Strategy Creating Positive Word of Mouth Based on Relationship Quality

Kata kunci: Kualitas Informasi; Keamanan; Responsivitas; Kualitas Hubungan; WOM; The development of technology had a great impact on the company in introducing or marketing their products to consumers, marketing has become the world's interest. They market of goods, services, property, people, places, events, information, ideas, and organizations. In order to maintain its position in the eyes of consumers, marketers strive to create an ideas and innovation that can attract the interest of market participants. The purpose of this research is done in order to examine the information of variable quality, assurance and responsiveness to the quality of relationships that affect consumer promotions lazada mouth to mouth in Semarang. This data research using questionnaires with a sample of 148 respondents in the city of Semarang. The sampling technique is to use probability sampling, with one of the methods used is purposive sampling. The data obtained and analyzed by Structural Equation Modeling (SEM) with AMOS application 21. The results of this study are: (1) the quality of information, assurance and responsiveness positive and significant impact on the quality of the relationship. (2) the quality of information, assurance, responsiveness and quality of the relationship positive and significant effect on word of mouth.


INTRODUCTION
The dynamics of increasingly fierce business competition requires businesses to think and act smart to be able to compete with competitors. One strategy that can be used by businesses that promotion through word of mouth (WOM). Word of mouth can provide input for a brand used by businesses because often WOM made the most honest opinion and what their consumers. Responses that do could be quite significant, such as trying to change the image or reposition itself, to make changes to the basic products or introducing new products, changing service or customer support for its customers after the sale, to communicate directly to customers about the evidence about the performance of their superior (Mulyadi, 2013).
Research on word of mouth has been conducted by previous research, but there are still inconsistencies in the results of the research. As research has been done by Felix, (2017) which states that the responsiveness, assurance no effect on consumer intentions to make WOM, but there are differences in the results of research conducted by Liu and Lee (2016) which says that the responsiveness and assurance to create WOM. In others research about WOM still there are differences in the results of studies such as the study carried out by the sukia et al, (2016) which says that the quality of information has no effect on WOM, but there are differences in the results of research conducted by Hyuong and Hyunjoo (2011) which says that greatly affect the quality of information consumers to make WOM.
Lazada is one of the largest online marketplace sites in Indonesia. Lazada was launched in March 2012 and expanded rapidly until today. Lazada Indonesia is one part of a network of online retail Lazada Group which operates in six countries in Southeast Asia, which consists of Lazada Indonesia, Lazada Malaysia, Lazada Thailand, Lazada Vietnam, Lazada Singapore and Lazada Philippines with total users 550 million users of the total six countries.
Based on the research results of the market share of e-commerce by Nusa Research in 2014 as many as 864 samples, Lazada in the top position as a popular brand, and OLX is in the second position. This was followed by trade, FJB Kaskus, Qoo10, Zalora, Tokopedia, Rakuten, Unity, and the last Elevenia. These results are calculated based on the sites visited respondents during the last 3 months (Accessed fromhttps://dailysocial.net.) Based on the table 1 above, from 2014 and 2015 Lazada became the Top Brand, but in 2016 and 2017 Lazada was in second place after OLX, this data proves that Lazada has experienced a decline in marketing performance. Based on the existing problems, this research includes the formulation of the problem as follows: Research on positive word of mouth has been conducted by previous research. However, no one has conducted research that uses information quality, assurance, responsiveness, to positive word of mouth using relationship quality as the intervening variable.

LITERATURE REVIEW
Research conducted by MS Balaji et al. (2017) which says that the quality of information significant effect on the quality of the retail banking client relationships in Malaysia. The study is in line with research conducted by McKnight et al. (2017) which states that the quality of information affects the relationship of trust that affects business to business commitments. H1: The quality of the positive impact of information on the quality of the relationship. Research conducted by Budi (2013) states that assurance positive effect on the quality of interpersonal relationships. The study is in line with research conducted by Giovanis et al. (2011) states that service quality (assurance) have a significant positive impact on the quality of the relationship. H2: Assurance positive effect on the quality of the relationship. Research conducted by Wei-Ming Ou et al. (2011) stated that the quality of service (responsiveness) have a significant positive impact on the quality of research relationship is consistent with the study conducted by MS Balaji et al. (2017) which states that responsiveness significant effect on the quality of the customer relationship banking.

H3:
Responsiveness positive effect on the quality of the relationship. Research conducted by Young and Hyunjoo (2012) stated that the quality of information significant influence by word of mouth. The study is in line with research conducted by Katerina et al. (2012) which states that information security was significantly to positive word of mouth.   Handayanto et al. (2017), Felix. (2017, MS. Balaji et al. (2017), Vaerenbergh and Jonas (2014).

RESEARCH METHOD
The population in this study is consumers who've made a purchase online in Lazada andin this study the authors took a sample using purposive sampling technique. In this regard, Arikunto (2010) explains that purposive sampling is done by taking the subject is not based on strata, random or region but based on their specific purpose. Characteristics of the respondents as follows: 1. Lazada customers in Semarang who has ever bought a product in Lazada 2. Consumers in Semarang who has ever bought a product in Lazada In this study used data collection method is by using a questionnaire / questionnaire. This technique premiere using to collect data. Answer given appraisal from 1 to 10 for the range of 1-10 vote was seen as a simple and commonly performed by researchers in Indonesia.
This research technique using two approaches: 1. Confirmatory factor analysis, The SEM using AMOS computerized package 21 in this case to confirm the factors predominant in one group of variables. 2. Regression Weight SEM were used to investigate how much the relationship between variables. By looking at the complexity of the data measurement techniques proposed is a multivariate technique that is SEM (Structural Equation Modeling).
To create a complete modeling some steps that need to be done (Ferdinand, 2005), namely: 1. development of a model 2. Development of flow charts to show causality. 3. Conversion flowchart into a series of structural equations and measurement model specifications. 4. Selection of the input matrix and estimation techniques on models built. 5. Assessing the problem of identification. 6. Goodness of fit model evaluation. 7. Interpretation and modification of the model.

Analysis of Structural Equation Model 1. Development of Model-Based Theory
The model developed in this study consists of four variables or constructs, that information quality, assurance, relationship quality, and are based on the positive WOM.Model literature review and support previous studies. This built theoretical models will then be analyzed as a model 'researchable' using SEM (Ferdinand, 2000).

Development Flowchart (path diagram)
The theoretical model that was established by the theory will be displayed in the form of a flowchart (path diagram) with the help of SEM that run through the Amos program 21.0. The variables contained in the flow chart is basically divided into two groups, namely the exogenous and endogenous variables. Exogenous variables consists of three variables: information quality, responsiveness assurance. While endogenous variables consist of two variables: relationship quality, and positive word of mouth.

Conversion Flowchart In Eq. Model and Measurement Model
The model has been presented in the form of a path diagram above, then expressed in equations and structural equation model specification that states the measurement (measurement model). Structural equation formulated to express causality among different constructs or variable formations.

Table 5. Sample Covariant -Estimates
Sources: Primary data are processed (2018) Covariant sample estimates the above only shows the result of the conversion performed by the SEM program of input data into the SPSS results matrix form input which will then be used as input to the next process. Estimation techniques that will be used is the method of maximum likelihood estimation for the number of samples used ranged from 100-200.

Confirmatory Analisis Constructs Exogenous
From Figure 2 below appears that the relationship between the variable information quality with assurance variables have a correlation of 0.77, with a relationship quality assurance variables have a correlation of 0.74 and variable information quality with the variable relationship quality 0,68. Thus the correlation between each of these variables is still below the threshold level required correlation of less than 0.90 (Ferdinand, 2000).   (2018) Conformance test models of table 6 known as chi-square value small (20.892 <36.415) and a probability value that indicates a value above the limit of significance that is equal to 0.645 (p> 0.05). These results indicate that the null hypothesis that there is no difference between the sample covariance matrix with a population estimated covariance matrix is acceptable. From the results of tests of significance weighting factors (Table 6), it is also seen that each indicator or dimension respectively forming latent variables on exogenous construct showed results that meet the criteria of CR value is above 1.96, with P less than 0.05. Moreover, the loading factor values as shown in

Analysis Construct Endogenous
From Figure 3 below shows that in the relationship between variables there is no problem of identification. Suppose the relationship between variables correlation relationship quality with variable positive word of mouth has a correlation of 0.41. Furthermore, the results of the confirmatory factor analysis of the research model shown in Figure 3, Table 6 and Table 7.  Source: Data processing (2018) There are two basic tests in the confirmatory factor analysis to construct a model of exogenous i.e. conformance test ( Table 8) and tests of significance weighting factors (Table 9). Of conformance test models known as table 9 chi-square value small (15.507 <8.834) and a probability value that indicates a value above the limit of significance that is equal to 0.087 (p> 0.05). These results indicate that the null hypothesis that there is no difference between the sample covariance matrix with a population estimated covariance matrix is acceptable.
From the results of tests of significance weighting factors (Table 7) was also seen that each indicator or dimension respectively forming latent variables on endogenous constructs showed results that meet the criteria of CR value is above 1.96, with P less than 0.05. Moreover, the loading factor values as shown in Figure

Analysis Structural Equation Model (SEM) in Full Model
As in confirmatory factor analysis, Model testing is also done with two tests, which test the suitability of the model and test the significance of causality through regression coefficient test.

Figure 2. Structural Equation Model Testing Results
Sources: Primary data are processed (2018)  Results of testing the suitability of the model can be seen in Figure 2 and Table 10 by observing the results of the analysis that has been qualified. Chi-square value (78.914<103.009) and the probability value (0.482> 0.05) indicates that the null hypothesis that there is no difference between the sample covariance matrix with the estimated population covariance matrix can be accepted so that the construct of this study can be accepted. From these results, it can be concluded that the indicators forming the latent variables significantly is an indicator of latent factors are formed. Besides all the analysis results in Table 10 also shows the value of goodness of fit acceptable because it has met the requirements. Sources: Primary data are processed (2018) The result of the regression coefficients (Table 11) also indicate a value that meets the requirements that are above CR value of 1.96 with a probability value of <0.05. On the relationship between the variable information quality with relationship quality have value CR = 2.248 (> 2.0) with probability 0.025 (<0.05). Meanwhile, on the relationship between the variable responsiveness to the relationship quality have value CR = 2.514 (> 2.0) with probability 0.012 (<0.05) .At the relationship between the variables relationship quality with positive word of mouth has a value of CR = 2.513 (> 2, 0) with probability 0.012 (<0.05) .At the relationship between the variables relationship quality with positive word of mouth has a value of CR = 3.792 (> 2.0) with probability 0.000 (<0.05). On the relationship between the variable information quality with positive word of mouth has a value of CR = 2.299 (> 2.0) with a probability of 0.014 (<0.05). On the relationship between variables assurance with positive word of mouth has a value of CR = 2.102 (> 2.0) with probability 0.010 (<0.05). Besides the above criteria, indicators of variable information quality, assurance, responsiveness, relationship quality and positive word of mouth is valid. 3. The emergence of a very high correlation between the estimated coefficients obtained (> 0.9).

Evaluation Criteria Goodness of Fit Univariate Outlier Evaluation
Presence or absence of outlier univariate testing done by analyzing the Z score value of the data used in this research. If there is a larger Z score value ± 3.0 then it will be categorized as outliers. This outlier univariate test using SPSS 10. The results of data processing to test whether there is an outlier is presented in Table 12: The results of analysis of univariate outliers were done by watching the figures on the minimum and maximum column indicates the absence of Z score greater value ± 3.0. In the column the minimum value is -2.98685(<-3.0), while the maximum column greatest value was 2.05948 (<3.0). It can be concluded that there is no univariate outlier in the data of this study.

Evaluation of Multivariate Outliers
Manual calculation of the distance based on chi-square is on df: 81 (the number of independent variables) with p <0.001 obtaining a yield of 128.08. While the results of the processing of SEM showed the largest is distance 33.416. If both results are compared it is seen that the value of the processing results of SEM smaller than manual calculations (33.416 <128.08). It can be concluded that there is no multivariate outlier in the study.

Normality Test Data
The test results of the normality of the data shown in Table 13: From the data processing are shown in Table 13 shows that the greatest value to the columns of CR is 0.910 (<2.58). Similarly, the multivariate value of -2.488 also showed a smaller value of -2.58. Thus, the research data was normally distributed.

Evaluation of Multicollinearity and Singularity
From the data processing value is a sample covariance matrix determinant: Determinant of the sample covariance matrix = 99.002. From these results it can be seen the value of the determinant of the sample covariance matrix still remains well above zero. It can be concluded that there is no multicollinearity and singularity.

Conformance Test and Test Statistics
From the data processing is then compared with a predetermined statistical limits, test the suitability of the model shown in Table 8 of the test results known that out of the eight criteria required, are in good condition, With these results it can be concluded that the model study had the goodness of fit is good.

6.Interpretasi and Modification Model
Good models have standardized Residual Covariance small. Figures + 2.58 is the limit of allowable value of the standardized residuals. Standardized Residual Covariance Results are presented in Table 14.  Source: Data processing (2018) The analysis of this study does not indicate a value of standardized residual covariance exceeding + 2.58 (Ferdinand, 2002.

Reliability and Variance Uji Extract test Reliability
The minimum reliability value of dimensional forming latent variable that can be accepted is equal is 0.70.
Overall results of testing the reliability and variance extracted subsequently presented in the table 15.

Table 15. Test Reliability and Variance Extract
From observations in Table 15 appears that there are no reliability values smaller than 0.7. Similarly, the variance test extract was not found that the value is below 0.5. Thus, the indicators | 296 | used as an observed variable for the constructs or latent variables, can be said to have been able to explain the constructs or latent variables are formed.

RESULTS AND DISCUSSION
Hypothesis 1 From the data processing is known that the value of CR on the relationship between information quality of the relationship quality, as shown in Table 11 is equal to 2.249 with a P value of 0.025. Both of these values indicate the results that qualify, ie above 1.96 for CR and under 0.05 to P. It can be concluded that the hypothesis 1 is acceptable. Thus, this study supports the results of MS, Balaji et al, (2017) which addressed that the better the quality of information consumers of the company can improve the quality of the customer relationship with the service provider.
Hypothesis 2 From the data processing is known that the value of CR on the relationship between assurance to relationship quality, as shown in Table 11 is equal to 2.478 and a P value of 0.013. Both of these values indicate the results that qualify, i.e. above 1.96 for CR and under 0.05 to P. It can be concluded that the second hypothesis proposed in this study can be accepted. Thus, this study supports the results of Giovanis et al, (2015) which shows that the better the guarantees given by the service providers to consumers can improve the quality of the relationship between the consumer and the service provider.
Hypothesis 3From the data processing is well known that the relationship between responsiveness with CR on relationship quality, as shown in Table 11 is equal to 2.514 with a P value of 0.012. Both of these values indicate the results that qualify, i.e. above 1.96 for CR and under 0.05 to P. It can be concluded that the third hypothesis in this study is acceptable. Thus, this study supports the research that has been done by MS Balaji et al, (2017) which states the higher responsiveness, the greater the likelihood that consumers will make positive word of mouth.
Hypothesis 4 From the data processing is known that the value of CR on the relationship between information quality to positive WOM, as shown in Table 11 is equal to 2.299 with a P value of 0.014. Both of these values indicate the results that qualify, i.e. above 1.96 for CR and under 0.05 for a 5 P. Thus, the hypothesis in this study is acceptable. This study supports the research that has been done by Young and Hyunjoo (2012) which states that the positive effect on the quality of information positive word of mouth.
Hypothesis 5 From the data processing is known that the value of CR on the relationship between assurance to positive WOM, as shown in Table 11 is equal to 2.102 with a P value of 0.010. Both values are showing results that qualify, i.e. above 1.96 for CR and below 0.05 for P. Thus we can conclude that the hypothesis 6 proposed in this study can be accepted research supporting the conclusions obtained in the study who conducted by Handayanto et al, (2017) which states there is a positive relationship between assurance to the word of mouth marketing communications.
Hypothesis 6 in this study is the better responsiveness, the higher the positive WOM. From the data processing is known that the value of CR on the relationship between responsiveness to positive WOM, as shown in Table 11 is equal to 2.734 with a P value of 0.006. Both values are showing results that qualify, i.e. above 1.96 for CR and below 0.05 for P. Thus, we can conclude that the hypothesis 6 proposed in this study can be accepted and research supports the conclusion obtained in studies conducted by Awais et al, (2016) which says that the responsiveness positive effect on word of mouth.
Hypothesis 7 From the data processing is known that the value of CR on the relationship between relationship quality to positive WOM, as shown in Table 11 is equal 3,792dengan P value of 0.000. Both of these values indicate the results that qualify, i.e. above 1.96 for CR and below 0.05 for 7 P. Thus, the hypothesis in this study can be accepted and this research supports the conclusions obtained in the study conducted by Sandy et al, (2011) which says that the quality of influential relationship positively to word of mouth.

CONCLUSION
In increasing positive word of mouth, Lazada can make a strategy to improve the quality of its relationships with consumers, of course this is not a young thing because establishing relationships with consumers who have a variety of thoughts certainly requires a big effort. In this study, it is confirmed that the existence of good quality relationships between companies and consumers can make consumers make positive word of mouth, another thing that can be done is to improve the quality of the products they sell and provide easy access to applications, another important thing is Consumers need good assurance Felix and responsiveness, because consumers also have the principle that they are kings and must be really listened to, because when they are not listened to they will easily make bad reviews or negative word of mouth on a market place, the results of this study provide a new perspective in the marketplace world because sometimes the market place does not think about the quality of their relationship and considers consumers to be the second priority, which in fact is the most crucial thing in the world of business. The word consumer is that the king needs to be studied properly and applied in a proper manner. set it up, because when consumers find one gap from the market place the other good will disappear.
It was supposed to be a top priority Lazada to improve the quality relationship that will positively affect the positive word of mouth is the increased assurance to users Lazada. Because in this case study was shown to significantly influence their high-impact quality relationship with the creation of positive WOM. Huge influence was greater assurance than information quality and responsiveness. Some technical issues that should be considered by basing on the results of this study are: Ensuring that consumers feel safe shopping in Lazada, ensure that consumers receive the original product from the seller, Ensuring that the consumer orders quickly in the process.
If efforts to increase assurance that do not have an impact as expected, then Lazada should improve information quality is expected to improve the quality relationship that will have an impact on positive word of mouth. Some technical issues that should be considered by basing on the results of this study are: Lazada need to provide full information to consumers (prices of products, stock products, size, product images, product descriptions and other information. And Besides Lazada should provide clear and accurate information to consumer. If the information quality efforts that do not have an impact as expected in enhancing relationship quality that will affect the positive word of mouth on the consumer then Lazada should improve the responsiveness becomes better built primarily on the quality of the seller, followed by the quality of the product. Some technical issues that should be considered by basing on the results of this study are: Lazada need to improve filtering seller that still there is dishonest in selling is selling a product that is not intact (broken). Lazada must perform maintenance on the application, to be more accessible and easy error.