Within research on information systems (IS), data governance is often theorised as a structured institutional framework that enables organisations to manage data assets effectively and support analytics capabilities (Bena et al. 2024). Several widely adopted frameworks reflect this perspective, including the Data Management Body of Knowledge of the Data Management Association (DAMA) (DAMA International 2017), the Big Data Analytics Capability (BDAC) model (Wamba et al. 2017), Control Objectives for Information and Related Technologies (COBIT) governance frameworks (Information Systems Audit and Control Association [ISACA] 2018) and the Data Management Capability Assessment Model (DCAM) (Enterprise Data Management [EDM] Council Inc 2020). These models emphasise the need for clearly defined roles, policies, standards and decision rights that coordinate the management and use of data within organisations.

Many of these frameworks implicitly assume a linear progression in which governance maturity precedes the development of analytics capabilities. For example, capability-based models conceptualise governance as an organisational antecedent that enables data quality management, integration and analytical insight generation (Abraham et al. 2019).

Similarly, hybrid governance approaches that combine centralised oversight with decentralised flexibility aim to support analytics adoption while maintaining coherence across organisational units (Fadler & Legner 2021). Maturity models used in public sector contexts further reinforce this sequential logic by linking analytics capability to predefined stages of organisational readiness (Okuyucu & Yavuz 2020).

While these frameworks provide important guidance for governance design, they often assume conditions of institutional stability, organisational capacity and resource availability that are characteristics of developed economies but not necessarily present in developing contexts (Alhassan et al. 2019). As a result, the linear relationship between governance and analytics capability proposed in many models may not fully capture the realities of organisations operating under institutional constraints.

An alternative perspective within research of information systems emphasises the emergent nature of organisational structures. Scholars such as Orlikowski (2000) and Leonardi (2011) argue that technologies and governance mechanisms are enacted through situated practices rather than solely through formal institutional design. From this perspective, governance may evolve through everyday organisational routines, improvisation and informal coordination mechanisms. Otto (2011) illustrates this dynamic by showing how organisations negotiate governance responsibilities informally when formal structures are weak or ambiguous.

Research on data governance in the public sector highlights complexities that influence its implementation. Public sector organisations operate within institutional environments characterised by legal mandates, political oversight and multilayered accountability structures (Pollitt & Bouckaert 2017). These conditions often complicate efforts to develop coherent governance arrangements.

Studies identify several structural and cultural challenges that hinder governance institutionalisation (Alhassan et al. 2019; Allard et al. 2018; Asensio et al. 2025). Organisational silos limit data sharing across departments, while legacy systems and fragmented information infrastructures reduce interoperability (Janssen et al. 2017). In some cases, responsibility for data governance is subsumed under information and communication technologies (ICTs) governance structures, creating ambiguity over ownership and stewardship responsibilities (Smith 2019). Business units may expect ICT departments to manage data assets, while ICT professionals lack the contextual knowledge required to ensure data quality and usability (Almeida et al. 2022).

Political and organisational dynamics further complicate governance implementation. Bureaucratic decision-making processes often move more slowly than technological developments, delaying the creation of governance frameworks (Goshovska et al. 2021). Overlapping mandates and shifting administrative responsibilities can also generate uncertainty around data roles and accountability structures (Thompson, Ravindran & Nicosia 2015). In developing countries, donor-funded digital platforms and vendor-driven system architectures may inadvertently become de facto governance mechanisms in the absence of locally institutionalised frameworks (World Bank 2021).

These challenges contribute to situations where governance structures remain incomplete even as digital systems are introduced into government operations. Mudzunga et al. (2025) describe this condition as one of continuous institutional evolution, in which agencies operate through provisional arrangements and workarounds because formal governance structures cannot be established quickly enough to keep pace with technological change.

Big data is commonly characterised by the three foundational dimensions proposed by Laney (2001): volume, variety and velocity, often referred to as the ‘3Vs’ (Kitchin & McArdle 2016). Khine and Shun (2017) define big data analytics in the public sector as the analysis of high-volume, high-velocity and heterogeneous data from multiple sources to generate reliable insights and create organisational value for improved public service delivery and decision-making.

Public sector organisations generate large volumes of administrative, demographic and transactional data that have the potential to support improved policy design and service delivery (World Bank 2017). When effectively leveraged, big data analytics can enhance decision-making in areas such as taxation, healthcare, urban planning, labour markets and emergency response (Fredriksson et al. 2017). Analytics systems can also support transparency and citizen trust by enabling evidence-informed policymaking and proactive service delivery (Suominen & Hajikhani 2021).

Research on big data analytics in government has examined several dimensions, including adoption (Langseth & Haddara 2024; Wiafe et al. 2020), organisational capability (Brinch, Gunasekaran & Wamba 2021; Marfo 2017), readiness (Seko et al. 2022), performance outcomes (Faridoon, Liu & Spence 2023; Mikalef et al. 2019) and ethical implications (Ghasemaghaei & Turel 2026; Richardson, Petter & Carter 2021). However, many of these studies implicitly assume that governance structures exist to support analytics deployment.

In Kenya, platforms such as eCitizen and Huduma Kenya illustrate the growing digitalisation of public services. While national legislation such as the Data Protection Act (Republic of Kenya 2019) and the Computer Misuse and Cybercrimes Act (Republic of Kenya 2018) provide an overarching regulatory framework, many agencies lack operational governance mechanisms such as data stewardship roles, standardised data management procedures or cross-agency coordination structures (Seko et al. 2022). As a result, analytics initiatives often evolve within institutional environments characterised by fragmented governance arrangements.

This literature is an indication that data governance is treated as a foundational prerequisite for analytics capability. Formal governance frameworks and maturity models are emphasised, while public sector scholarship highlights institutional and political constraints that complicate governance implementation. At the same time, studies on big data analytics frequently assume governance readiness despite empirical evidence that governance structures may remain incomplete or fragmented in many public organisations.

Addressing this gap requires shifting attention from governance as a formal structural prerequisite towards examining how governance is enacted in practice during the operationalisation of data analytics systems.

The study focuses on autonomous state agencies in Kenya, which operate under high accountability pressures, fragmented governance structures and varying levels of analytics sophistication. These agencies manage heterogeneous, high-volume, high-velocity and high-variety data, while facing performance demands and constrained resources. This setting provides a context to examine how big data analytics are deployed when formal governance mechanisms are weak.

This study adopts a qualitative multiple case study design to investigate big data analytics use in state agencies in Kenya. Grounded theory coding procedures were applied to inductively identify patterns of governance emerging from the data. This approach is appropriate for exploring complex, practice-based phenomena and understanding how governance emerges in environments where formal structures are weak or absent (Eisenhardt 1989). The research followed an inductive paradigm, focusing on data governance as enacted through decisions and practices of organisational actors.

A multiple case study approach was adopted, involving eight state agencies selected for their deployment of big data analytics systems. Eight cases of a population of approximately 66 active commercial government entities were considered sufficient to enable theoretical replication while maintaining analytical depth typical of qualitative case study research (Yin 2017). Agencies were selected purposively to capture variations in governance and analytics sophistication. This allowed theoretical replication and identification of recurring governance-in-use patterns across different organisational settings.

Additionally, purposive sampling was used to identify departments with engagement in analytics. Theoretical sampling followed, whereby participant selection was guided iteratively by emerging concepts and categories that required further probing from previous interviews. Selection of successive participants was therefore informed by their ability to provide information on issues being studied, their experience and respective roles. The interviews continued until conceptual saturation was achieved at 30 respondents, where no new concepts emerged from the interviews. Table 1 lists the respondents across Information Technology, Finance, Strategy, Programmes, Monitoring and Evaluation and Analytics who were interviewed.

Data were collected through semi-structured interviews lasting an average of 60 minutes. All interviews were audio-recorded with participants’ consent, and transcribed verbatim upon completion. A semi-structured interview guide was used to ensure consistency across interviews while allowing flexibility to probe emerging issues. The interview guide sort to elicit detailed accounts of respondents’ knowledge and experiences. The five sections covered were: Participants’ roles and organisational context; current use of big data analytics in the organisation; data governance practices and governance challenges; informal interventions, workarounds and enabling conditions adopted in response to governance gaps; and organisational consequences, perceived outcomes and lessons learned.

Although the guide provided a consistent line of inquiry, interviews remained flexible, enabling participants to introduce issues not anticipated by the researchers. Respondents were encouraged to describe concrete experiences, facilitating the identification of emerging governance practices through grounded theory analysis. As analysis progressed, emergent concepts were explored in greater depth during subsequent interviews through theoretical sampling, enriching and refining the developing categories.

Interview transcripts were imported into Nvivo software and were analysed using Strauss and Corbin’s (1998) grounded theory coding framework, comprising open, axial and selective coding. Analysis involved constant comparison across cases and iterative movement between data and theory, ensuring that categories and constructs were grounded in empirical observations.

Open coding data were examined line-by-line which identified key phrases, incidents and concepts, grouping them into categories while staying close to respondents’ language. Analytical memos documented decisions, reflections and category evolution.

Axial coding was used to group-related open codes into higher-order analytical categories. These categories captured recurring organisational patterns including governance gaps, informal coordination mechanisms, operational adaptations and evolving role of analytics in decision-making. Selective coding integrated categories into a central integrative concept, identifying themes and dimensions of emergent governance under a data governance vacuum.

The coding process generated a hierarchical structure linking first-order concepts derived from interview data to higher-order analytical categories. These categories were subsequently integrated into a core theoretical theme describing the emergence of governance practices around analytics use. Table 2 presents the resulting coding structure with representative interview evidence. The respondents are identified using letter R and a number (e.g. R1), years of experience with big data analytics (e.g. 4yrs) and occupation (e.g. IT Manager).

Several strategies were employed to ensure rigour and trustworthiness in the research process. Firstly, data triangulation was used to enhance credibility. Evidence was drawn from multiple sources, including semi-structured interviews and organisational documents such as strategic plans and annual reports. For example, respondents’ descriptions of unclear data ownership and the absence of dedicated analytics structures were corroborated by organisational structures provided in the strategic plan and human resource policy that lacked clearly defined data governance roles.

Secondly, transparency and an audit trail were maintained throughout the research process. Coding procedures, category development and analytical decisions were systematically documented through coding matrices, analytical memos and evolving codebooks. This documentation enabled the research process to be traced from raw interview data to the conceptual categories presented in the findings.

Thirdly, member validation was conducted with selected participants to confirm the plausibility and accuracy of emerging interpretations. Participants were consulted on whether the interpretations reflected their experiences with data governance and analytics use within their organisations.

These strategies collectively strengthened the credibility and robustness of the study’s findings. Through iterative coding, cross-case comparison and triangulation of multiple data sources, the study identified recurring patterns of informal practices, negotiation mechanisms and emergent governance routines. These findings form the basis of the subsequent analysis, highlighting how governance structures evolve in practice to support analytics use under institutional constraints.

Ethical clearance to conduct this study was obtained from the Kenya School of Government, Institutional Scientific and Ethics Review Board (Ref. No. KSG-RAS-ISERC-0191025). Verbal consent was received and included in the beginning of all recorded audio interviews.

The initial context for analytics adoption in many agencies was characterised by limited formal governance structures, fragmented data infrastructures and weak institutionalisation of data-driven decision-making. This governance vacuum shapes the conditions under which analytics initiatives emerge.

A key feature of this environment is the absence of clearly defined institutional roles and support structures for analytics functions. In several agencies, analytics responsibilities were concentrated in a small number of individuals operating without dedicated organisational units or institutionalised processes. One respondent described this situation as follows:

Beyond structural fragmentation, several respondents raised the issue of weak institutional recognition of data governance:

This lack of formal structure was compounded by fragmented data infrastructures, where datasets were dispersed across multiple systems with limited interoperability. Respondents found that the absence of integrated data architectures often made it difficult to develop comprehensive analytical outputs. As one respondent observed:

These findings suggest that analytics initiatives often begin in contexts where governance frameworks are weak or absent, requiring organisations to adapt through alternative mechanisms.

In response to governance gaps, analysts and technical staff developed a range of operational workarounds that enabled analytics activities to continue despite institutional constraints. These practices illustrate how individual initiative and improvisation play a central role in the early stages of analytics adoption.

One common strategy involved hybrid analytical processes combining automated systems with manual inputs. Respondents reported that incomplete or inconsistent data often required manual intervention before analysis could proceed. One respondent described this process as follows:

Operational workarounds also included significant efforts in data preparation and cleaning, which respondents identified as a recurring challenge in analytics work. As one analyst observed:

In some instances, departments created specific data dashboards and had data mediators and gatekeepers. As one respondent said:

Another form of workaround involved individual improvisation through the use of personal tools and locally developed analytical solutions. Analysts frequently relied on independently acquired software or custom scripts to perform analytical tasks that were not yet supported by organisational systems. As one respondent explained:

These practices demonstrate how analytics capabilities can emerge even in the absence of formal governance structures, sustained largely through informal practices and individual initiative.

As analytics initiatives expand, organisations begin to develop informal coordination mechanisms to address the limitations associated with fragmented data environments. These practices represent a transition from individual workarounds towards more collective approaches to data management.

One such mechanism involves informal cross-departmental collaboration to facilitate data sharing. Where technical integration between systems was difficult or costly, agencies adopted simpler mechanisms to exchange information. As one respondent observed:

Another mechanism was the emergence of data champions and communities who bridged the gaps, substituting for formal governance:

Another emerging practice involved efforts to harmonise datasets across organisational units, particularly for reporting and decision-making purposes. Respondents described attempts to establish authoritative datasets that could serve as common reference points. One participant explained:

Coordination practices also extended beyond organisational boundaries, as agencies sought to improve data accuracy through linkages with external government systems. One example involved the use of external databases to validate internal records:

These developments indicate that governance practices begin to emerge organically through collaboration and data sharing, even before formal governance frameworks are fully established.

In some agencies, these emergent coordination practices gradually evolved into formal governance structures and processes. This stage represents the institutionalisation of data governance within organisational structures.

One key development involved the creation of dedicated governance and analytics units, separating analytical functions from broader ICT operations. As one respondent described:

Some interventions gradually evolved into formal structures, governance councils. As one respondent observed:

Another mechanism involved the introduction of data stewardship roles within departments to strengthen accountability for data quality and management. One participant explained:

There was the institutionalisation of internal data policies that included standards, harmonisation. As one respondent said:

Institutionalisation was also supported through capacity building initiatives and governance forums, which aimed to strengthen organisational awareness of data governance principles and coordinate analytics initiatives across units:

These developments illustrate how governance structures often emerge in response to the operational challenges of managing data, rather than being fully designed at the outset of analytics adoption.

Taken together, the findings reveal a progressive pathway through which data governance develops in resource-constrained public sector environments. Rather than being implemented as a predefined framework, governance evolves through a sequence of organisational responses to the challenges associated with operationalising analytics capabilities.

A central insight from this study is that data governance often develops after analytics activities begin rather than preceding them. In many agencies, analytics initiatives were initially driven by individual analysts or small technical teams operating without formal governance structures. These actors relied on improvisational practices and personal expertise to generate analytical outputs despite fragmented data environments and limited institutional support, as is presented in other IS literature. Similar patterns have been observed in research on analytics adoption in the public sector, where the absence of integrated governance frameworks can lead to localised experimentation and improvisation (McCoy & Rosenbaum 2019).

This finding challenges the assumption in much of the data governance literature that governance frameworks must be designed before analytics initiatives can be effectively implemented. The findings suggest that in resource-constrained public sector environments, governance may instead evolve incrementally through practice.

This process reflects what may be described as emergent data governance, where governance mechanisms develop in response to the operational challenges associated with managing and using data. Rather than being imposed through top-down institutional design, governance practices arise through iterative problem-solving, collaboration and learning within organisations.

This perspective extends existing work on data governance by highlighting the importance of adaptive institutional processes, particularly in contexts where formal governance structures are still evolving.

The findings illustrate how institutional constraints shape the ways in which organisations adapt to support analytics initiatives. The governance vacuum identified in the study reflects several structural challenges common in public sector digital transformation initiatives, including fragmented data infrastructures, limited technical capacity and established decision-making cultures that rely heavily on experience and intuition (Bullini Orlandi & Pierce 2020; Ku & Gil-Garcia 2018).

Rather than preventing analytics adoption, these constraints triggered the development of operational workarounds, where analysts created locally developed solutions to compensate for institutional limitations. These practices included the use of personal analytical tools, manual data integration processes and repeated data cleaning efforts.

Such practices demonstrate the role of agency and improvisation in the early stages of digital transformation. Individual analysts and technical staff often act as institutional entrepreneurs who introduce new analytical practices into organisational contexts that may not yet be fully prepared to support them. These findings resonate with research on digital innovation in public organisations, which emphasises that new technological practices are often introduced as pilots or by individuals or small teams (Osakwe 2025).

Over time, these localised practices begin to generate broader organisational awareness of the value of analytics, creating pressures for more systematic coordination and governance.

The findings further reveal how informal coordination practices gradually evolve into more structured governance arrangements. As analytics initiatives expand, agencies begin developing mechanisms to address data fragmentation and improve analytical reliability. These mechanisms include informal cross-departmental data sharing, data harmonisation efforts and linkages with external government datasets (Janssen et al. 2020).

These coordination practices represent an intermediate stage in the governance process, where organisations begin recognising the need for shared standards and collective data management approaches (Gujar 2025). Over time, these practices may become formalised through the creation of governance units, data stewardship programmes and governance councils responsible for overseeing data management processes.

This progression suggests that governance institutionalisation is often driven by operational necessities rather than purely strategic planning. As analytics initiatives generate demand for more reliable and integrated data, organisations are compelled to establish governance structures that support these requirements.

This finding highlights the importance of viewing governance not only as a set of formal policies but also as a dynamic organisational capability that evolves alongside analytics adoption.

Building on empirical findings, the conceptual model depicted in Figure 1 outlines the emergence of data governance during the adoption of big data analytics in public sector organisations.

The model conceptualises governance development as a four-stage institutional process:

The model therefore suggests that institutional readiness for analytics is not solely a precondition but also an outcome of analytics adoption. Governance capabilities develop progressively as organisations confront the practical challenges associated with data use.

This conceptualisation contributes to existing literature by framing data governance as an emergent institutional process rather than a static organisational structure.

The study offered important insights into how data governance is emerging within state agencies, recognising that qualitative case research is intended to stimulate deeper reflection rather than produce broadly generalisable findings. While the multiple-case approach expanded the analytical scope, differences in cultural and institutional contexts may limit applicability beyond the Kenyan state agencies examined, although the results remain robust and sufficiently address the research questions.

Future research could extend this work by examining how emergent governance processes unfold across different institutional contexts and levels of government, as well as how governance structures influence the long-term sustainability and impact of analytics initiatives.

The study contributes to the literature on digital governance and public sector analytics in three ways. Firstly, it demonstrates that institutional readiness for big data analytics should be understood as a dynamic process rather than a static organisational condition. Secondly, it introduces the concept of emergent data governance to explain how governance structures develop incrementally through organisational practice. Thirdly, it provides empirical insight into how decision-making culture interacts with governance mechanisms during analytics adoption in public sector contexts.

These findings have important implications for policymakers and public sector leaders seeking to promote data-driven decision-making. Rather than focusing solely on the design of formal governance frameworks, efforts to strengthen analytics capabilities should also support organisational learning, cross-departmental collaboration and capacity building, which play a critical role in enabling governance structures to evolve.