidexterity (balanced and combined dimensions), role in the transformation mechanism that turns BDC into SCA and the value transformation mechanism of BDC that operates through innovation strategy. This helpfully clarifies that, in the new big data context, EIIS and ERIS mediate in the causal chain of BDC that acts on SCA; they also provide SCA in the form of strategic synergy and answer the request for “business practice to pay adequate attention to the synergy between exploitative innovation and exploration innovation. This will help to avoid the undesired scenario of over-exploration or exploitation”. It expands the theoretical study of strategic synergy between enterprise exploitative and exploration innovation by clarifying the mediation mechanism that enables the value transformation of BDC through innovation strategy; in addition, it also provides theoretical and empirical support to the exploration of ambidexterity in a big data context.

5.3. Managerial Implications

The paper’s main findings provide new insight into how businesses can use BDCs to trigger innovation strategies, apply their ambidexterity to gain SCA and implement big data practice that help to overcome the “Big Data productivity paradox.” This framework has important practical implications.

First, our analysis shows that BDC can make a positive contribution to SCA. This finding improves managers’ understanding of the impact big data and data capabilities have on the sustainability of competitive advantage. Businesses must therefore consider how to rethink management in the digital era. A data swamp can easily emerge within an organization if they do not make proper use of the massive amounts of data they collect. In order to provide effective management support to big data practice and build SCA, they should concentrate on developing BDCs, acquiring accurate and efficient internal and external data resources and storing and processing data for analysis; they should then use big data algorithms to analyze the integrated data, transform hidden value information into knowledge and improve data re-use value. These steps will increase business revenue, profit, market share and market area expansion, enhance product launch degree and contribute in other aspects. With regard to market region expansion, market share, product introduction, profit and revenue, the data may help to provide better indicators of SCA.

Second, an innovation strategy is essential for enhancing competitive advantage and achieving long-term success. This finding of the research suggests that when the level of BDCs is higher, a firm will be more inclined to build SCA by implementing EIIS. EIIS adapts innovation practice on the basis of existing knowledge and learning about how to respond to user needs [25,26]; while ERIS breaks down old technological barriers to gain a competitive advantage by experimenting with new knowledge and fields [25,27]. In initially engaging with the strategy formulation stage, businesses should focus on the logic of exploitation and exploration in the development of innovation practice activities. BDCs can be used to gain a better understanding of the current state of business innovation and potential development space and can also track cutting-edge technology development trends and real-time competitive situations, which will contribute to the scientific and effective formulation of innovation strategies. Thus, managers can make appropriate strategic decision-making by developing BDC from this perspective.

Moreover, the accumulation and maintenance of an enterprise’s competitive advantage should not only focus on the present, and more specifically on the short-term improvement in performance created by an innovation strategy. This means that businesses should, in internalizing BDCs into SCA, focus on the utility of both EIIS and ERIS and, in so doing, place particular emphasis on their synergistic interaction. Our research further offers some interesting insights that CDAIS is found to partially mediate between BDC and SCA while BDAIS does not. This finding can be used to enhance innovation strategy formulation from the viewpoint of ambidexterity in the big data era. When an enterprise’s EIIS is slow, explorative innovation can be used to explore exploitative innovation’s improvement space. Enterprises should strengthen innovation exploration, constantly inject new innovative blood into the enterprise and shape the leading advantage of competitive/potentially competitive enterprises in ways that are difficult to imitate. When ERIS produces little performance output, its innovation energy can be sparked by increasing investment and support for EIIS in a way that builds core competitiveness and contributes SCA. These findings not only help in overcoming the challenges of exploitative or explorative innovation strategy formulation, but also in achieving BDAIS and CDAIS through BDC. Managers can consider factors of BDC for the further implementation of innovation strategy.

5.4. Limitations and Future Directions

Although the paper’s research design was scientific and systematic and made it possible to obtain some useful results, there are a number of ways in which it can be improved. First, this study only looks at firms based in a single country; regional sampling could limit the generalizability of the results obtained. Future research could seek to address this. Second, our collecting data is self-reported. Despite considerable efforts being undertaken to confirm data quality, the potential of biases cannot be excluded. Future research can collect more large-scale and more types of sample data to improve the quality and reliability of the research. Third, this study examined the mediating role of innovation strategy and its ambidexterity in the mechanism that influences the impact of BDCs on SCA. However, in doing so it only revealed one feasible “black box” mechanism. Additional research will hopefully help to identify other mechanisms.