Big data governance and knowledge creation: the role of big data management capabilities in the hotel industry

Big data governance (BDG) indicates the management of processes, procedures, people and technologies with the intent to generate big data for organisational decision-making (Moradi and Hepsø, 2024; Wang et al., 2018). The big data also supported the intelligence system technologies consisting of artificial intelligence etc. Bena et al. (2025). Organisations have growing attention at harvesting big data to gain greater insight about their customers, prepare better products and provide tailored services (Tan and Zhan, 2017). In the hospitality industry, hotels have been found to draw on knowledge generated from big data to track and predict customer behaviours and tourism markets (Li et al., 2018). In turn, learnings obtained from big data applications (e.g. information regarding guest satisfaction and experiences) have been said to enhance organisational performance and innovation (Busulwa et al., 2022). BDG has therefore been identified as a critical capability for all types of organisations (ranging from multinationals to SMEs) to attain a competitive advantage (Ferrigno et al., 2024; Li et al., 2018).

Considering value creation and performance benefits associated with BDG, scholars have repeatedly called for more research to understand how BDG can be operationalised to reap maximum benefits (Shamim et al., 2020b; Wang et al., 2018). As indicated by Hussain and Nisar (2022), the presence of skilled operational staff is significant for knowledge creation and the development of radical, incremental and ambidextrous innovation in the Pakistani hotel industry. Our investigation answers this call by explaining the underlying mechanism by which BDG relates to knowledge creation through big data (KCBD). BDG encompasses contractual and relational governance dimensions (Lee and Cavusgil, 2006). Contractual governance (CG) mechanisms emphasise using written, valid contracts to elaborate the functions and responsibilities between partners dealing in big data. In contrast, relational governance (RG) mechanisms are grounded on trust and social norms and imply a collective expectation that partners will eagerly be involved in knowledge sharing to support their relationships (Aben et al., 2021; Janssen et al., 2017). CG and RG mechanisms are both important as they facilitate organisational access to quality data and effective decision-making (Wang and Jiang, 2022). Overall, because BDG supports management to make the timely decisions through resource utilisation (Patel, 2022), we propose that CG and RG mechanisms promote big data management capabilities (BDMC), which successively lead to KCBD. Furthermore, given the critical role of a data-driven culture (DDC) in driving successful big data implementation and performance in organisations (Karaboga et al., 2023; Chatterjee et al., 2024), we put forward that DDC moderates the BDMC–KCBD relationship.

BDMCs have been said to empower organisations to strategically restructure and optimise their big data resources (Mao and Lu, 2023) as a response to growing competitive pressures and disruptive environments. BDMCs are dynamic capabilities that enable organisations to access and transform high volume, velocity and variety of raw information into valuable data for decision-making purposes (Gao and Sarwar, 2024). The emerging research on BDMC differentiates between strategic-level BDMCs (SLBDMCs) and operational-level BDMCs (OLBDMCs) (Shamim et al., 2021; Teece, 2007). Being future-focused, opportunity-seeking and value-creating in nature, SLBDMCs have been associated with organisational factors that promote organisational coordination, configuration and renewal. This includes functions that facilitate long-term and high-level planning and oversight, support the internal alignment of organisational mission, vision and goals, and provide strategic direction to organisational members (Gupta and George, 2016; Gao and Sarwar, 2024; Mao and Lu, 2023). Firms that orchestrate the right strategic resources and capabilities (skills, expertise, leadership, technologies, systems, culture and relationships) are in a better position to facilitate big data success and achieve competitive advantage in the long run (Braganza et al., 2017; Nisar et al., 2021).

In contrast, OLBDMCs are dynamic capabilities that support firms to run smoothly within the context of day-to-day operations (Zhou et al., 2024) and to translate strategic business ideas into reality (Gupta and George, 2016; Shamim et al., 2020a) – for example, through developing the skill set of operational staff (Drnevich and Kriauciunas, 2011). Although work on OLBDMCs is still at an embryonic stage, Zeng and Glaister (2018) established a framework classifying OLBDMCs as big data democratisation, contextualisation, experimentation and execution. These scholars argue that organisations’ ability to transform big data to create value does not depend on the data itself but rather on what firms do with the data. Adopting a process-oriented approach, these scholars show that firms’ big data value creation stems from managers’ (rather than data scientists’) ability to democratise, contextualise, experiment with data and execute data insights promptly to create value (Zeng and Glaister, 2018).

In many developing economies, hotels face unique challenges, such as limited infrastructure, skills shortages and resource constraints, that shape how they govern and leverage big data towards knowledge creation. Empirical evidence from a developing economy, Pakistan, demonstrates that strategic and operational big data management capabilities significantly enhance knowledge creation, ultimately translating into improved service innovation and higher online quality ratings (Shamim et al., 2021). In Malaysia’s hotel sector, big data analytics capability has been shown to drive both marketing and organisational innovation, which in turn elevates firm performance (Khalil and Abd Aziz, 2023; Khalil et al., 2023). Thus, these studies underscore that, in developing-country hospitality settings, BDG enables hotels to create actionable knowledge and pursue both innovative performance and long-term sustainable growth. Nisar et al. (2021) found the data management findings revealed that data management difficulties are significant initiatives for big data decisions capabilities in both public and private hospitals. Therefore, situating the proposed model is especially critical in these settings, where governance mechanisms must compensate for systemic limitations and catalyse value from fragmented data environments.

Our study is motivated by four gaps. First, prior studies on big data and BDG primarily focus on high-tech sectors (e.g. finance, healthcare and manufacturing) in developed and resource-rich economies like the USA, Europe and China (Lyu et al., 2022; Abdian et al., 2023). The lack of big data research in developing countries and hospitality means that those resource-constrained countries, which would benefit most from big data-driven value creation initiatives, are deprived of critical insights Abdian et al. (2023), thereby potentially exacerbating knowledge inequalities for under-resourced countries. Second, with the bulk of the literature focusing on the broader concept of big data analytics capabilities (Foroughi et al., 2024; Huynh et al., 2023), BDMC has received limited scholarly attention. This oversight is problematic given that management capabilities that support organisations to analyse big data are fundamental to make the big data initiatives successful and therefore warrant dedicated research Shamim et al. (2021). Third, research that explores the functions of management capabilities in knowledge creation under the influence of DDC is limited (Chatterjee et al., 2024). To date, no study has examined the impact of DDC on KCBD in the hotel industry. Amazingly, the DDC has been recognised as an integral part of big data initiatives and firm performance (Karaboga et al., 2023; Chatterjee et al., 2024), particularly within the hospitality context. Finally, a methodological gap exists, with cross-sectional research designs dominating the intersecting literature between big data and knowledge creation (Huynh et al., 2023).

While addressing these gaps, our study makes several important contributions. First, our comprehensive model provides novel evidence from a non-Western context, thereby facilitating progress towards the United Nations Sustainable Development Goals No. 9 (industry, innovation and infrastructure) and No. 10 (reduced inequalities). By positioning our work in the highly competitive Pakistani hospitality sector, we offer a pathway for Pakistani hotels to reflect on how to maximise the use of their scarce resources (Zaman et al., 2022), create value through better decision-making (Cappa et al., 2021) and respond to fierce challenges (Busulwa et al., 2022). In addition, our choice of a low-tech sector like hospitality is deliberate, as research suggests that the technology intensity of an industry impacts the outcomes of information systems capabilities, with high-tech intensity sectors being said to draw higher impacts from information systems capabilities than medium-tech sectors (Felipe et al., 2020). By extending big data research to hospitality, we clarify the big data implications in a sector traditionally considered low-tech.

Second, by embedding BDMC within our framework, we add to the sparse body of work on BDMC. Furthermore, given the interconnections between strategic-level and operational-level capabilities (Witcher and Chau, 2014), we specifically examine BDMC at the strategic and operational levels. Other than Shamim et al. (2021), who have made a rare attempt to differentiate between these two levels of analysis, prior studies have either treated BDMC as a monolithic construct (Wamba et al., 2017; Mao and Lu, 2023) or have focused on one BDMC level only (Zeng and Glaister, 2018). By distinguishing between strategic and operational levels, we:

1. show how different types of BDMCs complement and interact with each other in big data knowledge creation processes; and

2. contribute to governance literature by showcasing how various BDG mechanisms (RG and CG mechanisms) influence strategic and operational dynamic capabilities in the hospitality sector.

Our fine-grained, multi-level approach not only makes a significant conceptual advancement over existing single-level models but also sheds light on whether certain BDMCs (i.e. strategic or operational) are more effective in driving KCBD outcomes.

Third, by incorporating DDC as a moderator, we contribute to an understanding of when and under what conditions BDMCs lead to KCBD, thereby facilitating an understanding of why big data success varies across contexts. Understanding the cultural dynamics behind big data success is crucial to help organisations foster a cohesive culture that aligns with their digital transformation objectives (Orero-Blat et al., 2025). Finally, by collecting time-lagged data, we advance the literature from a methodological perspective. Our methodological approach allows for rigorous testing of mediation and moderation variables while simultaneously mitigating common method bias through temporal separation of predictor, mediator and outcome variables (Podsakoff et al., 2024).

This paper is structured as follows: the next section presents the theoretical background and hypotheses underpinning the study. This is followed by the research methodology and findings. The paper concludes with a discussion that includes the theoretical and practical contributions, study limitations and future research directions.

This study is underpinned by social capital theory (Putnam, 1993; Schuller and Theisens, 2010) and the knowledge-based dynamic capability view (Mikalef et al., 2018; Zheng et al., 2011), which stems from the dynamic capability perspective (Teece, 2007). The dynamic capability perspective represents an addition to the resource-based view (Barney, 1991). The resource-based view stipulates that organisations can achieve a cutthroat advantage if they exploit their strategic assets and resources (i.e. those that are valuable, rare, inimitable and non-substitutable). However, following criticisms that the resource-based view neglects external environmental factors, this theory was extrapolated to form the dynamic capability perspective (Horng et al., 2022).

Dynamic capabilities refer to organisations’ ability to point out opportunities in the market, develop new capabilities, as well as reconfigure present capabilities in line with external environmental conditions. In brief, the dynamic capability perspective contends that strategic resources must be accompanied by effective management practices for firms to achieve a competitive advantage (Teece, 2007; Zheng et al., 2011). Dynamic capabilities are inherently associated with knowledge and can be nurtured via organisational learning and knowledge management practices (Shamim et al., 2020b). The integration of these various research streams eventually led to the development of the knowledge-based dynamic capability view (Mikalef et al., 2018; Zheng et al., 2011).

Social capital deals with social networks, exchanges that arise from them and their value within the business environment. This concept is grounded in two theoretical perspectives. One is led by Bourdieu (1986), whose centre of attention was the functions of social and other forms of capital in generating unequal power relations. The other perspective is led by Putnam (1993), who regarded social capital as an aspect of a community and argued that social capital arise from networks, standards and trust that is created within a group. From this perspective, social capital is a relational resource (Nahapiet and Ghoshal, 1998) that lays the foundation for RG mechanisms (Janssen et al., 2017; Poppo and Zenger, 2002). Social capital theory has gained traction in knowledge sharing and management research, with shared values (trust and credibility) and shared language and narratives being seen as essential for collaboration and consensus across stakeholders (Sen et al., 2025).

We jointly draw on Putnam (1993) view of social capital and the knowledge-based dynamic capability view (Mikalef et al., 2018; Zheng et al., 2011). In our study, social capital acts as the enabling infrastructure for knowledge-based dynamic capabilities. The relational dimension of social capital (trust and reciprocity) reduces transaction costs in knowledge exchange, making organisations more efficient at developing sensing, seizing and reconfiguring capabilities (Gao and Sarwar, 2024). These theories have been supported by scholars as being useful in understanding the concepts of BDG and KCBD. RG and CG mechanisms (i.e. BDG) provide quality data and associated knowledge that can be leveraged to nurture dynamic capabilities like BDMCs (Janssen et al., 2017; Shamim et al., 2021; Wang and Jiang, 2022). The data driven culture also supported the performance of SMEs. (Chau et al. (2024). As supported by Janssen et al. (2017), social capital enables organisations to manage access to big data and provide big data-related knowledge to processing and analytics teams. Furthermore, BDMCs at the strategic and operational levels are key heterogeneous capabilities that promote KCDBShamim et al. (2021). Having discussed this study’s theoretical principles, we now present the hypotheses underpinning our conceptual framework.

Nisar et al. (2021) defined BDG in terms of data arrangement and utilisation, and taking advantage of efficient and effective decisions for value creation in organisations. Cost-effective decisions require big data capabilities in integrating novel data sets. It has two dimensions: relational and contractual governance. Janssen et al. (2017) described relational governance as trust and confidence building between organisations or entities and big data vendors for optimal knowledge-sharing and big data interpretation. In addition, contractual governance denotes the formal and legal agreement between organisations and BD vendors to maintain data quality and provide a sound understanding of big data to perform their responsibilities Shamim et al. (2020b). BDMCs are dynamic capabilities that equip organisations with the right resources to sense and seize opportunities and create knowledge and value from big data Shamim et al. (2019). At the strategic level, BDMCs support organisations in strategy formulation, implementation and evaluation (Gupta and George, 2016; Gao and Sarwar, 2024; Mao and Lu, 2023). Researchers and practitioners have identified interconnected dimensions of SLBDMCs, including leadership capability, talent management and technology management (Nisar et al., 2021; Zhang and Li, 2022).

Leadership capability relates to leaders’ ability to identify the need for change and make decisions on how to respond to such change. For example, this may involve reconfiguring organisational skills to accommodate new routines Nisar et al. (2021). Leaders with IT understanding and experience may be more prone to infusing employees with common big data goals as well as investing in IT infrastructure boosts organisational performance and innovation (Baharuden et al., 2019). As highlighted by scholars, availability of additional data does not assure successful big data initiatives. Instead, leaders’ commitment, involvement and visionary goals around big data are fundamental for the effectiveness of such initiatives (Chatterjee et al., 2024), including in the hospitality sector (Nguyen et al., 2024). In terms of talent management, organisations that attract, retain and train tech-savvy employees are more likely to overcome business challenges and meet changing market needs. Talent management is the proper planning, forecasting and managing of staff required to efficiently and effectively run the organisation (Carpenter, Bauer, Erdogan, & Short, 2013; Nisar et al., 2021). This is because tech-oriented employees can better interpret big data and translate implications into familiar business language for managers and other workers (McAfee et al., 2012). Finally, technology management denotes the use of technologies and information management for value creation through effective decision-making (Chen and Zhang, 2014). Without the appropriate IT infrastructure, big data cannot be sourced, stored and analysed (Zhang et al., 2022). In summary, organisations need to reconcile leadership capability and the staff’s technological expertise with technological resources to reap positive value-creation outcomes (Akhtar et al., 2020).

Turning to the connection between BDG and SLBDMCs, and mechanism of governance enable organisations to improve big data quality and generate the right conditions for data processing (Janssen et al., 2017; Yang et al., 2021). CG mechanisms upgrade the connectivity, harmony and integration between big data exchange partners (Wamba et al., 2017) and strengthen knowledge transfers through formally binding contracts and agreements (Wang and Jiang, 2022). Similarly, RG mechanisms provide social capital (e.g. mutual trust and reciprocity) that encourages knowledge sharing and big data processing and analysis (Yang et al., 2021; Aben et al., 2021; Janssen et al., 2017).

Theoretically, CGs and RGs’ impact on SLBDMCs (leadership capability, talent management and technology management) can be justified using social capital theory and the knowledge-based dynamic capability view. In the case of CG mechanisms, when parties are bound by formal contracts, they are aware of their rights and obligations and are less likely to engage in opportunistic and broader behaviours that amount to contract violation and attract legal ramifications (Wang et al., 2020; Yang et al., 2021). As such, CG mechanisms provide a sense of security between big data exchange partners regarding their data-sharing relationships, thereby creating a platform for organisations to better leverage their dynamic capabilities (i.e. SLBDMCs). For RG mechanisms underpinned by social capital theory, trust acts as a self-enforcing safeguard that discourages partners from behaving opportunistically to exploit the exchange relationship, while relational norms restrain opportunism through developing shared norms and values that guide reciprocal exchange and cooperative behaviour (Yang et al., 2021). In this way, RG mechanisms equally set the scene for firms to make optimum use of their strategic dynamic capabilities (Gao and Sarwar, 2024) for knowledge creation purposes.

In sum, CG and RG mechanisms are closely linked to big data collection and preparation (i.e. cleaning up, arranging and gathering data for investigating and decision-making) Shamim et al. (2019), which are important components of SLBDMCs. The argument here is that high-quality big data (accessed through BDG mechanisms) provides leaders with the impetus to make sound decisions regarding environmental changes, talent management strategies and technological resources. Taken together, strong governance mechanisms facilitate SLBDMCs. Based on this discussion, the given below the proposed hypotheses:

As indicated in section 1, OLBDMCs consist of big data democratisation, contextualisation, experimentation and execution (Zeng and Glaister, 2018). Democratisation refers to an organisation’s capability to transform big data into a more understandable language for employees involved in brainstorming activities. It is the management skillset that enables different organisational departments to collectively use data applications for data analysis and effective decision-making (Zeng and Khan, 2019). This requires cross-departmental collaborations where data scientists and non-data professionals work together from multidisciplinary perspectives. Contextualisation explains the organisation’s ability to make sense of the multiple data streams and build a holistic understanding of organisational operations and customer demands. It implies assigning meaning to data. Such meaningful analytics contribute to value creation and positively impact organisational performance (Zeng and Glaister, 2018). This is particularly important if organisations are to generate meaningful insights and knowledge. Experimentation entails employees’ ability to experiment with data and build scenarios. In cultures of learning and experimentation, employees are empowered to test new ideas and take calculated risks without fear of being reprimanded if they fail. In this way, these types of cultures promote creativity and innovation. Execution is the organisational capacity to convert data-generated understanding into actionable activities that generate value (Shamim et al., 2021; Zeng and Glaister, 2018).

In terms of the BDG–OLBDMC link, we argue that BDG supports OLBDMCs in that data available through CG and RG mechanisms can support leadership decision-making through resource utilisation (Patel, 2022). From a theoretical perspective, social capital theory and the knowledge-based dynamic capability view can explain this relationship. Building on earlier discussions (section 2.2.1), RG mechanisms (through emphasis on trustworthy and reciprocal relationships as per social capital theory) and CG mechanisms (through the legal enforcement of contractual arrangements between big data exchange partners) facilitate dynamic capability building in terms of data democratisation, contextualisation, experimentation and execution. For example, well-designed contractual CG frameworks with clear data sharing protocols may promote broader big data access and experimentation by promoting a shared understanding of what is allowed, thereby reducing legal and compliance risks associated with uncertainty. CG can also pave the way for contextually rich data, especially when contractual obligations exist for quality data. In contrast, trust-based relationships stemming from RG potentially reduce data hoarding behaviours and resistance to data sharing within and across organisations. RG supports trusting and informal networks that arguably provide users with tacit knowledge to better understand the data in business contexts, thereby driving data contextualisation (Zhang and Li, 2022; Aben et al., 2021; Shamim et al., 2020b). Following this discussion, the following hypotheses are suggested:

Empirical evidence from Shamim et al., (2021) suggests that BDMCs (both SLBDMCs and OLBDMCs) are positively associated with KCBD. This is because KCBD requires SLBDMCs to add value from big data as well as OLBDMCs to execute strategies through operational competencies Shamim et al., (2021). The relationship between BDMC and KCBD is informed by the knowledge-based dynamic capability view. As indicated by the knowledge-based dynamic capability view (Mikalef et al., 2018; Zheng et al., 2011), knowledge is essential for value creation, achieving an advantage through market competition (Grant, 1996) and developing dynamic capabilities. A significant body of work highlights the role of KCBD for value creation, performance, innovation and competitive advantage (Li et al., 2018; Busulwa et al., 2022; Lyu et al., 2022). In the hospitality sector, for example, KCBD helps organisations track guest-related data (e.g. customer loyalty, preferences, satisfaction and overall experiences), which can, in turn, be used to notify business strategy, decisions and performance (Busulwa et al., 2022; Li et al., 2018; Zarezadeh et al., 2022). To reap those benefits, however, BDMCs at the strategic and operational levels are critical dynamic capabilities for creating value from big data (Shamim et al., (2021); Horng et al., 2022). As noted by Zeng and Glaister (2018), big data value creation depends on what organisations do with the data rather than data access. Based on this discussion, the following hypotheses are proposed:

A firm’s culture consists of a set of beliefs, values, assumptions and symbols, which guide how the firm undertakes its strategic and operational tasks (Chatterjee et al., 2024). In big data contexts, the success of technology initiatives depends on organisation-wide cultural adoption Orero-Blat et al., (2025). As such, DDC acts as a cultural catalyser that enhances learning, communication, and information-sharing and supports the successful transformation of new ideas and processes. From a knowledge-based dynamic capability perspective, DCC influences both the dynamic capabilities and organisational outcomes of a firm (Karaboga et al., 2023). While BDMCs help firms to plan, coordinate and make decisions related to big data initiatives, these initiatives require a cultural adaptation to data-driven working habits to generate valuable knowledge (McAfee et al., 2012; Karaboga et al., 2023). This is why scholars attribute the failures of big data projects to the lack of a supportive organisational culture rather than a lack of technological infrastructure (Ghasemaghaei and Calic, 2020).

Moreover, in volatile market conditions, descriptive analysis is insufficient for decision-making. Instead, managers need predictive and even prescriptive analysis through data analytics to understand and forecast market situations for firm performance (Barlette and Baillette, 2022). Within this context, DDC enhances organisational performance by facilitating reliable and accurate decision-making processes and creating value-added services (Chatterjee et al., 2024; Vidgen et al., 2017). What this all means is that, whilst BDMCs support KCBD, this relationship is strengthened via DDC. For organisations to skilfully extract value from big data through their management capabilities, they need DDC (Shamim et al., 2021), where the predominant leadership cognitive models are centred on evidence-based and analytical decisions. In developing-economy contexts, such as Pakistan’s hotel industry, constraints, ranging from resource scarcity and institutional fragility to digital infrastructure gaps, create unique boundary conditions for BDG and capability building, thereby reinforcing the relevance of both strategic and operational-level mechanisms in the current study. Specifically, resource scarcity encompassing knowledge, finance, data and infrastructure necessitates adaptable governance structures and relational contracts, underscoring how contractual and relational governance mechanisms must compensate for institutional limitations, echoing that firms must innovate governance to navigate these contextual constraints (Leão et al., 2023). Furthermore, digital infrastructure deficits and heterogeneous adoption levels across firm sizes in developing countries heighten the importance of operational-level capabilities such as democratisation, contextualisation, experimentation and execution, to ensure that insights from big data can be effectively harnessed despite infrastructural unevenness (Torracca et al., 2023).

Moreover, the imperative to develop absorptive capacities, i.e. spanning technological, human, infrastructural and policy dimensions, is especially pressing in low- and middle-income economies and aligns with the model’s assertion that strategic-level big data capabilities facilitate long-term coordination and knowledge creation under constrained conditions (Khan, 2022). In addition, as developing hotels often interact with global value chains or foreign tour operators, these relationships can spur technological capability upgrading even under local constraints, demonstrating how governance mechanisms and strategically oriented BDMCs can enable firms to leapfrog and engage in knowledge creation through external stimuli (Romero et al., 2024). Therefore, these contextual features reinforce the study’s theoretical proposition: in developing economies, the interplay between governance (contractual and relational), strategic-level BDMCs, operational BDMCs and DDC crucially determines whether knowledge creation from big data can emerge and translate into innovation, despite systemic institutional, infrastructural and resource-related barriers. Thus, DDC moderates relationships differently in developing and developed economies. The following hypotheses can be proposed:

Data acquired through RC and CG mechanisms (i.e. BDG) requires robust management capabilities to interpret big data, make well-informed decisions (Patel, 2022), and create knowledge that adds value to products and processes (Mikalef et al., 2019). Existing literature suggests that strategic-level leadership capabilities can contribute to firms’ competitive advantage when leaders draw on their resources (e.g. connections and partnerships) and influence teams to align with the organisation’s vision and big data objectives (Mahdi and Nassar, 2021). Similarly, operational-level capabilities support organisations to translate big data (from BDG sources) into practical insights and knowledge (Zeng and Glaister, 2018). The mediating role of operational-level capabilities has been evidenced by Mikalef et al., (2020b). That study found that big data analytics capability can lead to a competitive advantage, provided that operational dynamic capabilities (namely, marketing and technological capabilities) exist.

Theoretically speaking and building on earlier arguments, the mediating role of BDMCs at strategic and operational levels can be explained using the knowledge-based dynamic capability view. While RG (underpinned by trust and reciprocity) and CG (bounded by contractual obligations) provide the sensing mechanisms (identifying data opportunities) and structural foundations for organisations to detect and access data (Yang et al., 2021), BDMCs represent the seizing and reconfiguring processes that actually transform data into knowledge (Shamim et al., 2021; Horng et al., 2022). In other words, BDMCs orchestrate the technical and social processes needed to extract insights from data, combining analytical tools with organisational knowledge and networks. This aligns with scholars’ arguments that BDG policies are not sufficient for knowledge creation – organisations need to develop the dynamic capabilities to actually execute on that governance through coordinated management practices that leverage both technical systems and social capital (Zeng and Glaister, 2018; Gao and Sarwar, 2024). Based on this discussion, the following hypotheses are proposed:

In summary, the hypotheses developed emerge from dialectical engagement with the framework. Following this literature review, we proceed with an overview of the research methodology. The following Figure 1 represents the conceptual framework of the study.

Highest number of studies on BDG–KCBD have been examined in advanced economies, leaving a significant gap in comprehending the dynamics of organisations operating in growing economies (e.g. Pakistan). While, research in developed economies has yielded significant insights (Ghasemaghaei and Calic, 2020), the relevance of these outcomes to developing economies is usually limited. This limitation is aggravated by the absence of empirical studies explicitly targeting developing economies, resulting in an inadequate comprehension of the factors affecting KCBD in these settings. Given the unique structural challenges faced by developing economies – i.e. lower IT adoption, resource constraints and exorbitant costs, resistance to change, lower digital literacy and traditional decision-making cultures that rely on experience rather than analytics (Lee et al., 2024) – findings from developed economies cannot be applied to developing economies. Besides decision-making, big data is also used to guide product innovation (Tan and Zhan, 2017). As per the Global Innovation Index (WIPO, 2024), developed countries like Switzerland, Sweden and the USA are the top three innovative countries, and Pakistan, a developing country, ranks at 91. Against this backdrop, this study contributes to broadening the theoretical understanding of BDG in developing markets by integrating the two key dimensions of strategic and operational BDMC as an underlying mechanism to explain KCBD in a neglected hotel sector.

Although Pakistan currently ranks 91st on the Global Innovation Index (WIPO, 2024), its hotel industry serves as a compelling microcosm for examining BDG and digital capability development in developing economies. Unlike advanced economies, institutional frameworks in Pakistan, characterised by protracted regulatory procedures and uneven digital infrastructure, shape how hotels adopt and benefit from digital systems. For example, lengthy licencing and approvals significantly hinder innovation uptake, while large rural–urban disparities in internet access limit uniform digital capability deployment (IIPS, 2021; IIPS, 2022). Moreover, the hotel sector in Pakistan largely consists of informal enterprises and small-scale operations that operate under resource constraints, yet these entities hold substantial potential for innovation if supported by better infrastructure and governance (Yaseen, 2025).

Technologically, the broader economy faces barriers such as inadequate ICT infrastructure, low awareness of digital opportunities and fragmented digital processes, all of which directly impact the hospitality domain’s ability to govern and leverage big data. Nonetheless, a study in comparable developing contexts such as Gilgit-Baltistan shows that when entrepreneurial ecosystems, cultural identity and digital innovation are aligned, even amidst infrastructural constraints, SMEs can foster resilient, digitally-enabled service models (Jan et al., 2025).

Culturally, norms around change acceptance, risk aversion and organisational support influence how hotels in Pakistan embrace digital transformation. Achieving ambidexterity through a knowledge-sharing culture and support mechanisms enhances service innovation, even within emerging markets (Hadi and Sheikh, 2024). Thus, these factors suggest that while Pakistan’s ranking signals innovation deficits, its hotel sector’s challenges, rooted in regulation, infrastructure gaps and cultural dynamics, mirror those in many developing economies.

Hence, insights drawn from Pakistan’s hotel industry can be meaningfully generalised to other contexts experiencing similar constraints. The interplay of governance, digital capability and cultural readiness observed here is likely to resonate across settings where institutional fragility, limited ICT access and informal entrepreneurial structures define the innovation ecosystem.

The survey questionnaire was conducted to collect data from four-star and five-star hotels in Pakistan using convenience sampling due to practical constraints and accessibility considerations such as high job demands, variable work shifts, and limited access due to operational and managerial protocols (Etikan et al., 2016). Consistent with the literature, the use of convenience sampling facilitated timely data collection from a sizable cohort of employees without breaching organisational boundaries or interfering with their day-to-day operations (Sarwar and Muhammad, 2020). Survey questionnaires were distributed at two different points in time, with a one-month time-lag. At Time 1, hotel managers were asked to complete the demographic information section along with the survey questionnaire, measuring the following variables: BDG and BDMCs (both at the strategic and operational levels). Notably, 250 out of 318 questionnaires were returned. At Time 2, the same managers completed the questionnaire, measuring the variables of DDC and KCBD. Forty-eight out of 250 questionnaires proved invalid. Thus, a total of 202 questionnaires were deemed valid for statistical analysis, resulting in a response rate of 63.5%. A priori power analysis was conducted using G*Power 3.1 to determine the required sample size. With an alpha level of 0.05, a power of 0.80 and a medium effect size (f² = 0.15), the minimum required sample size for a model with three predictors was n = 77. The final sample of 202 respondents exceeded this threshold, ensuring adequate statistical power to detect meaningful effects. To assess non-response bias, we compared respondents and non-respondents. A chi-square test revealed no significant difference in gender distribution between the two groups, χ²(1) = 0.54, p = 0.46. Similarly, independent-samples t-tests showed no significant differences in demographic variables and key study constructs (all p > 0.05), suggesting that non-response bias was minimal. The demographic information of the respondents is presented in Table 1. The largest cohort of participants identified themselves as male (F = 138; p = 68.3%), within the age bracket of 26–35 years (F = 98; p = 48.5%), having completed a master’s degree (F = 78; p = 38.6%) and with 5–10 years of work experience (F = 46; p = 22.8%). The details can be seen in the following table:

The questionnaire uses the items and constructs adopted from previously published work from well-established researchers. The survey consisted of a closed-ended questionnaire with a 5-point Likert scale (with number 1 denoting a “strongly disagree” response and number 5 indicating a “strongly agree” response). The 53 items used in the questionnaire were classified into five sections (from A to E). Section A of the questionnaire measured BDG, consisting of 4 items of CG adopted from (Shamim et al., 2020a) and 3 items of RG adopted from (Poppo and Zenger, 2002). Section B measured SLBDMCs consisting of 14 items adopted from Shamim et al., (2019). Strategic-level management capabilities were measured using three dimensions adapted from Shamim et al., (2019): (a) leadership; (b) talent management; and (c) technology management. Leadership was measured with 6 items, talent management with 4 items and technology management with 4 items.

Section C consisted of 22 items to measure OLBDMCs, adopted from Shamim et al., (2019). Operational-level management capabilities were measured using four dimensions adopted from Shamim et al., (2019): (a) data democratisation (6 items); (b) data contextualisation (5 items); (c) data experimentation (6 items); and (d) data execution (5 items). Section D measured DDC using 5 items adopted from Shamim et al., (2019). Section E included 5 items to measure KCBD, adopted from Shamim et al., (2021). A rigorous translation and back-translation procedure (Brislin, 1986) was employed to ensure linguistic and conceptual equivalence for the Pakistani context. Three subject matter experts verified the accuracy of the translated items. Content validity of the questionnaire was established through a pre-testing phase. Two experienced hotel managers and three discipline expert academics were involved in this phase. Based on their feedback, the questionnaire was slightly modified for better clarity. The following section presents the study results.

The data were analysed using PLS-SEM from Smart PLS 4, which is considered an advanced estimation technique for the hotel industry (Ali et al., 2018). PLS-SEM is considered a practical tool, as the main purpose of the study was to explain and estimate the constructs (Hair et al., 2021). Compared to other statistical estimation methods, PLS-SEM has fewer data processing requirements, particularly in terms of sample size, and does not require normally distributed data (Hair et al., 2021). Moreover, PLS-SEM is a simple and effective technique for model-building (Ringle et al., 2005). It also provides bootstrapping to calculate item loadings, the direct relationships among constructs and significance levels using the PLS algorithm. PLS-SEM also assumes the absence of multicollinearity among indicators. Data were analysed in two stages: first, the measurement model was run to evaluate reliability and validity of data; second, the structural model was run to evaluate the proposed associations.

PLS- SEM does not need normally curved data for processing, but the assessment of data, whether it is normally distributed or not, is very crucial before the application of inferential statistics (Hair et al., 2007). The data distribution was assessed with the help of skewness, kurtosis analysis and histogram plots as per guidance given by previous researchers (Munro, 2005). The skewness and kurtosis values ranged between acceptable ±two values. This showed that all the constructs were normally distributed. As such, analysis was possible with the help of PLS-SEM.

According to Podsakoff et al. (2024), collecting data for independent variable and criterion variables from distinct sources and introducing a time gap between their measurements significantly reduce common method bias (CMB). In our study, the unit of analysis is individual respondents, which can potentially cause CMB. Therefore, a full collinearity test was performed to investigate CMB in SEM. As per Kock (2015), we used the variance inflation factor (VIF) to point out CMB through a full collinearity test. VIF figures higher than 3.3 indicate the presence of CMB, and lower than 3.3 statistic values show that the model is without CMB. The CMB values in the present study were lower than 3.3, showing that CMB does not exist. In addition, a principal component analysis based on Harman’s (1976) one-factor criterion was employed. The results indicated that the largest factor explained just 26.13% of the variance, well below the critical 50% threshold, affirming that CMB was not a significant concern.

The results are derived for the present from two critical analyses:

1. The PLS algorithm, used to verify convergent and discriminant validity; as well as.

2. bootstrapping, conducted to determine the structural model. Convergent validity and discriminant validity were assessed to evaluate the measurement model. Convergent validity examines whether all items effectively capture the same underlying construct.

This was tested using average variance extracted (AVE), composite reliability (CR) and factor loadings. The analysis illustrated that all factor loadings for the constructs were greater than the recommended threshold of 0.50, except for one item from data experimentation with a loading < 0.50, which was dropped from the analysis (see Figure 2 and Table 2). We removed Exp4 because it exhibited a low standardised factor loading in the measurement model, indicating insufficient shared variance with its intended construct. Consistent with established guidance, items with loadings below commonly accepted thresholds (e.g. < 0.50–0.60) are typically deleted, and items in the 0.40–0.70 range may be removed when doing so improves reliability and convergent validity without sacrificing content coverage (Hair, 2009; Kline, 2023; Fornell and Larcker, 1981). Following a pre-specified decision rule, we (i) examined loadings and cross-loadings, (ii) reassessed composite reliability and AVE after deletion and (iii) ensured that sufficient, conceptually distinct indicators remain to preserve the construct’s breadth. The revised model shows improved measurement quality (higher CR/AVE and marginally better fit indices), and our substantive conclusions are unchanged. For transparency, we report Exp4’s loading and the before/after reliability and validity indices in the revision. Removal of the item improved model fit indices and increased the construct’s reliability (CR) and convergent validity (AVE), without compromising its conceptual integrity (Kline, 2023). This approach aligns with standard practices recommended in scale refinement literature (e.g. Hair, 2009; Kline, 2023). Cronbach’s alpha values ranged from 0.783 to 0.933, surpassing the acceptable threshold of 0.70. In addition, composite reliability values ranged from 0.849 to 0.940, also exceeding the minimum threshold of 0.70. The AVE values for all constructs were above the recommended benchmark of 0.50, ranging from 0.532 to 0.715, further confirming the model’s reliability and convergent validity (Hair et al., 2007).

To ensure the constructs were distinct, discriminant validity was assessed using the heterotrait–monotrait (HTMT) ratio, which is regarded as a more robust and reliable method (Henseler et al., 2015). As shown in Table 3, all HTMT values were below the recommended threshold of 0.90, confirming that discriminant validity was established for all constructs (see Table 3). To further verify discriminant validity using the Fornell and Larcker criterion (1981), we compared the square roots of the AVEs with the corresponding inter-construct correlations, ensuring that each square root is greater than its associated correlations, thereby confirming discriminant validity through this method as well. All the constructs exhibit positive associations, specifying that the variables proceed in the same direction, either increasing or decreasing together. The details can be seen in the following table:

The following table presents the HTMT ratios: The following figure illustrates the measurement model assessment.

The results of the direct, indirect and moderating hypotheses, as presented in Table 4, were evaluated based on the critical value (t > 1.96; p < 0.05). The findings indicate that relational governance (RG) had a positive and significant relationship with strategic-level big data management capabilities (SLBDMC) (β = 0.376, t = 5.254, LL = 0.255, UL = 0.489), thus supporting H1. Similarly, contractual governance (CG) was positively and significantly associated with SLBDMC (β = 0.455, t = 6.549, LL = 0.337, UL = 0.565), supporting H2. Furthermore, RG had a significant positive impact on operational-level big data management capabilities (OLBDMC) (β = 0.275, t = 3.771, LL = 0.153, UL = 0.392), supporting H3. Likewise, CG demonstrated a strong positive relationship with OLBDMC (β = 0.519, t = 7.204, LL = 0.396, UL = 0.631), supporting H4. However, the connection between SLBDMC and knowledge creation through big data (KCBD) was positive but not statistically significant (β = 0.086, t = 0.987, p = 0.162, LL = −0.056, UL = 0.228), and therefore H5 was not supported. In contrast, OLBDMC had a significant and positive relationship with KCBD (β = 0.403, t = 4.479, LL = 0.264, UL = 0.558), supporting H6.

The moderating effect of DDC on the association between SLBDMC and KCBD was found insignificant (β = −0.003, t = 0.040, p = 0.484, LL = −0.134, UL = 0.133), leading to the rejection of H7. However, DDC significantly moderated the relationship between OLBDMC and KCBD in a positive direction (β = 0.154, t = 2.090, p = 0.018, LL = 0.143, UL = 0.167), thus supporting H8. The mediation analysis showed that SLBDMC mediated the relationship between RG and KCBD (β = 0.187, t = 3.254, LL = 0.111, UL = 0.269), and between CG and KCBD (β = 0.239, t = 3.597, LL = 0.137, UL = 0.343), hence H9 and H10 were supported. Similarly, OLBDMC significantly mediated the connection between RG and KCBD (β = 0.111, t = 2.661, LL = 0.053, UL = 0.192), and between CG and KCBD (β = 0.209, t = 3.800, LL = 0.131, UL = 0.309), supporting both H11 and H12. To assess the robustness of the structural model, bootstrapping was conducted using different re-sample sizes, ranging from 1,000 to 10,000, as recommended by Hayat et al. (2025). The results remained consistent across these re-samples, with stable path coefficients and significance levels, thereby affirming the reliability and robustness of the model.

The explanatory power of the model was gauged using R², Q² and f². The findings yielded R² values of 0.591 for SLBDMC, 0.547 for OLBDMC and 0.569 for KCBD. Predictive capability of the model was assessed by running a blindfolding test. The Q² values for all endogenous constructs were greater than 0, supporting the model’s predictive capability. The effect sizes (f²) ranged from weak to robust, with values between 0.004 and 0.386. The outcomes substantiate the model’s sufficient explanatory power. The following table presents the path analysis

The following Figure 3 illustrates the structural model assessment:

Model fit was assessed using SRMR and NFI, with threshold values of below 0.08 and above 0.90, respectively. The outcomes yielded an SRMR value of 0.042 and an NFI value of 0.957, thereby supporting the model’s fit. In addition, non-significant d_ULS and d_G statistics (d_ULS = 0.44; d_G = 0.19) suggest that the model exhibits strong global fit and robustness (Hair et al., 2021).

Out-of-sample prediction, as recommended by Shmueli et al. (2019), is used to further assess the predictive power of PLS-SEM models. A model is considered to have high predictive power if the Q² predict values are greater than zero and the RMSE values are lower than those of a linear model (LM-RMSE). The results indicated that the Q² predict values were greater than zero, and the PLS-SEM model yielded lower RMSE values compared to the LM-RMSE, thereby confirming the high predictive accuracy of our model. The details can be seen in the following Table 5.

The following Figure 4 illustrates the moderation of DDC on the relationship between OLBDMC and KCBD.

It shows that DDC strengthens the positive association between OLBDMC and KCBD.

The study explored the influence of BDG and management capabilities (at the strategic and operational levels) on KCBD under the moderating effect of DDC. The study results (H1–H4) show that CG and RG mechanisms influence strategic-level and operational-level management skills and capabilities. While BDG research mainly concentrated on the functions of CG and RG in developing broader big data analytics capabilities (Janssen et al., 2017; Shamim et al., 2020b), our findings break new ground by linking BDG to BDMCs. These findings can be justified by the fact that RG mechanisms have the potential to nurture trusting relationships between hotels (and their workforce) and big data exchange partners. Similarly, CG mechanisms provide managers and operational-level staff with quality customer-related data which can be used to inform product and service customisation and strategic decision-making processes (Janssen et al., 2017; Zeng and Khan, 2019).

We theorised that SLBDMC have a positive and significant effect on KCBD. However, contrary to the findings of Shamim et al., (2020a), our results demonstrated no statistically significant impact of SLBDMC on KCBD; thus, H5 was not endorsed. The non-significance of H5 might be due to several contextual and organisational factors unique to developing economies such as Pakistan. Hotels in Pakistan operate in a low-tech environment, with limited investment in advanced analytics and weaker digital maturity compared to counterparts in developed markets (Lee et al., 2024). Hotel managers in developing economies often prioritise short-term operational survival and cost control over long-term data-driven planning, given the resource constraints and competitive pressures they face (Busulwa et al., 2022; Zaman et al., 2022). This contextual emphasis on immediate performance outcomes might explain why strategic-level capabilities do not directly translate into knowledge creation through big data.

Another plausible explanation relates to workforce readiness and cultural barriers. Employees in the Pakistani hotel sector frequently demonstrate low levels of digital literacy and a reliance on traditional, experience-based decision-making (Asghar et al., 2021). These findings align with arguments in the dynamic capabilities literature, which suggest that strategic-level capabilities must be complemented by leadership cognition, entrepreneurial orientation and integrative foresight mechanisms to yield meaningful knowledge outcomes (Teece, 2007; Horng et al., 2022). In the absence of such complementary mechanisms, strategic capabilities may remain underutilised, leading to their non-significant role in knowledge creation in this study.

Conversely, H6 supported the proposition that OLBDMC has a positive effect on KCBD, aligning with Colombari and Neirotti (2023), who noted that operational-level capabilities enhance decision-making and knowledge procedures. However, H7 was unsupported, suggesting that DDC does not interact with SLBDMCs to influence KCBD. This might be the case because the industry has not uniformly adopted a DDC, and the overall culture might still be developing, therefore not offering the moderating effect between SLBDMC and KCBD. It may be the case that developing economies (e.g. Pakistan) have limited access to technological infrastructure and resources, a lack of leadership mindset or employee training (digital literacy) needed to embed DDC. Moreover, the hospitality sector is a low-tech industry with limited adoption of BDG, as it primarily focuses on maintaining high service delivery standards. DDC might already be a part of governance and capabilities, which limits the role of DDC as a moderator in the given models. Chatterjee et al., (2024) also asserted that, although DDC is crucial, its effectiveness may not necessarily be evident or varied in moderating functions.

Contrarily, it was also found that DDC significantly and positively moderates the association between operational-level big data management capabilities and knowledge creation through big data, thus supporting hypothesis H8. Similarly, H9 and H10 supported the partial mediating role of SLBDMC in the relationships between BDG (e.g. RG and CG) and KCBD. H11 and H12 supported that OLBDMC mediates the relationship between BDG and KCBD, implying that operational enablers are essential for the creation of organisational knowledge out of governance structures, as found in other works, for example, in a study by Mikalef et al., (2020a). By providing empirical support to these relationships, this study shows that for hotels to successfully leverage their RC and CG mechanisms and generate KCBD, they need to invest in operational-level staff capabilities. As highlighted by Zeng and Glaister (2018), organisations’ ability to generate value-creation opportunities from big data does not rely on the data itself or individual data scientists, but rather on the data management process.

Our study suggests that hotels can effectively process big data sourced from RC and CG if operational staff are supported to accurately interpret and contextualise big data and to experiment with new ideas and make mistakes. At the same time, however, within the context of high hospitality employee turnover in Pakistan (Asghar et al., 2021), such operational staff familiarity with big data may be challenging. Differences between five and four-star hotel management practices also suggest that hospitality employees in Pakistan are not homogeneously equipped with the same level of big data skills and expertise. While five-star hotels are reported to heavily invest in their talent attraction and retention strategies (e.g. through ongoing employee training and development, and strategic reward management systems), four-star hotels generally lag in terms of HRM practices (Asghar et al., 2021). Because four-star hospitality employees are less exposed to cutting-edge workplace technology and training, they may require additional technical assistance and training to translate big data into the workplace.

The findings show that DDC significantly moderated operational-level capabilities but not strategic-level ones, indicating that cultural readiness exerts greater influence on day-to-day processes like experimentation and execution rather than on higher-level decision-making. At the strategic level, cultural emphasis alone appears insufficient to generate knowledge without complementary factors such as leadership foresight, absorptive capacity and organisational agility (Karaboga et al., 2023; Chatterjee et al., 2024). This suggests that culture is not a universal moderator but context-dependent, operating more effectively at the operational layer. Future research should therefore explore alternative moderators and mediators, such as digital literacy, employee readiness or organisational agility, and use longitudinal or comparative designs to capture the evolving influence of governance, culture and capabilities in developing economies (Mikalef et al., 2020a; Gao and Sarwar, 2024).

This study makes several important theoretical contributions by extending the application of the dynamic capability perspective to the context of developing economies and low-tech service industries. While dynamic capability theory (Teece, 2007) emphasises the role of sensing, seizing and reconfiguring in achieving competitive advantage, much of the empirical evidence has emerged from resource-rich, high-tech sectors in developed economies (Lyu et al., 2022; Huynh et al., 2023). By examining the Pakistani hotel industry, our findings demonstrate how dynamic capabilities operate under conditions of resource scarcity, lower digital maturity and cultural barriers. This contextual contribution advances understanding of how dynamic capabilities can be nurtured and leveraged in environments where strategic alignment is challenged by limited resources and institutional support (Gao and Sarwar, 2024).

A second contribution lies in differentiating between strategic-level and operational-level big data management capabilities. Prior research has often treated big data management capabilities as a monolithic construct (Wamba et al., 2017; Mao and Lu, 2023), thereby overlooking how different layers of capability interact. Our study empirically shows that operational-level capabilities (e.g. data contextualisation, experimentation and execution) have a stronger and more direct influence on knowledge creation than strategic-level capabilities. This indicates that, in low-tech and resource-constrained environments, operational-level competencies serve as the real drivers of organisational learning and innovation. These findings refine dynamic capability theory by illustrating that the effectiveness of different capability layers is contingent on contextual factors, with operational competencies proving especially critical in industries such as hospitality where immediate service delivery and customer-facing decisions dominate (Colombari and Neirotti, 2023).

Finally, the study contributes to ongoing debates about the role of organisational culture in digital transformation. While a DDC was expected to enhance the effect of both strategic and operational capabilities, our findings reveal its moderating role only at the operational level. This suggests that cultural support for data use does not universally benefit all organisational processes but is bounded by the nature of the task and the maturity of the organisation. Theoretically, this advances the understanding of boundary conditions in digital transformation, highlighting that DDC interacts differently with capability layers and may need to be complemented by leadership cognition, strategic foresight or entrepreneurial orientation at higher levels (Chatterjee et al., 2024; Karaboga et al., 2023). Together, these contributions refine and extend both the dynamic capability perspective and the emerging literature on digital capabilities in developing economies, underscoring the importance of context in theorising about data-driven transformation.

Our results show the importance of BDG in knowledge creation in the hotel industry in Pakistan. The positive influence of CG mechanisms highlights the need for formal agreements between hotels and big data vendors to promote access to quality data, without which relevant and necessary knowledge cannot be generated for new hotel services, products and processes. Similarly, considering the positive influence of RG mechanisms, trusting relationships are needed in the Pakistani hotel industry. This requires broader stakeholder discussions (e.g. hotel industry associations and experts, big data vendors and government) on how to formulate strategies and mechanisms for governance to promote a collaborative and safe data-exchange landscape in Pakistan.

By integrating BDG and BDMC, hotels can streamline their operations by optimising resource allocation (e.g. energy consumption, real-time tracking of inventory, staff scheduling) and minimising waste. This is particularly relevant to the low-tech hotel sector in Pakistan, facing financial constraints essential for sustainability. Above all, BDG enables hotels to make better-informed decisions and achieve a competitive edge over its rivals by understanding customer needs and offering unique and customised products by cutting costs and offering quality products at low prices compared to competitors. BDG facilitates hotels to evaluate guest preferences, allowing for the personalisation of services like dining options and room configurations, yielding an enhanced customer experience that, in turn, leads to customer retention and loyalty in a competitive business landscape.

A key managerial implication is the need for continuous investment in staff training and digital literacy. Hotel managers should design in-house workshops and targeted learning initiatives to strengthen employees’ ability to interpret and apply data in decision-making (McAfee et al., 2012; Busulwa et al., 2022). Beyond technical upskilling, managers must also balance operational data-driven practices with strategic foresight by embedding scenario planning and long-term vision into their digital initiatives (Teece, 2007). Cross-departmental collaboration and knowledge-sharing practices, such as data councils or inter-unit analytics teams, can further ensure that insights generated at the operational level are aligned with broader organisational goals (Zeng and Glaister, 2018).

Workforce readiness and organisational culture are equally critical. Many hotel employees in developing economies exhibit low digital confidence and resistance to technological change (Asghar et al., 2021). To address this, managers should adopt gradual change management strategies that reward digital innovation and experimentation while minimising fear of failure (Colombari and Neirotti, 2023). Retention strategies for IT-savvy employees, through career development opportunities, recognition systems and competitive compensation, are also essential for sustaining digital transformation in a sector with high staff turnover (Zaman et al., 2022).

Finally, broader policy interventions are required to support digital transformation at the industry level. Tourism ministries can subsidise digital training and technology adoption programs tailored for hospitality firms, thereby reducing the resource burden on individual hotels (Chatterjee et al., 2024). Regulators could provide tax incentives or certification frameworks for hotels that adopt robust BDG systems, encouraging compliance and innovation. In parallel, public–private partnerships aimed at enhancing digital infrastructure, such as high-speed internet access and shared data platforms, would enable hotels to leverage big data more effectively in resource-constrained settings (Karaboga et al., 2023). Together, these measures would create an enabling ecosystem for hotels to integrate big data into strategic and operational processes more successfully.

In hotels, DDC is especially valuable for operational capabilities; however, the lack of significance at the strategic level suggests that data culture alone cannot drive decisions such as market expansion, brand repositioning or adoption of disruptive technologies. These choices require leadership vision, competitor analysis and scenario planning that extend beyond operational data. Practically, hotel firms should take a layered approach: strengthen DDC for operational excellence, while at the strategic level adopting a data-informed model that combines analytics with foresight, innovation and managerial intuition. This balance ensures hotels capture value from data without overlooking the broader capabilities needed for long-term transformation.

This study concludes that BDG, through both contractual and relational mechanisms, plays a critical role in shaping big data management capabilities (BDMCs) in the hotel sector of Pakistan. However, the findings reveal that while BDG enhances both strategic-level and operational-level capabilities, only operational-level BDMCs consistently translate into knowledge creation through big data. This suggests that in resource-constrained, service-intensive environments such as hospitality, the daily routine of data contextualisation, experimentation and execution serve as the real drivers of organisational learning and innovation (Zeng and Glaister, 2018; Colombari and Neirotti, 2023). Moreover, cultural readiness, operationalised through DDC, is found to matter more at the operational level than the strategic level, reinforcing the argument that cultural influences in digital transformation are not universal but contingent upon organisational layers and maturity (Chatterjee et al., 2024; Karaboga et al., 2023).

Future research can extend these findings in several directions. First, longitudinal studies are needed to capture the evolution of digital maturity and the dynamic interplay of governance, culture and capabilities over time (Mikalef et al., 2020a). Second, cross-country comparative studies could contrast developing and developed economies to identify context-specific enablers and barriers to big data-driven knowledge creation (Lyu et al., 2022). Third, alternative methodological approaches, such as case studies, experiments or mixed-methods research, may provide deeper insights into how BDG and capabilities unfold in practice and allow for greater causal inference (Huynh et al., 2023). Finally, future research should investigate additional moderators and mediators such as employee digital literacy, absorptive capacity and leadership orientation, which may explain why strategic capabilities failed to directly influence knowledge creation in this study (Teece, 2007; Horng et al., 2022).

Despite its contributions, this study is not without limitations. The time-lagged survey design, although effective in mitigating common method bias, does not fully capture the dynamic nature of governance and capability development. Future studies adopting longitudinal or panel designs could provide a more comprehensive understanding of these evolving relationships. In addition, the reliance on self-reported data from hotel managers introduces potential bias, and future research could triangulate responses with archival data or observational methods to improve robustness. Finally, the relatively modest sample size, although statistically adequate, limits the generalisability of the findings. Expanding to larger and more diverse samples across different hotel categories and regions would strengthen external validity.

The authors would like to express their sincere gratitude to all respondents for their invaluable participation in the data collection of this research. Special thanks to and selected hotels for their support and encouragement.

The authors are grateful to the anonymous reviewers and editor for their insightful comments and suggestions. The research was supported by the 2024 Humanities and Social Sciences Research Project of the Ministry of Education (Grant number 24YJA790018); 2023 Natural Science Foundation of Fujian Province (Grant number 2023J011134); 2023 Social Science Fund of Fuzhou city (Grant numbers: 2023FZB88); 2024 Fujian Provincial Social Science Program (Grant number JAS23177); 2024 Fujian Province Fujian Association for science and technology innovation think tank Research Program (Grant number FJKX2024XKB040); 2024 Social Science Fund of Fujian Province (Grant numbers: FJ2024BF009); 2025 Natural Science Foundation of Fujian Province (Grant number 2025J08258).

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.