Usage of Social Media Platforms as a Mediating Variable in the Relationship Between e-WOM and Consumer Purchase Decisions: The Case of Retail Consumers of Saudi Arabia

H5a: Social media platform usage positively and significantly mediates the relationship between e-WOM information quality and consumer purchase decisions in the retail industry in Saudi Arabia.

H5b: Social media platform usage positively and significantly mediates the relationship between e-WOM information credibility and consumer purchase decisions in the retail industry in Saudi Arabia.

H5c: Social media platform usage positively and significantly mediates the relationship between e-WOM information usefulness and consumer purchase decisions in the retail industry in Saudi Arabia.

H5d: Social media platform usage positively and significantly mediates the relationship between e-WOM information adoption and consumer purchase decisions in the retail industry in Saudi Arabia.

2.4. Ethical Considerations

Ethical issues were considered during the research process to guarantee the protection of the rights and privacy of the participants. The study was voluntary, and the participants were briefed on the academic nature of the study prior to completing the questionnaire. All participants provided informed consent in an electronic format prior to data collection. The respondents were assured that their identity and consent would be kept confidential and used only for research purposes. Moreover, the participants were allowed to withdraw from the survey at any moment without repercussions. All ethical research principles regarding privacy, confidentiality, informed participation, and responsible data handling were followed.

3. METHODOLOGY

The current research uses a quantitative research design because the study seeks to empirically investigate the relationships between eWOM dimensions, social media platform usage, and consumer purchase decisions. As suggested by (Dykema et al., 2022), a structured questionnaire on a five-point Likert scale (strongly disagree to strongly agree) was used to collect primary data (Appendix A). The tool is subdivided into major constructs: Consumer Purchase Decisions, Information Quality adapted from (Erkan & Evans, 2018), Information Credibility adapted from (Prendergast et al., 2010), Information Usefulness was adapted from (Erkan & Evans, 2018), Information Adoption from (Cheung et al., 2009), and Social Media Platform Usage from (Leong et al., 2021). This design can be effectively used to test the hypothesis and extend the empirical application.

This study takes a theoretical approach that adheres to the Information Adoption Model (IAM) which posits that people decide to use information on the web predominantly based on perceived quality, credibility, usefulness, and adoption before making a behavioural intention. Correspondingly, in (Erkan & Evans, 2018), a similar set of dimensions was selected as directly influencing the information-processing aspects of IAM, with other dimensions such as trust, attitude, and involvement being viewed as antecedents or consequences. Moreover, as identified by (Leong et al., 2021), social media platforms are included as a mediator because of their dual role in both disseminating and actively shaping eWOM interpretation. To guarantee theoretical parsimony and consistency with the existing IAM–TRA integration, the model has targeted information-centric determinants.

The segmented population is the retail consumers of Saudi Arabia using social media, such as Instagram, Pinterest, and Facebook. A purposive sampling method was employed to gather data, which, in this case, is appropriate since the respondents were chosen according to particular traits that are important in the study area, as (Thomas et al., 2022) point out. Only consumers who had previous experience interacting with eWOM content were used in this case, which guaranteed the relevance and reliability of the responses.

The necessary sample size was calculated using G*Power analysis, with multiple regression, medium effect size (f2 = 0.15), power of 0.95 and 6 predictors, which gave the minimum necessary sample size of about 146 respondents, as suggested by (Kang, 2021) as well. To enhance the statistical rigor, 800 questionnaires were sent through the Internet, and 420 respondents returned the questionnaires, representing a response rate of 52%. Following data screening, 70 responses were eliminated because of missing values and outliers, and a final sample of 350 valid responses was achieved.

To control for selection bias, respondents were selected based on various demographic profiles and platforms. To evaluate non-response bias, early (n1 = 60) and late (n2 = 60) respondents were compared, and no statistically significant differences were found, as in the study of (Guyot et al., 2023). As recommended by (Castillo et al., 2026), the common method bias (CMB) was assessed through full collinearity VIF against a 3.3 benchmark value, where no single value was found to be higher than 3.3, as shown in Table 1, indicating no issue of CMB.

Table 1. Full collinearity VIF.

In addition, the reliability and validity of the measurement model were evaluated. According to (Mohd Dzin & Lay, 2021), reliability is the degree to which measurement items are consistent in measuring the intended construct and yield stable measurement results over observations. Reliability in PLS-SEM is frequently evaluated using Cronbach’s Alpha and Composite Reliability (CR). Higher values of these alpha and CR indicate greater consistency among the items within the measurement, as suggested by (Li & Lay, 2024). Validity is the degree of accuracy in measuring the concept measured by an instrument. Convergent validity was used to test the consistency of the indicators of the same construct by the Average Variance Extracted (AVE) and evaluate how much variance is shared between the indicators, as indicated in the findings of (Dash & Paul, 2021). If the AVE is above the recommended level, it indicates that the construct is valid enough to explain the variance of its indicators. In this case, consistency is defined as the coherence and uniformity of the answers to the items testing the same latent construct, such that the measure as a whole represents the same concept (Mohd Dzin & Lay, 2021). The measurement model is evaluated for consistency, reliability, and validity through Cronbach’s Alpha and Composite Reliability with a standard value of 0.7 and Average Variance Extracted (AVE), against a threshold of 0.5, as suggested by (Purwanto & Sudargini, 2021).

Partial Least Squares Structural Equation Modelling (PLS-SEM) was used to analyse the data, which is appropriate when the predictor has complex predictive models, as recommended by (Mohd Dzin & Lay, 2021) for discovering direct and indirect relationships. Structural model analysis was conducted to test the hypothesised relationships and determine the significance of the path coefficients. Partial Least Squares Structural Equation Modelling (PLS-SEM) was used instead of covariance-based SEM (CB-SEM) because the study does not seek confirmation of the theory. As recommended by (Sarstedt et al., 2021), PLS-SEM is more suitable for models involving multiple constructs, mediation paths, and exploratory relationships, especially where the data do not necessarily meet the multivariate normality requirements. Furthermore, PLS-SEM is more flexible in dealing with predictive research models in new contexts, such as Saudi digital consumer behaviour.

4. RESULTS

4.1. Demographic Profile

The states shown in Table 2 signify the demographic features of the study participants (N = 350). From these statistics, it can be observed that 57.14% of the participants were male and 42.86% were female. Among these participants, 34.29% were in the age brackets of 18-25 years, 22.86% were between 26-30 years, 17.14% were between 31-35 years and 14.29% were 40 years and above. In addition, 28.57% of the participants were high school graduates, 24.29% were college graduates, 31.43% were Masters, and 15.71% were Ph.D. In terms of monthly income of the study participants, most participants fell into the lower income categories, with 21.43% earning less than SAR 5,000 and 24.29% earning between SAR 5,000 and SAR 10,000. In contrast, 14.86% and 14.29% of participants reported monthly incomes of SAR 15,001–25,000 and SAR 35,000 or above, respectively. Regarding the occupation and profession of the participants, 25.71% of the study participants were students, 42.86% were employees in the corporate sector, and 17.14% were business owners or self-employed. Lastly, regarding social media usage for product search and reviews, 28.57% use Facebook, 15.71% use Pinterest, 24.29% use YouTube, and 31.43% depend on Instagram’s influencer E-WOM.

Table 2. Demographic profile analysis.

4.2. Measurement Model Assessment

In Table 3, the measurement model assessment is shown which reflects high factor loading which ranges between (0.821-0.935). In addition, the results reflect the establishment of internal consistency and reliability, as the Value of Cronbach’s alpha is found to be higher than alpha (α > 0.7). In addition, convergent validity was attained as the value of AVE was greater than 0.5 for all constructs of the measurement model.

Table 3. Measurement model evaluation.

4.3. Discriminant Validity

The discriminant validity which measures conceptual overlapping and separability among model constructs, was tested through the HTMT ratio, as per its standard threshold of 0.85 suggested by (Haji-Othman & Yusuff, 2022). The results in Table 4 indicate that all constructs have HTMT values below the standard threshold of 0.85 which ensures that discriminant validity is established and there is no issue of conceptual overlap.

Table 4. Discriminant validity.

4.4. Path Coefficient

The results of the path coefficient, as depicted in Table 5, offer robust support for the direct and mediation hypotheses. Information Adoption demonstrates a strong and significant positive predictive relationship with consumer purchase decisions (β= 0.741, p < 0.001), which supports H4. Other significant positive relationships were observed in Information Quality (β = 0.095, p = 0.015) and Information Credibility (β = 0.021, p < 0.021), validating H1 and H2. The relationship between Information Usefulness and consumer purchase decisions was found to be marginally significant (β = 0.100, p = 0.059), supporting H3 at the 10% level of significance. Moreover, the use of Social Media Platforms has a weak association with consumer purchase decisions (β = 0.013, p = 0.034), which supports H5.

Table 5. Path coefficient analysis.

Note: *: Significance at 10%; **: Significance at 5%; ***: Significance at 1%

In terms of mediation, all indirect relationships were positive and statistically significant, proving the partial mediating role of the use of Social Media Platforms. In particular, the adoption of Information (β = 0.002, p = 0.034), Information Credibility (β = 0.005, p = 0.018), Information Quality (β = 0.004, p = 0.025), and Information Usefulness (β = 0.003, p = 0.027) had a significant indirect role in purchase decisions which favours H5a, H5b, H5c, and H5d. This also implies partial mediation and supports the significance of social media in eWOM translation into consumer buying decisions.  The effect sizes revealed from the t-test results showed that there were varying effect sizes among the constructs. Information Adoption has a significant impact on consumer purchase decisions (f² = 1.341), surpassing the threshold set by Cohen’s for a large effect, specified in (Gignac & Szodorai, 2016), making it the strongest predictor. The credibility of information on social media platforms had a medium effect (f² = 0.166), and information quality (f² = 0.110) and information usefulness (f² = 0.045) had small effects. However, for information credibility on consumer purchase decisions, the effect was f² = 0.001, and for social media platform usage on consumer purchase decisions, the effect was f² = 0.000 which represents a small effect.

4.5. Model Explanatory Power

The explanatory power examined through the R-squared shows the percentage of variance in the target variables predicted by their predictors. As per the results depicted in Table 6, it can be said that 69% variance in consumer purchase decisions is predicted by E-WOM dimensions which is strong explanatory power. However, 47.7% of the variance in social media usage is predicted by the e-WOM dimensions which depicts moderate explanatory power.

Table 6. Model explanatory power.

4.6. Predictive Relevance

The results in Table 7 regarding predictive relevance demonstrate that the models have a high level of performance. Q2 of Social Media Platforms (0.458) and Consumer Purchase Decision (0.680) were confirmed to be highly predictive. In addition, the low values of RMSE (0.743; 0.572) and MAE (0.537; 0.421) indicate adequate or moderate predictive accuracy. Overall, the model has a strong out-of-sample predictive success, specifically with consumer purchase decisions.

Table 7. Predictive relevance.

5. DISCUSSION

The results of the current research affirm hypothesis H1 that the quality of information in E-WOM plays an important role in shaping consumer purchase decisions, as previous research has indicated, but also indicates contextual peculiarities. In agreement with (Erkan & Evans, 2018), high-quality eWOM, determined by detailed, relevant, and structured information, can be used to improve the decision-making process because. Similarly, according to the Information Adoption Model (IAM), high-quality eWOM also helps consumers process information in the central route suggested by (Akın & Öztürk, 2025). Nevertheless, this study shows that the quality of information has a significant direct relationship with purchase decisions, unlike in the case of (Al-Haddad et al., 2022), where the quality of information did not have a significant direct association. This increased dependency on curated online content, stemming from the high prevalence of smartphone usage and social media among consumers, may explain the stronger link in the Saudi retail context, as there are likely to be higher levels of digital literacy among the individuals in this study. Detailed and structured eWOM information can be more relevant when people evaluate products during purchase decision-making processes, as proposed by (Erkan & Evans, 2018) under these conditions.

Higher uncertainty avoidance can result in increased effects on Saudi retail consumers because they culturally give importance to detailed and transparent information. Industrially, the nature of competition among retailers in Saudi Arabia pushes consumers to be more critical when comparing alternatives, thus increasing the importance of information quality, as supported by the findings of (Leong et al., 2021). Theoretically, these results are generally consistent with the IAM, which posits that information quality corresponds to purchase decision tendencies but does not work through other channels. In practice, Saudi retailers need to invest in the production of the right, comprehensive, and properly structured online content to enable consumer trust and informed buying decisions in an increasingly digital market.

The findings support H2 and show that the credibility of information provided in E-WOM is positively and significantly related to consumer purchase decisions, confirming the literature that was in place, but also showing some contextual differences. This result is consistent with that of (Ngo et al., 2024), who pointed out that credibility is one of the primary factors of information adherence and ensuing behavioural consequences. However, this conflicts with (Akın & Öztürk, 2025), who found which credibility did not significantly affect eWOM engagement. This difference might be a consequence of the context in which Saudi consumers conduct their online business, where trust is a vital factor because of the fear of false information and online reviews.

Theoretically, the findings are also echoed in the Theory of Reasoned Action (TRA), in which believable information influences attitudes and subjective norms that ultimately determine behavioural intentions, as (Leong et al., 2021) also revealed. Conversely, research such as (Akın & Öztürk, 2025) in emerging economies like Indonesia shows that retail consumers are less sensitive to credibility, owing to varying degrees of digital scepticism. In online interactions, Saudi consumers also value trust and authenticity, which could reflect their general trust and authenticity values in social and cultural interactions.

These results suggest that Saudi retailers and policymakers should promote the transparency and authenticity of online reviews. Verification mechanisms and the development of trustworthy content can boost consumer confidence and improve the reputation of online retail settings.

The relationship between information usefulness and consumer purchase decisions was established to be positive and significant, supporting H3 and indicating congruency and variations with the existing literature. Although (Subekti & Nugroho, 2023) indicated a strong impact of usefulness on adoption and purchase intention, this study shows a direct relationship. This was also proposed by (Leong et al., 2021), who hypothesised that usefulness acts largely indirectly via information adoption; hence, the relatively less significant direct significance in this case.

This weak relationship suggests that the perceived sense of usefulness is not enough for Saudi consumers to make a purchase, even though it is considered one of the initial conditions for the evaluation process. In the IAM, this interpretation is valid and useful as part of the cognitive evaluation, which can be followed by information adoption before the acceptance of its message is reflected in some aspects of behaviour. In terms of IAM, (Nugroho & Subekti, 2023) suggested that usefulness is a middle cognitive assessment that does not necessarily translate into action, especially within complex digital settings. In addition, the wealth of online information in Saudi Arabia can decrease the ability to differentiate based on usefulness because most of the information presented is already viewed as sufficiently helpful. For Saudi retail practitioners, the findings imply that improving usefulness is not sufficient; it should be accompanied by credibility and engagement strategies to impact purchase decision-making in the retail industry.

In addition, the results indicate that consumer buying decisions are substantially related to information adoption, validating H4. This finding is in line with (Al-Haddad et al., 2022), who noted that information adoption is the strongest predictor of purchase intention. It also follows (Subekti & Nugroho, 2023), which strengthens the IAM’s basic assumption that adoption is the immediate precursor of behavioural outcomes.

The relatively strong association found in the Saudi context is related to high digital involvement and frequent exposure to peer generated content, in which the consumers seem to internalise and use online information in decision-making processes also highlighted by (Nugroho & Subekti, 2023) also. Adoption has a stronger influence than information quality or usefulness, which are measures of a more fundamental cognitive and behavioural commitment (Al-Haddad et al., 2022).

The results largely corroborate the IAM and confirm the sequential relationship between information evaluation, adoption, and behavioural tendencies toward purchase. This effect is further enhanced by cultural aspects, including collectivism and social influence, where consumers tend to accept information that conforms to group norms and experiences. Saudi retailers are required to employ strategies that encourage consumers to not only access information but also internalise and act on it. The increase in information adoption and eventual purchase decision can be greatly improved by making the information more interactive, personalised, and engaging to the user.

The mediating role of social media in the association between eWOM dimensions and consumer purchase decisions was affirmed, but with complex overtones validating H5. The results support the idea that social media is a strong mediator of the relationship of information quality, credibility, usefulness and adoption, which, to some extent, are consistent with (Nurhadi et al., 2023), who also focused on the role of social media as a mediator of brand awareness and the influence on the decision to purchase. In contrast to (Santy & Andriani, 2023), in this case, eWOM alone did not directly influence a purchase, which this study illustrates as showing that social media reinforces these relationships as an interactive and interpretative medium.

Similar to the integration of IAM and TRA, the findings are generally consistent in that the social media environment could enable information processing while reinforcing socially favoured behavioural tendencies. While IAM describes how the nature of information can lead to adoption, TRA describes the importance of social norms, which are enhanced by interactions with social media, as suggested by (Santy & Andriani, 2023). The large mediation values for all dimensions suggest that social media can serve as a tool for influencing how eWOM content is seen, interpreted, and attributed in relevance.

In practice, Saudi retailers can leverage social media platforms in various ways to facilitate user-generated content, interactions, and influencer engagement. Public institutions can also participate in programs to boost digital literacy and confidence in information environments. Public organisations can also facilitate digital literacy and control of online content to boost the level of trust. In general, the findings of this study highlight the importance of social media as a mediator in changing eWOM into consumer actions.