Chapter 4 Results
4.1 Testing hypotheses
1. General
First of all, it is necessary to check descriptive statistics. Total 13 employees including the author were analyzed and 15 types of problems were recorded during observation.
Pivot table shows how the data are distributed among the problems: which receptionist is responsible for solving particular problem. Additional factor is added showing how many problems each receptionist solved and did not solve. It helps to have a graphical idea of entire situation among problems and receptionists. From Appendix 3 (Please, refer to the page 50) it is clear that too many employees were responsible for the same problems. The author divides the results from Pivot table into 2 parts and summarizes them. These results reflect only descriptive statistics and do not show any causal effects. According to researcher’s personal experience under normal conditions there were 1~4 people in front line, therefore, it is advised to think that
20
if more than 5 persons (4 people in front plus 1 person from back office) are responsible for solving the same problem, it shows that delegating responsibilities’ level is too high and too many employees are working on the same problem.
Table 1 Problems included less than 5 people
Problem Number of people worked on this problem
General question 4
ARC & Visa 4
Attendance list 4
Stamp 4
Scholarship 5
Group problem 4
Unstated 2
From Table 1 it is analyzed what problems include less than 5 persons to work on.
Some of the problems, for example, general questions, attendance list, stamp, group problem are quite wide-ranging, therefore, employees from the front line can easily solve them without including back office into communication process. However, some narrow problems like ARC &
Visa, scholarship shows that only limited number of employees could work on that problem.
These results show good delegating level, when people know what they are responsible for and know whom they might ask to help.
Table 2 Problems included more than 5 people
Problem Number of people worked on this problem
Card problem 7
Refund 6
Special docs 9
Payment 8
TOCFL 6
Personal problems 10
Choose courses 6
Study at MTC for newcomers 8
21
In Table 2 it is reported what kind of problems include more than 5 persons for solving them. These results show that employees face some difficulties with managing this type of problems, probably, they need more guidelines or time for solving the problems. These problems focus on specific situations, therefore, more guidelines should be provided for the staff, as for now they do not have enough knowledge to manage it.
In most cases personal problems and refunds were not solved by the employees, it might have happened due to the misunderstanding from the students’ side or the staff was not capable to answer.
Hypothesis 1.1
(H_1.1) The satisfaction level can be explained by following variables: helpful answer, prompt answer, polite answer, the length of queue (measured in minutes) and how the problem is solved.
To test the hypothesis mentioned above it is advised to run regression, because it helps to evaluate whether these variables have any influence on entire Satisfaction from the service.
Table 3 H_1.1 Model Summary
Model R R Square Adjusted R Square
Std. Error of the Estimate
1 .841a .707 .701 .589
a. Predictors: (Constant), How the problem is solved, Length of queue (measured in minutes), Polite answer, Prompt answer, Helpful answer
According to the Table 3, R and R square are both high enough for further analyses, therefore, it is necessary to check significance level.
Table 4 H_1.1 ANOVA
Model Sum of Squares df Mean Square F Sig.
1 Regression 208.755 5 41.751 120.414 .000b
Residual 86.682 250 .347
Total 295.437 255
a. Dependent Variable: Satisfaction
22
b. Predictors: (Constant), Rate how the problem was solved, How long was queue (measure in minutes), How polite the answer is, How prompt the answer is, How helpful the answer is
From Table 4 Significance level α < 0.05, therefore it shows highly significant result meaning that there at least one variable has significant meaning and can be analyzed.
Table 5 H_1.1 Coefficients
Model
Length of queue (measured in
minutes) -.149 .037 -.140 -4.065 .000
How the problem is solved .114 .026 .180 4.327 .000
a. Dependent Variable: Satisfaction
From the table 5 it is clear that all the variables have meaningful results except for
“polite answer”, which α level > 0.05 and cannot be used. It means that polite answer has no impact on satisfaction while other variables have. Therefore, H_1.1 can be accepted, excluding
“polite answer” from this model.
It can be concluded that more helpful and prompt answers can reach higher satisfaction level. Moreover, the less time the receptionist needs for contacting with the client, the more the client feels satisfied. Also, longer queue can badly influence on satisfaction, therefore, the language center needs to arrange more space for the students or arrange more staff at the front desk.
2. Receptionist
In this section the author expects to find some relationship between receptionist and satisfaction level, therefore, first, it starts with checking the association level between them.
There are totally 12 employees being observed at information center and as the author herself helped to communicate with the students, she also included herself into the dataset. From the
23
table 6 it is shown that person 1, 2 and 4 more than others helped the student (excluding the author). According to the researcher’s observation person 1 and 2 are most often have been seen at the information desk, therefore, they got more observed cases. Person 4 has a lot of
communications with the students, because she is responsible for payment (check the table
“Nature of student’s problem” for more details), and this type of problem has been recorded 63 times, which is an absolute maximum among the problems.
Table 6 H_1.1 Receptionist's name
Frequency Percent Valid Percent
Cumulative Percent
Valid 0 35 13.7 13.7 13.7
person 1 56 21.9 21.9 35.5
person 2 23 9.0 9.0 44.5
person 3 2 .8 .8 45.3
person 4 53 20.7 20.7 66.0
person 5 10 3.9 3.9 69.9
person 6 23 9.0 9.0 78.9
person 7 12 4.7 4.7 83.6
person 8 14 5.5 5.5 89.1
person 9 3 1.2 1.2 90.2
person 10 7 2.7 2.7 93.0
person 11 6 2.3 2.3 95.3
person 12 12 4.7 4.7 100.0
Total 256 100.0 100.0
(H_2.1) There is a strong association among satisfaction level, helpful answer, prompt answer and polite answer controlled by receptionist.
24
Table 7 H_1.1 Correlations
Control Variables Satisfaction
Helpful
Satisfaction Correlation 1,000 .808 .614 .464
Significance
From the table 7 all the data illustrates highly significant results, but the first row is the most important for this research. It shows the more helpful answer for a client, the higher level of satisfaction a client can have comparing it within receptionist variable set. Correlation shows 0.808, which is quite close to 1 and represents high correlation between the variables. Prompt answer score cannot be counted as highly correlated, however, there is still a positive correlation among the variables and as the result is >0.5, thereafter, this variable can also prove the
hypothesis alongside with helpful answer. Even though polite answer has high significance level, however, it is score <0.5, then it again shows less impact on satisfaction, and this variable cannot prove this hypothesis.
To sum it up, H_2.1 can be accepted, excluding “Polite answer” variable from the final results.
25
(H_2.2) Satisfaction level, helpful answer, prompt answer and polite answer differ in terms of who provided the service (receptionist).
For testing this hypothesis, it is required to run one-way ANOVA, the results are as follows:
Helpful answer Between
Groups 35.008 12 2.917 3.049 .001
Polite answer Between
Groups 41.199 12 3.433 7.260 .000
Within Groups 114.910 243 .473
Total 156.109 255
All the variables from Table 8 show highly significant results, because α level < 0.05.
ANOVA states that there is at least one pair of means showing significantly different results, therefore, among all 13 receptionists there is at least 2 of them whose data are different in terms of satisfaction, helpful / polite / prompt answers. This hypothesis can be accepted.
(H_2.3) The satisfaction level can be explained by helpful answer, prompt answer, polite answer and the receptionist who provided the service.
To check how satisfaction can be explained by other variables, it is necessary to run regression. In this case variable “receptionist” is not metric, therefore, it cannot be directly entered into the model, it is advised to add dummy variables. According to the formula (n-1), the
26
dataset requires to add 12 dummy variables. All the date will be compared with the first receptionist, who is the author herself, it is coded 0 and is not entered into variables list.
It is advised to use “enter” method, as it helps to check all the variable and it does not exclude them from the model.
Table 9 H_2.3 Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .844a .712 .694 .595
a. Predictors: (Constant), Person 12, Person 3, Person 9, Person 11, Person 10, Person 5, Person 8, Person 8, Person 6, Person 2, Prompt answer, Person 4, Polite answer, Person 1, Helpful answer
From Table 9 both R and R square are high enough for further analyses, as both of them are close to 1.
Table 10 H_2.3 ANOVA
Model Sum of Squares df Mean Square F Sig.
1 Regression 210.384 15 14.026 39.577 .000b
Residual 85.053 240 .354
Total 295.437 255
a. Dependent Variable: satisfaction (1 to 7)
b. Predictors: (Constant), Person 12, Person 3, Person 9, Person 11, Person 10, Person 5, Person 8, Person 8, Person 6, Person 2, Prompt answer, Person 4, Polite answer, Person 1, Helpful answer
Table 10 shows that the level of significance is high and the data can be analyzed. The next table will show how the data distributed among the variables. In this model it is used 12 dummy variables (out of 13 receptionists).
27
Table 11 H_2.3 Coefficients
Model
a. Dependent Variable: Satisfaction
From the table 11 except for polite answer, prompt and helpful answers show significant results, meaning that only they can explain satisfaction level in terms of mentioned above
receptionists. Moreover, most of the receptionist do not show any significant difference in terms of satisfaction level. There is person 1, 2 and 12 only can be different by how the clients feel satisfied with the service, as their significance level < 0.5.
There is high probability why other receptionists do not have any difference, it might happen because sample size is not enough for analyzing the data. Nevertheless, this hypothesis should be rejected and null hypothesis should be accepted, meaning that satisfaction level cannot be explained by helpful answer, prompt answer, polite answer and the receptionist who provided the service.
From this section it can be concluded that satisfaction level differs in terms of who provided the service, because there is a strong correlation. However, it cannot be fully explained
28
by receptionist, as there are other factors, such as helpful and prompt answer, which are more important in service delivery process.
3. Language
After checking the relationship between receptionist and satisfaction, it is possible to find whether language barrier influenced on service delivery’s perception.
(H_3.1) There is a strong association among satisfaction level, helpful answer, prompt answer and polite answer controlled by the language used during service delivery.
Table 12 H_3.1 Correlations
Control Variables Satisfaction
The results in Table 12 show quite same results as hypothesis H_2.1, presenting that if language is controlled, there is high correlation between satisfaction and helpful answer (0.807) and between satisfaction and prompt answer (0.609). Since polite answer shows low positive correlation 0.466 (< 0.5) it is advised not to include into the final results. Summarizing all the notes above, H_3.1 can be accepted, however “Polite answer” is advised not to include into the results.
(H_3.2) Satisfaction level, helpful answer, prompt answer and polite answer differ in terms of what language is used during service delivery.
29
Table 13 H_3.2 ANOVA
Sum of Squares df Mean Square F Sig.
Satisfaction Between Groups 6.297 2 3.149 2.755 .066
Within Groups 289.140 253 1.143
Total 295.438 255
Helpful answer Between Groups 6.581 2 3.290 3.190 .043
Within Groups 260.947 253 1.031
Total 267.527 255
Prompt answer Between Groups 27.223 2 13.612 6.080 .003
Within Groups 566.386 253 2.239
Total 593.609 255
Polite answer Between Groups 2.602 2 1.301 2.144 .119
Within Groups 153.508 253 .607
Total 156.109 255
Unfortunately, after checking the table 13, only prompt answer shows highly significant results (0.003<0.05), by meaning that promptness depends on the language used during
conversation, however, it does not have any impact on satisfaction, meaning that satisfaction level does not change in terms of chosen language, only change by the speed of the answer.
There is also one more variable which shows slightly significant result, it is helpful answer (0.043<0.05). It might mean that using one language might be easier for a staff (or student) to provide an answer (or listen to the answer). Therefore, this hypothesis should be changed and accepted as helpful answer and prompt answer differ in terms of what language is used during service delivery. Besides, language does not affect entire satisfaction (which is important) and has no connection with polite answer. Speed and helpfulness can be kept in mind for further discussion.
(H_3.3) The satisfaction level can be explained by helpful answer, prompt answer, polite answer and the language used during service delivery.
Regression is required to test this hypothesis, however, language cannot be directly used for this test, because it is not a continuous variable. Firstly, language should be coded into
dummies. During observation there have been used 2 languages: English and Chinese, and also mixture of English and Chinese during the same conversation. The last one should not be entered
30
into the model, and other two variables could be compared with it. As a result, there are 2 dummies: D1 – English and D2 – Chinese.
Table 14 H_3.3 Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .818a .669 .662 .625
a. Predictors: (Constant), Chinese, Polite answer, Prompt answer, Helpful answer, English
From table 14 both R and R square are less than in previous cases, however, these results can still be used for further analyses.
Table 15 H_3.3 ANOVA
b. Predictors: (Constant), Chinese, Polite answer, Prompt answer, Helpful answer, English
ANOVA shows (check table 15) that the data can be used for further testing, there is a difference in the means.
Table 16 H_3.3 Coefficients
Model
31
The main purpose of this test (Table 16) is to show the difference in satisfaction by using 2 languages, however, as the data shows highly insignificant results (both > 0.05), therefore, this hypothesis should be rejected and null hypothesis should be accepted: the satisfaction level cannot be explained by helpful answer, prompt answer, polite answer and the language used during service delivery.
4. Nature of the problem
The last section remained is the problem’s type, the aim is to test hypotheses in this section and to find some differences in satisfaction level among the nature of the problems.
Table 17 Nature of students' problem
Frequency Percent
As it is shown in Table 17, there are totally 14 problems that have been recorded, plus 1 problem which is unstated and does not have any naming, it will be considered as missing value in some analyses. The most common problem happened is “Payment”, following by “Study at MTC for newcomers” and then by “Personal problem”. Payment includes payment itself and getting confirmation that the office received student’s payment. Study at MTC for newcomers
32
includes a variety of problems: applying for the courses, MTC online, one-on-one courses, short-term courses, questions about studying process, summer camp, and etc. Those question are combined together under general idea of studying at MTC. Personal problem also has a large scale of questions: bank account, changing some personal information in school database, printing, forgetting some personal belongings at school, and etc.
(H_4.1) Satisfaction level, helpful answer, prompt answer and polite answer differ in terms of what problem the students faced.
Table 18 H_4.1 ANOVA
Sum of Squares df Mean Square F Sig.
Satisfaction Between Groups 74.632 14 5.331 5.818 .000
Within Groups 220.806 241 .916
Total 295.438 255
Helpful answer Between Groups 68.872 14 4.919 5.968 .000
Within Groups 198.656 241 .824
Total 267.527 255
Prompt answer Between Groups 146.462 14 10.462 5.638 .000
Within Groups 447.148 241 1.855
Total 593.609 255
Polite answer Between Groups 46.465 14 3.319 7.295 .000
Within Groups 109.644 241 .455
Total 156.109 255
From Table 18 all the variables show highly significant level (< 0.05) and therefore it means that there at least on pair of problems which are different in satisfaction / helpful / prompt / polite answer. As there are no restrictions in this analysis, even polite answer has some
meaningful mean difference, it is concluded as accepting hypothesis H_4.1.
(H_4.2) The satisfaction level can be explained by helpful answer, prompt answer, polite answer and the problem’s type the students faced.
Same with previous regression analyses, it is compulsory to change nonmetric variables into metric ones by using dummies. In this case, problem “unstated”, which is initially coded as
“99”, would not be entered into the model and would be considered as missing value. This type
33
of problem was not properly observed by the author, but it was not excluded from the data set, as it still has some meaningful data for language and receptionist’s sections. Therefore, after
excluding “unstated”, there are 14 types of problems, and the first problem related to the card would be counted as 0 and not be included into regression. All other 13 problems would be compared with “Card problem”.
Table 19 H_4.2 Model Summary
Model R R Square
Adjusted R
Square Std. Error of the Estimate
1 .843a .711 .691 .598
a. Predictors: (Constant), Newcomers, Prompt answer, TOCFL, General, Scholarship, Attendance, Courses, Refund, Docs, ARC & Visa, Stamp, Group, Personal, Polite, Helpful answer, Payment
From Table 19 R and R square are quite high and represent meaningful results.
Table 20 H_4.2 ANOVA
Model Sum of Squares df Mean Square F Sig.
1 Regression 209.964 16 13.123 36.694 .000b
Residual 85.473 239 .358
Total 295.437 255
a. Dependent Variable: Satisfaction
b. Predictors: (Constant), Newcomers, Prompt answer, TOCFL, General, Scholarship, Attendance, Courses, Refund, Docs, ARC & Visa, Stamp, Group, Personal, Polite, Helpful answer, Payment
According to the Table 20, the results are highly significant, meaning there is at least one pair has significantly different means.
Despite all the tables are advisable for further testing, but significance level for majority of problems are > 0.05 (Please, check Table 21). There are only few types, for example: general, payment and personal problems that can be differentiated in terms of satisfaction level. It might be explained that for this problems staff has more structured guidelines or they are more
prepared for working on these problems. The reason why others do not show significant results might be the size of sample, it is too small for comparing. The data is not enough for the program
34
to compare results, therefore, it is advised to keep in mind this hypothesis for further researchers.
Nevertheless, this time H_4.2 should be rejected and null hypothesis should be accepted as follows: the satisfaction level cannot be explained by helpful answer, prompt answer, polite answer and the problem’s type the students faced.
Table 21 H_4.2 Coefficients
Model
a. Dependent Variable: Satisfaction
(H_4.3) There is a strong association among satisfaction level, helpful answer, prompt answer, polite answer and number of employees involved into communication process and explained by the problem’s type.
From the table 22 under normal conditions there are 1~2 persons being responsible for communicating with a student, however, there are 31 cases when 3 persons have been involved and even 4 cases with 4 persons.
35
Table 22 How many employees were involved into solving the problem
Number of
employees Frequency Percent Valid Percent
Cumulative
Control Variables Satisfaction
Helpful
36
The aim of the table 23 is test the correlation between number of employees and satisfaction, controlled by the problem’s type. All the data are highly significant and can be analyzed. First of all, it is necessary to check the variable “Number of employees”, its r coefficient is extremely low (-0.217), so the correlation with satisfaction is not strong enough.
However, as it was expected, it can be noted, that “Number of employees” has negative correlation with satisfaction, meaning that the less people involved into delivering process, the higher level of satisfaction can be. Number of employees also has low negative correlation with helpful / prompt / polite answer, therefore, it is better to communicate with less people for solving one problem. However, the correlation is not strong enough.
Helpful answer shows the highest level of correlation (0.811) with satisfaction. It means that even among the different problem helpful answer is always highly correlated with
satisfaction.
After checking all these results, the author advises to accept this hypothesis, by
emphasizing that there is low correlation among number of employees and all other variables in terms of problem types.