*4.1. Respondent Profile*

To provide more comprehensive data, a discussion of the respondent's description was carried out by descriptive analysis of the respondent including gender, age, education level and status of e-wallet usage. Analysis based on the respondents' answers can be seen in Table 2.


**Table 2.** Descriptive statistic of respondents' e-wallets.


**Table 2.** *Cont*.

Based on Table 2, the profile of 100 respondents in the research of intentions to use electronic wallets at Warunk Upnormal, Bandung, West Java was dominated by women (66%) and men (34%) aged 26–35 years (42%) with an undergraduate education level/bachelor (33%). Most of the respondents' reasons for using an e-wallet are not dominated by the cause of benefit offerings (53%), because respondents have used an e-wallet for more than one year (79%). People use e-wallets for transactions on almost all purchases (23%) on the grounds that e-wallets are practical, easier, faster without having to carry cash, can be used anywhere, and avoids risk (24%), as well as that they are many of the choices of merchants.

#### *4.2. Structural Equation Model*

When the results of the identification of the structural equation model show that the parameters of the SEM can be estimated, then further estimation of the parameters of the structural equation model can be achieved through various techniques (Table 3). However, based on Lisrel's output, which shows the estimation model used in estimating the parameters in this study, the maximum likelihood (ML) was used.


**Table 3.** Test results using the overall structural equation model (SEM).

Based on Figure 5, it is known that the error value of each question is smaller than the value of the relation so that research using the structural equation model (SEM) can be continued. In addition, the overall SEM model test can also be used to see in its entirety whether the use of the structural equation model is suitable for the sample data. This test was carried out by comparing the sample covariance matrix and the estimated covariance matrix using a structural equation model. There were three types of measures to test whether the use of the structural equation model as a whole fits the data (good fit). Those are absolute fit measures, incremental fit measures, and parsimony fit measures. Based on Lisrel's output for absolute fit measures, incremental fit measures, and parsimony fit measures, the results showed that the use the structural equation model as a whole has a good ability in terms of matching sample data (good fit). In other words, the estimated covariance matrix using the structural equation model is not statistically different from the sample data covariance matrix.

**Figure 5.** Structural equation output results.

The overall SEM model test showed that the SEM model as a whole is able to match the data (good fit). Meanwhile, the measurement model test showed that the measurement model has good convergent validity and discriminant validity. Furthermore, it was supported by the structural model test.

Based on Lisrel's output in Figure 6, the following structural equations are formed:

Usage Decision = 1.22 PP − 0.73 PM + 0.31 PR + e

**Figure 6.** Structural model test results.

The structural equations above are also presented in the following Lisrel output (Figure 6).

The following will be interpreted for each of these structural equations in Figure 7.


**Figure 7.** Output structural equations of Lisrel.

#### For Structural Equation

The path coefficient of the latent variable of product knowledge is 1.22. A positive path coefficient value indicates that the latent variable of product knowledge has a positive effect on satisfaction. The statistical value of the test for the path coefficient of the product knowledge latent variable is t = 2.61. The table value with the significance level is α = 5%. Therefore, it can be concluded that the effect that occurs between the latent variable of product knowledge and the latent variable of the decision is statistically significant at a significance level of 5% with tt = 2.61 tα = 5% t table = 1.985. This means that the more customers know the features in their Go-Pay service, the faster they will make a decision to use Go-Pay services.

The path coefficient of the perceived benefit latent variable is −0.73. A negative path coefficient value indicates that the latent variable of perceived benefits has a negative effect on the decisions. The statistical value of the test for the path coefficient of the product knowledge latent variable is *t* = −0.49. The table value with the significance level is α = 5%. Therefore, it is concluded that the effect that occurs between the perceived benefit latent variable and the decision latent variable is not statistically significant at the 5% significance level with *tt* = −0.49 tα = 5% t table = 1.985. This means that the more customers understand the benefits they will get in using Go-Pay services, they will be more interested in making decisions to use Go-Pay services.

Furthermore, the path coefficient of the risk perception latent variable is 0.31. A positive path coefficient value indicates that the risk perception latent variable has a positive effect on decisions. The statistical value of the test for the path coefficient of the risk perception latent variable is *t* = 2,98. Meanwhile, the table value with the significance level is α = 5%. Therefore, it is concluded that the effect that occurs between the latent variable of risk perception and the latent variable of decision is statistically significant at a significance level of 5% with *tt* = 2.98 tα = 5% t table = 1.985. This means that the more customers understand the risks they will accept in using Go-Pay services, they will be more interested in using Go-Pay services.

The coefficient of determination based on Lisrel's output (Figure 4) is 0.70. This value can be interpreted as 70% of the total variation (total variation) of the latent decision variables can be explained by the structural equation, and the remaining 30% is explained by other variables (R2).

#### **5. Discussion**

#### *5.1. The Effect of Product Knowledge on Usage Decisions*

The test results in this study prove that product knowledge has an effect on usage decisions. This means that all information that consumers have about the product can be easily understood by consumers. Product knowledge is defined as a collection of various kinds of information about products [24]. This knowledge includes product categories, brands, product terminology, product attributes and features, product prices and product beliefs. Consumer product knowledge is basically determined by the level of consumer familiarity with the product. Consumer knowledge is all information that consumers have about various kinds of products and other information related to its function as a consumer. Understanding consumer knowledge is very important for marketers. Information about what to buy, where to buy, and when to buy will depend on consumer knowledge. Consumer knowledge will influence purchasing decisions, and even repeat purchases. When

consumers have more knowledge, they will be better at making decisions, more efficient, more precise in processing information, and able to recall information better

The results of research by Kim, Mirusmonov, and Lee [12] suggest that product knowledge is an important factor in facilitating users in making mobile payments. On the other hand, Parastiti and Mukhlis [37] found that product knowledge has no significant effect on use of electronic money. This could be due to the lack of information, thus causing a person's low interest in using a product, as well as the culture of the Indonesian people who still feel comfortable using cash instead of electronic money.
