Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3, 660-679 2025 Publisher: Learning Gate DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate © 2025 by the authors; licensee Learning Gate History: Received: 6 January 2025; Revised: 10 February 2025; Accepted: 14 February 2025; Published: 8 March 2025 * Correspondence: Alifa.putri@binus.ac.id Predicting the effects of EWOM and product innovation on purchase decision and the impact on customer loyalty Alifa Savira Putri1*, Ricardo Indra2, Indah Dirgahayu Putri Pertiwi3, Ulfira Salsabilla4 1,2,3,4Bina Nusantara University, Indonesia; Alifa.putri@binus.ac.id (A.S.P.) Ricardo.indra@binus.ac.id (R.I.) Indah.pertiwi@binus.ac.id (I.D.P.P.) Ulfira.salsabilla@binus.ac.id (U.S.) Abstract: Purpose: The dynamics of consumer behavior have undergone a transformative shift, especially in today’s digital age where most customers read online reviews before making purchase decisions. Online reviews are one example of the many examples of Electronic Word of Mouth (EWOM). However previous research states that not only does EWOM have an impact on purchasing decisions, but product innovation has also played a big role in purchasing decisions and purchase decision has an impact on customer loyalty. Therefore, this study aims to see how EWOM could affect purchase decisions by involving product innovation through online reviews of body care products Iswhite on the e-commerce Shopee. Research design, data, and methodology: This research adopts a positivist paradigm and employs a quantitative approach that is explanatory through a cross-sectional survey with 100 participants. The study utilizes theoretical approaches such as the Planned Behaviour Theory. After all the data was collected through Google Forms all the data was processed Statistical Package for Social Science (SPSS) software. Results: The result shows that the variables EWOM have an insignificant effect and the variables product innovation have a significant effect towards purchasing decisions and purchasing decisions has a significant effect towards on consumer loyalty. Conclusions: The results have shown that EWOM has an insignificant influence on purchasing decisions. Product innovation also has a significant influence on purchasing decisions and purchasing decision variables have a strong influence on consumer loyalty. Keywords: Customer Loyalty, E-Commerce, Electronic Word of Mouth, Product Innovation, Purchase Decision. 1. Introduction In recent decades, the world of communication has undergone dramatic changes. The current digital revolution has significantly altered the global communication landscape. This causes the communication revolution to indirectly combine old and new cultures, which then creates friction between social and new technology [1]. This has a significant impact on the realm of marketing communication, particularly the evolution of word-of-mouth (WOM) into electronic word-of-mouth (EWOM). According to Mothersbaugh and Hawkins [2] and Litvin, et al. [3] before the emergence of EWOM, there was word-of-mouth (WOM), which is a communication process by consumers to share information about opinions aimed at potential buyers to buy or avoid certain products, brands, and services. Kotler and Amstrong [4] say EWOM is the internet version of word-of-mouth advertising. Even mentioned in Gruen, et al. [5] EWOM is a communication tool to exchange information about products or services that have been used by consumers who do not know each other. EWOM is unique in that the source of information is independent and not bound by an institution [6]. This proves that the information generated by EWOM is the personal opinion of consumers who are independent and not bound by a particular company or institution. As a result, EWOM is a new phenomenon in which 661 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate communication technology plays a role in sharing experiences, reviews, and recommendations related to certain products or services. EWOM can occur in several conditions, such as when consumers post their reviews and opinions about a product or service on blogs, websites, social networks, discussion forums, or review features [7, 8]. Given these conditions, a number of phenomena encourage the occurrence of EWOM today. These include the prevalence of e-commerce and its features, such as the review feature, which encourage consumers to rely on e-commerce online platforms for product and service search, comparison, debate, and purchase. In this context, e-commerce is one of the places where EWOM can occur, where consumer reviews, comments, and recommendations are shared through e-commerce [9]. Shopee e- commerce is one of the Indonesian e-commerce sites that attracts consumers' attention. Data from Hootsuite 2023 shows that Shopee is the marketplace category e-commerce site with the most visitors in Indonesia, surpassing its main rival, Tokopedia [10]. Business people also take advantage of the massive market opportunity in Shopee e-commerce, using it as a medium to sell their products or services online. This is utilized by one of Indonesia's body care brands, Iswhite. Not only that, Iswhite has made an achievement as a beauty product brand with the Top Innovation Choice Award in 2024. Through the innovation of its flagship product, the Whitening Ultimate Body Mask, Iswhite managed to sell more than 300,000 pieces in the Shopee marketplace while getting the award as the number one best body mask in Indonesia. This phenomenon sparked positive conversations and led to the spread of EWOM through e-commerce on the Iswhite Shopee account. Around 15 thousand to 26 thousand reviews about Iswhite's Whitening Ultimate Body Mask are scattered in the review feature on the Iswhite Shopee account. Moreover, it's important to note that both a brand and a company consider purchasing decisions as an integral part of their business operations. This is based on previous research conducted by Kapferer and Laurent [11] and Adnan [12] which found that the pleasure of buying a product is one of the five dimensions used to determine the involvement of a product purchase decision. Not only that, Adnan (2019) also proved that purchasing decisions affect consumer loyalty. Through this phenomenon, it can be interpreted that consumers do not only stop at the time of the purchase decision, but there are indications that consumers have pleasure in buying products. The observation of positive conversations on the Shopee e-commerce review feature, coupled with Iswhite's notable achievements in product innovation, serves as the objective basis for this study. It aims to investigate how the EWOM phenomenon and product innovation can influence purchasing decisions, and how these decisions can impact consumer loyalty. This research marks a significant departure from previous studies. This research focuses on how EWOM and product innovation play an important role in purchasing decisions and how purchasing decisions influence consumer loyalty in the beauty industry, especially for body care products in e-commerce. While previous studies have primarily focused on the food or service industry through social media, this research delves deeper into the role of EWOM and product innovation in purchasing decisions. Not only that, this research also focuses more on the Theory of Planned Behavior perspective, which is different from previous studies that used Diffusion of Innovation Theory and Marketing Behavior Theory. Based on the explanation that has been given, it can be concluded that this study wants to show the effect of EWOM and product innovation on purchasing decisions and how they impact consumer loyalty. 2. Literature Review 2.1. Theory of Planned Behavior The use of theory in research primarily aims to help predict and explain the phenomena that occur in the study. The theory used in this study is planned behavior theory. Indeed, planned behavior theory (TPB) is derived from the theories of multiattribute attitude (TMA) and reasoned action (TRA), which were discovered by Ajzen and Fishbein [13]. This theory was designed in 1975–1980 to predict 662 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate volitional behavior or decisions while understanding the psychological supporting factors [14]. Initially, the TRA model assumed that all individual behaviors have the ability to control themselves by involving attitude, normative belief, and behavioral belief [15]. But over time, this theory developed and expanded, giving birth to the Theory of Planned Behavior (TPB). It is known that TPB theory develops into six variables, including behavioral beliefs, normative beliefs, control beliefs, attitude toward the behavior, subjective norms, and perceived behavioral control [16]. This theory aligns with our research because it provides a comprehensive foundation for the subject under investigation. Additionally, the variables provided by the Theory of Planned Behavior (TPB) align well, providing a solid foundation for understanding the phenomenon under investigation. Departing from this, this study wants to see whether factors such as EWOM and product innovation are factors that are influential enough in determining individual behavior in purchasing decisions and how purchasing decisions impact consumer loyalty. 2.2. Electronic Word of Mouth Electronic Word of Mouth (EWOM) refers to the word Word of Mouth (WOM), which has been used frequently for centuries [17]. According to Westbrook [18] WOM is an informal communication addressed to other consumers or potential consumers regarding the ownership, use, or characteristics of certain products and services. The development of the times encourages WOM to continue to evolve in social networks, which then turns into online word of mouth [19]. EWOM itself is defined as a form of informal communication conveyed by consumers through internet-based technology regarding the use or characteristics of certain products, services, or sellers [8]. Not only that, EWOM is also defined as a two-sided information exchange between consumers and communicators regarding the use of products and services via the internet [3]. In previous research, it was said that EWOM has changed the buying environment by allowing customers to access comparative evaluations of product attributes on social networks with just one click [20]. Meanwhile, Yildiz, et al. [17] define EWOM as an online exchange of information between customers. According to EWOM's explanation, it can be said that EWOM is a form of informal information sharing by consumers regarding the use of products or services online. Several studies have classified several dimensions in EWOM, including source credibility, source homophily, message credibility, message quality, receiver's characteristics, and website credibility [21]. 2.3. Product Innovation Product innovation plays an important role in a company. This is based on maintaining the value of product benefits for consumers [22]. With technological advances that are increasingly fast and instant, companies are faced with intense competition to continue to create attractive products with the best solutions and quality for consumers. This is consistent with Rayi and Aras [23], who demonstrate that product quality, product modification, and meeting market needs are elements of product innovation that influence purchasing decisions. Departing from this phenomenon, of course, makes product innovation also play an important role in the company's strategy for running its business. Product innovation itself can be defined as the creation of products that involve processes and marketing to improve business mechanisms Kotler and Amstrong [4]. Kotler and Keller [24] also refer to product innovation as the outcome of diverse processes merging and influencing each other, which can be measured by its occurrence, activities, and results. According to previous research, there are several dimensions to product innovation, including technological innovation, service innovation, new products, and market breakdown [25]. 2.4. Purchase Decision According to Kotler and Keller [24] a purchasing decision is an action that consumers take to purchase products or services. Consumers make purchasing decisions by searching, providing, using, 663 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate evaluating, and spending products or services to meet their needs [26, 27]. In determining purchasing decisions, there are several factors to consider, such as product selection, brand selection, channel selection, purchase time, purchase amount, and payment method [24]. 2.5. Consumer Loyalty Indeed, consumer loyalty is crucial for a company. This not only shows that customers buy the products and services used, but also serves as a long-term symbol for a company. Loyalty itself is a feeling of attraction or affection towards people, companies, products, or services [28, 29]. Consumer loyalty, however, is the emotional feeling of customers towards a particular brand that leads them to repurchase products or services that are used regularly Mollah [29]. Griffin [30] also defines loyalty as routine purchasing behavior based on decision-making units. In achieving consumer loyalty, there are two steps that must be fulfilled, among others: (1) the company must be able to meet the needs and desires of customers; and (2) the company must maintain a relationship with customers so that they can carry out repurchase activities for the product or service [4]. According to Kotler and Keller [24] consumer loyalty consists of three dimensions repeat purchases, retention, and referrals. 2.6. Research Methods and Materials This research uses an explanatory quantitative approach, where the use of hypotheses is a very important issue in testing the sample so that researchers can draw conclusions about the sample that has been determined. The method used in this research is the survey method, in which researchers systematically ask the same questions to large numbers of people and then record their answers [31]. Researchers use a cross-sectional survey, in which data collection is carried out on a sample of selected respondents and is carried out once [31]. The data to be processed in this study will be analyzed using statistical methods, with the help of SPSS (Statistical Package for Social Science) software. 2.7. Hypothesis Development Figure 1. Framework Model. 2.8. The Effect of Electronic Word of Mouth (e-WOM) on Purchasing Decisions at Shopee According to previous research by Pratminingsih, et al. [32] EWOM influences purchasing decisions. Handranata, et al. [33] also conveyed the same thing: consumer reviews on EWOM have a huge influence on purchasing decisions. So based on previous research, it can be temporarily assumed that EWOM has a significant influence on purchasing decisions. Therefore, the following hypothesis was formulated: 664 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate H1: Electronic word of mouth on Shopee Iswhite e-commerce has a significant influence on purchasing decisions. 2.9. The Effect of Iswhite Product Innovation on Purchasing Decisions at Shopee In research conducted by Rayi and Aras [23] significant involvement was found between product innovation and purchasing decisions. The same results were also proven in research conducted by Pulungan, et al. [34] which found that product innovation provides significant results on purchasing decisions. Seeing the results of some of these studies, it can be temporarily assumed that Iswhite product innovation in Shopee e-commerce has a significant influence on purchasing decisions. H2: In Shopee e-commerce, Iswhite product innovation has a significant influence on purchasing decisions. 2.10. The Effect of Purchasing Decisions on Consumer Loyalty for Iswhite Products at Shopee In research conducted by Hariyadi, et al. [35] also proves that Purchasing Decisions have a significant influence on Consumer Loyalty. On the other hand, Adnan [12] stated the same thing that purchasing decisions have a significant influence on consumer loyalty. Seeing some of the results of these studies, it can be temporarily assumed that the decision to purchase Iswhite products in Shopee e- commerce has a significant influence on consumer loyalty. H3: Purchasing decisions for Iswhite products in Shopee e-commerce have a significant influence on consumer loyalty. 2.11. Data Collection and Research Method The quantitative research method was chosen because researchers wanted to get more accurate data through numbers. Another method used in this research is the survey method, which is the process of collecting information, facts, and analyzing social data in a structured and detailed manner through questionnaire instruments [36]. The questionnaire to be used is a closed questionnaire, meaning that data collection is carried out by asking respondents questions and providing answers in the form of choices. The study selects a population based on specific characteristics: individuals who use and purchase Iswhite body mask products through Shopee e-commerce in 2023. Respondents are made up of three generations, namely generation X, the Millennial generation, and generation Z, with an age range of 12–59 years, and the sample size was 100 people. The questionnaire, compiled in the form of a Google Form, will be distributed via the customer's WhatsApp number, which the researcher can obtain from the company's internal information. The researcher will then contact the customer. This questionnaire will be distributed to 100 respondents according to predetermined criteria. The Statistical Package for Social Science (SPSS) software will process all the collected data. Data collection through questionnaires can be seen in Table 1. 665 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Table 1. Questionnaire. Variable Dimensions Item Questionnaire Source Electronic word of mouth Source credibility XA1 I feel that the information provided about iswhite body mask in the product review colums on shopee is accurate information Rani and Shivaprasad [37] XA2 I trust review about iswhite body mask using the reviewer’s real name rether than a pseudon ym XA3 I trust review if the reviewer has knowledge about iswhite body mask products Source homophily XA4 I have more trust in product reviews where the reviwer is the same age as me XA5 I have more trust in reviews where the reviewer is the same gender as me XA6 I have more trust in reviewr where the geographical residence is the same as mine Message credibility XA7 I think positive review about iswhite body mask products have a big influence on my purchasing decision XA8 I think negative reviews about iswhite body mask products have a big influence on my purchasing decision. XA9 I think the number of reviews on iswhite body mask products influences my purchasing decision XA10 I think reviews with high ratings or likes on iswhite body mask products at shopee influence my purchasing decisions Message quality XA11 I think reviews about siwhite body mask products that are emotional in nature influence my purchasing decisions XA12 I think reviews with positive or negative arguments about iswhite body mask products influence my purchasing decision XA13 I think reviews uploaded along with photos or vidios about iswhite body mask products influence my purchasing decisions Receiver’s characteristics XA14 I prefer to do semething that doesn’t require much consideration XA15 I only read reviews that contain information that is relevant to me XA16 I read all the reviews and decide which ones are relevant to me XA17 I have the intention to leave a similar review regarding iswhite body mask products Website credibility XA18 Reviews on shopee iswhite e-commerce influence my purchasing decision. XA19 The popularity of shopee iswhite e-commerce in providing review services influences my purchasing decision Product innovation Technological innovation XB1 I feel iswhite body mask products are superior to other brand. Shiau [25] XB2 I feel that the performance of iswhite body mask products has satisfactory technological quality XB3 I feel that the technology in is white body mask products is able to adapt to new technologies XB4 I feel that is white body mask products have high quality technology Services Innovation XB5 I feel that the serbis provided by is white through shopee e-commerce is very statisfying. XB6 I feel that the service provided by is white through the shopee e-commerce is very resposive 666 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate XB7 I feal that the service method provided by is white through the shopee e-commerce is very efficient XB8 I feel that the quality of service provided by is white through shopee e-commerce is very effective New products XB9 I think is white body mask products are unique and worth using XB10 I think is white body mask products are superior to similar products from other brands XB11 I think is white body mask product are different from other similar products XB12 I think is white body mask products launch the latesh innovations compared to similar products from other brands Market breakthrough XB13 I think is white body mask products attactcertain consumer groups XB14 I think Is white body mask products update their products to meet market needs Product selection Y1 I think the validity of the is white body mask product is a suitable product for me Kotler and Keller [24] Y2 I think the quality of the is white body mask product is what I need Brand selection Y3 I think is white is a brand that has superior product quality to other brands Y4 I think is white has a pretty good reputation Distribution selection Y5 I chose to buy is white body mask products through Shopee e-commerce Y6 I choose to buy is white body mask products through resellers Y7 I choose to buy is white body mask products offline Time of purchase Y8 I buy is white mask products when I need them Y9 I buy is white body mask products when they are out of stock Y10 I buy is white body mask product when there is a discount Payment method Y11 When purchasig body mask is white products, I prefer to use mobile banking/shopee pay/other online payment method Y12 When purchasing is white body mask products, I prefer to make transactions with debit cards Y13 When purchasing is white body mask products I prefer to transact with cash Customer loyalty Repeat Z1 I buy is white body mask products on a recurring basis Kotler and Keller [24] Z2 I buy the is white body mask products more than once Z3 I use is white body mask products on a recurring basis Z4 I use is white body mask products several times a week Z5 I use is white body mask products every day Retention Z6 I feel that I am already suitable and loyal to is white body mask product Z7 I feel that is white body mask products have become part of my life Referrals Z8 I would recommend is white body mask products to ather people Z9 I would give a rating or review about is white body mask products 667 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate 3. Result and Discussion 3.1. Respondent Attributes The total number of respondents collected in this research survey was approximately 100, consisting of 63 female respondents and 37 male respondents. The proportion of respondents consisted of Generation Z, Millennials, and Generation X respondents. The majority of respondents live in Jakarta and other major Indonesian cities such as Surabaya, Palembang, Bandung, and Yogyakarta. The number of responses submitted total 100, and each question on this form has been filled in according to the opinion of each respondent. In this study, respondents' characteristics will be classified based on gender, age, education, domicile, occupation, and Shopee E-Commerce users. The demographic characteristics of these respondents can be seen in Table 2. Table 2. Demographic Characteristics of Respondents. Category Classification No. Sample Percentage (%) Gender Male 37 37.0 Famale 63 63.0 Age 12-27 years old 59 59.0 28-43 years old 36 36.0 44-59 years old 5 5.0 Province DKI Jakarta 36 36.0 West Java 28 28.0 East Java 9 9.0 Central Java 8 8.0 Banten 8 8.0 South sumatra 5 5.0 South Kalimantan 2 2.0 Riau 2 2.0 Bali 2 2.0 Job State Civil apparatus 6 6.0 Private employee 49 49.0 Entrepreneur 20 20.0 Student 22 22.0 Others 3 3.0 Education SMA 30 30.0 S1 51 51.0 S2 18 18.0 S3 1 1.0 Shopee E-Commerce usang periode < 1 Years 12 12.0 1-2 Years 39 39.0 >2 Years 49 49.0 3.2. Validity and Reliability Analysis In doing validity, of course, it is closely related to how well a questionnaire is able to measure each item of the questionnaire itself. The following are the results of the validity test in Table 3. 668 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Table 3. Convergent validity test result. XA XB XC Y Z XA1 0.607 XA10 0.764 XA11 0.640 XA12 0.750 XA13 0.815 XA14 0.538 XA15 0.741 XA16 0.741 XA17 0.817 XA18 0.859 XA19 0.817 XA2 0.692 XA4 0.805 XA5 0.741 XA6 0.811 XA7 0.761 XA8 0.574 XB9 0.823 XA XB XC Y XB1 0.595 XB10 0.785 XB11 0.814 XB12 0.824 XB13 0.548 XB14 0.737 XB2 0.847 XB3 0.812 XB4 0.832 XB5 0.872 XB6 0.849 XB7 0.816 XB8 0.847 XB9 0.823 XA XB XC Y Z Y1 0.539 Y10 0.759 Y11 0.725 Y12 0.469 Y13 0.519 Y2 0.817 Y3 0.791 Y4 0.836 Y5 0.832 Y6 0.845 Y7 0.828 Y8 0.834 Y9 0.574 XA XB XC Y Z Z1 0.588 Z2 0.844 Z3 0.846 Z4 0.674 Z5 0.672 Z6 0.850 669 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Z7 0.827 Z8 0.842 Z9 0.811 The Average Variance Extracted (AVE) is a numerical measure often used to assess convergent validity. AVE indicates how much variance from the indicators is explained by the construct, compared to the total variance of those indicators. A higher AVE, typically above 0.50, suggests that the construct explains more than half of the variability of its indicators, providing strong evidence of good convergent validity. This value helps evaluate whether the construct can adequately capture the variations among the indicators linked to it. In addition, factor loadings are also essential in assessing convergent validity. Factor loadings measure how closely an indicator relates to the construct being measured. Higher factor loadings indicate that the indicator strongly correlates with the construct, signifying good convergent validity. Generally, a factor loading above 0.50 is considered sufficient for convergent validity, while a loading above 0.70 is regarded as strong evidence that the indicator reliably measures the construct. In this study, several criteria are typically used to assess convergent validity based on factor loadings and AVE. A minimum factor loading of 0.50 is necessary to ensure that each indicator significantly contributes to the measurement of the construct. Factor loadings above 0.70 further strengthen the validity of the indicator in measuring the same construct. An AVE above 0.50 also signals that the construct explains more than 50% of the variance of its indicators, providing evidence that the construct has good convergent validity. Additionally, the correlation among indicators measuring the same construct is expected to be positive and significant, meaning that these indicators move in the same direction when measuring the intended construct. Low or even negative correlations between these indicators may indicate issues with convergent validity. Analyzing the data reveals how the convergent validity of each construct is evaluated. 1. Indicators for Construct XA: Based on the provided table, the indicators measuring construct XA (XA1 to XA9) generally show good factor loadings, with most indicators having loadings above 0.50. For example, indicator XA1 has a loading of 0.764, indicating strong convergent validity, and XA9 has a loading of 0.770, further supporting its convergent validity. These results suggest that construct XA has sufficient convergent validity, as most of its indicators show significant factor loadings. 2. Indicators for Construct XB: For construct XB, the indicators show varying factor loadings. For example, indicator XB1 has a lower loading of 0.595, suggesting weaker convergent validity for this indicator. However, other indicators like XB9 have a loading of 0.823, indicating strong convergent validity. Based on these results, although some indicators may not be as strong in measuring construct XB, overall, the construct has acceptable convergent validity. 3. Indicators for Construct Y: Construct Y shows a wide range of correlations between indicators. For example, Y2 has a loading of 0.817, indicating good convergent validity. However, Y12 has a lower loading of 0.469, suggesting that this indicator may not measure construct Y as well. The discrepancies among these indicators imply that some may need to be reevaluated to improve the convergent validity of construct Y. 4. Indicators for Construct Z: Indicators for construct Z show varied correlations among themselves. For instance, indicator Z2 has a loading of 0.844, indicating excellent convergent validity, while Z1 has a lower loading of 0.588, suggesting potential issues with its validity. Despite some indicators having weaker convergent validity, overall, construct Z demonstrates sufficient convergent validity. 670 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Based on these results, most constructs in this study exhibit good convergent validity, as indicated by factor loadings above 0.50. However, some indicators have factor loadings below the recommended threshold, indicating that their convergent validity may need to be improved or reassessed. Ultimately, strong convergent validity ensures that each indicator accurately and consistently measures the intended construct, which is critical in construct-based research involving multiple indicators. To determine the correlation between items, a reliability measurement instrument using Cronbach's alpha is required. According to Dempster and Hanna [38] cronbach alpha helps present the average correlation between a group of items. Ideally, the ideal Cronbach Alpha value should be greater than 0.7, which means it shows that the items are reliable [38]. Table 4. Discriminant validity test result. XA XB Y Z XA1 0.607 0.588 0.590 0.521 XA10 0.764 0.670 0.696 0.616 XA11 0.640 0.402 0.521 0.437 XA12 0.750 0.504 0.578 0.497 XA13 0.815 0.560 0.682 0.541 XA14 0.575 0.357 0.302 0.268 XA15 0.538 0.443 0.393 0.348 XA16 0.761 0.504 0.641 0.507 XA17 0.830 0.683 0.744 0.631 XA18 0.719 0.552 0.588 0.480 XA19 0.817 0.628 0.723 0.613 XA2 0.692 0.575 0.639 0.514 XA3 0.760 0.610 0.698 0.556 XA4 0.731 0.623 0.670 0.558 XA5 0.734 0.527 0.601 0.498 XA6 0.761 0.655 0.699 0.590 XA7 0.761 0.668 0.721 0.614 XA8 0.782 0.621 0.711 0.583 XA9 0.763 0.670 0.691 0.616 XA XB Y Z XB1 0.410 0.595 0.477 0.448 XB10 0.619 0.785 0.717 0.717 XB11 0.610 0.738 0.752 0.684 XB12 0.627 0.824 0.756 0.654 XB13 0.613 0.737 0.677 0.657 XB14 0.613 0.737 0.677 0.657 XB2 0.659 0.847 0.748 0.639 XB3 0.640 0.737 0.728 0.647 XB4 0.667 0.832 0.754 0.639 XB5 0.660 0.818 0.762 0.665 XB6 0.672 0.844 0.774 0.679 XB7 0.702 0.823 0.728 0.623 XB8 0.702 0.823 0.728 0.623 XB9 0.702 0.823 0.728 0.623 XA XB Y Z Y1 0.655 0.679 0.726 0.603 Y10 0.453 0.412 0.448 0.373 Y11 0.453 0.412 0.448 0.373 Y12 0.453 0.412 0.448 0.373 Y13 0.470 0.447 0.431 0.368 Y2 0.714 0.801 0.816 0.797 Y3 0.743 0.789 0.838 0.720 671 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Y4 0.748 0.773 0.785 0.720 Y5 0.707 0.743 0.765 0.693 Y6 0.697 0.841 0.810 0.720 Y7 0.693 0.741 0.802 0.710 Y8 0.419 0.515 0.704 0.496 Y9 0.592 0.609 0.738 0.574 Z1 0.565 0.676 0.702 0.693 Z2 0.264 0.526 0.448 0.672 Z3 0.592 0.409 0.634 0.697 Z4 0.675 0.574 0.702 0.784 Z5 0.675 0.574 0.702 0.784 Z6 0.675 0.574 0.702 0.784 Z7 0.675 0.574 0.702 0.784 Z8 0.675 0.574 0.702 0.784 Z9 0.684 0.729 0.729 0.811 The correlation matrix presented provides an analysis of divergent validity among four latent constructs (XA, XB, Y, and Z), where the goal is to assess the distinctiveness of each construct from one another. Divergent validity is supported when the correlation between different constructs is low, indicating that the constructs are measuring different concepts. In the analysis, the correlations between XA and the other constructs vary, with lower values such as 0.357 (XA8 with XB) and 0.268 (XA14 with Z) supporting divergent validity, while higher correlations, such as XA19 with Z at 0.850, raise concerns about conceptual overlap. Similarly, XB shows a mix of moderate to high correlations with other constructs, with XB1 and Z at 0.448 supporting distinctiveness, but higher correlations like XB8 and Z at 0.850 suggest overlap. The same trend is seen in the correlations between Y and other constructs, where Y12 and Z at 0.326 indicate divergent validity, but higher correlations such as Y8 and Z at 0.850 suggest less distinctiveness. While some items demonstrate good divergent validity, several high inter-construct correlations, particularly between constructs like XA, Y, and Z, suggest potential overlap, requiring further refinement of the items. A confirmatory factor analysis (CFA) would be beneficial for identifying problematic items that may need revision to ensure the constructs are indeed distinct. Overall, while many items support divergent validity, certain high correlations indicate the need for a deeper investigation to confirm the distinctiveness of these constructs. A description of reliability can be seen in Table 4. Table 5. Reliability Test Result. Construct reliability and validity-overview Cronbach's Alpha Composite Reliability (rho_a) XA 0.947 0.954 XB 0.953 0.958 Y 0.924 0.940 Z 0.917 0.933 Cronbach's alpha is the most commonly used measure of reliability to assess internal consistency between items in a scale, producing values between 0 and 1. A value above 0.9, as seen in constructs XA (0.947), XB (0.953), Y (0.924), and Z (0.917), indicates a very high reliability. However, overly high values may suggest item redundancy. Composite Reliability (rho_a), on the other hand, offers a more accurate reliability measure as it takes into account each item's weight, particularly in models like Structural Equation Modeling (SEM). The rho a values above 0.9 for all constructs—XA (0.954), XB (0.958), Y (0.940), and Z (0.933)—demonstrate excellent internal consistency, similarly high to Cronbach's alpha but slightly more precise due to its sensitivity to item weights. The key difference between these two measures is their approach: Cronbach's alpha is more sensitive to the number of 672 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate items and does not consider item weights, whereas composite reliability provides a more stable and accurate estimate, especially in SEM models. This high reliability indicates that the instrument used in the study is highly dependable and likely to yield consistent results across repeated measurements. However, it's important to note that high reliability alone is insufficient to guarantee instrument quality; the instrument must also have good validity to ensure that the items are genuinely measuring the intended constructs. The normality test determines the final values of the variable of interest in the data under consideration [39]. This is done to show the minimal possible bias of the distributed data [39]. The value of the K-S test statistic is 0.062, which is used to determine the significance of the test results. An asymptotic (2-tailed) significance value of 0.200 indicates that there is insufficient evidence to reject the null hypothesis that the residual data comes from a normal distribution. This value is greater than 0.05, indicating that the residual data is normally distributed. A description of the normality test can be seen in Table 5. Table 6. F Test Result. XA XB Y Z XA 0.154 XB 1.812 Y 2.480 Z It allows us to determine the relative importance of predictors in our model, based on their contribution to the variance in the outcome variable. The F² value is used alongside other key metrics like path coefficients and R² values to provide a complete understanding of how different factors influence the dependent variable. For each predictor, the interpretation of the F² value is categorized into three ranges: small, medium, and large. Specifically, values below 0.02 indicate a very small effect, suggesting the predictor's contribution to the outcome is negligible. Values between 0.02 and 0.15 indicate a medium effect size, implying that the predictor plays a modest yet notable role in explaining the outcome. Finally, values above 0.15 represent a large effect, where the predictor significantly influences the dependent variable. In the context of our research: • XA (F² = 0.154 to Y): This represents a small effect size, implying that XA’s contribution to the dependent variable Y is minimal. Though it plays a role, XA is not a dominant factor in predicting Y. However, its effect is still notable enough to be considered as contributing to the model, falling within the medium effect range. • XB (F² = 1.812 to Y): XB exhibits a large effect size, indicating it has a strong and significant influence on Y. This suggests XB is a key predictor, and its contribution to explaining Y is much more substantial compared to XA. The high F² value for XB aligns with the high path coefficient, emphasizing its dominant role in the model. • Y (F² = 2.480 to Z): This represents an extremely large effect size, which indicates Y is a dominant predictor of Z. The high F² value suggests that Y explains most of the variability in Z. This is supported by the high R² value for Z, showing that the majority of Z’s variance is accounted for by Y. In summary, analyzing the F-square effect size gives us a detailed understanding of how each independent variable influences the dependent variable. Variables like XB and Y show significant contributions to their respective outcomes, whereas XA demonstrate smaller but still meaningful effects. Through this analysis, we can better assess the significance and relative strength of each predictor in shaping the dependent variables in model. 673 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Table 7. Coefficient path result. Path coefficients - mean, STDEV, T values, p values Original Sample (o) Sample mean (M) Standard deviation (STDEV) T statistics (Io/STDEVI) P values XA-> Y 0.135 0.132 0.080 1.693 0.091 XB-> Y 0.556 0.552 0.093 5976 0.000 Y-> Z 0.845 0.650 0.029 28.769 0.000 Path coefficients measure the strength of the direct relationship between variables in a structural model. These values range from -1 to 1, where values close to 1 or -1 indicate a very strong relationship, and values close to 0 indicate a weak or non-existent relationship. P-values are used to test the statistical significance of each path coefficient. If the p-value is below 0.05, the relationship is considered statistically significant. • XA -> Y (Coef = 0.135, P = 0.091): The coefficient of the path between XA and Y is 0.135, which indicates that there is a weak positive relationship between the two. However, a p-value of 0.091 is greater than 0.05, which suggests that this relationship is not statistically significant. That is, we cannot conclude with confidence that XA has a significant influence on Y. • XB -> Y (Coef = 0.556, P = 0.000): A coefficient of 0.556 indicates a strong positive association between XB and Y. A very low P-value (0.000) indicates that this relationship is statistically significant. This means that XB is a very strong predictor for Y, consistent with the also large F- square results for this path. • Y -> Z (Coef = 0.845, P = 0.000): A coefficient of 0.845 indicates a very strong relationship between Y and Z. A very low P-value (0.000) indicates that this relationship is statistically significant. This shows that Y is the dominant predictor of Z, corresponding to a high R² result for Z and a very large F-square value for this path. 3.3. Hypotheses Analysis Hypotheses play a very important role in research, a hypothesis test is used to test an idea or parameter in a population using sample data [40]. A description of the hypothesis analysis can be seen in Table 8. Table 8. Hypothesis test. Path coefficients - mean, STDEV, T values, p values Original Sample (o) Sample mean (M) Standard deviation (STDEV) T statistics (Io/STDEVI) P values XA-> Y 0.135 0.123 0.085 1.578 0.118 XB-> Y 0.556 0.531 0.089 6.228 0.000 Y-> Z 0.845 0.845 0.031 27.626 0.000 The closer the coefficient is to these limits, the stronger the relationship. A positive coefficient indicates a positive relationship, and a negative one suggests an inverse relationship. For the XA -> Y path, the coefficient of 0.135 indicates a weak positive relationship, implying a small increase in XA results in a slight increase in Y. In contrast, the XB -> Y path has a coefficient of 0.556, representing a strong positive relationship, where an increase in XB significantly boosts Y, making XB a strong predictor of Y in this model. Meanwhile, the Y -> Z path, with a coefficient of 0.845, shows a very strong relationship between Y and Z, meaning that increasing Y will greatly impact Z, underscoring Y's dominant influence over Z. 674 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate T-statistics assess the significance of these path coefficients, where values greater than 1.96 at a 5% significance level indicate that the relationships are statistically significant. For the XA -> Y path, a T- statistic of 1.578 suggests the relationship is not significant, while a T-value of 6.228 for the XB -> Y path clearly indicates that the relationship is highly significant. For Y -> Z, the T-statistic of 27.626 strongly confirms statistical significance. P-values measure the likelihood that the path coefficients are due to random chance. Values below 0.05 suggest statistical significance. The XA -> Y path, with a P- value of 0.118, indicates that the relationship is not significant. Lastly, the Y -> Z path, with a P-value of 0.000, reinforces that this relationship is highly significant, highlighting Y as a powerful predictor of Z. 675 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Figure 2. Hypotheses Testing. 4. Discussion The F² value is used alongside other key metrics like path coefficients and R² values to provide a complete understanding of how different factors influence the dependent variable. P-values are used to test the statistical significance of each path coefficient. If the p-value is below 0.05, the relationship is considered statistically significant. • XA -> Y (Coef = 0.135, P = 0.091): The coefficient of the path between XA and Y is 0.135, which indicates that there is a weak positive relationship between the two. 676 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate However, a p-value of 0.091 is greater than 0.05, which suggests that this relationship is not statistically significant. That is, we cannot conclude with confidence that XA has a significant influence on Y. • XB - > Y (Coef = 0.556, P = 0.000): A coefficient of 0.556 indicates a strong positive association between XB and Y. A very low P-value (0.000) indicates that this relationship is statistically significant. This means that XB is a very strong predictor for Y, consistent with the also large F-square results for this path. • Y -> Z (Coef = 0.845, P = 0.000): A coefficient of 0.845 indicates a very strong relationship between Y and Z. A very low P-value (0.000) indicates that this relationship is statistically significant. This shows that Y is the dominant predictor of Z, corresponding to a high R² result for Z and a very large F-square value for this path. In structural equation modeling (SEM), path coefficients serve as key indicators for understanding the strength of relationships between variables. They quantify the direct impact one variable has on another within a model, and values typically range between -1 and 1. Positive coefficients suggest a direct positive relationship, while negative values indicate an inverse relationship. The closer the coefficient is to 1 (or -1), the stronger the relationship. However, this numerical relationship alone doesn't determine the significance of the effect; the accompanying p-values provide that insight. A p- value below 0.05 is generally considered to indicate statistical significance, meaning that the relationship between the variables is unlikely to have occurred by chance. In the case of the relationship between XA and Y, the path coefficient of 0.135 indicates a weak positive relationship between the two variables. A value closer to 0 suggests that XA has a minimal direct influence on Y. However, the more critical issue is the associated p-value of 0.091. Since this is greater than the commonly accepted significance threshold of 0.05, the relationship between XA and Y is not statistically significant. In practical terms, this means that while there is a slight positive relationship between XA and Y, the evidence is not strong enough to conclude that XA has a meaningful impact on Y in the population being studied. This finding suggests that XA may not be a reliable predictor of Y. In real-world applications, variables like XA might be relevant but could require further refinement, such as including moderating variables or exploring other potential interactions, to better capture their effect on Y. This lack of statistical significance could also imply that the sample size was too small or that there is insufficient variance in the data for XA to demonstrate its true impact on Y. In contrast, the path between XB and Y presents a much different picture. With a path coefficient of 0.556, the model suggests that there is a strong positive relationship between XB and Y. A coefficient above 0.5 indicates that XB has a substantial influence on Y, suggesting that when XB increases, Y tends to increase as well. More importantly, the p-value of 0.000 indicates that this relationship is statistically significant. With a p-value far below the 0.05 threshold, we can conclude with high confidence that the relationship between XB and Y is not due to random chance. XB serves as a robust predictor of Y, and its effect is both significant and substantial. In practical applications, this strong association might imply that factors represented by XB—whether they are customer behavior patterns, market conditions, or other influences—are critical drivers of the outcome represented by Y. This result aligns with the large F² value previously discussed, which also showed that XB contributes a considerable amount to the variance explained in Y. Organizations or decision-makers could focus on improving or leveraging XB-related activities to directly influence the outcome Y. The path coefficient between Y and Z is the strongest among the relationships considered, with a value of 0.845. This near- perfect coefficient suggests a very strong positive relationship between Y and Z. As Y increases, Z increases as well, and the strength of this relationship is unusually high. Furthermore, the associated p-value of 0.000 reinforces the significance of this relationship. Given the p-value is well below the 0.05 cutoff, we can confidently say that Y is a dominant predictor of Z, and this relationship is statistically significant. This finding is consistent with both the high R² value for Z and the large F² value discussed earlier, both of which indicate that Y explains a significant portion of the variance in Z. In practice, this could mean that the factors represented by Y are critical for understanding and predicting Z. For example, if Y represents customer satisfaction and Z represents 677 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate customer loyalty, then improving satisfaction would have a profound impact on loyalty. This result suggests that efforts to influence Z should be heavily concentrated on improving Y, as Y appears to be the primary driver of changes in Z. Together, these path coefficients and p-values provide a comprehensive picture of the relationships in this model. The relationship between XA and Y is weak and not statistically significant, indicating that XA may not play a meaningful role in influencing Y, or at least that its effect is not robust in this context. On the other hand, XB has a strong, significant relationship with Y, suggesting that XB is a major predictor of the outcome variable. The relationship between Y and Z is even stronger, indicating that Y plays a crucial role in determining Z. These findings suggest that, for decision-makers or researchers, it would be more productive to focus efforts on variables XB and Y, as they have the most substantial and statistically significant effects in this model. XA, while exhibiting a positive relationship, does not appear to be as influential or reliable as a predictor. This might indicate that XA is a less relevant factor or that its role is contingent upon other conditions not captured in the current model. The insights derived from this analysis also provide avenues for further investigation. For XA, future research might consider whether additional moderating variables or alternative measurement techniques could reveal a more significant relationship. Alternatively, XA might simply be less relevant in the context of this specific study and could be deprioritized in favour of other predictors. For XB and Y, these findings support their role as critical factors in the model. Future research could explore the mechanisms by which XB exerts its influence on Y, potentially examining whether XB has direct or indirect effects on Z. Additionally, the strong relationship between Y and Z opens opportunities to explore other mediating factors or to generalize these findings across different contexts, industries, or populations. In conclusion, path coefficients and their accompanying p-values offer a robust method for understanding the relationships between variables in structural equation modelling. In this study, the data shows that XB and Y are the most influential predictors in the model, while XA exhibits a weaker and non-significant effect. These findings provide valuable insights for both theoretical development and practical applications, especially for improving outcomes related to variables Y and Z. 5. Conclusion The results of the data analysis have shown that eWOM has a positive but weak and insignificant effect on the purchasing decision. This indicates that the concept of eWOM remains relevant in the context of digital marketing, where positive reviews can help enhance product credibility and generate initial interest from consumers. Meanwhile Product innovation plays an important role in purchasing decisions, because the results of this study show that product innovation has a significant influence on purchasing decisions. Practically, these findings can provide guidance to marketers to focus more on developing product innovation in order to increase consumer loyalty to the Iswhite Body Care brand in online sales channels. On the other hand, purchasing decisions also have a significant influence on consumer loyalty. This statement confirms that purchasing decisions strongly influence the level of consumer loyalty to Iswhite Body Care products on the Shopee e-commerce platform. So, it can be concluded that both EWOM variable is not have an impact but product innovation variable have a strong influence on purchasing decisions, and purchasing decision variables have a strong influence on consumer loyalty. Transparency: The authors confirm that the manuscript is an honest, accurate, and transparent account of the study; that no vital features of the study have been omitted; and that any discrepancies from the study as planned have been explained. This study followed all ethical practices during writing. 678 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate Copyright: © 2025 by the authors. This open-access article is distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). References [1] B. Kovarik, Revolutions in communication: Media history from Gutenberg to the digital age. USA: Bloomsbury Publishing, 2015. [2] D. L. Mothersbaugh and D. I. Hawkins, Consumer behavior: Building marketing strategy, 13th ed. New York: McGraw- Hill, 2016. [3] S. W. Litvin, R. E. Goldsmith, and B. Pan, "Electronic word-of-mouth in hospitality and tourism management," Tourism Management, vol. 29, no. 3, pp. 458-468, 2008. https://doi.org/10.1016/j.tourman.2007.05.011 [4] P. Kotler and G. Amstrong, Principles of marketing, 15th ed. Boston, MA: Pearson, 2018. [5] T. W. Gruen, T. Osmonbekov, and A. J. Czaplewski, "eWOM: The impact of customer-to-customer online know-how exchange on customer value and loyalty," Journal of Business Research, vol. 59, no. 4, pp. 449-456, 2006. https://doi.org/https://doi.org/10.1016/j.jbusres.2005.10.004 [6] S. Sen and D. Lerman, "Why are you telling me this? An examination into negative consumer reviews on the web," Journal of Interactive Marketing, vol. 21, no. 4, pp. 76-94, 2007. https://doi.org/10.1002/dir.20090 [7] C. M. Cheung and D. R. Thadani, "The impact of electronic word-of-mouth communication: A literature analysis and integrative model," Decision Support Systems, vol. 54, no. 1, pp. 461-470, 2012. https://doi.org/https://doi.org/10.1016/j.dss.2012.06.008 [8] E. Ismagilova, Y. K. Dwivedi, E. Slade, and M. D. Williams, Electronic word-of-mouth (eWOM) BT - Electronic word of mouth (eWOM) in the marketing context: A state of the art analysis and future directions (E. Ismagilova, Y. K. Dwivedi, E. Slade, & M. D. Williams, Eds.). Cham: Springer International Publishing. https://doi.org/10.1007/978-3-319-52459- 7_3, 2017. [9] H. Siagian, Z. Tarigan, and S. Ubud, "The effect of electronic word of mouth on online customer loyalty through perceived ease of use and information sharing," International Journal of Data and Network Science, vol. 6, no. 4, pp. 1155–1168, 2022. https://doi.org/10.5267/j.ijdns.2022.7.004 [10] We Are Social, "Digital 2023. Retrieved from We Are Social website," Retrieved: https://www.hootsuite.com/resources/blog/social-trends-report-2023, 2023. [11] J.-N. Kapferer and G. Laurent, "Consumer involvement profiles: A new and practical approach to consumer involvement," Journal of Advertising Research, vol. 25, no. 6, pp. 48-56, 1985. [12] A. Adnan, "The effect of product quality, brand image, and advertising on purchase decision and its impact on customer loyalty of morinaga milk in aceh," International Journal of Innovation, Creativity and Change, vol. 9, no. 4, pp. 234–249, 2019. [13] I. Ajzen and M. Fishbein, "Attitudinal and normative variables as predictors of specific behavior," Journal of Personality and Social Psychology, vol. 27, no. 1, p. 41, 1973. https://doi.org/10.1037/h0034440 [14] J. Kuhl and J. Beckmann, Action control: From cognition to behavior. Berlin, Heidelberg: Springer Science & Business Media, 2012. [15] A. W. Kruglanski, E. T. Higgins, and P. A. M. Van Lange, Handbook of theories of social psychology. London: SAGE Publications, Limited, 2011. [16] I. Ajzen, Constructing a theory of planned behavior questionnaire. Amherst, MA: University of Massachusetts, 2006. [17] G. K. Yildiz, S. Yildiz, M. A. Yoruk, and S. Sevgen, "A tablet game or training booklet? Two methods for evaluating symptom management and quality-of-life of children receiving chemotherapy," European Journal of Oncology Nursing, vol. 61, p. 102190, 2022. https://doi.org/10.1016/j.ejon.2022.102190 [18] R. A. Westbrook, "Product/Consumption-based affective responses and postpurchase processes," Journal of Marketing Research, vol. 24, no. 3, pp. 258–270, 1987. https://doi.org/10.1177/002224378702400302 [19] W. Duan, B. Gu, and A. B. Whinston, "Do online reviews matter?—An empirical investigation of panel data," Decision Support Systems, vol. 45, no. 4, pp. 1007-1016., 2008. [20] P. R. Varadarajan and M. S. Yadav, "Marketing strategy and the internet: An organizing framework," Journal of the Academy of Marketing Science, vol. 30, no. 4, pp. 296–312, 2002. https://doi.org/10.1177/009207002236907 [21] A. Rani, M. Toni, and H. Shivaprasad, "Examining the effect of electronic word of mouth (eWOM) communication on purchase intention: A quantitative approach," Journal of Content, Community and Communication, vol. 15, no. 8, pp. 130- 146, 2022. https://doi.org/10.31620/JCCC.06.22/10 [22] S. Syam et al., "The influence of social media marketing, product innovation and market orientation on Indonesian SMEs marketing performance," International Journal of Data & Network Science, vol. 6, no. 1, pp. 9–16, 2022. https://doi.org/10.5267/j.ijdns.2021.11.002 [23] G. Rayi and M. Aras, "How product innovation and motivation drive purchase decision as consumer buying behavior," Journal of Distribution Science, vol. 19, no. 1, pp. 49-60, 2021. https://creativecommons.org/licenses/by/4.0/ https://doi.org/10.1016/j.tourman.2007.05.011 https://doi.org/https:/doi.org/10.1016/j.jbusres.2005.10.004 https://doi.org/10.1002/dir.20090 https://doi.org/https:/doi.org/10.1016/j.dss.2012.06.008 https://doi.org/10.1007/978-3-319-52459-7_3 https://doi.org/10.1007/978-3-319-52459-7_3 https://doi.org/10.5267/j.ijdns.2022.7.004 https://www.hootsuite.com/resources/blog/social-trends-report-2023 https://doi.org/10.1037/h0034440 https://doi.org/10.1016/j.ejon.2022.102190 https://doi.org/10.1177/002224378702400302 https://doi.org/10.1177/009207002236907 https://doi.org/10.31620/JCCC.06.22/10 https://doi.org/10.5267/j.ijdns.2021.11.002 679 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 3: 660-679, 2025 DOI: 10.55214/25768484.v9i3.5281 © 2025 by the authors; licensee Learning Gate [24] P. Kotler and K. L. Keller, Marketing management, 15th ed. Boston, MA: Pearson Education, 2016. [25] H.-C. Shiau, "The impact of product Innovation on behavior intention: The measurement of the mediating effect of the Brand Image of Japanese anime dolls," The Anthropologist, vol. 17, no. 3, pp. 777–788, 2014. https://doi.org/10.1080/09720073.2014.11891492 [26] R. D. Santy and R. Andriani, "Purchase decision in terms of content marketing and e-WOM on social media," Journal of Eastern European and Central Asian Research, vol. 10, no. 6, pp. 921-928, 2023. https://doi.org/10.15549/jeecar.v10i6.1502 [27] L. G. Schiffman and L. L. Kanuk, Consumer behavior, 8th ed. New Jersey: Prentice Hall, 2013. [28] T. O. Jones and W. E. Sasser, "Why satisfied customers defect," Harvard Business Review, vol. 73, no. 6, pp. 88-100, 1995. [29] A. S. Mollah, "The impact of relationship marketing on customer loyalty at Tesco Plc, UK," European Journal of Business and Management, vol. 6, no. 3, pp. 21-55, 2014. [30] J. Griffin, "The internet’s expanding role in building customer loyalty," Direct Marketing, vol. 59, no. 7, pp. 50–53, 1996. [31] L. Neuman, Social research methods: Qualitative and quantitative approaches. Harlow, England: Pearson Education Limited, 2014. [32] S. A. Pratminingsih, E. Astuty, and V. Mardiansyah, "The influence of electronic word of mouth and brand image on buying decision," Journal of Advanced Research in Dynamical & Control Systems, vol. 11, no. 3, pp. 61-67, 2019. https://repository.widyatama.ac.id/items/f70641ec-9c55-4a57-95aa-592346b14151 [33] Y. W. Handranata, M. G. Herlina, E. L. Tanaya, and R. Setiawan, "Out-of-Group or In-Group E-WOM affects Consumer Purchasing Decision?," in E3S Web of Conferences, 2023, vol. 426: EDP Sciences, p. 02117. [34] D. Pulungan, A. Fauzi, and E. S. Rini, "The effect of product innovation, consumer attidude, and advertising at the purchase decision of yamaha motorcycle in CV Sejati mitra motor medan Marelan," International Journal of Business and Management Invention, vol. 7, no. 6, pp. 42-50, 2018. [35] S. Hariyadi, A. Ningsih, and J. L. Away, "The effect of effect of store image and store location on purchase decision and customer loyalty of modern retail in the City of Samarinda," International Journal of Scientific & Technology Research, vol. 7, no. 9, pp. 66-72, 2018. [36] C. G. Thomas, Research methodology and scientific writing. Cham: Springer, 2021. [37] A. Rani and H. Shivaprasad, "An empirical analysis of receiver’s psychological characteristics in eWOM engagement," International Journal of Cyber Behavior, Psychology and Learning, vol. 12, no. 1, pp. 1-19, 2022. https://doi.org/10.4018/IJCBPL.298686 [38] M. Dempster and D. Hanna, Research methods in psychology for dummies. Hoboken, NJ: John Wiley & Sons, 2015. [39] J. P. Verma and A.-S. G. Abdel-Salam, Testing statistical assumptions in research Hoboken: John Wiley & Sons, 2019. [40] G. J. Privitera, Statistics for the behavioral sciences. Thousand Oaks, CA: Sage Publications, 2023. https://doi.org/10.1080/09720073.2014.11891492 https://doi.org/10.15549/jeecar.v10i6.1502 https://repository.widyatama.ac.id/items/f70641ec-9c55-4a57-95aa-592346b14151 https://doi.org/10.4018/IJCBPL.298686