The Influence of Electronic Word of Mouth on Purchase Intention of Compass Shoe Customers in Padang City with Brand Image as a Mediating Variable

The aim of this research is to analyze: (1) The influence of E-WOM on Purchase Intention for Compass Shoes in Padang City. (2) The influence of E-WOM on the Brand Image of Compass Shoes in Padang City. (3) Influence of Brand Image on Purchase Intention for Compass Shoes in Padang City. 4) Influence of E-WOM on Purchase Intention for Compass Shoes in Padang City with Brand Image as a mediating variable. The type of data in this research is quantitative, the data source is primary data through distributing questionnaires via Google Form, the sampling technique is purposive sampling technique, the data analysis tool is SEM analysis using the Smart PLS VERSION 4 application.Research Results: (1) There is a significant influence between E-WOM on Purchase Intention for Compass Shoes in Padang City. 2) There is a significant influence between E-WOM on the Brand Image of Compass Shoes in Padang City. 3) There is a significant influence between Brand Image on Purchase Intention for Compass Shoes in Padang City. 4) There is a significant influence between E-WOM on Purchase Intention for Compass Shoes in Padang City through brand image


INTRODUCTION
The rapid development of the world of fashion in Indonesia, especially shoes, has caused various kinds of local shoe brands to become known and in demand among young people, especially with the viral hashtag #localpride adding to the premium impression of the local shoe industry.This is an opportunity for manufacturers to produce shoes.The competitive local shoe industry is shown in the various product choices available on the market.Several large companies produce Compass, Converse, Vans, Nike, Puma shoes and others.Among the products that researchers want to discuss at this time, one of the research objects discussed is the Compass shoe product.Purchase intention will continue to become a purchasing decision when customers feel they have confidence in an item.The Purchase Intention problem experienced by Brand Compass is that there is still low buying interest compared to other products listed in the data and image above.According to Mohammed T. Nuseir (2019), the factors that influence Purchase Intention are Electronic Word Of Mouth and Brand Image as variables mediation.(Nuseir, 2019) Electronic word of mouth (eWOM) also encourages people to collect information about desired products and services which is not only collected through people they know but also from other people in a large circle who have used the product or service and can collectively encourage this intention.consumer buying (Tariq et al, 2017).

Brand
Brand image is consumers' understanding of the brand as a whole, consumers' trust in the brand and how consumers view the brand.A brand is something unique and attractive that every product must have and is also a characteristic of the product and what differentiates it from other products.The importance of brand image is that a product has a value, which will later be used as a perception by consumers.This is why the Brand Image must be maintained.

RESEARCH METHODS
The type of research is quantitative descriptive research, the data source is primary data, the data collection technique is via a questionnaire distributed via Google Form, the research was conducted in January 2024.

Outer Model Evaluation
Data The questionnaires were collected and filled in completely by 189 respondents which were then processed using PLS-SEM with smart PLS 4 software.The image shows the initial form of the indicator validity measurement model on its parent variable.For more details, see the initial outer loading structure output

Reliability Test
Reliability testing is used to measure the consistency of measuring instruments in measuring concepts, and can also be used to measure the consistency of respondents in answering questions in questionnaires or research instruments.Reliability, while Composite reliability measures the true value of the reliability of a construct.Rule of thumb alpha or composite reliability value is > 0.7, although a value of 0.6 is still acceptable (Hair et al., 2018).can demonstrate accuracy.

Convergent Validity After removing several indicators
According to JF Hair (2014:45), "Validity testing in This research uses two methods of measurement, namely convergent validity and discriminant validity."The convergent validity criteria are the average variance extracted (AVE) value > 0.5 and outer loading > 0.6."Draw a path diagram of the causal relationship between constructs and the loading factor values after removing several indicators or questions.The reason is because we want the data to be valid because previously there was a problem with the AVE value.

Picture1Image of Outer Model Source: Processed Smart PLS 4 (2023)
The figure shows the initial form of the indicator validity measurement model on its parent variable.For more details, see the outer loading output table for the initial structure below Based on Table 15 above, each variable shows an AVE value above 0.5, so the conclusion is that the data is valid.

Reliability Test
Reliability testing is used to measure the consistency of measuring instruments in measuring concepts, and can also be used to measure the consistency of respondents in answering questions in questionnaires or research instruments.Reliability, while Composite reliability measures the true value of the reliability of a construct.Rule of thumb alpha or composite reliability value is > 0.7, although a value of 0.6 is still acceptable (Hair et al., 2018) Based on Table 16 above, each variable shows a Cronbach Alpha value above 0.6 and an AVE value above 0.5, so the conclusion is that the data is reliable.

Inner Model Evaluation
The Structural Model Test or (inner model) aims to predict causal relationships (cause-effect relationships) between latent variables or variables that cannot be measured directly.In testing the structural model (inner model) using the help of the Bootstrapping procedure in SMARTPLS.Tests on the structural model were carried out by testing the relationship between latent constructs using R-Square for the dependent construct and the significance of the structural path parameter coefficients.The following are the results of R-Square estimation using SMARTPLS version 4. Based on table 17, the R-square value can only be found in endogenous constructs.It can be seen that the Rsquare value for Purchase Intention is 0.592, this result shows that 59.2% means that the E-WOM variable contributes to the Purchase Intention variable while the rest is influenced by variables outside this research.Furthermore, the Rsquare value for Brand Image is 0.544.This result shows that 54.4% means that the E-WOM variable contributes to the Brand Image variable while the rest is influenced by variables outside this research.

Hypothesis test
Hypothesis testing is carried out if the data meets the measurement requirements using the bootstrapping method in SmartPLS 4 software.Bootstrapping is a re-sampling method that allows freely distributed data to be applied so that it does not require the assumption of a normal distribution and a large sample size (Ghozali&Laten, 2012).In this study, a re-sample of 189 respondents was used using the No sign change scheme.Hypothesis testing can be seen from the results of the significance test, where the level of significance in this study is 5%.At a significance level of 5% or 0.05, it is considered "significant" and the hypothesis can be accepted if the T-statistic value is greater than 1.96 (Hair, 2013).
The direct influence hypothesis test was carried out by looking at the statistics produced by the structural model (Inner Model).The hypothesis can be accepted if the T-statistic is >1.96.The following are the results of the direct influence hypothesis test.The following is a direct hypothesis using SEM (Structural Evaluation Model) analysis using the Smart PLS 4 application which is in the following table.Based on table 16, the output path coefficient value shows that E-WOM has an effect on Purchase Intention with a parameter coefficient of 0.718 and is significant at 0.000 (P Values 0.000).This can be proven by looking at the t statistic of 11,389 which has a value greater than the t table value (t count 11,389 > t table 1.96).

Total Effect
Furthermore, the output path coefficient value shows that E-WOM has an effect on Brand Image with a parameter coefficient of 0.737 and is significant at 0.000 (P Values 0.000).This can be proven by looking at the t statistic of 17,066 which has a value greater than the t table value (t count 22,176 > t table 1.96).
Furthermore, the output path coefficient value shows that Brand Image influences Purchase Intention with a parameter coefficient of 0.409 and is significant at 0.000 (P Values 0.000).This can be proven by looking at the t statistic of 4,139 which has a value greater than the t table value (t count 4,139 > t table 1.96).

The Influence of E-WOM on Purchase Intention
Based on table 16, the output path coefficient value shows that E-WOM has an effect on Purchase Intention with a parameter coefficient of 0.718 and is significant at 0.000 (P Values 0.000).This can be proven by looking at the t statistic of 11,389 which has a value greater than the t table value (t count 11,389 > t table 1.96).This means that if the communication carried out by Compass is by word of mouth, it will increase the intention to buy Compass shoe products.
Based on Faza Research, (2019) that electronic word of mouth has a positive influence on purchase intention.This research shows consistent results according to research conducted by Kumar & Kudeshia (2017) that electronic word of mouth has a positive influence on purchase intention by shaping consumer behavior and ultimately creating purchase intention in consumers.(Zhao et al., 2020) This research is also supported by the results of research by Jalivland, (2012) which shows that electronic word of mouth has a very strong direct influence on purchase intention, because electronic word of mouth has a role in increasing popularity and consumers can read product recommendations online and will creating Purchase intention (Gede et al., 2018).

The Influence of E-WOM on Brand Image
Furthermore, the output path coefficient value shows that E-WOM has an effect on Brand Image with a parameter coefficient of 0.737 and is significant at 0.000 (P Values 0.000).This can be proven by looking at the t statistic of 17,066 which has a value greater than the t table value (t count 17,066 > t table 1.96).E-WOM is closely related to brand image.e-WOM can have a very strong influence on consumer perceptions of a product.Research conducted by Dosti et al. 2016) shows that the e-WOM variable has a significant effect on the brand image variable.This opinion is in line with Torlak et al. (2012) which shows that consumers consider reviews obtained from e-WOM channels and use these reviews to form perceptions regarding brand image.(Chu & Chen, 2019) Validity before removing several indicators According to JF Hair (2014:45), "Validity testing in This research uses two methods of measurement, namely convergent validity and discriminant validity."The convergent validity criteria are the average variance extracted (AVE) value > 0.5 and outer loading > 0.6."The path diagram of the causal relationship between constructs and the loading factor values for each indicator can be seen as follows.

table below : Outer Loading table before removing several indicators or variables Indicator E-WOM Brand Image
Based on Table12above, each variable shows an outer loading value above 0,6 then the conclusion is that the data is valid.The following is a table of AVE values for each variable.

Processed Smart PLS 4 (2023)
Based on Table12above, each variable shows an outer loading value above 0,6 then the conclusion is that the data is valid.The following is a table of AVE values for each variable.This is in accordance with according to JF Hair (2014:45), "Validity testing in this research uses two measurement methods, namely convergent validity and discriminant validity."The convergent validity criteria are the average variance extracted (AVE) value > 0.5 and outer loading > 0.6."The path diagram of the causal relationship between constructs and the loading factor values for each indicator can be seen as follows.

Table Hypothesis
Based on table 17, the output coefficient value shows that E-WOM influences Purchase Intention through Brand Image with a parameter coefficient of 0.302 and is significant at 0.000 (P Values 0.000).This can be proven by looking at the t statistic of 4,060 which has a value greater than the t table value (t count 4,060 > t table 1.96).