The term SOR was expanded by Mehrabian and Russell (1974) within the framework of environmental psychology. They posited that external stimuli (S) impact an individual’s cognition, emotions and internal states (O), which subsequently affect their attitudes and behaviours (R). The SOR theory highlights the significance of both external stimuli and the individual’s internal perceived state in understanding behaviour. External stimuli influence the internal psychological state, which then drives various behavioural responses (Kim et al., 2021).

Stimulus-organism-response theory has been extensively applied in online user behaviour research, particularly in traditional shopping and e-business contexts. Key studies include Luqman et al. (2017), who investigated the impact of aesthetic features on consumers’ internal states and subsequent purchasing behaviours. Additionally, SOR theory has been applied to social networking sites, with Qiu et al. (2023) finding that exhaustion and technological pressure can decrease Facebook use. Research also identifies negative information factors, such as vulgar content and rumour flooding, that contribute to social media burnout and reduced user engagement (Wolter et al., 2023).

A virtual community is a group formed on the Internet where members share knowledge and information through interactive communication, creating interdependent and relatively stable social networks (Capdevila & Mérindol, 2024). The widespread use of the Internet has accelerated and enhanced information transfer, facilitating the growth of virtual communities. These communities transcend traditional time and space constraints, enabling members to interact and communicate through online platforms (Zhu & Yang, 2023).

Grzelka (2020) defined a virtual community as an Internet-mediated platform initiated by e-business platforms, consumers or third parties, where groups of consumers engage in continuous social interactions. These interactions, including information exchange and knowledge sharing, establish stable social relationships among members. Practically, platform managers recognise that virtual communities facilitate the maintenance of relationships between e-business platforms and consumers and that knowledge sharing within these communities can address management issues, thus improving platform usability (Zhang et al., 2024).

With the advancement of the Internet, online shopping has surged in popularity among Chinese consumers. However, the virtual nature of the Internet introduces risks and uncertainties in the shopping process. Online reviews help mitigate information asymmetry and reduce purchasing risks, garnering significant attention from the academic community (Ikhsan et al., 2024).

Online reviews consist of both subjective evaluations and objective information about purchased goods from e-business or third-party review sites, significantly influencing consumer purchase decisions. Unlike traditional word-of-mouth, online reviews benefit from Internet technology, allowing for intuitive display and enabling consumers to assess credibility based on review quality (Xu, 2021). However, online reviews face issues such as information overload and integrity concerns, making it challenging to evaluate their usefulness. To address this, e-business platforms are increasingly focusing on improving review content. For instance, Dianping.com recruits ‘experiencers’ to post reviews, and many communities award medals to highlight valuable contributions from reviewers (Burtch et al., 2021).

The platform market institution trust refers to the extent to which buyers trust the regulatory body and regulatory norms in the platform transaction process (Beck et al., 2024). In online transactions, the creation and maintenance of institutional trust relies mainly on various documents and credentials, especially third-party authentication and third-party payment escrow. McKnight et al. (1998) divided institutional trust into two dimensions, one is situational normality, that is, the transaction is based on normal transaction links and processes, and the interaction process meets the expectations, which leads to trust; the other is structural assurances, that is, the success of the transaction is guaranteed through contractual agreements, rules and procedures, warranties, etc.

Self-efficacy, as defined by Bandura et al. (1999), is the belief in one’s ability to achieve goals and complete tasks. Research indicates that self-efficacy can stem from observational learning, verbal persuasion and past successes (Graham, 2022). In platform transactions, characterised by high frequency and uncertainty, self-efficacy is reflected in buyers’ confidence in successfully shopping through e-business platforms. This is termed platform shopping efficacy, which describes buyers’ belief in their ability to complete online transactions, such as acquiring information, purchasing correctly, finding suitable sellers and resolving issues (Martínez-López et al., 2020). Higher platform shopping efficacy suggests that buyers are more confident in finding quality sellers and achieving successful transactions (Livan et al., 2022).

In marketing, Zott et al. (2000) introduced the concept of ‘stickiness’, specifically ‘Internet stickiness’ in the e-business context. This concept has been applied in studies as ‘website stickiness’, focusing on a website’s ability to retain customers, and as ‘user stickiness’ or ‘consumer stickiness’, highlighting customer retention on platforms. Unlike ‘customer loyalty’, which has distinct characteristics, most research on stickiness employs models such as the information system success model (ISSM), technology acceptance model (TAM), commitment-trust theory (CTT) and expectation-confirmation theory (ECT). These studies suggest that stickiness results from factors such as perceived quality and perceived usefulness, reflecting individual rational decision-making processes (Deodhar & Gupta, 2023).

This article draws on Kankam et al. (2023) to deconstruct the conceptual connotation and formation mechanism of ‘customer stickiness’, and divides customer stickiness into two dimensions, visit stickiness and purchase stickiness. Visit stickiness refers to the behaviour of customers who keep browsing for a long period of time, repeat browsing over a period or browsing in depth on an e-business platform (Lee et al., 2021). Purchase stickiness refers to the behaviour of customers who maintain a high purchase frequency, purchase type or purchase quantity on an e-business platform over a period (Goyal & Dutta, 2021).

Reputation is the overall trust and perceived evaluation of consumers in a company, product or service based on past experience and word-of-mouth communication (Kuppelwieser et al., 2022). Reputation is the digital asset of an online platform, which brings quantifiable business value and sustainable growth drivers to the platform by building trust, boosting conversions, optimising search rankings and enhancing competitiveness (Fatorachian & Ramesh, 2025). The perceived reputation of platform sellers explored in this article refers to the rational perceptions such as understanding, judgement or evaluation that buyers hold about all sellers of platform-based e-business as a marketplace (Rohn et al., 2021). During platform transactions, buyers will not only choose high-reputation sellers but also high-reputation platform-based e-business companies (Li et al., 2019). However, previous studies have primarily concentrated on the signalling effect of seller reputation but neglected the halo effect of platform-based e-business reputations (Gala & Mueller, 2024).

Online reviews, a key form of online word-of-mouth, significantly influence consumer purchase decisions (Lei et al., 2022). Consumers routinely consult these reviews before purchasing, benefiting from their ability to overcome time and space constraints and enhance communication between merchants and consumers (Hou & Ma, 2022). However, the Internet’s indirect contact, low cost and anonymity lead to information overload and uneven quality, including misleading content from online trolls. This makes consumers to demand higher usefulness from reviews, requiring them to sift through numerous reviews to find useful information, increasing their information processing costs (Rosário & Raimundo, 2021).

Consequently, academic research is shifting focus from the impact of online reviews on purchase decisions to improving review content and perceived usefulness (Harrigan et al., 2021). Current research primarily addresses the quantitative and textual characteristics of reviews, with less emphasis on content (Ghorbani et al., 2022). This article explores the relationship between online reviews and the perceived reputation of platform sellers, examining how customer stickiness moderates this relationship and the chain mediation roles of platform market system trust and platform shopping efficacy. This study offers new insights into word-of-mouth research and practical implications for guiding effective consumer evaluations in e-business.

As a major force in global e-commerce, the expansive and rapidly evolving Chinese market plays a crucial role in the global economy (Lundvall & Rikap, 2022). Survey data from Chinese CBEB platforms are vital because of the market’s size and its rapid technological and business model innovations. These platforms are often at the forefront of adopting new e-commerce technologies, such as artificial intelligence and advanced analytics for consumer behaviour prediction and personalised marketing, setting benchmarks that influence global e-commerce trends (Micu et al., 2021).

Incorporating data from China’s diverse and dynamic market refines theoretical models on consumer behaviour, market entry and digital strategy (Gomes et al., 2022). This enhances the inclusivity and applicability of these models across various economic environments and deepens understanding of unique factors driving consumer engagement in digital marketplaces. Insights into Chinese consumer behaviour offer a perspective on global market trends and preferences, providing valuable lessons for international businesses targeting this lucrative market (Brunzel, 2024). Studying Chinese consumers on cross-border platforms reveals broader trends, such as the rapid adoption of mobile payments and social commerce, which influence global consumer expectations and business strategies. These insights can predict emerging e-commerce practices and technologies with potential global relevance.

China’s cross-border e-commerce market is developing rapidly and has a large number of users. According to relevant data, China’s cross-border e-commerce import and export scale will reach 2.72 trillion yuan in 2023, and the scale of users is also expanding. Such a large user group, covering different ages, genders, regions, consumption habits and other characteristics, provides a rich and diverse sample for the study of international business theory, which can more comprehensively reflect various phenomena and laws in international business activities. So, leveraging Chinese CBEB user data is crucial for advancing international business theory (Yan et al., 2023). It allows scholars and practitioners to challenge and expand traditional theories, adapting them to accommodate the complexities of the digital age and cross-cultural consumer interactions (Papadopoulos & Cleveland, 2023). These data not only inform business strategies tailored for the Chinese market but also enhance global strategic planning, providing a competitive edge in the fast-paced world of international e-commerce.

The contributions and the gaps with previous research are elaborated upon in the following aspects. Firstly, previous research mainly focused on domestic e-commerce platforms or single cultural contexts, examining consumer trust and security using models such as TAM (Hanakata & Bignami, 2023; Wen et al., 2024) and proposing strategies to build consumer trust (Wang et al., 2020). In contrast, this study addresses the complexities of CBEB platforms serving a global audience. Online consumer behaviours, including review habits and evaluation criteria, vary significantly across cultures, impacting perceptions of platform reputation. Our survey incorporated diverse cultural perspectives, languages and consumer habits to better understand how these factors affect online reviews and the reputation of international e-commerce platforms, providing a more accurate global consumer feedback depiction.

Secondly, prior studies primarily investigated simpler online reputation systems, such as eBay’s rating system, and the role of online conversations in word-of-mouth transmission to build consumer trust (Oraedu et al., 2021). These studies analysed forums and discussion groups, focusing on how review content quality and structure affected purchasing behaviour (Palalic et al., 2021). They did not extensively explore the impact of social media reviews on platform reputation. In contrast, this study examined how technological advancements, including social media and intelligent systems, have reshaped e-commerce by accelerating consumer feedback spread and enhancing review analysis accuracy. This approach addresses a gap in understanding how these technological changes influence consumer perceptions and platform reputation.

Lastly, earlier studies largely focused on domestic traditional retailers and e-commerce platforms, emphasising product characteristics within a single market and customer loyalty in traditional e-commerce (Al-Adwan et al., 2020; Cao et al., 2020). They did not explore the role that customer loyalty plays in reviews and platform reputation. This study enhances the discourse by examining how customer stickiness moderates the relationship between online reviews and platform reputation. This perspective is crucial for developing strategies tailored to the diverse needs and behaviours of users in CBEB platforms.

In summary, by bridging these gaps, the article not only extends the existing body of knowledge on the influence of online reviews in e-commerce but also provides actionable insights that can help CBEB platforms enhance their strategic approaches to managing consumer perceptions and strengthening their market position globally. The inclusion of cultural diversity, technological progression and the unique characteristics of CBEB platforms contributes to a more comprehensive framework for understanding the complex dynamics of global e-commerce.

In this article, consumers who browse online reviews of virtual communities and make online purchases on CBEB platforms are taken as survey respondents, and samples are selected using random sampling method. On the basis of the questionnaires of previous research, the corresponding contextual adjustments were made by combining the characteristics of virtual communities and China’s CBEB platforms (e.g., regarding platform features, users can be asked whether they have used the search functions of Chinese shopping apps such as Taobao or JD.com, and to evaluate their convenience; regarding community interaction, users can be asked whether they share shopping experiences on Chinese social apps such as Little Red Book or Douban, and to assess the impact on their purchasing decisions). The electronic questionnaires were distributed to virtual community platforms (e.g. Zhihu, Little Red Book, Weibo, Today’s headlines, Amazon 1688, Hugo.com, Fubu foreign trade forum, Global foreign trade forum, Chuanglang forum, Wear seller community, DHgate.com and Ennews.com, etc.) to find the data. The questionnaire consisted of three parts: survey background introduction, consumer profile survey and consumer behaviour survey, and the survey items were assessed by employing a five-point Likert scale for measurement.

There are 41 items in this study, including 8 personal information items as well as 33 consumer behaviour survey items. The scale is contextually adapted to the characteristics of Chinese CBEB platforms, and the specific variable measures and literature sources are provided in Table 1.

Pre-survey is a necessary step before the formal research; through the pre-survey, the problem can be identified in the questionnaire design, and corresponding changes and adjustments can be made to address these problems to ensure the influence of the formal research. In the pre-survey, a total of 120 valid questionnaires were returned.

Based on the pre-study data, this study conducted a reliability test by calculating Cronbach’s α coefficient. The results showed that the Cronbach’s α coefficients of each variable and the total scale were larger than 0.8000, denoting that the scale had a high level of reliability. The validity analysis used the principal component analysis framework in the exploratory factor analysis method, and the factors were extracted by the rotation with the largest variance, and the selection standard of the factors was the eigenvalue greater than 1. The outcome showed that the structure of the eight factors was stable, there was no cross-factorial situation, and the loadings of each factor were all greater than 0.5000, reliant on the scale of who? Wen, Ma, & Lyu, (2024), which pointed out that the validity level of the scale was good. In summary, the questionnaire used in the pre-survey can be used in the formal studies.

In this article, a total of 576 questionnaires were returned by means of electronic questionnaires distributed over the Internet, and the returned questionnaires are cleaned and rejected according to the following two basic principles (The questionnaire question type is in Chinese) (Wen et al., 2024). Firstly, whether the IP addresses of the questionnaires are the same to screen for duplicate responses. Secondly, whether the answers to similar questions are the same to screen for false responses. In the end, 529 qualified questionnaires are retained, with a validity rate of 91.84%.

The sample demographics indicate that 41.67% are male and 58.33% are female. Most participants are aged 22–34 years (63.72%). The majority are university undergraduates (50.35%), with school students (48.61%) and office workers (27.43%) being the primary occupations. Exposure to CBEB platforms ranged mainly from 1 year to 3 years (43.23%). The most frequently used virtual communities are Little Red Book (35.07%), Weibo (32.29%), Fubu foreign trade forum (21.88%), Amazon 1688 (19.27%) and Hugo.com (18.06%), with other platforms being used by less than 5% of respondents.

To control for homogeneous variance among data subjects from the same population, this study employed both procedural and statistical controls. Procedurally, questionnaire randomisation and psychological isolation methods were used. Statistically, factor analysis with added latent factors was conducted to test for homoscedasticity bias. Table 2’s Panel A shows that model indicators remained stable after including unmeasured latent factors, indicating that serious homoscedastic bias was absent.

For the problem of multicollinearity in the regression equation, it is mainly judged by the variance inflation factor (VIF) and the tolerance. From the table, it can be seen that the tolerance and VIF values of these four factors are close to 1, and the VIF value is less than 10, so it can be considered that the multicollinearity between these variables is not serious. The specific covariance test is shown in Table 2’s Panel B.

In this study, Cronbach’s alpha coefficient was used to assess the scale’s reliability and internal consistency. The reliability coefficient ranges from 0 to 1, with values closer to 1 indicating better reliability. A value below 0.7000 suggests poor reliability (Liu et al., 2025). The scale, consisting of six dimensions and 33 items, achieved a total Cronbach’s alpha of 0.8910, indicating high reliability. Each subscale also had reliability coefficients above 0.7000, confirming robust reliability (Wen et al., 2024). Overall, the scales demonstrated strong reliability and high internal consistency for all measured variables.

SPSS 28.0 was used to conduct the KMO and Bartlett’s test of sphericity to assess the data’s suitability for factor analysis. The overall KMO value was 0.851, and Bartlett’s test showed a chi-square significance level of 0.000, which is below the 0.050 threshold, rejecting the null hypothesis. Each subscale met the KMO criterion with values exceeding 0.700 and Bartlett’s test significance levels under 0.050 (Wen et al., 2024). These results indicate that the data are suitable for factor analysis and qualifies for further validation.

AMOS 28.0 was used to validate the questionnaire for convergent and discriminant validity. The standardised factor loadings for all six factors exceeded 0.500, and the composite reliability (CR) values were all above 0.800, with personal inclination and perceived usefulness surpassing 0.9000, indicating strong CR (Wen et al., 2024). The average variance extracted (AVE) values ranged from 0.532 to 0.643, exceeding the threshold of 0.500. These results confirm that the scales exhibit excellent convergent validity, as detailed in Table 3’s Panel A.

Discriminant validity was assessed by comparing each variable’s correlation coefficient with the square root of its AVE value. If the square root of the AVE for each variable exceeds the correlation coefficients with other variables, the scale demonstrates good discriminant validity. Analysis revealed that the square root of the AVE for each variable was indeed greater than its correlation coefficients with other variables, indicating strong discriminant validity and effective differentiation between dimensions. Details are presented in Table 3’s Panel B.

Following the reliability and validity tests, this article uses an SEM approach for modelling and hypothesis testing. Figure 2 shows the structural equation model.

Before conducting path analysis, the goodness-of-fit of the model needs to be tested, and this article is based on the discriminant criteria for the goodness-of-fit of structural equation models compiled by Ying et al (2025) and Zhou et al (2025).

In absolute fit metrics, RMSEA denotes the square root of the sum of squares of the asymptotic residuals and is usually used to measure the degree of model misfit. The results of the SEM analysis showed that the RMSEA in this study was 0.0330 less than 0.0800, and RMR, which denotes the fitted residuals, was 0.0250, which is less than 0.0500, indicating that the model was well fitted. CMIN/df denotes the degree of fit of the model’s covariance matrix to the observed data, and when the value is between 1 and 3, it indicates that the fit is very good. The CMIN/df value of this study is 1.5860, which is a good fit. In the relative fit indices, NFI is 0.9500, RFI is 0.9430, TLI is 0.9780, CFI is 0.9810 and IFI is 0.9810, and it is usually considered that NFI, RFI, TLI, CFI and IFI greater than 0.9000 indicate goodness-of-fit, and values greater than 0.7000 indicate acceptability. In addition, when the model was subjected to the validation factor in the previous section, it can be seen that the combined reliability of the latent variables is all greater than 0.7000, and the mean variance extracted are all greater than 0.5000, which indicates that the model has an intrinsic goodness-of-fit. It can be seen that all the indicators meet the criteria for the goodness-of-fit of the model, indicating that the model is well adapted, and hypothesis testing can be carried out accordingly.

The article uses the maximum likelihood method to estimate path parameters, with results summarised in Table 4’s Panel A. Five out of six hypothesised paths are confirmed, except for H3. The findings demonstrate that online reviews, platform market institution trust and shopping efficiency significantly impact platform sellers’ perceived reputation. Specifically, online reviews influence both platform market institution trust and shopping efficiency, which in turn affect the sellers’ reputation. However, H3, which posits that online reviews impact shopping efficiency, was not supported. This may be because of the competitive landscape of China’s CBEB platforms, where high-quality products and reliable logistics services reduce consumers’ attention to shopping efficiency. At the same time, consumers pay more attention to individual sellers and place greater emphasis on personalised experiences and overall service quality. In addition, the dynamics of the competitive landscape and the adaptability of consumer behaviour also make the impact of shopping efficiency on seller reputation not significant. This phenomenon may be temporary; as the market environment changes and consumer demands further develop, shopping efficiency may once again become an important factor influencing seller reputation in the future.

Further results suggest that the effect of online reviews on the perceived reputation of platform sellers is mediated through three mediators: an independent mediator through the platform market institution trust, an independent mediator through perceptions of platform shopping efficacy and a chain mediator through the platform market institution trust and perceptions of platform shopping efficacy. In this study, platform market institution trust and platform shopping efficacy mediate the effect of virtual community online reviews on the perceived reputation of platform sellers, respectively.

As shown in Table 4’s Panel B, the mediating effect was further tested using the bias-corrected nonparametric percentile Bootstrap method. The results show that the mediating effect of platform market institution trust and platform shopping efficacy is significant, with a mediating effect value of 0.3590. Specifically, the mediating effect emerges via three chains of mediation. Firstly, the indirect effect lnd1 (0.1939) consisting of online reviews → platform market institution trust → platform sellers’ perceived reputations, with a Bootstrap 95% confidence interval that does not contain 0, indicates that the mediating effect of platform market institution trust has a significant mediating effect. Secondly, the indirect effect lnd3 (0.0330) consisting of online comments → platform market institution trust → platform shopping efficacy → platform sellers’ perceived reputation, with Bootstrap 95% confidence intervals not containing 0, suggests that the chain mediating role of platform market institution trust and platform shopping efficacy between online comments in the virtual community and platform sellers’ perceived reputation is significant. Thirdly, the indirect effect lnd2 (0.1321) consisting of online reviews → platform shopping efficacy → platform sellers’ perceived reputation, with Bootstrap 95% confidence intervals not containing 0, suggests that the mediating role of platform shopping efficacy is significant. The specific path of online reviews acting on platform sellers’ perceived reputation is shown in Figure 3.

As shown in Table 5, mediation effect analyses were conducted using the SPSS macro program Process, prepared by Hayes, to analyse the mediating role of platform market institution trust and platform shopping efficacy in the relationship between online reviews in virtual communities and perceived reputation of platform sellers, controlling for gender and age.

Regression analyses showed that online comments had a direct positive predictive effect on platform market institution trust (β = 0.8863, p < 0.01) and platform shopping efficacy (β = 0.7201, p < 0.01). Platform market institution trust had a direct positive predictive effect on sense of platform shopping efficacy (β = 0.2029, p < 0.01). When online reviews, platform market institution trust and platform shopping efficacy simultaneously predicted platform sellers’ perceived reputation, online reviews, platform market institution trust and platform shopping efficacy all had a significant positive predictive effect on platform sellers’ perceived reputation (β = 0.5504, p < 0.01; β = 0.2188, p < 0.01; β = 0.1834, p < 0.01).

In this study, consumers were assigned a value of 1 for browsing online reviews and –1 for not browsing online reviews, and customer stickiness (divided into access stickiness and purchase stickiness) was centred. To test the moderating role of customer stickiness between online reviews and platform sellers’ perceived reputation in virtual communities, stepwise regression analyses were conducted with online reviews, customer stickiness and the product of online reviews and customer stickiness as the independent variables, and platform sellers’ perceived reputation as the dependent variable, respectively. The specific results of the analyses are shown in Table 6.

For CBEB platforms, the following specific measures can be taken based on the outcome of this article to effectively enhance the reputation of platform sellers.

Firstly, CBEB platforms should implement measures to incentivise users to write reviews, such as offering virtual badges, points or monetary rewards, while preventing manipulation of the review system to maintain credibility. Secondly, to improve accessibility and avoid information overload, platforms could cap review word counts and encourage image-based reviews, alongside developing guidelines for detailed and relevant feedback. Thirdly, incentivising sellers to positively respond to reviews through an after-sales service attitude points mechanism can enhance decision-making for potential users. Fourthly, platforms should focus on increasing visit and purchase stickiness by optimising website experiences, providing valuable content, and personalising recommendations. Finally, high-reputation platforms should refine reviewer information systems to attract consumers, while lower-reputation platforms should focus on improving product quality and service to build trust rather than relying on review systems alone.