You are on page 1of 10

Summary of the case:

AgriComp is a supplier of computer systems for the farmers. This system can be useful
for purposes as crop rotation planning and spreadsheet analysis for financial planning.
Both the hardware and software of this system were sold through their 350 affiliated
dealers nationwide. Currently local dealers handle warranty services for customers via
local repair followed by a reimbursement claim to AgriComp. Denied claims follow an
internal company appeal process. But recently the company is facing several problems
in settling the reimbursement claim of the dealers as they are re!uiring the advantage
for particular repair that was not covered by the company"s warranty. As a result dealers
have been complaining about the fairness of the appeal process. To resolve this
problem the company has surveyed its dealers and collected the dealer"s support for
replacing the e#isting warranty appeals system with a mediator system $%&'( and the
number of times the dealer used the appeals process $)*&(. +ow it is assigned us to
review survey results of those two variables and determine whether the costly e#ternal
mediator process would be worth implementing to ,eep the dealers happy.

Objective of the analysis:
The main problem !uestion of this case is-
.hether the company should maintain their e#isting internal appeal process or
should replace it with an e#ternal mediator process/
To get the answers of this problem !uestion the relation between the two variables i.e.
%&' $The dealer"s support for replacing the e#isting warranty appeals system with a
mediator system( )*& $The number of times the dealer used the appeals process(
were tried to be developed. 0or this purpose we have used some statistical tools and
tried to derive-
The correlation between the variables measure of sampling ade!uacy $123( and
percentage of variance
The relation between the individual )*& and their average %esponse rate for the
corresponding )*& and the deviation of this average %&'
Average and weighted average of %&' depending on the relationship of )*& with
average %&'
The percentage of different level of dealer"s response $%&'( and the relationship
between the levels of response and their corresponding )*&
The correlation between these %&' and )*&
The regression between %&' and )*&
Data analysis methodology:
This case analysis is basically dependent on the numerical primary data found from the
survey of 454 dealer respondents. And there were two variables. These are-
6. %&' $The dealer"s support for replacing the e#isting warranty appeals system with a
mediator system(7 where6 8 9*trongly agree: 48 ;"Agree""3 8 9+either agree nor
disagree: <8:Disagree: 5 8 9*trongly disagree:
4. )*& $The number of times the dealer used the appeals process( were used7 it stands
from =ero to 3 or more
To determine the relationship between two variables several statistical analyses has
been performed. These are given bellow-
Factor analysis: To measure the correlation between the variables measure of
sampling ade!uacy $123( and percentage of variance the factor analysis was done
in *tatistical 'ac,age for *ocial *cience $*'**( 6>.0
Average calculation: To get an idea about the average response level the
arithmetic mean of %&' is calculated. ?n addition to this the relation between the
individual )*& and their average %esponse rate is derived by calculating arithmetic
mean of %&' for individual uses time. 0or calculating the arithmetic mean we have
used the following formula-

@ere A 8 %&' for a particular respondent of an individual )*&
+ 8 Total number of respondent for those indentified %&'
The weighted average of the %&' also calculated to get a better result of the average
%&' as the normal arithmetic mean can be affected by the different weight of each
)*& category. The weighted average is calculated by using the following formula-

Standard deviation: To determine how the averages calculated for every )*&
category is e#pressing the data set7 the standard deviation for each average %&' is
calculated. The standard deviation basically shows the deviation of the individual
data from its average and it has been calculated with the following formula-
@ere Wi 8 ?ndividual weight of )*&
Xi8 Average %&' for particular )*&

Correlation analysis: To determine whether the %&' and )se has positive or
negetive relationship and what is the intensity of this relationship the corelation
analysis has been performed. ?n the corelation the following formula has been used-
Regreession analysis: To do a hypothesis testing on 9%&' has no relationship with
)*&: a regression analysis between those two variables were performed. The
overall calculation is based on the following formula-
Bit 8 CiD E Ait D eit
.here
F Bit is dealer"s response level $%&'(
F Ci refers to timeGinvariant firmGspecific effects
F Ait is the independent variable $)*&(
F E coefficients
F eit is a random disturbance.
Solution:
0or solving the case a systematic process is followed. ?t is included in the following step
by step-
Factor analysis: This survey had used two variables i.e. response for replacing the
e#isting warranty appeals system with a mediator system $%&'( and the number of
times the dealer used the appeals process $)*&( for the 454 respondents. Two
determine the correlation between the variables measure of sampling ade!uacy
$123( and percentage of variance of the variables the factor analysis of the
variables is done. The result of this analysis is given bellow-
Hariable 0actor Correlation 2easure of
sampling
ade!uacy$ 123
I of variance
@ere Ai8 ?ndividual %&'
J8 Average %&' for a particular )*&
(
4 6 .<44 .5 Hariabl
e 6
Hariabl
e 4
K6.666 4L.LL5
Table-01: Factor analysis of REP and USE
@ere the results shows that the %&' and )*& has a positive coG relation of .<44 which
indicates the dealers number of uses time of the appeal process has a positive
correlation with the response level. That means the dealers who have used the appeal
process more is responding the high levels of %&'. The high level of %&' indicates that
the dealers who have used the appeal system in more time are disagreeing or strongly
disagreeing to replace the e#isting appeal system with the mediator system. 3n the
other hand the measure of sampling ade!uacy $123( of factor analysis indicates that
whether the number of the respondents is sufficient enough to e#press the number of
the variables used and the standard is greater than e!uals to .3. That means if the
result of the 123 is derived .3 or greater than this than the respondent number will be
sufficient enough to e#press two variables properly. As here the result is found .5M.3 so
we can say that in this survey the 454 respondent is sufficient enough to properly
determine the variation of %&' and )*&.
3n the other hand the percentage of variance calculated in for Hariable one that means
%&' is K6.66I which means the overall characteristics of this data set that includes
the correlation regression variation is effected K6.66I by the dealer"s response levels
$ %&'( and the other 4L.LL5I is effected by the number of the uses time of the dealer
$)*&( .
ealer!s res"onse level #REP$ de"ending on t%e nu&ber of t%e usage ti&e of
e'isting "rocess: To determine that whether the appeal process should be
replaced by the e#ternal mediator system depending on the dealer"s response level
$%&'( and using time $)*&( it is important to ,now the average %&' average
)*& and the %&' variation with the different )*&. The Average %&' found in that
survey was 3.43 which indicates that on average the dealers are either neutral or
disagree to replace the e#isting system.
To get a more concrete result data was sorted by their uses time $starts from 0G3( and
then their average response level was calculated. Those results have been given
bellow-
Average %&' 3.43
Average )*& 6.55
)*& Total Average %&' Correspondin
g Comment
*tandard
Deviation
0 36 4.445L0> Agree 6.6L3<L
6 KL 4.K>5436 +eutral 6.4L0455
4 5L 3.603<<L +eutral 6.4<63K5
3 645 3.L6> Disagree 6.60550>
%&' weighted
Average
3.45
Table-0(: T%e average) corres"ondent average) *eig%ted average of REP
The average response rate for every category of the no of usage time is shown in the
graph bellow-
C%art-01: Average REP *it% corres"onding USE
0rom the calculation it was found that who have used the e#isting appeal process
more have given the high level which is more or less greater three. That means
depending on the levels they are either neutral or disagree to replace the e#isting
system. 3n the other hand it was also found that the response level is simultaneously
increasing with the uses time. That means the people who have used this appeal
process more is willing to sting with this and not to replace this.
The standard deviation is also calculated for the average response level of every
individual using time. The standard deviation for each classification was found
6.60550>G6.4L0455 which is very low. That means the result found from the average
calculation is not very much deviated which results that the average is moderately
e#pressing the whole data.As the standard deviation is found moderate so the average
is used to calculate the weighted average of the %&' for getting a more perfect result.
The weighted average was found 3.45 which are also indicating that on average the
dealers are neutral to disagree to replace this e#isting system.

ealer!s res"onse level "ercentage and corres"ondent USE: The dealers were
as,ed to response according to a level of6 8 9*trongly agree: 48 ;"Agree""3 8
9+either agree nor disagree: <8:Disagree: 5 8 9*trongly disagree: and based on the
condition of the level the decision of replacing the e#isting appeal process can be
made. The percentage of the different provided by the dealer is given bellow-

%&' *trongly
Agree
Agree +eutral Disagree *trongly
disagree
+o of
respondent
3< >3 >0 K3 >4
'ercentage 66.><I 46.5LI 40.55I 45I 46.43I
Table-0+: Percentage of REP
0or better representation the percentage of response level $%&'( is shown in the pie
chart-
C%art-0(: Percentage of REP
0rom the table and chart it is found that the ma#imum people that are 45I dealer is
disagree to replace the e#isting appeal system. The lowest percentage was found in
level 6 that means only 66.><I people is willing to replace the appeal system with the
mediator system.

0or every percentage of %&' the number of the uses time of the appeal process is
calculated to determine each of that level is dominated by which )*&. The following
table shows the correspondent )*& for every level of response.
%&'
*trongly Agree Agree +eutral Disagree *trongly
Disagree
)*& +umbe
r
I +umbe
r
I +umbe
r
I +umbe
r
I +umbe
r
I
0 64 35.4
5
> 5.54 L 63.3
3
< 5.<L 6 6.>0
6 64 35.4
5
4K <4.L
>
6L 30.0
0
5 64.3
3
64 65.3
>
4 < 66.K
K
6L 4L.5
K
6> 4>.>
K
L 60.5
>
64 65.3
>
3 > 6K.>
5
64 65.0
5
6L 30.0
0
54 K6.4
3
3K 55.>
L
Total 3< 600 >3 600 >0 600 K3 600 >4 600
Table-0,: -o and "ercentage of USE in eac% level
The percentage of the number of uses rate for every level of response is given in the
bar chart-
C%art-0+: Percentage of USE in every level of res"onse
0rom the table it is found that among the people who have strongly agreed were the
dealers who have never used the e#isting appeal process. As they did not use the
e#isting process anytime so complain about the process from them is Nust the
malcontent.
3n the other hand the people who have strongly disagreed to replace the e#isting
system were the highest user that means three time user of the e#isting process. That
means they are satisfied with the e#isting process.
T%e correlation bet*een t%e variables: To determine the nature and degree of
relationship between the using times $)*&( of the response level $%&'( the
correlation analysis is done. The result found from the analysis is given bellow-
%&' )*&
%&' 6 0.<44445
)*& 0.<44445 6
Table-0.: Correlation &atri' bet*een REP and USE
0rom the correlation matri# it is found that the dealer"s response level has a positive
correlation with the no of the uses time of the e#isting process. That means when the
using rate increases than the response level also increases. As a result it can be
concluded that the respondents with ma#imum level of use are giving the higher levels
of response i.e. neutral disagree strongly disagree. 3n the other hand the relationship
between the %&' and )*& is .<4445 which indicates that this two variable has a
moderate positive relationship.
/y"ot%esis testing t%roug% regression analysis: ?n this section to investigate the
relationship among the response level of the dealers $%&'( and the number of uses
time of the e#isting process $)*&( a linear regression is done. ?ndependent variable
was )*& $The number of times the dealer used the appeals process( and the
dependent variable was %&' $The dealer"s support for replacing the e#isting
warranty appeals system with a mediator system(.Depending on the relationship
between the dependent and independent variables there can be identified the
following hypothesis-
+ull @ypothesis- @08%&' has no relationship with )*&
Alternative @ypothesis- @68%&' has a relationship with )*&
The result of the regression between the %&' and )*& is presented in table-
Hariable ?ndependent Dependent
)*& %&'
2ultiple % .<44445
% *!uare .6KL4K<
0 value >4.565>>
?ntercept 4.4044L
Coefficient .5453>L
*tandard &rror .0>>43<
' value <.K4&G6<
Table-00: Regression result bet*een REP and USE
0rom the table it was found that the multiple regressions between the responses level
$%&'( and the no of uses time $)*&( is .<44445 which is indicating that the %&' and
)*& has a positive correlation. The % *!uare value indicates that only 6K.L4I changes
in the %&' can be e#pressed by the no of uses time of the e#isting process $)*&(. And
the 0 value founded from the analysis is >4.56 which is more than the appendi# 0 value
3.L< founded for the degree of freedom of regression 6 and residual value 450. This
results show that the null hypothesis is reNected that means the %&' has a relationship
with )*&. ?n addition to that the intercept value is found 4.4044L which indicate that the
effects of other variables than )*& have a constant value of 4.4044L.
3n the other hand the )*& has a coefficient of .5453>L which is indicating that %&'
and )*& has a positive relationship and with the number of usage time increase by 6
the response level will increase by .5453>L. The standard error in this analysis was .
0>>43< and ' value of <.K4&G6< which is less than .05 indicating that the relationship
between %&' and )*& is not significant.
Findings:
0rom the overall statistical analysis the following things are fond-
The factor analysis shows that the 454 respondents are sufficient enough to
e#press two variables and the overall variation of the data can be e#pressed
K6.666I by response level $%&'( and 4L.LL5I by the no of uses time $)*&(
The average %&' is 3.43 and weighted %&' is 3.45 each of which indicates a
response level of neutral to disagree as the level used in the variable was 6 8
9*trongly agree: 3 8 9+either agree nor disagree: and 5 8 9*trongly disagree:.
That means the dealer"s support for replacing the e#isting warranty appeals
system with a mediator system was disagreed on average.
The average %esponse rate for the individual )*& was found in a positive trend.
That means the level given by the respondents was getting higher with the more
time user of the e#isting process. ?t indicates that the dealer who has used the
e#isting appeal process more is e#perienced with the system and satisfied with
their e#perience.
The percentage calculation of the different level of %&' showed the ma#imum
respondents that are 45I disagree to replace the e#isting appeal system. The
lowest percentage that is only 66.><I respondent is willing to replace the appeal
system with the mediator system.
Along with the percentage of every level it is found that the lower levels is highly
distributed by the lower user people and vice versa. ?n the level of strongly agree
the more respondent was =ero time user that means they have no any
e#perience with the e#isting process. *o their response has no any impact on
the decision. 3n the other hand the in the level of strongly disagree the more
respondent was highest time user that means they are e#perienced and happy
with the e#isting process
The correlation analysis reveals a positive co relation between the %&' and )*&
that means if )*& increase than %&' also increases. The amount of correlation
was found 0.<44445.
?n the regression analysis the null hypothesis was reNected that means it was
proved that the response level has a relationship with number of uses time and
with the number of usage time increase by 6 the response level will increase by .
5453>L.
*o from the analysis it is proved that Agricomp should not replace the e#isting appeal
system and implement costly e#ternal mediator process as the user of this e#isting
system is already happy.

Recommendation:
The dissatisfaction on the e#isting process was mostly found from the non users
so the company can ta,e a step to train up the dealers about the e#isting appeal
system.
The dealer"s should also be trained up on the warranty features provided by the
company and should inform any changes in those features time to time.
Agricomp has got 350affiliated dealers but only 454 of them had responded to
the company"s recent survey. *o the data found from the survey is not properly
reflecting the decisions of the every dealer. The additional 54 dealers can change
the result of the survey. *o Agricomp can do another survey on whole 350
respondents to get the appropriate result.
3nly two variables were used in this survey which is not properly reflecting the
decision. The cost of implementing the e#ternal mediator process should be
calculated with those variables for proper decision ma,ing
Agricomp should maintain the present !uality of the appeal process or improve it
to retain the dealers happy so that no discontent can arise.

You might also like