An ad hoc questionnaire was designed for university teaching staff, which measures: firstly, the use of a digital platform for various processes associated with in-person teaching; secondly, variables that may be related to the use of the platform; and finally, circumstances that could encourage the use of teaching platforms. An initial version of the questionnaire, which included a final open-ended question for suggestions about the instrument, was answered by a group of 11 lecturers from different areas (education, statistics, mathematics, business economics, law, languages), selected for their knowledge of the subject. Each member contributed their own insights, from perspectives such as specialisation in the application of information and communication technologies to higher education, expertise in questionnaire development, or extensive experience in using platforms for their own teaching and team coordination. The final questionnaire, described below, incorporates the suggestions received.

In formulating the questions about the use of each process, studies by Al-Hamed (2022), Esnaola- Arribillaga and Bezanilla (2020), Garrote and Pettersson (2007), Rhode et al. (2017), as well as the teaching experience of the authors and suggestions from the group commenting on the initial version, were considered. The questionnaire included 17 questions (see Table 1), all evaluated on a Likert scale from 1 (“Never”) to 5 (“Usually”). The questions correspond to tasks carried out with the support of teaching platforms. The first three questions (P1, P2, P3) relate to the distribution of materials. The next four pertain to personalised assessment: (P4, P5, P6) asynchronous assessment, and (P7), synchronous assessment. The subsequent questions deal with bidirectional communication: (P8 and P9), synchronous communication, (P10 and P11), asynchronous communication. Then, two further questions refer to organisational tasks (P12 and P13); two to monitoring tasks (P14, P15); and two questions to unidirectional communication (P16 and P17).

Regarding the variables that may be related to use of the platform, the literature shows multiple classifications (Salgado-Chamorro et al., 2023; Ziraba et al., 2020). According to the literature review by Al- Nuaimi and Al-Emran (2021), the most frequently considered variables are, in order, the TAM (Davis, 1989); the contribution of DeLone and McLean (2003); and the UTAUT, developed by Venkatesh et al. (2003). It was decided that it would be desirable not to discard any potentially relevant factors in the descriptive phase of the cluster analysis. Therefore, the variables proposed in the extensions of these theories were included, particularly for the use of teaching platforms in higher education, along with teaching vocation, suggested by the authors to provide a more comprehensive view of the participants’ professional self-perception. All the variables that can be directly influenced by external agents were grouped under factors that could promote the use of teaching platforms, to focus the data on the third objective of the work. As usual, demographic variables were collected and, less commonly (Al-Nuaimi & Al-Emran, 2021), so were contextual variables. The selected variables are shown in Table 2, along with the questions on the questionnaire and the most relevant references.

Non-probabilistic convenience sampling was used. After verifying the identity of the recipients, emails were sent with a link to the survey in Microsoft Forms. A total of 303 responses were obtained. After the data cleansing process, in which incomplete or invalid observations were eliminated, the final sample consisted of a total of 288 complete, valid, anonymised records corresponding to 288 lecturers who had accepted the corresponding informed consent.

Twenty-two Spanish universities are represented in the sample, with a high percentage of responses corresponding to teaching staff at the Complutense University of Madrid (37.8%) and the Pontifical Comillas University (37.5%). 18.1% correspond to other public universities, and 6.6% to other private universities.

Those responding have an average age of 50.7 years, with a standard deviation of 8.7 years. 54% identify as female and 46% as male. Their average experience is 19.8 years, with a standard deviation of 10.6 years. 25% have a part-time contract at the university.

The data analysis was performed in two different stages. In the first, the aim was to identify patterns based on the platform use dimensions indicated in Table 1. The methodology applied for this objective is hierarchical agglomerative clustering, an option chosen since it offers greater interpretability than non-hierarchical clustering algorithms. We used Euclidean distance and opted for Ward’s linkage method, one of the most commonly used in practice, since besides being robust to outliers, it tends to generate compact clusters. The presence of significant differences was confirmed with a Kruskal-Wallis test for differences between clusters, with post-hoc analysis for those variables where differences were detected. A non-parametric test was chosen, given that the conditions to assume a normal distribution are not met.

In the second stage, a descriptive analysis of each cluster was carried out. This included, firstly, the demographic, contextual, self-perception, and platform perception variables, with the aim of offering insights into their relationship with the use of teaching platforms. We identified which of these variables showed significant differences between clusters using a Kruskal-Wallis test, with post-hoc analysis in those variables where differences were detected. Finally, we described the circumstances mentioned by the teaching staff as possible drivers for promoting the use of teaching platforms within each cluster and tested the equal distribution among clusters of the selected circumstances using chi-square tests and likelihood ratios.

The analysis was performed using IBM SPSS and the R programming environment, employing various libraries: “dplyr” (Wickham et al., 2021), “dunn.test” (Dinno, 2024), “fmsb” (Nakazawa, 2019), “gplots” (Warnes et al., 2019), and “NbClust” (Charrad et al., 2014). The selection of these two tools is due to specific reasons. SPSS was used for the more basic analyses since it is an accessible and reliable tool that allows for quick and efficient statistical analyses. In contrast, for more advanced analyses, such as cluster analysis, R was chosen because it offers greater flexibility in parameterisation and allows for the generation of highly customised graphics.

The results of the cluster analysis are shown in Figure 1, based on the 17 previously standardised questions from Table 1. The presence of four different groups of lecturers is observed (horizontal interpretation of Figure 1). To facilitate the interpretation of the results, Figure 2 shows the radar chart of the four identified clusters.

Cluster 1 can be characterised as the group that tends to make the least use of the platforms. It includes only 8.7% of the teaching staff who responded. They regularly use the platform for tasks such as distributing materials (P1) or checking for plagiarism in submissions (P6). In contrast, Cluster 4, which comprises 31.9% of the teaching staff, is the group that tends to use all the features of the platforms more than the others.

Clusters 2 and 3, with 32.3% and 27.1% of the responses respectively, make intermediate use of the platforms. Differences between them emerge depending on the type of task and tool. The noticeable difference between the two clusters is in the habit of sharing material with other teachers through the platform (P3), which is more frequent in Cluster 2 than in Cluster 3. An interpretation for these results will be proposed in the discussion section of the article.

As expected, the Kruskal-Wallis test shows significant differences for all the questionnaire’s variables. The post-hoc analyses allow comparisons between specific processes (see Table 3, which includes the difference and the p-value with Bonferroni adjustment):

The description of the clusters concerning demographic variables is summarised in Tables 4 and 5.

The Kruskal-Wallis tests indicate that there are significant differences for both demographic variables (see Table 5). The post-hoc tests show that, in the case of age, Cluster C3 consists of younger teaching staff than the rest of the groups. Regarding gender, in the cluster that makes greatest use of the technologies (C4), there is a significantly higher percentage of women than in clusters C1 and C2. In both variables, we observe evidence contrary to some social stereotypes.

Table 6 summarises the description of the clusters based on the variables for the perception of the teaching platforms. Note that more than half of the teaching staff in Cluster C4 consider that the platforms decrease their workload, while only 20% of Cluster C1 share this opinion, which is consistent with the fact that these groups use them to a greater or lesser extent.

The Kruskal-Wallis test shows significant differences regarding the three platform perception variables (see Table 7). The post-hoc tests reveal that such differences are always related to clusters C1 and C4, which have the least and greatest use of the platforms. However, there is no significantly different perception between clusters C2 and C3, consisting of teaching staff who make intermediate use of the platforms, with different way they are use. These results seem to suggest that perceptions of the platforms among teaching staff are associated with their level of utilisation.

The scores for the self-perception variables for each cluster are presented in Table 8.

According to the Kruskal-Wallis test, no significant differences are observed for the teaching vocation variable, which has a very high score in all groups.

There are significant differences in the self-perception of technological competence. The post-hoc tests show that the score is significantly higher for Cluster C4 than for the rest of the clusters (see Table 9), which is consistent with it being the group that makes the greatest use of the platforms.

Table 10 summarises the measurements of the contextual variables from the questionnaire for each cluster. According to the Kruskal-Wallis test, significant differences appear only based on whether the university is public or private. Table 11 shows the results of the post-hoc analyses for this variable.

The clearest differences by type of university are observed between clusters C2 and C3, which do not differ very much in intensity of platform use but do differ in the way they use platforms.

Finally, the questionnaire asked the teaching staff to identify the circumstances that could promote the use of teaching platforms. Table 12 indicates the percentage of teaching staff in each cluster who selected each factor (the majority response in each cluster, which in all cases exceeds 40%, is highlighted in bold). The responses given to the open-ended question did not provide new avenues but tended to make their answers more specific, indicating, for example, the type of features they wanted from the platform.

One can observe the low rating that increased training receives across all the groups. When the proposal is for such training be more practical, the score increases considerably, and having technical support to respond to queries is valued even higher. No significant differences are observed between clusters in any of these three demands (see Table 13).

Overall, the most common demand among faculty members is having more time for teaching work. This is also the most frequent response in clusters C2 and C3.

In Cluster C1, however, the most limiting circumstance for the use of platforms is the difficulty of use (selected by 44% of the faculty members), which is consistent with the low perceived technological competence score of this group.

In Cluster C4, the most common response is greater recognition by the University (selected by 47.8% of the teaching staff). It is striking that, in the rest of the groups, this circumstance scores between 16% and 21.5%. In fact, the chi-square and likelihood ratio tests of equiprobability that have been conducted show significant differences at a 1% significance level, as can be seen in Table 13.