Pull and Push Factors of Data Analytics Adoption and Its Mediating Role on Operational Performance

Abstract

Despite the influx of data analytics (DA) practices among firms, their impact on operational performance remains ambiguous. This study examined the pull and push factors affecting the data analytics adoption (DAA) from the theoretical perspectives of the technology–organization–environment (TOE) model, theory of perceived risk (TPR), and resource-based view (RBV). The study analyzed data from 169 firms on the basis of the positivist paradigm and employed the partial least square to run the reflective–formative two-stage analysis. Accordingly, the results indicated that the three TOE model aspects exhibited a positive direct impact on DAA and indirectly impacted operational performance through DAA. However, the perceived risk did not display a similar effect in both situations. This study further revealed that the environment push factor had more explanatory power than the perceived risk pull factor, suggesting that a conducive TOE environment would motivate DAA, subsequently enhancing operational performance. The study provided valuable empirical evidence on the determinants of DAA and its subsequent effect on firms’ operational performance. Uniquely, the study also contributed to the literature from the perspective of higher-order-construct analysis in examining the determinants of DAA and its effect on operational performance. Furthermore, the mediation analysis covered the interaction of indirect-path coefficients, minimizing errors in interpreting the mediation effect.

1. Introduction

The era of knowledge-driven economies has altered how firms compete [1,2], prompting them to utilize data analytics (hereafter DA) as a ‘game-changer’ to improve performance [3]. Data analytics serves as a tool for firms to transform data into meaningful information and subsequently make an informed decision. Firms that successfully integrate DA will reap results through improved predictive capabilities and enhancing operational performance [4,5]. Without DA, it is challenging for managers to learn from the past or future. In essence, DA improves the assessment of business parameters in dealing with market dynamics.

Previous studies focused on the potential of DA in various aspects, such as improving financial performance [6], marketing efficiency [7], and the decision-making process [8]. Despite DA being frequently associated with improved performance, contradictory evidence is found [6,9,10]. These contradictions may cast doubt on the benefits of DA and cloud the understanding of DA’s benefits. Such misunderstandings, especially among smaller firms where resources are often limited, would hinder them from embarking on the process of DA adoption (hereafter DAA). Furthermore, the association between DAA and operational performance remained uncharted in the past. Hence, this study enriches the body of knowledge through simultaneous analysis of the pull factors (determinants that attract) and push factors (determinants that repel) acting on DAA and examines the mediating role of DAA on operational performance using firm-level data of a developing country [11]. Accordingly, the study aims: (1) to examine if the pull factors of the TOE (technological, organizational, and environmental context) and the push factor (perceived risk) of the theory of perceived risk (TPR) can be associated with DAA, (2) to determine if DAA is related to firms’ operational performance, and (3) to investigate the central role of DAA in the relationship between the pull/push factors and firms’ operational performance.

2. Literature Review

The cornerstone of strategic management involves creating and sustaining business values and performance [9]. The resource-based view (RBV) theory suggests that the firms’ performances are influenced by the capabilities to exploit strategic resources [12]—tangible resources, intangible resources, and organizational capabilities [13]. DAA is essential for business development [14], enabling firms to increase their competitive advantages [15,16,17]. Thus, it is justifiable to examine the impact of DAA through the lens of the RBV.

Essentially, data are meaningless if not being converted into information and, subsequently, knowledge. Thus, firms embracing DA can acquire critical knowledge that leads to informed business decision making. In other words, they possess better insight into dealing with issues and challenges from the knowledge gained, which provides them with leverage in competitive advantages. However, for most firms, setting up a DA platform can consume substantial resources. Rationally, firms would usually engage in such endeavors if the expected improved performance or return exceeds the anticipated risks and costs.

Firms’ resources and capabilities are two vital RBV elements that received significant attention in empirical studies, especially in technological innovation [9,18]. Previous literature reported the direct and indirect influence of technological capabilities on firms’ performance [19,20]. Similarly, recent findings suggested that Information Technology (IT) investment [21,22] enables firms to stay competitive [23]. Even though these works were grounded on the RBV, they focused exclusively on financial or marketing performance. Because of the insufficient evidence from the perspective of the firm’s operational performance, which deserves equal attention as financial and marketing performances, this study examined the impact of DAA on operational performance.

The operational performance is measurable from several viewpoints: managing demand, supply chain, customer relationships, and new product development [24]. Demand forecast and supply chain are frequently combined in operations management [25,26] related to DA [27]. This idea is based on the viewpoint that operations and the supply chain are intrinsically linked [28]. Thus, the current study measured operational performance in terms of demand forecast and supply chain. Chae et al. [29] explained that efficiency in data handling and analysis would support management with relevant and quality information, and these aspects are deemed vital to enhance a firm’s planning, control, and overall operational performance. The study employed a higher construct for operational performance to ensure a parsimonious and interpretable model [30].

Previously, pull and push factors that may implicate DAA were often investigated separately. Specifically, there are more studies on pull factors than push factors. The disunification of these factors in previous research somewhat impedes understanding the firm’s technology adoption. Simultaneously, the motivating and hindering factors are affecting firms’ DAA. Hence, this study contributes to the literature by merging the two groups of regressors in a research model.

The technology–organization–environment (TOE) framework posited that the firms’ technology adoption, DAA included, is related to external and internal factors [31]. The first context of the TOE encompasses various technologies relevant to firms, including those in the marketplace, regardless of their implementation. Accordingly, three essential factors consistently employed to measure technological aspects are compatibility, complexity, and trialability [31,32]. The second TOE context is organizational, which focuses on the firms’ characteristics and resources. Measures used in this context include top management support, organizational readiness, communication process, and organizational structure. The linkage mechanism within firms affects innovation adoption, and the process that connects these subunits tend to promote this adoption further. Notably, other primary factors for innovation adoption include top management support and organizational readiness [6]. The final TOE context is environmental, which refers to the business operating environment directly impacting innovation adoption. This context includes external aspects related to industry structure, the presence of technology providers, and the regulatory system. Observing recent trends [2,6,33], it is opined that competitive pressure and external support are relevant for the current study.

While past studies on technology adoption mainly include the pull factor from the TOE framework, the push factor perspective, on the other hand, can be examined in Bauer’s (1960) Theory of Perceived Risk (TPR). The TPR includes a two-dimensional construct of uncertainties and negative consequences. These constructs explain why the decision to act is affected by one’s subjective judgments about risk, a critical concern that firms must address. Accordingly, the greater the decision-makers-perceived adverse possibilities related to DAA, the lower the tolerance to undertake these initiatives. Thus, any rational decision makers would be concerned to consider investing in DA and evaluate their readiness to face challenges arising from at least three fronts: erroneous, legal, and ethical. The TPR is closely related to the concept of partial ignorance, in which neither the negative consequences nor the probability of incidence is precisely known. Other than the two dimensions, TPR’s flexibility enables researchers to include the types of uncertainties according to the research context [34,35]. Hence, the TPR is widely used in various research areas, and scholars have proposed numerous types of risks, including psychological and social, and related to performance and resources. The current study includes several dimensions under perceived risk: privacy and security, data quality, and resource risk.

Substantial existing evidence was determined using the repeated-indicators approach, where the manifest items of the lower-order constructs are used again for the higher-order constructs. As scholars recommend higher-order constructs examination, the study was motivated to investigate DAA among firms using a two-stage approach. There are two statistical stages under this approach where the lower and higher stages are separated. The former examines the relationship between deals constructs and indicators, while the latter defines the relationship among various constructs. In other words, hypothesis testing will be based on a higher-level construct. There are at least three advantages that arise from the reduction in measurement in lower-order constructs. Firstly, higher-order construct utilization minimizes multicollinearity among indicators [36]. Secondly, the approach prevents indicators from double counting [37]. Thirdly, it allows the higher-order construct to be placed at the endogenous position within a research framework [38]. Overall, instead of specifying relationships between multiple independent and dependent constructs in a path model, higher-order constructs help reduce the number of path model relationships, thereby achieving model parsimony.

3. Research Framework and Hypotheses Development

The study merges three theoretical perspectives: the TOE framework, the TPR, and the RBV. The TOE framework provides illumination of the potential pull factors of DAA and operational performance. Meanwhile, the TPR grounds the inclusion of the push factor by examining the correlation of perceived risk, DAA, and operational performance. The RBV provides support to investigate the association between the DAA and the firm’s operational performance. Though literature shows that DAA mediates between TOE contexts and firms’ performance, these studies did not specify operational performance. Thus, it is crucial to determine if DAA would display a persistent mediating role under this context. The research framework showing the nine hypotheses of the study is given in Figure 1.

3.1. Technological Context

Technological context comprises various technologies relevant to a firm, including those not implemented by the firm but available in the marketplace [32]. These technologies can significantly impact their future adoption process [39] and will set the pace and capacity to deal with technological change. For most firms, there are three primary types of technological changes related to innovation adoption: (1) incremental (minimal risk and changes due to new features of existing technology), (2) synthetic (moderate change due to existing technologies being integrated in a novel way), and (3) discontinuous (massive change due to paradigm shift) [32]. Therefore, firms must consider the potential technological changes resulting from innovation adoption.

Technological context can be measured based on several aspects such as compatibility, complexity, and trialability [32]. While compatibility and complexity are said to be the most consistent determinants [6], trialability is recognized as a crucial determinant, especially for early adopters [40]. Hence, the study examined this context from the angle of compatibility, complexity, and trialability. Compatibility refers to the degree of consistency related to the firm’s current condition [41], which is measurable upon the firm’s innovation adoption. Innovation provides tremendous success when the firm can adopt them seamlessly. The firm will exploit competitive advantage from the innovation, given that the adoption is seamless and appropriate to its values and practices. However, incompatible innovation could be costly and prohibit them from gaining the advantage.

Complexity is the difficulty level in utilizing innovation [42], and complex innovation can cause integration difficulty that potentially makes it incompatible with the existing system [18,43]. Even if innovation can be easily applied, at times, it may require a different approach from the existing practice. In other words, a simple innovation does not necessarily imply higher compatibility. Hence, it is fair to argue that complex innovation hinders firms from innovation adoption [44]. Accordingly, compatibility can only be confirmed at the postadoption phase, and complexity may discourage a firm from innovation adoption. Meanwhile, the adoption of data analytics would be favorable if firms had the opportunity to pretest the technology innovation without any significant commitment [6]. This reflects the degree to which an innovation can be put on trial, and it is described as trialability [45]. Trialability encourages the adoption of technology innovation as it increases a firm’s innovation exposure and reduces its level of insecurity. Hence, H1 postulates:

H1. 

Technological context has significant influences on DAA.

3.2. Organizational Context

The firm’s characteristics and resources describe the organizational context, which influences the innovation adoption propensity through formal or informal Intra organizational mechanisms for communication and control [46]. Innovation adoption is closely related to the firm’s internal connection, where the mechanism that links the internal subunits promotes this adoption [32]. There are multiple studies on the factors under this context, such as top management support, organizational readiness, communication process, and structure type [32,47]. Among them, top management support and organizational readiness are the most critical in innovation adoption [6]. Hence, top management support and organizational readiness are examined in this study.

Top management is usually referred to as senior executives or individuals who hold most responsibilities in the decision-making process [48], while support refers to physical and mental assistance. Senior executives who are optimistic about the potential advantages of sustainable innovation adoption are more likely to take action to implement them [49]. Top management’s support (e.g., financial support) can effectively build positive values, culture, and environment towards this adoption [42,50]. However, the extent of support is also highly dependent on resource availability [51]. This term is known as organizational readiness [51], or the firm’s capabilities in investing and managing an innovation adoption [52]. The resources cover various types of the firm’s assets, including the skills and talents of employees. If organizational assets are inadequate, it will be challenging for the firm to properly utilize and maintain innovation [32,41]. Therefore, H2 suggests:

H2. 

Organizational context has significant influences on DAA.

3.3. Environmental Context

In the TOE model, the environmental context refers to the external business environment that directly impacts innovation adoption. Previous studies have included various aspects of environmental context. Factors that have been linked to impacting business activities include government intervention, policies, legislation system, geographic location, cultural, and consumer behavior. In the context of technology adoption, researchers have regarded competitive pressure and external support competitive pressure as preeminent antecedents of innovation adoption [53,54,55,56].

Competitive pressure is the singular antecedent of innovation adoption by an organization [53]. Early adopters of DA can strategically increase competitive pressure on their peers, accelerating the innovation adoption process [57]. Hence, firms that realize their competitors are adopting DA may feel threatened, especially when the competitiveness gap between them becomes widened.

Meanwhile, scholars explain that external support can be personal development and technical assistance provided by external vendors or third parties [2,6]. External support has been proven critical toward innovation success [55]. Therefore, firms that garner more support from external parties would be more likely to adopt DA [6]. In other words, the degree of external support can impact DAA. Nevertheless, the firm’s willingness to adopt DA depends on its perception of the vendor [6], meaning there is no specific positive or negative relationship between external support and DAA. Thus, H3 proposes:

H3. 

Environmental context has significant influences on DAA.

3.4. Perceived Risk

Perceived risk refers to the uncertainties and unfavorable consequences associated with one’s expectations [58]. Its level depends on the subjective uncertainty degree of outcomes [59]. Companies that decide on adopting and investing in innovation may sense higher risk if they find any dispute between goals and experiences. Risks that are related to DAA, including privacy and security, data quality, and resource risk, are the risk investigated from the perspective of firms.

Privacy and security become an issue when the data storage is outsourced to an external party [2]. The main concern of data outsourcing is to safeguard the security of personally identifiable data and customer information. New adopters often encounter this issue, especially when they do not have sufficient capability and capacity to sustain the DA system. Consequently, they may feel insecure because of a lack of complete control over the data, which subsequently leads to concern over data quality. Firms that lack experience, tools, and talent to screen and analyze data may feel less certain that their data are credible, complete, accurate, relevant, and timely [60]. In other words, the perception of DA risk would influence DAA. Therefore, H4 is:

H4. 

Perceived risk has significant influences on DAA.

3.5. The Firm’s Data Analytics Adoption and Operational Performance

The current study examines the relationship between the DAA and the operational performance based on RBV. The firm’s operational performance is measured using two constructs: demand forecast and supply chain. Structured demand forecasting and practices in supply chain management improve the firm’s operational performance [61,62]. Specifically, DA contributes to a firm’s values as DA potentially enhances efficiency, flexibility, and cost reduction [9,63]. It transforms conventional operation practices and makes a paradigm shift to data-driven decision making [64]. In other words, firms that practice DA can efficiently transform data into information that is valuable for strategic decision making. Hence, improvement in demand forecasting and supply chain management would lead to alleviation of operational performance [42]. Accordingly, H5 proposes:

H5. 

DAA has a significant influence on operational performance.

3.6. Mediating Role of Data Analytics

One of the key successes of innovation adoption depends very much on firms’ capabilities to effectively assimilate the adoption into their business process. A precondition for successful innovation adoption includes the technological, organizational, and environmental factors [65], enabling firms to achieve effective business performance [18]. In the past, scholars claimed that DAA mediates the effect between sustainable business performance and other factors not included in the TOE model [6,66]. Drawing from this concept, the TOE model may potentially serve as the factor of DAA, and DAA would, in turn, enhances firms’ performances. In other words, this framework can be extended to examine the factors of DAA, which will subsequently improve the firm’s performance.

Meanwhile, a study by Khayer et al. [67] showed that perceived risk has a negative impact on innovation adoption. The reason for the reverse relationship is the uncertain outcomes of the adoption, which causes concern and unease among managers at firms, ultimately hindering the adoption of innovation. Although Khayer et al.’s study was based on cloud adoption, a similar scenario may also be observed in the context of DAA. Hence, perceived risk may negatively impact the adoption of data analytics, which subsequently impedes the firms’ performance, including operational performance [68].

Another study found that DA fully mediates the relationship between environmental practice and supply chain performance in the manufacturing sector but only partially mediates the linkage between supply chain performance and total quality management [69]. Additionally, a study that investigated the management control system of family SMEs and technological innovation found that technological adoption exhibited mediating effects between family management and firm performance [70].

This study differs from the previous by examining the DAA’s determinants and the consequence of DAA on operational performance. Hence, four hypotheses in which DAA is a mediator are proposed:

H6. 

DAA mediates the relationship betweentechnological context and operational performance.

H7. 

DAA mediates the relationship between organizational context and operational performance.

H8. 

DAA mediates the relationship between environmental context and operational performance.

H9. 

DAA mediates the relationship between perceived risk and operational performance.

Investigation of the literature also revealed that the interpretation of indirect effect mostly ignored the path coefficients of a, b, and c for a simple mediation model, which may have led to erroneous conclusions [71]. Hence, this study extends the mediation analysis by examining the impact of indirect (a and b) coefficients (i.e., the indirect effect) on the total mediation effect, which has been largely ignored in previous works.

4. Methodology

An online survey (refer

Table A1. Research Instrument for the Study (After Pre and Pilot Tests).

Symbol	Question
TCPT1	The adoption of data analytics is consistent with our business practices.
TCPT2	The adoption of data analytics fits our organizational culture.
TCPT3	It is easy to synchronize external data with the internal data of our firm.
TCPT4	Data analytics adoption provides a flexible architecture that is compatible with numerous analytical tasks.
TCPT5	Overall, it is uncomplicated to incorporate data analytics into our organization.
TCPX1	Our firm manages to operate data analytics without difficulty.
TCPX2	Our firm manages to maintain data analytics without difficulty.
TCPX3	Our firm manages to extract useful information without difficulty.
TCPX4	Our firm manages to normalize a huge amount of unstructured data from multiple sources without difficulty.
TCPX5	Our employees manage to learn data analytics without difficulty.
TTRL1	It is easy for our firm to quit after testing a data analytics package.
TTRL2	The start-up cost for our firm to use data analytics is low.
TTRL3	Our firm has the opportunity to try out different types of data analytics software packages.
TTRL4	Our firm has the opportunity to try out allowed data analytics until its capabilities are revealed.
TTRL5	Our firm has the opportunity to have a free trial before deciding adoption of data analytics.
OTMS1	Our top management promotes the use of data analytics in the organization.
OTMS2	Our top management creates support for data analytics initiatives within the organization.
OTMS3	Our top management promotes data analytics as a strategic priority within the organization.
OTMS4	Our top management is interested in the news about using data analytics adoption.
OTMS5	Our top management provides the necessary resources to support the use of data analytics.
OORD1	There are adequate financial resources for our firm to adopt data analytics.
OORD2	There are adequate IT infrastructures for our firm to adopt data analytics.
OORD3	There is adequate analytics capability for our firm to exploit data analytics.
OORD4	There are adequate employee skills to conduct data analytics in our firm.
OORD5	There is adequate training provided to the analyst about data analytics and its applications.
ECPR1	Our choice to adopt data analytics would be strongly influenced by what competitors in the industry are doing.
ECPR2	Our firm is under pressure from competitors to adopt data analytics.
ECPR3	Our firm would adopt data analytics in response to what competitors are doing.
ECPR4	There is a trend in the industry to enhance the utilization of data analytics for business-related activities and decision making.
ECPR5	Our firm could lose customers to the competitors if we do not adopt new technologies.
EESP1	Community agencies who provide training encourage our firm to adopt data analytics.
EESP2	Community agencies who provide effective technical support encourage our firm to adopt data analytics.
EESP3	Community agencies who provide valuable assistance at the postimplementation stage encourage our firm to adopt data analytics.
EESP4	The governmental policies encourage our firm to adopt data analytics.
EESP5	The government provides incentives if our firm adopts data analytics.
RPNS1	Our firm has the ability to secure data.
RPNS2	Our firm has complied with Internet security.
RPNS3	Our firm has given top priority to the privacy of customer data.
RPNS4	The need to outsource data analytics creates concerns about data security and privacy.
RPNS5	The need to outsource data analytics creates risks through excessive dependency on the vendor.
RDQT1	There are risks where data used for analytics are incomplete.
RDQT2	There are risks where data used for analytics are inconsistence.
RDQT3	There are risks where data used for analytics are inaccurate.
RDQT4	All data (ordering details, material inventory, etc.) are managed the same way throughout our firm.
RDQT5	A common definition of the main data source (order forms, component information, etc.) is being applied in our firm.
RRSR1	There are uncertainties about the costs involved in data analytics adoption.
RRSR2	There are uncertainties about the potential benefits of data analytics adoption.
RRSR3	There are uncertainties about the employee to adapt to changes.
RRSR4	The use of data analytics creates concerns about technological risks.
RRSR5	The use of data analytics creates capital outlay with no guarantee of financial returns.
ASTR1	The adoption of data analytics helps our firm gain competitive advantages.
ASTR2	The adoption of data analytics helps our firm improve customer relations.
ASTR3	The adoption of data analytics helps our firm establish useful links with other organizations.
ASTR4	The adoption of data analytics helps our firm respond more swiftly to changes.
ASTR5	The adoption of data analytics helps our firm provide better products or services to customers.
ATSC1	The adoption of data analytics helps our firm increase savings in supply chain management.
ATSC2	The adoption of data analytics helps our firm reduce operating costs.
ATSC3	The adoption of data analytics helps our firm reduce communication costs.
ATSC4	The adoption of data analytics helps our firm enhance employee productivity.
ATSC5	The adoption of data analytics helps our firm increase its return on financial assets.
ATFM1	The adoption of data analytics helps our firm improve employees’ skill levels.
ATFM2	The adoption of data analytics helps our firm develop new business opportunities.
ATFM3	The adoption of data analytics helps our firm expand its capabilities.
ATFM4	The adoption of data analytics helps our firm improve organizational structure and processes.
ATFM5	The adoption of data analytics helps our firm improve its business models.
AIFM1	The adoption of data analytics enables our firm to access data faster.
AIFM2	The adoption of data analytics enables our firm to access data easier.
AIFM3	The adoption of data analytics enables our firm to improve data management.
AIFM4	The adoption of data analytics enables our firm to improve data accuracy.
AIFM5	The adoption of data analytics enables our firm to generate data in more useable formats.
PDMF1	The adoption of data analytics enables our firm to perform demand forecasts accurately.
PDMF2	The adoption of data analytics enables our firm to access demand forecasts conveniently.
PDMF3	The adoption of data analytics enable our firm to have reliable demand forecast.
PDMF4	The adoption of data analytics enables our firm to increase the number of goods delivered on time.
PDMF5	The adoption of data analytics enables our firm to decrease in inventory levels.
PSPC1	The adoption of data analytics enables our firm to reduce order fulfilment lead times.
PSPC2	The adoption of data analytics enables our firm to reduce order-to-delivery cycle time.
PSPC3	The adoption of data analytics enables our firm to enhance customer response time.
PSPC4	The adoption of data analytics enables our firm to enhance the ability to respond to customer/supplier query time.
PSPC5	The adoption of data analytics enables our firm to enhance the ability to help customers by providing the services needed.

) was employed after the research instrument had been pretested (for content clarity and validity) and piloted (ensuring internal consistency). Respondents were selected according to judgmental sampling, involving both local and multinational firms from the services and manufacturing sectors utilizing DA. As the study was based on firm-level data, it was vital to ensure respondents of the study were firm owners, chief executives, senior managers, executives, or DA authorized personnel. Accordingly, 169 responses were collected and entered for analysis. It fulfilled the minimum sample required at 129 calculated responses based on the G*Power criterion, and it exceeded the suggested minimum of 100 for path analysis [72].

Data were analyzed using the SmartPLS-SEM that can deal with non-normal distribution and a relatively small sample between 100 to 200 [73,74]. Notably, this variance-based SEM approach can verify a hierarchical model by considering relevant solutions with its flexible assumptions feature [75]. Subsequently, Harman’s Single Factor test was performed to ensure there were no detrimental effects caused by the common method bias [76].

5. Data Analysis and Findings

The data in

Table 1. Description of Respondents (n = 169).

Demographic Profile	Frequency	Percentage
Job Position:
CEO/Owner	25	14.79
Senior Manager	5	2.96
Executive	80	47.34
Authorized Personnel	59	34.91
Number of Employees:
5–29	17	10.06
30–75	68	40.24
>75	84	49.70
Business Sector:
Service Sector	74	43.79
Of which:
Business Services	7	9.46
Constructions and Related Engineering	18	24.32
Education	20	27.03
Health and Related Social Services	18	24.32
Tourism and Related Travel	11	14.86
Manufacturing Sector	95	56.21
Of which:
Automotive Industry	11	11.58
Electrical and Electronics Industry	32	33.68
Machinery and Equipment Industry	19	20.00
Iron Steel Industry	11	11.58
Food Industry	10	10.53
Pharmaceutical Industry	12	12.63

show there is a good proportion in terms of representation from the service (43.70%) and manufacturing (56.21%) sectors. As intended, all the data for the research were obtained from relevant personnel (CEO/Owner, Senior Manager, Executive, and other authorized personnel) with authority for valid feedback about DAA.

5.1. Measurement Model Assessment

The measurement model is divided into two parts. First is the assessment of the reflective measurement model, which is then followed by the formative measurement model.

The reflective measurement model assessment is to examine the internal consistency reliability, indicator reliability, convergent validity, and discriminant validity.

Table 2. Factor Loadings (Explanation for abbreviation provided in Abbreviations part).

Constructs	Items	Initial Loading	Modified Loading	CR	AVE	VIF
Compatibility	TCPT1	0.859	0.865	0.877	0.644	1.126
	TCPT2	0.761	0.766
	TCPT3	0.861	0.876
	TCPT4	0.659	0.688
	TCPT5	0.195	Deleted b
Complexity	TCPX1	−0.146	Deleted a	0.873	0.776	1.022
	TCPX2	0.644	Deleted c
	TCPX3	0.591	0.939
	TCPX4	0.565	0.819
	TCPX5	0.241	Deleted b
Trialability	TTRL1	0.657	0.657	0.840	0.514	1.104
	TTRL2	0.678	0.678
	TTRL3	0.712	0.713
	TTRL4	0.807	0.808
	TTRL5	0.719	0.719
Top Management	OTMS1	0.724	0.725	0.882	0.600	1.144
	OTMS2	0.760	0.760
Support	OTMS3	0.687	0.687
	OTMS4	0.832	0.832
	OTMS5	0.857	0.857
Organizational Readiness	OORD1	0.374	Deleted b	0.935	0.782	1.144
	OORD2	0.877	0.879
	OORD3	0.869	0.877
	OORD4	0.880	0.895
	OORD5	0.878	0.885
Competitive	ECPR1	0.734	0.734	0.874	0.582	1.195
Pressure	ECPR2	0.784	0.784
	ECPR3	0.702	0.703
	ECPR4	0.820	0.820
	ECPR5	0.769	0.769
External Support	EESP1	0.571	0.572	0.867	0.570	1.195
	EESP2	0.782	0.782
	EESP3	0.824	0.824
	EESP4	0.816	0.816
	EESP5	0.752	0.751
Privacy and	RPNS1	0.701	0.697	0.832	0.502	1.037
Security	RPNS2	0.738	0.740
	RPNS3	0.822	0.822
	RPNS4	0.560	0.560
	RPNS5	0.695	0.696
Data Quality	RDQT1	0.826	0.826	0.887	0.612	1.036
	RDQT2	0.817	0.816
	RDQT3	0.747	0.748
	RDQT4	0.753	0.754
	RDQT5	0.765	0.765
Resource Risk	RRSR1	0.393	Deleted b	0.858	0.603	1.010
	RRSR2	0.762	0.830
	RRSR3	0.736	0.802
	RRSR4	0.572	0.675
	RRSR5	0.820	0.789
Strategic Value	ASTR1	0.643	0.641	0.848	0.532	1.164
	ASTR2	0.761	0.758
	ASTR3	0.820	0.823
	ASTR4	0.815	0.817
	ASTR5	0.576	0.575
Transactional	ATSC1	0.874	0.877	0.944	0.808	2.099
Value	ATSC2	0.936	0.936
	ATSC3	0.898	0.900
	ATSC4	0.879	0.881
	ATSC5	−0.111	Deleted a
Transformational	ATFM1	0.844	0.851	0.918	0.738	1.982
Value	ATFM2	0.849	0.845
	ATFM3	0.892	0.894
	ATFM4	0.840	0.845
	ATFM5	−0.155	Deleted a
Informational	AIFM1	0.815	0.815	0.917	0.690	1.832
Value	AIFM2	0.893	0.891
	AIFM3	0.890	0.889
	AIFM4	0.709	0.711
	AIFM5	0.832	0.833
Demand	PDMF1	0.832	0.831	0.924	0.708	1.557
Forecast	PDMF2	0.842	0.842
	PDMF3	0.849	0.849
	PDMF4	0.873	0.874
	PDMF5	0.811	0.812
Supply Chain	PSPC1	0.829	0.828	0.908	0.664	1.557
	PSPC2	0.780	0.780
	PSPC3	0.859	0.859
	PSPC4	0.783	0.784
	PSPC5	0.821	0.821

Deleted ^a: Item removed due to negative loadings. Deleted ^b: Item removed due to loading less than 0.40. Deleted ^c: Item removed due to AVE less than 0.50.

shows the composite reliability of constructs ranging between 0.832 and 0.944, suggesting good internal consistency reliability. The AVEs of constructs are between 0.502 to 0.808, which are all exceeding the 0.50 threshold, thus, signifying adequate convergent validity [77]. The discriminant validity assessment showed that the constructs are free from discriminant validity issues. In other words, the constructs do not overlap with one another [78] and are measuring what they are designed to measure. Three approaches were employed to check discriminant validity, namely cross-loading, the Fornell and Larcker Criterion, and the Heterotrait-Monotrait (HTMT) ratio of correlations [79]. All results confirmed that the discriminant validity of the constructs is sufficient (Full results for discriminant validity are not included in the current paper).

The study subsequently assessed a formative measurement model for the higher-order construct. Specifically, this procedure assessed the convergent validity, indicator collinearity, indicator significance, and model relevance.

Table 3. Paths Coefficient and R².

Construct	Path Coefficient	R2
Technological Context	0.725	0.526
Organizational Context	0.838	0.702
Environmental Context	0.809	0.655
Perceived Risk	0.831	0.691
DAA	0.906	0.821
Operational Performance	0.880	0.775

illustrates the redundancy analysis, revealing that the constructs’ path coefficients exceed 0.70 [80] and the R² values of endogenous constructs are higher than 0.50 [81]. These results provide the confirmation that the formative constructs successfully meet the convergent validity evaluation, meaning that the constructs are, in fact, measuring what they are theoretically expected to measure. Meanwhile, high collinearity in the measurement models may cause severe issues for the model’s estimation because of the standard error increment within the model. Evidently, in Table 2, the model’s variance inflation factors (VIFs) range between 1.010 and 2.099, which are less than 3.3, signifying that the constructs were not correlated. Thus, this confirms that the model does not exhibit multicollinearity issues [82].

The outer weights can be estimated by measuring the multiple regression between indicators and latent variable scores [81]. The indicators’ relative contribution to the construct and the outer loadings were assessed with t-values. The outer weights should not be outright treated as a poor measurement if their indicator is insignificant. Instead, further assessment based on the outer loading can be assessed.

Table 4. Decision in Retaining or Removing Formative Indicators.

Construct	Items	Outer Weight t-Value (A)	Outer Loading	Outer Loading t-Value (B)	Decision
Technological Context	Compatibility	3.819	0.654	6.133
	Complexity	1.087	−0.065	0.449	Deleted
	Trialability	8.312	0.910	17.095
Organizational Context	Top Management Support	10.962	0.896	21.882
	Organizational Readiness	5.714	0.733	11.335
Environmental Context	Competitive Pressure	4.151	0.760	9.153
	External Support	7.066	0.902	16.854
Perceived Risk	Privacy and Security	4.494	0.727	6.225
	Data Quality	5.037	0.753	6.191
	Resource Risk	1.929	0.259	1.702	Deleted
DAA	Strategic Value	8.911	0.912	21.508
	Transactional Value	0.546	0.483	4.198	Retained
	Transformational Value	2.285	0.628	6.496
	Informational Value	2.250	0.648	6.115
Operational Performance	Demand Forecast	9.355	0.945	40.226
	Supply Chain	4.907	0.828	16.908

shows the measurement properties for the formative construct. Initial investigation shows that the outer weights (column A) of the three constructs are insignificant (t < 1.96), namely complexity, resource risk, and transactional value. Further examination of their outer loadings (column B) indicates that the outer loading t-value of transactional value is significant (t = 4.198). Therefore, transactional value is retained as a dimension of DAA as it reflects the innovation’s ability to create operational benefits [83]. Meanwhile, the outer loading t-values for complexity and resource risk were insignificant, and the two dimensions were removed from subsequent estimation.

5.2. Structural Model Assessment

The structural model assessment (SMA) has two parts. The first is to assess the lateral collinearity, and the second is for hypothesis testing, determination coefficient (R²), effect sizes (f²), and predictive relevance (Q²). Lateral collinearity in models can contradict the findings, specifically for variables that are hypothesized to be causally related, especially in measuring the common construct [84].

Table 5. VIFs for Lateral Collinearity Assessment.

Construct	DAA	Operational Performance
DAA		1.748
Technological Context	1.444	1.556
Organizational Context	1.820	1.918
Environmental Context	1.736	1.805
Perceived Risk	1.515	1.549

shows that the VIFs are all under the value of 3.3, confirming that the SMA is free from lateral multicollinearity concerns. The multivariate skewness and kurtosis were also assessed. Although results based on Mardia’s multivariate skewness (β = 53.726, p < 0.01) and multivariate kurtosis (β = 339.514, p < 0.01) suggest that the data were not multivariate normal, the estimation approach approximates the data’s normality using a 5,000-subsample bootstrapping to estimate the t-value for path analysis [73].

Table 6. Results of Direct Path Analysis (H1 to H5).

Hypothesis	Relationship	Std. Beta	Std. Error	t-Value	p-Value	BCa-CI0.025	BCa-CI0.975	Decision	R2	f2	Q2
H1	TC→DAA	0.253	0.094	2.676	0.007	0.089	0.445	Supported	0.428	0.077	0.176
H2	OC→DAA	0.236	0.095	2.498	0.013	0.057	0.423	Supported		0.054
H3	EC→DAA	0.199	0.087	2.286	0.022	0.024	0.372	Supported		0.040
H4	PR→DAA	0.141	0.093	1.515	0.130	−0.047	0.308	Not Supported		0.023
H5	DAA→OP	0.404	0.088	4.611	<0.001	0.243	0.591	Supported	0.567	0.216	0.418

Note: Bootstrapping of 5000 at 95% confidence interval.

shows the results of H1 to H5, the direct path analysis for pull and push factors on DAA. Evidently, all three TOE constructs are significantly related to DAA. The beta coefficients (β) for H1, H2, and H3 are 0.253 (t = 2.676, p = 0.007), 0.236 (t = 2.498, p = 0.013), and 0.199 (t = 2.286, p = 0.022), respectively, confirming the crucial positive role of TOE constructs on DAA. Meanwhile, H4, which examines if the perceived risk is associated with DAA, is not supported (t = 1.515, p = 0.130). The results for H5 indicate that DAA is significant and positively related to operational performance (β = 0.404, t = 4.611, p < 0.001).

The effect of the four factors on DAA is substantial, with an R² of 0.428, exceeding 0.26 [85]. In other words, these factors are responsible for explaining 42.8% of variances in DAA. Similarly, DAA has a substantial effect on operational performance, as evident by R² of 0.567, implying that DAA explains 56.7% of the variance in operational performance. Individually, each of the TOE constructs exhibited a negligible effect on the DAA, measured by f², as the values are between 0.02 to 0.15 [85]. Meanwhile, DAA with an f² value of 0.216 has a medium effect (within the threshold of 0.15 to 0.35) in producing the R² for operational performance. Overall, the model presented sufficient predictive relevance, indicated by the Q² values exceeding zero both for DAA (0.176) and operational performance (0.418).

Table 7. Results of Mediation Analysis (H6 to H9).

Hypothesis	Relationship	Std. Beta	Std. Error	t-Value	p-Value	BCa-CI0.025	BCa-CI0.975	Decision
H6	TC→DAA→OP	0.102	0.044	2.313	0.021	0.028	0.197	Supported
H7	OC→DAA→OP	0.095	0.043	2.918	0.028	0.023	0.192	Supported
H8	EC→DAA→OP	0.080	0.044	1.822	0.068	0.009	0.178	Supported
H9	PR→DAA→OP	0.057	0.040	1.413	0.158	−0.010	0.151	Not Supported

Note: Bootstrapping of 5000 at 95% confidence interval.

illustrates the mediation results (H6 to H9). The study shows that DAA significantly mediates the relationships between the three TOE constructs and operational performance. The t-values and p-values for H6, H7, and H8 exceed the threshold of 1.645 and are lower than 0.05, respectively. This result is further confirmed by the bias-corrected confidence interval (BCa-CI), where the lower limit and upper limit confidence intervals are not straddled between the value of zero [86]. The result for H9, however, revealed that DAA does not significantly mediate the link between perceived risk and operational performance, as indicated by all the indicators (t-value, p-value, and BCa-CI).

The examination of the mediation’s path coefficients was carried out, and the results are shown in

Table 8. Path Coefficients Effects of Mediation Model.

Mediation Pathway	a Coefficient	b Coefficient	Indirect Effect (IE) (a × b)	Direct Effect (DE) (c’)	Total Effect (TE = IE + DE)	IE/TE
H6: TC→DAA→OP	0.253	0.404	0.102	0.161	0.263	38.8%
H7: OC→DAA→OP	0.236		0.095	0.306	0.401	23.6%
H8: EC→DAA→OP	0.199		0.080	0.104	0.184	43.5%
H9: PR→DAA→OP	0.141		0.057	−0.095	−0.038	−150.0%

Note: H9 is unsupported.

. The indirect pathway TC→DAA→OP (H6) shows that technological context encourages DAA (a = 0.253), and in sequence improves operational performance (b = 0.404). The indirect effect of TC→DAA→OP is 0.102 (0.253 × 0.404). As the total effect of the mediation is 0.263, the indirect pathway explains approximately 38.8% (0.102/0.263) of the total effect. Following a similar approach, the pathways OC→DAA→OP (H7) and EC→DAA→OP (H8) revealed that they contributed 23.6% and 43.5%, respectively, to the total effect. As all coefficients are positive, it is fair to conclude that efforts to enhance TOE contexts at firms would promote DAA, which subsequently improves operational performance. Although the pathway PR→DAA→OP (H9) yielded insignificant results, its total negative effect may serve as a precaution that perceived risk could not be taken lightly as it has the potential to hinder DAA and impede operational performance growth.

6. Discussion

The resultant outcomes of the study demonstrated the significant association of the pull factors under the TOE context with the DAA, in sequence affecting operational performance. It also reaffirmed the role of technological context influence on innovation adoption. With respect to the research hypotheses, all three pull factors representing the technological, organizational, and environmental contexts are positively associated with DAA. From the technological viewpoint, innovation and solution providers must ensure that data analytic system designs are user friendly and easily integrated with systems that are commonly used by industry players. Moreover, the solution providers can provide trial versions to new adopters to promote further innovation adoption. In essence, as firms will assess multiple technological criteria when deciding on DAA, system trialability would allow firms to conduct necessary system evaluation, which is vital to avoid postadoption incompatibility. Meanwhile, firms should have proactive initiative and lend support to encourage the adoption of innovative solutions, including searching for new solutions for data analytics.

Additionally, the study revealed organizational context as an influential pull factor toward DAA, which covers internal and external organizational aspects. Internally, top management support and organizational readiness are vital to DAA. It plays a central role in managing the changes in norms, values, and cultures and facilitating firm members to accept innovation [50] fully. Firms must also pay attention to organizational readiness in terms of resource availability and capabilities. For example, internal fund shortages and inadequacy of human capital would hinder DAA. Roles of external agencies also potentially affect DAA. For instance, financial institutions may offer support through the provision of a unique lending scheme; educational organizations could contribute in terms of developing and nurturing DA talents; governmental agencies may support by offering incentives for DAA. Accordingly, support and cooperation both from the internal and external organizational aspects would encourage a shift towards the DAA sphere. Overall, an inference from the organizational context describes the constructive internal and external attitude that would enhance the DAA.

The obtained results proved the significance of the pull factor under the environmental context. From the viewpoint of competitive pressure, firms may lose market shares if they are not alert and fail to respond to rivals’ strategies. Thus, tense competition within the industry induces firms’ reactions to adopt new technologies [87]. From our findings and concurring with Gangwar [2], the study exerts critical consideration for external support toward DAA. This form of support prompts firms’ DAA, encompassing reinforcement, knowledge sharing, and problem resolution. The experiential learning offered by external parties, i.e., DA solution providers, would promote DAA; hence, more channels must be established for this idea.

The study showed that DAA is associated with TOE pull factors and an antecedent of operational performance. TOE enhances operational performance through DAA, as evident by three mediation analyses showing significant results. Their level of effect on operational performance, indirectly via DAA, varies from one to another. Based on the empirical evidence presented, the technological context of TOE exhibited the most robust effects on DAA and, subsequently, the operational performance, followed by organizational and environmental contexts.

The role of perceived risk on DAA was not established in the study. At the time of the study, the DAA among firms in Malaysia was relatively low [88], and DAA among the firms was still in its infancy. Perhaps, these early adopters may view DA as a game changer, and its adoption would offer more benefits than inherent risk. Thus, they were more focused on the prospects of a successful adoption. Furthermore, it is also fair to assume that early adopters generally are risk takers and thus have a higher tolerance to risk [89]. The unsupported finding on the association of perceived risk with DAA may be viewed as encouraging as there was no evidence to support the claim that perceived risk would repel firms from adopting DA among the firms in the manufacturing and services sectors.

7. Conclusions

The study provided empirical evidence using data collected from 169 firms in the manufacturing and services sectors. This study shows that pull factors are more dominant than push factors in associating with data analytic adoption. Although the study covered firms in two economic sectors, it does support the proposition that data analytic adoption plays a strong contribution toward operational performance. From the viewpoint of technological context (compatibility and trialability), it supports the argument that the technological context was positively related to the adoption of data analytics. Meanwhile, organizational context (the top management support and organizational readiness perspective) and environmental support (the competitive pressure and external support perspective) demonstrated a similar link to data analytic adoption. The indirect effects between the three TOE contexts and operational performance, via data analytic adoption as moderator, have also been confirmed. Evidently, initiatives that increase support in terms of technological, organizational, and environmental contexts would promote the adoption of data analytics among firms, which in turn would improve operational performance.

As the TOE and TPR are considered flexible contextual theories, future research can include additional dimensions, which would broaden the scope of interpretation and provide a comprehensive view of the higher-order construct. New research may also explore if the size of firms would influence the relationship between the pull and push factors on the adoption of data analytics. This is because firm size potentially delivers varying challenges in priorities or strategy formulations. As the current study does not distinguish between new and old adopters, examination using multigroup analysis may unravel new insight as firms at different stages of adoption may encounter different challenges in each phase of data analytics. Lastly, future studies may adopt a mixed-method approach to understand how the progression of data analytics adoption affects firm performance.