Can travel apps improve tourists’ intentions? Investigating the drivers of Chinese gen Y users’ experience

The stimulus-organism-response (S-O-R) model

The S-O-R model is one of the most popular theories used to understand human behavior (Lim et al., 2022a; Mehrabian and Russell, 1974). As its name implies, the model has three primary constructs, namely “stimulus,” “organism,” and “response.” It examines how the stimulus influences individuals’ internal states (organism) and how they respond to it (response). Unlike other behavioral intention models that only fit online or offline contexts (Chen and Yao, 2018), the S-O-R model encompasses a variety of stimuli and organisms (Mehrabian and Russell, 1974) that grants researchers the flexibility to study various behaviors in different settings.

In practice, this model has been widely used in examining atmospheric factors in both information systems and e-commerce contexts (Fang et al., 2017), such as online shopping apps (Peng and Kim, 2014; Rauschnabel et al., 2019), social commerce apps (Hu et al., 2016; Zhang et al., 2014), and branded apps (Lim et al., 2021). Concurrently, myriad tourism studies have also applied the S-O-R model in exploring the intention to use travel websites (Ali, 2016; Gao and Bai, 2014). However, evidence of the S-O-R model's application in the travel app context remains limited (refer to Table 1). To bridge this literature gap, this study adopted the S-O-R model as a theoretical foundation to gain a better understanding of travel app users’ repurchase and recommendation intentions, on the basis of the model's inclusiveness and flexibility in incorporating context-bound constructs. In doing so, we suggested the stimuli to be the drivers of users’ experiences, which in this study were information quality, relative advantages, ease of use, user interface attractiveness, compatibility, and personalization. These drivers were predicted to potentially influence perceived value. Accordingly, the organism in this study was represented by users’ cognitive (perceived value) and affective (user engagement) evaluations (Zhang et al., 2014). Lastly, both repurchase and recommendation intentions were specified as the final outcomes (i.e., responses) of this study.

Psychological reactance theory (PRT)

The PRT assumes that people experience a psychological reactance if something impairs or takes away their specific freedom (Brehm, 1966). This theory is highly useful in describing and contextualizing the technological resistance provoked by the forced disclosure of personal data (Feng and Xie, 2019). For example, users may feel uncomfortable with constant requests for personal information when using travel apps. In other words, reactance occurs when users believe they are being compelled to do something (i.e., forced to disclose data) rather than acting voluntarily. In this regard, reactance is an unpleasant “motivational state that drives freedom restoration” (Rosenberg and Siegel, 2018: 281). However, it is noteworthy that the concept of freedom is not unrestricted freedom, but rather unique, restricted, and concrete freedom (i.e., behavioral realities) (Brehm, 1981). This type of freedom often does not interact with other freedoms. To illustrate, the threatened or eliminated freedom to sleep will not affect a person's freedom to drink a cup of coffee.

The PRT has been used in a variety of extant research to explain users’ decision-making inconsistencies, especially since the interactive nature of the current digital era considerably empowers users. The issue is that more freedom and empowerment also raise the risk of being negatively affected. In the virtual environment, users are frequently compelled to face online pop-up ads (Edwards et al., 2002), online user recommendations (Fitzsimons and Lehmann, 2004; Lee and Lee, 2009), personalized online recommendations (Bleier and Eisenbeiss, 2015), and online information privacy concerns (Chen et al., 2019), all of which result in reactance. These potential threats might lead to negative behavioral patterns such as avoidance, anger, and destructive cognition and affect (Amarnath and Jaidev, 2020). However, some studies have also shown that highly tailored communication with obviously beneficial perception seems to generate little reactance (White et al., 2008). These contrasting results, therefore, highlight the ambiguous nature of threatened freedom (i.e., privacy concerns) and the need for further research in this area. Despite its theoretical implications, empirical research on the influence of the PRT in the tourism domain remains scarce. Hence, this study included privacy concerns as a potential moderator that may affect the relationships between (i) user engagement and repurchase intention and (ii) user engagement and recommendation intention. The following sections expound on these relationships in depth.

Drivers of travel app user experience and perceived value

The term “experience” refers to “perceptions, feelings, and thoughts that consumers have when interacting with products and brands” (Schmitt, 2010: 60). In response to Lemon and Verhoef's (2016) recommendation to analyze user experience in a holistic manner, this study examined the effectiveness of experiential drivers in travel apps that are directly tied to real usage experience. A collection of factors that impact usage experiences, including information quality, relative advantages, ease of use, interface attractiveness, compatibility, and personalization, were included (c.f., Sun et al., 2016).

Information quality. refers to the accuracy and reliability of the information provided by a device (Seck and Philippe, 2013). Due to the intangible nature of the online environment and tourism services, information quality is critical for enhancing users’ experience (Jeong and Shin, 2020). Mobile applications that are accompanied by high-quality information can help users minimize the risk of information traps and allow for the transmission of authentic information. In fact, information quality was identified as a key predictor of perceived value by Chung and Koo (2015) and Sun et al. (2016). Thus, the hypothesis was formulated as:

H1a: Information quality is positively associated with perceived value.

Relative advantages. are defined as the user-perceived benefits of utilizing a travel app (Fang et al., 2017). Usability is one of the most important considerations for mobile travel businesses (Sakshi et al., 2020). For example, online hotel booking allows users to search and compare different hotel companies and rooms, analyze cost-effectiveness, and make bookings to save both money and time (Mathwick et al., 2001). Bilgihan et al. (2014) also emphasized the impact of online channel usefulness on travel provider–client relationship marketing. To create a pleasant experience and make users perceive apps’ value positively, travel apps should offer various benefits to users. The following hypothesis was thus proposed:

H1b: Relative advantages are positively associated with perceived value.

Ease of use. refers to how easy a travel app is to comprehend and utilize (Fang et al., 2017). According to Ozturk et al. (2016), ease of use is the key factor that influences users to accept and adopt a travel app. In practice, users may not be able to positively evaluate the value of switching from a website to an app if the new platform is difficult to understand and navigate. Generally, ease of use reflects the desire of travel apps to understand and care for their users (Bilgihan et al., 2016). Easier usage leads to less strain, thereby influencing users’ experience as well as their value perception toward the apps (Assaker et al., 2020; Lim et al., 2022b). Thus, it is vital to consider ease of use as one of the key drivers that influence the perceived value of travel apps. Therefore, this study postulated that:

H1c: Ease of use is positively associated with perceived value.

User interface attractiveness. measures the visual appeal of a mobile travel app (Fang et al., 2017). A visually presentable interface is widely recognized as an effective technique to attract users and generate a positive experience (Coursaris and Van Osch, 2016). An attractive interface may increase users’ perceived value, while an uninteresting interface may reduce the motivation of users to continue using the app (Wu et al., 2017). Hence, a travel app with an attractive user interface tends to be perceived as valuable by users. Accordingly, this study proposed that:

H1d: User interface attractiveness is positively associated with perceived value.

Compatibility. is the degree to which a travel app matches users’ travel interests (Fang et al., 2017). In the travel app setting, compatibility with users’ travel styles and preferences might foster quick adoption (Al-Jabri and Sohail, 2012; Lim et al., 2022c). Specifically, a good match between an app's technical features and task execution affects users’ perceived app performance and satisfaction (Lin and Wang, 2012; Ray et al., 2014). Travel apps that fit the needs of their users are more likely to be viewed favorably by users (Kapoor et al., 2014). Thus, when a travel app is compatible with users’ expectations, it will be perceived as having great value (Kang et al., 2015; Wang, 2019). The following hypothesis was postulated to reflect this relationship:

H1e: Compatibility is positively associated with perceived value.

Personalization. is the degree to which a travel app caters to a user's specific request or need (Jeong and Shin, 2020). Bhalla (2014) considers personalization as a top-tier need that can assess user experience. With the advancement of technology, the data stored in apps enables service providers to assess and provide individualized products and services (Godfrey et al., 2011; Oh and Teo, 2010). As a result, personalization can be utilized to offer users great deals that match their preferred consumption style, thus decreasing ambiguity and saving time (Gao and Su, 2017; Reis et al., 2020). This makes users feel convenience and value, resulting in positive perceived value (Schramm-Klein et al., 2011). Thus, the following hypothesis was proposed:

H1f: Personalization is positively associated with perceived value.

Perceived value and user engagement

Perceived value refers to an individual's cognitive-affective evaluation of a product or service based on a comparison between its benefits and costs; in this case, both utilitarian and hedonic values are considered the most important when using travel apps (Kim et al., 2021; Pandža Bajs, 2015). In practice, utilitarian value often relates to the functionalities offered by the travel app, while hedonic value refers to the enjoyment travelers derive from using the travel app. User engagement is extremely context-based, such that each digital setting has its distinctive features to stimulate a degree of engagement (O’Brien et al., 2018). Meanwhile, perceived value is a summary assessment of the distinctive features of different settings. When users perceive value in a service, they are more inclined to engage with it (Cheung et al., 2022; Pansari and Kumar, 2017). This suggests that a user's perceived values have a discernible effect on engagement. For example, O’Brien (2010) established hedonic and utilitarian incentives as antecedents of client engagement. Claffey and Brady (2015) likewise discovered a favorable connection between cognitive evaluation and affective commitment. Therefore, perceived value is expected to influence user engagement. This research formulated the hypothesis as:

H2: Perceived value is positively associated with user engagement.

User engagement and repurchase and recommendation intentions

User engagement is defined as the level of interaction and association between a service provider and its users, which can be measured using three dimensions: social connection, enthusiastic participation, and conscious attention (Vivek et al., 2014). Repurchase intention relates to how likely a user is to repurchase products/services from a company, while recommendation intention refers to how likely a user is to tell others about favorable product/service reviews (Lee et al., 2019). Studies in the marketing field have demonstrated unequivocally that customer engagement is positively correlated with repurchase intention and positive word-of-mouth (e.g., Zheng et al., 2022; Bitter and Grabner-Kräuter, 2016; Ran et al., 2021; Cheung et al., 2015). In the tourism sector, highly engaged users are also presumably more likely to buy products from travel apps and promote favorable word-of-mouth about such apps. Therefore, this study predicted that:

H3: User engagement is positively associated with repurchase intention.

The mediating role of user engagement

Currently, the concept of engagement is defined from a consumer-centric perspective rather than the earlier product-centric perspective, so as to examine additional internal mechanisms of behavioral intention (Kumar et al., 2019; Pansari and Kumar, 2017). In this regard, user engagement is seen as a critical component of tourism marketing owing to its close association with a positive user experience (Grewal et al., 2017). By providing pleasant experiences to users before, during, and after their journey via online and offline communication channels (Hwang and Seo, 2016), travel businesses may develop long-term relationships with them, resulting in engaged states and increased user loyalty (Lai, 2014). Specifically, the existing literature has established that engagement is an important mechanism to strengthen the relationship between perceived value and behavioral responses (Dabbous and Barakat, 2020; Zhang et al., 2017). Consequently, the following hypotheses were developed:

H5a: User engagement mediates the relationship between perceived value and repurchase intention.

The moderating role of privacy concerns

This study postulated that the effect of user engagement on both repurchase intention and recommendation intention may vary due to the potential influence of boundary conditions like privacy concerns. The PRT explains the underlying reason for including the boundary effect of privacy concerns. According to Brehm (1966), the PRT suggests that people are likely to experience psychological reactance when they feel that their control over free choice is threatened or eliminated (e.g., forced disclosure of private data).

Users may oppose technology that compels them to divulge personal information (Feng and Xie, 2019). To mitigate privacy concerns, they would be motivated to restore their control of data freedom after losing it. Nowadays, threats to one's sense of “control” via online and offline channels such as feedback forms, smartphone apps, and social media sites can result in undesirable behavioral patterns, including avoidance, rage, and negative cognition (Amarnath and Jaidev, 2020). Since privacy is a crucial determinant of the quality of an experience, privacy threats can significantly alter users’ decision-making when using travel apps (Rahman et al., 2022). Individuals with higher privacy concerns would be warier and more anxious about the use of personal data (Mpinganjira and Maduku, 2019). These concerns undermine users’ confidence and inclination to continue with a mobile travel app, resulting in unfavorable reactions (e.g., low ratings and negative reviews) (Cheah et al., 2020). Therefore, this study formulated the hypotheses:

H6a: Privacy concerns moderate the relationship between user engagement and repurchase intention, such that low privacy concerns strengthen the relationship.

Finally, the literature has reported that demographic factors, including both gender and age, impact how individuals respond to the use of technology devices. For instance, female users are more likely to buy travel goods offered by travel apps and to recommend them to others, while male users are more careful in their purchases and recommendations (Allan et al., 2022). Also, a decent job with a high monthly income might increase spending, and vice versa (Mancino et al., 2018). The research model, therefore, included gender, marital status, age, education level, occupation, and monthly income as control variables to ensure there were no confounding variables or erroneous results. The research model is depicted in Figure 2, and the methodology is explained in the section that follows.

OPEN IN VIEWER

Figure 2.

Research model. Note: dashed line box represents lower-order constructs. PV = perceived value; UE = user engagement; RPI = repurchase intention; RDI = recommendation intention.

Data collection procedures

A self-administered questionnaire was employed to collect data from Gen Y travel app users in Mainland China. In this study, the Ctrip app was selected as the main setting for two main reasons. First, Ctrip is the largest tourism provider in China, owning over 50% of the travel market share (Kate, 2017). Second, as of December 2021, the Ctrip app recorded the highest number of active users in China (Statista, 2021a). An online survey was designed using the Wenjuanxing platform (a Chinese Qualtrics-like platform) due to COVID-19 concerns. The purposive sampling technique was used to ensure that the survey was answered by valid respondents. Notably, we established the inclusion requirements for the survey platform prior to data collection, whereby the respondents had to be China nationals aged 20 to 40 (i.e., Gen Y) with at least six months of usage and purchase history with the Ctrip app. In total, 401 responses were collected between October 2021 and January 2022, of which 14 straight-lining responses were deleted using the case-by-case method. Of the 387 remaining responses, the majority were women (55.7%), aged 31 to 35 (45.2%), and had completed their undergraduate degree (76.5%). A large proportion of them were working in the private sector (60.7%) and earned between 11,001 and 13,000 yuan ($1700 and $2000) per month (20.2%) (see Table 2).

OPEN IN VIEWER

Table 2.

Respondent profile.

Category	Item	Frequency (n = 387)	Percent (%)
Gender	Male	171	44.2
	Female	216	55.8
Marital status	Single	105	27.1
	Married	282	72.9
Age	20 years old and below	9	2.3
	21–25 years old	32	8.3
	26–30 years old	119	30.7
	31–35 years old	175	45.2
	36–40 years old	52	13.4
Education	Undergraduate degree (B.Sc., B.A., etc.)	296	76.5
	Graduate Degree (M.Sc., MA, MBA, etc.)	74	19.1
	Postgraduate degree or professional certificates (PhD, DBA, etc.)	17	4.4
Occupation	Employee - Government sector	114	29.5
	Employee - Private sector	235	60.7
	Self-employed	21	5.4
	Student	16	4.1
	Unemployed	1	0.3
Monthly income	Below RMB 3000	7	1.8
	RMB 3001–5000	16	4.1
	RMB 5001–7000	36	9.3
	RMB 7001–9000	63	16.3
	RMB 9001–11,000	65	16.8
	RMB 11,001–13,000	78	20.2
	RMB 13,001–15,000	69	17.8
	RMB 15,001 and above	53	13.7

Measures

All measures were adapted from past studies and were modified to the context of travel apps. To validate the measurements, the questionnaire was pre-tested with a panel of ten members. Based on the comments received, minor changes were made to the survey. Subsequently, a pilot test was conducted with 50 Ctrip app users before the actual survey. Perceived value was measured using two dimensions, i.e., utilitarian and hedonic (Jahn and Kunz, 2012). User engagement was measured using three dimensions: social connection, enthused participation, and conscious attention (Vivek et al., 2014). Information quality was assessed using the scale from Seck and Philippe (2013), while relative advantages, ease of use, user interface attractiveness, and compatibility were measured using Fang et al.'s (2017) scale. The scales for personalization (Jeong and Shin, 2020), privacy concerns (Kim, 2020), and user intention (i.e., repurchase and recommendation intentions) (Lee et al., 2019) were also adapted from previous studies.

Common method variance (CMV)

This study endeavored to reduce CMV with two widely accepted approaches. To begin, the procedural remedy was applied, in which exogenous and endogenous variables were measured using two separate Likert scales (5-point and 7-point). For the statistical remedy, we first adopted Kock and Lynn's (2012) full collinearity approach. As shown in Table 3, the variance inflation factor (VIF) values for all constructs were estimated to be less than 3.33 (between 1.073 to 3.32), indicating that CMV had a negligible effect in the current dataset (Kock and Lynn, 2012). Subsequently, Harman's single-factor method was used, wherein the primary extracted component explained only 37.679% of the total variance (less than 40%), suggesting no single dominant factor (Fuller et al., 2016).

OPEN IN VIEWER

Table 3.

Results of reliability, convergent validity, and full collinearity.

Construct	Loading
Information Quality (IFQ) [α = 0.743; rho_A = 0.746; CR = 0.854; AVE = 0.661; Full Collinearity = 1.073]
IFQ1: The Ctrip travel app provides me with useful information.	0.827
IFQ2: The Ctrip travel app meets my information needs.	0.842
IFQ3: Information on the Ctrip travel app is well-updated.	0.769
Relative Advantages (RA) [α = 0.741; rho_A = 0.746; CR = 0.837; AVE = 0.563; Full Collinearity = 2.140]
RA1: The Ctrip travel app enhances my traveling experience.	0.734
RA2: The Ctrip travel app makes it easier for me to understand travel destinations.	0.782
RA3: The Ctrip travel app helps me save a lot of time.	0.710
RA4: The Ctrip travel app makes my traveling more effective.	0.773
Ease of Use (EOU) [α = 0.724; rho_A = 0.730; CR = 0.844; AVE = 0.644; Full Collinearity = 1.134]
EOU1: The Ctrip travel app is easy to use.	0.821
EOU2: Learning how to use the Ctrip travel app is easy for me.	0.818
EOU3: I would imagine that most people would learn to use the Ctrip app very quickly.	0.767
User Interface Attractiveness (UIA) [α = 0.877; rho_A = 0.878; CR = 0.908; AVE = 0.622; Full Collinearity = 3.153]
UIA1: The interface design of the Ctrip travel app looks clean.	0.654
UIA2: The interface design of the Ctrip travel app is sophisticated.	0.806
UIA3: The interface design of the Ctrip travel app is fascinating.	0.818
UIA4: The interface design of the Ctrip travel app is aesthetically pleasing.	0.805
UIA5: The interface design of the Ctrip travel app is visually appealing.	0.818
UIA6: The interface design of the Ctrip travel app is attractive.	0.818
Compatibility (CMP) [α = 0.773; rho_A = 0.776; CR = 0.868; AVE = 0.688; Full Collinearity = 1.546]
CMP1: The Ctrip travel app is compatible with my travel preferences.	0.812
CMP2: The Ctrip travel app suits my travel needs.	0.825
CMP3: The Ctrip travel app fits well with my travel needs.	0.85
Personalization (PSN) [α = 0.761; rho_A = 0.762; CR = 0.863; AVE = 0.677; Full Collinearity = 1.890]
PSN1: The Ctrip travel app allowed me to receive tailored information.	0.796
PSN2: I could interact with the Ctrip travel app to get personalized information.	0.837
PSN3: The personalized information provided by the Ctrip travel app met my needs.	0.835
Perceived Value (PV)
(i) Utilitarian Value (UTV) [α = 0.792; rho_A = 0.794; CR = 0.866; AVE = 0.617; Full Collinearity = 1.416]
UTV1: The Ctrip app is helpful for me.	0.834
UTV2: The Ctrip app is useful for me.	0.774
UTV3: The Ctrip app is functional for me.	0.733
UTV4: The Ctrip app is practical for me.	0.799
(ii) Hedonic Value (HDV) [α = 0.864; rho_A = 0.864; CR = 0.907; AVE = 0.710; Full Collinearity = 2.545]
HDV1: The Ctrip app is fun.	0.842
HDV2: The Ctrip app is exciting.	0.833
HDV3: The Ctrip app is pleasant.	0.842
HDV4: The Ctrip app is entertaining.	0.853
User Engagement (UE)
(i) Conscious Attention (CA) [α = 0.781; rho_A = 0.785; CR = 0.872; AVE = 0.695; Full Collinearity = 1.252]
CA1: Anything related to Ctrip grabs my attention.	0.857
CA2: I like to learn more about Ctrip.	0.798
CA3: I pay a lot of attention to anything about Ctrip.	0.845
(ii) Enthused Participation (EP) [α = 0.782; rho_A = 0.795; CR = 0.860; AVE = 0.606; Full Collinearity = 2.189]
EP1: I spend a lot of my discretionary time on Ctrip.	0.792
EP2: I am heavily into Ctrip.	0.812
EP3: I am passionate about Ctrip.	0.824
EP4: My days would not be the same without Ctrip.	0.678
(iii) Social Connection (SC) [α = 0.785; rho_A = 0.788; CR = 0.875; AVE = 0.700; Full Collinearity = 2.144]
SC1: I love Ctrip with my friends.	0.874
SC2: I enjoy Ctrip more when I am with others.	0.827
SC3: Ctrip is more fun when other people around me do it too.	0.808
Privacy Concerns (PR) [α = 0.943; rho_A = 0.946; CR = 0.957; AVE = 0.815; Full Collinearity = 3.320]
PR1: I am concerned that Ctrip is collecting too much personal information from me.	0.891
PR2: I am concerned that Ctrip will use my personal information for other purposes without my authorization.	0.928
PR3: I am concerned that Ctrip will share my personal information with other entities without my authorization.	0.918
PR4: I am concerned that unauthorized persons (e.g., hackers) may have access to my personal information collected by Ctrip.	0.885
PR5: I am concerned about the privacy of my personal information during e-transactions with Ctrip.	0.892
Repurchase Intention (RPI) [α = 0.866; rho_A = 0.868; CR = 0.909; AVE = 0.714; Full Collinearity = 1.313]
RPI1: I intend to continue to purchase from Ctrip.	0.868
RPI2: I intend to acquire products and services from Ctrip.	0.868
RPI3: I intend to choose Ctrip as the preferred brand for my future purchases.	0.816
RPI4: Except for any unanticipated reasons, I intend to continue to purchase from Ctrip as usual.	0.825
Recommendation Intention (RDI) [α = 0.909; rho_A = 0.910; CR = 0.936; AVE = 0.785; Full Collinearity = 2.149]
RDI1: I say positive things about Ctrip to other people.	0.858
RDI2: I recommend Ctrip to anyone who seeks my advice.	0.904
RDI3: I encourage friends to purchase from Ctrip.	0.883
RDI4: I refer my acquaintances to Ctrip.	0.899

Note: CR = composite reliability; AVE = average variance extracted.

Reflective measurement model

The measurement model was assessed with various metrics. First, this study examined the reliability of the constructs. The results demonstrated that all values of Cronbach's α, rho_A, and composite reliability (CR) exceeded 0.70 (Hair et al., 2019) (see Table 3). Second, the outer loadings and average variance extracted (AVE) were evaluated to assess the constructs’ convergent validity. As presented in Table 3, most items’ outer loadings satisfied the recommended threshold of 0.40 (Bagozzi et al., 1991) ¹ , whereas all AVE results were above the 0.50 minimum threshold (Bagozzi and Yi, 1988; Fornell and Larcker, 1981), thereby confirming convergent validity. Finally, discriminant validity was evaluated using the Heterotrait-Monotrait (HTMT) ratio. As shown in Table 4, all HTMT values fell below the threshold of 0.85 (Kline, 2011), demonstrating satisfactory discriminant validity.

OPEN IN VIEWER

Table 4.

Discriminant validity result using the heterotrait-monotrait (HTMT) ratio of correlations.

	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1. Compatibility
2. Conscious attention	0.468
3. Enthused participation	0.616	0.775
4. Hedonic value	0.714	0.684	0.749
5. Information quality	0.616	0.371	0.526	0.514
6. Ease of use	0.617	0.452	0.543	0.636	0.710
7. Relative advantages	0.754	0.581	0.581	0.749	0.737	0.764
8. Personalization	0.748	0.550	0.660	0.666	0.603	0.620	0.615
9. Privacy concern	0.356	0.426	0.526	0.456	0.256	0.250	0.252	0.391
10. Recommendation intention	0.642	0.618	0.743	0.679	0.567	0.567	0.647	0.633	0.527
11. Repurchase intention	0.648	0.539	0.631	0.639	0.633	0.678	0.686	0.623	0.451	0.768
12. Social connection	0.695	0.727	0.821	0.727	0.590	0.638	0.684	0.743	0.520	0.850	0.803
13. User interface attractiveness	0.694	0.742	0.742	0.849	0.586	0.647	0.813	0.708	0.495	0.689	0.621	0.736
14. Utilitarian value	0.795	0.460	0.646	0.685	0.741	0.814	0.804	0.702	0.289	0.752	0.839	0.805	0.681

Higher-order constructs (HOCs)

Both perceived value and user engagement were specified as reflective-formative HOCs (comprising several lower-order constructs - LOCs) and were assessed with a split two-stage approach (Becker et al., 2022; Sarstedt et al., 2019). First, convergent validity demonstrated satisfactory results under the single-global item approach proposed by Cheah et al. (2018). The single global items for perceived value (i.e., “Overall, the service of the travel app provides decent value to users”) and user engagement (i.e., “Overall, I am engaged with the service of the travel app”), generated path coefficients of 0.775 and 0.812, respectively. Second, collinearity did not pose a problem among the LOCs, since the VIF values ranged from 1.472 to 1.986 (see Table 5) (Becker et al., 2015). Finally, both utilitarian value ( = 0.546) and hedonic value ( = 0.584) had a significant influence on perceived value (p < 0.05), while the three sub-dimensions of user engagement (conscious attention = 0.099; enthusiastic involvement = 0.328; and social connection = 0.689) exhibited significant effects (p < 0.05). Thus, it was again confirmed that both perceived value and user engagement were formatively constituted by several LOCs.

OPEN IN VIEWER

Table 5.

Assessment of higher-order constructs.

Higher order construct	Sub-dimension for LOC	Convergent validity	Outer VIF	Outer weights	t-value	p-value	Confidence interval
Perceived value	Utilitarian value	0.775	1.472	0.546	12.376**	0.000	(0.470,0.611)
	Hedonic value		1.472	0.584	12.702**	0.000	(0.510,0.660)
User engagement	Conscious attention	0.812	1.718	0.099	1.966*	0.025	(0.016,0.181)
	Enthused participation		1.986	0.328	5.250**	0.000	(0.226,0.436)
	Social connection		1.873	0.689	11.684**	0.000	(0.587,0.780)

Note: *p < 0.05; **p < 0.01, VIF = variance inflation factor.

Structural model

All the VIF values of the exogenous constructs were less than 3.33 (from 1.766 to 2.534), indicating that inner collinearity was not an issue in this study (Becker et al., 2015). The bootstrapping technique (with 10,000 sub-samples) was then utilized to assess the significance of the path coefficients (Becker et al., 2022). The results indicated insignificant effects of the six control variables (i.e., gender, marital status, age, level of education, occupation, and monthly income) across the model (see Table 6).

OPEN IN VIEWER

Table 6.

Results of the structural model.

Path relationship	Std. Beta	Std. Error	t-value	p-value	CI (2.5%, 97.5%)	Decision	VIF	f2	R2	Q2_predict
H1a) Information quality -> PV	0.055	0.045	1.237	0.108	(−0.018,0.129)	NS	1.663		0.719	0.704
H1b) Relative advantages -> PV	0.156	0.048	3.244**	0.001	(0.075,0.232)	S	2.276	0.038(S)
H1c) Ease of use -> PV	0.194	0.041	4.780**	<0.001	(0.125,0.259)	S	1.727	0.078(S)
H1d) User interface attractiveness -> PV	0.317	0.062	5.124**	<0.001	(0.218,0.421)	S	2.219	0.162(M)
H1e) Compatibility -> PV	0.229	0.042	5.517**	<0.001	(0.164,0.300)	S	1.908	0.098(S)
H1f) Personalization -> PV	0.109	0.043	2.536**	0.006	(0.041,0.184)	S	1.812	0.023(S)
H2) PV -> UE	0.744	0.028	26.818**	<0.001	(0.697,0.787)	S	1.000	1.239(L)	0.553	0.524
H3) UE -> RPI	0.673	0.039	17.103**	<0.001	(0.606,0.733)	S	1.000	0.829(L)	0.453	0.382
H4) UE -> RDI	0.754	0.029	26.115**	<0.001	(0.704,0.799)	S	1.000	1.315(L)	0.568	0.413
H5a) PV -> UE -> RPI	0.501	0.043	11.745**	<0.001	(0.429,0.571)	S
H5b) PV-> UE -> RDI	0.561	0.037	15.219**	<0.001	(0.500,0.619)	S
H6a) Privacy concern*UE -> RPI	−0.053	0.063	0.847	0.199	(−0.160, 0.047)	NS
H6b) Privacy concern*UE -> RDI	−0.089	0.050	1.795*	0.036	(−0.169, −0.004)	S
Control Variable
Gender -> RPI	0.001	0.038	0.016	0.987	(−0.072,0.079)
Marital status -> RPI	−0.059	0.041	1.439	0.150	(−0.135,0.026)
Age -> RPI	0.009	0.046	0.204	0.839	(−0.082,0.097)
Education -> RPI	0.030	0.039	0.763	0.446	(−0.045,0.109)
Occupation -> RPI	−0.060	0.040	1.493	0.136	(−0.139,0.017)
Income -> RPI	0.016	0.043	0.373	0.709	(−0.068,0.101)
Gender -> RDI	0.061	0.034	1.810	0.070	(−0.006,0.126)
Marital status -> RDI	0.022	0.041	0.539	0.590	(−0.058,0.106)
Age -> RDI	−0.011	0.042	0.258	0.796	(−0.099,0.070)
Education -> RDI	0.010	0.036	0.273	0.785	(−0.062,0.080)
Occupation -> RDI	0.004	0.034	0.108	0.914	(−0.064,0.069)
Income -> RDI	−0.014	0.041	0.341	0.733	(−0.094,0.066)

Notes: *p < 0.05; **p < 0.01; PV = perceived value; UE = user engagement; RPI = repurchase intention; RDI = recommendation intention; CI = confidence interval; NS = not supported; S = supported; effect size; S = small; M = medium; L = large.

User interface attractiveness was revealed to be the strongest predictor of users’ perceived value (β = 0.317; p < 0.001), supporting H1d. Information quality, contrary to the hypothesis, did not influence perceived value (β = 0.055; p = 0.108) in a significant way; therefore, H1a was not supported. Further, relative advantages (β = 0.156; p = 0.001), ease of use (β = 0.194; p < 0.001), compatibility (β = 0.229; p < 0.001), and personalization (β = 0.109; p = 0.006) all demonstrated a significant positive impact on perceived value, supporting H1b, H1c, H1e, and H1f. Next, perceived value (β = 0.744; p < 0.001) was shown to affect user engagement positively, supporting H2. The results also confirmed the positive effect of user engagement on repurchase intention (β = 0.673; p < 0.001) and recommendation intention (β = 0.754; p < 0.001), supporting both H3 and H4. The mediating role of user engagement was assessed with Hayes’s (2018) method (see Table 6). The results indicated that user engagement significantly mediates the paths linking perceived value to repurchase intention (β = 0.501; p < 0.001) and recommendation intention (β = 0.561; p < 0.001), supporting H5a and H5b (refer to Figure 3 for a structural model illustrated with path coefficient results). Moreover, according to Lachowicz, Preacher, and Kelley (2018) ² , the effect sizes of these two mediations were large at 0.251 and 0.315, respectively.

OPEN IN VIEWER

Figure 3.

Structural model (with path coefficient results). Note: *p < 0.05; **p < 0.01, ns = not significant; PV = perceived value; UE = user engagement; RPI = repurchase intention; RDI = recommendation intention.

Overall, 45.3% of the variance in repurchase intention and 56.8% of the variance in recommendation intention were explained by user engagement, while 55.3% of the variance in user engagement was explained by perceived value. Notably, information quality, relative advantages, ease of use, user interface attractiveness, compatibility, and personalization were able to explain 71.9% of the variance in perceived value (see Table 6). Following Cohen’s (1992) guidelines, user engagement exhibited a large effect on repurchase intention (ƒ² = 0.829) and recommendation intention (ƒ² = 1.315), as did perceived value (ƒ² = 1.239) on user engagement. Relative advantages (ƒ² = 0.038), ease of use (ƒ² = 0.078), user interface attractiveness (ƒ² = 0.162), compatibility (ƒ² = 0.098), and personalization (ƒ² = 0.023) all reported small effect sizes on perceived value.

Finally, predictive relevance was examined (Geisser, 1975; Stone, 1974). The Q²_predict values for perceived value (0.704), user engagement (0.524), repurchase intention (0.382), and recommendation intention (0.413) were greater than zero, validating the predictive relevance of the model (Chin et al., 2020). The PLSpredict technique developed by Shmueli et al. (2019) was used to further assess the endogenous constructs’ predictive relevance. As shown in Table 7, except for RPI1, RPI2, RDI1, and RDI3, both the RMSE and MAE of the PLS model for repurchase intention and recommendation intention were lower than those of the linear model (LM), indicating a smaller prediction error. Therefore, both repurchase intention and recommendation intention demonstrated medium predictive potential (Shmueli et al., 2019).

OPEN IN VIEWER

Table 7.

Assessment of PLS predict.

	PLS			LM			PLS-LM			Predictive power
Methods	RMSE	MAE	Q2_Predict	RMSE	MAE	Q2_Predict	RMSE	MAE	Q2_Predict
Repurchase Intention										Medium
RPI1	0.917	0.715	0.347	0.856	0.663	0.431	0.061	0.052	−0.084
RPI2	1.046	0.809	0.349	1.007	0.766	0.397	0.039	0.043	−0.048
RPI3	1.160	0.913	0.245	1.163	0.873	0.240	−0.003	0.040	0.005
RPI4	1.015	0.814	0.246	1.042	0.818	0.207	−0.027	−0.004	0.039
Recommendation Intention										Medium
RDI1	0.988	0.780	0.357	0.983	0.774	0.364	0.005	0.006	−0.007
RDI2	1.166	0.936	0.355	1.170	0.893	0.352	−0.004	0.043	0.003
RDI3	1.131	0.920	0.347	1.118	0.883	0.362	0.013	0.037	−0.015
RDI4	1.058	0.850	0.362	1.075	0.856	0.341	−0.017	−0.006	0.021

Notes: RMSE and MAE metrics in PLS must produce smaller values than that of LM, thus generating negative values in PLS-LM; Q²_predict metric in PLS must produce larger values than that of LM, thus generating positive values in PLS-LM.

Moderating effect

The moderating effect of privacy concerns was examined using a two-stage latent interaction approach (Becker et al., 2018). The results indicated that users’ privacy concerns have no effect on the link between user engagement and repurchase intention (β = −0.053; p-value = 0.199); thus, H6a was rejected. In contrast, privacy concerns were found to significantly impact the relationship between user engagement and recommendation intention (β = −0.089; p-value = 0.036), supporting H6b. The interaction plot further depicted that the relationship is weaker when a high level of privacy concerns exists and vice versa (see Figure 4: the slope of the solid line is steeper than that of the dotted line).

OPEN IN VIEWER

Figure 4.

User engagement*privacy concerns on recommendation intention.

Discussion of results

This study's first objective was to examine the experiential drivers of travel apps that influence users’ perceived value. Drawing on the S-O-R model (Mehrabian and Russell, 1974), this study found that relative advantages, ease of use, user interface attractiveness, compatibility, and personalization are the significant factors that determine a travel app's perceived value, while information quality appears to have no such effect (H1a was rejected). The insignificant finding is inconsistent with previous studies that reported a positive relationship between information quality and user behavior (Kullada and Kurniadjie, 2021; Majeed et al., 2020). This could be due to the fact that users tend to perceive and assess the quality of online tourism information differently, resulting in different emotions and cognitive responses (Kim et al., 2017). The inconsistency might also be caused by the fact that information quality is an essential component of any contemporary travel app. In the era of big data, a single travel app can rarely hold exclusive travel-related information. In addition, it may appear that users seldom rely on a single resource when searching for travel-related information. For example, they prefer to evaluate the best offers from multiple travel service providers, both online and offline.

Next, user interface attractiveness reported the highest path coefficient score (β = 0.317) and largest effect size (f ² = 0.162) towards perceived value (H1d was supported). This once again evidences that visual attractiveness is a critical feature for enhancing value perception when using a travel app, especially in the current digital age where users are exposed to countless apps. An interface's visually pleasing page, user-friendly functions, and well-organized information soothe the emotions of impatient users, thereby enhancing their perception of value. Gursoy (2019) also reaffirmed that an attractive interface not only helps relieve users’ moods and mental stress but also enhances value perception due to its straightforwardness and intuitiveness.

Apart from that, the compatibility between users’ travel interests and the travel app positively affects perceived value as well (H1e was supported). This result suggests that users of travel apps appreciate the feeling of mutual choice. Travel app users have diverse beliefs and backgrounds, resulting in varied tastes. A single travel app may not be able to accommodate all preferences, but one thing is certain—the greater the compatibility between a travel app's offer and its users’ needs, the higher the chance of users valuing the app. This is consistent with Amaro and Duarte’s (2015) finding that users will be interested in a travel app if it fits their preferences and interests. In addition, both relative advantages (H1b was supported) and ease of use (H1c was supported) exert positive influences on perceived value, which is consistent with prior studies (see Lee et al., 2021; Bravo et al., 2021; Morosan et al., 2021). The ease of use determines how quickly a user can become familiar with the travel app, while the relative advantages inform users of the benefits provided. By reducing learning costs and saving energy, ease of use helps users reduce mental burdens. Similarly, relative advantages relate to the tangible benefits that users enjoy, such as saving money and increasing efficiency in planning trips. It is always the case that a useful travel app can make users feel the worth of spending time using and exploring it, resulting in a positive perception. Lastly, personalization was found to be positively correlated with perceived value (H1f was supported). Similar to the finding of Wang et al. (2020), the low path coefficient (β = 0.109) indicates that personalization may not be the panacea for increasing user value under the threat of privacy violations, as users are often required to disclose more personal data for accurate and personalized offers.

For the second objective, this study affirms that user engagement is a crucial mechanism linking perceived value to repurchase and recommendation intentions (H5a and H5b were supported). Underpinned by the S-O-R model, external environmental stimuli (i.e., travel app experiential features) trigger cognitive (perceived value) and affective (user engagement) evaluations, which in turn, lead to responses (i.e., repurchase and recommendation intentions). It is further revealed that it is important for an app to be able to engage with Gen Y users to stimulate long-term results, such as repurchasing travel offers from the app or recommending the app to others. Similarly, our findings indicate that users who engage with travel apps tend to have a higher level of perceived value (H2 was supported), leading to the desirable states of intention to repurchase (H3 was supported) and intention to recommend the travel app to others (H4 was supported). This finding is similar to that of Dabbous and Barakat (2020), who discovered the mediating effect of customer engagement between perceived value and purchase intention. In practice, users who engage the most time and energy in travel apps are always those who have had pleasant experiences, and it is true that this group of users is more likely to buy from and recommend travel apps.

For the last objective, we evidenced that privacy concerns do not influence engaged users’ repurchase intention (H6a was rejected), while those who are engaged and have low privacy concerns are more likely to recommend the travel app to others (H6b was supported). This conclusion is corroborated by the PRT, wherein the freedom threat of privacy concerns triggers the psychological reactance of travel app users, resulting in inconsistent reactions. As predicted, users’ weaker perception of privacy invasion contributes to their willingness to spread favorable word-of-mouth. That is, vigilant users may alter their behavior to avoid potential loss due to excessive engagement when they believe they are facing high privacy risks. This significant conclusion corresponds to Ladhari et al.’s (2019) research, which found that the Gen Y segment (i.e., the target group of this study) is more skeptical and pragmatic than previous generations. Alternatively, our interaction analysis found that regardless of whether privacy concerns are low or high, the effect of user engagement on repurchase intention is similar, indicating an insignificant moderating effect. This again demonstrates that the positive effect of user engagement on repurchase intention is not affected by the degree of privacy concern, corroborating earlier findings that user engagement remains the key factor in stimulating positive repurchase intention (Lee et al., 2019).

Theoretical implications

This study presents several significant contributions to the extant literature. First, we expand the theoretical boundaries of the S-O-R model by utilizing user experience to better comprehend perceived value in the travel app context. In particular, this research delves into the stimuli (app features) and mechanisms that contribute to users’ intention enhancement. To date, limited studies have identified the consistent drivers that influence travel app experiences (Sota et al., 2020). Prior research (e.g., Wang et al., 2014; Kennedy-Eden and Gretzel, 2012) has mainly highlighted “product” and “function” from the perspective of a company (e.g., navigation, information, and transaction). By addressing the limitations of previous studies, the findings of this study have revealed a collection of travel app experience-related factors that significantly impact perceived value. In addition, the S-O-R model was used to widen the range of travel app experience drivers derived from different theories (e.g., (extended) TAM, (extended) UTAUT, and innovation diffusion theory). This set of findings is expected to bring better and more comprehensive insights to the scholarly literature aiming to discover the external drivers that favorably affect app usage experience.

Second, our findings demonstrated a mediation effect supported by the S-O-R model, wherein perceived value influences user intentions (i.e., repurchase intention and recommendation intention) via user engagement. Previous research on travel apps has often used user perception or user engagement as a single organism. This study, on the other hand, characterized the organism process of travel app users by clarifying the influence of drivers on user intention via both cognitive (perceived value) and affective (user engagement) evaluation processes. Interestingly, it underscores that the modern user has become more informed, empowered, and sophisticated as a result of increased interconnectivity, making him or her more demanding (Claffey and Brady, 2015). Consequently, increasing users’ emotional bond and engagement with businesses is extremely crucial (Rahimi et al., 2017). Future tourism research should therefore place greater emphasis on the critical role of relationship management (i.e., user engagement) to maintain sustainable and profitable relationships with the emerging segment of Gen Y users.

Third, most prior research has overlooked the boundary condition of privacy concerns and has rarely employed the PRT to understand travel app usage. Our study extended the PRT to the travel app context, offering new theoretical options for research on travel app users’ decision-making inconsistency. As mentioned, the findings indicate that privacy concern is a key contingency factor that strengthens the positive association between user engagement and recommendation intention. Thus, future research must examine the influence of privacy concerns when studying the behavior of innovative tool users (such as mobile app users).

Practical implications

This study also delivers significant managerial implications. In essence, it is imperative for travel companies to design app features that are compatible with the needs and wants of users rather than focusing on the company's perspective. For instance, the interface design of travel apps and their compatibility with users should be prioritized when targeting Gen Y users. Specifically, the greatest coefficient value of user interface attractiveness indicates that most travel app users are visually stimulated. It is thus recommended that travel apps have a multi-faceted, visually pleasing interface to heighten user preferences. Generally, the interface's aesthetic elements should adhere to certain standards, i.e., basic yet attractive, clear yet concise, and familiar yet creative. Additionally, the interactivity of the interface may be improved by introducing different orientation modes to enhance the value of the app. For example, the interface of the ‘pre-planning’ mode should direct users to what they need, such as itinerary browsing, hotel reservations, and car rental. Meanwhile, the ‘on-trip’ mode should integrate a cartoonish travel map that enables flexible itinerary creation and specifies how to access itineraries.

Compatibility is the second most important feature to increase the value of travel app usage. Again, this demonstrates that it is important for travel app developers to better comprehend their users by knowing the features or contents that are prioritized by users. In practice, many existing travel apps attempt to deliver an exhaustive list of products and services without carefully considering user preferences from the demand perspective. To better leverage compatibility appreciated by users, travel apps should undertake two types of efforts. First, at the strategic level, they should create a specialist task force for user profile analysis with yearly updates or outsource annual user profile research to consultancies. Second, at the operational level, they should introduce new communication channels (with human personnel as the primary point-of-contact and automated analysis tools as supplements) for gathering user feedback. Automated analysis tools may even count the number of taps to generate a heat map of characteristics and summarize users’ keywords. This enables travel apps to obtain a comprehensive understanding of their users without violating the latter's privacy, hence improving compatibility.

In terms of relative advantages, apps must continue to be updated to attract early adopters and keep current users by offering value-added discounts and making trip planning more efficient. Specifically, travel apps must build unique strategic partnerships with vertical and horizontal businesses to offer lower prices on certain items or services. In addition, travel apps are encouraged to offer advantageous utilities that need a certain level of user engagement, so as to raise the bar for rivals’ imitation.

Furthermore, ease of use can increase the value of a travel app. This indicates that app developers should pilot test an app before launching it in the market to reduce the possibilities of complexity. Also, for a travel app to be more user-friendly, it necessitates an advanced search tool with an intuitive-to-use filtering and sorting system. For instance, with the assistance of a powerful search and a comprehensive sorting system, it is easier for the user to discover what they need, thereby significantly enhancing their ease of use.

Lastly, personalization shows a significant but relatively low path coefficient. This finding implies that although travel companies intend to offer personalization as an active caring service, Gen Y travel app users in China may dislike it due to excessive personal data requests. When attempting to provide tailored services as a value-added service, travel apps should exercise caution and conduct a more precise analysis of their users. In addition, travel apps can categorize their users into four types, i.e., new users, regular users, engaged users, and savvy users. Depending on the type, the travel app can prompt engaged and savvy users with a pop-up asking whether they desire more tailored services that require relatively more data collection.

Although providing a great user experience can result in positive value perception, tourism providers are advised to concurrently implement more productive relationship management strategies (user engagement) to achieve profitable goals (e.g., repurchase and recommendation intentions). To enhance emotional ties with users, tourism providers are recommended to encourage users’ active participation behaviors. This can be done by combining their social media plan with their overall business strategy. For instance, they can investigate the most popular social media platforms among their users and concentrate on a single key platform, such as Zhihu, the Chinese version of Quora. Consequently, social media efforts will not be overshadowed by complementary social media initiatives. Then, the travel app should feature a reward mechanism to encourage the sharing of user-generated content on the targeted social networking platform. Since Zhihu is a question-and-answer platform, relevant answers may be accurately delivered to the appropriate audience. With purposeful education, users may engage in the company's activities with more enthusiasm. Next, actively engaged and outstanding users (e.g., travel influencers with a medium to a large following) can be designated as brand ambassadors, thereby greatly boosting their sense of pride and duty towards the travel app. Last but not least, these dedicated and engaged users should be rewarded with flexible and individualized incentives. This set of procedures may be used to increase the user engagement of various user groups.

Privacy concerns are another important factor that may deter users’ engagement and behavioral intention. Increasingly, users are expected to supply personal information (e.g., phone number, location permission, and credit card information) to complete a transaction. Therefore, travel apps’ capacity to handle and alleviate users’ privacy concerns will determine their effectiveness in establishing relationship marketing (i.e., user engagement). For example, users’ freedom to delete their account and their associated data might help relieve privacy concerns. Also, travel apps may explicitly clarify how their online systems and offline services will use users’ information from the onset. A handy mechanism (e.g., a 24-h support hotline or an online privacy complaint system) for users to raise privacy concerns is encouraged. If properly executed, this additional service would likely reinforce user–app relationships and reduce potential privacy concerns. Nonetheless, as indicated by previous findings, privacy concerns have a differing effect on repurchase intention and recommendation intention respectively. Therefore, travel apps prioritizing positive recommendations should exercise extra caution in protecting users’ privacy and make every effort to safeguard information privacy. On the other hand, travel apps with repurchase priority may occasionally be able to take riskier steps with privacy-related big data technologies to better serve users who are more engaged and less concerned about their privacy.