Dynamic Capability theory evolved from the resource-based view (RBV) theory (Lin & Wu, 2014; Peteraf et al., 2013; Schilke et al., 2018), which suggested prior to DC that firms can gain a sustained competitive advantage based on the value, rarity, inimitability and non-substitutability of their resources and capabilities (Barney, 1991). Subsequently, DC diverged from the static capabilities-based RBV theory by arguing that competitive advantage is not an equilibrium state, but rather the result of firms’ continuous efforts to build, integrate and reconfigure their resources and capabilities to respond to changes in the market (Teece et al., 1997). Although it is worth mentioning that Eisenhardt and Martin (2000) believe that DC can only provide a competitive advantage to a limited extent in high velocity contexts. Consistent with extant DC literature (Mousavi et al, 2018; Peteraf et. al., 2013; Schilke et al., 2018; Shepherd et al., 2019), this study is aligned to the approach to dynamic capabilities as espoused by Teece et al. (1997).

Resources refer to the assets a firm has to create and implement strategies (Barney, 1991), while capabilities refer to the effective recombination of resources to achieve business objectives (Dutta et al., 2005). Thus, DCs are defined as ‘higher-level competences that determine the firm’s ability to integrate, build, and reconfigure internal and external resources and/or competences to address, and possibly shape, rapidly changing business environments’ (Teece, 2012, p. 1395). This enables organisations to direct their efforts towards higher pay-off endeavours (Teece, 2014).

Dynamic Capability can be further segmented into base and higher-order tiers (Teece, 2007, 2018). Base DCs facilitate the recombination and adjustment of resources and capabilities and the creation of new ones (Teece, 2007), which results in innovation, market expansion and astute managerial decision-making (Teece, 2018; Winter, 2003). Higher-order DCs, which are based on repeated routinisation (Helfat & Peteraf, 2003, 2015; Mousavi et al., 2018) build upon this to inform and guide business strategy and operations at a macro level for long-term sustainability (Schoemaker et al., 2018; Teece, 2018). This includes sensing future opportunities and designing business models and organisational structures based on the optimum configuration for its existing and future paths (Teece, 2018).

The process by which DCs enable change in organisations is summarised by its disaggregation into sensing, seizing and transforming processes (Teece, 2007). Prior to the introduction of DC, Day (1994) suggested that sensing refers to the ability of a firm to gain awareness of its market environment and respond to this new knowledge by taking action. Dynamic Capabilities support this process by routinising the scanning, searching, exploring, learning and interpreting processes that transform information into knowledge that can be acted upon to create new opportunities (Helfat & Peteraf, 2003; Teece, 2007). Sensing also enables firms to improve their capability to predict future sources of competitiveness by assessing the value and rarity of capabilities and resources (Matysiak et al., 2018). Seizing refers to the ability to mobilise resources to capture value from opportunities (Teece, 2014). This is done by leveraging a firm’s products, services and processes, and where required, investing in the development and commercialisation of opportunities (Teece, 2007). Finally, transforming refers to firms’ capacities to enhance, combine and reconfigure the organisations’ tangible and intangible resources and capabilities to remain competitive (Teece, 2012). This is required when a major change occurs in the market and technological environment.

According to the DC theory, successful sensing, seizing and transforming activities allow firms to sustain their competitive advantage by continuously reconfiguring and evolving their resources and capabilities to deviate from unfavourable, towards favourable, path dependencies and remain competitive in an ever-evolving environment of business (Matysiak et al., 2018; Teece, 2007).

In the last two decades, digital networks have connected more people, devices and sensors, which has led to a notable increase in the scale and speed of data generation, and in turn to the big data phenomenon (Grover et al., 2018). Big data is defined according to its size (volume), the speed at which it is generated (velocity), its diverse formats (variety), its bias and abnormality (veracity) and the value it delivers (value) (Fosso Wamba et al., 2015), referred to as the 5V model.

As big data has become more prevalent, firms have increased efforts to develop capabilities to derive value from it, which is referred to as BDAC (Mikalef et al., 2020). Big Data Analytics Capability is defined as organisations’ inimitable and distinctive capabilities to deploy technology and human resources to collect, store, process and analyse big data (Fosso Wamba et al., 2017; Gupta & George; 2016; Mikalef et al., 2019). The capability places equal focus on technical assets, human competences and organisational culture to generate and disseminate insights effectively across the business (Fosso Wamba et al., 2017; Gupta & George, 2016; Mikalef et al., 2019).

Technical assets, referred to as tangible resources, comprise the data, infrastructure and financial resources required to fund big data projects (Gupta & George, 2016; Mikalef et al., 2020). Human capabilities comprise the human skills related to technical, managerial and relational capabilities of big data employees (Gupta & George, 2016; Mikalef et al., 2020). Finally, organisational culture refers to the intangible resources related to embracing data-driven insights and organisational learning (Gupta & George, 2016; Mikalef et al., 2019).

Through this process of managing diverse internal and external information, BDAC enables organisational sensing, which supports both operational and strategic decisions regarding the effectiveness and efficiency of firms (Fosso Wamba et al., 2017; Fosso Wamba, Queiroz et al., 2020). From an operational perspective, BDAC supports firms’ efforts in securing future revenue from existing customers and reducing operating costs by enhancing the efficiency of production operations (Fosso Wamba, Dubey et al., 2020; Grover et al., 2018; Holmlund et al., 2020; Rialti et al., 2019).

To enable the former, BDAC supports customer experience and customer satisfaction efforts by generating insights related to customers’ behaviour, attitudes and preferences by integrating data from social, digital and physical interactions (Grover et al., 2018; Holmlund et al., 2020). These insights can be used to enhance user touch points, identify product optimisation opportunities and influence the personalisation of purchase recommendations and discounts (Grover et al., 2018). When customer experience is enhanced, it improves firms’ business relationships with their customers, driving customer satisfaction and retention (Grover et al., 2018; Holmlund et al., 2020; Sorescu, 2017).

From a production perspective, efficiency can be achieved through the collection of historical data using complex BDAC infrastructure, which allows processes to be fine-tuned and fast-response mechanisms to be developed to address issues proactively (Grover et al., 2018; Rialti et al., 2019). This enhances the agility and adaptability of supply chains, as less time is spent on breakdowns, which contributes to optimised execution and competitive performance (Fosso Wamba, Dubey et al., 2020; Grover et al., 2018).

From a strategic perspective, BDAC facilitates the generation of insights that influence the value creation, value proposition and value capture strategies of firms (Ciampi et al., 2021). This is enabled by continual scanning, analysing and interpreting market, customer and competitor trends that highlight opportunities for growth (Grover et al., 2018). In fact, having access to information enables firms to gain a first-mover advantage, by reacting quickly to high potential opportunities (Côrte-Real et al., 2017). Furthermore, Mikalef et al. (2019) found that BDAC not only enables the development of incremental innovations based on existing products and services but also radical ones that result in new value propositions. Lehrer et al. (2018) had similar findings and found that it also enhances the likelihood of success of innovations.

An organisation’s business model is the architecture that summarises its structure, content and governance processes used to extract value from transactions and business opportunities (Teece, 2010; Zott & Amit, 2010). When major changes occur in the environment, such as shifts in technological innovation, customer demand and competitor intensity, it can result in transformational activities to remain competitive (Bhatti et al., 2021; Foss & Saebi, 2017; Zhang et al., 2021). This process is referred to as BMI, which is defined as the deliberate, novel and non-trivial modification, adaption or innovation of key components of the business logic (Foss & Saebi, 2017; Zhang et al., 2021). This leads firms to rally efforts towards reconfiguring and recombining their resources, assets and structures, and in some cases, creating entirely new break-out structures to maintain competitiveness (Teece, 2007).

Clauss (2017) suggested that BMI can lead to competitive advantage by optimising firms’ value creation, value propositions and value capture strategies. New value creation opportunities can be achieved by enhancing the firms’ resources and capabilities by launching new technology or infrastructure or creating new strategic partnerships (Clauss, 2017; Teece, 2010). Value propositions are related to firms’ products and services that support targeting of new customer segments with higher willingness to pay or market expansion into more lucrative segments (Clauss, 2017; Teece, 2010). Finally, value creation refers to the strategies taken by firms to generate revenue and profit from business activities, which can be enhanced by launching new revenue models or adopting flexible processes that lead to cost reduction (Clauss, 2017; Foss & Saebi, 2017). In all instances, the goal of BMI is to transform key components of the firm’s business logic to remain competitive in the long term (Foss & Saebi, 2017; Teece, 2018; Zhang et al., 2021).

In his seminal research, Porter (1985) suggested that competitive advantage is core to firms’ performance in competitive markets and can be achieved through two competitive strategies. These strategies, namely, cost leadership and differentiation, consider the gap between the price charged for a product or service and its perceived value from customers relative to competitor strategies (Porter, 1985).

In the years following, RBV theory suggested an alternative approach to achieving competitive advantage, which was less related to the competitive strategy and more focused on firms’ abilities to attain rare and valuable resources (Barney, 1991). Resource-based view theory argued that when these resources were not evenly distributed among competitors, firms could sustain a competitive advantage. Dynamic Capability theory shared the focus of RBV theory, based on high-quality resources and capabilities, but argued that competitive advantage was achieved through their continual reconfiguration to more complex, inimitable and distinctive states to remain competitive (Eisenhardt & Martin, 2000; Teece et al., 1997). The emergence of generative AI technologies made possible through new business models that make such technologies freely available to customers and competitors, while giving their developers a competitive advantage, further emphasises that rare and non-inimitable resources are not a sustainable source of competitive advantage (Budhwar et al., 2023). Unsurprisingly, the DC perspective has been endemic in management literature in the last three decades and has formed the theoretical foundation of this research (Eisenhardt & Martin, 2000; Teece 2007, 2010, 2012, 2014, 2018; Teece et al., 1997).

To measure competitive advantage, this research assessed competitive performance indicators related to financial and market performance (Behl et al., 2022). These indicators are commonly used in academia to assess firms’ performance relative to competitors and were, thus, deemed relevant for this study (Akter et al., 2016; Behl et al., 2022; Fosso Wamba et al., 2017).

To test the hypotheses, data were collected using a cross-sectional approach via an online questionnaire. The units of analysis were organisations that use data and analytics as part of their business processes, while the sampling unit was employees that either work in data and analytics departments or interact regularly with them. Regular interaction was required for respondents that do not work in data and analytics departments because of the focus on big data analytics questions within the survey that necessitated an understanding of the topic. Respondents with seniority of manager level and above were targeted as these respondents have a broad understanding of strategic decisions and business processes to assess the effect of advanced analytics and BMI on performance (Ciampi et al., 2021). No restrictions were placed on the country of residence, because of the pervasive nature of BDAC and the ubiquitous adoption of big data analytics technologies, tools and infrastructure supported globally by the growth in mobile phones, social media, cloud-enabled platforms and the Internet of Things (Grover et al., 2018). This resulted in responses from 42 countries, which allowed greater generalisability of the results (Ciampi et al., 2021).

The questionnaire was administered online from August to September 2022. A total of 277 surveys were attained. During data screening, five respondents were eliminated as the variability in their responses did not exceed the threshold of 0.25 standard deviation (Collier, 2020). All responses exceeded the minimum completion time threshold of two seconds per question (Huang et al., 2012). No missing values were present in the questionnaire because of limitations imposed in the questionnaire design. Thus, the final sample size comprised 272 valid responses.

Three constructs were measured as part of the conceptual model by using previously validated scales from academic literature (see Table 1). Big Data Analytics Capability was measured as a third-order formative construct consisting of 25 indicators based on the validated scale of Mikalef et al. (2019). Big Data Analytics Capability comprised three second-order formative constructs, namely, tangible resources, human skills and intangible resources, with 10, 8 and 7 indicators, respectively. Tangible resources comprised three first-order formative constructs including data, technology and basic resources with three, five and two indicators, respectively. Human skills comprised two first-order reflective constructs including managerial skills and technical skills both with four indicators. Finally, intangible resources comprised two first-order reflective constructs including data-driven culture and organisational learning with three and four indicators.

Business Model Innovation was assessed as a first-order reflective construct with indicators evaluating organisations’ foci on new or existing activities related to the nine elements of the business model canvas (Bhatti et al., 2021; Osterwalder & Pigneur, 2010). These included key resources, activities, partnerships, channels, customer relationships, customer segments, value propositions, cost structure and revenue streams (Bhatti et al., 2021; Osterwalder & Pigneur, 2010).

The final construct, competitive performance, was also a first-order reflective construct comprising five financial and market indicators. These included revenue, profit, operating costs, market share and service and product quality (Behl et al., 2022). Respondents assessed the indicators of the constructs on a seven-point Likert scale. Screening and demographic questions were assessed using categorical responses.

The analysis was conducted using Partial Least Squares Structural Equation Modelling (PLS-SEM) using SmartPLS 3.0 software. The same approach was taken by Ciampi et al. (2021), who assessed the relationship of BDAC with BMI and Akter et al. (2016), Mikalef et al. (2020) and Fosso Wamba et al. (2017), who assessed the relationship of BDAC with competitive performance.

Partial Least Squares Structural Equation Modelling was selected as it is recommended for research where structural models contain both formative and reflective constructs and where latent variable scores are required to develop higher-order constructs (Hair et al., 2011). The assessment of the hypotheses required evaluating the significance of the total, direct and indirect path coefficients (Hair et al., 2017). This was assessed using a bootstrapping technique with 5000 random samples with a 95% confidence interval.

To mitigate the effects of common method bias, the confidentiality and privacy of respondents were protected by using an anonymous survey (Podsakoff et al., 2003). Furthermore, indicator vagueness was avoided by using validated scales from literature, dividing the questionnaire into separate parts for each of the constructs and conducting a pilot on the questionnaire before administering it to the research respondents (Ciampi et al., 2021).

The pilot was conducted in two phases. The initial phase included five respondents where feedback was received that the questionnaire length, which exceeded 80 questions, was too long. This resulted in shortening the questionnaire by shifting to more succinct scales for BDAC and BMI (Bhatti et al., 2021; Mikalef et al., 2019). In the second pilot, comprising 10 respondents, the questionnaire length was deemed appropriate. Feedback was received to add definitions for uncommon terms to ensure understanding across industries. This was implemented to strengthen the accuracy of responses.

Harman’s single factor test was also conducted to assess whether the first factor accounted was below the threshold of 50% (Fuller et al., 2016). This was the case as the first factor accounted for 35.94% variance, concluding that common method bias was not a risk.

The ‘ten times rule’ was used to determine the minimum sample size, which suggests that the size should be 10 times greater that the number of indicators of the largest construct (Peng & Lai, 2012). In this research, BDAC has the greatest number of indicators with 25 measures, which led to a minimum sample size of 250. This sample size exceeded the critical sample size of 200 for Structural Equation Modelling, which provides stable parameter estimates and adequate power to test structural models (Collier, 2020). The final sample size of 272 exceeded both criteria.

The validation and reliability criterion used to assess the measurement model differed for reflective and formative constructs. This was because the internal consistency assessments that underpin reflective model assessment are not relevant for formative constructs as they do not necessarily co-vary (Hair et al., 2017).

The reliability and validation of the reflective constructs, which comprised all first-order constructs apart from data, basic resources and technology, were verified by conducting tests for multi-collinearity, factor loadings, internal consistency reliability, convergent reliability and discriminant validity. Variance inflation factor (VIF) was used to evaluate whether multi-collinearity exists. All constructs fell below the threshold of five, indicating that no collinearity issues existed (Hair et al., 2017). The factor loading results demonstrated that most indicators exceeded the threshold of 0.7, apart from five factors that fell between 0.6 and 0.7. For complex constructs, as is the case in this research, it is recommended that indicators that have factor loadings that exceed 0.6, and average variance extracted (AVE) values higher than 0.5 should be retained in the model (Collier, 2020). While these indicators may have a weaker contribution, they may capture a unique component of the construct (Collier, 2020). Table 1 summarises the factor loadings for reflective indicators and VIF results.

The internal consistency reliability was tested via Cronbach’s alpha and composite reliability. In both cases, the constructs exceeded the threshold of 0.7, verifying their reliability (Nunnally, 1978; Nunnally & Bernstein, 1994). The convergent validity was assessed through AVE where all constructs exceeded the threshold of greater than 0.5 verifying convergent validity (Fornell & Larcker, 1981).

The discriminant validity was assessed through three tests, namely, Fornell and Larcker criterion, cross-loadings and the Heterotrait-Monotrait (HTMT) ratio. The cross-loadings demonstrated that the indicators loaded most strongly on their parent constructs (Farrell, 2010). The Fornell and Larcker criterion illustrated that the square root of the AVE of a construct was greater than its correlation with other constructs (Fornell & Larcker, 1981). Finally, the HTMT ratio tested that the ratio of between-trait and within-trait correlations of constructs had a threshold of less than 0.85 (Henseler et al., 2015). In all three cases, the discriminant validity was verified. Table 2 and Table 3 summarise the construct correlation matrix, internal consistency reliability, convergent reliability and discriminant validity (specifically HTMT ratio) results.

To assess the measurement model of the formative constructs, namely data, basic resources, technology, tangible resources, human skills, intangible resources and BDAC, the significance of the outer-weights, the size of the outer-loadings and the collinearity of the constructs was assessed. Most of the outer weights were significant apart from the indicators T2 and T3 and the construct human skills. This required further assessment of the size of the outer loadings to establish whether they exceeded 0.5. As this was the case, the constructs were retained in the measurement model (Hair et al., 2017). A VIF assessment was conducted to test for multi-collinearity. All values fell below the threshold of five, indicating an absence of multi-collinearity (Hair et al., 2017) (see Table 4). Based on these assessments, all reflective and formative models met their required criterion, and the measurement model was verified to be reliable and valid.

Figure 2 summarises the structure of the measurement model and its indicators, illustrating the BDAC third-order construct dimensions, the factor loadings for reflective constructs and the significance of outer weights for formative constructs.

Ethical clearance to conduct this study was obtained from the University of Pretoria’s Gordon Institute of Business Science Research Ethics Committee.

The empirical results of the study are based on 272 respondents whose demographics are summarised in Table 5. The demographics section comprised seven respondent questions and two company questions. The dominant respondent groups were males, aged 30 years to 39 years, with post-graduate education, with more than 10 years’ work experience, at management position, working in data and analytics departments, and residing in the Netherlands, United States, United Kingdom or South Africa. From a company perspective, the dominant industries were professional services and technology industries, and the dominant company size was large.

Figure 3 and Table 6 summarise the structural model from the PLS-SEM analysis by providing the standardised path coefficients (β), their significance (t-value) and the explaining variance of the endogenous variables (R²). The analysis was conducted using a bootstrap method based on 5000 random samples.

The results demonstrated that all the hypotheses were supported. Big Data Analytics Capability had a significant relationship with competitive performance (β = 0.343, t = 4.768, p < 0.001; Hypothesis 1 was confirmed). Big Data Analytics Capability also had a significant relationship with BMI (β = 0.640, t = 15.672, p < 0.001; Hypothesis 2a was confirmed). Business Model Innovation had a significant relationship with competitive performance (β = 0.245, t = 3.048, p < 0.01; Hypothesis 2b was confirmed).

Finally, the indirect relationship of BDAC→BMI→CP was also found to be significant indicating a mediation path (β = 0.156, t = 2.974, p < 0.01; Hypothesis 2c was confirmed). The combined effect of BDAC on competitive performance, based on the sum of the direct and indirect effect, was 0.500 (t = 10.603, p < 0.001). Thus, the direct effect and indirect effect accounted for 69% and 31%, respectively. As both the direct and indirect effects were in the same direction and significant, the nature of the median is complementary partial mediation (Collier, 2020; Hair et al., 2017).

The structural model explained a variance rate of 41% for BMI (R² = 0.409) and 29% for competitive performance (R² = 0.285). This indicated a weak to moderate predictive accuracy as the values exceeded 25% but were less than 50% (Hair et al., 2017). To further understand the predicative relevance of the exogenous variable, BDAC, on the endogenous variables, BMI and competitive performance, a Q² analysis was conducted using a blindfolding procedure (Shmueli et al., 2019). The results indicated that satisfactory predictive relevance was achieved (Shmueli et al., 2019). Big Data Analytics Capability had a large predictive relevance for BMI as the result of 0.406 exceeded 0.35, and a medium predictive relevance for competitive performance as the result of 0.246 exceeded 0.15 (Hair et al., 2017).

The study has theoretical contributions in BDAC, BMI and DC literature. From a BDAC perspective, it validated direct and indirect positive relationships between BDAC and competitive performance. This brings about the opportunity for organisations to foster the development of BDAC by specifically investing in big data infrastructure, attaining relevant human skills and the development of a data-driven corporate culture to translate insights to action (Gupta & George, 2016; Mikalef et al., 2019). It also demonstrated the central role that BDAC has in informing strategic decisions aimed at sustaining competitive advantage in the long term. This supported emerging research that suggests that big data analytics is not subordinate to strategy, but a key enabler in defining the future path of the organisation, illustrating the fusion and integration between data, technology and strategy (Mikalef et al., 2020).

Furthermore, the study contributed to BMI literature by illustrating, for the first time, the positive mediating role of BMI in the BDAC-competitive performance relationship. This finding demonstrated two principles; firstly, that the availability of BDAC resources and capabilities advanced the alertness of firms to internal and external triggers, thus allowing them to proactively respond to opportunities (Ciampi et al., 2021). This was because of the scanning, searching, researching, interpreting, experimenting and discovering that sensing enables (Day, 1994; Teece, 2007), which leads to transformational activities that drive BMI (Teece, 2018). Secondly, when business models were effectively reconfigured, competitive performance was enhanced (Bhatti et al., 2021; Foss & Saebi, 2017; Teece, 2018; Zhang et al., 2021). While the relationship found by this research was weak, it remains significant.

Finally, the validated mediation path extended DC literature by demonstrating the role that BDAC-enabled sensing had in generating insights that influence the transformation of business models, which in turn enhance competitive performance. This reflects the sensing, seizing and transforming process of DCs (Teece, 2007).

From a managerial perspective, the results suggested that to support the effective transformation of business models, organisations should foster the development of BDAC. This includes hiring and training skilled employees, adopting organisation knowledge sharing and learning practices and supporting the development and adoption of a data-driven culture (Mikalef et al., 2019). This requires the support from leadership to develop and execute dedicated plans regarding investment in big data infrastructure, the recruitment of relevant skills to advance the big data analytics department and relevant training programmes to shift the corporate culture.

Once BDAC resources and capabilities are established, leadership are encouraged to create a strong integration between BDAC departments and the strategy of the organisation. As a result, it is recommended that BDAC management should form part of the leadership team responsible for strategy development, and big data analytics employees should be incentivised to balance the generation of insights that benefit both the short term and long term.

The research presented some limitations that might be areas to explore in future research. When considering the structure of the research, the survey approach, the scales and the cross-sectional nature of the research, each introduced limitations. The questionnaire relied on self-reported responses, which were subjective, and might have undermined the objectivity of the results because of bias (Ciampi et al., 2021). In terms of the scales, a shorter set of scales were selected to measure BMI to optimise the survey length based on feedback from the pre-testing phase. While this had a positive impact on drop-out rate, it may have reduced the comprehensiveness of the questionnaire related to BMI. An opportunity exists to strengthen insights related to the dimensions of BMI, namely, value proposition, value creation and value capture, using the more comprehensive scales by Clauss (2017).

The cross-sectional nature of the study provided insights that were relevant to a specific point in time and would not adapt with time (Saunders & Lewis, 2018). Longitudinal research could be conducted in the future to overcome this limitation, as it is less likely to be impacted by temporary factors, and therefore would be more robust (Köhler et al., 2017). Finally, the generalisability of the results might be limited because of the purposive non-probability sampling approach used where judgement was applied to source respondents.

Despite the limitations discussed, the research brought about possible future research opportunities. As discussed, while the research validated that BDAC has an influence on competitive performance, the mechanisms through which it acts could be investigated in the future (Mikalef et al., 2020). Furthermore, based on the focus on BMI, an opportunity exists to assess the effect of moderating variables on the structural model. Foss and Saebi (2017) suggested that macro-level, firm-level and micro-level moderators might influence the antecedents and outcomes of BMI. Macro-level influence includes changes to the external environment, firm-level influence refers to organisational culture, structure and leadership characteristics, and micro-level influence refers to management cognition, open mindedness and adversity to change (Foss & Saebi, 2017).