Public higher education institutions in South Africa continue to grapple with low student retention and throughput rates, a challenge that disproportionately affects students from historically disadvantaged backgrounds (Chiramba & Ndofirepi 2023). Although reforms aimed at widening access to education post-apartheid, the higher education sector still suffers systemic inequalities, manifesting in underpreparedness for university-level study by first-time entering students, limited financial resources, and insufficient academic and psychosocial support (Department of Higher Education and Training [DHET] 2021a). Findings by the Council on Higher Education (CHE) indicate that only about 27% of students are able to complete their degrees within the minimum time, and at least 55% fail to complete at all. Black African students account for the lowest completion rates (CHE 2021). These outcomes underscore deep-rooted disparities in educational readiness and institutional responsiveness.

Although universities have implemented interventions such as extended curriculum programmes, National Student Financial Aid Scheme (NSFAS) support, and academic development services (Makura, Skead & Nhundu 2011; Manik & Ramrathan 2021), these efforts are often fragmented and not systematically linked to the timely identification of student risk (Cele 2021). In this regard, Big Data, with its predictive analytics capabilities, offers universities an important opportunity. Through the use of academic, financial, and engagement data to identify at-risk students earlier, these institutions can move from delayed support for students to a more proactive and targeted intervention. Although Big Data analytics provides the wider infrastructure for integrating and analysing diverse student datasets, predictive analytics is particularly important because it enables early warning, risk profiling, and more responsive decision-making for student retention (Cele 2021).

The world over, higher education institutions are using predictive analytics to strengthen student success strategies (Hollingworth et al. 2012; Umezuruike & Ngugi 2020; Villegas-Ch, Palacios-Pacheco & Luján-Mora 2020). However, in South Africa, adoption lags and is often constrained by fragmented systems, weak data integration, and governance concerns (Cele 2021). This study, therefore, asks and answers the following question: How can the use of Big Data analytics foster increased learner retention rates in public higher education institutions in South Africa? The study aimed to develop a Big Data framework that supports predictive identification of risk, personalised intervention, and broader institutional improvement in student retention.

Student retention remains a major concern in South African higher education, where expanded access has not translated into equivalent gains in student success. Enrolment rates have increased since 1994 to date, but throughput and completion rates remain weak, with many students failing to complete within the minimum time and others dropping out before graduation (CHE 2022a; DHET 2021, 2024). This challenge can, to a larger extent, be closely linked to structural inequality, uneven schooling quality, financial hardship, and differences in institutional capacity, particularly for students from historically disadvantaged and rural backgrounds (Hlongwane & Langa 2020; Kemp & Bush 2018; Mokhele 2021; Mouton & Bezuidenhout 2019). This literature shows that the central issue in South Africa is not only access to higher education, but also success within it.

While traditional retention interventions, such as first-year experience programmes, tutoring, mentoring, psychosocial support, and financial aid, remain important in improving student persistence, these interventions are often fragmented, labour-intensive, and insufficiently responsive to emerging risk (Manik & Ramrathan 2021; Tinto 1993). Cele (2021) finds that many institutions in South Africa still identify vulnerable students too late for effective intervention, despite the growing availability of student data through learning management systems (LMS), student information systems and administrative platforms.

Within this context, Big Data analytics has emerged as a promising means of strengthening student retention. Defined by volume, velocity, and variety, Big Data enables institutions to collect, analyse, and act on student-level data in real time (Madamshetty, Suresh & Naidu 2019; Philip Chen & Zhang 2014). When applied to education, Big Data can support predictive modelling, risk profiling, and personalised interventions, shifting retention strategies from reactive to proactive (Baker & Siemens 2014; Bichsel 2012). International and local studies suggest that Big Data analytics can support early warning, predictive modelling, risk profiling, and more targeted interventions by drawing on academic, behavioural, and financial data (Baker & Siemens 2014; Bichsel 2012; Daniel 2015; Kumar 2018). Yet the South African literature shows that adoption remains limited by fragmented data systems, weak interoperability, low data literacy, infrastructure constraints, and ethical and regulatory concerns, including compliance with the Protection of Personal Information Act (POPIA) (Cele 2021; Matto 2022). Existing frameworks also tend to emphasise technical processes while paying insufficient attention to governance, organisational readiness, and institutional context. This creates a need for a more context-sensitive framework capable of integrating diverse data sources, supporting timely intervention, and responding to the socio-technical realities of South African universities.

This research employed an interpretivist qualitative approach, supported by Design Science Research (DSR) methodology. Design Science Research was deemed suitable because the goal was not only to comprehend the problem identified in the study but also to design, demonstrate, evaluate, and refine a framework as an artefact for institutional implementation (Hevner et al. 2004; Peffers et al. 2008). The research problem stemmed from a comprehensive literature review, document analysis, and interviews. An outcome of the research emerged from the thematic analysis of data collected through interviews and institutional policy documents, strategic plans, and reports.

Thirty interviews were conducted at four South African universities: Fort Hare University, Walter Sisulu University, the University of South Africa, and Rhodes University. The participants consisted of academic personnel, student support professionals, and key institutional stakeholders focused on student success. Thematic analysis was used to analyse the collected data. Triangulation among interviews, documents, and existing literature informed the design of the framework. Finally, expert review represented a formal stage in the DSR evaluation, enhancing the relevance, clarity, and practical applicability of the developed framework.

Across the data analysed in the study, several interrelated themes emerged. These include student vulnerability and attrition risk, fragmented institutional support, limited analytics readiness by university institutions, and ethical and governance concerns. These themes were triangulated against documentary and literature sources. These themes are discussed in detail in the sections that follow.

The research results reveal that student retention in South African universities is influenced by a complex interplay of academic, financial, social, and institutional elements. Analysis of interviews and documentary evidence indicates that students face a heightened risk of dropping out because of issues such as lack of preparedness, financial difficulties, insufficient engagement, and inadequate institutional support. These results underscore that retention is not merely a single-factor issue but a multifaceted challenge that requires coordinated institutional approaches.

The study further discovered that existing institutional responses tend to be reactive, disjointed, and poorly coordinated. Although participants recognised tutoring, mentoring, counselling, financial assistance, and student support services as crucial interventions, these were often reported as inconsistently applied or disconnected from early identification processes. Concurrently, participants highlighted the potential usefulness of data on academic performance, LMS activity, attendance records, financial aid status, and engagement metrics for identifying at-risk students sooner. However, this information frequently resides in separate systems and is not consistently acted upon in a timely manner.

In addition, it was found that obstacles to analytics-driven retention strategies are not solely technical but also organisational and ethical in nature. Participants found issues such as outdated infrastructure, weak interoperability among systems, isolated departments, low levels of data literacy among staff, poor coordination efforts, and concerns regarding privacy rights and compliance with POPIA regulations. These insights suggest that the challenge lies not just in the lack of student data but also in the absence of an integrated framework governed by ethical standards for utilising such data effectively in retention strategies.

Guided by the TOE framework alongside Tinto’s (1993) model, the proposed framework in this study was directly informed by the empirical themes identified throughout the research. These themes were translated into concrete institutional and analytical requirements. Table 1 illustrates how key themes derived from existing literature were both validated and expanded upon through the study’s findings, thereby establishing a theoretical and empirical basis for designing the proposed framework.

This research finds that applying Big Data analytics can enhance student retention at South African universities. Using the TOE framework and Tinto’s (1993) model, the findings indicate that student dropout is driven by academic, financial, social, and institutional factors. In addition, the successful application of analytics depends on the preparedness of organisations, the capacity of staff, and adherence to ethical data handling. Therefore, the proposed framework helps in promoting better early detection, intervention strategies, and decision-making. For university leadership, the findings imply that student retention should be viewed as a shared endeavour at the level of the institution, supported by data integration, employee training initiatives, and collaboration among university departments, information and communication technology (ICT) services, and student support services. In terms of policy, this study highlights the need for governance mechanisms, equity considerations, and compliance with POPIA as tools for analytics to optimise student success. Thus, further research should investigate how such a framework can be applied across a variety of institutional contexts and how analytics-driven interventions can be effectively adopted and sustained over time.