The Role of Electronic Word of Mouth on Customer Purchase Intention in Social Media Instagram

Abstract


INTRODUCTION
Indonesia is the country with the fourth most Instagram users in the world, as quoted from tirto.id, the discount portal CupoNation released that as many as 56 million Indonesians use the Instagram social media service.As a new commercial avenue for social commerce, Instagram has attracted the attention of marketers.Among the main social media networks, marketers' use of Instagram has climbed from 36 to 44 percent since 2015, according to a 2016 social media marketing industry research (Subiyakto et al., 2018).
The growth of the Internet also offers a promising venue for electronic word of mouth (eWOM).People are starting to use web 2.0 (e.g., online forums, web blogs, social media) as a place to exchange information and opinions about a product, experiences in using sharing goods or services on social media with friends and acquaintances (Cakim, 2009).Electronic word-of-mouth is referred to as a type of marketing communication that includes statements made by current or former customers regarding a company or product that are made available to numerous individuals or organizations via the Internet (Croasmun & Ostrom, 2011).Some authors consider eWOM as a virtual development of traditional word of mouth (Sen & Lerman, 2007).Although comparable to wordof-mouth (WOM), electronic word-of-mouth (eWOM) can offer fresh options for discreetly and anonymously disseminating information and can also transcend geographical borders (Charo et al., 2015).The act of consumers sharing information about companies, products, and services with other consumers in a non-commercial way is known as word of mouth.Word of mouth is considered one of the most influential communication tools between message senders and receivers, more reliable and convincing than conventional media like print ads, personal selling, radio ads, and TV ads (Abdillah, 2018;Sangadji & Sopiah, 2010).
The differences between WOM and eWOM are in several aspects including that information shared in traditional WOM occurs in a small group of individuals or in private conversations, making it difficult to pass on information to people who are not present when WOM occurs (Goyette, 2010).Whereas eWOM is disseminated virtually, so information spreads very quickly and reviews are shared with more people (Mustakini, 2009), and it doesn't need to happen at the same time when everyone is present in the conversation (Chatterjee, 2001).Moreover consumer reviews on the Internet are available for a long time from which a large number of users can access them (Roscoe, 1975).Nonetheless, in terms of employee performance, the Employee Engagement Index (EEI) data is at a score of 4.1, or the same as the accomplishment in 2019, according to the annual sustainable reports produced by the company in 2020 and 2021 (Cheung et al.,2008).There was no discernible improvement in the Employee Engagement Index score (EEI).Yet, on a scale of 1 to 5, this value shows that employees at PT. IKAD appreciate their work environment and feel a strong sense of loyalty to the business.The goals of this study are to understand and analyze the effects of organizational commitment, job satisfaction, and remuneration on employee performance as well as the effects of compensation and organizational commitment on employee performance based on the background information provided above.
Electronic word of mouth (eWOM) has undoubtedly become a powerful marketing strategy and is influential in shaping consumer attitudes towards a product or service.Since social media facilitates the dissemination of information to a large number of people, users can even share similar reviews that they agree.For this reason, consumers are increasingly using social media to obtain information about brands.Conversations on social media often refer to brands, therefore it naturally has an effect on consumers' purchase intentions.

LITERATURE REVIEW
Word of mouth (WOM) has been known for a long time with the term word of mouth communication which is used to exchange information so that it can be conveyed widely.The information disseminated can be in the form of news, invitations, and even recommendations for a product or service that aims to provide personalized and noncommercial information based on consumers' direct experience of the selected product or service.61% of consumers choose to buy the same brand that their friends buy, because WOM comparatively provides more accurate and reliable information.Advertising is starting to lose its appeal to consumers due to reliability issues.Meanwhile, Word of Mouth occurs in two people who interact to share information without any desire to sell products that have previously been used (Cheung et al., 2012).
The eWOM process begins when consumers recognize a need or want, and know the problem at hand.Consumers seek information from internal or external sources.Seeking opinions and online reviews from other customers has become a large part of purchasing behavior by social media users.Following an information search, eWOM now enables customers to compare alternatives for items or services.Moreover, eWOM lowers risks that have a significant influence on customer decisions at the stage of the purchase decision process where the final product or service, brand, and store decisions are made (Chevalier & Mayzlin, 2006).
Like traditional WOM, eWOM can be created by marketers as well as by consumers.Online forums and company-created websites that allow users to share their opinions with others by producing media assets like photographs, videos, or text are common places where marketers distribute information.eWOM platforms include blogs, shopping sites, forums, discussion boards, and social media websites.The three main concepts in shaping S-Commerce are Web 2.0 technologies, social media and eCommerce.S-Commerce, broadly speaking, refers to the use of Internet-based media to make it possible for people to market, sell, compare, curate, buy, and share products and services both online and off, as well as in their communities.Customers can share their information with friends and other customers in a more social and participatory setting thanks to s-commerce.Moreover, social objectives including networking, teamwork, and knowledge sharing are a focus of social commerce.Social media has altered how individuals interact with one another.Social media has also become very popular over the past few years.Millions of users have integrated these websites into their daily live.As many as 18% of Indonesians use the internet to access social media.There are five social media sites most frequently visited by Internet users in Indonesia, namely Facebook, Instagram, Youtube, Twitter, and LinkedIn.These social media look the same as one another.But apparently, each social media has its own advantages and functions.Some researchers categorize these websites as networking sites, video sharing sites, and microblogs (Chu & Kim, 2011).

METHODOLOGY
This research was conducted using a quantitative approach, where data was collected through a survey in the form of a questionnaire, and using statistical data analysis techniques.The research population is all residents of DKI Jakarta who are in the age range of 15-39 years.With the number of samples in this study after considering various sources was 100 people who were taken using purposive sampling technique.Observations were carried out for 10 months.
Data collection was carried out in several stages, namely in the form of literature studies and questionnaires distributed online using whatsapp, twitter, and LINE social media, where the Google Form feature was used for making questionnaires.This questionnaire was distributed to active Instagram users.The data analysis process is divided into two, namely demographic analysis and static analysis.In demographic analysis, respondents' data were grouped based on gender, age, frequency of using Instagram social media, frequency of doing onlineshopping on Instagram, frequency of reading consumer reviews on Instagram social media, and length of time using the Internet.While in statistical analysis, researchers conducted inferential statistical analysis with the PLS-SEM approach using the SmartPLS 3.0 tool by conducting two stages of analysis, namely outer model and inner model.

RESULTS AND DISCUSSION
The highest respondents were 52% women and the lowest choice was 48% male respondents.For age, the highest choice is 19-24 years as much as 61% and the lowest is the age range of 30-34 and 35-39 as much as 1%.Then, 94% of respondents chose the use of the internet for more than 6 years, and the smallest choice fell to a span of 1-3 years by 2%.Respondents who are active users of Instagram as many as 96% chose to use Instagram social media every day, and the lowest choice is to use Instagram 4-5 times a week as much as 4%.The most choice in the category of online-shopping on Instagram is online-shopping more than once in one month as much as 38% while the smallest choice with a percentage of 1% is never doing online-shopping on Instagram.The last category is reading consumer reviews on Instagram, the highest percentage of choice with 65% is the choice that states that respondents always read consumer reviews on Instagram, on the contrary 35% of respondents stated that quite often read consumer reviews on Instagram about certain products.
From the 100 data collected, the results of the analysis show that 34 people very rarely do online-shopping on Instagram, and only one person has never done online transactions through Instagram social media.However, 38 people actually do online shopping on Instagram more than once a month, and 27 others do online-shopping once a month.However, many added that online shopping is done based on referrals from sources such as Instagram or similar apps, on the other hand, they also use affiliates from some accounts but purchases are made on other apps.At this stage, the measurement model analysis is carried out which consists of the outer and inner models.Where the outer model measurement is assessed from the reliability and validity of each indicator block related to the latent variable.The outer model consists of four stages of testing, namely, indicator reliability, internal consistency reliability, convergent validity, discriminant validity.Then, in the second stage, an inner model measurement is carried out which consists of the coefficient of determination (R2), path coefficient (β), t-test through the bootstrapping method, effect size (f 2), predictive relevance (Q2).It was found that there were four indicators that had a value below 0.7, namely indicators IC1, PI2, RV1, and RV2 which were declared invalid.
At this stage, reliability testing is carried out by reviewing the composite reliability (CR) value with a value limit of 0.7.From the reliability test results, all variables have a Composite Reliability value of more than 0.7.So all variables in this study have been declared reliable and are eligible for use and continue at the next stage.Another outer model testing stage is to review the AVE value, or Average Variance Extracted.This number demonstrates the range or variety of manifest variables that latent constructs may accommodate.More than 0.5, which is the optimal AVE value, indicates good convergent validity.From the measurement results obtained, it is concluded that the RV variable has an AVE value of less than 0.5, while the other variables have met the requirements to continue to the next stage.Discriminant validity occurs if scores from two distinct instruments measuring two constructs that are supposed to be uncorrelated will indeed be uncorrelated.Discriminant validity tests can be JEMSI (Jurnal Ekonomi, Manajemen, dan Akuntansi) E-ISSN: 2579-5635, P-ISSN: 2460-5891 Volume 9 (2) April Tahun 2023, Hal 226-231.checked by cross loading or Fornell-Lacker's cross loading, namely by comparing the AVE roots for each construct.Model has sufficient discriminant validity if the root AVE for each construct is greater than the correlation between one construct and another.
The analysis found that the AVE value for the Information Adoption (IA) variable is higher than the relationship between other variables.However, if you look closely, the AVE value for the Review Valence (RV) variable is lower than the correlation between other variables.With values that do not meet the applicable provisions, the outer model analysis must be reviewed before proceeding to the inner model analysis stage.Taking into account the four indicators that do not meet the requirements with the loading value, which is lower than 0.7, as well as the AVE value below 0.5, a retest is carried out by removing these four indicators.The rule of thumb explains that indicators with loading measurement values in the range of 0.40 to 0.70 should be removed if they increase the composite reliability value.
Based on four stages of testing, namely Indicator Reliability, Internal Consistency Reliability, Convergent Validity, and Discriminant Validity, the model proposed in this study has met statistically good characteristics.With all variables having a loading value and composite reliability above 0.7, the AVE value is above the 0.5 threshold, which means that the latent variable can explain on average more than half of the variance of its indicators.In addition, the root AVE value in the Fornell-Lacker's Cross Loading is higher than the correlation value between other latent variables, which means that discriminant validity is good.
The conclusion that can be drawn is that the results of the above analysis have shown that the proposed model has met the requirements to proceed to the structural model analysis test stage (inner model).To ascertain the link between constructs that academics have proposed as being related, structural model measurement is used.At this stage of the analysis, six stages of testing were carried out, namely, path coefficient (β), coefficient of determination (R2), t-test with bootstrapping calculation, effect size (f2), predictive relevance (Q2), and relative impact (q2).Path coefficient is calculated to see the significance of the relationship between constructs.The path coefficient measurement has a threshold value above 0.1 this is to state that the path in question has an influence in the model.Based on the test results, there are 6 paths that have a value above 0.1, which means that they have a significant effect in the model under study.Meanwhile, the path between IQ and IU is not significant with the limit value of 0.075.Result of determination test explains that the coefficient of determination or R2 value of the IA, IU, and PI variables are both at a moderate level.With each value of 0.542 (54.2%), 0.440 (44%), and 0.577 (57.7%) which is below the value of 0.67 and above the value of 0.33.
Judging from the results of the T-test, there is one value of the relationship between the IQ and IU variables with a value of 0.704 which means less than 1.96, indicating that of the 7 hypotheses in this study, one is not accepted.In order to demonstrate that specific variables employed in a model have predictive relevance with other variables in the model with a measurement threshold value above zero, Q2 (predictive relevance) testing employs the blindfolding method.Based on the test results, the Q2 value is obtained at the threshold value of 0, then each variable has a predictive relationship.The q2 (Relative Impact) test uses blindfolding also to measure the relative influence of a link between the predictive of a particular variable and other variables, which has a threshold value of 0.02 for a small influence, 0.15 for a medium influence, and 0.35 for a large influence.

DISCUSSION
Referring to the results of the path coefficient analysis of Information Quality (IQ) on Information Usefulness (IU), the value obtained is 0.075 which is far below the value of 0.1, so the path from IQ to IU is not significant.In addition, the t-test results show that this hypothesis is rejected with evidence that the path value of IQ to IU does not meet the t-test value requirement of 1.96.Where Information Quality (IQ) affects Information Usefulness (IU).However, it does not rule out the possibility that users are unable to translate the quality of information received, so that the value in addition to the path coeffiecient and t-test values that support that IQ does not have a significant effect on IU, the effect size and relative impact of IQ on IU are also relatively small with a value of 0.005 and -0.003 respectively.
The path coefficient and t-test values indicate that Information Credibility (IC) has a significant effect on Information Usefulness (IU).It is at the specified threshold, although it does not have a strong effect.The t-test value of 2.493 indicates that the IC path to IU is acceptable in accordance.Therefore, Information Credibility (IC) has a significant effect on Information Usefulness (IU) with an effect size value of 0.055 which means that the effect is small.However, the IC path to IU is acceptable and has a significant effect.Social media users who are dominated by users aged 19-24 with a fairly frequent frequency and always read other consumer reviews on Instagram social media are likely to find out the credibility of the information received before using it as information needed to make purchasing decisions.
From the results of the previous inner model analysis.Shows that Need of Information (NI) or the need for information has a significant effect on Information Usefulness (IU) with a path coefficient value of 0.406 and a ttest value of 3.371, both of which are above the required value that must be had for the hypothesis to be accepted.The need for prospective customers or existing customers who want to make online purchases on Instagram social media for a product turns out to make the information that has been received into useful information before making a decision.In addition, the results of the effect size along with the relative impact of the Need of Information (NI) variable on Information Usefulness (IU) have a moderate influence, with values of 0.173 and 0.126 respectively.Therefore, the hypothesis stating that Need of Information has a significant effect on Information Usefulness.
Based on the results of the analysis, the Information Usefulness variable has a significant effect on the Information Adaption variable.The t-test value above 1.96, namely 10.974, proves that the IU path to IA is acceptable and has a significant effect.Although the result of the effect size value is -0.238 which means it has a small influence, this result is in line with previous research conducted by Erkan & Evans (2016) as well as Hussain et al. ( 2018) which proves that the results of customers who feel information about a product is useful, then indirectly, adopting useful information will affect purchasing decisions.
Based on the results of the analysis, the Review Quantity variable has a significant effect on the Purchase Intention variable.The t-test value above 1.96, namely 2.448, proves that the RQ path to PI is acceptable and has a significant effect.Previously, Erkan & Evans did not examine the variable quantity of consumer reviews, but Teng et al. examined the relationship between Review Quantity and Purchase Intention.Consumers feel that the large number of reviews written by other consumers on Instagram social media influences their purchasing decisions.Which means that many other consumers have made purchases, so this gives encouragement to potential customers that the product has high popularity, is used by many people, thereby reducing the risk which ultimately influences customer decisions to make decisions whether to make a purchase or not.
Information Adaption has a significant effect on the Purchase Intention variable.The t-test value for the IA path to PI is 4.481, which means that the path is accepted, with a small influence according to the effect size value of 0.066.The information that has been adopted has an influence on customer decisions in determining whether to continue purchasing a product or not.In this case, the customer gets information from Instagram social media, then the information needed is adopted as a consideration before making an online transaction.It can be concluded that customer buying interest is directly influenced by the information that has been adopted, as well as the number of reviews written by other consumers, as well as the types of reviews (positive and negative) written by other consumers on Instagram social media.

CONCLUSION
Based on the research results, it can be concluded that a total of 7 hypotheses were tested in this study.Six hypotheses have been accepted.Electronic word of mouth has significantly affects customer purchase intention on Instagram social media.The Information Quality variable does not significantly affect Information Usefulness even though the study conducted by Erkan & Evans shows that Information Quality is significantly related to consumer interest in buying certain products.The Need of Information variable has a big influence because the stage of a person before making an online purchase activity is to find information on the information needs of a product.In addition, the large number of users from 19-29 years old is proof that users do not just buy a product, but look for additional information before adopting the information to make a purchase.Meanwhile, Information Usefulness has little effect on Information Adoption because the information spread allows users to immediately adopt the information without the need to filter whether the information they get is useful or not.Based on these findings, it can be concluded that the research.