The Effect of Service Quality on Customer Retention: Insights from Travel Agency Managers in the KRI

Customer retention, as a behavioral phenomenon, has undergone minimal evolution over the past three decades, underscoring the need for illuminating frameworks (Dias et al., 2026). Customer retention is the relationship between the company and the customer, reflecting the customer’s desire to participate in business and continually purchase from the company rather than its competitors (Israel, 2022). In addition, it is a fundamental concept in today’s highly competitive markets (Banabo & Perearau, 2025). It has become a critical strategic priority, making strong customer relationships essential (Datta, 2018). Retention helps companies manage growing competition and should be considered from both the company’s and consumers’ viewpoints (Trenggana et al., 2022). Satisfied, retained customers are a company’s most valuable asset for sustaining reputation and market edge (Al-Hashem, 2022). On the other hand, the advantages of customer retention have been highlighted, including reduced sensitivity among loyal customers to service/product pricing and competitors’ marketing strategies (Loureiro et al., 2025). Resultantly, loyal customers tend to accumulate more goods and services over time and are associated with lower marketing expenses, as they indirectly attract new customers through favorable word of mouth. Consequently, elevated customer retention rates lead to enhanced firm profits (Carvalho dos Reis Silveira et al., 2022).

Customer retention hinges on customers’ attitudes and actions toward service quality (Kabasinguzi et al., 2024). It also reflects how effectively a company connects with its customers (Almohaimmeed, 2019). Adzhigalieva et al. (2022) define retention as the customer’s intention to sustain a business relationship. Maladi et al. (2019) argue that a successful retention strategy involves comprehensive marketing actions to retain and maximize long-term value for both new and existing customers, strengthening mutual relationships (Lovemore et al., 2023).

However, Trenggana et al. (2022) note there are two main strategies: retaining existing customers and acquiring new ones. Acquiring new customers can be difficult and expensive, making retention essential for a brand’s growth and sustainability (Al-Adwan et al., 2025). Studies consistently show retaining existing customers costs significantly less than recruiting new ones (Hanaysha, 2018; Sağlam & El Montaser, 2021). However, Sikri et al. (2024) observed that offering comprehensive facilities to all customers is neither practical nor cost-effective, as such an approach poses significant challenges. Companies may instead forecast potential churners and focus on them with targeted retention strategies (Sikri et al., 2024). Hence, Hanaysha (2018) demonstrated that retention has financial consequences, as acquiring new customers typically requires significant investments to influence buying behavior, reinforcing the value of retaining existing customers. As a result, customer retention is a vital goal for service enterprises within relationship marketing. Their growth and survival in a competitive environment are critically contingent upon retention (Zhengmeng et al., 2024). As an increasing number of customers become the focal point of marketing efforts, interest in their individual preferences intensifies (Rîpa & Nicolescu, 2023).

As per Harriet et al. (2024), service quality is a consumer’s comprehensive evaluation of an organization and its services, reflecting the level of excellence or lack of value. In other words, the value of a service from the customer’s viewpoint defines service quality and quality in general. However, the study by Ifedi et al. (2024) mentioned service quality as the company’s comprehensive service system. It describes how the company’s delivery system can meet customers’ expectations. This supports the idea of overall service quality, looking beyond just direct results.

On the other hand, service quality management ensures that services provided to customers meet or exceed their expectations. It also evaluates the efficacy of a service, identifying issues and addressing them to enhance future quality and boost customer satisfaction (Jing et al., 2020). Therefore, the manager’s primary objective is to enhance the organization’s fundamental skills to deliver a higher standard of service. They achieve this by differentiating their products and services from competitors and effectively managing efficiency and responsiveness (Al Kurdi et al., 2023). Unquestionably, quality services involve delivering a service that aligns with customer expectations and the organization’s values. The company’s high-quality services and products enhance customer satisfaction (Harriet et al., 2024).

Furthermore, service quality is essential for achieving a competitive advantage in the service sector (Chang et al., 2024). Nonetheless, the determinants of service quality in offline services differ from those in online services. Alongside outcome-oriented factors, the physical environment and personal interactions, which rely on counsellors’ expertise, conduct, and problem-solving abilities, significantly influence service quality (Baumgartner et al., 2024). Prior to the rise of the internet in the early 1990s, travelers relied heavily on travel agencies to make travel arrangements. This was because travel companies were the only intermediaries from which they could obtain reliable, up-to-date travel information (Wei, 2022).

Relationship marketing has garnered growing attention in marketing literature. It is defined as the process of enhancing quality to foster customer satisfaction and retention by cultivating long-term relationships with customers and stakeholders, thus promoting future operational success (Mokhtaruddin et al., 2018). The RM theory supports customer retention by providing components that explain how and why customers develop long-term relationships with companies. Service quality positively affects both relationship quality and customer loyalty (Dung & Tu, 2023; Tun, 2023). An improved relationship with customers can ultimately lead to greater loyalty and retention (Rîpa & Nicolescu, 2023).

Furthermore, Morgan and Hunt (1994) identify trust, commitment, and communication as essential to RMT. These elements align with service quality dimensions such as reliability, responsiveness, and empathy (Parasuraman et al., 1985). A lack of trust is a significant barrier to customer acceptance, especially when products are less tangible (Shi et al., 2025). High trust leads to greater customer satisfaction and service success, making it crucial for business success in service companies (Wahyudi et al., 2025). Similarly, commitment enhances emotional and psychological attachment, which reduces the likelihood of switching (Ekejiuba et al., 2025). Continuous communication also sustains relationships and addresses customer needs effectively (Phyu, 2019). In this study, RM theory underpins the idea that the strength of the provider-customer relationship shapes how service quality is perceived and maintained over time.

Several scholars (Adzhigalieva et al., 2022; Adzinyo et al., 2024; Al-Gasawneh & Dalain, 2023; Harriet et al., 2024; Kabasinguzi et al., 2024; Lovemore et al., 2023; Maladi et al., 2019; Ogonu & Elvis, 2023; Pezeshki, 2009) have examined the link between service quality and customer retention. Their studies consistently show a direct link, identifying service quality as key to customer retention (Sari et al., 2018). Furthermore, customer retention is a crucial factor in the success of services provided by various companies (Nallaluthan & Al Falasi, 2023). Service quality directly influences customer satisfaction, which in turn affects consumer loyalty intentions and repeat purchases (Etim et al., 2021). Nallaluthan and Al Falasi (2023) examined how service quality affects retention in tourism companies in the United Arab Emirates. They found a significant relationship between the two.

Tangibility was positively correlated with customer retention in a study by Tsokwa Wunuji et al. (2023) regarding hotels in Nigeria. The study identified factors such as the aesthetics of physical amenities, employee tidiness, seating comfort, and room decor as influencing customer retention. Additionally, Mng’ong’o (2023) investigated the impact of tangibility and assurances on customer retention in community banks in Tanzania, using a mixed-methods survey. Although the contexts differed, hospitality versus banking, the study concluded that tangible elements significantly affected customer retention, advocating the strategic use of tangible assets to enhance retention.

Moruku et al. (2025) studied airline customers in Nigeria. They found that companies should emphasize reliability and assurance to enhance customer retention, with reliability having a positive impact on customer retention. In a related vein, responsiveness, defined by Negassa and Japee (2023) as the perception of prompt assistance, was linked to improved customer retention in Nigerian hotels (Tsokwa Wunuji et al., 2023). Through these service industries, studies consistently found positive associations among responsiveness, reliability, assurance, and customer retention, though in different cultural and sectoral contexts. Rajagukguk et al. (2024) posited that trust and loyalty boost referrals, while Adzinyo et al. (2024) in Nigeria further confirmed the importance of assurance in retention. Moreover, Harriet et al. (2024) identified empathy as crucial, emphasizing its benefit across private service sectors. These findings collectively suggest that while each study centers on a specific dimension, multiple aspects of service quality are similarly vital across sectors.

Consistent with the above discussions, the following hypothesis is formulated:

• H1: Service quality positively influences customer retention.

• H1a: Tangibility has a positive impact on customer retention.

• H1b: Reliability has a positive impact on customer retention.

• H1c: Responsiveness has a positive impact on customer retention.

• H1d: Assurance has a positive impact on customer retention.

• H1e: Empathy has a positive impact on customer retention.

Accordingly, drawing from the literature, the authors suggest the subsequent conceptual framework (Figure 1) for the current research:

Figure 1.

Conceptual framework of the study

Source: own processing

Purposive sampling was used to distribute the questionnaire. This technique enabled the researcher to select participants most likely to provide relevant information to the study’s objectives, as described by Fahlevi (2025). The questionnaire targeted travel agency managers in the KRI governorates, resulting in 509 questionnaires distributed based on population size. Of these, 353 were returned, yielding a response rate of 70.6%. According to Sekaran and Bougie (2016), a 30% response rate is considered acceptable for survey research, making this study’s rate suitable for further analysis.

All constructs were scored using five-point Likert scale, with “1” denoting “strongly disagree” and “5” denoting “strongly agree”. The service quality scale is derived from multiple sources, such as Parasuraman et al. (1985), Adzinyo et al. (2024), Ogonu and Elvis (2023), Alemu (2023). This 25-item scale evaluates travel agency services across five dimensions: tangibility, reliability, responsiveness, assurance, and empathy. the (Farquhar, 2005)’s scale was used to measure customer retention. It comprises 22 items in five dimensions: direction (3), structure and culture (5), relationship strategy (4), information system (5), and staff (5). These items have also been used as a measure of customer retention by (Al-Hadrawi & Al-Shammari, 2017).

IBM SPSS 27 was used for categorization, cleaning, frequency analysis, and descriptive statistics, whereas SmartPLS 4.0 was used for SEM. Given its popularity in social science research, this study uses PLS-SEM rather than covariance-based SEM. Second, it benefits small-sample analyses when other applications are limited. Furthermore, PLS-SEM does not assume normally distributed data. Thus, empirical data do not need a normal distribution (Sann & Siripipattaworn, 2024). This made PLS-SEM suitable for our investigation. PLSSEM’s ability to handle complex models, such as those in this research, with second-order (customer retention) constructs and five independent variables, as well as non-normal data, lends support to its use (Hair et al., 2019).

A t-test of early and late responses showed no non-response bias (Armstrong & Overton 1977). This proves the lack of non-response bias. Data were also checked for common method bias. Improving data collection can reduce the CMB by anonymizing respondents or simplifying survey questions. Harman’s single-factor test (Podsakoff et al., 2024) and the comprehensive collinearity test were used to assess common-method bias. The initial component accounted for 37.25% of the variation, indicating scale independence (Singh et al., 2017). All construct VIF values were far below the recommended level of 3.3 (Kock, 2015). An assessment committee verified adherence to ethical standards, and ethical approval confirmed that participants understood the study’s goals and rights.

The largest proportion of respondents was from Sulaymaniyah (36.3%), followed by Erbil (36.8%), Duhok (26.9%), and Halabja (0.3%; note: Table 1 reports 3%, please reconcile). In terms of sex, men comprise 75.4% of managers in the study, while women account for 24.6%. Regarding age distribution, the data show that the 31–45 age group is the largest in the sample, accounting for 59.20%. The next age group, 30 and younger, makes up 27% of managers in this sector.

Looking at educational backgrounds, the majority of managers (74.5%) hold bachelor’s degrees, and 14.20% hold professional diplomas. By contrast, only 7% have a high school diploma or less, underscoring their underrepresentation in managerial positions. Postgraduate (Master’s or PhD) is the least represented at 4.20%. Examining work experience, the majority of managers (50.7%) have 6–15 years of experience. Consequently, 42.5% of managers have five years or less experience. Only 6.80% of the sample have over 15 years of management experience.

In this study, we used SmartPLS 4.0 to examine composite reliability and Cronbach’s alpha for construct validation (Bujang et al., 2018). Table 2 indicates that the composite reliability is acceptable, as the coefficients rho_c and rho_a are > 0.70. The factor loading in this study exceeds 0.50, thus satisfying the criteria for convergent validity (Peterson & Kim, 2013). According to Hair et al. (2017), the AVE for each construct should be greater than 0.5; the present study indicates that the AVE for all constructs exceeds 0.7, which is considered very good. The AVE of the study presented ranges from 0.754 to 0.850. Table 2 presents the AVE for all constructs, while Cronbach’s alpha exceeds 0.70.

Table 2 indicates that all variables met the measurement criteria, including the AVE, which exceeded 0.5. As per Hasan et al. (2020), the extent to which a test is independent of other tests assessing different constructs is termed discriminant validity. To ascertain this, the square root of AVE is necessary. To establish discriminant validity, the square root of the AVE of the construct must exceed the variance shared between the construct and other constructs (Sann & Siripipattaworn, 2024).

The current study utilized different methods to assess discriminant validity: the Fornell and Larcker (1981) approach, the factor loading technique outlined by Chin (1998), and the (HTMT) as described by Henseler et al. (2015) (Table 3 and Table 4).Table 3 displays the discriminant validity of all 15 constructs. The square root of the AVE for each latent construct must exceed the correlations of any other latent construct (Sann & Siripipattaworn, 2024).

VIF is commonly used to assess collinearity among formative indicators. Collinearity is evaluated when more than one formative measurement item exhibits high correlation and conveys the same information and tendencies (Hair et al., 2017). In PLS-SEM, a VIF of ≥5 is considered to indicate collinearity (Hair et al., 2014). In the same way, VIF values of 5 or more indicate serious collinearity issues among the indicators of constructs that are assessed (Becker et al., 2015), while optimal VIF values should be approximately three or below. Therefore, as shown in Table 5, all VIF values are below 5, suggesting no significant collinearity among the indicators.

In structural model prediction, path coefficients, p-values, t-values, and bias confidence intervals were employed to validate the proposed nexus. This section primarily examines the hypotheses about both direct and indirect effects. Consequently, employing Smart PLS and adhering to the procedures established by esteemed researchers (Hair et al., 2011; Hair et al., 2014; Hair et al., 2017), the PLS standard bootstrapping function was executed utilizing 353 cases and 10,000 bootstrapped samples, in line with (Al-Adwan et al., 2025; Singh & Manohar, 2024).

Consequently, it is recommended to use a minimum of 10,000 bootstrap samples, as shown in Figure 2. Given the rapid development of contemporary computing technologies, 10,000 bootstrap samples will provide minimal problems (Hair et al., 2022; Streukens & Leroi-Werelds, 2016). For a one-tailed test, a value of 1.645 is deemed significant. Conversely, for two-tailed tests, the suggested significant t-values must exceed 1.967 (Hair et al., 2014). The significance level of the t-value was evaluated using a one-tailed distribution. For a one-tailed statistical test, the t-value thresholds are ≥ 2.33, ≥ 1.65, and ≥ 1.28 for significance levels of 1%, 5%, and 10%, respectively. Values below the specified thresholds are deemed non-significant (Hair et al., 2017).

Figure 2.

Structural model

Source: own processing

The relationship between tangibility and customer retention was determined to be positive and statistically significant (β = 0.116, t = 2.388, p = 0.008, LL= 0.030, UL= 0.189). Consequently, H1 = SQTGB -> CR is supported. Nonetheless, the H2 = SQRLB -> CR hypothesis was rejected as the findings indicated that the relationship between reliability and customer retention (beta= 0.053, t=1.312, p > 0.095, LL= −0.005, UL= 0.128) was not significant. A notable and positive correlation was identified between responsiveness and customer retention (β = 0.184, t = 4.094, p = 0.000, LL= −0.029, UL= 0.227). Moreover, the findings indicate that in the context of H4 = SQASN -> CR, there exists a substantial positive correlation between assurance and customer retention (β = 0.093, t=2.040, p=0.021, LL= 0.021, UL= 0.172). Consequently, the hypothesis was supported. Similarly, for the study H5 = SQEMP -> CR, a significant positive correlation exists between the independent variable (IV) service quality (empathy) and the dependent variable (DV) customer retention (beta = 0.188, t = 3.977, p = 0.007, LL= 0.111, UL= 0.267). Thus, the hypothesis SQEMP -> CR was fully substantiated.

The predictive significance of the measurement model was subsequently evaluated using PLS prediction with a 10-fold cross-validation, a holdout-sample-based technique that facilitates case-level prediction for each item (Table 6).

Shmueli et al. (2019) argued that if all constructs of the PLS model are lower than those of the linear regression model, this may indicate significant predictive power. The predictive relevance cannot be confirmed if all PLS model constructs exceed the LM constructs. If the majority of item differences (PLS-LM) were lowered, this may suggest that the model possesses moderate predictive power. If the minority of item variances (PLS-LM) are reduced, the model’s predictive power will decrease. According to Hair et al. (2017), a Q² value above zero indicates that the exogenous constructs are predictive of the associated endogenous variables. The Q² result for CRDIR is 0.233, signifying strong predictive relevance. CRIS, CRRS, CRSC, and CRSTF have satisfactory predictive relevance, with Q² values of 0.251, 0.278, 0.264, and 0.260, respectively. Table 6 indicates that the PLS model exhibited greater accuracy than the LM model. Therefore, the predictive model has strong predictive power because all PLS-LM values are smaller.

This study examines service quality and customer retention using RM theory. Relationship marketing strengthens communication and improves service quality. Customer loyalty and retention are the main goals (Antwi, 2021). The first hypothesis examined the role of tangibility in customer retention. The results confirmed H1 and align with previous research (Ta & Chang, 2026; Mng’ong’o, 2023; Tsokwa Wunuji et al., 2023). As in earlier studies (Parasuraman et al., 1988), this study shows that tangibility is a core service-quality feature influencing travel agency retention.

The second hypothesis examines the link between reliability and retention. The results show no significant correlation, so H2 is rejected. Prior literature also reports little effect of reliability on retention. Sari et al. (2018) found no direct effect, though some researchers argue that a positive link exists (Moruku et al., 2025). Theoretical connections may remain despite the lack of statistical support.

Testing H3 found no significant link between responsiveness and retention, so H3 is rejected. Many studies report a strong positive effect of responsiveness on retention, but these results differ. Adzinyo et al. (2024), Tsokwa Wunuji et al. (2023) also report that employee responsiveness improves retention. Here, only a theoretical connection exists.

The fourth hypothesis examines the connection between assurance and retention. The results demonstrate a positive, significant correlation, supporting H4. As in previous research, these findings reinforce the notion that assurance is closely tied to retention (Lam et al., 2023; Adzinyo et al., 2024). In addition to the findings on assurance, the data also supports a relationship between empathy and customer retention. Findings show a strong positive correlation, supporting H5. These results align with earlier research suggesting that empathy is key to retention (Harriet et al., 2024).