Evolving Consumer Preferences: The Role of Attribute Shifts in Online Travel Agency Satisfaction and Loyalty

Abstract

The attributes that people consider when selecting an online travel agency (OTA) change over time, affecting how consumer satisfaction and loyalty evolve. However, attribute weight shifts in subsequent OTA visits cannot be determined using a cross-sectional approach. Thus, this study investigates the linkage dynamics between satisfaction and loyalty intentions as OTA attribute weights shift over time. We also assess the moderating effect of sales promotions on this linkage across subsequent OTA visits. Using a two-time-lag survey methodology (T1 and T2), we obtain 329 usable data. Our findings demonstrate that the link between satisfaction and loyalty intentions strengthens as customers gain more experience with an OTA. While price and post-service quality weights increase sharply, the weights of security and ease of use decrease. Furthermore, our findings show that the moderating effect of sales promotions does not appear in the early stages of a consumer’s experience with an OTA. Our research is the first to offer a complete understanding of the role of sales promotions based on the satisfaction–loyalty mechanism, considering a broad range of OTA selection attributes.

1. Introduction

Historically, service and tourism experts have been confident in understanding their customers. However, Shepherd [1] suggests that, following the pandemic, travel and hospitality brands must reimagine how they engage with customers and promote loyalty, underscoring the need for tourism and service researchers to develop a more nuanced understanding of customer behavior by examining how loyalty evolves.

Although researchers have emphasized satisfaction’s critical impact on loyalty and behavioral intentions [2,3,4], they have limited their focus to tourist consumption systems from a longitudinal perspective [5,6,7]. For example, a consumer’s experience with an online travel agency (OTA) changes over time, and their satisfaction is often based on an OTA’s specific attributes [8,9,10,11]. Thus, researchers have called for capturing the dynamic nature of service constructs [12,13].

To respond to this call, we present the following research questions: (1) How do OTA selection attributes change over time? (2) How do these changes affect the relationship between satisfaction and loyalty intentions in the tourism industry? (3) If OTAs implement sales promotions, do consumer responses to these promotions change across subsequent OTA revisits?

These questions highlight the need for research evaluating attribute changes, satisfaction–loyalty mechanism dynamics, and sales promotion responses. For example, Mittal et al.’s [14] examination of temporal changes in vehicle purchase attributes and satisfaction using the consumption-system approach only focused on attribute weight shifts without discussing how these shifts respond to the satisfaction–loyalty mechanism. Furthermore, although Guizzardi and Mariani [15] address the importance of customer behavior dynamics in the tourism industry, they do not account for attribute weight shifts. This could lead to temporal changes in the satisfaction–loyalty mechanism being misinterpreted, which can especially be detrimental in the service and tourism sectors.

Wu et al.’s [7] time-varying parameter demand system framework supports the argument that service research is currently limited. They examine the evolution of various parameters across subsequent service visits, focusing on changes in customers’ perceptions of price elasticity or dynamics. However, they fail to consider that many individual attributes may contribute to determining consumption experience satisfaction when consumers make repeat visits to an OTA [16]. Shifts in individual attribute weights based on prior consumption experiences should determine satisfaction in subsequent visits. This logic further underscores the need for studies investigating how changes in various attribute weights impact satisfaction over time.

Based on this logic, our main framework, which comprises OTA attributes, can be influenced by sales promotions. Particularly, when choosing OTAs, consumers prefer functional values (e.g., monetary and benefit values) [17]; however, the literature has placed relatively less emphasis on the role of sales promotions during consumption periods. For example, when consumers use an OTA for the first time, did they do so due to sales promotions offered at the point of first use? If not, would they continue to use an OTA if sales promotions were presented after experiencing the services offered?

In filling these gaps, this study makes two incisive contributions. First, it extends the scope of the current literature by explaining how attribute changes, which are determinants of satisfaction, vary across multiple OTA visits. In considering how the effects of these attributes shift, we demonstrate that attribute weight changes are a function of the consumption experience.

Second, we refine the satisfaction-based loyalty cycle developed by Oliver [18], explaining how the effects of the satisfaction–loyalty linkage evolve across OTA consumption experiences. Third, our study provides insights into when OTAs should implement sales promotions to achieve business goals. Thus, this study contributes to the academic discourse by discussing how the relationship between satisfaction and loyalty can manifest distinctly within the consumption system.

This study is organized as follows: After reviewing the theoretical foundations, we present our hypotheses. We then describe our methodology and data analysis techniques. Next, we report our hypothesis testing results and discuss theoretical and practical contributions. Finally, we provide our limitations to be addressed in future studies.

2. Theoretical Foundations

According to the general living systems theory [19], a consumption system involves a cycle of service provision, consumer evaluation, and repeated purchases across events. Mittal et al. [14] suggest the subsystems and evaluations of a specific service within the consumption-system approach, while Oliver’s [18] satisfaction cycle model, based on the expectancy disconfirmation theory, emphasizes the carryover effects and their impact on loyalty and the resulting long-term purchase episodes. Additionally, behavioral mechanisms identified in the marketing and tourism literature have highlighted how intentions evolve throughout this process [20,21]. We incorporate these theoretical frameworks into this study’s model, as shown in Figure 1.

Our proposed model includes three fundamental consumption system elements: selection attributes, customer satisfaction, and loyalty intentions. A commonality among these characteristics is their evolution across consumption episodes [12]. From a dynamic perspective, Huifeng and Ha [22] found that online review efficacy increased throughout their model’s initial period but decreased in subsequent periods. In consideration of their findings, our model focuses on changes in temporal and carryover effects, specifically from a natural evolutionary perspective. A carryover effect occurs when a particular construct (T1) affects the same construct (T2) at a later time. For instance, Johnsen et al. [20] investigated the carryover effects of value, brand equity, and loyalty intentions concurrent with mobile contract renewals in Germany; however, comparable research on specific constructs in the travel industry is rare.

Meanwhile, a dominant premise of the service and tourism literature is that positive experiences enhance visitor satisfaction, resulting in customer loyalty [23,24,25]. Oliver [26] established a customer loyalty stage model by adding a cognitive–affective–conative loyalty framework. We consider this stage model theoretically acceptable; thus, we have extended it by examining which changes in attributes strengthen or weaken customer behavior or loyalty dynamics. Furthermore, our conceptual model encompasses a sequence of events linking OTA service experience, satisfaction, and loyalty intentions. As examples of events, sales promotions are essential for OTAs as consumers tend to be price-sensitive [27], suggesting that a deeper understanding of the satisfaction–loyalty mechanism should comprise consumers’ overall service experiences with particular OTAs.

Furthermore, consumer behavior logic suggests a chain of cause-and-effect relationships between OTA service satisfaction and loyalty intentions. For example, the logic that OTA satisfaction at T1 → OTA loyalty at T1 → OTA service experience decisions at T2 → OTA service satisfaction at T2 → OTA loyalty at T2 may align with our proposed model. In particular, sales promotions reflect consumer preferences and can lead to changes in the proposed mechanism.

2.1. The Linkage Between OTA Selection Attributes and Satisfaction over Time

OTA selection attributes are determinants that consumers carefully consider when choosing an agency for booking accommodation, air tickets, or other travel products. Although many studies on the hospitality, tourism, and service sectors consider similar attributes, we specifically analyze research published after the early 2000s. OTA attribute importance evolves as consumption experiences alter users’ selection criteria. Although more than 30 attributes have been identified [10,11,28,29], we concentrate on the five used most frequently: security, information, finding low fares (price), ease of use, and post-service quality. Notably, researchers have underscored the importance of post-service quality in relation to OTA revisits as it determines how service failures are handled [30,31,32].

As OTA attribute evaluations depend on consumption experiences, researchers and managers are interested in the relationship between attributes and satisfaction [2,8,11]. Despite existing studies identifying the key attributes linked to satisfaction based on an assessment using importance-performance analysis (IPA), hospitality and tourism researchers have yet to clearly define customer satisfaction using individual-level attributes. However, defining satisfaction as a function of OTA attribute-level evaluations resulting from performance-based customer experiences may be possible and valuable [14,18].

Such a definition offers two distinct advantages. First, it can account for consumers’ memory-based perceptions of consumption experiences. For example, Mittal et al. [14] found that consumers’ perceptions of products’ individual-level attributes across consumption periods affect satisfaction evaluations. Second, this definition can facilitate an examination of changes in individual-level attributes over consumption episodes while differentiating level-specific responses to attribute changes [33]. That is, the proposed definition can help elucidate the effects of changes in individual-level attributes before and after consumption experiences. In doing so, it enables us to demonstrate how to optimize the allocation of OTA marketing resources.

Most studies have investigated attribute evaluations using a cross-sectional orientation, highlighting that our approach to attribute weights is timely and significant. Specifically, evaluating shifts in individual-level weights may produce a more accurate diagnosis than analyses that view all attributes as a single construct, mainly because changes in these weights can drive attribute salience [34]. For example, in the hospitality industry, food and drink prices are critical for satisfaction judgments at T1; however, this attribute’s weight drops significantly over subsequent visits. Conversely, Zhang and Ha [35] show that cultural and historical attractions are not significant at T1; however, the weight of this attribute substantially increases when revisiting a destination. These results suggest that attribute weight shifts are an essential consideration in OTA research.

Customer satisfaction is determined by many attributes in the context of OTA services. Users tend to choose an OTA by combining several attributes or considering specific attribute weights. In the latter case, investigating which attributes decisively influence OTA satisfaction is more critical than considering all attributes. Although most studies have proposed future research directions from a cross-sectional perspective, our approach highlights the need for a robust theoretical mechanism that accounts for changes in individual-level attribute weights over time.

In recognizing that customers respond differently to individual-level attributes, we aim to demonstrate that a complete understanding of attribute weight changes is invaluable in predicting satisfaction and behavioral directions, as these changes can determine service satisfaction judgments from T1 to T2, driving consumption experience evaluations [36]. Consequently, improving the consumption experience may increase attribute weights when choosing an OTA. As satisfaction is based on attribute-level evaluations [14], increasing individual attribute weights can be considered a decision-making function that may lead to OTA revisits. Thus, our first hypothesis is:

H1. 

OTA attribute weights increase over time as a customer’s consumption experience improves.

2.2. The Satisfaction-Loyalty Intentions Linkage over Time

Research shows that satisfaction’s long-term effects are linked to repeated consumption experiences, as it accumulates following initial positive experiences with specific services [18]. Thus, increased satisfaction across consumption can foster brand or service loyalty [14,18,20]. However, studies addressing customer loyalty development over time are sparse within the tourism literature, despite one-off transactions not tending to cultivate loyalty. For example, although a carefully designed one-time consumption event when booking a summer vacation at a particular OTA can be a starting point for developing loyalty, we seek to demonstrate that subsequent consumption episodes can increase it. Our argument aligns with the satisfaction cycle [18] in highlighting the importance of the consumption-system approach, beginning with an initial encounter and moving through continued consumption experiences.

This study defines satisfaction as a consumer’s fulfillment response to an OTA’s individual-level attributes. Based on consumers’ experiences with OTA services, modified satisfaction judgments are associated with behavioral intentions and loyalty development—a proposition supported by adaptation-level theory [37,38]. Such a satisfaction judgment process emphasizes the importance of reactions based on the expectancy disconfirmation process and suggests that existing judgments and intentions are the basis for forming future ones. While these two variables do not have an absolute effect at T2 [14], they may create a relatively significant connection from T1. Thus, as consumption experiences evolve, adaption-level adjustments may cause satisfaction judgments to differ [18], influencing future loyalty intentions.

As the relationship between satisfaction judgments and loyalty intentions varies systematically as consumption experiences evolve [18], we contend that this concept should be understood as unfolding naturally. Therefore, we define loyalty intentions based on customers’ dispositions toward revisiting a particular OTA, aligning with the notion that changes in behaviors and conditions foster subsequent consumption episodes [39] or OTA revisits [40]. For example, positive (negative) customer satisfaction can strengthen (weaken) loyalty intentions. In other words, the former can strengthen the satisfaction–loyalty relationship over time, while the latter can weaken it in subsequent consumption episodes [41].

A robust corpus of tourism literature has demonstrated that customer satisfaction is fundamental in shaping or reinforcing loyalty intentions [42]; however, we specifically focus on two points in the satisfaction–loyalty relationship: before and after consumption. That is, once satisfaction is achieved, we ask: How do loyalty intentions evolve with satisfaction? In other words, we argue that the satisfaction–loyalty intention linkage cycle changes across subsequent consumption episodes.

While we predict that the linkage between satisfaction and loyalty intentions can strengthen over time, empirical studies find that, specifically in the vehicle industry, this relationship weakens incrementally [14]. By logical extension, changes in individual-level attribute weights that undermine customer satisfaction could dilute vehicle consumers’ loyalty. Similarly, Kumar et al. [43] emphasized that this relationship can potentially change over consumption periods. However, in the tourism industry, where consumers face many alternatives, we expect visitors’ prior OTA satisfaction to support repeated visits or subsequent service purchases.

From the perspective of the satisfaction cycle theory, consumers’ updated loyalty intentions across consumption episodes influence their resulting satisfaction [18]. Thus, the satisfaction–loyalty linkage strengthens over time. For example, Expedia stresses its member promotions before purchases by announcing, “Members save up to 30% when you add a hotel to a flight”, centered on the website, attempting to enhance customer satisfaction and loyalty. Based on adaptation level adjustments like OTA selection attributes, the relationship between satisfaction and loyalty intentions may be stronger in T2 than in T1. Thus, we conjecture that the satisfaction–loyalty relationship strengthens until loyalty peaks due to many alternatives. Therefore, we propose the following hypotheses:

H2. 

Satisfaction (T1) directly and positively affects loyalty intentions (T1).

H3. 

The temporal effect of the satisfaction-loyalty intentions linkage at T2 increases as changes in individual-level attribute weights improve.

2.3. The Moderating Role of Sales Promotions

As an example of a marketing mix, sales promotions (monetary vs. nonmonetary) are directly related to consumer choice behavior [44,45]. For example, researchers have demonstrated that monetary promotions are effective when consumers purchase a specific product for the first time [46], as in the context of this study. However, consumer behavior can also depend on the presence or absence of promotions over time [47,48]. Consumers using OTAs for the first time may experience uncertainty or perceive risks [49]. At this time, if an OTA offers additional sales promotions (e.g., included breakfast or room upgrades) in addition to general price discounts, consumers may be suspicious of the product or avoid the OTA. However, actually experiencing an OTA’s service determines the likelihood of revisits, where promotions may be preferred. Thus, presenting a sales promotion may be more effective at a later stage than an early one. Our logic leads to the following hypothesis:

H4. 

The presentation of a sales promotion moderates the satisfaction–loyalty intentions linkage more effectively at later stages of the consumption experience.

3. Methodology

3.1. Data Collection

We selected OTAs in Korea, as this sector has dramatically recovered since the COVID-19 pandemic. For example, in 2023, Korea’s travel and tourism penetration rate had grown by 207.8% compared to the beginning of COVID-19 in 2020 [50]. As such, Korea’s tourism industry is representative of the growth of global tourism [51,52].

We tested our hypotheses using surveys administered by a Korean online research firm with two time-lag interval samples of participants who had booked summer vacations. We used the following criteria to select our sample to achieve our study’s aims. The respondents (1) must have booked at least one service with an OTA for their summer vacation before participating in the survey, and (2) had not previously used the OTA’s service. To better understand the evolution of the satisfaction–loyalty intention linkage due to attribute level changes, tracking attribute weight shifts for new customers may be more beneficial than for existing ones with more experience in individual-level attribute evaluations, as their evaluations may strengthen, weaken, or remain unchanged. Additionally, despite OTA’s strategic use of sales promotions [53], customer judgments may vary depending on the time of purchase or during revisits [18].

Our initial survey (T1) was conducted in June 2023, and the second (T2) followed six months later. Both surveys were administered to the same respondents. To carry out the study’s purpose, participants at T2 were limited to those using the same OTA as at T1. The sample was also limited to tourists who had opted for international travel, as Korean demand for overseas travel increased dramatically at the end of COVID-19.

We conducted a panel survey that repeatedly questioned the same participants to ensure they were consistent across the two time-lag intervals. However, we excluded those who did not travel during the second survey period (T2). Survey participation at T1 was guided through mobile and email messages to 590 respondents who met the study’s sample criteria. Additional notification messages were delivered to increase participation rates. A total of 507 people participated in the survey at T1, but 26 were excluded due to careless responses, leaving 481 usable responses. Our survey methodology was similar to prior studies [14,22].

Approximately six months later, we conducted a second survey targeting the respondents (n = 481) from T1, employing the same method as the initial survey. A total of 375 people participated, but 46 were excluded due to cancellations, reservation changes, or careless responses. Ultimately, we collected 329 usable responses (response rate = 55.9%). We determined our sample size using an alpha level of 0.05 [54] and a t-value of 1.96. As the minimum sample size needed to be 120 or above, our sample size was deemed appropriate.

Our data were collected using a consumption-system approach, emphasizing consumer behavior evolution via two time-lag intervals. This method allows for conservative and robust testing of the model’s temporal dynamics [14]. The respondents who participated in both surveys received a voucher for Starbucks. Of the 329 participants, 43% were male, and the sample’s mean age was 34.6. In particular, 80.4% of the sample were between 25 and 44 years old, similar to the distribution (81.1%) of Korean OTA users [53]. Among them, 62.6% were office workers, and the sample’s average annual salary was approximately 48,200 USD.

In the following sections, we introduce measurements that test key constructs and examine their mean differences to determine whether these constructs have changed over time. Next, we check for common method bias, an error that can occur during measurement. We then use control variables to check for exogenous risk. Finally, we introduce an analytic approach to test our hypotheses.

3.2. Measures

Table 1. Measures and factor loading by period.

	Outer Loading		AVE
	T1	T2	T1	T2
Satisfaction (CR: T1 = 0.87; T2 = 0.85)			0.68	0.65
The vacation booked through the OTA exceeded my expectations.	0.78	0.87
My decision to choose this OTA was wise.	0.84	0.74
Overall, I am satisfied with my experience with this OTA.	0.86	0.81
Loyalty intentions (CR: T1 = 0.90; T2 = 0.91)			0.75	0.76
If I get a chance, I will use this OTA again.	0.87	0.88
I will revisit this OTA.	0.89	0.87
I would recommend this OTA to others.	0.83	0.85

Note: CR = Composite reliability.

details the survey measures used to test the proposed model. We measured satisfaction using three items adapted from El-Adly [54] and loyalty intentions using three items adapted from Johnson et al. [20] and El-Adly [55]. Five individual OTA selection attributes were used: finding low fares (price), security, ease of use, information content, and post-service quality. We defined these attributes as follows. Finding low fares (price) refers to price rate appropriateness [56,57]. Security is defined as an OTA’s security capabilities in ensuring that customers can process transactions and that information is private [58]. Ease of use is customers’ ability to conveniently use a specific OTA [59]. Information content refers to overall OTA information quality, including individual product descriptions [60]. Post-service quality represents an OTA’s ability to adequately address service complaints through recovery offerings [30].

As the measurement items for satisfaction and loyalty were initially developed in English, we used a two-stage process proposed by Almanasreh et al. [61] to obtain content validity. First, a marketing professor and three doctoral students designed and reviewed all items at the questionnaire development stage. After the initial version was developed, it was judged by two marketing and tourism experts fluent in English and Korean. Following the judgment stage, we concluded there were no issues with content validity.

All attributes were measured as one single item. Overall, researchers tend to prefer using multi-item measures; however, Diamantopoulos et al. [62] demonstrated that certain conditions perform equivalently to multi-item scales, such as examining individual-level attributes. Similarly, a recent study argued that single-item measures are valid and reliable in explaining psychological phenomena [63]. Therefore, we identified individual attributes that previous studies considered essential. Similarly, sales promotions were also measured using a single item: “When I choose an OTA, the price discounts it offers are important”. Respondents rated all measures using a 5-point Likert scale, ranging from 1 = ‘strongly disagree’ to 5 = ‘strongly agree’, or 1 = ‘unimportant’ to 5 = ‘very important’.

3.3. Measurement Comparison

We compared changes in OTA satisfaction, loyalty intentions, and the five OTA selection attributes across periods. As presented in

Table 2. Means and reliabilities for scales by period.

Variable	T1: Mean (SD)	T2: Mean (SD)	Alpha T1 (T2)	Mean Difference
Satisfaction	4.14 (0.65)	4.24 (0.57)	0.78 (0.79)	−0.10 *
Loyalty intentions	4.05 (0.83)	4.25 (0.72)	0.82 (0.84)	0.19 **
Online travel agency selection attributes
Security	4.24 (0.67)	4.04 (0.66)		−0.20 **
Information	4.12 (0.53)	4.09 (0.50)		−0.03 (ns)
Finding low fares	4.16 (0.61)	4.38 (0.53)		0.16 **
Ease of use	3.79 (0.74)	3.52 (0.62)		−0.27 **
Post-service quality (price)	4.09 (0.69)	4.43 (0.61)		0.34 **
Moderator
Sales promotions	3.88 (0.84)	4.33 (0.62)		0.45 **

Note: *, p < 0.50, **, p < 0.01, (ns) = not significant.

, we conducted a pairwise t-test to determine whether the differences in the means of the constructs and attributes were significant. The results indicated that these differences had increased significantly, suggesting respondents’ consumption experiences had evolved. Meanwhile, the Cronbach’s alphas of the two constructs remained stable across periods.

3.4. Common Method Bias

Satisfaction and loyalty intentions are highly correlated in the tourism sector, highlighting the potential for common method bias. To test this, we performed Harman’s single-factor procedure [64] to compare the one-factor measurement model to the two-factor oblique model. The results showed that the latter was more efficacious with acceptable fit indices like CFI, TLI, and RMSEA. Additionally, the chi-square difference (Δχ²(2) = 19.482. p < 0.01) was significant, suggesting these two constructs are inter-related yet distinct [65].

3.5. Control Variables

Our study included three control variables: gender, travel frequency, and travel level. These variables may significantly relate to participants’ perceptions of attributes, satisfaction, and loyalty. To this end, respondents indicated their gender as (1) male or (2) female. Furthermore, respondents’ travel frequency (e.g., frequency of domestic and international travel, excluding business trips per year) was measured using the following three indicators: (1) rarely, (2) sometimes, and (3) often. Their travel level (basic, average, or luxurious) was controlled to reduce the risk of endogeneity. The proposed model included these control variables to reduce variability in the model’s dependent measures [66]. Consequently, our sample became more homogenous, resulting in greater estimation precision.

3.6. Analytic Approach and Measurement Validity

We used Smart-PLS (hereafter: PLS) to test our proposed model. This program is especially useful for analyzing larger complex models that predict consumer behavior [67]. Furthermore, PLS modeling is more robust for specific structural paths than other covariance-based structural equation modeling programs [68]. It is also ideal for analyzing single-item variables [69].

Before testing the proposed model, we assessed its reliability and validity. Table 1 shows that all factor loadings ranged from 0.74 to 0.89, above the baseline of 0.7. The composite reliability (CR) and Cronbach’s alphas exceeded the 0.7 cut-off value. Furthermore, the average variance extracted (AVE) exceeded the baseline value of 0.5 [70]. These results verified that this study’s convergent validity was acceptable. Next, we verified discriminant validity using two approaches. First, we used Fornell and Larcker’s [71] criterion to determine whether the AVE root-squared values were greater than the construct correlations. As demonstrated in

Table 3. Discriminant validity.

Fornell and Larcker’s Criterion
	1	2	3	4
OTA satisfaction (T1)	0.82
OTA loyalty intentions (T1)	0.23	0.86
OTA satisfaction (T2)	0.01	0.02	0.81
OTA loyalty intentions (T2)	0.02	0.11	0.62	0.87
Heterotrait Monotrait Ratio (HTMT)
OTA satisfaction (T1)
OTA loyalty intentions (T1)	0.27
OTA satisfaction (T2)	0.05	0.06
OTA loyalty intentions (T2)	0.04	0.13	0.68

Note: Italicized numbers are the square root of AVE.

, all the AVE root-squared values were greater than the suggested correlation values, indicating that discriminant validity was satisfied.

Second, we used the Heterotrait Monotrait Ratio (HTMT) method due to criticisms of Fornell and Larcker’s [70] criterion. From a conservative perspective, HTMT values greater than 0.85 may cause discriminant validity issues [72]. The results in Table 3 clearly show that all the HTMT values were lower than the threshold (< 0.85), confirming that discriminant validity was satisfied.

4. Results

Model Estimates

We checked the R² values to assess the four constructs’ overall explanatory power. As detailed in Figure 2, all R² values were satisfied (OTA satisfaction at T1: R² = 0.20; OTA loyalty intentions at T1: R² = 0.33; OTA satisfaction at T2: R² = 0.46; OTA loyalty intentions at T2: R² = 0.37). Establishing adequate explanatory significance requires an additional step in the PLS approach to obtain the proposed model’s predictive relevance using a Q² value. Exogenous variables in structural models are vital in determining their predictive relevance. In this case, the Q² value provides a solution if this value is greater than zero [71]. The Q² values of OTA satisfaction (T2) and OTA loyalty intentions (T2) were 0.51 and 0.42, respectively, suggesting our model has strong predictive relevance.

Furthermore, we applied a cross-validated predictive ability test (CVPAT) to assess the model’s predictive accuracy. As shown in

Table 4. CVPAT results.

	PLS Loss	Indicator Average (IA)	Average Loss Difference	p-Value
Satisfaction (T2)	1.245	1.269	−0.024	0.024
Loyalty intentions (T1)	1.894	1.906	−0.012	0.049
Loyalty intentions (T2)	1.821	1.830	−0.009	0.093
Overall model	1.653	1.668	−0.015	0.005

, CVPAT was conducted on the overall model level and for each of the three constructs [73]. A negative average loss value difference suggests the indicator average is smaller and thus preferable [74]. The overall model level demonstrates its strong predictive validity.

Regarding H1, these results aligned with the shifting weights of individual attributes. Among them, two attributes (finding low fares and post-service quality) increased, security and ease of use decreased, and the information attribute remained unchanged (see

Table 5. Attribute weight shifts over time.

Attribute	Weight at T1	Weight at T2	Difference (Δ)	Significant?
Security	0.16 **	0.02	−0.14	Yes
Information	0.09 *	0.08	−0.01	No
Finding low fares (price)	0.02	0.15 **	0.13	Yes
Ease of use	0.18 **	0.02	−0.16	Yes
Post-service quality	0.14 **	0.35 **	0.21	Yes

Note: *, p < 0.50, **, p < 0.01.

). Specifically, the weights of post-service quality (Δ = 0.21) and finding low fares (Δ = 0.13) increased noticeably, indicating these attributes are more critical than others after the initial OTA consumption experience. Interestingly, the importance of security and ease of use was diluted compared to the initial OTA visit. In the case of information, it was significant during the initial visit, but its impact somewhat weakened across subsequent visits. However, no significant changes were observed over time. Thus, H1 was partially supported.

H2 tested the extent to which OTA attribute evaluations (T1) determined attribute satisfaction, positively affecting loyalty intentions (T1). The results showed that the proposed relationship was positively significant (β = 0.29, p < 0.01). Thus, H2 was supported. H3 focused on how the linkage between satisfaction and loyalty intentions changed over time. Figure 2 demonstrates that this linkage grew significantly stronger with accumulated consumption experiences (T1: β = 0.29, p < 0.01 vs. T2: β = 0.42, p < 0.01), supporting H3. Notably, this result contradicts Mittal et al.’s finding, which suggested that the connection between satisfaction and loyalty intentions [14], as highlighted in the consumption-system approach, weakens over time. Our results show that this mechanism becomes stronger, at least in Korean OTA services.

H4 tested the moderating effect of sales promotions on the linkage between satisfaction and loyalty intentions across OTA visits. In the early stage, the moderating effect of sales promotions was not significant. However, in the later stage, the same effect was negatively significant. For customers using OTAs, sales promotions may be perceived suspiciously on initial visits. However, after actually experiencing an OTA, customers tended to more actively accept sales promotions. As shown in Figure 3, the more positive sales promotions were perceived, the more satisfaction and loyalty improved. Thus, H4 was supported.

5. Discussion

5.1. Summary and a Brief Literature Discussion

Drawing on the consumption-system approach and the satisfaction cycle mechanism, we investigated the linkage dynamics between satisfaction and loyalty intentions as OTA attribute weights shift over time. We also assessed the moderating effects of sales promotions on the linkage of satisfaction–loyalty intentions across OTA visits. Using a two-time-lag design, we generated the following three key results.

First, our first research questioned whether OTA selection attributes change over time. The answer is ‘yes’. Individual attribute weights increased, decreased, or remained unchanged over time. Specifically, the importance of post-service quality increased sharply, indicating this specific attribute has a greater impact than others. In contrast, the weights of security and ease of use became diluted over time. Researchers have demonstrated that attribute utility functions are concave to some degree [75]; however, our findings show that these functions may vary for individual attributes; thus, they are asymmetrical and may not produce a standardized curve, such as a concave or convex one.

Second, we also asked whether these attribute changes affect the relationship between satisfaction and loyalty in the tourism industry. Our answer is significant. Specifically, the satisfaction–loyalty intentions linkage strengthened over time, suggesting it is a function of attribute weight changes. Bacon [76] found that improving a poorly performing attribute notably boosts a customer’s overall satisfaction. However, even without an attribute improvement process, the increase in attribute weights within the consumption system strengthens overall satisfaction, improving loyalty intentions over time. These findings suggest that the temporal effect of the link between the two constructs depends on increases in overall attribute weights. Thus, we conclude that larger attribute weight increments can help improve the relationship between satisfaction and loyalty intentions.

Third, our final question was whether consumers’ responses to these promotions would change over multiple OTA visits. Our answer is yes. Sales promotions have little impact on customers visiting OTAs for the first time; however, this impact strengthens across subsequent visit stages. In other words, customers may question the effects of sales promotions during their initial OTA visits, but they actively accept these promotions when revisiting. Our findings demonstrate that the tolerance zone of sales promotions varies depending on the customer OTA experience stage.

Further, with respect to the consumption-system mechanism [14], we advance the understanding of attribute weights in the satisfaction–loyalty linkage [6,23,34]. Researchers have highlighted the importance of sales promotions, a characteristic of online services [45,77]. Thus, this study extends the online commerce and OTA literature by demonstrating that the effects of sales promotions are reinforced by the satisfaction–loyalty mechanism in OTA revisit periods. In particular, the moderating role of sales promotions aligns with recent findings suggesting that firms can use sales promotions to improve satisfaction and target specific loyalty groups [78].

5.2. Theoretical Implications

This study contributes to the tourism literature on the satisfaction–loyalty mechanism in three ways. First, our analysis reveals how individual OTA selection attributes impact the evolution of the satisfaction–loyalty relationship during initial and repeat visits. While previous research has indicated that behavioral intentions change over time [6,14,20], our outcomes demonstrate that the satisfaction–loyalty mechanism is influenced by consumers’ experiences during subsequent OTA visits. This result aligns with previous findings, highlighting the importance of consumers’ service experiences over time [79].

More importantly, while Mittal et al. [14] find that the satisfaction–loyalty relationship gradually weakens, our analysis indicates the opposite. Specifically, we show that this linkage strengthened when overall individual attribute weights increased. These outcomes do not simply mean all selected individual attribute weights increase. Rather, they indicate that the increase in individual attribute weights is relatively greater than before. These changes can influence consumers’ OTA selections, thereby shifting their satisfaction judgments and repeating behaviors with time. Thus, our findings suggest that the conditions that shift individual attributes are dynamic.

Second, regarding loyalty evolution, the general living systems theory and the consumption-system approach enable us to introduce a new perspective on OTA attribute dynamics. Marketing literature argues that theories concerning attributes, satisfaction, and loyalty intentions should focus on constructs’ dynamic nature [12,14,20,80]. Our findings provide ample evidence that the relationship between the overall performance of OTA selection attributes and satisfaction with these attributes is dynamic, impacting loyalty intentions over time. This outcome extends the scholarly understanding of existing theoretical mechanisms from the marketing domain to the tourism domain by responding to the demands introduced by these theories.

Finally, not all consumers are persuaded by sales promotions such as price discounts. As price discounts offered by OTAs can be easily compared through search engines, promotions presenting vastly different prices can increase perceived risks or be viewed suspiciously by customers when initially visiting an OTA. In other words, these promotions may have a worse effect than not offering a discount [47]. However, once a customer uses a specific OTA, the impact of sales promotions changes dramatically, suggesting that the tolerance zone for sales promotions differs over time. These results may fill a research gap revealed in previous studies. Specifically, firms can increase efficiency by targeting loyal customer groups, rather than expending marketing resources on indiscriminate sales promotions for OTA services.

5.3. Managerial Implications

The prevailing view is that changes in the importance of individual OTA selection attributes depend on the consumption experience [3,81], suggesting that managers should continuously monitor and invest in OTA selection attributes. This view requires tracking how individual attribute weights increase, decrease, or remain stable. As such, changes in attribute weights stemming from consumption experiences must be managed differently for existing and new customers. For example, while the weights of post-service quality and finding low fares (price) increased sharply, those of security and ease of use decreased across consumption episodes. Specifically, post-service quality is essential when deciding to revisit a particular OTA, while finding low fares (price) is central to choosing an OTA for the first time. While prior research has viewed security as key when consumers select an OTA [56], our findings demonstrate that this attribute is unimportant to customers throughout the consumption experience. Therefore, managers should inform OTA users of continuous post-service improvements (e.g., announcing active recovery efforts after a service failure on their website) and price offerings and implement strategies to increase adoption levels.

OTA managers must also consider improving their understanding of individual attributes with unchanged weights. For example, we found that the consumption experience did not alter the weight of information. This is likely because information is critical for choosing an OTA, at least in the Korean context, demonstrating its centrality to customer evaluation criteria regardless of the passage of time. Thus, managing valuable information for new and repeat customers is vital [82]. Indeed, it is fundamental to maintaining customer focus from the OTA search stage to the repeat visit consideration stage. Therefore, we recommend that managers provide salient information that captures customers’ attention and enhances real-time customer service, as these efforts may prove to be the most effective for all customer types.

Finally, sales promotions are a strategic tool for OTAs, but managers must enhance their understanding of how they are applied. Our findings demonstrate that promotions are more effective for repeat customers than first-time ones. Therefore, managers should frame prices differently when repeat visitors search for products or services compared to first-time visitors. In particular, designing promotions targeting repeat customers to foster loyalty can help improve satisfaction and encourage revisits. For example, presenting a price similar to that of a competitor on the first page of the OTA’s site and then clearly stating, “Please log in to check for more detailed prices”, may correspond to repeat customers’ expectations. Researchers can implement these price promotions to foster and retain loyal customers.

5.4. Limitations and Future Directions

Despite providing considerable insights on the impacts of OTA selection attributes regarding the dynamics of the satisfaction–loyalty intentions mechanism, this study has some limitations that future research should address. First, while our findings can help OTA operators improve satisfaction and loyalty when implementing specific processes and personalized services, future studies should combine big data and AI to assist management services.

Second, two customer types use OTA services: business and leisure travelers. This study only focused on individuals taking vacations (leisure travelers); however, traveler behavior may depend on the purpose and nature of the trip. Furthermore, OTAs can be categorized as either hotel sellers (or airline ticket sellers) or comprehensive travel product sellers (e.g., group tours) [83]. Therefore, as OTA selection attributes may depend on customer type and travel purpose, future research is needed to compare how these factors affect attribute weight changes.

Finally, this study only focused on the Korean tourism context, indicating geographical limitations. To generalize our findings, future studies must conduct cross-cultural analysis. For example, by comparing the OTA consumption-system approach in China and the United States over time, researchers may discover similarities and differences in our study’s findings.