THE EFFECT OF ELECTRONIC WORD OF MOUTH AND SOCIAL MEDIA MARKETING ON THE PURCHASE DECISION OF BILLIONAIRE'S PROJECT PRODUCTS THROUGH PRODUCT QUALITY

: The purpose of this study was to analyze the effect of Electronic Word of Mouth and Social Media Marketing on the Purchase Decision of Billionaire's Project clothing with Product Quality as the Mediation variable. The research design used is quantitative research with associative methods, namely methods that are formulated to provide answers to problems that have a relationship or influence between 2 or more variables. The data used is primary data obtained through distributing questionnaires to 105 respondents with the criteria that they are buyers of Billionaire's Project products. The results of the study indicate that the Social Media Marketing variable has a direct or indirect effect on purchasing decisions. Meanwhile, the Electronic Word of Mouth variable has no direct effect on the Purchase Decision. However, through the Product Quality variable as a mediating variable, Electronic Word of Mouth has an effect on Purchase Decision.


Introduction
Technology and the internet are fast advancing in this era of globalization. This will undoubtedly result in cultural changes in the environment and in society. According to the Indonesian Internet Service Providers Association (APJII), there were 196 million internet users in Indonesia from 2019 to 2020, accounting for roughly 73 percent of the country's total population of 266 million people.  Performance, is the main operating characteristics of the purchased core product.  Features (additional features or characteristics), namely secondary or complementary characteristics.  Reliability (Reliability), which is less likely to be damaged or fail to use.  Conformance to Specifications, namely the extent to which the design and operating characteristics meet pre-determined standards.  Durability, which is related to how long the product can be used.  Serviceability, including speed, competition, convenience, then direspati; and handling complaints satisfactorily.  Aesthetics, namely the attractiveness of the product to the five senses.  Perceived Quality, namely the image and reputation of the product as well as the company's responsibility towards it.

Purchase Decision
Consumers have increasing wants and needs for goods and services. To meet these wants and needs, consumers must make purchases. However, making the decision to buy goods or services is not always easy, especially with the growing number of business people and technological advances that provide consumers with a variety of options to meet their wants and needs. The purchase decision, according to Kotler and Keller (2016), is the step of the purchase decision process where the final consumer purchases the product. The buying decision is divided into five stages:  Recognizing the Need. The buying process, according to Kotler and Keller (2016), begins when the buyer recognizes a problem or need through internal and external stimulation, whereas, according to Hakwins & Motherbaugh (2010), realizing a need is the result of a mismatch between the desired state and the actual situation, which is sufficient to arouse and activate the purchasing process. The real state is a person's current perception of feelings and events. Those who want to feel the current feeling or condition define the ideal state.  Information Search.
People begin to start an active information search or look for content to read at the next level, according to Kotler and Keller (2016). They call friends, search online, and visit businesses to see things in person. According to Kotler and Keller (2016), the primary sources of customer information are split into four categories: a. Personal i.e. family, friends, neighbors and relatives b. Commercial, namely advertising, website sales, intermediaries, packaging, displays. c. Public i.e. mass media, consumer rating organizations d. Experiential namely handling, researching and using the product.  Evaluation of Alternatives.
Customers try to meet their requirements. Both consumers are looking for certain benefits from product solutions. Third, customers consider each product to be a collection of attributes with varying abilities to give value.  Purchase Decision.
In the evaluation stage, consumers form preferences among the preferred brands and can also form intentions to buy the most preferred brand. At this stage the customer makes a decision to buy the most preferred brand.
 Repurchase. After the purchase process, consumers may experience dissonance from realizing that the product purchased is good and has advantages so that they want to buy it again. Figure 2 Research Model

Research Hypothesis
The hypothesis is a speculative answer to the topic given, and it must be verified with more full and supporting facts. The purpose of this study was to see how social media marketing, electronic word of mouth, and product quality influenced purchase decisions for Billionaire's Project clothes. The following is the formulation of the research hypothesis:

Research Method
The authors employed quantitative methodologies in this investigation. Quantitative methods, according to Sugiyono (2018), are based on the positivist philosophy and are used to examine specific populations or samples. Sampling techniques are generally done at random, data collection is done with research instruments, and data analysis is quantitative or statistical with the goal of testing hypotheses that have been established. set. The goal of this study is to investigate the relationship between research variables, such as Electronic Word of Mouth and Social Media Marketing variables, and product quality and purchase decision variables.
The population used in this study are followers of the Billionaire's Project. The sample appointed is those who have purchased products from the Billionaires Project. In this study, the analysis tool uses the Structural Equation Model (SEM), if the sample size is too large, the minimum representative sample size is determined (Hair, et al, 2010). A suitable number or sample size is suggested between 100-200 respondents or an estimated 5 to 10 times the number of indicators. Therefore, the number of samples is determined based on the results of the calculation of the smallest minimum sample.
Based on Hair, et al. (2010) there is no exact sample size, but it can be seen that the appropriate sample size is between 100-200 samples. Therefore, according to the SEM requirements, the minimum number of respondents in this research sample is 95 respondents and the largest number of samples is 190 respondents. However, the sample used in this study was 105 respondents. Collecting data using a questionnaire using a Likert scale as a measuring tool. Sugiyono (2018) states that the Likert scale is a scale that can be used to measure a person's attitudes, opinions and perceptions about a particular object or phenomenon. With a Likert scale, the variables to be measured are translated into variable indicators. The Likert scale consists of numbers 1 (strongly disagree) to 5 (strongly agree) for all variables.
The analysis technique is carried out using Structural Equation Modeling (SEM) based on components or variants (component based), which is popular with Partial Least Square (PLS) with the help of the SmartPLS program. This technique was chosen on the grounds that the PLS technique does not require many assumptions. The data does not have to be multivariate normally distributed and the number of samples does not have to be large. The research PLS technique applies two components to the causal model, namely the measurement model and the structural model. A complete model basically consists of a measurement model and a structural model. The measurement model is carried out to produce an assessment of the validity and discriminant validity, while the structural model is a model that describes the hypothesized relationships.

Results and Discussion
The results of data collection through questionnaires from 105 respondents have been grouped into the characteristics of gender, age, and last education. Of the 105 respondents, 61% or 64 respondents were women and 39% or 41 respondents were men. From these data it can be said that there are more women than men. From the age factor, it can be shown that 60% or 63 respondents aged between 21-25 years, 26.7% or 28 respondents aged between 26-30 years, 9.5% or 10 respondents aged between 16-20 years, 3 ,8% or 4 respondents aged between > 30 years. From these data, it can be said that there are more respondents in the productive age group, with a total of 105 respondents. In terms of education, the latest shows that 54.3% or 57 respondents are high school graduates or equivalent, and 45.7% or 48 are undergraduate graduates. From these data, it can be said that the number of respondents with the latest education is high school graduates or the equivalent, namely 57 respondents.  Based on the table above, the results of the convergent validity test are shown after reestimating the recalculation. It can be concluded that each indicator of the research variable has a factor loading value greater than 0.7. Based on this, it can be stated that all indicators have met the requirements of convergent validity and are declared valid according to and have high validity and according to standards. The next step is to see the score or value of the Average Variance Extracted (AVE).
In contrast to factor loading, which is the value owned by each indicator, while the Average Variance Extracted AVE is the value owned by each variable. The recommended AVE value must be greater than 0.50, each variable in this study can be seen in table 2 as follows:  (2022) Based on the table data above, all AVE values for each research variable are greater than 0.5, so this study is in accordance with Chin's recommendation in Ghozali (2020), which indicates that the minimum AVE limit has been met. Therefore, it can be concluded that the convergent validity based on the factor loading value and the AVE value has been met, thus all indicators in this study can be declared valid and suitable for further research..  (2022) The criteria for measuring cross loading are expected that each indicator variable has a higher loading value for each measured latent variable compared to indicators for other latent variables. Based on table 3 above, it can be seen that each value of the cross loadings variable indicator in the gray block has a higher loading value than the indicators for other variables, so it can be said that it has a good discriminant validity value.

Reliability Test
There are two methods of measuring the reliability of a construct with reflexive indicators, namely Cronbach's Alpha and Composite Reliability. If the composite reliability value of the variable is > 0.7 then the variable can be declared to meet composite reliability. It can be said to be reliable if a construct is if the value of Cronbach's alpha > 0.6. In this study, the results of the reliability test on Composite Realiability and Cronbach's Alpha are shown in Table 4. below: represent the amount of variance of the constructs described by the model (Chin et al, 1998 in Ghozali andLatan, 2020). In this study the results of R² are shown in Table 5 (2022) Based on the output results in Table 5 above, there is an R² value of the user satisfaction variable of 0.826 and 0.810 which means it is in the strong category. The value obtained explains that the variable Product Quality and Purchase Decision, and then the remaining 17.4% and 19% is influenced by other variables outside of this research variable.

Hypothesis Testing (T Test)
Furthermore, hypothesis testing is carried out by looking at the significance value to determine the effect between variables through the bootstrapping procedure in SmartPLS. The bootstrap procedure uses the entire original sample to re-sampling, then analyzed by looking at the value of t-statistics on the bootstrap path coefficient. The T-statistic in the new path coefficient indicates the level of significance in hypothesis testing. The t-statistic assessment criteria can be seen if the t-statistic > 1.96 at a significance level of p-value 0.05 (Hair et al, 2010).

H2: E-WoM has a significant effect on product quality.
The original sample value shows a positive number, then the t-statistic value is 2,952 > 1.96 and the P-Value is 0.003 < 0.05. From these results it was concluded that the E-WoM variable had a positive and significant effect on product quality, so based on this, H2 was accepted.
H3: Social Media Marketing has a significant effect on product quality. The original sample value shows a positive number then the t-statistic value is 4.889 > 1.96 and the P-Value is 0.000 < 0.05. From these results, it can be concluded that the Social Media Marketing variable has a positive and significant effect on product quality, so based on this, H3 is accepted.

H4: Social Media Marketing has a significant effect on Purchase Decision.
The original sample value shows a positive number, then the t-statistic value is 2.211 > 1.96 and the P-Value is 0.027 < 0.05. From these results it is concluded that Social Media Marketing variable has a positive and significant effect on Purchase Decision, so based on this H4 is accepted.
H5: Product quality has a significant effect on Purchase Decision.. The original sample value shows a positive number, then the t-statistic value is 4.081 > 1.96 and the P-Value is 0.000 < 0.05. From these results, it is concluded product quality variable has a positive and significant effect on Purchase Decision, so based on this, H5 is accepted.

Intervening Effect Test
To test the hypotheses H6 and H7, the Intervening Effect Test (mediation) was conducted. The mediating effect shows the interaction between the mediating variable and the exogenous variable in influencing the endogenous variable. The magnitude of the influence between constructs and the effect of interaction (intervening) is measured by the path coefficient value. This mediating effect can be seen in the Specific Indirect Effects menu on SmartPLS after bootstrapping. It is shown that the T-statistic value is 3.384 or 1.96 and the P-value is 0.001 or 0.05, based on this, H7 is accepted.

Conclusion
 Electronic Worth of Mouth has no effect on purchasing decisions, the original sample value shows a positive number, but the P-Value value of 0.402 > 0.05 means that the variable does not affect. So even though the Electronic Worth of Mouth written in the comments column is good, it does not necessarily increase the purchase decision of the Billionaire Project product.  Electronic Worth of Mouth has a positive and significant effect on product quality, the original sample value shows a positive number meaning that the influence of the variable is unidirectional, if the Electronic Worth of Mouth is good, it will improve the quality of the Billionaire Project product.  Social Media Marketing has a positive and significant effect on product quality, the original sample value shows a positive number meaning that the influence of the variable is in the same direction, if promotion in Social Media is increased, it will improve the quality of Billionaire Project products.  Social Media Marketing has a positive and significant effect on purchasing decisions.
The original sample value shows a positive number, meaning that the influence of the variable is in the same direction, the better the promotion on Social Media, the better the purchase decision on Billionaire Project products.  Product quality has a positive and significant effect on purchasing decisions, the original sample value shows a positive number meaning that the influence of the variable is in the same direction, the better the quality of the product, the better the purchase decision for Billionaire Project products.  Electronic Worth of Mouth has a positive and significant effect on purchasing decisions through product quality mediation, the original sample value shows a positive number. This shows that product quality mediates the relationship between the Electronic Worth of Mouth variable on the purchase decision of the Billionaire Project product.  Social Media Marketing has a positive and significant effect on product purchasing decisions through product quality mediation, the original sample value shows a positive number. This shows that product quality is an intermediary (mediation) between the relationship of Social Media Marketing variables to the decision to purchase Billionaire Project products.