• Tidak ada hasil yang ditemukan

CHAPTER 4: Research Result and Discussion

4.3 Research Result

4.3.1 Multiple Linear Regression Analysis

55

Figure 0-7: Why visit Shell station.

56

Figure 0-5: Boxplot without outliers 4.3.1.2 Test of Normality

In the above table 8 presented, Shipro-Wilk’s test score is (p < 0.5) and Kolmogorov-Smirnova.

The results indicate that this data is not follow normally distribution for service visibility, service reliability, service customization service effectivity and efficiency, and customer satisfaction.

In table 4-7, the results of test of normality is presented for every variable. Moreover, as shown in figure 4-7 to figure 4-14, the Normal Q-Q plots for all variables determined the same pattern, with the assumption that the normality is violated for all of them.

57 Table 0-9: Test of Normality

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

Service visibility .116 192 .000 .941 192 .000

Service reliability .087 192 .001 .961 192 .000

Services Effectivity Efficiency

.118 192 .000 .950 192 .000

Service customization .089 192 .001 .979 192 .006

Customer satisfaction .144 192 .000 .922 192 .000

a. Lilliefors Significance Correction

Figure 0-8: Service visibility Normal Q-Q plot

58

Figure 0-9: Service reliability Normal Q-Q plot

Figure 0-10: Service effectivity and efficiency Normal Q-Q plot

59

Figure 0-11: Service customization Normal Q-Q plot

60

Figure 0-12: Customer satisfaction Normal Q-Q plot

4.3.2.2 Scatter Plot

These graphs presented in figure 4-12 to figure 4-16 show that the scatter points were roughly in diagonal as illustrated from left bottom to right upper of the scatterplot. The service visibility and service reliability have a straight line with customer satisfaction. However, service effectiveness and efficiency and service customization did not show straight line but a slope in a diagonal way with customer satisfaction. So, this assumption is not violated in any scatterplot.

61

Figure 0-13: Scatter plot linear regression of services visibility and customer satisfaction

Figure 0-14: Scatter plot linear regression of services reliability and customer satisfaction

62

Figure 0-15: Scatter plot linear regression of services effectivity and efficiency and customer satisfaction

63

Figure 0-16: Scatter plot linear regression of services customization and customer satisfaction

4.3.2.3 Multicollinearity Between Independent Variables

The first premise is to observe the multicollinearity among independent variables and this would be problematic if the correlations value is more than 0.8. The table 4-8 show Pearson correlations to describe that service visibility, service reliability, service customization, services effectivity and efficiency have positive association with each other.

Services reliability has weak positive association with service customization (r = .196, p < .01), and services visibility has a strong positive association with customer satisfaction (r = .701, p <

.01). The first premise of assumption related multicollinearity is also met.

64

Table 0-10: Pearson correlations between independent variables

Service visibilit

y

Services reliabilit

y

Service customizatio

n

Services Effectivenes

s &

Efficiency

Customer satisfactio

n

Service visibility

Pearson Correlatio

n

1 - - - -

Services reliability

Pearson Correlatio

n

.633** 1 - - -

Service customizatio

n

Pearson Correlatio

n

.359** .196** 1 - -

Services Effectivenes

s &

Efficiency

Pearson Correlatio

n

.522** .415** .318** 1 -

Customer satisfaction

Pearson Correlatio

n

.701** .623** .391** .528** 1

** Correlation is significant at the 0.01 level (2-tailed).

Note: r denoted the Pearson coefficient value which represented the positive association and r value explained the effect of correlation between variables. P value denoted the significance value which we used to explain the association is significant or insignificant while comparing with the alpha value .05 or .01 (as above in table 2-tailed).

In table 4-9, coefficients are captured. The second premise of this assumption is to see the tolerance value and VIF scores. If the VIF score is greater than 10, or the Tolerance is less than 0.1, then you have concerns over multicollinearity. The following table coefficient explained

65

that the data met the assumption of multicollinearity as service visibility has (tolerance = .500, VIF = 1.99), service reliability (tolerance = .589, VIF = 1.69), services effectivity and efficiency (tolerance = .698, VIF = 1.43) and service customization (tolerance = .847, VIF = 1.18) have more tolerance value and less VIF score.

Table 0-11: Coefficients table

Model

Unstandardized Coefficients

Standardized

Coefficients t Sig.

B Std. Error Beta

1

(Constant) .268 .222 1.208 .229

Service visibility .381 .063 .400 5.997 .000

Service reliability .266 .061 .267 4.346 .000

Services Effectiveness &

Efficiency

.198 .066 .169 3.002 .003

Service customization .103 .043 .124 2.413 .017

a. Dependent Variable: Customer satisfaction

4.3.2.3 Homoscedasticity

To test the fifth assumption, you need to look at the final graph of the output. This tests the assumption of homoscedasticity, which is the assumption that the variation in the residuals (or amount of error in the model) is similar at each point of the model. The points of standardized predicted value and standardized residuals were scattered around 0 to both sides -3 and +3 equally and randomly. As the values of standard residuals increased the values of standard

66

predicted varied roughly and overlapped shown in above scatterplot. The dots did not show funnel like shape but look like random array from -3 to +3 range (see figure 4-17).

Figure 0-17: Test of homoscedasticity

4.3.2.4 Influential cases for model bias

The table 4-11 illustrates the study’s assumption confirmed by checking Data File and examining the Cook’s Distance values. The SPSS must create the new column of cook’s distance statistic values for every participant. We use the descriptive statistics to present the minimum and maximum among values to observe the values that were over 1. The interpretation here is that there is a likelihood that these values are significant outliers and that they might influence the presentation and analysis of the model. In that effect, they are removed so that the whole analysis is rerun. For the results of the analysis performed for this case, cook’s values did not exhibit such instances corresponding to the minimum value shown as 0.00 and the maximum

67

value identified as 0.16, which means none of these values is more than 1. This assumption is also met.

Table 0-12: Cook’s distance statistics

N Minimum Maximum Mean Std. Deviation

Cook's Distance 192 .00000 .16162 .0057415 .01393059

Valid N (listwise) 192

4.3.2.4 Residuals values are independent

We have to interpret the model summary, which is shown in table 4-12. Durbin-Watson statistic is referenced to test assumption that corresponds to noting that the residuals are independent (or uncorrelated). Ideally, the score for this statistic is identified to vary, ranging from a value of 0 to 4. Relatively normal expected for the values of this test statistic are 1.5 to 2.5. Nevertheless, the source of concern is evident when the values captured are outside this range.

Table 0-13: Model summary using Durbin-Watson statistic

Model R R Square Adjusted R Square

Std. Error of the Estimate

1 .764a .584 .575 .70504

a. Predictors: (Constant), Service customization, Services Effectivity & Efficiency, Service reliability, Service visibility

68

The table model summary identified the variation in the dependent variable due to the independent variable where the value of R determined the strength of prediction between independent and dependent variables, here (0.764) is good level of prediction. R square (coefficient of determination) showed that almost 58.4% variation was observed in customer satisfaction due to service visibility, service reliability, service effectivity and efficiency, and service customization. In addition, 57.5% adjusted r square determined the honest variability between variables by generalizing these results to the whole population. (see table 4-12).

Table 0-14: Summary of model prediction: Independent and dependent variables

Model R R Square Adjusted R Square

Std. Error of the Estimate

1 .764a .584 .575 .70504

a. Predictors: (Constant), Service customization, Services Effectivity & Efficiency, Service reliability, Service visibility

In the ANOVA results shown in table 4-14, a significant value of 0.000 determined that our regression model was statistically significant, which indicated that service visibility, service reliability, service effectivity and efficiency, and service customization predicted customer satisfaction.

69 Table 0-15: ANOVA results

Model Sum of Squares df Mean Square F Sig.

1

Regression 130.679 4 32.670 65.723 .000b

Residual 92.955 187 .497

Total 223.634 191

a. Dependent Variable: Customer_satisfaction1

b. Predictors: (Constant), Service customization, Services Effectivity & Efficiency, Service reliability, Service visibility

The F value presented the fit of the model by considering the inaccuracy in the model according to our value it is considered as an efficient model (F (4,187) = 65.723), p < .01 whereas significance value is less than alpha value.

Table 0-16: Model coefficients

Model

Unstandardized Coefficients

Standardized

Coefficients t Sig.

B Std. Error Beta

1

(Constant) .268 .222 1.208 .229

Service visibility .381 .063 .400 5.997 .000

Service reliability .266 .061 .267 4.346 .000

Services Effectiveness & Efficiency .198 .066 .169 3.002 .003

Service customization .103 .043 .124 2.413 .017

a. Dependent Variable: Customer satisfaction

70

In the coefficient table, the unstandardized value of B determined the expected change in the dependent variable by changing one unit in the independent variables. While increasing one point of service visibility predicted positive .381 units in customer satisfaction and this positive prediction was found statistically significant (β = .40, t = 5.99, p < .05). This positive effect of service visibility on customer satisfaction concluded that the first alternate hypothesis (H1) was accepted that service visibility has a significant effect on customer satisfaction. Service reliability also predicted positively and found statistically significant (β = .26, t = 4.346, p < .05) and with the increase of one unit in service reliability predicted .266 units in customer satisfaction. This concluded that the second alternate hypothesis (H2) was accepted that service reliability has a significant effect on customer satisfaction.

In the same way, one-unit increase of service effectiveness and efficiency predicted .198 units in customer satisfaction. Customer satisfaction predicted positively by service effectiveness and efficiency and it was found statistically significant (β = .16, t = 3.00, p < .05). This concluded that the third alternate hypothesis (H3) was accepted that service effectiveness and efficiency have significant effect on customer satisfaction. Service customization predicted customer satisfaction positively with .103 units. This positive effect of service customization on customer satisfaction was found statistically significant (β = .12, t = 2.41, p < .05) and the fourth alternate hypothesis (H4) was accepted that service customization has a significant effect on customer satisfaction.

4.3.2.6 Achievement of Research Objectives

The overarching aim of this study was to investigate the impact of customer service modernization on end-user satisfaction with Oman-based Shell petrol stations’ service offerings, specifically focusing on the four dimensions of service quality – service visibility, service

71

reliability, service efficiency and effectiveness, and service customization. In particular, the research investigated the impact of service visibility on customer satisfaction, assesses whether service reliability significantly influences customer satisfaction, and determine the extent to which service efficiency and effectiveness affect consumer satisfaction at the filling stations.

Moreover, the study hoped to examine if service customization substantially affects customer satisfaction. The results of the regression coefficients indicate that service visibility (β = .400, p

= 0.000) and service reliability (β= 0.267, p = 0.000) have a significant positive impact on customer satisfaction, confirming the hypotheses H1 and H2. Based on their coefficient, service effectivity and efficiency (β= 0.169, p = 0.003) also have a statistically significant impact on the dependent variable, rejecting the null hypothesis H3. Further, the coefficient for service customization (β= 0.124, p = 0.017) implies that it has a statistically significant positive impact on customer satisfaction, supporting hypothesis H4. Thus, the research findings suggest that service visibility, service reliability, service effectivity and efficiency, and service customization have a significant positive impact on customer satisfaction in Shell petrol stations in Oman.

Overall, the study achieves the specific objectives, as the results suggest the need to invest in service modernization, especially in the domains of visibility, reliability, service effectivity and efficiency, and customization, to enhance customer satisfaction in the petroleum sector.

72