Factors Associated with the Success in the Adoption of Business Intelligence in Organizations: an Exploratory Approach Based on Multivariate Analysis.

Lucas Brandão Garcia; Miguel Ângelo Lellis Moreira

DOI: 10.22167/2675-6528-202603176

Article derived from a Course Conclusion Work (TCC), with content based on the student’s original work and adapted to the editorial format of the E&S Magazine with the support of the ResumeAI tool, an artificial intelligence solution developed by Instituto Pecege for textual synthesis and organization.

Summary

The adoption of Business Intelligence (BI) solutions in corporate organizations has expanded in recent years; however, many projects have not achieved the expected results. In this context, factors associated with the success of BI adoption were investigated, considering that the process involves technological, organizational, and behavioral aspects. The research adopted an applied character, a quantitative approach, and an exploratory nature, using a simulated database with technical and organizational variables related to the use of BI solutions. For the analysis, Multiple Correspondence Analysis (MCA) and clustering techniques were employed, aiming to identify distinct adoption profiles. The results indicated that organizations with similar technical structures may exhibit distinct success levels, associated with factors such as user engagement, data culture maturity, and solution utilization intensity. It was observed that a larger operational scale and intensity of use did not, by themselves, guarantee greater adoption success. The findings reinforced the multidimensional nature of BI adoption success and the contribution of exploratory multivariate approaches to organizational diagnostics and the improvement of the management of these initiatives.

Keywords: Organizational adoption; Multivariate analysis; Business Intelligence; Data culture; Analytical maturity.

1. Introduction

The use of data to support decision-making has become increasingly common in the corporate environment. In this context, Business Intelligence solutions play a central role by enabling the integration, processing, and visualization of strategic information from different data sources (Wixom; Watson, 2010; Mikalef et al., 2021; Grover et al., 2022). Despite technological advancements and the wide availability of analytical tools, many organizations still face difficulties in extracting effective value from these solutions.

The literature on Business Intelligence indicates that the success of these initiatives does not depend exclusively on the implemented technological infrastructure. Organizational factors, such as a data-driven culture, information governance, and user engagement, are also crucial (Yeoh; Koronios, 2010; Ransbotham et al., 2021; Olszak, 2022). The competitive advantage associated with the use of data lies less in the technology itself and more in the organizations’ ability to incorporate quantitative analyses and evidence into decision-making processes (Davenport; Harris, 2007; Davenport, 2023).

Recent studies, both in the Brazilian and international context, reinforce that the success of analytical and Business Intelligence initiatives results from the articulation between technical, organizational, and management factors. Analytical maturity and a data-driven culture are important for generating value from BI solutions (Mikalef et al., 2021; Wamba et al., 2021; Olszak, 2022). It is observed that organizations with similar technological structures and resources can present quite distinct levels of use, engagement, and perceived return from analytical tools.

Professional experience in the technical support of corporate organizations during the adoption and maintenance processes of Business Intelligence solutions shows that the success of these initiatives results from the interaction between technical and human factors. This complexity suggests that successful adoption is a multidimensional phenomenon, which cannot be understood through isolated analyses of variables.

Given this scenario, it becomes relevant to employ analytical approaches capable of simultaneously examining multiple factors associated with the adoption of Business Intelligence. Multivariate statistical methods allow for the identification of patterns, associations, and behavioral profiles among variables, enabling the analysis of complex organizational phenomena that would hardly be captured by isolated univariate or descriptive analyses (Hair et al., 2019; Shmueli et al., 2020). Thus, this work aims to identify factors associated with the success in adopting Business Intelligence solutions in corporate organizations through the application of multivariate statistical analysis techniques on a simulated database.

2. Material and Methods

This research has an applied character, with a quantitative approach and exploratory nature. Studies of this nature are appropriate for understanding relationships between variables and identifying patterns in complex organizational contexts, especially when the objective involves the exploration of phenomena that are still little structured in the literature (HAIR et al., 2019).

The study was conducted from the construction of a simulated database, designed to represent organizational units using corporate Business Intelligence solutions. The adoption of simulated data resulted from the restriction of access to real organizational databases, even after anonymization attempts. The use of synthetic data is recognized as a pertinent methodological alternative in contexts of confidentiality and privacy protection (PATEL, 2024; BEDUSCHI, 2024; MENDES et al., 2025).

The database was generated in a Python environment, using the NumPy libraries for generating statistical distributions and Pandas for structuring. Continuous usage behavior variables were simulated from plausible distributions, and categorical variables, such as organizational maturity and engagement, were generated to represent different levels. Fictitious records were created with technical and organizational attributes observed in BI projects. To ensure reproducibility, a fixed random seed was defined in the generation procedures. The complete script is documented in Appendix A of the original work.

The database contained records of 500 fictitious organizational units, representing different adoption profiles of Business Intelligence solutions. The variables were divided into two groups: technical and organizational. Technical variables included solution cost, number of users, platform utilization volume, number of support calls, implementation time, and account age. Organizational variables included user engagement level, data culture maturity, risk of solution discontinuation, and adopted support model.

To complement the analysis, a synthetic indicator of success in the adoption of Business Intelligence was constructed, combining the adoption rate of the solution among potential users, the utilization time of the analytical platform, and the risk of discontinuation (churn risk). The continuous indicators were normalized through the Min-Max transformation, according to equation (1), so that the values varied between zero and one (SHMUELI et al., 2020). After normalization, the three dimensions were combined by simple arithmetic mean, according to equation (2).

Before applying multivariate statistical analysis techniques, variables were prepared and categorized. Simulated quantitative variables were transformed into ordinal categories, such as low, medium, and high levels, to represent different observed levels in the organizational context. Variables such as number of users, usage volume, and solution cost were categorized based on defined intervals derived from the distribution of simulated data. Indicators of engagement, data culture maturity, and risk of discontinuation were represented by categorical levels.

After preparation and categorization, Multiple Correspondence Analysis (MCA) was applied, a multivariate statistical technique to examine associations between multiple categorical variables and represent them in a reduced dimensional space (GREENACRE, 2017). MCA starts with the construction of a complete indicator matrix or Burt matrix, which synthesizes the cross-frequencies between categories, according to equation (3). The technique performs a singular value decomposition (SVD), according to equation (4), to identify latent dimensions that synthesize the main association patterns (GREENACRE, 2017; HAIR et al., 2019).

Based on the factorial coordinates obtained by PCA, a cluster analysis was applied with the K-means algorithm to identify homogeneous profiles of Business Intelligence adoption. K-means seeks to minimize the internal variability of groups and maximize the separation between distinct clusters (HAIR et al., 2019; KAUFMAN; ROUSSEEUW, 2005). The optimization criterion consists of minimizing the sum of the squared distances between each observation and the centroid of its respective group, according to equation (5).

The definition of the adequate number of groupings was supported by the analysis of the average silhouette coefficient and the elbow method. The silhouette coefficient assesses the internal cohesion of the groups and the separation between distinct clusters, according to equation (6). The joint analysis of these metrics showed a trade-off between statistical quality and analytical detail, leading to the choice of eight clusters (k = 8) for segmentation, aiming for greater granularity and managerial interpretation.

3. Results and Discussion

The exploratory analysis of the simulated data revealed consistent patterns related to the adoption and use of Business Intelligence solutions in corporate organizational units. The application of Multiple Correspondence Analysis (MCA) proved to be a suitable technique for examining associations between categorical variables, allowing for the reduction of dataset complexity and the preservation of the most relevant relationships among the analyzed attributes. This multivariate approach was fundamental to understanding the complex nature of the phenomenon, as highlighted in the methodology, aligning with the need to explore multifaceted organizational phenomena.

The interpretation of the dimensions generated by the ACM indicated the existence of two main axes of differentiation between the organizational units studied. The first dimension showed a predominant association with variables related to scale and operational complexity. Units characterized by high cost categories, greater volume of solution usage, a larger number of active users, and a greater quantity of support calls were concentrated at one extreme of this axis, reflecting the size and intensity of use of Business Intelligence. At the opposite pole, units with lower intensity of use, a smaller user base, and lower operational complexity were grouped, suggesting clear structural differences between the analyzed environments.

The second dimension showed a distinct pattern, mainly associated with organizational and behavioral factors. It was observed that categories related to low user engagement, initial analytical maturity, and higher risk of solution discontinuation contributed more to one pole of the dimension. In contrast, units with higher levels of engagement, more consolidated analytical maturity, and lower perceived risk of solution discontinuation were concentrated at the opposite pole. These results indicate that, regardless of operational scale, factors related to analytical culture and the organizational relationship with the solution play a relevant role in differentiating the observed profiles, corroborating the literature that points to the importance of non-technological aspects for BI success (Yeoh; Koronios, 2010; Ransbotham et al., 2021).

Together, the two dimensions obtained by MCA show that the adoption of Business Intelligence cannot be understood from a single isolated factor. While the first dimension reflects structural and operational differences between organizational units, the second dimension indicates the relevance of organizational factors associated with data culture and user engagement. The projection of organizational units in the two-dimensional MCA space showed a consistent visual separation between different groups, indicating that the associations between categorical variables are effective in distinguishing distinct patterns of Business Intelligence adoption, which was a crucial step for the segmentation stage.

To identify organizational profiles with similar characteristics of analytical solution adoption, clustering methods were applied to the dimensions obtained by Multiple Correspondence Analysis. The definition of the adequate number of clusters was supported by two complementary approaches: the average silhouette coefficient and the elbow method. The silhouette coefficient, which assesses internal cohesion and separation between clusters, indicated that smaller values of k, especially k equal to two and k equal to three, presented higher silhouette indices, suggesting better statistical separation between the groupings. However, the silhouette coefficient showed a gradual reduction as the number of clusters increased, reflecting greater complexity and overlap between the formed groups.

The elbow method, in turn, allowed for the analysis of the reduction in internal variability of the groupings as the number of clusters increased. A sharp reduction in the internal variability of the clusters was observed up to approximately k equal to four or k equal to five, after which the curve began to show more stable behavior. This pattern indicated an inflection point that suggests decreasing marginal gains in segmentation quality for higher values of k. The joint analysis of these two metrics highlighted a trade-off between the statistical quality of the segmentation and the level of analytical detail.

Given this scenario, the final choice for k equal to eight was based on the combination of statistical and analytical criteria. Although the graphs indicated better statistical performance for smaller values of k, the configuration with eight clusters allowed for greater granularity in identifying organizational profiles, favoring the managerial interpretation of the results. Configurations with a smaller number of clusters would tend to group organizational units with heterogeneous characteristics into broad categories, reducing the capacity for differentiation between adoption patterns. On the other hand, very high values of k could compromise the stability of the groupings and hinder their practical interpretation, which justified the choice of an intermediate number of clusters to optimize the practical relevance of the findings.

After defining the number of clusters, the distribution of organizational units among the identified clusters was analyzed. This distribution allowed for observing the quantity of organizations allocated to each group and evaluating the balance between the groupings obtained by the model. The cluster analysis revealed the existence of heterogeneous profiles of Business Intelligence adoption, indicating that organizations do not follow a uniform path in the implementation and use of these solutions.

Groups characterized by high scale and elevated operational complexity were identified, as well as medium-sized groups with more stable use of the solution. Groupings associated with organizational environments with limited adoption were also observed, in which low levels of engagement and analytical maturity were associated with greater instability in BI use. These results indicate that the adoption of Business Intelligence cannot be explained exclusively by technical or structural factors, being strongly influenced by the interaction between organizational and behavioral characteristics, a perspective aligned with DeLone and McLean’s (2003) view on information systems success.

The cluster analysis also allowed the groupings to be interpreted as different organizational profiles of Business Intelligence adoption, representing distinct stages of maturity and intensity of use of the analytical solution. To examine the differences in operational scale between the groupings, the average monthly cost of the solution in each cluster was analyzed. The distribution of average costs among the clusters reinforces the interpretation that the different adoption profiles also distinguish themselves in terms of operational scale and intensity of use of the analytical solution, with Cluster 2 presenting the highest average monthly cost, indicating greater investment and, presumably, greater scale.

The average success score for each identified cluster revealed that certain groupings concentrate organizational units with higher levels of the indicator, suggesting environments with greater analytical maturity and greater consolidation of the use of Business Intelligence tools in decision-making processes. On the other hand, some clusters presented significantly lower values of the indicator, reflecting organizational contexts in which the adoption of the solution occurs in a more limited way or is still in an initial stage. For example, Cluster 7 presented the highest average success score, while Cluster 5 presented the lowest.

To understand the characteristics of each cluster in more detail, an analysis of the average profile of the main variables considered in the study was performed, including the number of users, solution usage volume, support frequency, and performance indicators. Cluster 2, for example, concentrated the largest number of organizations (122 units) and presented the highest average values for users (260.02), usage volume (1470.78 monthly hours), and support tickets (9.96 per month), indicating environments of greater operational complexity. However, its average success score was 0.45, remaining relatively lower than observed in clusters with smaller operational scale.

In contrast, Cluster 7, with 38 organizations, presented the highest average success score (0.72), along with a high average adoption rate (0.74) and low volume of support tickets (0.24 per month), suggesting more stable organizational environments with greater maturity in the use of the analytical solution. Cluster 5, with 49 organizations, presented the lowest average success score (0.27) and shortest average account age (21.00 months), indicating a more recent organizational profile that is still not well consolidated in terms of analytical adoption. These direct comparisons between clusters reveal the diversity of adoption scenarios.

A particularly relevant result emerges from the analysis of Cluster 2. Although this grouping presents the highest average values for the number of users, platform usage volume, and frequency of support calls, its average success score remains lower than that observed in smaller clusters. This result suggests that greater operational scale and higher intensity of platform use do not, in themselves, guarantee greater success in the adoption of Business Intelligence. In more complex organizational environments, the expansion of usage scale may occur without the corresponding consolidation of analytical maturity, user engagement, and stability of solution adoption, which highlights the importance of qualitative factors.

Another aspect analyzed in this study was the relationship between the adoption rate of the analytical solution and the risk of churn, understood as the probability of discontinuing the use of the tool. It was observed that higher adoption levels tend to be associated with a lower probability of discontinuing the use of the platform. However, the observed dispersion indicates that this relationship does not occur in a perfectly linear way, suggesting that other organizational factors, such as the maturity of the analytical culture, the quality of technical support, and the integration of analytical tools into decision-making processes, also exert relevant influence on the continuity of the solution’s use, as pointed out by Ransbotham et al. (2021) and Olszak (2022).

In general, the identified clusters can be interpreted as representing three main patterns of organizational behavior regarding the use of Business Intelligence. The first corresponds to organizational environments of advanced adoption, characterized by greater stability in the use of the solution and higher analytical maturity. The second pattern corresponds to environments with intermediate adoption, in which Business Intelligence presents relatively stable use, but with moderate levels of engagement and maturity still under development. Finally, groupings were identified associated with organizational contexts of initial or limited adoption, characterized by lower intensity of use and higher risk of solution discontinuation, evidencing the multidimensional nature of adoption success.

These results indicate that success in the adoption of Business Intelligence does not depend exclusively on operational scale or solution utilization intensity, but also on the presence of organizational factors related to data culture and user engagement. The combination of Multiple Correspondence Analysis and clustering methods allowed for the reduction of dataset complexity, identification of relevant differentiation dimensions, and segmentation of organizational units into distinct adoption profiles. The analysis highlighted that technical factors, such as scale of use and operational complexity, coexist with organizational factors, such as user engagement and data culture maturity, in configuring the different levels of success observed, confirming the study’s objective.

4. Conclusion

This study aimed to identify factors associated with the successful adoption of Business Intelligence solutions in corporate organizations, employing an exploratory approach based on multivariate analysis on a simulated database. It was found that success in the adoption of Business Intelligence is a multidimensional phenomenon, influenced by a complex interaction of factors. The application of Multiple Correspondence Analysis revealed two main axes of differentiation between organizational units: one related to operational scale and complexity, and another associated with organizational and behavioral factors, such as user engagement and data culture maturity. It was observed that organizations with similar technical structures can exhibit distinct levels of success, and that greater operational scale or intensity of solution use do not, in themselves, guarantee greater adoption success. Cluster segmentation allowed for the identification of heterogeneous organizational profiles, from environments of advanced and stable adoption to contexts of initial or limited adoption, with a higher risk of discontinuation.

The main contribution of this work lies in demonstrating that multivariate analytical approaches, such as the combination of Multiple Correspondence Analysis and clustering methods, are effective in producing relevant diagnostics on Business Intelligence adoption patterns. This capability to identify complex profiles and associations supports managerial decision-making and the formulation of strategies to consolidate the analytical culture within organizations. However, the study presents important limitations, such as the use of a simulated database, which implies that the identified patterns should be interpreted as plausible associations rather than definitive empirical evidence for real-world contexts. Furthermore, the exploratory nature of the methodology does not allow for establishing direct causal relationships between the analyzed variables, and subjective dimensions, such as users’ perceived value, were not incorporated. For future studies, it is suggested to apply the methodology to real databases, include new organizational variables, and develop explanatory models that deepen the analysis of factors associated with the success of Business Intelligence adoption.

Bibliographic References

BEDUSCHI, 2024 [Referência completa não encontrada no documento original]

DAVENPORT, T. Competing on analytics revisited. Harvard Business Review, 2023.

DAVENPORT, Thomas H.; HARRIS, Jeanne G. Competing on analytics: the new science of winning. Boston: Harvard Business School Press, 2007.

DELONE, William H.; MCLEAN, Ephraim R. The DeLone and McLean model of information systems success: a ten-year update. Journal of Management Information Systems, Armonk, v. 19, n. 4, p. 9–30, 2003.

GREENACRE, 2017 [Referência completa não encontrada no documento original]

GROVER, V. et al. Creating strategic business value from big data analytics. MIS Quarterly Executive, 2022.

HAIR, Joseph F. et al. Multivariate data analysis. 8. ed. Boston: Cengage Learning, 2019.

KAUFMAN; ROUSSEEUW, 2005 [Referência completa não encontrada no documento original]

MENDES et al., 2025 [Referência completa não encontrada no documento original]

MIKALEF, P. et al. Big data analytics capabilities and firm performance. Information & Management, 2021.

OLSZAK, C. Business intelligence systems and organizational performance. Information Systems Management, 2022.

PATEL, 2024 [Referência completa não encontrada no documento original]

RANSBOTHAM, S. et al. Expanding Al’s impact with organizational learning. MIT Sloan Management Review, 2021.

SHMUELI, Galit; BRUCE, Peter; GEDECK, Peter; PATEL, Nitin. Data mining for business analytics. Hoboken: Wiley, 2020.

WAMBA, S. F. et al. Big data analytics and firm performance. Journal of Business Research, 2021.

WIXOM, Barbara H.; WATSON, Hugh J. The BI-based organization. International Journal of Business Intelligence Research, v. 1, n. 1, p. 13–28, 2010.

YEOH, William; KORONIOS, Andy. Critical success factors for business intelligence systems. Journal of Computer Information Systems, Stillwater, v. 50, n. 3, p. 23–32, 2010.

Article originating from the Final Course Project of the Specialization in Data Science and Analytics of the MBA USP/Esalq

To learn more about the course, click here and access the MBX Academy platform