Service quality experience in the context of informal paratransit, shared taxi and related ride-sharing services is formed on the foundation of synthesising many classical theories on service quality, service experience and perceived value. Foundational models such as SERVQUAL (Parasuraman, Zeithaml & Berry 1988), the servicescape of Bitner (1992) and the perceived value scale (PERVAL) (Sweeney & Soutar 2001), along with approaches on trust in the service relationships (Mayer, Davis & Schoorman 1995), provide a core theoretical foundation to form the lower-order constructs (LOCs) of SQE. At the same time, lots of recent studies on service quality in public transport and informal paratransit have also shown the demand of conceptualising service quality at a proper level rather than merely considering each isolated dimension (Joewono & Kubota 2007; Peters & Bhusal 2020; Rahman et al. 2023; Tiglao et al. 2020). First and foremost, the SERVQUAL model was used widely with five classical dimensions including tangibles, reliability, responsiveness, assurance and empathy (Parasuraman et al. 1988). A considerable number of studies in the field like public transport, informal paratransit and ride-sharing services were based on SERVQUAL to build and inspect service quality models. Consequently, they showed the remarkable role of dimensions related to the serving behaviour of front-line employees (Chan et al. 2021; Joewono & Kubota 2007; Luke & Heyns 2020). Accordingly, this research inherits and specifies two dimensions, assurance and empathy, into an LOC called staff assurance and empathy (SAE). This structure reflects behaviours such as safe driving, polite attitude, willingness to provide support and ability to understand a passenger’s demand during the entire journey (Luke & Heyns 2020; Ojekunle et al. 2021). Within the informal paratransit context, drivers and people directly providing service are regularly the main agents contacting customers. In addition, SAE serves an especially vital role because it can dominate a customer’s feeling about safety, trust and decision to return to using the service. Next, the concept servicescape of Bitner (1992) highlighted the role of the physical environment in shaping general service experience. In public transport and informal paratransit, physical servicescape not only consists of vehicle interior but also includes pickup and drop-off points, waiting areas, signage, lighting and sanitation conditions. All of these things contribute to establish initially a passenger’s feeling about service quality and their accessibility (Tiglao et al. 2020). Besides, the PERVAL of Sweeney and Soutar (2001) was quickly developed with the aim of emphasising that customers often experienced numerous kinds of different values in a service, including functional value, social value, emotional value and economic value. With the setting of informal public transport services, the reality indicates that passengers often evaluate a service not only based on absolute cost but also on the level of fairness between price and received value and by comparing those things with alternative modes of transport such as buses, traditional taxis and technology-based motorbikes (Bhuyan & Roy 2024; Jen & Hu 2003; Vu et al. 2024). Therefore, the research adjusts the centre of PERVAL into a price-value (PV) fairness structure that reflects level of fairness regarding ticket price and reasonability when comparing with other choices. It also presents the overall assessment of the correlation between price and received benefit. The trust aspect is also highlighted in several studies on shared service and public transport, especially when passengers must depend on the booking system, schedule information and mechanisms of guaranteeing safety to make decisions (Hartl, Penz & Schuessler 2025; Mas-Machuca, Marimon & Jaca 2021; Nguyen-Phuoc et al. 2021). Service system trust (SST) within the informal paratransit context comprises believing in transparency regarding information, personal data security if using digital platforms, the reliability of service providers and drivers, as well as the level of compliance with basic commitments like journey, pickup, drop-off time and price. When the passenger has to share their moving space with strangers, SST is able to reduce perceived risks, hence enabling them to accept and maintain using service (Nguyen-Phuoc et al. 2021; Tiglao et al. 2020). In addition to the above classical theories, a considerable amount of recent research on transport service has suggested supplementing another crucial dimension, that is, in-vehicle amenities quality. This structure also reflects the level of adequacy and quality of in-vehicle amenities such as air conditioner, seat system, Wi-Fi connection, power socket, beverage or other simple amenities that are able to make the journey more comfortable (Ittamalla & Srinivas Kumar 2021; Oubahman & Duleba 2024). Even in the context of informal paratransit, the reality shows that service providers tend to increase some basic amenities in an effort to attract customers, especially on interprovincial travel or journeys with long travel times. These amenities not only improve the instant experience but also contribute to increase perceived value and maintain positive feedback about the general service quality (Rahman et al. 2023). Despite the fact that each LOC was tested and confirmed in many independent studies, the integration of those into an HOC, which specifically is SQE, has especially important meaning in terms of both theoretical and methodological aspect. In view of theory, the approach of HOC allows conceptualising service experience in a more general direction. Within the setting of informal public transport, service quality that passengers feel is rarely decided by a single dimension such as price, safety or amenities, but it is the integrative result of numerous factors such as physical servicescape, in-vehicle amenities, serving behaviour of drivers and staff, perceived value and SST (Firzan et al. 2025; Peters & Bhusal 2020). With respect to methodology, conceptualising SQE as an HOC in the framework of PLS-SEM brings lots of benefits. This approach enables to decrease the number of observed variables that are directly brought into the structural model. It, moreover, limits problems regarding multicollinearity among service quality dimensions that may overlap and concurrently allows testing the efficiency of the HOC for the purpose of explaining the change of outcome variables (Hair & Alamer 2022). More importantly, the usage of HOC reflects more practically because the passengers often form the overall evaluation about service quality, and then discuss satisfaction, intention to return and referral tendency, instead of assessing separately every single aspect. Based on those arguments, this study conceptualises SQE as a reflective–reflective HOC whose lower-order dimensions capture both tangible and intangible aspects of informal paratransit SQE. Such an approach both ensures the theoretical stability when based on classical models like SERVQUAL, PERVAL, trust and servicescape and increases the implications when analysing customer behaviour using informal paratransit in the context of developing economies.

Within the context of informal public transport and related ride-sharing services, satisfaction is usually viewed as a core mediating variable, explained based on the ECT. According to ECT, the level of satisfaction of passengers is established when they compare an actual service experience with their initial expectation. If the actual service experience meets or goes beyond the expectation, confirmation will happen, thereby leading to satisfaction. In reverse, if the actual service experience is lower than the expectation, passengers tend to show dissatisfaction (Fu, Zhang & Chan 2018; Wang, Lin & Liu 2021). Multiple studies within the context of shared service also present that service quality, perceived value and trust in the booking system are vital factors to confirmation. So, they influence on satisfaction and continuance intention (Arteaga-Sánchez et al. 2020; Lim et al. 2022). Furthermore, satisfaction is not only a final result but also acts as an intermediate mechanism connecting the SQE with loyalty intention and referral behaviour, advocacy or WOM. There are many studies on public transport and shared service indicating that when passengers highly evaluate service quality, they tend to be more satisfied, and thus, easily forming continuance intention, as well as willingness to recommend the service for other passengers (Quy Nguyen-Phuoc et al. 2022; Van Lierop & El-Geneidy 2016; Widianti et al. 2015). The SOR frame enables to specify this role. According to SOR, stimuli from the service environment such as service innovation, in-vehicle amenities, SST or destination image will impact on a passenger’s internal psychological state, especially satisfaction and positive emotion; then these states will lead to behavioural responses like loyalty intention and WOM behaviour (Asyraff, Hanafiah & Zain 2024; Su, Swanson & Chen 2018). In the informal paratransit context, SQE can be considered as an integrative stimulus group, reflecting how passengers show their overall perception about service and thus form satisfaction or dissatisfaction. Loyalty intention in this research is comprehended as the level of a passenger’s commitment to continuing choosing informal paratransit service, instead of switching to other alternative modes of transport like public bus, traditional taxi, technology motorbike or personal vehicle. Previous findings demonstrated that satisfaction was the strongest predictive variable to loyalty in the urban transport context, especially when combined with trust and perceived value (Khanh Giao & Vuong 2021; Mai Le & Park 2023; Van Lierop & El-Geneidy 2016). With the context that has many alternative choices regarding price and relatively similar amenities, loyalty intention shows the competitive ability of informal paratransit service on the market. On the contrary, referral behaviour, advocacy or WOM are grasped as a higher behavioural outcome, when passengers are not only satisfied but also willing to recommend the service to their acquaintances and share positive experiences through both offline WOM and online WOM, e-WOM (Hamzah et al. 2023; Quy Nguyen-Phuoc et al. 2022). Studies in the sharing economy field present that WOM is often affected indirectly by service quality through satisfaction and loyalty, trust and social value that the service carries at the same time (Arteaga-Sánchez et al. 2020). When WOM is spread broadly, the attraction of services increases, marketing cost can decrease, and a passenger’s trust, therefore, is maintained; this is especially meaningful within the context of informal paratransit that is quite flexible and informal. Consequently, recommendations from current users become a form of important social proof.

The SOR frame, along with the ECT, creates a stable foundation to explain the relationships among variables in the research model. At the stimulus step, the entire SQE is conceptualised as an HOC forming five LOCs, including physical servicescape, in-vehicle amenities, SAE, price–value fairness and SST. At the organism step, the process of a passenger’s internal cognition and emotion is reflected through confirmation and satisfaction. According to ECT, confirmation takes place when the perceived quality is suitable or exceeds initial expectation, thus leading to satisfaction (Fu et al. 2018; Wang et al. 2021). Empirical evidence in the context of carpooling indicates that confirmation acts as a mediating role between service quality and continuance intention, tightly connecting with satisfaction at the same time (Arteaga-Sánchez et al. 2020; Wang et al. 2021). The SOR frame adds an emotional perspective when pointing out that positive internal states like satisfaction, comfortable emotion or emotional attachment can be a bridge ranging from service stimuli to behavioural responses (Asyraff et al. 2024; Su et al. 2018). In relation to response, two main behavioural outcomes considered are loyalty intention and referral behaviour, WOM. Multiple studies in the public transport and shared service field prove that satisfaction is the most crucial predictive factor to loyalty, and it is also a premise of positive WOM behaviour (Hamzah et al. 2023; Mai Le & Park 2023; Quy Nguyen-Phuoc et al. 2022; Van Lierop & El-Geneidy 2016). Within the informal paratransit context, where trust between passengers and service providers is being reinforced, recommendations from satisfied customers can significantly affect a new user’s decision, therefore enabling to stabilise service and develop in an increasingly competitive environment among many diverse modes of transport. In conclusion, the combination between the SOR frame and ECT creates a relatively coherent logical flow: SQE operates as a stimulus, the process of confirming expectation and satisfaction functions as a living entity; loyalty intention and WOM are behavioural responses. This integration relatively exactly reflects both cognition and emotion procedures in service experience and is suitable for empirical proofs related to public transport, shared service and tourism in Asian countries.

There are multiple studies in the field of public transport, informal paratransit and shared service confirming that service quality (SQ) is one of the core factors that decides satisfaction (SA), loyalty intention (LI) and WOM of passengers. Based on SERVQUAL, dimensions such as assurance, empathy, tangible elements and reliability are often grasped as directly effective factors on satisfaction as well as intention to return or continuance intention (Chan et al. 2021; Luke & Heyns 2020; Rahman et al. 2023). Existing work on servicescape and perceived value further underscores the importance of physical environment, price–value fairness and trust for long-term usage. When combining these dimensions, SQE is conceptualised as an HOC including five LOCs: physical servicescape, in-vehicle amenities quality, SAE, price–value fairness and SST. According to the SOR frame, SQE acts as a stimulus role in forming service experience, thereby leading to satisfaction as a psycho-biological state. On the basis of ECT, when practical experience is suitable or goes beyond expectation, the process of confirmation takes place and increases satisfaction (Fu et al. 2018; Wang et al. 2021). Next, satisfaction directly affects loyalty intention and WOM behaviour, which are merely viewed as crucial behavioural responses within the context of informal paratransit (Joewono & Kubota 2007; Quy Nguyen-Phuoc et al. 2022; Van Lierop & El-Geneidy 2016). Remarkably, some recent studies have also noted that service quality can directly influence loyalty and WOM behaviour without necessarily going through satisfaction (Lim et al. 2022; Widianti et al. 2015). Therefore, inspecting both the direct and indirect effect of SQE at the same time is essential to clearly understand the mechanism of forming customer behaviour in the context of informal public transport. On the foundation of the above theoretical claims and empirical proofs, this study suggests seven specific hypotheses. Firstly, SQE is expected to have a positive impact on satisfaction (SA), because SQE reflects a general evaluation about service quality that passengers experience. This is demonstrated in hypothesis:

Secondly, SQE has potential to influence on LI, when positively assessing the service quality that makes passengers willing to continue using informal paratransit service; this content is mentioned in hypothesis:

Thirdly, SQE is also expected to directly affect advocacy or WOM, showing through the willingness tendency to introduce the service to others when passengers have favourable experience; this is generalised in hypothesis:

Next, according to ECT and SOR, satisfaction is considered as a bridge between SQE and behavioural outcomes. The research expects that the more satisfied passengers feel, the higher the probability that they will continue to use the service, mentioned in hypothesis:

Simultaneously, satisfaction is also anticipated to have necessary contribution to form referral behaviour, WOM, expressed in hypothesis:

Finally, based on the viewpoint that SQE can affect LI and WOM through the intermediate psychological mechanism that is satisfaction, this research suggests two intermediate hypotheses:

Thus, the proposed synthesis model considers SQE as a multidimensional HOC, influencing on satisfaction, loyalty intention and referral behaviour, WOM of passengers within the setting of informal public transport. At the same time, it tests both direct and indirect impacts of SQE in the framework of the two-stage approach PLS-SEM.

This research uses the quantitative research design in which data are collected through survey questionnaires and analysed by the PLS-SEM. Partial least squares structural equation modelling is particularly suitable when testing complex models that have latent constructs and multivariate relationships, and concurrently puts emphasis on the prediction of the model (Byon & Jang 2024; Sarstedt et al. 2019). With the actual sample size of 333 observations, the data satisfactorily meet the statistical power requirements in the context of PLS-SEM; the experienced rule called the 10-times rule and current sample size guidelines show that this sample size is suitable for estimating model parameters with high arrow-to and for performing stable bootstrapping (Byon & Jang 2024; Hair & Alamer 2022). In the research model, SQE is conceptualised as a reflective–reflective HOC. To estimate the HOC, this study adopts a two-stage approach to reduce multicollinearity, enhance flexibility when synthesising LOCs and ensure a rigorous evaluation of reliability and validity at both the LOC and HOC levels (Crocetta et al. 2021; Sarstedt et al. 2019). The higher model of SQE is formed from five LOCs, including physical servicescape, in-vehicle amenities, SAE, PV fairness and SST. The dependent variables consist of satisfaction, loyalty intention and advocacy or WOM, which are suitable for the research line in terms of the service quality, pleasure and post-consumption behaviour in transportation (Ha et al. 2019; Kang et al. 2019; Wang et al. 2019).

The scales adopt a reflective measurement structure and use a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The content of the questionnaires is adjusted appropriately with the context of informal paratransit in Vietnam and presented in detail in Online Appendix 1. In practical data, observed variables are coded consistently with the appendix as follows: PS1–PS5 for physical servicescape, AM1–AM4 for in-vehicle amenities quality, SAe1–SAe5 for SAE, PV1–PV4 for price–value fairness, SST1–SST4 for SST, SA1–SA5 for satisfaction, LI1–LI4 for loyalty intention and WOM1–WOM4 for advocacy or WOM. The process of building questionnaires is conducted through forward–backward translation to ensure semantic accuracy. After that, the questionnaires are reviewed in terms of the language and content according to cultural and travel behaviour characteristics in Vietnam, based on the instruction about the adaptability of the measuring tools in the context of multi-diverse culture (Beaton 2000). Next, a pilot test is implemented to test the transparency and the clarity of the research questionnaires, thereby adjusting how to express and interpret before the official survey.

The concepts of HOCs in PLS-SEM can be estimated by two popular approaches that are repeated indicator and two-stage (Crocetta et al. 2021; Sarstedt et al. 2019). The repeated indicator approach assigns all observed variables to the lower-order concepts and to the higher-order concepts at the same time. The advantage is simple and direct, especially suitable when both the higher-order and lower-order have a reflective-reflective structure, in spite of the fact that it will increase the likelihood of multicollinearity (Sarstedt et al. 2019). The two-stage approach proceeds to estimate the factor scores of the LOCs in stage 1, then use them as indicators for HOCs in stage 2. This method helps to reduce the complexity, and it is suitable for large-scale models or when the higher-order structures have mixed nature (Cataldo et al. 2023). Because SQE in this research is modelled as a reflective–reflective HOC, we adopt the two-stage approach. This approach includes two estimations: (1) estimate the lower-order model to extract latent variable scores for LOC; (2) use these scores as indicators to estimate HOC, thereby evaluating directly the contribution level of each LOC to SQE and mitigating the risk of multicollinearity (Sarstedt et al. 2019).

The survey participants in this research are passengers who used informal paratransit services in Vietnam, including various ride-pooling or shared-seat vehicle categories from 4–7 seat cars to 16–45 seat vehicles. The sample criteria are people who are 18 years of age or older and have used informal paratransit at least once during the 12-month period between September 2024 and September 2025. The data were collected through the online questions built on the Google Forms platform combining with direct survey randomly at some big pickup and drop-off points, so as to increase the diverse level of the survey sample. Before participating, respondents are given full information about the research objectives and guaranteed that all their private data information would be anonymous.

During the data collection process, 333 valid responses were obtained and included in the analysis. There are no excluded answers because of the lack of data or inconsistency, thereby making sure the completeness of the entire input data for analysing the PLS-SEM.

The data analysis procedures are processed by using SmartPLS 4.0, with the bootstrap technique that repeats 5000 resamples to estimate the reliability of the scale (Hair & Alamer 2022). Firstly, the data are screened to assess and mitigate common method bias (CMB). Secondly, this research uses the full collinearity index to evaluate the ability of CMB appearance. Additionally, inner VIF indicators for endogenous variables are all less than 3.3, thereby reinforcing evidence that CMB does not significantly affect the research outcomes.

The reflective measurement model is evaluated based on three basic criteria. Firstly, the internal reliability is determined by Cronbach’s alpha and composite reliability (CR), with a minimum acceptance threshold of 0.70. Secondly, the convergent validity is tested by average variance extracted (AVE); the accepted requirement is above 0.50. Concurrently, outer loadings must be above 0.708 (Vishnoi et al. 2024). Thirdly, the discriminant validity with the square root of mean variance extracted (AVE) condition of each concept must be higher than the correlation coefficient between that concept and other concepts in the model. For the SQE-HOC, the evaluation focuses on reliability indicators, convergent value, as well as statistical significance of path coefficients from five LOCs to HOCs (Crocetta et al. 2021; Sarstedt et al. 2019). In the structural model, multicollinearity among variables is tested by inner VIF. These path coefficients are estimated with high reliability that are confirmed thanks to the bootstrap (Hair & Alamer 2022). The explaining potential of the model is reflected through R² indicators. Meanwhile, the effect size is estimated through f², ability of predictive relevance is evaluated by Q², and the overall model fit is tested through SRMR indicators (Afthanorhan, Awang & Aimran 2020). Eventually, the mediating role of satisfaction in the relationship between SQE → loyalty and SQE → WOM is analysed by the indirect bootstrapping method with bias-corrected confidence intervals. At the same time, the remaining direct effects are compared to identify whether the mediation is a full mediation or a partial mediation.

The demographic profile and travel behaviour of the respondents indicate a relatively diverse sample that broadly reflects typical users of informal paratransit services in both urban and interprovincial contexts in Vietnam (see Table 1).

The descriptive results show that the gender structure of the sample is quite balanced, with 55.0% being female and 45.0% being male. The age group focuses mainly on the working-age and middle-aged population, of which 59.2% are in the 30–45 age group and 27.0% are under 30 years old; the remaining 13.8% are over 45 years old. In terms of frequency of use, passengers are distributed relatively evenly among the three groups; however, the group using the service 2–3 times in the past 12 months accounts for the highest proportion (47.7%). This suggests that the majority of respondents have a certain level of familiarity with informal transport services (informal paratransit), rather than just being first-time users. In terms of trip purposes, family travel accounts for an overwhelming proportion of 52.0%, followed by corporate travel at 21.3%, while purposes such as business trips, car rental for weddings or other reasons account for a lower proportion. This structure reflects the significant role of informal transport (informal paratransit) in domestic tourism activities in small and medium groups. In terms of experience in using different types of vehicles, small 4-seat and 7-seat vehicles have a higher experience rate (58.0% and 63.4%, respectively) compared to large 16–45 seat vehicles with rates ranging from 20.1% to 42.6%. This pattern reflects both the common demand of small family groups and short-term business trips, and shows that there is still a significant proportion of passengers who are used to using large vehicles for large tours or long-distance group trips. In Indonesia, paratransit users are primarily found in the working-age population, with a moderate to high frequency of use and service quality is proved to be a core factor that increases satisfaction and loyalty (Joewono et al. 2021). In the Philippines, studies on two-wheeled vehicles in the informal paratransit areas recorded that the gender structure was quite balanced and the purpose of the journey was biased towards daily commute to work. They affirmed the intermediate role of satisfaction in the relationship between service quality and loyalty (Suhartanto et al. 2020). Meanwhile, in China and other Southeast Asian countries, research on the online booking systems and ride-sharing services showed that trust, service quality and perceived value directly affected passengers’ satisfaction and loyalty intention (Chou et al. 2014; Mas-Machuca et al. 2021). In general, the sample structure in this research has many compatible points with the above regional evidence, especially existing in both small and large vehicles in the ecosystem of informal paratransit, as well as the characteristics of users, who are mainly from the working-age group that has a moderate to high frequency of use.

The measurement model testing results for the LOCs show that all measurement scales meet the requirements in terms of reliability and convergent validity. As illustrated in Table 2, all of the scales have Cronbach’s alpha indicators above 0.850, in which the highest indicator is SAE (0.916) and the lowest one is PV fairness (0.850). Simultaneously, composite reliability (CR/rho_c) indicators all exceed the recommended threshold of 0.70 to 0.80, thereby confirming the stability and the internal consistency of the scales (Byon & Jang 2024; Hair & Alamer 2022). On the other hand, AVE indicators of the scales are above 0.50, fluctuating in the range of 0.672–0.755. For SQE in particular, AVE reaches 0.582; even though it is lower than other scales, it still exceeds the recommended threshold for convergent validity in studies adopting PLS-SEM (Afthanorhan et al. 2020; Hair & Alamer 2022). Therefore, it is reasonable to jump to the conclusion that all of the latent constructs in this research, including advocacy, loyalty intention, satisfaction and all components of SQE, meet the standards regarding reliability and convergent validity (Calero Valdez et al. 2023). In parallel, the discriminant validity value is tested through the standards of the Fornell–Larcker criterion and is presented in Table 3. The results demonstrate that the square root of AVE are the values on the diagonal for each construct, which are higher than the off-diagonal correlation coefficients. For example, advocacy = 0.851, loyalty intention = 0.869, satisfaction = 0.853, SAE = 0.866, all are suitable with the recommended threshold of PLS-SEM in relation to ensure discriminant values (Afthanorhan et al. 2020; Byon & Jang 2024). To have a steadier foundation for further evidence, the analysis of cross-loadings indicates that each indicator has the highest loading factor on the constructs it measures. For instance, WOM1–WOM4 variables have loading factor from 0.826 to 0.871 on advocacy; LI1–LI4 variables are from 0.862 to 0.880 on loyalty intention; and SA1–SA5 variables are from 0.836 to 0.877 on satisfaction. This confirms that only indicators are suitable for the original construct of their own, thereby making sure the indicator-level discriminant validity (Afthanorhan et al. 2020). In conclusion, the analysis findings in stage 1 show that LOCs in this research meet all reliability, convergent validity and discriminant validity standards.

The analysis results in stage two show that the second-order variable SQE is stably measured through five LOCs. The outer loadings fluctuate from 0.886 to 0.916, all exceeding the recommended threshold of 0.70 with the reflective model (Hair & Alamer 2022). The synthesis indicators also confirm reliability and convergent value at the HOC level, when Cronbach’s alpha = 0.947, CR = 0.959 and AVE = 0.824. These findings present that SQE fully meets the criteria of a reflective–reflective HOC when being implemented according to the procedure of two-stage in PLS-SEM (Byon & Jang 2024; Sarstedt et al. 2019). In addition, the collinearity test in the inner model shows that all variance inflation factors are below 5; meanwhile, the highest is 4.768 with SQE → advocacy. This is absolutely suitable with the standards of PLS-SEM and not creating any barriers for explaining path coefficient interpretation indicators (Afthanorhan et al. 2020; Hair & Alamer 2022).

The results from the bootstrap analysis with 5000 resamples only point out that all the paths in this model have statistical significance. The standardised path coefficients and significance levels are summarised in Figure 2 and Table 4. Service quality experience is strongly impactful on satisfaction (β = 0.861, p < 0.001), and it concurrently significantly affects loyalty intention (β = 0.499, p < 0.001) and advocacy (WOM) (β = 0.458, p < 0.001). Moreover, satisfaction plays a mediating role, continuing to transfer the influence from SQE to loyalty intention (β = 0.383, p < 0.001) and advocacy (β = 0.255, p = 0.001). Besides, loyalty intention also affects advocacy (β = 0.216, p = 0.001), thus reflecting a hierarchical mediation mechanism which is often seen in transportation service behaviour models (Byon & Jang 2024; Calero Valdez et al. 2023). The adjusted R² coefficients for the three endogenous variables demonstrate a very good explanatory power: advocacy = 0.772, loyalty intention = 0.723 and satisfaction = 0.741 (see Table 5). All of these values exceed common thresholds in PLS-SEM applied studies (Hair & Alamer 2022). Regarding the effect size (f²), the relationship from SQE to satisfaction yields f² = 2.865, which far exceeds the conventional 0.35 threshold for a large effect in PLS-SEM; therefore, it can be classified as very large. This implies that SQE accounts for a substantial additional portion of the variance in satisfaction, thereby reinforcing SQE’s central role in the model. Meanwhile, the SQE → loyalty (f² = 0.233) and SQE → advocacy (f² = 0.194) relationships are medium effects. Satisfaction shows a significant mediating role with f² = 0.138 for loyalty and f² = 0.065 for advocacy, and loyalty → advocacy has f² = 0.057, reflecting a small but significant effect (Afthanorhan et al. 2020; Hair & Alamer 2022). Overall, the structure of results reinforces and supports the centre point of this study: service quality experience not only directly affects loyalty intention and forms advocacy or WOM behaviours but also indirectly goes through satisfaction. This is consistent with recent empirical evidence in the transportation field and digital services when using PLS-SEM (Byon & Jang 2024; Calero Valdez et al. 2023).