Identifying and Prioritising Factors for Effective Decision-Making in Data-Driven Organisations: A DEMATEL Approach

Abstract

The strategic transformation of increasing data volumes into managerial decisions is critical for organisational performance and sustainability; yet, it faces hurdles like poor data quality, technological deficiencies, and skill gaps. This study investigates the causal interdependencies among key factors influencing data-driven decision-making within data-driven organisations. Utilising the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method, a robust structural and Multi-Attribute Decision-Making (MADM) technique, expert judgments from five management-level professionals were analysed to construct direct and total-relation matrices. The results classify Data Analytics Literacy (DAL) and Business-Strategy Alignment (BSA) as primary causal factors, while Data Quality (DQ), Data Infrastructure & Technology (DIT), and Data Culture & Governance (DCG) emerge as effect factors. These findings provide a structured framework for prioritising managerial interventions, suggesting that strengthening foundational elements (DAL and BSA) will significantly enhance analytical capabilities and strategic alignment. A limitation is the small, expert-based sample, indicating the potential for future validation with larger, more diverse panels or Fuzzy-DEMATEL applications.

1. Introduction

In the contemporary economic and technological landscape, which is characterised by an exponential increase in the volume, velocity, and variety of data (Big Data) [1], the ability of organisations to efficiently leverage these information resources has become a key factor in performance and sustainability [2,3,4]. Transforming raw data into actionable insights and, subsequently, into well-founded managerial decisions is a strategic imperative that confers a distinctive competitive advantage in a volatile and dynamic business landscape [2,5]. Within the Business Process Management (BPM) disciplinary framework, the judicious integration of data into decision-making processes is a systemic necessity for optimising operational procedures, stimulating innovation, and ensuring organisational adaptability [6,7].

However, the process of transitioning from data collection to making strategic decisions based on empirical evidence is often fraught with complex, multidimensional challenges [8]. Academic literature [5,9] and managerial practice highlight significant obstacles, such as suboptimal data quality, technological infrastructure deficiencies [10], cultural resistance to change, information fragmentation, and gaps in digital and analytical literacy among staff [11]. These systemic impediments can substantially diminish the return on investment in advanced data analytics solutions, creating a critical gap between an organisation’s aspiration to be ‘data-driven’ and its actual operational reality [12].

In order to address these dysfunctions holistically and optimise the process of transforming data into decisions, it is essential that we have a profound understanding of the causal and interdependent relationships among the critical factors influencing this endeavour [13]. Traditional analytical methods often struggle to model the complexity of nonlinear interactions and reciprocal influences among multiple variables [14]. In this context, the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method is a robust tool capable of overcoming these limitations [15]. DEMATEL not only identifies critical factors, but also precisely maps cause-and-effect relationships, offering a systemic and nuanced perspective on the dynamics of reciprocal influences [15,16]. DEMATEL is chosen because it can quantify experts’ qualitative judgements, transforming them into structured information that clarifies which factors have the most significant causal influence (cause factors) and which factors are most strongly influenced (effect factors). This differentiation is crucial for the strategic prioritisation of managerial interventions and the efficient allocation of resources [16,17].

Unlike other MADM methods that often focus on selecting an optimal option or identifying a single critical factor, DEMATEL provides a comprehensive matrix framework for analysing causal relationships and interdependencies between various factors [15]. This allows for a more holistic and dynamic approach, which is particularly beneficial in complex domains like data management where the factors are highly interconnected and variable [14]. An additional advantage of DEMATEL is its ability to facilitate the indirect improvement of critical factors when direct influence is not feasible, by strengthening their determinants to achieve systemic enhancement [18]. This makes DEMATEL a valuable strategic management tool, as it offers an intuitive understanding of system dynamics and enables the differentiation of cause and effect factors for the strategic prioritisation of interventions. Furthermore, it quantifies influence intensity, providing a comprehensive view of both direct and indirect reciprocal impacts among factors [19].

The primary aim of this study is to identify, analyse, and prioritise the causal relationships between the key factors that influence the success of the process of transforming data into managerial decisions in data-driven organisations. This study is particularly useful for organisations seeking to enhance their data-driven culture and optimise business processes, ultimately contributing to their long-term sustainability by ensuring more informed and effective decision-making. By applying the DEMATEL method, this study aims to provide a structured decision-making framework and contribute to the development of effective strategies for adopting a data-driven culture and optimising business processes [17,19].

This paper is structured as follows: Section 2, “Materials and Methods,” provides a detailed explanation of the DEMATEL methodology, outlining its fundamental steps and the advantages that justify its selection for this research. Section 3, “Results,” presents the empirical findings obtained from the application of the DEMATEL method, including the direct-relation matrix, total-relation matrix, and the assessment of factor influence and causality. Section 4, “Discussion,” interprets these results in the context of previous studies and discusses their implications for organisations striving to become data-driven. Finally, Section 5, “Conclusions,” summarises the main contributions of this study and suggests avenues for future research.

2. Theoretical Framework

The effective transformation of raw data into actionable managerial insights is contingent upon a constellation of interrelated organisational, technological, and human factors. This section delineates the principal constructs examined in this study, each substantiated by contemporary scholarly literature, to elucidate their individual and synergistic contributions to data-driven decision-making efficacy. By critically reviewing the dimensions of Data Quality (DQ), Data Infrastructure and Technology (DIT), Data Culture and Governance (DCG), Data Analytics Literacy (DAL), and Business-Strategy Alignment (BSA), this section establishes a conceptual foundation for understanding how organisations can systematically leverage data assets to enhance strategic and operational performance in increasingly complex and dynamic environments.

Table 1. Factors included in the DEMATEL analysis.

Factor	Factor	References
A	Data Quality (DQ)	[2,5,9,20,21,22]
B	Data Infrastructure & Technology (DIT)	[2,4,20,21,23,24,25]
C	Data Culture & Governance (DCG)	[9,20,21,26]
D	Data Analytics Literacy (DAL)	[20,21,23,26]
E	Business-Strategy Alignment (BSA)	[9,21,27]

presents the factors included in the analysis.

This selection of these five factors (DQ, DIT, DCG, DAL, and BSA) is based on a comprehensive review of the existing literature that consistently identifies them as pivotal to successful data-driven decision-making. While previous research often examines these factors individually or as a collective, our framework is distinct in its specific focus on uncovering the causal interdependencies among them using the DEMATEL method, rather than simply identifying their importance or offering a prescriptive model. This allows us to understand how these factors influence each other, providing a more nuanced and actionable understanding for organisations.

DQ refers to the degree to which data are fit for their intended use in operations, decision-making, and strategic planning. It is generally operationalised through dimensions such as accuracy (truthfulness), completeness (extent of missing data), consistency (absence of contradictions), timeliness (currency), and validity (adherence to defined formats and rules) [5,9]. Inadequate data quality compromises the integrity of analytical processes, introduces cognitive and computational biases, and may result in decisions based on erroneous assumptions or misleading trends [5].

From a systems theory perspective, data quality represents an upstream determinant in the information value chain: errors introduced at the point of data capture or during preprocessing can propagate through successive stages of analysis and modeling, magnifying their detrimental effects [21]. Moreover, high-quality data underpin organisational trust in analytics, thereby influencing user adoption, decision confidence, and the overall return on data investments. Consequently, data quality is not merely a technical attribute but a strategic asset integral to sustaining a competitive advantage in information-rich contexts [2].

DIT encompasses the technical systems, platforms, and tools that support the entire data lifecycle—from ingestion and storage to processing, analysis, and dissemination [28,29]. Core components include databases, data lakes, cloud-based architectures, ETL (extract–transform–load) pipelines, APIs, analytics platforms, and real-time processing engines [24,25]. The efficiency, scalability, and interoperability of this infrastructure determine an organisation’s capacity to respond to data velocity, volume, and variety—key attributes of the so-called “Big Data” paradigm [20,23].

In a dynamic business environment characterised by rapid technological evolution and increasing data complexity, a resilient and adaptable DIT enables not only efficient data handling but also accelerates innovation and supports predictive, prescriptive, and automated decision-making. A robust data infrastructure also facilitates integration across heterogeneous sources [30,31], enabling more comprehensive and context-rich analyses. Organisations that lack adequate technological frameworks risk delayed insights, fragmented knowledge bases, and systemic inefficiencies, thereby reducing the strategic utility of their data assets [2,21,32].

DCG refers to the set of organisational norms, values, policies, and institutional mechanisms that regulate how data are perceived, managed, and utilised. A strong data culture is characterised by normative support for evidence-based decision-making, cross-functional collaboration around data, and widespread data literacy [26]. It reflects an organisational mindset in which data are regarded as a critical resource and a driver of continuous improvement [21].

Complementarily, data governance involves the formalisation of processes and responsibilities that ensure data quality, security, privacy, accessibility, and compliance with regulatory standards [9]. Effective governance structures delineate ownership, stewardship, and accountability across the data lifecycle. From an institutional theory perspective, governance mechanisms legitimise data practices, mitigate risks, and institutionalise ethical and lawful data usage [2,5]. Without a coherent culture and governance framework, organisations may experience informational asymmetries, resistance to change, fragmentation of datasets, and reduced alignment between analytical outputs and managerial needs [2,33,34]. Therefore, DCG is fundamental not only for ensuring ethical and efficient data use but also for embedding analytics within organisational routines and decision architectures.

DAL denotes the cognitive and technical competencies that enable individuals within an organisation to understand, interpret, critique, and apply the data and analytical results [26]. This construct encompasses a spectrum of skills, including numeracy, statistical reasoning, data visualisation, tool proficiency, and the contextual interpretation of findings. DAL is a human capital asset that conditions the extent to which analytical insights are effectively integrated into decision-making processes [21,35].

Organisations with high DAL levels tend to experience better cross-functional communication, a reduced dependence on data specialists for routine tasks, and greater agility in responding to analytical findings. Conversely, deficits in analytics literacy create interpretive bottlenecks, reduce trust in insights, and can lead to the misapplication of results [20]. From the perspective of organisational learning theory, DAL represents a key capability for absorptive capacity—that is, the ability to recognise, assimilate, and apply new knowledge. In increasingly automated and AI-augmented environments, developing advanced DAL across hierarchical levels ensures that technological potential is matched by interpretive competence, thereby maximising the value extracted from data systems [23].

BSA refers to the degree of congruence between an organisation’s data-driven initiatives and its overarching strategic objectives, priorities, and performance metrics. This alignment ensures that data efforts are purposefully directed toward solving core business problems, enhancing value creation, and enabling strategic differentiation [21]. BSA acts as a contextual anchor, connecting analytical projects with desired business outcomes.

Strategic misalignment, by contrast, leads to analytics initiatives that are disconnected from operational realities, resulting in wasted resources, unutilised insights, and a diminished organisational impact. From a strategic management perspective, BSA is essential for ensuring that analytics serve as a source of sustainable competitive advantage rather than a cost center. It involves the integration of analytics within strategic planning processes, KPI formulation, and performance monitoring systems [9]. Organisations that embed data practices within their strategic frameworks can more effectively optimise processes, identify emergent opportunities, and adapt to environmental volatility, thereby enhancing organisational resilience and innovation capacity [36].

Together, the five constructs presented—DQ, DIT, DCG, DAL, and BSA—form an integrated framework that captures the multifaceted nature of data-driven decision-making. These dimensions interact dynamically, with deficiencies in one area often undermining progress in others. Understanding their interdependencies and contextual relevance enables organisations to design more coherent, adaptive, and performance-oriented data strategies. In the following sections, this conceptual model will guide the empirical analysis of how organisations operationalise data capabilities to achieve sustained strategic and operational excellence.

3. Data Collection and Applied Method

The Decision-Making Trial and Evaluation Laboratory (DEMATEL) method is a structural and multi-criteria analysis technique founded on the theory of directed graphs (digraphs). This method is designed to analyse the complex causal and interdependent relationships among a set of factors or criteria, being particularly pertinent for studying complex systems where reciprocal interactions are prevalent. DEMATEL enables not only the identification of critical system elements but also the visualisation of their causal structure, differentiating between factors that exert a predominant influence and those that are primarily influenced.

The methodological advantages of DEMATEL, which justify its selection for the present research, include visualising the systemic structure and facilitating the construction of a causal map that illustrates direct and indirect relationships between factors. This offers an intuitive understanding of system dynamics. Furthermore, it enables the differentiation of cause and effect factors by classifying them into cause and effect categories, which is essential for the strategic prioritisation of interventions. Another advantage is the ability to quantify the influence intensity and measure both the direct and indirect influences among factors, providing a comprehensive view of the reciprocal impact. Lastly, the method has been shown to be effective in complex decision-making situations, efficiently analysing multi-criteria decision problems with interconnected criteria, which are specific to strategic management, quality management, and, in this case, data management.

While other MADMs like Analytic Hierarchy Process (AHP), Analytic Network Process (ANP), or Fuzzy-DEMATEL exist, DEMATEL was specifically chosen for this study due to its unique ability to map causal relationships and interdependencies among factors, rather than just ranking them [16,37]. AHP and ANP are excellent for hierarchical decision problems and dependency modeling, respectively, but they do not explicitly distinguish between cause and effect factors in the same quantitative manner as DEMATEL. Fuzzy-DEMATEL extends the method to handle uncertainty in expert judgments, but, for this initial exploration of causal relationships, the crisp DEMATEL method was deemed sufficient to provide clear insights into the core interdependencies. This explicit mapping of cause and effect is crucial for identifying actionable intervention points, which is the core aim of this research.

This qualitative research study is based on the expertise of professionals with relevant experience. Respondents were selected from individuals with management experience in data-driven organisations. The DEMATEL methodology specifically requires four to five respondents to assess the interactions between factors [15,16,18]. The factors were evaluated based on responses from five management-level respondents from organisations with a strong data orientation, each with at least 2 years’ experience, in order to assess the degree of mutual influence between the relevant factors. Data were collected in June 2025. The questionnaire was structured as a relationship matrix containing six factors, and participants were asked to rate the influence of each pair of factors (36 arrangements).

The selection of five experts for this DEMATEL study, while seemingly small, is consistent with established methodological guidelines for DEMATEL, which suggest a panel of four to five experts is generally sufficient given the iterative nature of consensus-building and the depth of qualitative judgment required [15,16]. This approach prioritises the depth and quality of expert insights over a larger, potentially less cohesive, sample. To ensure the robustness and credibility of our findings, we specifically chose management-level professionals from organisations with a strong data orientation, each possessing a minimum of two years of relevant experience. These criteria ensured that all participants had firsthand knowledge and practical understanding of the challenges and opportunities in data-driven decision-making within organisational contexts (

Table 2. Expert panel.

Expert	Role /Position	Industry Sector	Years of Relevant Experience	Key Area of Expertise
E1	Operations manager	Financial services	3	Data governance
E2	IT manager	Manufacturing	5	Data infrastructure
E3	Expert accountant	Financial services	2	Data governance
E4	Business development lead	Engineering and management	7	Strategic planning
E5	Head of Operations	Retail	10+	Data-driven decision-making

).

While this sample size might limit the statistical generalizability to all organisations, the goal of DEMATEL is to provide a systemic understanding of complex interdependencies based on informed expert opinion, which this panel adequately provided. Future research could explore validating these findings with a larger, more diverse sample.

The fundamental steps of applying the DEMATEL method are rigorously structured and involve the following operations. First, the set of factors is defined by explicitly delimiting and validating the relevant factors for the research problem, based on a review of the specialised literature and preliminary expert consultation. Next, the initial direct-relation matrix $A$ is constructed by collecting expert judgements on the direct influence of each factor on the others. This is performed using a predefined numerical scale:

• 0 = no influence;
• 1 = weak influence;
• 2 = moderate influence;
• 3 = strong influence;
• 4 = very strong influence.

Expert opinions are aggregated statistically to form a consolidated initial matrix. The specific aggregation method used was the arithmetic mean of individual expert ratings for each pairwise influence, ensuring a consolidated representation of the collective expert judgment.

The next step is to normalise the matrix of direct relations $Y$, by scaling the values of matrix $A$ to ensure compatibility with subsequent calculations.

Normalisation is performed according to the following formula:

$$Y=A·k$$

(1)

$$\mathrm{Where}: A=⌊\begin{array}{cccccc}0& a_{12}& \dots & a_{1j}& \dots & a_{1n}\\ a_{21}& 0& \dots & a_{2j}& \dots & a_{2n}\\ \dots & \dots & \dots & \dots & \dots & \dots \\ a_{i1}& a_{i2}& \dots & a_{ij}& \dots & a_{in}\\ \dots & \dots & \dots & \dots & \dots & \dots \\ a_{n1}& a_{n2}& \dots & a_{nj}& \dots & 0\end{array}⌋$$

(2)

$$k={\displaystyle \frac{1}{{max}_{1\le i\le n}\left(\sum _{j=1}^n{a}_{ij}\right)}} (i,j=1, 2, \dots , n)$$

(3)

where $n$ is the number of factors and ${max}_{1\le i\le n}\left(\sum _{j=1}^n{a}_{ij}\right)$ is the maximum value among the sum of elements in rows and the sum of elements in columns. For example, if the maximum row sum in matrix $A$ is 10 and the maximum column sum is 12, then $k={\displaystyle \frac{1}{12}}$. Each element in matrix $A$ is then multiplied by this k to obtain the normalised matrix Y. This step ensures that all values in the normalised matrix fall within a range suitable for further matrix operations, typically between 0 and 1.

Then, the total-relation matrix $T$ is calculated, determining the total influences (both direct and indirect) among factors. This matrix is derived from the normalised matrix $Y$ and the identity matrix $I$ (unit matrix), using the following relationship:

$$T=Y·{\left(I-Y\right)}^{-1}$$

(4)

This formula calculates the total influence by considering all possible paths of influence between factors, direct and indirect. The ${\left(I-A\right)}^{-1}$ term represents the inverse of the difference between the identity matrix and the normalised direct-relation matrix, effectively capturing the cumulative impact of influences throughout the system.

The threshold for assessing the α-factor is established using the following formula:

$$\alpha ={\displaystyle \frac{\sum _{i=1}^n\sum _{i=1}^n\left[t_{ij}\right]}{N}}$$

(5)

where $N=n^2$ denotes the total number of elements or relationships within the influence matrix.

The $\alpha$ threshold is typically used to filter out negligible influences in the total-relation matrix, focusing on the more significant relationships for constructing the causal map. Relationships with values in the total-relation matrix greater than α are considered significant. For this study, the $\alpha$ threshold of 2.296 was calculated using Formula (5) based on the total-relation matrix. This calculated positive value ensures that only influences above the average total influence are considered, providing clarity and emphasising the most significant causal relationships in the map.

Once the comprehensive influence matrix is computed, two key metrics can be derived for each criterion: the impact score $D_i$ and the causal score $R_j$.

The impact score $D_i$ of a given factor represents the cumulative influence it exerts on all other factors, and is calculated as follows:

$$D_i= {\left[\sum _{j=1}^n{t}_{ij}\right]}_{nx1}={\left[t_i\right]}_{{nx1}^{‣}}$$

(6)

$$\mathrm{for} \mathrm{each} i\in \{\mathrm{1,2},\dots ,n\}.$$

Conversely, the causal score $R_j$ reflects the total influence that all other factors exert on a particular factor, and is given by the following:

$$R_j= {\left[\sum _{j=1}^n{t}_{ij}\right]}_{1xn}={\left[t_j\right]}_{{nx1}^{‣}}$$

(7)

$$\mathrm{for} \mathrm{each} j\in \{\mathrm{1,2},\dots ,n\}.$$

4. Results

The interdependencies among the identified factors were systematically examined using the DEMATEL methodology. This approach facilitated the construction of a series of matrices that capture the structure and intensity of mutual influences between the factors. Table 2 displays the direct influence matrix, which quantifies the extent to which each factor directly impacts another. The values within this matrix are derived from expert assessments, where each numerical entry reflects the perceived magnitude of direct influence exerted by one factor upon another, based on structured evaluative input from the study’s participants.

Table 3. Direct-relation matrix.

Factor	A	B	C	D	E
A	0	2.5	2.75	2.5	2.5
B	3	0	2.75	2.75	2.75
C	2.75	3	0	2.5	2.5
D	2.5	3.25	3.25	0	2.25
E	2.75	3.25	3	2.75	0

displays the direct influence matrix. This matrix quantifies the extent to which each factor directly impacts another. The values within this matrix are derived from expert assessments, where each numerical entry reflects the perceived magnitude of direct influence exerted by one factor upon another, based on structured evaluative input from the study’s participants. This step corresponds to the construction of the initial direct-relation matrix.

Each cell in the direct-relation matrix indicates the direct influence of factor $i$ on factor $j$. For instance, the value of 2.5 in row A, column B signifies that data quality (A) has a direct influence of 2.5 on data infrastructure and technology (B) as perceived by the experts. The diagonal elements are 0, as a factor does not directly influence itself.

Table 4. Total-relation matrix highlighting (*) factors with significant influence $(\alpha >2.296)$.

Factor	A	B	C	D	E
A	2.017	2.328 *	2.305 *	2.109	2.038
B	2.376 *	2.325 *	2.472 *	2.276	2.199
C	2.284	2.441 *	2.203	2.186	2.112
D	2.349 *	2.539 *	2.499 *	2.089	2.171
E	2.444 *	2.626 *	2.571 *	2.354 *	2.088

presents the total-relation matrix. This matrix encapsulates the full extent of influence that each factor exerts on and receives from all other factors, accounting for both direct and indirect relationships. For example, the values in row ‘A’ indicate the total influence of data quality on all other factors, including indirect effects.

The $\alpha$ threshold of 2.296 signifies that only total influence values from the total influence matrix (Table 4) greater than this threshold are considered significant for the purpose of constructing the causal map, highlighting the most impactful relationships within the system. This positive value for $\alpha$ ensures that only influences demonstrating a clear, substantial positive impact are included, thereby providing a more focused and actionable causal representation.

The total-relation matrix shows the comprehensive influence, both direct and indirect. For example, the value of 2.017 for factor A on itself (A,A) indicates its cumulative self-influence or feedback loop. The positive values across the matrix indicate a consolidated direct influence and an amplification effect due to complex interdependencies. Thus, indirect influences tend to amplify the initial direct impact, reflecting robust interconnectedness. The asterisks highlight factors that have a total influence greater than $\alpha =2.296$, indicating particularly strong and noteworthy relationships for further analysis.

Subsequently, the row and column sums are calculated. For each factor, two key indicators are determined from the total-relation matrix $T$: the sum of influences given ($D$), determined by summing the elements of each row in matrix $T$, which quantifies the total influence a factor exerts on all other factors; and the sum of influences received ($R$), calculated by summing the elements of each column in matrix $T$, which indicates the total influence a factor receives from all other factors (

Table 5. Assessment of the influence and causality of factors.

Factor	D	R	D + R	D − R	Dominant Characteristic
A	10.798	11.470	22.269	−0.671	Effect
B	11.649	12.260	23.909	−0.610	Effect
C	11.227	12.053	23.281	−0.825	Effect
D	11.649	11.016	22.665	0.632	Cause
E	12.085	10.609	22.694	1.476	Cause

). Based on the values of $D$ and $R$, the prominence and relation indicators are determined. The prominence indicator ($D+R$), the sum of $D$ and $R$, measures the total importance of a factor within the system, indicating its degree of involvement in general interactions. The relation indicator ($D-R$), the difference between $D$ and $R$, classifies factors into cause and effect categories. A positive D-R value designates a cause factor, while a negative value indicates an effect factor.

Finally, the causal map is constructed, which is a graphical representation of the factors in a cartesian coordinate system, with $D+R$ on the horizontal axis and $D-R$ on the vertical axis (Figure 1). This allows for a visual inspection of the causal relationship structure and the rapid identification of the role of each factor.

In the causal map, DAL and BDA (D, E) are positioned in the upper portion of the $D-R$ axis, indicating their role as cause factors. These, when fully accounted for, can lead to an increase in the overall perceived impact, or the expert ratings, when aggregated and processed through the DEMATEL formulae, might result in these numerical outcomes. This could also imply that the system is highly integrated and that strong influences in one area might necessitate significant effort across multiple factors to yield positive net change. The causal map therefore provides an intuitive understanding that, to enhance DQ, DIT and DCG, organisations should primarily focus on improving DAL and BDA. The current study used the DEMATEL method to identify and prioritise factors for effective, data-driven decision-making within organisations. The analysis revealed that DAL and BSA are causal factors, while DQ, DIT, and DCG were identified as effect factors. This suggests that interventions aimed at improving DAL and BSA are likely to positively influence DQ, DIT, and DCG, thereby enhancing data-driven decision-making and organisational performance. These findings provide a nuanced understanding of the causal relationships within the complex ecosystem of data-driven decision-making and offer actionable insights for strategic management.

5. Discussion

Comparing these results with the relevant literature from the last five years reveals significant convergence and distinct nuances, particularly in the classification of factors as causes or effects. The methodology employed in the example study by Estiri et al. [15], which used DEMATEL-MABAC for High-Performance Work Systems, demonstrates its ability to identify critical interdependencies—a strength mirrored in our current analysis.

Our findings regarding the cause factors (DAL and BSA) largely align with the prevailing academic discourse, which often emphasises these elements as prerequisites for successful data-driven initiatives.

Classifying DAL as a primary causal factor aligns with much of the literature. For example, Koltay [23] describes the evolution from data literacy to AI literacy, highlighting their growing importance, while Taş [26] emphasises the pivotal role of data literacy in university–industry collaborations. Our results suggest that DAL is a fundamental driver that influences the subsequent stages of data utilisation and decision-making accuracy, including improvements in DQ, DIT, and DCG. Our study reinforces the idea that DAL is not just a desirable attribute, but a critical factor that determines the effectiveness of all other data-related endeavours. For instance, Airbnb’s internal training initiative, “Data University,” offers a compelling example of how data analytics literacy transforms organisational decision-making. By teaching non-technical staff how to read and act on analytical dashboards, Airbnb has made access to insights more democratic, enabling employees in all roles to make evidence-based decisions. The programme has led to a significant increase in data adoption rates [38], demonstrating how DAL can improve data quality, culture, and infrastructure throughout an organisation.

The identification of BSA as a causal factor is well-supported by studies focusing on the strategic underpinnings of data-driven transformation. In the literature on strategic management and data governance [2,5,20], BSA is often presented as an important precondition for the success of data-driven initiatives, ensuring that data efforts are directed towards achieving strategic objectives. Our results indicating BSA’s causal role emphasise that a clear and strong alignment between business strategy and data initiatives is determinative of an organisation’s capacity to process, store, and utilise data effectively, thereby accelerating innovation. This reflects the wider consensus that inadequate BSA can severely hinder an organisation’s ability to extract timely insights, leading to systemic inefficiencies.

A pertinent example is a payroll software firm preparing for an IPO which partnered with Kenway Consulting to design a data governance roadmap aligned with its strategic goals. Through cross-departmental interviews and stakeholder engagement, the organisation defined roles, KPIs, and metrics that linked their data systems closely to business priorities. Consequently, their governance framework not only supported compliance and risk mitigation, but also directly enabled the essential investor-readiness analytics capabilities [39].

While the ‘cause’ factors (DAL and BSA) demonstrate a high degree of alignment, classifying DQ, DIT, and DCG as ‘effect’ factors provides a more nuanced perspective, departing from the way in which some recent literature often implicitly or explicitly positions these elements as primary drivers or preconditions.

Many studies emphasise the importance of DQ as a fundamental element. For example, Delinschi et al. [5] emphasise that robust DQ assessment methodologies are foundational as well as procedural, and directly impact the reliability and trustworthiness of any data-driven outcome. However, our DEMATEL analysis provides a different perspective: it suggests that DQ is an outcome significantly influenced by robust cause factors such as DAL and BSA. Unlike studies that treat DQ as an isolated attribute, our findings suggest that high DQ is fostered and facilitated by preconditions such as strong DAL and effective BSA. This implies that efforts to improve DQ might be ineffective if the human and strategic aspects (DAL and BSA) are not addressed first.

This is illustrated by a longitudinal case study in the aviation sector, in which a European operator implemented a structured programme to improve data quality over three years. By introducing cross-departmental feedback loops, audit mechanisms, and employee training, the organisation reduced the variability in maintenance reporting and improved the accuracy of predictive scheduling. These quality improvements were enabled by investments in employee analytics literacy and a clearer alignment between maintenance strategy and data architecture [40].

Employees’ ability to interpret and utilise data effectively (DAL) appears to be a consequence of, or an emergent property of, a well-established data foundation. This offers valuable insight for managerial interventions: rather than simply promoting quality initiatives, organisations should first ensure that the foundational elements of DAL and BSA are in place.

To further illustrate these causal relationships with real-world examples from the literature [39,41], consider the challenges faced by organisations in ensuring DQ. A common issue is inconsistency in the data entry or interpretation across departments. If employees lack sufficient DAL, they may not understand the importance of consistent data formats or how their input impacts downstream analysis. Similarly, if there is no clear BSA guiding data collection and usage, different departments might collect data in disparate ways, leading to quality issues when trying to integrate information for a holistic view. For instance, a report on the data literacy impact highlights how a retail corporation, by providing data literacy training to its marketing team, enabled them to better analyse customer data, leading to more targeted campaigns and improved customer satisfaction, an outcome intrinsically linked to higher data quality in their customer relationship management efforts [42]. This showcases how improvements in DAL and BSA directly lead to better DQ outcomes.

Similarly, although DIT is often considered essential, our study classifies it as an effect factor. Chakraborty et al. [20] discuss the evolution towards data-driven paradigms in critical sectors such as medicine and healthcare. They emphasise that such advancements fundamentally rely on sophisticated and robust technological infrastructures capable of handling vast and complex datasets. However, our results suggest that DIT, while important, is influenced by DAL and BSA. An organisation with a well-trained analytical workforce (DAL) and a clearly defined data strategy (BSA) will be better positioned to invest in and effectively implement the necessary infrastructure. This pattern is evident in a case study of 32 regional university hospitals in France that implemented clinical data warehouses to support research and analytics. The maturity and usefulness of these warehouses depended not only on their technical features, but also on the hospitals’ ability to align the technology with their strategic vision and provide appropriate training for their staff. Those that prioritised DAL and cross-functional governance achieved more scalable [31], interoperable DIT environments, facilitating real-time analytics [43]. Therefore, a resilient, scalable, and interoperable DIT appears to be an outcome of a sound strategic approach and strong human capability, rather than a standalone driver. For example, it has been observed that, despite investments in advanced data science infrastructure, businesses often fail to achieve desired outcomes if there is an absence of a common ‘medium of communication’ between the technical and business staff, underscoring how data literacy facilitates the effective utilisation of such technologies [44]. Concurrently, case studies on strategic technology and business alignment, such as in a credit bureau, demonstrate how organisations that achieve this alignment manifest better performance through the effective integration of IT infrastructure with business objectives.

In a similar vein, DCG is classified as an effect factor. Chaudhuri et al. [2] explore how emerging technologies can foster a data-driven culture and enhance innovation capabilities, implicitly acknowledging that a conducive culture is a prerequisite. Furthermore, in their discussion of digital academic leadership in higher education, Jing et al. [21] highlight that effective data utilisation in such contexts necessitates a strong underlying data culture. However, our study confirms that robust DCG, encompassing shared data values and formal management policies, is the result of sound DAL and effective BSA. Formalising processes for data quality, security, and compliance aligns with the view that a strong data culture is successfully embedded when literacy and strategic alignment are well-established, allowing for the effective integration of analytics into an organisation’s routine operations and strategic thinking. This is illustrated by the example of GE Aviation, where the implementation of centralised data governance teams, such as data stewards, quality owners and database administrators, enabled cohesive, cross-departmental practices and shared taxonomies. However, these structures only became effective once the organisation had established a robust DAL foundation and prioritised analytics within its business strategy. The governance framework formalised data quality, ethics, and access management, thereby creating a scalable, collaborative data culture [38].

Empirical evidence from the public service sector indicates that data literacy competence has a crucial mediating effect in the relationship between data governance and a data-driven culture [45], affirming that a more data-literate workforce actively contributes to the development and adherence to governance frameworks [46]. Complementarily, research exploring the impact of business-IT alignment on organisational culture provides clear indications that such alignment significantly influences the successful evolution and implementation of a desired data culture [46].

In conclusion, while the current study confirms the universally acknowledged significance of DAL and BSA in data-driven decision-making, its application of the DEMATEL method provides a unique causal understanding. Identifying DQ, DIT, and DCG as effect factors suggests that efforts to improve these areas may be more fruitful when preceded by robust advancements in DAL and BSA. This offers a structured framework for prioritising managerial interventions, suggesting a sequential approach to fostering an effective, data-driven culture and optimising business processes for long-term sustainability and enhanced decision-making capabilities.

This study makes some relevant key contributions. Theoretically, it advances our understanding of the causal relationships among critical factors in data-driven decision-making, providing a more nuanced perspective than studies that merely identify their importance. By applying the DEMATEL method, we move beyond simple correlations to illustrate direct and indirect influences, offering a systemic view of how these factors interact. Practically, the findings offer actionable insights for strategic management by prioritising interventions. Organisations can strategically allocate resources by focusing on the identified cause factors (DAL and BSA) to achieve broader improvements in effect factors (DQ, DIT, and DCG). This suggests a sequential approach to developing a data-driven culture, moving from foundational elements to more outcome-oriented aspects. The originality of this research lies in its specific application of DEMATEL to this particular set of factors, challenging conventional assumptions about the independent importance of all factors and revealing a hierarchical dependency.

6. Conclusions

This study employed the DEMATEL method for identifying and prioritising the key factors necessary for effective, data-driven decision-making within organisations. The analysis revealed that DAL and BSA are causal factors. In contrast, DQ, DIT, and DCG were identified as effect factors.

These results suggest that improving DAL and BSA is likely to positively influence DQ, DIT, and DCG, thereby improving data-driven decision-making and organisational performance. The study provides a nuanced understanding of the causal relationships within the complex ecosystem of data-driven decision-making, offering actionable insights for strategic management.

Compared to the recent literature, classifying DAL and BSA as causal factors is consistent with the prevailing academic discourse [41], which often considers them prerequisites for successful data-driven initiatives. For instance, a strong DAL is essential for accurate data-driven outcomes, and an effective BSA drives the overall direction of the data efforts. Furthermore, DAL’s causal role is supported by studies focusing on the human and strategic foundations of data-driven transformation, which emphasise that a skilled workforce and clear strategic goals are essential for leveraging complex datasets. BSA as a causal factor is also consistent with literature recognising the essential role of strategic planning and alignment in achieving successful data outcomes.

Classifying DQ, DIT, and DCG as effect factors provides a more nuanced perspective. While many studies emphasise the importance of DQ, the DEMATEL analysis in this study indicates that DQ is significantly influenced by robust causal factors like DAL and BSA. This implies that efforts to improve DQ may be ineffective if the underlying human capabilities (DAL) and strategic direction (BSA) are not favorable. DQ appears to be a consequence of a well-established foundation in DAL and BSA. Similarly, while DIT and DCG are often presented as prerequisites for the success of data-driven initiatives, the DEMATEL results suggest that effective DIT and DCG can be achieved through the proper functioning of strategic alignment and data literacy. Improvements in DIT and DCG occur as the fundamental causal factors (DAL and BSA) are strengthened.

In conclusion, this study suggests that efforts to improve DQ, DIT, and DCG may be more effective if preceded by solid progress in DAL and BSA. This provides a structured framework for prioritising managerial interventions and suggests a sequential approach to cultivating an effective, data-driven culture, as well as optimising business processes for long-term sustainability and improved decision-making capabilities.

For organisations seeking to enhance their data-driven culture, these findings offer concrete guidance for both government and enterprises. Instead of simultaneously investing equally across all factors, a phased approach is recommended. The initial focus should be on ensuring a high DAL and establishing a strong BSA. For example, organisations should prioritise initiatives like comprehensive training programs for data interpretation and analysis, and developing clear data strategies that are tightly integrated with overarching business objectives. Once these foundational “cause” factors are strengthened, efforts to improve DQ through validation processes, enhance DIT through system upgrades, and foster a strong DCG will likely yield more significant and sustainable results. This “cause-first” strategy ensures that resources are efficiently allocated and that improvements in the underlying human and strategic ecosystem genuinely enable the desired outcomes of the enhanced DQ, DIT, and DCG.

Regarding limitations, this study’s findings are based on a small panel of expert judgments, which inherently introduces a degree of subjectivity. While this is a common practice in DEMATEL studies aiming for in-depth qualitative insights, it may affect the statistical generalizability of the findings across all organisational contexts [47]. Future research could address this by incorporating a larger and more diverse expert panel, or by utilising more advanced DEMATEL methodologies, such as Fuzzy-DEMATEL, to account for the inherent uncertainties and vagueness in human judgment, or integrating with other MADMs like ANP for a more comprehensive network analysis. Future research could also include longitudinal case studies in various industries to validate this sequential intervention strategy in real-world settings or explore a hybrid Fuzzy-DEMATEL–ANP approach to incorporate uncertainty and network dependencies more comprehensively, offering both theoretical and practical advancements.