A dynamic information technology capability model for fostering innovation in digital transformation

Yoshikuni, Adilson Carlos et al. Article A dynamic information technology capability model for fostering innovation in digital transformation Journal of Innovation & Knowledge (JIK) Provided in Cooperation with: Elsevier Suggested Citation: Yoshikuni, Adilson Carlos et al. (2024) : A dynamic information technology capability model for fostering innovation in digital transformation, Journal of Innovation & Knowledge (JIK), ISSN 2444-569X, Elsevier, Amsterdam, Vol. 9, Iss. 4, pp. 1-14, https://doi.org/10.1016/j.jik.2024.100589 This Version is available at: https://hdl.handle.net/10419/327491 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ A dynamic information technology capability model for fostering innovation in digital transformation Adilson Carlos Yoshikuni a , Rajeev Dwivedi b, *, Muhammad Mustafa Kamal c , Duanning Zhou d , Pradeep Dwivedi e ,S  ergio Apolin ario a a Center of Applied Social Science, Program of Postgraduate in Controllership and Corporate Finance, Mackenzie Presbyterian University (UPM), Brazil b Visiting Associate Professor, Department of Information Systems and Business Analytics, School of Business, Eastern Washington University, WA, USA c Associate Professor in Sustainable Operations and Analytics, Head of Sub-Department - Operations & Analytics, Department of Management, University of Exeter Business School Faculty of Environment, Science & Economy Building One, Streatham Ct, Rennes Drive, Exeter, EX4 4PU, UK d Professor, Department of Information Systems and Business Analytics, School of Business, Eastern Washington University, WA, USA e Researcher and Doctoral Candidate of Information Systems Area, Sage University, Indore, MP, India ARTICLE INFO Article History: Received 25 July 2024 Accepted 28 September 2024 Available online 6 November 2024 ABSTRACT The advent of the digital era has transformed the way businesses create, compete, and maintain their existence, particularly in the aftermath of the COVID-19 pandemic. To maintain resilience in the ever-changing business landscapes of today, especially in emerging economies, businesses utilize dynamic information technology capabilities (DITC) to cultivate organizational capacities that foster innovation. This article argues that Dynamic Information Technological Capabilities (DITC) allows companies to develop flexible and adaptive skills to foster innovation. The results from 684 Brazilian businesses showed that implementing DITC improved their ability to come up with both new ideas and ways to use existing ones. This was possible because DITC improved the firms’dynamic and improvisational skills. The post hoc analysis of FIMIX PLS and PLS-POS reveals that DITC plays a greater role in fostering innovation through dynamic capabilities (DCs) rather than improvisational capabilities. The research on unobserved heterogeneity showed that a high level of DITC hasBig and strong effects on developing dynamic and improvisational skills for coming up with new ambidexterity ideas. The results indicated that companies should integrate the potential of digital technology and information to establish organizational capacities that can effectively compete within developing economies. © 2024 Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Keywords: Digital era Dynamic information technology capability, Improvisational capabilities Explorative innovation Ambidexterity JEL classification: 000 014 030 031 035 036 Introduction Innovation has consistently been a critical driver of organizational growth, sustainability, and the enhancement of competitive advantage (Mehralian et al., 2024). Recent studies show that enterprises must pursue explorative and exploitative innovations to thrive in the digital era (Al Dhaheri et al., 2024;Wei et al., 2024), even more so in developing economies (Yoshikuni, 2024;Yoshikuni et al., 2024). Explorative innovations are considered radical, targeting emerging markets with new designs and channels, often requiring new knowledge (Aboelmaged et al., 2023). In contrast, exploitative innovations are incremental, enhancing existing products and processes to serve current markets better, building on established knowledge and skills (Jansen and Tempelaar et al., 2009). Emerging economies face significant challenges in driving innovation in the digital era after the COVID-19 crisis (Aftab et al., 2022), as they often struggle with limited access to advanced technologies, underdeveloped digital infrastructure, and more significant regulatory uncertainty, making it difficult for companies to compete with the innovation capabilities of developed markets (CIA, 2024). However, enterprises in this context must recognize and act upon the uncertain environment’s opportunities and mitigate threats, building organizational capabilities and competing through innovation (Dutta et al., 2022;World Economic Forum, 2016). Organizational capabilities refer to an enterprise’s ability to use resources and competencies to achieve business goals effectively (Grant, 1991;Porter, 1998;Wolf & Floyd, 2017). According to Winter (2003) and Daniel et al. (2014) zero-order ordinary and first-order * Corresponding author E-mail addresses: [email protected] (A.C. Yoshikuni), [email protected] (R. Dwivedi), [email protected] (M.M. Kamal), dzhou@ewu. edu (D. Zhou), [email protected] (P. Dwivedi), [email protected] (S. Apolin ario). https://doi.org/10.1016/j.jik.2024.100589 2444-569X/© 2024 Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Journal of Innovation & Knowledge 9 (2024) 100589 Journal of Innovation &Knowledge https://www.journals.elsevier.com/journal-of-innovation-and-knowledge capabilities can be not sufficient to compete in this environment, and second-order capabilities should be developed by enterprises. Organizational capabilities include ordinary capabilities for routine operations (Kohli & Grover, 2008;Melville et al., 2004), DCs for adapting to market changes (Helfat et al., 2023;Teece, 2018), and improvisational capabilities for quickly reconfiguring resources in response to unexpected situations (Pavlou & El Sawy, 2010). Hence, firms that substantially invest in IT to enhance organizational capabilities through (DCs (Steininger et al., 2022) and improvisational capabilities (ICs) (Yoshikuni & Dwivedi, 2023) can exploit opportunities, mitigate risks, and reconfigure their resource base, enabling them to better navigate and innovate within this dynamic context. New research studies say that businesses need to adopt digital technologies and build dynamic IT capabilities and flexible IT skills (Li & Chan, 2019;Steininger et al., 2022;Yoshikuni et al., 2024)to stay competitive in today’s fast-paced business world, where customer tastes change quickly, boundaries shift, and there is competition from around the world. Therefore, dynamic IT capability (DITC) is a first-order dynamic capability combining IT infrastructure (ITI) and IT knowledge management (ITKM). ITI involves the integration of digital technologies—such as data, computing, communication, and connectivity—to enhance customer understanding and optimize processes (Bharadwaj et al., 2013b;Li & Chan, 2019;Vial, 2019). ITKM focuses on strategic investments in developing digital skills, fostering cross-functional collaboration, and cultivating a digital culture while establishing governance and empowering employees to effectively utilize digital technologies (Li & Chan, 2019;Warner & W€ ager, 2019;Yoshikuni et al., 2024). Hence, in the digital era, enterprises face increasing pressure to develop second-order capabilities that leverage digital technologies to support and enhance a wide range of business activities (Li & Chan, 2019) and foster innovation, particularly in developing economies (Dutta et al., 2022). Furthermore, recent research on innovation highlights the need for future studies to clarify how enterprise characteristics are necessary to compete in uncertain markets in developing economies through innovation (Aboelmaged et al., 2023;Leal-Rodríguez et al., 2023). Innovation studies indicate that investigating the pathways through which digital adoption and knowledge management influence innovation is essential to fully understand the role of these mechanisms in driving innovation (Chen & Pan et al., 2024). Moreover, a recent systematic review on process innovation emphasizes the overlooked role of digital investments in empirical research, including technology infrastructure and digital knowledge, in adapting resources for innovation projects, leading to faster time-to-market, improved product quality, and enhanced customer satisfaction, and suggests that future research should address this gap (Goni & Van Looy, 2022). Additionally, recent studies call for further empirical research to explore different configurations of DITC components to validate the construct and enhance the understanding of it (Li & Chan, 2019). Finally, future research in information systems should investigate how IT helps organizations by using DCs (Mikalef et al., 2020) and improvisational capabilities (ICs) (Yoshikuni, 2022). The research should focus on how to make IT business value through digital technologies (Steininger et al., 2022). The overall purpose of the paper is to investigate the relationship between DITC, and organizational capabilities to create Innovation, to fill knowledge gaps mentioned by recent studies (Aboelmaged et al., 2023;L.Chen et al., 2024;Goni & Van Looy, 2022;Leal-Rodríguez et al., 2023;Li & Chan, 2019;Luftman et al., 2015; Mikalef, Pateli, et al., 2020;Steininger et al., 2022;Yoshikuni, 2022) and propose the following research question: (RQ) In developing economies, how can dynamic IT capabilities serve as a precursor to building dynamic and improvisational capabilities that drive explorative and exploitative innovation? This study makes significant contributions to the existing literature by addressing key gaps. Firstly, it addresses a gap in DITC research by operationalizing, validating, and deepening the understanding of this novel construct as a measure. First-order DITC enhances the IT-business value literature. Second, it advances the literature on organizational capabilities, particularly dynamic and improvisational capabilities, by demonstrating their critical role in driving innovation through DITC in the Digital Era. Lastly, it contributes to innovation literature by showing that explorative and exploitative innovations are viable strategies for competing in developing economies. The structure of the paper is as follows: The next section reviews the literature and outlines the research hypotheses and proposed model. Following is a detailed description of the research methodology, including variable measurement and statistical techniques. The subsequent section presents the empirical findings. Post hoc analysis examines the interrelationships between latent variables and explores potential unobserved heterogeneity within the proposed model. The final section discusses the findings, highlights contributions to the literature and practical implications, addresses limitations, and offers recommendations for future research. Fig. 1 presents the detailed research design. Theoretical background and hypotheses According to Teece et al. (2016), the resource-based view (RBV) underscores that organizations possess distinct resources and capabilities to leverage to secure a competitive edge. This view forms the foundation for understanding organizational capabilities as essential for achieving enterprise success (Amit & Schoemaker, 1993). The RBV suggests that firm capabilities lead to a competitive advantage when organizational assets can be valuable, scarce, difficult to replicate, and irreplaceable in their unique combinations (Taher, 2012). Previous research has significantly advanced the understanding and definitions of different types of organizational capabilities, particularly by distinguishing between operational (or ordinary) capabilities and DCs (Helfat & Winter, 2011;Winter, 2003). Operational capabilities, such as zero-order capabilities, allow an enterprise to sustain its current operations (Winter, 2003). They enable ongoing activities using consistent methods and scale to support existing products and services for the same customer base, thereby maintaining the status quo (Amit & Schoemaker, 1993;Dwivedi et al., 2023). On the other hand, a dynamic capability allows an enterprise to change its existing operations, enabling it to adapt or enhance its functions, whether by modifying operational capabilities, the organization’s resource base, or aspects of the external environment or ecosystem (Helfat et al., 2023;Teece et al., 1997). According to metaanalyses (Helfat et al., 2023) and theoretical studies (Teece, 2018)on organizational capabilities, the DCs view (DCV) is a better way to explain why companies do better in the market than the rigiditybased view (RBV). This is because DCs make a company more flexible over time and less rigid. This study uses Teece et al., 1997 definition of dynamic capabilities, which means that an organization can combine, develop, and change its internal and external skills to adapt to a world that is changing quickly and effectively. Dynamic IT capabilities (DITC) When these IT capabilities exhibit characteristics such as rarity, appropriability, non-replicability, and non-substitutability, they can create a competitive advantage, as outlined by the RBV (Taher, 2012; Wade & Hulland, 2004).. However, when IT-enabled operational capabilities are grounded in the RBV framework, they can be insufficient for adapting operational routines to address external challenges (Bharadwaj, 2000). Thus, first-order dynamic IT capabilities by the DCV become essential to empower organizational capabilities to integrate, build, and reconfigure internal and external IT competencies to A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 2 address rapidly changing environments (Li & Chan, 2019;Steininger et al., 2022). The company possesses strong organizational capabilities, with IT playing a crucial role in enhancing them, leading to significant improvements in operational excellence and competitive responsiveness (Mikalef & Pateli, 2017;Yoshikuni et al., 2024). Therefore, DCs are characterized by IT-embedded DCs, incorporating IT resources within their structure (Li & Chan, 2019). These IT-embedded DCs include general DCs and more specific forms, such as ITenabled agility, flexibility, exploitative innovation, explorative innovation, ambidexterity, business process performance, and other routines (Steininger et al., 2022). Hence, IT-enabled DCs emphasize capacities directly supported by specific technical assets, including IoT data, enterprise resources planning, business data analytics, business intelligence, strategic enterprise management, big data, etc. IT capabilities include the use of IT resources, including physical and human assets, as well as complementary organizational resources, to optimize business processes and improve immediate and long-term outcomes (Melville et al., 2004). When these IT capabilities exhibit characteristics such as rarity, appropriability, non-replicability, and non-substitutability, they can create a competitive advantage, as outlined by the RBV (Taher, 2012;Wade & Hulland, 2004). However, the RBV framework-based IT-enabled operational capabilities may not be sufficient for adapting operational routines to external challenges (Bharadwaj, 2000). Thus, the DCV’sfirst-order dynamic IT capabilities become essential to empower organizational capabilities to integrate, build, and reconfigure internal and external IT competencies to address rapidly changing environments (Li & Chan, 2019;Steininger et al., 2022). The company possesses strong organizational capabilities, with IT playing a crucial role in enhancing them, leading to significant improvements in operational excellence and competitive responsiveness (Mikalef & Pateli, 2017;Yoshikuni et al., 2024). Therefore, DCs are characterized by IT-embedded DCs, which incorporate IT resources into their structure (Li & Chan, 2019). These DCs with-IT built-in have both broad and specific DCs, such as IT-enabled agility, flexibility, exploitative and explorative innovation, ambidexterity, business process performance, and other routines (Steininger et al., 2022). Hence, IT-enabled DCs emphasize capacities directly supported by specific technical assets, including IoT data, enterprise resources planning, business data analytics, business intelligence, strategic enterprise management, big data, etc. This research is based on the definition of dynamic IT capability [DITC, (Li & Chan, 2019)], operationalized as the first-order IT dynamic capability created by IT units to remain responsive to evolving technology and market conditions in today’s business environment. Broadly, DITC is described as the ability of an IT unit to obtain, utilize, integrate, restructure, and modify the organization’sIT resources to enable organizational capabilities and achieve business goals (Bharadwaj, 2000;Bharadwaj et al., 2013a). In this study, DITC represents a first-order DCs composed of two interconnected and interdependent DCs linked to the ordinary capabilities of IT infrastructure (ITI) and IT knowledge management (ITKM), as proposed by Li and Chan (2019) in their conceptual study on DITC. The ITI refers to how an enterprise integrates IT infrastructure through digital technologies, including data, computing, communication, and connectivity (Bharadwaj et al., 2013a;Vial, 2019), to enable organizational capabilities to understand customers better, connect customer-facing and operational processes, market and sell products and services, deliver customer service, and provide a comprehensive view of critical operational and customer information (Li & Chan, 2019). The ITKM refers to how an enterprise makes a strategic investment in developing an innovative mindset (Li & Chan, 2019) through cross-functional collaboration between business and IT teams, cultivating a digital culture by enhancing digital skills and mindset, advancing digital initiatives and collaborative learning across local and corporate units (Warner & W€ ager, 2019), establishing clear roles and responsibilities for governing digital efforts, and empowering all members to engage in discussions around the advantages of utilizing digital technologies (Vial, 2019). Organizational capabilities enabled by DITC Organizational capabilities refer to an enterprise’s ability to effectively utilize its resources and competencies to achieve specific business objectives (Grant, 1991). These capabilities are categorized into Fig. 1. Research design process A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 3 (1) ordinary (or operational) capabilities, which involve routine activities that support the ongoing production, sales, and service of products or services (Winter, 2003); (2) DCs, which focus on the strategic adaptation and transformation of the firm’s ordinary capabilities and resources in response to changing market conditions and opportunities (Helfat & Winter, 2011;Helfat et al., 2023); and (3) improvisational capabilities, which reflect the firm’s ability to spontaneously reconfigure existing resources to develop new operational capabilities that address urgent, unpredictable, and novel environmental situations (Pavlou & El Sawy, 2010;Yoshikuni, 2022;Zhang et al., 2023). Together, these capabilities enable an organization to maintain competitiveness and drive long-term success. DITC is defined as first-order DCs that enable organizational capabilities as a second-order into DCs (Daniel et al., 2014)and improvisational capabilities (Pavlou & El Sawy, 2010). Past research on DCs is grouped into three main organizational capacities: (1) sensing, involving the recognition and evaluation of potential opportunities and risks; (2) seizing, allocating resources to respond to these opportunities or risks and deriving benefits; and (3) transforming, focused on ongoing adaptation and renewal (Teece, 2007,2018;Teece et al., 2016). Thus, DITC helps firms to capture, harness, and understand customer data (Aydiner et al., 2019;Knabke and Olbrich, 2018;Pappas et al., 2018;Mikalef et al., 2019) through digital technologies, such as social, and mobile, analytics and cloud [ITI, (Frishammar et al., 2018)]. According to Warner and W€ ager, (2019), digital innovation comprises new capabilities in digital scenario planning, digital scouting, and digital mindset crafting to pinpoint new technological, customer, and competitor-based trends to create innovation. This study proposesthefollowinghypothesistoaddresstheknowledgegapin IS literature (Li & Chan, 2019;Steininger et al., 2022)onhow firms develop organizational DCs through digital and knowledge capabilities to enable second-order DCs. H1a. DITC drives the building of dynamic capabilities. Improvisation capability (IC) is the capacity to act spontaneously in trying to respond to problems or opportunities in a novel way (Vera et al., 2016;Yoshikuni, 2022). IC utilizes existing resources in real time to build new operational capabilities that better match novel environmental situations (Hadida & Tarvainen, 2015;Tseng et al., 2015;Zhang J. et al., 2023). Digital innovation fosters identification and the rapid embrace of changes, allowing organizations to gain the best performance through faster innovation (Levallet & Chan, 2018;Nambisan et al., 2017). The authors (Levallet & Chan, 2018;Yoshikuni, 2022)defined digital culture as the improvisational ability to act promptly in an unplanned manner (i.e., spontaneously) and/or creatively in the face of uncertainty. DITC enabled by analytics and big data may play a central role in decision-making in the face of highly unstructured tasks being carried out in high uncertainty (Aydiner et al., 2019;Knabke & Olbrich, 2018;Mikalef, Krogstie, et al., 2020;Steininger et al., 2022). Hence, IT infrastructure associated with IT knowledge management leveraging improvisation capabilities among managers brings about more success in decision-making as they move forward. Therefore, DITC’s leverage on IC promotes effective improvisation when carrying out our activities, dealing with unanticipated events on the spot, and responding quickly to unexpected problems (Levallet & Chan, 2018).. This study hypothesizes that: H1b. DITC drives the building of improvisational capabilities. Organizational capabilities for innovation Most research findings provide different perspectives on the variations of innovation (Benner & Tushman, 2015;Mehralian et al., 2024; Gholamhossein Mehralian et al., 2024). However, most of the studies claim that innovation is anything original or an improved idea, recombination of old ideas, approaches, or methods, or anything that is perceived as new or improved (Alves et al., 2017;Goni & Van Looy, 2022;Jha & Bose, 2016). The exploration-exploitation framework of organizational learning gained prominence within organizational, strategic innovation (Benner & Tushman, 2015;Konlechner et al., 2018) and IS literature (Yoshikuni & Galv~ ao, 2023;Yoshikuni et al., 2024). According to Jansen et al. (2009), exploratory and exploitative innovation are classified into two domains: proximity to existing technologies, products, and services and proximity to existing customers or market segments. However, enterprises that create together explorative and exploitative innovation are considered ambidexterity (Yoshikuni, 2024). Exploratory innovations help meet new demands for products and services, embrace challenges to serve new markets, and develop new distribution channels, units, and production lines (Jansen, et al., 2009). Previous studies of innovation (Tsai, 2015;Yoshikuni et al., 2024)defined exploration innovation as a radical innovation that creates innovation value in a novelty dimension that renders the new product radically differentiated from the conventional product specifications. Exploitative innovation is defined as the improvement of existing products and services with minor and frequent adjustments to the value proposition to maintain and/or expand their current customer and market share (Behnam & Cagliano, 2019;Goni & Van Looy, 2022;Jansen, Vera, et al., 2009). DCs reflect an organization’s ability to respond timely, rapidly and flexibly to product innovation to achieve new and innovative forms of competitive advantage given path dependencies and market positions (Lau & Lo, 2019;Lu & Ramamurthy, 2011;Warner & W€ ager, 2019;Zhang et al., 2023). The dynamic capability framework recognizes the importance of evolutionary innovation (bottom-up) (Konlechner et al., 2018) as the creation of new knowledge to develop breakthrough products, i.e., exploration innovation (Helfat et al., 2023;Teece, 2018). Previous studies demonstrated that DC was positively and highly significant to exploration innovation (effectiveness to create new products) and exploitation innovation (adaptive efficiency to renew existing products) (Pavlou & El Sawy, 2006,2010;Protogerou et al., 2012;Wilhelm et al., 2015). To expand the knowledge and contribute to innovation management research, it is recommended that further studies investigate new constructs associated, with exploration and exploitation innovation (Benner & Tushman, 2015;Jansen, Vera, et al., 2009) leveraged by digital technologies (Nambisan et al., 2017)andspecific context factors, such as developing economies (Konlechner et al., 2018;Warner & W€ ager, 2019). This study declares the following hypotheses: H2a. Dynamic capabilities positively influence explorative innovation. H2b. Dynamic capabilities positively influence exploitative innovation. ICs are spontaneous and creative abilities (Vera et al., 2016)to reconfigure existing resources, build new operational capabilities, and address urgent, unpredictable, and novel environmental situations (Pavlou & El Sawy, 2006;J.Zhang et al., 2023). Exploratory innovation entails the search for new knowledge to create products and services for emerging markets and customers. On the other hand, exploitative innovation builds on existing knowledge resources and improves existing products and services for current markets (Jansen, Vera, et al., 2009). Previous studies on IC demonstrated that they have positive effects on innovation and are relevant to knowledgebased processes (Vera et al., 2016), information systems development (Du et al., 2019), decision support systems (Mendon¸ca, 2007; Mishra et al., 2023), IT-enabled organizational virtues (Chatterjee et al., 2015) by digital innovation (Levallet & Chan, 2018;Pavlou & El Sawy, 2006,2010). A growing number of studies are investigating the effects of IC on innovation, but few studies focus on identifying the relationship between improvisational capabilities and exploration/ A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 4 exploitation innovation. Hence, the study proposed the following hypotheses: H3a. Improvisational capabilities positively influence explorative innovation. H3b. Improvisational capabilities positively influence exploitative innovation. The proposed research model presented (Fig. 2) illustrates our hypotheses. Research methodology Sample size Based on a comprehensive review of the relevant literature, a survey instrument was developed, and a pre-test survey was conducted to check the clarity of the items’contents, response time, and related observations (Yayla and Hu, 2012). The pre-test respondents were four IT/IS researchers and three senior executives who commented on improving the clarity of the measurement items. See Appendix A, which summarizes the measures and sources of the variables used in the analyses. Convenience sampling was utilized to gather data from Brazilian organizations. Following established practices in management and IS research, respondents were contacted through diverse channels, such as social media, industry association lists, personal networks, and academic directories curated by the authors. Additionally, the questionnaire instructions encouraged respondents to seek input from other knowledgeable members within their organization if they required additional expertise on the survey constructs. The targeted respondents were chosen based on their position, experience, and professional knowledge of IS and management. To motivate and encourage executive respondents to participate in the IT capabilities survey, they were provided with feedback on how their company performed in the assessed aspects of the proposed model, along with general recommendations on addressing any potential issues identified. The study used an online data collection approach that did not permit missing data entry, as participants had to respond to a particular question before proceeding to the next question. The research instrument was prepared before the onset of COVID-19. However, due to the pandemic, respondent participation was initially low, and most responses were collected after the pandemic began. The final response was received in mid-2023. Additionally, the study examined suspicious response patterns as recommended by Hair et al. (2022). A visual inspection of the data was analyzed and verified to see if respondents marked the same response for all questions or alternated between extreme pole responses, and none of the cases were removed. The total sample size of respondents in the online platform was 684 firms. The minimum study sample size was tested using the G*Power v.3.1.9.2 software (Faul et al., 2007) with a median effect size [f 2 ]of 0.15 and statistical power of not less than 0.80. The minimum sample size was 92 cases, which indicated the sample size validation. Outliers were analyzed by the Mahalanobis square distance (DM 2 )(Cousineau & Chartier, 2017), and five cases presented high DM 2 values (range of statistical residual of the DM variable = 17,878 until 22,799 and, with a probability value of p >0.001) indicating multivariate outliers. The outcomes were obtained from the database and the final sample was 684 cases. The sample size exceeded the requirements. Data distribution was verified, and the variables’normality was evaluated using coefficients of skewness (Sk) and kurtosis (Ku). Univariate and multivariate variables showed no severe violation of the normal distribution assumption (|Sk| <3 and |Ku| <10) (Hair et al., 2022). Finally, statistical techniques were applied to detect and (where possible) control the common method bias. Consistent with Chin et al. (2013), the measured latent marker variable (MLMV) technique was used for the model as the dependent variable. Measurement of the variables DITC was a second-order (reflective construct) measured through two constructs (Li & Chan, 2019): IT knowledge management (ITKM) was measured as a first-order construct (reflective) based on the conceptual studies from Li and Chan (2019),Warner and W€ ager (2019), and Chanias et al., (2019); while IT infrastructure (ITI) was a firstorder construct (reflective) based on the conceptual studies from Bharadwaj et al., (2013a);Li and Chan (2019);Vial, (2019). Organizational capabilities through DCs were based on Pavlou and El Sawy (2010) as a first-order construct, and improvisational capabilities as a first order-construct from Chatterjee and colleagues (Chatterjee et al. 2015), Innovation of exploration and exploitation adopted by Jansen and colleagues (Jansen, Vera, et al., 2009). Measured latent marker variable (MLMVformative) (Almeida et al., 2022;Yoshikuni, 2024) were also determined, see Appendix A. The scale content was Fig. 2. Proposed model A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 5 validated by an IS field specialist with more than 10 years of experience (researchers and professors), and validity and reliability were assessed by statistical tests (see Table 3), as indicated by Morgado et al. (2018). The primary constructs were measured through sevenpoint Likert-type scales ranging from 1 = “strongly disagree”to 7=“strongly agree.”.Table 1 measures and data sources for the constructs used in the proposed model. Control variables According to Melville et al. (2004) and Kohli and Grover (2008),an enterprise’s characteristics, including size, age, industry sector, and other factors, can influence how IT generates business value. Previous information systems (IS) studies have highlighted these aspects (Mikalef & Pateli, 2017;Yayla & Hu, 2012;Yoshikuni, 2022). Firm size (SIZE) was measured by the number of employees. It was measured as an ordinal variable with the following values: micro (1−9 employees), small (10−49 employees), medium (50−249 employees), and large (above 250 employees). To control the effect of the industry sector (SECTOR), four categories were created: agribusiness, manufacturing, service, and government. Responses were collected from agribusiness (3 %), manufacturing (30 %), services (63 %) and government (4 %). The survey was predominantly completed by 31 % C-level executives (e.g. chief executive officers), 33 % management and coordination personnel, and 36 % supervisors with decision-making powers. The firm size (number of employees) was grouped into large (65 % >500 employees), medium (17 % - 50-249 employees), small (12 % - 10-49 employees), and micro (5 % - 0-9 employees) classes. Statistical Techniques Hypotheses Hypotheses were tested using partial least squares structural path modeling (PLS-PM) by the SmartPLS software package (Ringle et al., 2015). PLS-PM was particularly appropriate for this study because the structural model was complex (many constructs and many indicators). Formative measured constructs are part of the structural model, which helps comprehend the FIMIX-PLS and PLS-POS approaches for identifying and treating unobserved heterogeneity. Common Method Bias in the Survey There is potential for common method bias (CMB) in this study. The bias is controlled during the research design phase by using priori approaches (Schwarz et al., 2017), such as choosing respondents who are able to answer the questionnaire, items constructed in clear and concise language, counterbalancing the order of questions and anonymity of the respondent, and applied technical remedies as suggested by MacKenzie and Podsakoff (2012) and Fuller et al. (2016). Finally, the statistical technique was applied to detect where possible to control CMB in line with Chin et al., (2013). The MLMV technique was used for the model of exploitation and exploration innovation constructs. According to Chin et al., (2013), four items designed to have the lowest possible correlation with the other constructs under investigation were used. The MLMV was adapted through formative indicators used for MLMV analysis by Yoshikuni et al. (2024). The model with MLMV variables revealed a more fitting sight than the original one (difference less than 1 % in all variance explanation −R 2 ), see Table 4 Thus, this suggests that CMB is not a severe concern in this study. Measurement of the first-order model The reflective latent variables were subjected to the tests of reliability, convergent validity, and discriminant validity. (Akter et al., 2017;Hair et al., 2022), and all latent variables were connected for these analyses as recommended by Bido and Silva (2019).Table 3 shows that the indicators have higher factor loadings on their assigned constructs, above 0.70, and indicate the discriminant validity (Akter et al., 2017;Hair et al., 2022). Items with lower loadings were omitted from the measurement model, as indicated by Bido and Silva (2019) and Hair et al. (2022). Convergent validity was assessed by examining whether AVE > 0.50 and each construct’s AVE square root was greater than its highest correlation with any other construct (Fornell-Larcker criterion). Composite reliabilities, Cronbach’s alpha, and Dijkstra-Henseler’s indicator (Rho_A) are >0.68 (Bido & Silva, 2019;Hair et al., 2017)of all constructs (Table 2) and all values of HTMT confidence interval were lower than 0.85 which supported the HTMT criterion (Henseler et al., 2015). The second-order DITC variable yielded an AVE value of 0.535 and a CR estimate of 0.919. A comparison of the Fornell −Larcker criterion with the square root of DITC (0.731) AVE values showed the criterion to be satisfied. Empirical results Structural model Fig. 3 and Table 4 summarize the structural model from the PLS analysis: the effect size of path coefficients (f 2 ), path coefficients (b), and coefficient of determination (R 2 ). The significance of estimates (tstatistics) is obtained by performing a bootstrap analysis with 5000 resamples. From the observation of Table 4, the result of the direct effects was presented (value p <0.05) to the relationship between DITC !DC, DITC !IC, DC!ER, DCI !EP, IC !ER, and IC !EP, thus, hypotheses H1a, H1b, H2a, H2b, H3a, and H3b were supported. The structural model accounts for 44,8 % of the variance in dynamic capability (R 2 = 0.448), 19.4 % of the variance in improvisational capability (R 2 = 0.194), 35.3 % of the variance in the innovation of exploration (R 2 = 0.353), and 45.8 % of the variance the innovation of exploitation (R 2 = 0.458). For the control variables, only Sector was found to significantly influence EP and ER (p-value <0.05), demonstrated in Table 4. Table 1 Summary of constructs, measures and sources Constructs Measures Sources Dynamic IT capabilities IT Infrastructure Li and Chan (2019), Chanais et al. (2019) IT knowledge manage Bharadwaj et al. (2013a),Li and Chan (2019),Vial (2019). Organizational capabilitiies Dynamic capability Pavlou and El Sawy (2010) Improvisational capability Chatterjee et al. (2015) Innovation Exploration Jansen et al. (2009),Yoshikuni et al. (2024) Exploitation Jansen et al. (2009),Yoshikuni et al. (2024) Control variables Firm-sized Melville et al. (2004) Industry sector Melville et al. (2004) Measured latent marker variable MLMV Yoshikuni et al. (2024) A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 6 Post-hoc analysis Several post hoc analyses were conducted to clarify (1) the interrelationships between DITC, DC, IC, EP, and ER further and explore (2) potential unobserved heterogeneity within the proposed model. The first post hoc analysis assessed statistical differences between parameter estimates in Partial Least Squares Path Modeling (PLS-PM) by comparing the effects of two-parameter estimates with two separate tests (Rodríguez-Entrena et al., 2018). Table 5 highlights the differences in the relationships between DITC and DC, and DITC and IC, as well as the differences in the effects of DC on EP and ER, and finally, the differences in the relationships between IC and EP, and IC and ER. The (1) difference of 0.208 between the path coefficients of DITC !DC and DITC !IC was statistically significant (p-value < 0.05). Similarly, the (2) difference of 0.059 between the path coefficients of DC !ER and DC !EP was also statistically significant (pvalue <0.05). However, the (3) difference of 0.002 between the path coefficients of IC !EP and IC !ER was not statistically significant (p-value >0.05). All tests were conducted following the methodological framework recommended by Rodríguez-Entrena et al. (2018). Second, post hoc analysis allowed us to identify and interpret possible “unobserved heterogeneity”in the relationship within the proposed model, verifying the existence of factors not included in the original analysis, which may explain the differences between various groups. For these companies, the final blending technique (FIMIXPLS) was used as recommended by Hair et al. (2016) and Matthews et al. (2016) On the PLS-PM analyses. To identify the number of unobservable segments the FIMIX-PLS algorithm [SmartPLS v3 software (Ringle et al., 2015)] was used and executed 10 times for g = 2-5 segments using the Akaike Information Criteria (AIC), Factor 3 Modified AIC (AIC3), Bayesian Information Criteria (BIC), Consistent AIC (CAIC), Hannan-Quinn Criteria (HQ) and Standard Entropy Statistics (EN) that presented satisfactory criteria for segment selection. Indicators that presented lower values in specific information criteria portrayed the best segment solution and showed EN values above 0.50 (Hair et al., 2019). According to the authors (Hair et al., 2022), the first criterion verified that the AIC3 and CAIC values were smaller per segment tentatively considering the lower values together between AIC3 and BIC, plus the segment number as indicated by AIC4 and BIC (lowest values). Generally, the study chooses the smallest segment indicated by AIC and more segments than indicated by MDL5 in cases where the segment size meets the minimum sample size (92 cases). Table 6 shows the fit indices for a oneto five-segment solution. According to the segmentation’s criteria, the three segments were adequate and thus preceded the prediction-oriented segmentation Table 2 Assessment of convergent and discriminant validity of reflective constructs. Constructs 123456 1. IT Infrastructure 0.784 2. IT Knowledge Management 0.697 0.806 3. Dynamic Capabilities 0.656 0.567 0.811 4. Improvisational Capabilities 0.408 0.334 0.576 0.798 5. Exploitative Innovation 0.599 0.619 0.650 0.446 0.765 6. Explorative Innovation 0.525 0.550 0.573 0.411 0.741 0.817 Cronbach’s Alpha 0.842 0.865 0.895 0.732 0.857 0.834 Rho_A 0.855 0.870 0.897 0.774 0.863 0.841 Composite Reliability 0.888 0.902 0.920 0.840 0.894 0.889 Average Variance Extracted (AVE) 0.615 0.649 0.657 0.636 0.585 0.668 Table 3 Factor loadings (bolded) and cross-loadings of reflective constructs. First Latent Variable Items DCI ICI ITI IKM EPIA ERIA Dynamic Capabilities in Innovation (DCI) DC1 0.809 0.423 0.530 0.487 0.579 0.520 DC2 0.804 0.412 0.539 0.488 0.551 0.458 DC3 0.837 0.431 0.570 0.487 0.515 0.436 DC4 0.776 0.495 0.522 0.414 0.450 0.390 DC5 0.816 0.559 0.515 0.419 0.503 0.459 DC6 0.820 0.492 0.513 0.454 0.551 0.510 Improvisational Capabilities in Innovation (ICI) IC1 0.600 0.821 0.467 0.361 0.457 0.414 IC5 0.325 0.765 0.177 0.175 0.239 0.223 IC6 0.370 0.805 0.243 0.206 0.309 0.293 IT Knowledge Management (ITKM) ITKM1 0.534 0.368 0.816 0.544 0.517 0.487 ITKM2 0.569 0.325 0.861 0.627 0.554 0.498 ITKM3 0.549 0.318 0.819 0.671 0.520 0.423 ITKM4 0.454 0.278 0.744 0.452 0.368 0.316 ITKM5 0.453 0.315 0.664 0.395 0.356 0.303 IT Infrastructure (ITI) ITI1 0.484 0.270 0.636 0.866 0.518 0.453 ITI2 0.372 0.235 0.453 0.773 0.444 0.409 ITI3 0.359 0.244 0.476 0.796 0.430 0.412 ITI4 0.510 0.312 0.545 0.773 0.540 0.477 ITI5 0.539 0.283 0.670 0.815 0.550 0.462 Innovation of Exploitation (EPI) EP2 0.348 0.241 0.343 0.305 0.660 0.562 EP3 0.558 0.447 0.506 0.491 0.758 0.557 EP4 0.588 0.393 0.507 0.488 0.809 0.599 EP5 0.519 0.344 0.438 0.440 0.787 0.545 EP6 0.473 0.335 0.436 0.517 0.808 0.647 Innovation of Exploration (ERI) ER2 0.564 0.462 0.459 0.460 0.594 0.785 ER3 0.475 0.312 0.478 0.497 0.647 0.881 ER4 0.383 0.307 0.361 0.416 0.525 0.772 ER5 0.429 0.249 0.401 0.416 0.646 0.826 A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 7 (PLS-POS). The PLS-POS algorithm [SmarPLS v3 software, (Ringle et al., 2015)] was run to evaluate the differences in the segments and workarounds. The sample sizes defined by PLS_POS were 1-segment, 253 (38 %), 2-segment, 264 (40 %), and 3-segment, 157 (24 %). The PLS-POS R 2 values of the variables were compared. Table 8 shows the coefficient of determination (R 2 ) for Original, 1-segment, 2-segment, 3-segment, and PLS-POS Average. It demonstrates high differences between the original sample R 2 and segments 1, 2, and 3 and the PLSR 2 Average, see Table 7. The values were compared, and evidence showed clear differences between the path coefficient estimates of three segments in the structural models’effects, see Table 8. All segmentation (1, 2, and 3) Fig. 3. Final model. Note: p-value <0.5 *; p-value <0.01 **; p-value <0.001 ***; no significant NS. Table 4 Relationships between all latent variables. Variables Relationship f 2 effect Size Path Coefficient Standard Error tvalue p-value R 2 R 2 with MLMV DITC !DC 0.724 0.685 0.026 26.353 0.000 0.448 0.463 DITC !IC 0.241 0.477 0.037 12.858 0.000 0.194 0.194 DC !ER 0.223 0.484 0.039 12.308 0.000 IC !ER 0.020 0.140 0.040 3.515 0.000 0.353 0.356 SIZE !ER 0.001 -0.030 0.046 0.644 0.520 SECTOR !ER 0.009 0.078 0.035 2.207 0.028 DC !EP 0.336 0.543 0.040 13.742 0.000 IC !EP 0.023 0.138 0.038 3.662 0.000 0.458 0.464 SIZE !EP 0.023 -0.114 0.083 1.375 0.169 SECTOR !EP 0.015 0.090 0.036 2.483 0.013 Note: DITC: Dynamic IT capabilities; DC: Dynamic Capability; IC: Improvisational Capability; ER: Explorative Innovation; EP: Exploitative Innovation; SECTOR: Firm Sector; SIZE: Firm Size. Table 5 Results of parameter differences based on different confidence intervals (CI). Type of Confidence Interval (a=5 %) Differences of path coefficients (1) b1a 6¼ B1b (2) b2a 6¼ b2b (3) b3a 6¼ b3b Lower bound Upper bound Lower bound Upper bound Lower bound Upper bound Standard 0.202 0.206 0.316 0.332 - 0.008 - 0.003 Percentile 0.149 0.273 0.215 0.471 - 0.075 0.065 Basic 0.148 0.272 0.221 0.477 - 0.077 0.063 Note: b1a: DITC !DC; BH1b: DITC !IC; b2a:DC !ER; b2b:DC !EP; b3a:IC !ER; b3b:IC !EP Table 6 Fit indices for a oneto five-segment solution. S AIC AIC3 AIC4 BIC CAIC HQ MDL5 EN Relative Segment Sizes g=1 g=2 g=3 g=4 g=5 s=2 7,418.57 7,463.57 7,508.57 7,622.33 7,667.33 7,497.42 8,797.36 0.66 0.604 0.396 s = 3 7,221.90 7,289.90 7,357.90 7,529.80 7,597.80 7,341.05 9,305.41 0.65 0.479 0.312 0.209 s = 4 7,010.35 7,101.35 7,192.35 7,422.39 7,513.39 7,169.80 9,798.57 0.72 0.339 0.33 0.211 0.120 s = 5 6,834.31 6,948.31 7,062.31 7,350.50 7,464.50 7,034.06 10,327.25 0.75 0.317 0.203 0.177 0.159 0.144 Legend: AIC - Akaike’s Information Criterion, AIC3 - Modified AIC with Factor 3, AIC4 - Modified AIC with Factor 4, BIC - Bayesian Information Criteria, CAIC - Consistent AIC, HQ - Hannan Quinn Criterion, MDL5 - Minimum Description Length with Factor 5, and EN - Entropy Statistic (Normed) A.C. Yoshikuni, R. Dwivedi, M.M. Kamal et al. Journal of Innovation & Knowledge 9 (2024) 100589 8