South Africa’s banking industry comprises 65 registered banks. The banks with the largest market share, accounting for 90% of the total banking assets in the country, are Absa, First National Bank, Capitec, Nedbank and Standard Bank (Techpoint.Africa 2025). These banks serve an estimated 46 million retail customers. Despite South Africa’s persistent economic challenges, including low economic growth and high unemployment, these banks have reported stable financial performance and strong capital resilience. However, the competitive landscape is gradually shifting. Challenger banks, such as TymeBank, Bank Zero, Discovery Bank and South African Postbank, have gained traction since their market entry in 2019, capitalising on digital platforms and financial inclusion strategies to attract underserved segments (BusinessTech 2024; PWC 2025). This growing competition has contributed to a slow, but noticeable erosion of the ‘Big Five’s’ market dominance.

The convergence of economic pressures, the imperative to expand financial accessibility and evolving consumer expectations have forced traditional banks to re-evaluate and adapt their value propositions. Today’s banking customers demand increasingly personalised, transparent and seamless service experiences, valuing convenience as much as competitive pricing (KPMG 2025). In this rapidly evolving environment, banks must proactively understand and meet customer service expectations to maintain loyalty and market relevance. For banks to succeed, they must offer customers superior service quality, embrace change and new products/services and favour approaches that emphasise customers. A failure to adjust to these shifts could accelerate customer attrition towards more agile, digitally native competitors, further transforming South Africa’s banking sector in the coming years.

According to fairness theory, customers assess the fairness of events by considering accountability, reasonableness and justice. The theory posits that individuals evaluate outcomes by establishing whether impairment was preventable, intentional or justified. Fairness theory has been extensively applied in organisational behaviour research to explain how individuals respond cognitively, emotionally and behaviourally to perceived inequities and unfair treatment (Gilliland 1993). Therefore, the application of fairness theory in this study is appropriate, as banking services require customers to continuously evaluate not only service outcomes but also the processes and interpersonal treatment that accompany service delivery. These evaluations directly influence customers’ trust, satisfaction and subsequent relationship decisions.

Marketing literature identifies three core dimensions of fairness theory: Distributive, procedural and interactional fairness. Distributive fairness focuses on the perceived equity of outcomes, while procedural fairness concerns the rules, procedures and mechanisms used to arrive at those outcomes (Brockner et al. 2000). Interactional fairness reflects the quality of interpersonal treatment and relational engagement between service employees and customers (Chiu et al. 2009). Prior studies validate the relevance of these dimensions by demonstrating that customers do not assess fairness in a fragmented manner. Instead, they integrate outcome, process and interpersonal cues when forming overall fairness judgements. This validates the use of fairness theory in service research, as it captures the multidimensional nature of customer evaluations in banking environments where service encounters are frequent and relationally embedded.

Furthermore, scholars suggest that the three fairness dimensions do not exert equal influence in shaping customers’ fairness perceptions. This is particularly evident in service contexts, where experiences unfold over time and across multiple touchpoints. Marketing research confirms that customers develop an overall perception of service fairness based on their holistic experience with service providers (Lemon & Verhoef 2016; Törnblom & Vermunt 1999). Within service recovery contexts, distributive justice is primarily linked to customers’ evaluation of whether the recovery outcome adequately compensates for the service failure (Smith, Bolton & Wagner 1999). However, procedural and interactional fairness have been shown to play a more decisive role in shaping post-recovery satisfaction and trust, as customers often place greater emphasis on transparency, empathy and respectful treatment than on the outcome (Liao et al. 2022). These findings validate the relevance of fairness theory for this study by highlighting its explanatory power in understanding how service delivery practices influence customer perceptions and relational responses in banking.

In contrast, relationship marketing theory provides a broader relational lens by explaining how service interactions translate into long-term customer-firm relationships. Multiple scholars have confirmed the importance of relationship marketing in the development and management of enduring customer relationships, particularly in service industries (Naalchi Kashi 2024). Azhari and Utari (2023) argued that relationship marketing theory is grounded in a deep understanding of customers’ relational needs, enabling service providers to manage customer expectations more effectively. Empirical studies applying relationship marketing theory in banking and financial services contexts demonstrate that relational practices, such as trust building, personalised communication and perceived commitment, significantly enhance customer engagement and long-term loyalty (Alvionita et al. 2024). These findings validate the relevance of relationship marketing theory for this study, as they confirm its ability to explain customers’ intentions to maintain long-term relationships beyond transactional considerations.

Against this background, fairness and relationship marketing theories are theoretically complementary. Fairness theory is grounded in the principle of accountability and focuses on how customers evaluate the fairness of specific service encounters through distributive, procedural and interactional justice (Carr 2007). Contrastingly, relationship marketing theory explains how repeated fair and respectful interactions accumulate over time to strengthen relationship quality, trust and commitment (Morgan & Hunt 1994). Previous studies confirm this complementarity by demonstrating that fairness perceptions are critical antecedents to relationship marketing outcomes, including trust, satisfaction and relationship commitment (Sari, Hartoyo & Muflikhati 2025). As such, the integration of these two frameworks is appropriate for the study’s objectives, as fairness theory explains how customers evaluate service experiences at an interactional level, while relationship marketing theory explains how these evaluations translate into sustained relational outcomes. The frameworks provide a robust and holistic lens for understanding the development and management of long-term banking relationships in a service-orientated context.

The multidimensional nature of service fairness has been widely confirmed in marketing literature, particularly in service and relationship-orientated contexts. Scholars (Aguilar-Rojas, Fandos-Herrera & Pérez-Rueda 2024; Samoggia et al. 2023) confirm that service fairness consists of interactional, procedural and distributive dimensions. Interactional fairness reflects politeness, respect, empathy and professional communication during customer–employee interactions, which is relevant in relationship marketing where repeated interpersonal exchanges form the basis of long-term relationships. Procedural fairness refers to customers’ perceptions of the fairness of service processes, including transparency, consistency, opportunities for voice and professional competence (Tyler 1988). Distributive fairness relates to the perceived equity of service outcomes (Baumann et al. 2023). Therefore, service fairness’s inclusion in this study is justified, as relationship marketing is fundamentally concerned with how customers experience and evaluate ongoing interactions, rather than isolated transactions. Prior studies validate this relevance by demonstrating that fairness perceptions shape trust, satisfaction and willingness to continue relationships, particularly in high-contact service environments, such as banking (Kaweesi 2023; Rusydi et al. 2024).

Within a relationship marketing context, service fairness is an important relational signal that influences customers’ emotional and cognitive bonds with service providers. Affective commitment captures this emotional bond and reflects customers’ attachment to service providers based on trust, identification and perceived relational value (El-Adly, Souiden & Khalid 2024). Abid et al. (2023) noted that affective commitment is rooted in confidence in the brand, rather than convenience or contractual dependence. Empirical studies confirm the relevance of affective commitment by showing that emotionally committed customers are more likely to remain loyal, recommend service providers and forgive occasional service failures (Wang & Binti Omar 2023). These findings validate the inclusion of affective commitment in relationship marketing research, as they demonstrate that emotional bonds stabilise long-term relationships even when service performance is imperfect (Petzer & Roberts-Lombard 2024). In service contexts, affective commitment is critical for reducing churn and strengthening future loyalty intentions (Kigen & De Villiers 2024).

Contrastingly, calculative commitment reflects a more rational attachment based on cost-benefit considerations. Customers remain in relationships because switching would involve financial, time-related or effort-related costs (Petzer & Roberts-Lombard 2022). Scholars (Susriyanti et al. 2023) caution that calculative commitment often reflects obligation rather than genuine loyalty. Prior studies validate its relevance by demonstrating that customers driven by calculative commitment are more likely to defect when superior alternatives become available (Kipkogei, Misoi & Biru 2022). This confirms its importance in relationship marketing research, as it highlights the vulnerability of relationships that lack emotional attachment (Geyskens, Steenkamp & Kumar 1999). Consequently, service providers must continuously deliver value, while strengthening emotional bonds to transition customers from calculative to affective commitment (Roberts-Lombard et al. 2024).

The ultimate outcome of relationship marketing efforts is customer loyalty, which is increasingly viewed as a multidimensional construct. Scholars (Balducci & Marinova 2022; Diallo, Moulins & Roux 2021; Nunkoo et al. 2025) argue that loyalty should be assessed through attitudinal and behavioural dimensions. Attitudinal loyalty reflects emotional attachment, beliefs and preference for a service provider (Dikcius et al. 2024), while behavioural loyalty captures repeat-purchase intentions and resistance to switching (Anwar et al. 2024). Extant research validates this dual conceptualisation by showing that attitudinal loyalty often precedes and strengthens behavioural loyalty in relationship-driven services (Zikienė, Kyguolienė & Kisieliauskas 2024). This supports the inclusion of both loyalty dimensions, as relationship marketing seeks not only repeated patronage, but also deep relational attachment (Zikienė et al. 2024). Consequently, it is argued that collectively service fairness shapes customers’ emotional and rational commitment, which drives both attitudinal and behavioural loyalty, making these constructs theoretically complementary and highly relevant for exploring relationship marketing outcomes in banking.

Marketing literature extensively validates that customers become more attached to their service provider when they experience respectful and courteous service perceived as fair (Kwong et al. 2023; Rusydi et al. 2024). Boakye et al. (2023) concurred, noting that professional problem resolution and respectful treatment foster emotional connection. However, over the past two decades, some scholars have debated the uniformity of this effect, suggesting that cultural context, customer personality and prior experiences can moderate the relationship between perceived fairness and attachment (Mattila & Patterson 2004; Truong 2025). Additionally, inconsistencies exist regarding which fairness dimension (interactional, procedural or distributive) exerts the strongest influence on emotional attachment, with some studies reporting stronger effects for procedural fairness and others highlighting interactional cues (Bahri-Ammari & Bilgihan 2017). However, fair service procedures can strengthen confidence, deepen attachment and foster belonging (Krolikowska & Kuenzel 2024). As such, the following hypothesis is proposed:

Scholars (Farooq & Moon 2020; Lim et al. 2025) have endorsed the direct relationship between service fairness and calculative commitment. According to Chang (2023), customers experiencing fair processes and reliable service quality are more likely to recognise the value of maintaining their relationship with providers. Similarly, when financial charges are perceived as fair relative to service delivered, and customers feel they receive superior advantages, their commitment tends to strengthen (Calder, Malthouse & Omatoi 2023; Ozuem et al. 2024). However, some studies over the past two decades highlight inconsistencies, reporting that the strength of this relationship can vary depending on customers’ prior experiences, perceived switching costs and service context, with fairness not always significantly predicting calculative commitment. This suggests that while fairness can encourage calculative commitment, its effect is contingent and context-dependent (Al-Abdi 2010; Roberts-Lombard et al. 2022). Accordingly, the following hypothesis is proposed:

Marketing literature has extensively confirmed the importance of affective commitment as a driver of future loyalty intention. Scholars (Wang & Yang 2025) suggest that a deep emotional connection with a service provider, such as a bank, strengthens customers’ intentions to maintain the relationship. Ho and Wong (2023) similarly noted that customers with a strong sense of affiliation are more likely to continue their association. However, some studies debate the uniformity of this effect, indicating that affective commitment’s influence on loyalty may vary depending on cultural context, prior experiences and individual customer personality traits, with certain customer segments showing weaker emotional responses. These findings suggest that while affective commitment generally enhances future loyalty, its strength is context-dependent (Bothma 2020; Mgiba & Ndlazi 2025). Thus, fostering affective commitment can support banks in encouraging long-term customer retention (El-Adly et al. 2024). As such, it is hypothesised that:

Marketing literature confirms that calculative commitment is an important driver of future loyalty intent. For instance, Abid et al. (2023) highlighted that Pakistani banking customers are more likely to continue their relationship when they perceive greater economic benefits compared to competitors. Similarly, Kigen and De Villiers (2024) found that high switching costs reinforce South African customers’ commitment and loyalty intentions. Interestingly, some studies debate the uniformity of this effect, suggesting that calculative commitment’s influence on loyalty can vary depending on individual customer differences, contextual factors and perceived alternatives, with some segments prioritising emotional bonds over economic considerations (Bothma 2020; Truong 2025). Furthermore, in contexts where switching costs are low or alternatives are readily available, calculative commitment may weaken as a predictor of sustained loyalty (Mgiba & Ndlazi 2025). Consequently, when customers believe their bank delivers economic value, convenience and a satisfying service experience, their overall commitment and long-term loyalty can be reinforced (Kim, Jindabot & Yeo 2024). Hence, it is hypothesised that:

The validation of affective commitment as a mediator between antecedents and loyalty in the marketing field has been widely confirmed. For instance, Bahadur, Aziz and Zulfiqar (2018) found that Chinese customers who develop strong emotional bonds with their service providers are more likely to remain committed over time. Similarly, research in Indonesia’s banking sector (Karim et al. 2023) showed that an emotional connection increases customers’ likelihood of maintaining the relationship. Hence, when banks understand and address customers’ affective commitment needs, overall commitment is enhanced (Ali & Song 2023; Naalchi Kashi 2024). However, studies debate the consistency of affective commitment’s mediating role, suggesting that its strength can vary across cultures, service contexts and customer personality traits, with some segments showing weaker emotional effects on loyalty outcomes (Putera & Famiola 2024; Petzer & Roberts-Lombard 2024). These inconsistencies indicate that affective commitment may not always operate uniformly as a mediator in all settings. Consequently, the following hypothesis is proposed:

Extensive marketing research confirms the mediating role of calculative commitment in financial services. Banking customers are said to expect economic value from their relationship with their bank (Samuel et al. 2025), and greater economic benefit strengthens willingness to remain in the relationship (Abid et al. 2023). Moreover, Petzer and Roberts-Lombard (2024) affirmed that calculative commitment mediates loyalty when economic advantages outweigh costs. Nevertheless, some studies debate its consistency, finding that calculative commitment may be less predictive of loyalty when emotional bonds or perceived service quality are stronger drivers, especially in high trust contexts (Khan et al. 2020; Marshall 2010). Additionally, inconsistency arises when customers prioritise convenience or brand reputation over pure economic benefit, suggesting that mediating role of calculative commitment is context-dependent and may vary across customer segments (Roberts-Lombard et al. 2022; Shin et al. 2019). Thus, banking customers may view their relationships as wise when benefits outweigh efforts (Ekejiuba, Bohari & Ab Hamid 2025). Therefore, it is hypothesised that:

An extensive literature review has confirmed that trust can function as a moderating variable (Abid et al. 2023; Biswas, Jaiswal & Kant 2023). For example, a banking study in Indonesia found that when customers believe their product and service needs are satisfied, trust in their bank increases (Novitasari et al. 2024). Anggraini (2024) concurred, noting that customers are more likely to trust banks perceived as warm, responsive and professional. Since the dawn of the new millennium, scholars have debated the consistency of trust’s moderating role, suggesting that its strength may vary depending on cultural norms, prior experiences and service contexts, with trust sometimes exerting weaker moderation in high-service-quality environments where fairness alone predicts commitment (Liu, Cheng & Ouyang 2021; O’Connor & Assaker 2024). Additionally, evidence indicates that in commoditised banking segments, customers may prioritise convenience or brand reputation over trust, thereby weakening its moderating effect (Khan et al. 2023). Therefore, trust is proposed as a moderator in the model linking service fairness with affective and calculative commitment. Considering the above, it is hypothesised that:

Based upon the discussion, the model in Figure 1 is proposed.

The study’s sample comprised individuals over 18 who resided in South Africa’s Gauteng province and banked with one of the country’s five leading retail banking institutions. Purposive sampling was utilised to select suitable participants for the study (Malhotra 2019) with the assistance of a professional field services agency, whose trained fieldworkers found qualifying respondents. Initial screening questions ensured individuals fulfilled the study’s criteria. Data were collected via interviewer-administered questionnaires, which included a consent form and screening questions to confirm eligibility, alongside items capturing respondents’ demographic details and retail banking practices. Additionally, the survey collected data concerning the study’s primary constructs. Prior to analysis, all collected data underwent a rigorous process of editing and cleaning. Furthermore, ethical considerations were applicable throughout the data collection process, following the guidelines approved by the Ethics Committee who granted ethical clearance for this study.

A total of 550 valid responses were obtained and analysed. Among the participants, 57.8% were men and 44.7% were aged over 50 years. The majority (60%) held post-secondary qualifications and primarily spoke English (94.3%). Most respondents were employed full-time (64%), were in marital or cohabiting relationships (61%) and banked primarily with Standard Bank (61.5%). Moreover, 58.3% had been with their main bank for 7 years or less, while 57.8% held cheque accounts, with 56.3% reporting monthly banking fees of R250 or less.

The study employed a seven-point Likert scale to evaluate the main constructs, with one indicating ‘strongly disagree’ and seven indicating ‘strongly agree’. The measurement instruments were sourced from established literature: Service fairness, affective commitment, calculative commitment and trust were drawn from Giovanis, Athanasopoulou and Tsoukatos (2015), while attitudinal and behavioural loyalty were drawn from Mandhachitara and Poolthong (2011). Table 1 presents the details of these measurement scales.

In the study, SmartPLS 3.2.7 was applied to test the direct effect (H1–H6) and the Hayes PROCESS Macro for SPSS (Model 4) was used to test the indirect effects (H7a and H7b). Furthermore, the normality of individual items was assessed using the Kolmogorov–Smirnov and Shapiro–Wilk tests to guide the selection of the most appropriate model estimation technique. In line with Muthén and Muthén (1998–2017), the maximum likelihood method (MLM) was employed when the data did not meet normality assumptions, while standard maximum likelihood estimation was applied when normality was confirmed. Given that service fairness was conceptualised as a second-order construct comprising three primary dimensions (interactional, procedural and distributive fairness), two competing measurement models were tested. The first treated the dimensions as separate, but correlated first-order reflective factors, while the second model positioned service fairness as a higher-order construct with the three dimensions as reflective indicators. To compare the models, the Bayesian information criterion (BIC) was utilised, with preference given to the model displaying the lower BIC value, which indicated a superior fit.

Following model selection, confirmatory factor analysis was conducted to assess the measurement model’s psychometric robustness. Reliability was evaluated using composite reliability (CR) and Cronbach’s alpha, with a threshold of 0.7 for acceptability (Bagozzi & Yi 1988; Hair et al. 2014). Factor loadings were also analysed, with acceptable loadings exceeding 0.7 at a significance level of p < 0.01 to establish convergent validity. Additionally, average variance extracted (AVE) values were reviewed to ensure they were above 0.5. Discriminant validity was confirmed using the Fornell–Larcker criterion, which requires that the square root of a construct’s AVE surpass its correlations with other constructs (Fornell & Larcker 1981). Furthermore, to evaluate model fit for the measurement and structural models, the researchers used four key indices: The Satorra–Bentler chi-square divided by degrees of freedom (acceptable if < 3), root mean square error of approximation (RMSEA; < 0.08), standardised root mean squared residual (SRMR; < 0.08) and comparative fit index (CFI; > 0.90), consistent with the recommendations of Hu and Bentler (1999) and Muthén and Muthén (1998–2017).

After testing the direct hypotheses (H1–H6), indirect effects (H7a–H7b) were explored through mediation analysis using the Hayes PROCESS Macro in SPSS (Model 4). The theoretical rationale supported affective and calculative commitment as mediators between service fairness and customer loyalty. A bootstrapping procedure with 5000 samples generated 95% bias-corrected confidence intervals (BCCIs). These intervals were examined to determine whether zero was included between the lower-level confidence interval (LLCI) and upper-level confidence interval (ULCI), indicating the presence or absence of significant indirect or direct effects. The classification of mediation types followed the guidelines established by Zhao, Lynch and Chen (2010). In addition, moderation was evaluated using bootstrapping estimation to determine whether the confidence intervals for the interaction term (x*m) contained zero. If zero was not present within the BCCIs, the interaction effect was considered significant. In cases where a significant interaction was found, a simple slope analysis was conducted to explore the nature of the interaction in more detail. Conclusively, loyalty was measured as a higher-order construct, reflecting attitudinal and behavioural loyalty (refer to Table 2).

Ethical clearance to conduct this study was obtained from the Humanities and Social Science Research Ethics Committee of the University of the Western Cape (No. [HS25/5/45]).

The data failed to meet the assumption of univariate normality. As a result, the MLM estimator was deemed most appropriate for estimating the study’s models, as it produces parameter estimates along with a mean-adjusted Satorra–Bentler chi-square statistic and robust standard errors that are resilient to violations of normality (Muthén & Muthén 1998–2017).

The first model specified interactional, procedural and distributive fairness as separate first-order reflective constructs, yielding a BIC value of 6437.110. In contrast, the second model conceptualised service fairness as a second-order reflective construct, with its three dimensions (interactional, procedural and distributive fairness) serving as indicators, resulting in a BIC of 6710.210. A comparison of the two models revealed that the second model had a lower BIC value, supporting the interpretation of service fairness as a second-order construct composed of the three first-order dimensions. Consequently, service fairness was modelled accordingly in the study’s measurement and structural analyses.

The measurement model demonstrated strong reliability and validity, as indicated by CR, internal consistency and convergent validity (see Table 2). Both CR and Cronbach’s alpha values exceeded the threshold of 0.7. Regarding convergent validity, all factor loadings for the items measuring the main constructs were above 0.7 (two-tailed p < 0.01), and the AVE for each construct surpassed 0.5.

Table 3 demonstrates discriminant validity, as the square root of the AVE for each construct is greater than its correlations with the other constructs (Fornell & Larcker 1981).

The model fit statistics indicated an acceptable fit between the data and the model: Satorra–Bentler χ²/df ratio = 2.77, RMSEA = 0.071, CFI = 0.912, SRMR = 0.047. Therefore, it was appropriate to proceed with the evaluation of the structural model.

The structural model demonstrated a good fit to the data overall, although the Satorra–Bentler χ²/df ratio slightly exceeded the acceptable threshold, registering at 3.03. The other fit indices indicated an acceptable model fit: RMSEA = 0.079, CFI = 0.932, SRMR = 0.049. The results of the structural model are summarised in Table 4. Service fairness was found to be positively and significantly related to affective (β = 0.943; p < 0.01) and calculative (β = 0.754; p < 0.01) commitment. Furthermore, both affective (β = 0.713; p < 0.01) and calculative (β = 0.654; p < 0.01) commitment had a significant and positive effect on loyalty. As such, hypotheses H1–H4 were supported. These findings confirm that service fairness serves as a key driver of affective and calculative commitment, while affective and calculative commitment serve as determinants of customer loyalty in the given context.

Table 5 outlines the results pertaining to the two hypotheses addressing indirect effects (H5 and H6). The analysis indicated that affective commitment partially mediated the relationship between service fairness and loyalty, with an indirect effect of 0.110 and a 95% BCCI of 0.070–0.155. Similarly, calculative commitment served as a partial mediator between service fairness and loyalty, with an indirect effect of 0.210 and a 95% BCCI of 0.155–0.266. Following the mediation criteria proposed by Zhao et al. (2010), both instances demonstrated evidence of partial mediation. Consequently, both H5 and H6 were supported.

Table 6 summarises the moderation analysis examining whether trust moderates the relationships between key antecedents and relational outcomes. The interaction effect in the relationship between service fairness and affective commitment was negative and statistically significant (B = –0.055; LLCI = –0.090; ULCI = –0.021; t = –3.060; p < 0.05). Similarly, the interaction between customer orientation and perceived value on trust was negative and statistically significant (B = –0.060; LLCI = –0.095; ULCI = –0.024; t = –3.315). Despite these significant interaction terms, the overall moderation hypotheses were not supported, as indicated by the non-significant two-tailed p-values reported for the moderation tests. The negative coefficients suggested that higher levels of trust weaken the strength of these relationships, indicating a buffering rather than an enhancing effect. Overall, the findings imply that trust does not function as a true moderator in these relationships, but may operate as an independent relational mechanism influencing customer evaluations and commitment formation. As such, H7a and H7b were not supported.

Table 7 and Table 8 provide the results for the conditional effects of service fairness on affective and calculative commitment at different levels of trust. Table 7 reflects the conditional effect of service fairness on affective commitment at varying levels of trust. The results indicated that service fairness had a positive and statistically significant effect on affective commitment across low, moderate and high trust levels (p = 0.0001). However, the magnitude of this effect decreased as trust increased. At low trust (4.402), service fairness exerted the strongest influence (β = 0.397), followed by moderate trust (β = 0.325), with the weakest effect observed at high trust (β = 0.255). This pattern suggests that perceptions of fairness are particularly critical for strengthening affective commitment when trust is relatively low, highlighting a compensatory role of fairness in relationship development.

Table 8 presents the conditional effect of service fairness on calculative commitment at different levels of trust. The findings revealed a positive and statistically significant effect of service fairness on calculative commitment across low, moderate and high trust levels (p = 0.0001). The strength of this relationship decreased as trust increased. Specifically, the effect was strongest at low trust (β = 0.485), followed by moderate trust (β = 0.431), and weakest at high trust (β = 0.373). This trend suggests that service fairness plays a more prominent role in reinforcing calculative commitment when trust is limited, underscoring its importance in shaping rational, cost-benefit evaluations within customer–bank relationships.

The study has identified significant positive relationships between service fairness, affective and calculative commitment and customer loyalty in retail banking. These relationships are evident when customers perceive their bank’s service as fair, dependable and transparent; feel an emotional connection to the bank; recognise economic benefits from staying; and view the bank as trustworthy in its operations. Consequently, the findings enhance understanding of retail banking customers’ service fairness expectations in an emerging market, potentially fostering stronger perceptions of fairness and greater loyalty intentions. Regarding limitations, the study centres on service fairness as a precursor to affective and calculative commitment. Additionally, the sample was selected using purposive sampling, a non-probability method, which limits the ability to generalise the results broadly. Furthermore, the structural model’s Satorra–Bentler χ²/df ratio slightly exceeds recommended guidelines, as this implies the possibility that additional variables may be needed to improve model fit.

Future research could adopt a multidimensional and interdisciplinary approach by integrating behavioural economics, digital trust and data ethics to assess how fairness perceptions evolve in tech-driven service environments. Longitudinal studies could also examine how loyalty develops over time in response to evolving customer experiences, particularly during crises or institutional changes. Additionally, incorporating psychographic profiling and artificial intelligence-driven sentiment analysis could uncover deeper emotional responses to perceived fairness. Moreover, comparative studies across generational cohorts (e.g. Generation Z versus Generation X) may also reveal how digital nativity influences commitment and loyalty. Finally, qualitative explorations into fairness narratives in historically underserved communities could deepen contextual insights in the emerging market banking landscape.