The transformation within Higher Education Institutions (HEIs) in South Africa has led to challenges regarding the impact and success of the online registration process (Gaffoor & Van der Bijl 2019; Ngwato 2020). According to Salisu (2020), online registration in HEIs is an electronic format application that consists of specialised programmes that contain students’ information details and procedures that administrative employees and students require from first-time entry until graduation. Online registration simplifies the registration process by digitising registration business processes and tools in an online portal using a user-friendly wizard. Because institutional systems require information and documentation, the online registration solution allows HEIs to reduce paperwork processing, protect registration procedures and control access to files (Salisu 2020). According to Amini Valashani and Abukari (2020), Enterprise Resource Planning (ERP) systems that are most widely used for online registration are the integrator technology system, management information system, College Technology, Student Online Portal system, Microsoft Dynamics family, Oracle and SAGE system solutions. Higher Education Institutions in South Africa recognise the powerful transformational potential of these application portals and have developed and implemented institutional systems for online registration (Salisu 2020).

However, students at Technical Vocational Education and Training (TVET) Colleges in South Africa have challenges associated with the radical transformation of these ERP systems and the transformation to online registration (Nundkumar & Subban 2018). The transformation necessitated that TVET Colleges change their registration systems from manual to online registration, allowing students to register, cancel or add courses within the registration period. Access to online registration at TVET Colleges is accessible from anywhere and at any time, and is dependent on an Internet connection. Digitalising registration allows TVET Colleges to manage electronic records easily (Capuc & Atibuni 2018; Dhlamini 2018). Portz et al. (2019) claim that understanding users’ opinions and user experiences about the online registration system may improve the accessibility, acceptability and acceptance of students’ applications thereof.

This research aims to assess online student registration in HEIs in South Africa. The study intends to address the research gap by examining the impact of ERP systems on transforming online student registration at a TVET College using the DeLone and McLean Information Systems (D&M IS) success model. The D&M IS success model will interpret the perceptions of students regarding online registration and provide a basis for measuring the ERP systems implementation success.

Higher Education Institutions rely on fragmented ERP systems, which strengthens the ability to optimise processes such as online student registration (Ngonzi & Jisaba 2021; Shatat & Al Burtamani 2019). Research indicates that user perceptions, particularly in terms of ease of use, reliability and accessibility, are critical for the successful adoption and sustained use of ERP systems for online registration (Jegundo et al. 2020; Portz et al. 2019). Moreover, service quality and information accuracy significantly influence students’ satisfaction and willingness to engage with ERP platforms when registering online (Safsouf et al. 2020).

This study adopts and extends the D&M IS success model to evaluate ERP systems in online student registration within HEIs. The model integrates dimensions of information, system and service quality with user satisfaction, behavioural intention, and net benefits as key indicators of ERP success (Figure 1) (DeLone & McLean 1992, 2003). The D&M IS success model guided the investigation into the influence that ERP systems had on the online registration of students at HEIs.

The study adopted an objectivist ontology, suggesting that social phenomena are external factors that exist beyond influence or control. The positivist epistemological approach was adopted, where data were collected to establish a foundation for universal propositions through induction, aiming to uncover the truth through empirical means and testing hypotheses. A quantitative research approach was employed, and a survey was used to gather data. A cross-sectional research design was used to assess a selected population across all significant variables.

Demographic data revealed that the majority of respondents were young, with 72.55% aged between 18 years and 29 years. The largest group consisted of 1st-year students, making up 35.85% of the participants. Most students, 95.24%, were aware of the online registration system, and 87.96% had used it. Although 83.19% identified themselves as intermediate or advanced technology users, opinions on the system’s performance were mixed, with 45.65% rating it poorly. Nonetheless, 85.15% reported being familiar with the institutional registration support services.

The study assessed nine hypotheses to explore the relationships among system quality, information quality, service quality, user satisfaction, behavioural intention and actual usage in the context of ERP systems for online student registration. The study tested the research hypothesis with confirmatory factor analysis (CFA) and through the Kruskal–Wallis equality-of-populations rank test to determine whether there are statistically significant differences between the medians of three or more independent groups and the Two-sample Wilcoxon rank-sum (Mann–Whitney) test to compare two independent groups. Confirmatory factor analysis was employed to test the hypothesis among a set of observed variables on a specific number of factors that explain the relationship patterns. The utilisation of CFA confirms whether the data fit the hypothesised measurement model, which assesses the scales’ construct validity and determines the extent to which the observed variables represent the underlying theoretical constructs. Table 1 reflects the hypothesis testing.

In this study, all variables with Cronbach’s alpha values were included in the analysis. The results reveal that the Cronbach’s alpha coefficient consistently produced an average scale of 0.8262 and 0.9440, surpassing the minimum acceptable level of 0.7.

The goodness-of-fit analysis evaluates how well the proposed framework fits the actual data and provides insight into how closely the model aligns with the observed data (Idkhan & Idris 2023; Pavlov & Ferraz 2023). Table 2 reflects the goodness of fit for the model.

The goodness-of-fit statistics indicate that the model demonstrates a strong fit for system quality. The likelihood ratio tests show that the model significantly outperforms the baseline model, with the Chi²-bs test of 663.324 yielding a p-value of 0.000 compared to the Chi²-ms of 4.244 with a p-value of 0.120. Furthermore, population error statistics, such as root mean squared error of approximation (RMSEA) value of 0.056, 90% confidence interval (CI), lower bound 0.000 and 90% CI, upper bound 0.132 and p-close at 0.344. While the baseline comparison and size of residuals, such as Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) values of 0.997 and 0.990, respectively, also support a good fit. The fit statistics suggest that the model effectively represents system quality. In addition, the Standardised Root Mean Squared Residual (SRMR) value of 0.014 and CD value of 0.878 further confirm the model’s strong fit for system quality.

Upon reviewing the fit statistics of information quality, the proposed model significantly differs from the Chi²-ms and Chi²-bs models. The likelihood ratio Chi²-ms test of 143.842 resulted in a p-value of 0.000. Similarly, the Chi²-bs value of 1050.807 with a p-value of 0.000 suggests a noteworthy distinction between the baseline and saturated models. Although the proposed model shows some explanatory power beyond chance, it is important to consider other fit indices, such as population error, RMSEA 0.205, 90% CI, lower bound of 0.176 and 0.235 of 90% CI upper bound and 0.000 p-close. These indices suggest that the model’s fit to the data, particularly regarding information quality, may be suboptimal. For instance, the RMSEA value of 0.205, with a 90% confidence interval between 0.176 and 0.235, and the probability value of 0.000, suggest a substantial difference between the models, indicating a mediocre fit. The CFI and TLI values of 0.870 and 0.783, the SRMR value of 0.073 and the CD of 0.876 also indicate a less-than-ideal fit to the data. These results suggest that the model may not adequately represent the relationships between the observed variables and the information quality construct.

The likelihood ratio test on service quality indicates that the model fits the data well. The Chi²-ms value of 3.830 with a p-value of 0.147 did not reach statistical significance compared to the Chi²-bs value. Similarly, the Chi²-bs value is 1093.788 with a p-value less than 0.05, showing that the baseline model significantly differs from the saturated model. Regarding population error, the RMSEA value of 0.051 provides strong evidence of a high-quality fit. The RMSEA, comfortably below the threshold of 0.08, signifies a close fit between the model and the observed data with 90% CI lower bound at 0.000, 90% CI upper bound at 0.127 and p-close at 0.386. The baseline comparison results, with CFI and TLI values of 0.998 and 0.995, respectively, confirm a near-perfect fit. Lastly, the size of residuals, as indicated by the SRMR value of 0.009 and the CD value of 0.934, both point to an excellent fit. These results collectively demonstrate the model’s good fit for the data across likelihood ratio, population error, information criteria, baseline comparison and residual size, leaving no room for doubt about its performance.

Upon analysing the user satisfaction results, it is evident that the model demonstrates a strong fit across various statistical measures. The likelihood ratio Chi²-ms test 1.962 indicates a non-significant p-value of 0.375, suggesting that the model effectively captures the data compared to the saturated model, while Chi²-bs resulted in 1341.727 with the p-value 0.000. The RMSEA value of 0.000, with a 90% CI lower bound and upper bound between 0.000 and 0.104, support the conclusion of a well-fitting model. Moreover, the associated p-close (0.636) further confirms the model’s good fit. The baseline comparison shows perfect scores of 1.000 for both the CFI and TLI, indicating an excellent fit. This excellent fit should instil confidence in the model’s performance. Additionally, the SRMR value of 0.005 and the CD value of 0.946 underpin the argument for a well-fitting model. The results across likelihood ratio, population error, information criteria, baseline comparison and size of residuals collectively support the notion that the model accurately represents user satisfaction.

On the behavioural intention to use, the likelihood ratio results on the Chi-square test statistic show significant differences between the Chi²-ms 20.776, with p < 0.000 and between the Chi²-bs 585.758 and p < 0.000. The RMSEA is 0.162, with a 90% CI ranging from 0.104 to 0.229. The p-value for the RMSEA is 0.001, close to the information criteria, indicating its statistical significance.

The CFI is 0.968 and the TLI is 0.903. The SRMR is 0.043 and the CD is 0.892. The analysis reveals significant differences in likelihood ratio tests and suggests a reasonable model fit based on the population error and size of residual statistics. Equally important, the information criteria and baseline comparison play a significant role in assessing the model’s goodness of fit, providing a comprehensive understanding of the model’s performance.

The model demonstrates a good fit across various goodness-of-fit measures on construct actual usage. The likelihood ratio test suggests a good fit, with the Chi²-ms statistic indicating a value of 0.000 and 182.807 for the Chi²-bs comparison. The population error statistics, such as RMSEA with a value of 0.000, its 90% CI lower bound and upper bound resulted in both 0.000 and the p-close to 1.000, indicating a minimal error and a good fit of the model to the data. The CFI and TLI have high values of 1.000, indicating an excellent fit of the model to the data. The size of residuals, including SRMR with a value of 0.000 and CD with a value of 1.000, suggests minimal residual error and a good fit for the model. In conclusion, based on these measures, it may be inferred that the model exhibits a good fit based on the criteria of actual usage.

Results for the framework, as per Figure 2, indicate that while system quality positively correlates with user satisfaction, it paradoxically shows a negative relationship with users’ intentions to utilise the online registration system. Conversely, the quality of information emerges as a fundamental driver of user satisfaction and engagement with the online ERP systems. Users who find the information provided by the ERP system to be clear, relevant, and accurate demonstrate a significantly higher propensity to continue using the system. This observation is consistent with established theories, such as the theory of DeLone and McLean (1992:88) and DeLone and McLean (2003:23), which underscore the pivotal role of information quality in shaping user perceptions.

This study applied the D&M IS success model to evaluate the factors influencing the success of ERP systems in supporting online student registration in HEIs at a TVET College. The analysis focused on key constructs including system quality, information quality, service quality, user satisfaction, behavioural intention to use and actual system usage. The findings reveal that information quality significantly and positively influenced both behavioural intentions to use and user satisfaction, highlighting the critical role of accurate and relevant information in promoting ERP systems adoption. System quality and service quality significantly influenced user satisfaction but showed no significant effect on behavioural intention to use, suggesting that technical performance and support services contribute more to user contentment than to usage intentions. Notably, behavioural intention to use emerged as the strongest predictor of actual use. At the same time, user satisfaction demonstrated a significant but negative relationship with actual use, suggesting that higher satisfaction does not necessarily equate to increased system engagement. These results offer empirical support for the theoretical propositions of the D&M IS success model in a higher education context and provide a foundation for improving online registration relating to ERP systems. Enhancing information accuracy, system usability, and support services can contribute to more effective and sustainable online registration use in HEIs.