Determinants of Student Loyalty and Word of Mouth in Dual VET Secondary Schools in Bulgaria

Abstract

In response to the growing importance of vocational education for youth employability, this study examines students’ perceptions of dual vocational education and training (dVET) in Bulgaria, focusing on the following determinants of student loyalty (SL) and word-of-mouth communication (WOM) in the secondary education context: brand associations, brand relevance, brand image, image of dVET, service quality, and student satisfaction, based on previously validated scales adapted to the Bulgarian context. A structured questionnaire was administered to a target population of 608 students across nine vocational secondary schools in the Plovdiv region. A total of 507 usable surveys were collected from students in 11th and 12th grades who were actively participating in work-based learning. Data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) with the SmartPLS 4 software. The findings indicate that brand image is the strongest direct predictor of the image of dVET. Furthermore, student satisfaction stands out as the most influential antecedent of WOM. The indirect pathways from service quality to both SL and WOM, mediated by student satisfaction, underscore the pivotal role of satisfaction as a transmission mechanism. The study contributes to the limited empirical research on branding in dVET and offers insights for policymakers, school administrators, and employers seeking to improve the attractiveness of these pathways.

1. Introduction

In recent years, dVET, a specific form of vocational education and training (VET; see consolidated list of abbreviations), has emerged as a focal point for educational reforms across Europe. This model has gained traction as a strategic approach to reducing youth unemployment and aligning education with labor market demands. Among the internationally recognized countries that have successfully implemented dVET through the integration of theoretical and practical learning are Germany and Austria (Cedefop & ibw Austria, 2023; Schwede et al., 2025). By contrast, Bulgaria is still in the early phases of adopting dual training within its upper secondary vocational schools (ISCED 3), having introduced it less than a decade ago.

Bulgaria’s VET system operates using two main models: traditional VET and dVET. The traditional model is implemented mainly within the secondary schools, with practical training taking place in school workshops or through short internships, and the connection with employers is often episodic and limited in scope. In contrast, dVET combines school-based learning with systematic and prolonged practice in a real work environment provided by partner companies. Employers play a key role not only in providing practical training, but also in adapting curricula to labor market requirements and often participating in the process of student assessment. The share of the practical lerning component is significantly higher in dual education, and students receive an employment or civil-law contract, as well as remuneration for their work. These structural differences imply different expectations regarding the professional preparation, motivation, and graduate outcomes. In 2023, several important measures were undertaken to support dVET (European Commission, 2024), but no public evaluation or outcome indicators have yet been released to assess the progress. In the 2022–2023 school year, dVET was conducted in 182 educational institutions, with a total of 11,609 students enrolled in the dual training. However, a significant issue was noted—a relative decrease in the number of students enrolled in dVET during the first year of training (8th grade), as well as the decline in the number of dVET students who, after completing the first stage of upper secondary education (10th grade), continue into the second stage (11th grade). This trend underscores the need for further analysis of these phenomena (Ministry of Education and Science, Bulgaria, 2023).

For more than two decades, numerous studies have highlighted the growing importance of branding in education. This evolution is propelled by globally prevalent phenomena such as intensifying competition (Eger et al., 2018; Sholihah, 2023), rising parental expectations (Klomtooksing & Sato, 2023), marketization of education (Lundahl et al., 2013; Waslander et al., 2010), ranking pressures and standardized performance metrics (Nichols & Berliner, 2007; Verger et al., 2016), projections of demographic decline (Organisation for Economic Co-Operation and Development, 2024), fragmentation of regulatory responsibilities in the education system (United Nations Educational, Scientific and Cultural Organization, 2021), and the digitization of media and public communication (Capriotti & Zeler, 2023; Perera et al., 2022; Peruta & Shields, 2016; Zeqiri et al., 2023). Together, these factors challenge institutions to build a coherent brand identity alongside a desired brand image; manage online reputation and crises; translate the brand promise into tangible improvements in teaching and learning; monitor satisfaction and loyalty levels among students and other stakeholders; mobilize staff as brand ambassadors; and uphold ethics, inclusion, and equitable access for all. There is ample empirical evidence that the classic brand dimensions (including awareness, associations, perceived quality, and loyalty) as well as satisfaction and word of mouth are valid and predictive, predominantly in the higher education sector (Alves & Raposo, 2010; Bruce & Edgington, 2008; Chen, 2016; Greenacre et al., 2014; Helgesen & Nesset, 2007; Kairat et al., 2024; Mourad et al., 2011; Pinar et al., 2014), whereas evidence from secondary technical and vocational education and training (TVET) remains limited (Nursetiani & Wahyuni, 2024; Satria & Hidayat, 2018). Still, several previous studies should be mentioned that focus on secondary school students’ perceptions of TVET image and loyalty (Awang et al., 2011; Dang & Hathaway, 2014; Dang & Hathaway, 2015); student loyalty as a mediator linking service quality and institutional image to WOM (Andriana et al., 2023); the direct and indirect (via the mediating role of student satisfaction) effect of service quality on WOM (Nursetiani & Wahyuni, 2024; Rusdi & Ali, 2020); and a moderating effect of WOM on the relationship between brand image and the decision to choose school institutions (Pradana & Nurali, 2020).

This paper seeks to partially address this gap, with a focus on the Bulgarian context. Specifically, it aims to explore the interrelationships between brand associations, brand relevance, the brand image of secondary dVET schools, the image of dVET, service quality, student satisfaction, student loyalty, and word-of-mouth communication using structural equation modeling (SEM). The research was conducted in vocational secondary schools in the Plovdiv district, yielding 507 valid questionnaires from students in the final two years of their education who were actively engaged in work-based learning.

Grounded in consumer-based brand equity (CBBE), SERVQUAL, and the satisfaction–loyalty paradigm, we link branding scholarship to student outcomes in dVET by arguing that classic brand dimensions shape satisfaction and, in turn, loyalty and WOM. We also address unresolved issues in the upper-secondary TVET literature by clarifying the integration of brand associations and brand relevance, the relationship between schools’ brand image and the system-level image of dVET, and the pathway from service quality, mediated by satisfaction, to loyalty and WOM. This study offers two key contributions. First, it expands the understanding of the phenomenon of students’ loyalty and WOM to vocational secondary schools offering dual education by providing empirical data on the perceptions of students enrolled in dVET programs. Second, by applying this conceptual framework for the first time in Bulgaria, the study not only enriches the existing research with a new cultural context but also identifies key factors that can inform the development of effective educational policies and communication strategies to promote student motivation and retention within the dual education system.

2. Theoretical Framework

2.1. Theoretical Foundations

Our study integrates constructs from three influential theoretical traditions: brand equity (BE), service quality, and the satisfaction–loyalty paradigm, and tests them in the specific context of dVET. In essence, BE measures the value of a brand to both the organization and the consumer (Mourad et al., 2011). The two foremost models of consumer-based brand equity were conceptualized by Aaker (1991) and Keller (1993) and have been widely applied and supported as a multidimensional construct (Rojas-Lamorena et al., 2022). According to Aaker’s account, BE arises from five categories of brand assets and liabilities: brand awareness, brand associations, perceived quality, brand loyalty, and other proprietary brand assets. In Keller’s knowledge-based model of CBBE, the core dimensions are brand awareness and brand image, which together shape consumer responses.

Conceptualized within the services literature, service quality is now widely accepted as critical for organizations because it drives marketing and financial performance (Calvo-Porral et al., 2013). One of the earliest scholars to emphasize the need to study service quality was Grönroos (1984), who provided an early conceptualization of the construct. Over the years, numerous models and instruments have been proposed to measure service quality, but reaching consensus remains difficult because of the concept’s complexity, with the lack of agreement partly attributed to the so-called ‘multidimensional problem’ (Martínez & Martínez, 2010). Parasuraman et al. (1988) argue that perceived service quality reflects the degree and direction of the gap between consumers’ expectations and perceived performance.

Within the satisfaction–loyalty paradigm, customer loyalty is often perceived as the main consequence of customer satisfaction (Helgesen & Nesset, 2007). Foundational research shows that satisfaction translates into retention (Anderson & Sullivan, 1993; Oliver, 1999) and positive word of mouth (Fornell, 1992), thereby providing the behavioral logic that links quality perceptions to loyalty outcomes. In services, the pathway from perceived service quality to satisfaction and, in turn, to customer loyalty is well established (Bloemer et al., 1998; Caruana, 2002; Dahiyat et al., 2011; Leninkumar, 2017).

In the following subsections, we define the constructs we employ, state the hypotheses, and present the research model. Relative to prior work focused on higher education, our model is tailored to the secondary-level dual VET context. This allows us to theorize a cross-level link from a school’s brand image to the system-level image of dVET and to test satisfaction as the transmission mechanism from perceived service quality to loyalty and WOM.

2.2. The Influence of Brand Associations (BA) and Brand Relevance (BR) on Brand Image (BI) of Secondary dVET Schools

Brand image can be defined as a subjective mental picture of a brand shared by a group of consumers (Riezebos et al., 2003). It is expressed through perceptions about a brand as reflected by the brand associations held in consumer memory (Keller, 1993). In the context of education, “brand image” and “corporate image” are often used interchangeably or tend to overlap in meaning. The better the brand image, the easier it will be for educational institutions to attract customers and gain public attention. Therefore, the brand image strategy in education is essential because the development and growth of education are determined by the ability to manage educational institutions effectively (Sholihah, 2023). The brand image established by schools refers to the overall perception of a particular vocational secondary school as a brand, shaped by its material attributes (e.g., building facilities, environmental resources, and curriculum delivery) and symbolic elements (e.g., teaching philosophy, school culture, and student success), based on the conceptualization of educational brand image proposed by Chen (2016). The findings of Sintani (2020) indicate that students’ motivation to choose an SMK (Sekolah Menengah Kejuruan, i.e., a vocational high school in Indonesia) is significantly influenced by the school’s image.

Previous research supports the positive and significant influence of brand association on brand image (Isnawijayani et al., 2025; Khan & Jalees, 2016). BA are distinct attributes that highlight and create positive perceptions in the mind of customers (Latif et al., 2015). In the education sector, brand associations among students can be formed, shaped, and transformed through daily interactions with teachers or professors, administrative staff, mentors, employers, alumni, parents, and other relevant stakeholders, promotional activities, and campaigns conducted by educational institutions. These associations may represent the underlying motives behind students’ educational choices. They also facilitate the perception, processing, interpretation, and dissemination of information related to the institution’s brand. Empirical evidence is available confirming the significant influence of brand associations on students’ decision-making process in choosing a secondary school (Irpansyah et al., 2023).

Another antecedent of the brand image of an educational institution, for which there is empirical evidence of a statistically significant and positive influence, is brand relevance (Le, 2021; Tran & Duc, 2022). BR in the context of VET schools reflects the commitment of the vocational organization to meet the needs and aspirations of students and broader stakeholders, demonstrating alignment with shared values and long-term goals, such as sustainable development and the empowerment of future generations.

Based on the above, the following hypotheses are proposed:

Hypothesis 1 (H1).

Brand associations have a positive effect on the brand image of secondary dVET schools.

Hypothesis 2 (H2).

Brand relevance has a positive effect on the brand image of secondary dVET schools.

2.3. The Influence of Brand Image of Secondary Vocational Schools on Image of dVET (IDVET)

The image of vocational education and training is one of the factors that plays an influential role in students’ decisions to enroll in these programs (Awang et al., 2011). Based on the conceptualization proposed by Dang and Hathaway (2015), we define IDVET as the sum of participants’ beliefs and attitudes toward dual vocational education and training as a system, shaped by their experiences within the educational process at their school and their overall perception of dVET as a learning model. Developing a framework that accurately represents the image of VET and can measure students’ perceptions is a very complex and tedious task (Dang & Hathaway, 2014). Nevertheless, among the notable efforts in this area is the work of Awang et al. (2011), who examined the perceptions of secondary vocational students in TVET schools in Malaysia. Their model of the image of VET consisted of seven dimensions used to measure students’ perceptions, namely: low entry qualification, trainer credibility, applicability of course content, quality of training facilities and equipment, recognition of qualification, career and job potential, and work ethics and social value. To the best of our knowledge, the relationship between the brand image of vocational secondary schools and the overall perceptions of dVET has not been explored so far. However, given the established effect of school image on students’ perceptions (Eger et al., 2018; Klomtooksing & Sato, 2023), we assume that BI will have a positive effect on the IDVET.

Accordingly, we propose the following hypothesis:

Hypothesis 3 (H3).

The brand image of secondary dVET schools has a positive effect on the image of dVET.

2.4. The Influence of dVET Image and Service Quality (SQ) on Student Satisfaction (SS)

Student satisfaction in the context of dVET can be described as an emotional and cognitive response to the fulfillment or non-fulfillment of prior expectations regarding the institution’s performance, based on the experiences during the study. This understanding aligns with the view that students are treated as customers of an educational institution, and their satisfaction is essentially regarded as synonymous with customer satisfaction (Haverila et al., 2021). It has been argued that customer satisfaction depends on the assessment of the price and quality of the product or service provided. A substantial body of empirical research in the education sector has consistently supported a statistically significant positive effect of service quality on student satisfaction (Doan, 2021; Hasan et al., 2008; Hassan et al., 2019b; Hassan & Shamsudin, 2019; Nursetiani & Wahyuni, 2024). Monitoring the quality of educational services within vocational context is pivotal for enhancing educational outcomes and aligning with government priorities (Guo et al., 2024). While Doan (2021) describes service quality as a complex and multifaceted concept in the field of higher education, this perspective is equally relevant to dVET institutions, where educational services also encompass various interrelated components. Perhaps one of the most popular instruments for measuring service quality in the service industries is SERVQUAL (Parasuraman et al., 1988), which has been frequently adapted for use in educational contexts, including vocational schools (Hassan et al., 2019a; Nursetiani & Wahyuni, 2024). It consists of five dimensions: tangibles, reliability, assurance, responsiveness, and empathy.

Regarding the relationship between the construct of dVET image and SS, it should be noted that previous empirical studies have confirmed that certain dimensions of the image of vocational education and training have a direct positive effect on students’ loyalty (Awang et al., 2011; Dang & Hathaway, 2015). However, to the best of our knowledge, research in this specific direction remains highly limited. We assume that the image of dual vocational education and training may influence students’ satisfaction as an independent outcome, regardless of whether it results in loyalty toward their vocational school.

Thus, the following hypotheses are proposed:

Hypothesis 4 (H4).

The image of dVET has a positive effect on student satisfaction.

Hypothesis 5 (H5).

Service quality has a positive effect on student satisfaction.

2.5. The Influence of Student Satisfaction on Student Loyalty and Word-of-Mouth Communication

As noted by Usman and Mokhtar (2016), loyalty represents a complex factor of consumer behavior. This complexity has been acknowledged in earlier research—for example, Oliver (1999) conceptualizes loyalty as a “deeply held commitment” to a particular product or service, which comprises four elements: cognitive, affective, conative, and behavioral. When applied in the educational context, this form of commitment implies the development of a lasting relationship between the student and the institution. In this sense, achieving a certain level of loyalty toward the educational institution among learners has important practical implications (Nesset & Helgesen, 2009), as a loyal student will remain until completion of the study, encourage others, and spread positive word of mouth (Hassan et al., 2019b). This commitment may manifest both during the study and after its completion. Upon graduation, loyal individuals are likely to continue supporting their academic institution in a financial sense, through encouraging reports to current or former students, or even through some form of cooperation (Dang & Hathaway, 2015). Understanding the predictors of student loyalty in dVET may be crucial for enhancing institutional engagement and long-term student support. Consistent with prior findings in the HES, where researchers have established that student satisfaction is positively related to student loyalty (Helgesen & Nesset, 2007; Martin & Nasib, 2021), we assume that student satisfaction will likewise have a positive impact on student loyalty in the dVET context.

The basic idea behind word-of-mouth communication is the dissemination of information about products or services from one customer to another (Satria & Hidayat, 2018). Some authors define WOM as a powerful method in marketing (Andriana et al., 2023) because it is live, direct, experiential, and face-to-face (Chen, 2016). It is widely accepted that WOM can be categorized into positive and negative (Richins, 1983). Although both types of WOM are important, it is the positive WOM that should attract the interest of educational institution managers, as it stems from a sense of satisfaction (Yasa et al., 2021). Based on the literature (e.g., Elahinia & Karami, 2019), we define WOM from the learners’ perspective in dVET as the voluntary and informal interpersonal transmission of information by students, based on their impressions and experiences related to their vocational education and training. It is expressed through a sense of pride in belonging to the school, active discussions with other stakeholders about the institution, and a willingness to share positive experiences. Empirical research confirms the effect of student satisfaction on student WOM (Nguyen et al., 2021).

Therefore, the following hypothesis is developed:

Hypothesis 6 (H6).

Student satisfaction has a positive effect on (a) WOM, and (b) student loyalty.

2.6. Brand Image, dVET Image, and Student Satisfaction as Mediators

In addition to the proposed direct effects, this study further predicts that indirect effects (mediation effects) exist between these constructs. Prior research has indicated that brand image is a crucial mediator in consumer decision-making (Chang, 2025). Empirical evidence supporting this conclusion also exists in the field of education. For example, Jiewanto et al. (2012) found that university image mediates the relationship between students’ perceived service quality and their WOM intention. In line with this, Mesta (2019) confirmed that students’ satisfaction with service quality indirectly affects student loyalty through the university’s BI. Furthermore, Hassan et al. (2019a) demonstrated that corporate image indeed functions as a mediator, albeit a partial one, explaining 49.7% of the effect of service quality on students’ loyalty in TVET higher learning institutes (HLIs), as measured through the Variance Accounted For (VAF) analysis. Finally, Jamaludin and Sugiyanto (2024) reported that the influence of service quality on purchase decision is significantly greater when mediated by the brand image of vocational high schools.

Regarding the role of student satisfaction as a mediator, there is also empirical evidence supporting its indirect effect (Martin & Nasib, 2021). For instance, Sultan and Wong (2019) demonstrated that SS mediates the relationship between perceived service quality, brand performance, brand image and behavioral intention in the higher education sector (HES). Additionally, Hassan et al. (2019a) further supported this role—though partially—in the link between service quality and student loyalty in TVET HLIs, reporting a VAF value of 68.7%. Similarly, Moslehpour et al. (2020) found that SS acts as a bridge between the academic and non-academic aspects of service quality and institutional reputation. Lastly, Nursetiani and Wahyuni (2024) observed that service quality has a significant positive influence on WOM, and this relationship is mediated by SS in vocational high schools.

With regard to the indirect effect of dVET image, this study makes a novel contribution, as—to the best of our knowledge—it is the first to empirically examine this mediating role.

Based on the above, four mediating hypotheses are developed for testing:

Hypothesis 7 (H7).

There are significant and positive indirect effects of (a) brand associations and (b) brand relevance on the image of dVET, mediated by brand image.

Hypothesis 8 (H8).

There is a significant and positive indirect relationship between the brand image of secondary dVET schools and student satisfaction, mediated by the image of dVET.

Hypothesis 9 (H9).

There are significant and positive indirect effects of the image of dVET on (a) WOM and (b) student loyalty, mediated by student satisfaction.

Hypothesis 10 (H10).

There are significant and positive indirect effects of service quality on (a) WOM and (b) student loyalty, mediated by student satisfaction.

The proposed research model has been visualized in Figure 1.

3. Materials and Methods

3.1. Measures

After an in-depth review of the relevant literature, we generated an initial pool of potential items based on previously validated scales used in prior research. The resulting questionnaire was professionally translated into Bulgarian by a bilingual qualified expert. A standard back-translation procedure was then applied to ensure that the translated content accurately reflected the original English version. Subsequently, the questionnaire was reviewed by five experts working in the dVET system in Bulgaria. They were asked to evaluate the content of each construct for semantic redundancy and appropriateness for the specific characteristics of the target population. After the removal of several items due to content redundancy or incompatibility with the cultural context, as well as following a pilot study involving 30 students, during which some items were excluded due to low internal consistency values (Cronbach’s alpha < 0.7), the questionnaire was refined. As a result, it includes 56 statements designed to measure eight constructs: brand associations, brand relevance, brand image, image of dVET, service quality, student satisfaction, student loyalty, and word-of-mouth communication. Brand associations were measured with three items from the study of Girard and Pinar (2020). Brand relevance (five items) and brand image (three items) were adapted from Tran and Duc (2022). The second-order construct Image of dVET was adapted from the scale originally developed by Awang et al. (2011) and the works of Dang and Hathaway (2014). To measure the multidimensional construct of service quality, twenty-seven items were taken from the study by Elahinia and Karami (2019), as well as from the work of Hasan et al. (2008). Student satisfaction (five items) and word of mouth (three items) were also adapted from Elahinia and Karami (2019). The student loyalty scale, applied and validated by Nesset and Helgesen (2009), had two items. Appendix A (

Table A1. Final measurement items for the main study.

Variables	Items                           1 2 3 4 5	Sources
Service Quality		(Elahinia & Karami, 2019; Hasan et al., 2008)
Tangibility	The learning material we study is modern and up to date (TAN1) Number of vocational programs offered (TAN2) Adequacy of computers provided in the classrooms for students (TAN3) Appearance of the building and grounds (TAN4) Degree to which the classrooms and study rooms are comfortable (TAN5) Access to the Internet/e-mail (TAN6) Overall cleanliness (TAN7)
Assurance	Friendly and courteous teachers (ASSR1) Teachers are innovative (ASSR2) Communication skills: lessons are well taught by the teachers in my vocational secondary school (ASSR3)
Reliability	Teachers are punctual and rarely cancel classes (REL1) Teaching capability of mentors (REL2) Teachers’ sincere interest in solving the student’s problem (REL3)
Responsiveness	Teachers’ capacity to solve problems when they arise (RES1) Staff’s capacity to solve problems when they arise (RES2) I seldom get the “run-around” when seeking information from this school (RES3) Channels for expressing student complaints are readily available (RES4)
Empathy	The administration has students’ best interest at heart (EMP1) Access to computer facilities is suited to students’ convenience (EMP2) Access to study rooms is suited to students’ convenience (EMP3) Staff are willing to give students individual attention (EMP4)
Brand Relevance	My school looks modern and up-to-date (BR1) My school provides learning value tailored to my needs (BR2)	(Tran & Duc, 2022)
Student Satisfaction	I am satisfied with my decision to attend this school (SS1) I am happy with my decision to enroll in this school (SS2)	(Elahinia & Karami, 2019)
Image of dVET		(Awang et al., 2011; Dang & Hathaway, 2014)
Low Entry Qualifications	Vocational students have low learning abilities (LEQ1) My school gives access to all secondary school students (LEQ2) My school has low and flexible entry requirements (LEQ3)
Training Facilities and Equipment	The company where I do my vocational training has proper equipment and the latest technology (TFE1) The company where I train has enough space for good training (TFE2)
Mentors’ Credibility	Helpful mentors (MC1) Experienced mentors (MC2) Qualified mentors (MC3)
Recognition of Vocational Qualification	The dVET qualification is recognized by many private companies in Bulgaria (RVQ1) The dVET qualification is recognized by many foreign companies (RVQ2) The salary is on par with the academic qualifications (RVQ3)
Career and Job Potential	The dVET provides students with strong practical skills (CJP1) The dVET enables the integration of academic knowledge and technical skills (CJP2) The dVET offers a wide range of career selections for graduates (CJP3) The dVET system prepares students to meet the needs of the nation (CJP4)
Curriculum	The theory we learn at school is closely connected to our practical training (CU1) My school maintains strong connections with various industries (CU2) My school offers many interesting opportunities for dual education (CU3) My school is well connected with the local community (CU4)
Soft Skills	DVET graduates have communication skills (SOF6) DVET graduates have leadership skills (SOF7)
Student Loyalty	There is a high probability that I would recommend this school to my friends or acquaintances (SL1) If I had to start over, there is a high chance I would choose the same school again (SL2)	(Nesset & Helgesen, 2009)
Brand Image	The name of the school is well known in the Plovdiv region (BI1) The name of the school is associated with high-quality teaching (BI2) My school has a strong brand image (BI3)	(Tran & Duc, 2022)
Word of Mouth	I always speak well of this school to people (WOM1) I usually talk about this school with my friends (WOM2) I am honored to tell people that I am studying in this school (WOM3)	(Elahinia & Karami, 2019)
Brand Associations	My school offers many opportunities for vocational training (BA1) My school organizes events related to the presentation of various professions and employers (BA3)	(Girard & Pinar, 2020)

Note: Following reliability and validity analyses, some items were excluded from the final model (see Section 4.1). Please indicate your level of satisfaction or agreement with each of the following statements about your school or the dVET system on a scale from 1 to 5, where 1 = strongly disagree and 5 = strongly agree.

) provides the final items for each construct used in the study, measured on a 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree.

3.2. Participants and Data Collection

This study focused on vocational secondary schools in the Plovdiv region—the second most populous area in Bulgaria after the capital, Sofia, and a key manufacturing hub. Institutions that offer dual education but are not officially classified as vocational secondary schools were excluded from the sample. Likewise, schools with students enrolled in the dVET program who had not yet reached grades 11th and 12th—the final two years of secondary education in the country—and therefore lacked meaningful experience in work-based learning, were not included in the target population.

Data requested from the Education Information Assurance Center (CIAC) of the Ministry of Education and Science show that, during the 2023/2024 academic year, nine vocational secondary schools in the Plovdiv district met the study’s criteria. Together, these schools enrolled 608 students, forming the total target population for this study.

Before conducting the survey, meetings were held with the principal of each school explaining the scope, purpose, and confidentiality conditions of the research. The administrations consented to the distribution of the surveys and notified the target classes before the visitation date. During the visit, the authors informed students about the nature of the questions and their intended use. All participants were over the age of 16 and took part voluntarily. Informed consent was obtained directly from the students, under national educational regulations for low-risk academic surveys. Additionally, the school principals provided official permission for the study, and a school representative was present during data collection.

The data collection was conducted on-site at nine schools between 21 April and 30 April 2024. A total of 559 questionnaires were obtained, of which 52 were excluded due to incomplete or incorrectly completed responses. Thus, 507 valid questionnaires were analyzed. The study used a convenience sampling approach, based on the accessibility of the participating schools and the physical presence of students in the relevant classes. All students who were present on the day of the visit voluntarily participated in the study, resulting in an exceptionally high response rate of 92% of the target group. This strong level of engagement increases confidence in the reliability and representativeness of the results. However, as the sample was based on convenience (i.e., students who happened to be available during data collection), some important limitations should be noted. For example, absent students may differ in certain ways. Furthermore, since the study is cross-sectional and captures data from a single point in time, any causal interpretations should be made cautiously, as they rely on the proposed model rather than observed changes over time.

4. Results

In the present study, partial least squares structural equation modeling (PLS-SEM) was employed using SmartPLS 4. The choice of PLS-SEM was based on several considerations. This analytical approach is well known for its ability to estimate and evaluate complex models with multiple interrelated constructs. For this reason, it is considered to be highly suitable for the present research. First, the theoretical framework is in an early stage of development, and the model is partly exploratory. It is applied in a novel empirical context—the Bulgarian dual vocational education and training (dVET) system—where no stable conceptual foundation currently exists. In such circumstances, PLS-SEM offers greater flexibility for theory development and exploration compared to CB-SEM (Hair et al., 2022). Secondly, the proposed model is characterized by its substantial complexity, encompassing multiple constructs, paths, mediators, and two higher-order reflective–reflective constructs. PLS-SEM has been demonstrated to be a particularly suitable approach for the estimation of such models. This is due to its ability to manage higher-order constructs with greater efficiency and fewer restrictive assumptions compared to CB-SEM (Sarstedt et al., 2019; Hair et al., 2022).

In the context of structural equation modeling analysis, two primary components are responsible for different aspects of the model: the measurement model and the structural model. The measurement model assesses the reliability and validity of the constructs through observed indicators, checking whether the data adequately reflect the theoretical concepts. Conversely, the structural model is responsible for the modeling of causal relationships between constructs and the examination of the strength and significance of these relationships. This duality enables comprehensive testing of theoretical hypotheses and provides a foundation for elucidating complex models in both practical and academic contexts.

The proposed conceptual model included two second-order reflective-reflective constructs and six first-order reflective constructs. In the scientific literature, several well-established approaches have been proposed for evaluating models in PLS-SEM, especially when higher-order constructs are included in the model. There is a consensus that these approaches produce relatively similar results in terms of robustness and precision of estimates, given sufficiently large sample sizes (Sarstedt et al., 2019). In the context of the present study, given the nature of the reflective-reflective construct design and the number of observations, a disjoint two-stage approach was chosen. This approach not only allows for a more precise operationalization of higher-order constructs but also minimizes the risk of multicollinearity among indicators, thereby providing a reliable and conceptually clear basis for interpreting the structural relationships (Sarstedt et al., 2019; Hair et al., 2022).

In the first phase of the disjoint two-stage approach, the model was assessed for reliability and validity using only first-order constructs. Once the model was estimated, construct scores for the first-order constructs of the two second-order constructs, IDVET (Image of dVET) and SQ (Service Quality), were also calculated. These include the following subconstructs: LEQ_LV, TFE_LV, MC_LV, SCJP_LV, QC_LV, SOC_LV and SOF_LV for the IDVET construct, and TAN_LV, ASSR_LV, REL_LV and RES_LV for the SQ construct. The suffix “_LV” was used to denote the corresponding latent variables, which were subsequently used as indicators in the second-order constructs to estimate the structural model in the second phase of the disjoined two-stage approach.

4.1. Measurement Model

The measurement model was evaluated per the procedures recommended by Hair et al. (2022). The reliability of the indicators was assessed via outer loadings, the internal consistency was verified using Cronbach’s Alpha and Composite Reliability, and convergent validity was established through the Average Variance Extracted (AVE). Discriminant validity was examined using the Fornell-Larcker criterion and the HTMT ratio. Furthermore, the Variance Inflation Factor (VIF) values were examined to ascertain the absence of multicollinearity among the indicators.

In the assessment of the measurement model, it was necessary to exclude certain indicators due to their outer loadings falling below the recommended threshold of 0.5. Their inclusion in the analysis would have adversely affected the reliability and validity of the measurement model. The following indicators were removed from the model: TAN4, REL2, RES4, RVQ3, and CU3. Moreover, the entire subconstruct Empathy was excluded due to similarly low loadings and unsatisfactory psychometric properties.

Following the elimination of these indicators, the evaluation of the measurement model (

Table 1. Results of measurement model.

Constructs and Measurement Items	Loadings
(LEQ) Low Entry Qualification (Cronbach’s alpha = 0.715, CR = 0.837, AVE = 0.633)
LEQ1. Vocational students have low learning abilities.	0.692
LEQ2. My school gives access to all secondary school students.	0.836
LEQ3. My school has low and flexible entry requirements.	0.849
(TFE) Training Facilities and Equipment (Cronbach’s alpha = 0.880, CR = 0.944, AVE = 0.893)
TFE1. The company where I do my vocational training has proper equipment and the latest technology.	0.945
TFE2. The company where I train has enough space for good training.	0.946
(MC) Mentors Credibility (Cronbach’s alpha = 0.905, CR = 0.941, AVE = 0.841)
MC1. Helpful mentors.	0.910
MC2. Experienced mentors.	0.926
MC3. Qualified mentors.	0.915
(CJP) Career and Job Potential (Cronbach’s alpha = 0.870, CR = 0.911, AVE = 0.719)
CJP1. The dVET provides students with strong practical skills.	0.818
CJP2. The dVET enables the integration of academic knowledge and technical skills.	0.855
CJP3. The dVET offers a wide range of career selections for graduates.	0.861
CJP4. The dVET system prepares students to meet the needs of the nation.	0.856
(CU) Curriculum (Cronbach’s alpha = 0.725, CR = 0.879, AVE = 0.784)
CU1. The theory we learn at school is closely connected to our practical training.	0.875
CU2. My school maintains strong connections with various industries.	0.896
(RVQ) Recognition of Vocational Qualification (Cronbach’s alpha = 0.830, CR = 0.921, AVE = 0.854)
RVQ1. The dVET qualification is recognized by many private companies in Bulgaria.	0.934
RVQ2. The dVET qualification is recognized by many foreign companies.	0.915
(SOF) Soft skills (Cronbach’s alpha = 0.784, CR = 0.902, AVE = 0.822)
SOF6. DVET graduates have communication skills.	0.918
SOF7. DVET graduates have leadership skills.	0.895
(TAN) Tangibility (Cronbach’s alpha = 0.841, CR = 0.883, AVE = 0.558)
TAN1. The learning material we study is modern and up to date.	0.758
TAN2. Number of vocational programs offered.	0.721
TAN3. Adequacy of computers provided in the classrooms for students.	0.741
TAN5. Degree to which classrooms and study rooms are comfortable.	0.796
TAN6. Access to the Internet/e-mail.	0.690
TAN7. Overall cleanliness.	0.771
(ASSR) Assurance (Cronbach’s alpha = 0.889, CR = 0.931, AVE = 0.818)
ASSR1. Friendly and courteous teachers.	0.894
ASSR2. The teachers are innovative.	0.910
ASSR3. Communication skills: lessons are well taught by the teachers in my vocational secondary school.	0.910
(REL) Reliability (Cronbach’s alpha = 0.736, CR = 0.883, AVE = 0.791)
REL1. Teachers are punctual and rarely cancel classes.	0.893
REL3. Teachers’ sincere interest in solving the student’s problem	0.886
(RES) Responsiveness (Cronbach’s alpha = 0.823, CR = 0.894, AVE = 0.738)
RES1. Teachers’ capacity to solve problems when they arise.	0.878
RES2. Staff’s capacity to solve problems when they arise.	0.859
RES3. I seldom get the “run-around” when seeking information from this school.	0.840
(BA) Brand Association (Cronbach’s alpha = 0.817, CR = 0.916, AVE = 0.845)
BA1. My school offers many opportunities for vocational training.	0.925
BA3. My school organizes events related to the presentation of various professions and employers.	0.914
(SL) Student Loyalty (Cronbach’s alpha = 0.859, CR = 0.934, AVE = 0.876)
SL1. There is a high probability that I would recommend this school to my friends or acquaintances.	0.938
SL2. If I had to start over, there is a high chance I would choose the same school again.	0.934
(BR) Brand Relevance (Cronbach’s alpha = 0.722, CR = 0.878, AVE = 0.782)
BR1. My school looks modern and up to date.	0.874
BR2. My school provides learning value tailored to my needs.	0.894
(SS) Student Satisfaction (Cronbach’s alpha = 0.939, CR = 0.970, AVE = 0.942)
SS1. I am satisfied with my decision to attend this school.	0.973
SS2. I am happy with my decision to enroll in this school.	0.968
(BI) Brand Image (Cronbach’s alpha = 0.757, CR = 0.858, AVE = 0.669)
BI1. The name of the school is well known in the Plovdiv region.	0.796
BI2. The name of the school is associated with high-quality teaching.	0.789
BI3. My school has a strong brand image.	0.868
(WOM) Word of mouth (Cronbach’s alpha = 0.831, CR = 0.898, AVE = 0.747)
WOM1. I always speak well of this school to people.	0.894
WOM2. I usually talk about this school with my friends.	0.787
WOM3. I am honored to tell people that I am studying in this school.	0.906

) indicated that all remaining reflective constructs satisfied the established threshold values. The majority of outer loadings exceeded 0.708, indicating satisfactory indicator reliability. The composite reliability and Cronbach’s α values for all constructs exceeded 0.70, thereby confirming the internal consistency of the data. Furthermore, the AVE values for all constructs exceed the threshold of 0.50, thereby confirming convergent validity.

Discriminant validity was confirmed through the Fornell-Larcker criterion, where the square root of each construct’s AVE was greater than its correlations with other constructs. The Heterotrait-Monotrait (HTMT) ratios were all below the threshold of 0.90 (Henseler et al., 2015), indicating no issues with discriminant validity (

Table 2. Discriminant validity (Fornell–Larcker criterion and HTMT).

	BA	LEQ	TFE	MC	CJP	CU	RVQ	SOF	BI	SL	TAN	ASSR	REL	RES	BR	SS	WOM
BA	0.919	0.203	0.560	0.528	0.704	0.798	0.562	0.594	0.658	0.656	0.616	0.566	0.600	0.586	0.678	0.551	0.812
LEQ	0.165	0.796	0.290	0.206	0.203	0.251	0.150	0.176	0.357	0.197	0.385	0.150	0.170	0.233	0.244	0.203	0.256
TFE	0.475	0.253	0.945	0.841	0.720	0.634	0.528	0.516	0.517	0.453	0.570	0.457	0.503	0.449	0.477	0.484	0.574
MC	0.454	0.184	0.752	0.917	0.693	0.583	0.474	0.476	0.498	0.423	0.534	0.445	0.535	0.420	0.466	0.426	0.526
CJP	0.596	0.178	0.632	0.615	0.848	0.837	0.543	0.549	0.583	0.557	0.657	0.544	0.607	0.580	0.589	0.598	0.720
CU	0.616	0.193	0.508	0.474	0.668	0.885	0.635	0.707	0.719	0.633	0.734	0.642	0.663	0.684	0.757	0.629	0.797
RVQ	0.463	0.117	0.451	0.411	0.463	0.493	0.924	0.838	0.584	0.493	0.540	0.462	0.493	0.475	0.523	0.410	0.639
SOF	0.475	0.146	0.430	0.404	0.456	0.533	0.678	0.907	0.636	0.530	0.620	0.511	0.582	0.520	0.579	0.443	0.657
BI	0.534	0.263	0.429	0.420	0.494	0.552	0.482	0.506	0.818	0.694	0.711	0.563	0.723	0.610	0.740	0.561	0.770
SL	0.549	0.169	0.394	0.374	0.482	0.499	0.415	0.436	0.578	0.936	0.578	0.545	0.629	0.603	0.649	0.657	0.801
TAN	0.513	0.307	0.494	0.470	0.566	0.575	0.453	0.507	0.581	0.495	0.747	0.730	0.784	0.750	0.867	0.589	0.715
ASSR	0.483	0.131	0.404	0.399	0.479	0.517	0.396	0.429	0.484	0.477	0.632	0.905	0.833	0.754	0.735	0.642	0.684
REL	0.465	0.136	0.405	0.437	0.487	0.485	0.388	0.444	0.554	0.500	0.617	0.674	0.890	0.791	0.807	0.635	0.741
RES	0.481	0.188	0.381	0.363	0.491	0.529	0.391	0.417	0.497	0.509	0.624	0.648	0.619	0.859	0.789	0.593	0.687
BR	0.522	0.184	0.380	0.377	0.468	0.549	0.406	0.437	0.563	0.511	0.678	0.591	0.591	0.613	0.884	0.652	0.726
SS	0.483	0.172	0.441	0.394	0.542	0.520	0.364	0.381	0.493	0.592	0.529	0.588	0.529	0.525	0.538	0.971	0.683
WOM	0.674	0.208	0.495	0.460	0.619	0.626	0.527	0.531	0.637	0.688	0.608	0.598	0.587	0.573	0.568	0.615	0.864

The square roots of the AVE values have been marked in bold and positioned diagonally. The HTMT values have been presented above the diagonal elements.

).

To ensure that multicollinearity did not bias the estimation of the path coefficient in the model, Variance Inflation Factor (VIF) values were calculated for all indicators in the measurement model. According to Hair et al. (2022), the presence of multicollinearity can be assessed by examining the VIF values. Values below 5 indicate that multicollinearity is not a problem; values between 3 and 5 are regarded as moderate but still acceptable, while values above 5 suggest potential multicollinearity issues that may compromise the reliability of the results. As demonstrated in

Table 3. Multicollinearity diagnostics using VIF values.

Indicators	VIF Values	Indicators	VIF Values	Indicators	VIF Values
BA1	1.911	CJP2	2.280	ASSR2	2.720
BA2	1.911	CJP3	2.243	ASSR3	2.638
BR1	1.468	CJP4	2.085	REL1	1.514
BR2	1.468	CU1	1.477	REL2	1.514
BI1	1.470	CU2	1.477	RES1	1.926
BI2	1.562	RVQ1	2.010	RES2	1.986
BI3	1.554	RVQ2	2.010	RES3	1.707
LEQ1	1.309	SOF1	1.712	SL1	2.309
LEQ2	1.553	SOF2	1.712	SL2	2.309
LEQ3	1.433	TAN1	1.848	SS1	4.587
TFE1	2.620	TAN2	1.890	SS2	4.587
TFE2	2.620	TAN3	1.681	WOM1	2.201
MC1	2.693	TAN4	2.176	WOM2	1.627
MC2	3.340	TAN5	1.673	WOM3	2.252
MC3	2.937	TAN6	1.744
CJP1	1.901	ASSR1	2.436

, the VIF values for all indicators range from 1.309 to 4.587, indicating that they fall within an acceptable range.

To assess the potential presence of common method bias (CMB), Harman’s single-factor test was conducted on the final set of indicators retained in the measurement model. The analysis was performed using unrotated principal axis factoring. The results showed that the first factor accounted for 37.7% of the total variance, which is below the critical threshold of 50% (Fuller et al., 2016). These findings suggest that common method bias is not a substantial concern in the present study.

Thus, the results of the measurement model evaluation confirmed its reliability and validity, providing a sound basis for subsequent analysis of the structural relationships in the model.

4.2. Structural Model

Following the confirmation of the reliability and validity of the measurement model, structural model estimation was performed to test the empirical support for the postulated hypotheses. In addition to hypothesis testing, the model’s explanatory power (R²), predictive relevance (Q²), and collinearity diagnostics via inner VIF values were assessed. A bootstrapping method with 5000 resamples was employed to assess the significance of the path coefficients.

4.2.1. Assessment of Direct Relationships

Table 4. Direct effects between latent constructs.

Hypotheses	β-Value	2.5% CI	97.5% CI	Decision
H1: Brand Associations → Brand Image	0.332	0.247	0.418	Supported
H2: Brand Relevance → Brand Image	0.391	0.310	0.471	Supported
H3: Brand Image → Image of dVET	0.627	0.570	0.682	Supported
H4: Image of dVET → Student Satisfaction	0.253	0.144	0.366	Supported
H5: Service Quality → Student Satisfaction	0.463	0.356	0.559	Supported
H6a: Student Satisfaction → WOM	0.616	0.549	0.679	Supported
H6b: Student Satisfaction → Student Loyalty	0.592	0.525	0.656	Supported

All reported path coefficients β are significant at p < 0.01.

presents the results of the analysis of the direct effects between the latent constructs, including path coefficients, 95% confidence intervals, and hypothesis testing outcomes. Following established benchmarks (Hair et al., 2022), path coefficients above 0.50 are considered large, those around 0.30 are medium, and those near 0.10 are small. The results showed that the strongest effect in the model was the effect of brand image on image of dVET (β = 0.627, large effect), followed by the effect of student satisfaction on word of mouth (β = 0.616, large effect) and on student loyalty (β = 0.592, large effect). Service quality has a medium-to-large effect on student satisfaction (β = 0.463), while brand relevance (β = 0.391) and brand associations (β = 0.332) exert medium effects on brand image. The weakest direct effect is that of image of dVET on student satisfaction (β = 0.253), which is small-to-medium. All path coefficients were statistically significant at p < 0.01, and the 95% confidence intervals reported in Table 4 exclude zero, further confirming the significance of these relationships. Therefore, Hypotheses 1 through 6 were supported.

4.2.2. Assessment of Indirect Relationships

Table 5. Indirect effects between latent constructs.

Hypotheses	β-Value	2.5% CI	97.5% CI	Decision
H7a: Brand Associations → Brand Image → Image of dVET	0.208	0.149	0.273	Supported
H7b: Brand Relevance → Brand Image → Image of dVET	0.245	0.191	0.304	Supported
H8: Brand Image → Image of dVET → Student Satisfaction	0.159	0.089	0.236	Supported
H9a: Image of dVET → Student Satisfaction → WOM Communication	0.156	0.084	0.235	Supported
H9b: Image of dVET → Student Satisfaction → Student Loyalty	0.150	0.082	0.225	Supported
H10a: Service Quality → Student Satisfaction → WOM Communication	0.285	0.214	0.352	Supported
H10b: Service Quality → Student Satisfaction → Student Loyalty	0.274	0.207	0.336	Supported

All reported path coefficients β are significant at p < 0.01.

presents the indirect effects between the latent constructs. In PLS-SEM, path coefficients for indirect effects indicate the strength of the mediated relationship between constructs, with higher absolute values denoting stronger mediation effects (Hair et al., 2022). The strongest indirect effect was observed for service quality on word of mouth mediated by student satisfaction (β = 0.285), followed closely by its mediated effect on student loyalty (β = 0.274). The indirect impact of brand relevance (β = 0.245) and brand associations (β = 0.208) on the image of dVET through brand image were also statistically significant. There were weaker, yet still statistically significant, mediation effects of brand image on student satisfaction via image of dVET (β = 0.159), and of image of dVET on word of mouth (β = 0.156) and student loyalty (β = 0.150) via student satisfaction, all of which can be classified as small-to-medium effects. All indirect relationships were statistically significant at p < 0.01, and the 95% confidence intervals reported in Table 5 excluded zero, confirming the robustness of these mediation effects.

These results confirm the mediating roles of brand image, image of dVET, and student satisfaction within the model structure. There is partial mediation as the direct effects between the respective constructs are also statistically significant (as shown in Table 4), which means that the mediators do not fully but partially explain the relationships between the independent and dependent variables. These findings suggest the presence of both direct and indirect pathways of influence, thereby strengthening the validity of the theoretical model. Hence, Hypotheses 7 through 10 were supported.

4.2.3. Model Quality and Collinearity Assessment

The following subsection evaluates the structural model’s in-sample explanatory power, out-of-sample predictive performance, and potential multicollinearity. In-sample explanatory power is assessed using R², out-of-sample predictive performance is evaluated with Q²_predict from PLSpredict, and potential multicollinearity is examined through inner VIF diagnostics.

The explanatory power of the model was assessed using the coefficient of determination (R²) for the endogenous constructs. The findings indicated that the model accounts for 39.8% of the variance in BI, 39.4% in IDVET, 35.0% in SL, 44.0% in SS, and 37.9% in WOM. In accordance with the guidelines established by Hair et al. (2022), these values indicate a low-to-moderate level of explanatory power across the endogenous constructs.

Next, predictive relevance was evaluated using the Q²_predict statistic from PLSpredict (Shmueli et al., 2019). All Q²_predict values were positive and substantially above zero (BI = 0.391, IDVET = 0.394, SL = 0.285, SS = 0.410, WOM = 0.369), providing evidence of out-of-sample predictive relevance for all endogenous constructs.

Finally, collinearity among predictor constructs was examined by inspecting inner VIF values, which reflect the degree of multicollinearity in the structural model and help assess the stability of path coefficients (Hair et al., 2022). As shown in

Table 6. Collinearity assessment of the structural model via inner VIF values.

Structural Path	VIF Values	Structural Path	VIF Values
BA → BI	1.374	SQ → SS	1.911
BI → IDVET	1.000	SS → SL	1.000
BR → BI	1.374	SS → WOM	1.000
IDVET → SS	1.911

, the range of inner VIF values was from 1.000 to 1.911, all well below the recommended threshold of 3.3 (Kock, 2015), indicating the absence of collinearity issues. As the VIF values for the indicators (outer VIF) were previously reported in the measurement model assessment, the inner VIF values presented here complement that analysis by focusing on construct-level relationships within the structural model.

5. Discussion

5.1. Theoretical and Managerial Implications

Our results confirm that brand associations, brand relevance, brand image, image of dVET, service quality, and student satisfaction are significant predictors of student loyalty and word of mouth. The effect of BA on BI is well established in commercial contexts (Dey & Rahman, 2023; Khan & Jalees, 2016), but to our knowledge, this is the first statistical confirmation within dual VET, which underscores the strategic role of associations in building a school’s brand image. Brand relevance also emerges as a strong determinant of BI, in line with the findings from higher education (Le, 2021; Tran & Duc, 2022), suggesting that a closer alignment between program offerings and students’ expectations improves the perceived image. Additionally, brand image mediates the effects of both brand associations and brand relevance on the image of dVET, extending earlier demonstrations of the mediating role of brand image in educational contexts (Jiewanto et al., 2012; Mesta, 2019; Jamaludin & Sugiyanto, 2024).

The image of dVET shows both direct and indirect effects on SS: it raises satisfaction in its own right and mediates the relationship between brand image and student satisfaction. This builds on the conclusions of Awang et al. (2011) and, in the Bulgarian setting, suggests that all dimensions of the system image can contribute positively to satisfaction; therefore, maintaining a strong, positive image of dVET is a precursor to loyalty.

Theoretically, the results are consistent with TVET evidence on the central role of student satisfaction as the mechanism that transmits the effects of perceived service quality to subsequent outcomes: in our model, SS has strong direct effects on WOM (β = 0.616) and SL (β = 0.592), while SQ substantially influences SS (β = 0.463); the strongest indirect route to recommendations runs through satisfaction (β = 0.285), in line with Nursetiani & Wahyuni (2024). At the same time, the presence of significant direct and indirect paths adds nuance to studies in which direct effects dominate (Rusdi & Ali, 2020). Finally, brand image emerges as the strongest predictor of the image of dVET (β = 0.627), which complements the conclusions of Andriana et al. (2023) regarding the weight of institutional image and, in parallel, supports the idea that in TVET settings the key transmission mechanism to word of mouth is satisfaction rather than loyalty (e.g., Nursetiani & Wahyuni, 2024).

Practically, these findings point to three groups of actions that can be implemented as routine management practices. First, to raise brand relevance and improve the alignment between programs and expectations, it is appropriate to hold annual joint meetings between dVET school administrators, local government representatives, and employers. To support teaching, schools can introduce teacher externships in firms and provide systematic training for mentors in host organizations so that firsthand knowledge continues to flow into classrooms. Second, to strengthen brand image and the image of dVET, schools can use jointly branded communication with employers and industry bodies, including showcasing student projects, report placement rates and starting wages, keep visual identity and message pillars consistent across channels, host open lab days and evening parent events with live demonstrations of dual training, and publish public dashboards with key outcomes to increase transparency and engagement. Third, to enhance service quality—and, through it, satisfaction—schools can define standards for reliability, responsiveness, assurance, and the physical environment; conduct periodic surveys of different stakeholders and publicly report specific results for greater transparency; and monitor identified bottlenecks in the learning process (scheduling, equipment availability, assessment clarity), tracking changes in satisfaction and recommendations. Together, these interventions can improve the interconnection between school programs and students’ aspirations (relevance), create an authentic and consistent image of the school and the dual system (brand image and the image of dVET), and institutionalize quality routines that reliably raise satisfaction, which is the most effective lever for loyalty and word of mouth in dual VET.

5.2. Limitations and Directions for Future Research

Like any research, this study has its limitations, which also suggest meaningful directions for future inquiry. First, it adopted a quantitative approach to test the developed hypotheses. Future investigations might combine this survey-based method with qualitative techniques, which could offer deeper insights into how students perceive their dVET experience. In addition, tracking students over time through longitudinal studies could offer valuable insights into how their loyalty develops as they progress through the various stages of education and training (from enrollment to graduation). Second, this study used a cross-sectional design. As such, it limits the ability to draw strong conclusions about cause-and-effect relationships between the examined variables. The results reveal associations and dependencies based on data gathered at a single point in time and within a specific context, and unmeasured or external factors may have influenced the outcomes. Moreover, since the study was conducted in a single geographic area and within a specific category of educational institutions, these contextual factors should be considered when assessing the broader applicability of the findings. Third, we focused on the opinions of 11th- and 12th-grade students in dVET programs, i.e., those who have already passed through the earlier, higher-risk stages of vocational education where dropout is more likely. Therefore, further research in Bulgaria is needed to include students in the early stages of dVET to better understand the drivers of dropout. Fourth, the current analysis was limited to the students’ standpoint. The viewpoints of additional stakeholders, such as parents, teachers, school principals, mentors, employers, and educational policymakers, are equally important. Observations of these groups could provide a more holistic understanding of the vocational education and learning system. Fifth, it should be noted that the present study was focused exclusively on the perceptions of students enrolled in the dVET. Future research could conduct a comparative analysis of the influential factors of SL and WOM among students in both dual and traditional forms. The lack of such comparison here restricts our ability to fully assess the effects of the dual model relative to the traditional one. A future study in this direction could offer a more grounded understanding of the benefits, challenges, and actual effects of both forms, supporting stronger recommendations for future policy and practice in the field of VET in Bulgaria. Finally, this research framework focused on selected variables that act as predictors of student loyalty and word-of-mouth communication. Future studies may consider exploring other under-researched factors in the context of VET, such as brand love, brand trust, brand resonance, brand engagement, etc.