Professional Documents
Culture Documents
M J XAVIER
PREFACE
This practical guide on data analysis has been prepared specifically for the business
students majoring in marketing who have an aversion for numbers and statistical
methods. The simple step-by-step approach used in the guide should enable students to
gain insight into statistical tools and help them develop their skills in interpreting and
making meaning out of numbers.
The entire range of statistical tools has all been explained using a single data set from a
questionnaire on tooth-paste market. The tools covered range from simple frequencies,
mean, median etc. to multivariate techniques like factor, cluster and discriminant
analysis.
The questionnaire, the code sheet and the final report are all given in the appendix. The
first chapter on simple analytical methods start with SPSS data preparation and go on to
explain the use of descriptive statistics to prepare summary results for each question in
the survey data. It also highlights the use of charts for displaying data. The second
chapter goes into the use of brand rating data for making snake charts, and positioning of
brand using factor analysis. The third chapter introduces the concept of correlation
coefficient and its sue for getting derived importance weights used for construction of
Kano diagram.
Chapter 4 uses the importance scores for benefits to do benefit segmentation using cluster
analysis. Chapter 5 introduces Correspondence analysis and its use for mapping brand-
personality association data.
The data files referred to in the text are all available in the attached 3½” disk.
I am grateful to the Graduate and undergraduate students who enrolled for the marketing
research course during the Fall 2003 term for their co-operation in developing the
questionnaire and collection of data. I am grateful to Dr. R Krishnan, Director Graduate
Program and Dr. Norm Borin, Marketing Area Chair for their support and
encouragement for this project.
M J Xavier
December 2003
1
CONTENTS
2
Chapter 1
INTRODUCTION AND SIMPLE ANALYTICAL METHODS
SESSION OBJECTIVES:
At the bottom left hand corner you will see TWO BUTTONS:
DATA VIEW AND VARIABLE VIEW
Click on the variable view. You will see the complete definition of each variable.
3
STRING VS NUMERIC VARIABLES
Go to Data View and see that String variables are made up of letters or letters and
numbers (alpha numeric) while the numeric variables are made up of numbers only.
Go to variable view again and study the columns LABEL and VALUES.
Click at the right hand corner of VALUES corresponding to the VARIABLE `class’
A small window shown below will open.
VIEW
VALUE LABEL
You will notice that labels corresponding to the numerical codes appear on the data sheet.
4
FREQUENCIES AND CHARTS:
Let us start with the age profile. Run the following SPSS Commands to get the
distribution of respondents by age.
ANALYZE
DESCRIPTIVE STATAISTICS
FREQUENCIES
Drag variable `Age[q12a]’ on to the VARIABLE(S) Box
CHART
PIE CHARTS
PERCENTAGES
CONTINUE
OK
Check if you get the following table and the Pie chart from in a new window.
Age
Cumulative
Frequency Percent Valid Percent Percent
Valid Under 18
58 82.9 82.9 82.9
years
18-24 years 12 17.1 17.1 100.0
Total 70 100.0 100.0
5
Age
18-24 years
17.1%
Under 18 years
82.9%
Now repeat the analysis with other demographic variables, namely Household Income,
Gender, and Race. You can drag all three variables to the variables box and have the
charts made simultaneously.
Now do the frequencies with other variables, awareness of Brands and also with trial of
brands.
Now go the variable Current Brand and change the chart from PIE to Bar and see if you
get the following chart.
6
Current Brand
40
30
20
10
Percent
0
Aqua-Fresh Crest Arm & Hammer
Colgate Mentadent Others
Current Brand
Try the following commands to create a new variable called aware which is derived from
other variables.
TRANSFORM
COMPUTE
Type `aware’ in the TARGET VARIABLE Box
Move variable `q01a’ into the NUMERIC EXPRESSION Box
Click on +
Move variable `q01b’ into the NUMERIC EXPRESSION Box
Click on +
Move variable `q01c’ into the NUMERIC EXPRESSION Box
Click on +
Move variable `q01d’ into the NUMERIC EXPRESSION Box
Click on +
Move variable `q01e’ into the NUMERIC EXPRESSION Box
Click on +
Move variable `q01f’ into the NUMERIC EXPRESSION Box
Click on +
7
Move variable `q01g’ into the NUMERIC EXPRESSION Box
Click on +
OK
Notice that a new column has been created by SPSS called `aware’. While the original
variables are called raw variables, the new one formed out of raw varaiables is called a
transformed variable.
Go to variable view and type `No. Of Brands Aware’ in the LABEL column
corresponding to the variable `aware’
Now perform a Frequency analysis on the new variable `aware’ and get a bar chart as
shown below.
Cumulative
Frequency Percent Valid Percent Percent
Valid 2.00 1 1.4 1.4 1.4
3.00 4 5.7 5.7 7.1
4.00 30 42.9 42.9 50.0
5.00 28 40.0 40.0 90.0
6.00 6 8.6 8.6 98.6
7.00 1 1.4 1.4 100.0
Total 70 100.0 100.0
8
No. Of Brands Aware
50
40
30
20
10
Percent
0
2.00 3.00 4.00 5.00 6.00 7.00
In the same way create a new variable called trial (no. Of brands tried by each person)
using the following expression.
Note that the two new transformed variables, namely, aware and trial are different from
the other variables we have seen earlier. These are ratio scaled variables whereas the
other variables are only nominally scaled.
We shall see how to calculate mean, standard deviation and variance for a ratio scaled
data.
ANALYZE
DESCRIPTIVE STATISTICS
DESCRIPTIVES
drag the variable `aware’ to VARAIABLES Box
OPTIONS
9
VARIANCE
CONTINUE
OK
Descriptive Statistics
This shows that on an average a respondent is aware of 4.53 brands and the standard
deviation of the same variable is 0.85.
Suppose we want to know whether the use of a particular brand depends on whether the
person is a male or female, we need to use the type of analysis called cross-tabulation.
Crosstabs is used to explore the relationship between two nominal or categorical
variables.
ANALYZE
DESCRIPTIVE STATITICS
CROSSTABS
Drag Gender to ROW(S)
Drag Current Brand to COLUMN(S)
CELLS
ROW
CONTINUE
STATISTICS
CHI-SQUARE
OK
10
Gender * Current Brand Crosstabulation
Current Brand
Aqua- Mentade Arm &
Fresh Colgate Crest nt Hammer Others Total
Gender Male Count 7 9 8 3 3 5 35
% within
20.0% 25.7% 22.9% 8.6% 8.6% 14.3% 100.0%
Gender
Female Count 3 11 17 3 0 1 35
% within
8.6% 31.4% 48.6% 8.6% .0% 2.9% 100.0%
Gender
Total Count 10 20 25 6 3 6 70
% within
14.3% 28.6% 35.7% 8.6% 4.3% 8.6% 100.0%
Gender
It is a convention to keep the independent variable in the row, dependent variable in the
column and get row percentages in the cells. In this case, we are trying to explore
whether gender has an impact on brand choice.
To interpret the Table always look column wise and see if the percentages vary
drastically. In the Aquafresh column, there is a larger percentage of males. Colgate has
marginally large percentages of females. Crest has substantially large percentage of
females compared to males. Mentadent has equal following among males and females.
Arm & Hammer appears to be an exclusive male brand.
There appear to be some relationship between gender and brand used. To check whether
the relationship is significant or not, we need to look at the chi-square value.
Chi-Square Tests
Asymp. Sig.
Value df (2-sided)
Pearson Chi-Square 10.707(a) 5 .058
Likelihood Ratio 12.230 5 .032
Linear-by-Linear
1.452 1 .228
Association
N of Valid Cases
70
a 6 cells (50.0%) have expected count less than 5. The minimum expected count is 1.50.
Note that the cell frequencies should be 5 or more for chi-square test. However SPSS
applies a correction factor to take care of this deficiency and makes it all the more
difficult to attain significance.
11
The same rule for the significance level of 0.05 or less applies to all the tests that we are
going to learn, be it t-test, f-test or any other test.
The same way construct cross tabs for age, income, and race against Current Brand and
check if the relationships are significant using chi-square values.
COMPARE MEANS
Suppose we want to know whether the mean number of brands aware across males and
females, we could use the following commands.
ANALYZE
COMPARE MEANS
MEANS
Drag variable `aware’ to DEPENDENT LIST
gender to INDEPENDENT LIST
OK
Report
4.6 and 4.2 are very close values. The difference between Male and Female appears to be
very marginal.
Now do the analysis with other classification variables, race, income, age and class.
INDEPENDENT t-TEST:
Suppose we want to know whether the difference of 0.4 in the number of brand aware of
between male and female populations is statistically significant, we need to conduct a t-
test.
12
Run the following SPSS Commands
ANALYZE
COMPARE MEANS
INDEPENDENT SAMPLES t-TEST
Drag variable aware to the TEST VARIABLE(S)
Drag variable q12c to GROUPING VARIABLE
DEFINE GROUPS
GROUP 1 (Type 1)
GROUP 2 (Type 2)
CONTINUE
OK
Levene's Test
for Equality of
Variances t-test for Equality of Means
95%
Confidence
Std. Interval of the
Mean Error Difference
Sig. (2- Differe Differe
F Sig. t df tailed) nce nce Lower Upper
No. Of Equal -
.6038
Brands variances 3.608 .062 .988 68 .327 .2000 .20239 .2038
6
Aware assumed 6
Equal
-
variances 58.53 .6050
.988 .327 .2000 .20239 .2050
not 5 4
4
assumed
Our Null Hypothesis is that the means for males and females are the same. The alternate
Hypothesis is that the means are significantly different. As we do not know which one
should be greater we use a 2-tailled significance test.
Notice that the t-value of 0.988 at 68 degrees of freedom has significance of only 0.327.
The corresponding confidence level is 67.3% which is too low. Unless the significance
level is less than 0.05 the mean values are not significantly different.
Note that the degrees of freedom in this case are No. of observations minus two.
Conduct the t-test for other variables – age, race and income to see whether the mean
brands aware of vary by any of these categories.
13
Chapter - 2
Snake Chart, Factor Analysis and Brand Positioning
Objectives:
To understand how to compute mean ratings of brands and to construct snake charts
To learn how to run factor analysis and understand the following concepts
- variance explained
- factor loading
- eigen value
- communality
- rotation
- factor score
Data Structure:
Study the file structure. This new file has been created out of the master data file by re-
arranging the variables q06a01 to q06d11 as indicated below.
q06a01 …... q06a11 q06b01 …... q06b11 q06c01 …... q06c11 q06d01 …... q06d11
1. 1. 1. 1.
2. 2. 2. 2.
. . . .
. . . .
. . . .
. . . .
. . . .
70 70 70 70
Original Data
14
Brand q06a01 …... q06a11
Code
Data
Note that blank rows have been deleted in the new file and a new variable brand code has
been created.
Snake Chart:
We need to calculate the mean ratings of brands, before we can construct the snake chart.
Use the following commands to obtain the mean ratings.
ANALYZE
COMPARE MEANS
MEANS
Highlight and Drag `q06_01…q06_10’ to DEPENDENT LIST
Highlight `brand’ and drag to INDEPENDENT LIST
OPTIONS
Uncheck NUMBER OF CASES
Uncheck STANDARAD DEVIATION
CONTINUE
OK
15
Fighting Whitening Cleaning Good Likeable Freshening Brand Attractive Innovative
Brand Cavities Teeth Stains Taste Flavor Breath Image Color Packaging features
Aqua-Fresh 7.2 6.1 6.7 6.8 6.9 7.4 6.6 7.0 7.1 6.4
Colgate 8.1 6.9 7.5 6.9 7.0 7.7 7.9 7.0 7.0 6.8
Crest 8.4 7.6 7.9 8.2 8.0 8.3 8.7 7.8 7.8 7.4
Mentadent 9.3 8.1 8.9 9.3 9.6 9.6 7.9 8.7 8.3 7.4
Arm &
Hammer 9.0 8.3 8.3 6.0 5.0 8.7 7.3 7.7 7.3 8.0
Others 8.6 6.2 6.8 9.0 8.8 9.4 5.8 7.2 6.0 8.4
Right click on the table, copy and paste onto an excel worksheet.
Highlight the relevant portions and click Chart Wizard. Choose Line, press Next and
Finish to get the following snake chart.
12.0
10.0
8.0
6.0
4.0 Aqua-Fresh
2.0 Colgate
0.0 Crest
Mentadent
s
ng vor
h
s
at Pac lor
h
e
r
nt
rta
itie
et
st
ag
at
in
Br Br e
Te
e s kag
a
Im
av
ed
Fl
/T
Li o d
ng n g
W gC
gr
ns
e s ble
Others
an
/In
o
ni
ai
ni
t in
G
te
St
ke
ive
he
gh
hi
ur
ct
Fi
va tra
ni
Fr
fe
ea
At
e
Cl
t iv
no
In
This chart can be used to study the relative positioning of brands on different attributes.
We can see that Mentadent and Arm&Hammer are rated highly on selected attributes
while Crest scores consistently higher rating on all the attributes. Aquafresh has lower
rating and Colgate is stuck in the middle.
Here the points are very cluttered and it is difficult to see finer distinctions. Factor
analysis will help us do sharper positioning.
16
Factor Analysis
ANALYZE
DATA REDUCTION
FACTOR
Highlight and Drag `q06_01…q06_10’ to VARIABLES
DESCRIPTIVES
Check COEFFICIENTS
CONTINUE
EXTRACTION
SCREE PLOT
CONTINUE
ROTATION
VARIMAX
CONTINUE
SCORES
SAVE VARIABLES
CONTINUE
OPTIONS
SORTED BY SIZE
CONTINUE
OK
Take a look at the Correlation Matrix and notice that the variables are correlated among
themselves. For example the correlation between good taste and likeable flavor is as high
as 0.928. The correlations are sufficient for conducting a factor analysis is confirmed by
Bartlett’s Test of Sphericity which is significant.
17
Correlation Matrix
Attracti
Likeabl ve
Fighting Whitening Cleaning Good e Freshenin Brand Packag Innovativ
Cavities Teeth Stains Taste Flavor g Breath Image Color ing e features
Fighting
Cavities
1.000 .385 .610 .221 .203 .450 .367 .142 .210 .228
Whitening
Teeth
.385 1.000 .613 .278 .223 .403 .354 .198 .249 .501
Cleaning
Stains/Tartar
.610 .613 1.000 .218 .161 .470 .461 .188 .314 .488
Good Taste .221 .278 .218 1.000 .928 .475 .313 .436 .303 .366
Likeable
Flavor
.203 .223 .161 .928 1.000 .481 .276 .436 .300 .318
Freshening
Breath
.450 .403 .470 .475 .481 1.000 .384 .349 .358 .390
Brand Image .367 .354 .461 .313 .276 .384 1.000 .576 .624 .373
Color .142 .198 .188 .436 .436 .349 .576 1.000 .650 .419
Attractive
Packaging
.210 .249 .314 .303 .300 .358 .624 .650 1.000 .413
Innovative
features .228 .501 .488 .366 .318 .390 .373 .419 .413 1.000
Read Component as Factors in the Table. Technically the 10 original variables can be
converted into 10 new factors which are orthogonal to each other (i.e. will have zero
correlation among them). The first such factor will account for 44.378 percent variance in
the original data, second one will account for 15.773 percent and so on. In statistics
variance is information. As 72.543 percent of information (variance) is summarized by
three variables, it is enough to work with three factors.
We shall see what does the eigen value and rotation mean later.
18
Now go to the data view in the SPSS data file. You will notice that three new variables,
namely, fact1_1, fact2_1 and fact3_1 have been added by the system. The values that
these variables take are called factor scores. Basically the original 10 inter-correlated
variables have been converted to 3 new factors which are orthogonal to each other. To
check the orthogonality do the following analysis.
ANALYSE
CORRELATE
BIVARIATE
Highlight and drag `fact1_1, fact2_1 and fact3_1’ to VARIABLES
OK
You will get the following output which shows that the factors have zero correlation between them.
Correlations
Now the problem is to find out what these factors mean. Obviously the three new factors
summarize the information present in the original ten variables. We need to establish
which variables go into which factor.
Look at the rotated component matrix
19
Rotated Component Matrix(a)
Component
1 2 3
Cleaning Stains/Tartar .878 .203 .014
Fighting Cavities .763 .051 .103
Whitening Teeth .758 .150 .126
Freshening Breath .550 .220 .484
Innovative
features/Ingredients .482 .437 .234
Attractive Packaging .154 .867 .115
Color .013 .830 .323
Brand Image .364 .757 .077
Likeable Flavor .098 .183 .950
Good Taste .152 .195 .933
The cells contain factor loadings, i.e. correlation coefficients of original variables with
the new factors. Conduct the following analysis to confirm the above statement.
Conduct a correlational Analysis of the first variable Cleaning of Stains/Tartar with the
three new factors (fact1_1, fact2_1 and fact3_1) to get the first row in the rotated
component matrix.
The variable `cleaning stains/tartar has correlation coefficient of 0.878 with the first
factor, 0.203 with the second factor and 0.014 with factor 3. What it means is that the
variable `cleaning stains/tartar’ belongs to first factor. The same way the variables
highlighted in the column corresponding to factor 1 belong to the same factor. For the
moment ignore the variable `innovative features/Ingredients’ as it is highlighted in two
columns.
Looking at the variables that go into each factor we can name them as Dental Hygiene,
Visibility and Sensory Benefits.
The variable `Innovative features/ingredients’ has a high correlation with Dental Hygiene
as well as visibility. Suppose a brand claims in its advertisements that it has a new
ingredient that whitens the teeth, it contributes to Dental Hygiene as well the visibility of
the brand. That is how it features in two factors.
20
Now take a look at the communalities matrix.
Communalities
Initial Extraction
Fighting Cavities 1.000 .596
Whitening Teeth 1.000 .613
Cleaning Stains/Tartar 1.000 .812
Good Taste 1.000 .931
Likeable Flavor 1.000 .945
Freshening Breath 1.000 .585
Brand Image 1.000 .712
Color 1.000 .794
Attractive Packaging 1.000 .789
Innovative
features/Ingredients 1.000 .478
Communalities refer to the amount of information that has been extracted from each
variable. Notice that more than 90 percent of information (variance) has been extracted
from variables Good taste and Likeable Flavor whereas less than 50% is extracted from
Innovative features/Ingredients. If we work with large number of variables, say more
than 20, it may be a good idea to leave out variables with low communality while naming
factors.
In the same way the Eigen values are directly proportional to the amount of variance
explained by each factor. The sum of all Eigen values always equals the total number of
variables. Hence the proportion of variance explained by each factor can be calculated by
dviding the corresponding eigen value by the total number of variables. Now take a look
at the variance explained table to verify the same. As the first eigen value is 4.438, the
variance explained by the first factor can be calculated by diving 4.438 by 10 (total
number of variables) and multiplying bv 100.
21
Scree Plot
5
2
Eigenvalue
0
1 2 3 4 5 6 7 8 9 10
Component Number
Scree plot gives an idea as to how many factors to extract. The rule normally applied is to
stop at where the arm bends. In this case it is three factors. After three factors the curve
gets flat indicating that the gain will be marginal if we go beyond three factors. The
default in SPSS is that it stops when the eigen value gets less than one.
To understand the concept of rotation, take a look at the unrotated component matrix. If
we plot factor2 and factor3 we get the graph shown below.
22
.6
Likeable Flavor
.4 Good Taste
Freshening Breath
Fighting Cavities
.2 Whitening Teeth
Cleaning Stains/Tart
FACTOR3
-.0
Innovative features/
-.2
Attractive Packaging
-.6
-.6 -.4 -.2 -.0 . .4 .6 .8
FACTOR2
It is difficult to interpret this type of data as we find that the cluster of variables likeable
flavor and good taste are mid-way between factor 2 and factor 3. If we rotate the y-axis
so as to pass through the cluster we can interpret the y-axis as `sensory benefits. Same
way the x-axis can be rotated to pass through the cluster that corresponds to Dental
Hygiene. Rotation is done to make it easy to interpret the output.
Note that the angle between X and Y axis in our rotation is more than 90 degrees. If the
angle is maintained at 90o it is called an orthogonal rotation otherwise it is known as
oblique rotation. Note that we used Varimax rotation which is an orthogonal rotation
method.
What we have achieved by conducting a factor analysis is that we have converted the
original ten variables into 3 new factors. Now we can use these three new variables to do
brand positioning.
We can bring both the variables and the brands on to the same map.
23
Brand Positioning Using Factor Scores:
We shall now find out the mean ratings of brands for the three new factors. First of all go
to the variables view and label those new factors as:
1. Dental Hygiene
2. Visibility
3. Sensory Benefits
ANALYZE
COMPARE MEANS
MEANS
Highlight and Drag `Dental Hygiene’, `visibility’, and `oral sensation’ to
DEPENDENT LIST
Drag `brand’ to INDEPENDENT LIST
OPTIONS
Highlight `number of case’ and `standard deviation’ and send back to STATISTICS
CONTINUE
OK
Dental Sensory
Brand Hygiene Visibility Benefits
Aqua-Fresh -.5622039 -.1069474 -.1044542
Colgate .0723164 -.0603488 -.2591244
Crest .3170127 .2179352 .2061569
Mentadent .7780827 .0560046 .9165143
Arm & Hammer .9514446 .0165516 -.9228734
Others .0827401 -.8240165 1.1187287
Total .0000000 .0000000 .0000000
We already have the co-ordinates of the variables in the rotated components matrix.
Create combined table, which has the co-ordinates of both the brands and attributes as
below.
24
Dental Sensory
Brand/Attribute Hygiene Visibility Benefits
Aqua-Fresh -0.56 -0.11 -0.1
Colgate 0.07 -0.06 -0.26
Crest 0.32 0.22 0.21
Mentadent 0.78 0.06 0.92
Arm & Hammer 0.95 0.02 -0.92
Others 0.08 -0.82 1.12
Cleaning Stains/Tartar 0.88 0.2 0.01
Fighting Cavities 0.76 0.05 0.1
Whitening Teeth 0.76 0.15 0.13
Freshening Breath 0.55 0.22 0.48
Innovative
features/Ingredient 0.48 0.44 0.23
Attractive Packaging 0.15 0.87 0.12
Color 0.01 0.83 0.32
Brand Image 0.36 0.76 0.08
Likeable Flavor 0.1 0.18 0.95
Good Taste 0.15 0.2 0.93
To get the positioning map of the first two factors use the following commands.
GRAPHS
SCATTER
SIMPLE
DEFINE
Drag `Dental Hygiene’ to X-AXIS
Drag `Visibility’ to Y-AXIS
Drag `Brand/Attribute’ to LABEL CASES BY
OPTIONS
Check DISPLAY CHART WITH CASE LABELS
OK
The resulting plot can be taken to Power point to have it annotated as shown below. Note
that the attributes are represented as vectors and brands as points.
25
1.0
Attractive Packaging
Colo Brand Image
.5 Innovative
-.5
Others
-1.0
-.6 -.4 -.2 -.0 .2 .4 .6 .8 1.0
Dental Hygiene
While Arm & Hammer and Mentadent are seen as better in Dental Hygiene, Crest score
better on Visibility.
In the same way get the other two plots, namely, Dental Hygiene Vs Sensory Benefits
and Visibility Vs Sensory Benefits.
26
Chapter 3
KANO Model
Objectives:
To learn how to plot the Kano Model and interpret the same
Data Files:
1. cluster.sav
2. factor.sav
Stated Importance:
This will be the mean importance rating given to attributes by the respondents. To
calculate the means, open the file cluster.sav and run the following commands.
ANALYZE
DESCRIPTIVE STATISTICS
DESCRIPTIVES
Highlight and drag variables `q05a….q05j’
OPTIONS
Check DESCENDING MEANS
CONTINUE
OK
27
Descriptive Statistics
You find that Freshening Breath is the most important attribute with a mean rating of 6.3
on a 1-7 scale. Attractive Packaging is the least important attribute. To convert the means
into importance weights, we need to normalize the means.
Take the above table to Excel and find the sum of means. Then calculate:
Importance
Attribute Mean Weight
Freshening Breadth 6.30 12.68
Fighting Cavities 6.11 12.30
Cleaning Stains/Tartar 5.67 11.41
Whitening Teeth 5.64 11.35
Good Taste 5.34 10.75
Likeable Flavor 5.17 10.41
American Dental Association
Recommendation 4.49 9.03
Innovative Feature/new ingredient 4.07 8.19
High Prestige Brand 3.47 6.98
Attractive Packaging 3.43 6.90
Sum 49.70 100.00
Stated Importance
28
Derived Importance:
In order to obtain the derived importance we are going to use the file factor.sav. By
correlating rating of attributes with the overall rating we get the derived importance of
attributes.
ANALYZE
CORRELATE
BIVARIATE
Highlight variables q06_01 to q06_11 and drag to VARIABLES
OK
You will get a 11 by 11 matrix of correlations. We are interested in the last column only
which has the correlation of individual attributes with the overall rating.
As correlations can range from -1 to +1, take the r2 value for derived importance. Once
again these values can be normalized by taking the sum of all the r2 values. The resultant
table is given below.
Importance
Attribute r r2 Weights
Freshening Breath 0.65 0.43 13.26
Good Taste 0.63 0.39 12.22
Likeable Flavor 0.59 0.35 10.84
Brand Image 0.59 0.35 10.81
Innovative
features/Ingredients 0.57 0.33 10.22
Color 0.57 0.32 9.96
Cleaning Stains/Tartar 0.54 0.29 9.04
Whitening Teeth 0.53 0.28 8.77
Fighting Cavities 0.52 0.27 8.46
Attractive Packaging 0.45 0.21 6.41
Sum 3.21 100.00
Bring stated and derived importance to a common table as shown below. Now plot
derived Stated Importance against Derived importance to develop the Kano Model.
29
Stated Derived
Attribute Importance Importance
Attractive Packaging 6.90 6.41
Cleaning Stains/Tartar 11.41 9.04
Fighting Cavities 12.30 8.46
Freshening Breadth 12.68 13.26
Good Taste 10.75 12.22
High Prestige Brand 6.98 10.81
Innovative Feature/new
ingredient 8.19 10.22
Likeable Flavor 10.41 10.84
Whitening Teeth 11.35 8.77
GRAPH
SCATTER
Drag stated importance to X-AXIS
Drag derived importance to Y-AXIS
Drag attribute to LABEL CASES BY
OPTIONS
Check DISPLAY CHART WITH CASE LABELS
CONTINUE
OK
On the graph use 10 as a cut off for High and Low values of importance and illustrate by
taking it to PowerPoint.
30
1
Freshening breath
1 Delight
High
Attributes
Good Taste
1
Derived Importance
Cleaning Stains/Tartar
9
Whitening teeth
Fighting cavities
8
Low Minimum
7 Expected
Attractive packaging
Attributes
6
6 7 8 9 1 1 1 1
Low High
Stated Importance
KANO’s Model
According to Kano’s model attributes that have a high stated and low derived importance are Minimum
expected attributes. Attributes like whitening teeth, fighting cavities , cleaning stains are minimum expected
in a tooth paste.
Attributes with low stated and high derived importance are called Delight attributes. The marketers should
concentrate on these attributes. In this study the innovative ingredients and brand image emerge as the
delight attributes.
Others are linear attributes. If they are important then pay attention. The most important attribute is
freshening breath, as the stated and derived importance are high. If they have low importance one should
not do over engineering of those attributes. In this case spending too much on packaging may not produce
commensurate returns.
31
Chapter 4
Cluster Analysis and Benefit Segmentation
Objectives:
CLUSTER ANALYSIS:
ANALYZE
DESCRIPTIVE STATISTICS
DESCRIPTIVES
Drag variables q05a to q05j to VARIABLES
OK
If you sort the output according to descending values of Standard deviation you will get
this output.
Std.
Attribute N Mean Deviation
American Dental Association
Recommendation 70 4.49 1.89
Attractive Packaging 70 3.43 1.68
Likeable Flavor 70 5.17 1.52
Good Taste 70 5.34 1.50
High Prestige Brand 70 3.47 1.50
Fighting Cavities 70 6.11 1.44
Whitening Teeth 70 5.64 1.41
Innovative Feature/new ingredient 70 4.07 1.37
Cleaning Stains/Tartar 70 5.67 1.35
Freshening Breadth 70 6.30 0.77
32
From the output, it is clear that the five variables with high standard deviation are:
• American dental Association recommendation
• Attractive Packaging
• Likeable Flavor
• Good Taste
• High Prestige Brand
• Fighting Cavities
• Whitening Teeth
These are the attributes where the opinion of the respondents vary much. Hence for
clustering and segmentation we shall use only these seven variables.
Hierarchical Clustering:
ANALYZE
CLASSIFY
HIERARCHICAL CLUSTER
Highlight and drag the seven variables to VARIABLE(S)
Drag nickname to LABEL CASES BY
PLOTS
Check DENDROGRAM
Check NONE
CONTINUE
If you look at the output you will find a tree structure. If you leave out three cases 38
bearing the Nick name `Warden’, 2 (Bud) and 9 (Hank) there are three major branches.
That gives us some idea about how many clusters to ask for when we go to K-means
clustering.
K-Means Clustering:
In this method the respondents will get allocated to different clusters based on the number
of clusters the researcher asks for. Based on the results of the hierarchical cluster we
have decided to ask for three clusters.
ANALAYZE
CLASSIFY
K-MEANS CLUSTER
Highlight and drag the five variables to VARIABLES box
Drag `nickname’ to LABEL CASES BY
NUMBER OF CLUSTERS change from 2 to 3
SAVE
CLUSTER MEMEBERSHIP
CONTINUE
33
OPTION
CLUSTER INFORMATION FOR EACH CASE
Yellow filling indicates rank one across the row and green indicates rank 2.
Benefit Segmentation:
Cluster 1 members are primarily concerned with fighting cavities and are also interested
in American Dental Association recommendation. So the benefit sought is `Medically
proven cavity fighter.
Cluster 2 is primarily interested in Whitening teeth. They have also given relatively high
rating for likeable flavor and good taste. Though they get the second rank on attractive
packaging and high prestige brand, the ratings themselves are low in absolute terms. the
benefit sought by this group is white teeth plus good taste and flavor.
Cluster 3 wants everything; they look for a balanced paste that provides dental care and
sensory benefits (taste & flavor).
From the table on number of cases in each cluster we find that 19 are in cluster 1, 13 are
in cluster 2 and 38 are in cluster 3. Majority (54%) of the people wants a balanced paste,
27% want a cavity fighter and 19% are for white teeth.
34
Number of Cases in each Cluster
Cluster 1 19.000
2 13.000
3 38.000
Valid 70.000
Missing .000
Now take a look at the cluster membership table. This gives the information about the
cluster membership of each individual. The same information is also stored in the SPSS
data file as a new variable crated qcl_1.
Cross tabulation of the new variable qcl_1 Vs race will produce the following table.
Crosstab
Race/Ethicity
White Others Total
Cluster Number of Cavity fighter Count 16 3 19
Case % within Cluster
Number of Case 84.2% 15.8% 100.0%
White teeth Count 9 4 13
% within Cluster
Number of Case 69.2% 30.8% 100.0%
Balanced Paste Count 30 8 38
% within Cluster
Number of Case 78.9% 21.1% 100.0%
Total Count 55 15 70
% within Cluster
Number of Case 78.6% 21.4% 100.0%
While cavity fighting is important for more proportion of whites, white teeth seems to be
of more importance to non-whites. However the chi-square does not show a significant
relationship between benefits segments and race.
35
Chi-Square Tests
Asymp. Sig.
Value df (2-sided)
Pearson Chi-Square 1.036(a) 2 .596
Likelihood Ratio 1.005 2 .605
Linear-by-Linear
.097 1 .755
Association
N of Valid Cases
70
a 2 cells (33.3%) have expected count less than 5. The minimum expected count is 2.79.
Same kind of analysis can be done with age, income, and gender.
Crosstab
Current Brand
Aqua-
Fresh Colgate Crest Others Total
Cluster Cavity fighter Count 4 5 6 4 19
Number of % within
Case Cluster
21.1% 26.3% 31.6% 21.1% 100.0%
Number of
Case
White teeth Count 1 7 2 3 13
% within
Cluster
7.7% 53.8% 15.4% 23.1% 100.0%
Number of
Case
Balanced Count 5 8 17 8 38
Paste % within
Cluster
13.2% 21.1% 44.7% 21.1% 100.0%
Number of
Case
Total Count 10 20 25 15 70
% within
Cluster
14.3% 28.6% 35.7% 21.4% 100.0%
Number of
Case
Aqua Fresh has large proportion of cavity fighters; Colgate has a large proportion of
White teeth seekers; And Crest ahs a large proportion of Balanced paste segment. Once
again the chi-square is not significant. We need to take these results with a pinch of salt.
36
Chi-Square Tests
Asymp. Sig.
Value df (2-sided)
Pearson Chi-Square 7.213(a) 6 .302
Likelihood Ratio 7.038 6 .317
Linear-by-Linear
.674 1 .412
Association
N of Valid Cases
70
a 6 cells (50.0%) have expected count less than 5. The minimum expected count is 1.86.
37
Chapter 5
Correspondence Analysis
Objectives:
Brand
Brand A Brand B Brand C Brand D Total
Income Less than
5 2 1 1 9
$1000
$1000 to
2 4 1 2 9
$3000
$3001 to
2 2 4 1 9
$5000
Above $5000 2 1 1 5 9
Total 11 9 7 9 36
Data Preparation:
To get a better insight on the relationship between income and brand used, we could use
correspondence analysis. To run correspondence analysis using SPSS we need to have
the data organized in the following format.
38
$3001 to $5000 3 Brand B 2 2
$3001 to $5000 3 Brand C 3 4
$3001 to $5000 3 Brand D 4 1
Above $5000 4 Brand A 1 2
Above $5000 4 Brand B 2 1
Above $5000 4 Brand C 3 1
Above $5000 4 Brand D 4 5
SPSS Commands:
The same data has been used to create a data file corres1.sav. Use the following SPSS
commands to run correspondence analysis.
DATA
WEIGHT CASES
WEIGHT CASES BY
Drag `freq’ to FREQUENCY VARIABLE
OK
ANALYZE
DATA REDUCTION
CORRESPONDENCE ANALYSIS
Drag `income’ to ROW VARIABLE
DEFINE RANGE
MINIMUM VALUE `1’
MAXIMUM VALUE `4’
UPDATE
CONTINUE
Drag `brand’ to COLUMN VARIABLE
DEFINE RANGE
MINIMUM VALUE `1’
MAXIMUM VALUE `4’
UPDATE
CONTINUE
OK
39
Interpretation of Results:
The data we used as input for the analysis is printed in the correspondence table.
Correspondence Table
BRAND
INCOME Brand A Brand B Brand C Brand D Active Margin
Less Than $1000 5 2 1 1 9
$1000 to $3000 2 4 1 2 9
$3001 to $ 5000 2 2 4 1 9
Above $5000 2 1 1 5 9
Active Margin 11 9 7 9 36
From the summary Table we can infer that the first two dimensions account for 81.4% of
inertia. This is pretty similar to the eigen values in factor analysis. It is enough to work
with two dimensions.
Summary
Correspondence analysis decomposes the original matrix into row and column points.
Score in
Dimension Contribution
Of Point to Inertia Of Dimension to Inertia of
of Dimension Point
INCOME Mass 1 2 Inertia 1 2 1 2 Total
Less Than
.250 .310 .801 .080 .056 .470 .128 .682 .810
$1000
$1000 to
.250 .020 .245 .053 .000 .044 .001 .096 .097
$3000
$3001 to $
.250 .717 -.764 .106 .301 .428 .518 .469 .987
5000
Above $5000 .250 -1.047 -.282 .127 .642 .058 .920 .053 .974
Active Total 1.000 .367 1.000 1.000
40
The score in Dimension gives the co-ordinates for the row variables in the joint plot. In
the same way you will find the co-ordinates for the column variables in the net matrix.
Using the co-ordinates for the row and column co-ordinates the program produces a joint map which is given
below.
1.0
Less Than $1000
Brand A
.5
Brand B
$1000 to $3000
Dimension 2
0.0
Brand D
Above $5000
-.5
$3001 to $ 5000 BRAN
Brand C
-1.0 INCOM
-1.5 -1.0 -.5 0.0 .5 1.0
Dimension 1
41
From the chart it is very clear that there is a one-to-one relationship between brand used
and the income category. It also shows that Brand A and B are more close to each other
than the other brands.
Toothpaste Data:
Now let us turn our attention to the brand-personality association data collected in the
tooth-paste study. The data has been arranged in the file correspond.sav
Open file corresponds, study the way the data is arranged and run the following SPSS
Commands.
DATA
WEIGHT CASES
WEIGHT CASES BY
Drag `freq’ to FREQUENCY VARIABLE
OK
ANALYZE
DATA REDUCTION
CORRESPONDENCE ANALYSIS
Drag `attri’ to ROW VARIABLE
DEFINE RANGE
MINIMUM VALUE `1’
MAXIMUM VALUE `11’
UPDATE
CONTINUE
Drag `brand’ to COLUMN VARIABLE
DEFINE RANGE
MINIMUM VALUE `1’
MAXIMUM VALUE `3’
UPDATE
CONTINUE
OK
42
Notice that 100 percent inertia has been accounted for the first two dimensions. It is
enough to work with two dimensions.
Summary
Confidence Singular
Proportion of Inertia Value
Standar Correlati
d on
Dimensio Singular Chi Accounted Cumulati Deviatio
n Value Inertia Square Sig. for ve n 2
1 .286 .082 .793 .793 .035 .043
2 .146 .021 .207 1.000 .036
Total .103 72.795 .000(a) 1.000 1.000
a 20 degrees of freedom
The resulting correspondence Map can be taken to Powerpoint and annotated as given below.
.8 Hedonist
Masculine
.6
Aqua Fresh
.4
Fun Loving
Colgate
Overcautious
Dimension 2
.2
Outgoing Sensuous
Romantic
0.0 Feminine
Traditional
-.2
Crest
-.4
Ambitious
-.6 Brand
Achiever
-.8 Personality
-1.0 -.5 0.0 .5 1.0
Dimension 1
43
Correspondence analysis is a powerful tool for visualization of data from contingency
tables. It has no restrictions on the sample size or on the scale used. Association between
any two categorical variables can be easily analyzed using this technique.
44
Chapter 6
Regression Analysis
Objectives:
Simple Regression:
Study the file structure to understand that it is the same field which was used for factor
analysis along with three new variables corresponding to the factor scores that we created
using factor analysis.
Variables q06_01 to q06_10 refer to rating scores for different brands on different
attributes. Variable q06_11 correspond to overall rating given to different brands. We are
going to fit regression equations with overall rating as dependent variable and attribute
ratings as independent variables.
Linear regression refers to fitting of a linear mathematical model between one dependent
variable and one or more independent variables.
We shall first conduct a regression analysis using just two variables. Use the following
SPSS commands to fit a regression model with Fighting Cavities (q06_01) as
independent variable and Overall Rating (q06_11) as the dependent variable.
ANALYZE
REGRESSION
LINEAR
Drag Fighting Cavities (q06_01) to INDEPENDENT
Drag Overall Rating(q06_11) to DEPENDENT
OK
45
Coefficients(a)
Unstandardized Standardized
Coefficients Coefficients
Y = 3.771 + 0.505 X
Y - Overall rating
X – Fighting Cavities
The strength of relationship between the independent and dependent variables is given by
the correlation coefficient R given in Model Summary Table.
Model Summary
The value of R2 0.272 is somewhat low. Normally an R2 value of 0.7 and above is
supposed to signify a strong relationship. In the present case we cannot rule out that there
is no relationship between the two variables as the F value is significant in the ANOVA
Table and the t-value corresponding to the varaiable Figiting Cavities is significant in the
coefficients Table.
ANOVA(b)
Sum of
Model Squares df Mean Square F Sig.
1 Regressio
108.880 1 108.880 73.540 .000(a)
n
Residual 291.672 197 1.481
Total 400.553 198
a Predictors: (Constant), Fighting Cavities
b Dependent Variable: Overall rating
46
Coefficients(a)
Unstandardized Standardized
Coefficients Coefficients
Same way conduct simple regression of each rating variable with the dependent variable
and interpret the results.
Multiple Regression:
Now we shall conduct regression analysis with all the 10 attribute ratings as independent
variables.
ANALYZE
REGRESSION
LINEAR
Drag variables q06_01 to q06_10 to INDEPENDENT
Drag Overall Rating(q06_11) to DEPENDENT
OK
Model Summary
From the coefficients table, we can construct the following mathematical model to depict
the relationship between the 10 independent variables and the overall rating.
The negative sign for variable nine (Attractive Packaging) connotes that the same has a
negative relationship with overall rating. That is, the paste receiving higher rating on
attractive packaging has a diminishing effect on the overall rating.
47
Coefficients(a)
Unstandardized Standardized
Coefficients Coefficients
Now the question arises as to which variable is contributing more to the dependent
variable. This can be inferred by looking at the standardized coefficients. From the β
values we can see that the most important variable is Color (0.238) and the second
important variable is fighting cavities (0.215) and so on
From a statistical stand point also see which are the variables that are significant. The
significance of β is given by the t-values. Note that only the following variables are
significant at 95 percent confidence level.
• Fighting cavities (0.000)
• Freshening breath (0.001)
• Brand Image (0.000)
• Color (0.000) and
• Innovative features (0.000)
Other variables have significance values above 0.05. Can we conclude that other
variables are not important? We cannot say that unless we are sure that the independent
variables are not correlated among them.
48
Multi-collinearity:
When the independent variables are correlated among themselves, we call it a problem of
multi-collinearity. Consider the variables Good Taste and Likeable Flavor. They are
highly correlated among themselves.
ANALYZE
CORRELATE
BIVARIATE
Drag Good Taste and Likeable Flavor to VARIABLES
OK
Correlations
Likeable
Good Taste Flavor
Good Taste Pearson
1 .932(**)
Correlation
Sig. (2-tailed) . .000
N 202 201
Likeable Flavor Pearson
.932(**) 1
Correlation
Sig. (2-tailed) .000 .
N 201 201
** Correlation is significant at the 0.01 level (2-tailed).
Now we shall use only these two variables as independent variable and conduct multiple
regressions.
ANALYZE
REGRESSION
LINEAR
Drag variables Good Taste and Likeable Flavor to INDEPENDENT
Drag Overall Rating(q06_11) to DEPENDENT
OK
The output indicates that only the variable good taste is significant (t = 3.762 and sig. =
0.000) and Likeable flavor is not significant (t = 0.431 and sig. = 0.667). Can we
conclude here that likeable flavor is not an important attribute?
No. Since the two variables are highly correlated, the β values get distorted. As these two
are highly correlated, only one of them attains significance.
Coefficients(a)
49
Unstandardized Standardized
Coefficients Coefficients
To get over the problem of multi-collinearity, we could factor analyze the independent
variables and work with the resulting new factors.
Refer to the module on factor analysis where we reduced the original ten variables to 10
new factors, which were named as Dental Hygiene, Visibility and Sensory benefits. The
resulting factor scores for these new variables have been saved in the regression.sav file.
We shall now conduct a multiple regression with these three new variables as
independent variables and overall rating as the dependent variable.
ANALYZE
REGRESSION
LINEAR
Drag variables Dental Hygiene, Visibility and Sensory Benefits to INDEPENDENT
Drag Overall Rating(q06_11) to DEPENDENT
OK
We get an R2 value of 0.710 which shows that the relationship is quite strong between
the three factors and the overall rating.
Model Summary
50
From the coefficients table we see that all three variables are significant. From the β
values we can conclude that dental hygiene is the most important attribute, followed by
sensory benefits and then visibility.
Coefficients(a)
Unstandardized Standardized
Coefficients Coefficients
In this case we used Overall rating as the dependent variable. If we get respondents
Intention to buy score on rating scale, the same could also be used as dependent variable.
We could then build a regression equation model to predict intention to buy rating from
the attribute rating scores.
51
Chapter 7
DISCRIMINANT ANALYSIS
Objectives:
Basic Ideas:
The variables 1 – 10 in the file refer to attribute ratings for the brand Aquafresh. 11-13
are the factor scores obtained from earlier analysis (refer factor analysis). We need this
information as we may face the same problem of multi-collinearity as we experienced in
multiple regression (Refer to the section on Multiple regression).
We are going to use the first 10 variables as independent variables and later use 11-13,
the factor scores for the three new factors. The dependent variable is whether the
respondent is an user of Aquafresh or not. This is a transformed variable developed from
the current and previous brand variables in the master data file.
Note that the dependent variable is dichotomous and we are going to fit a model that will
help us predict whether someone is an user or non user of Aquafresh. Use the following
SPSS commands.
52
SPSS Commands:
ANALYZE
CLASSIFY
DISCRIMINANT
Drag usage to GROUPING VARIABLE
DEFINE RANGE
Type 1 MINIMUM VALUE
Type 2 MAXIMUM VALUE
CONTINUE
Drag Variables 1-10 to INDEPENDENTS
STATISTICS
MEANS
Check UNSTANDARDIZED COEFFICINETS
CONTINUE
CLASSIFY
Check COMPUTE FROM GROUP SIZES
Check SUMMARY TABLE
CONTINUE
SAVE
Check PREDICTED GROUP MEMBERSHIP
Check DISCRIMINANT SCORE
CONTINUE
OK
Attra- Inno-
Whit- Cleaning Fresh- ctive vative
Fighting ening Stains/ Good Likeable ening Brand Pack- feat-
Groups Cavities Teeth Tartar Taste Flavor Breath Image Color aging ures
Users 7.95 6.65 7.20 7.10 7.30 7.30 6.25 6.85 6.95 6.20
Non
Users 6.74 5.76 6.39 6.82 6.87 7.45 6.95 7.16 7.29 6.53
Overall 7.16 6.07 6.67 6.91 7.02 7.40 6.71 7.05 7.17 6.41
Notice the difference in the ratings given by users and non-users. In the first five
attributes, the users have given a higher rating for Aqua Fresh and in the last five
attributes non-users have given a marginally higher rating for Aquafresh.
In the next Table find something called Canonical correlation. The interpretation of this
is similar to the R value we had in Regression analysis.
Eigenvalues
53
Canonical
Function Eigenvalue % of Variance Cumulative % Correlation
1 .662(a) 100.0 100.0 .631
a First 1 canonical discriminant functions were used in the analysis.
Next Table shows Wilke’s λ, which can be interpreted the same way as the F-test in the
Multiple regression output. The discriminate equation that has been constructed is
significant.
Wilks' Lambda
Wilks'
Test of Function(s) Lambda Chi-square df Sig.
1 .602 25.919 10 .004
Function
1
Fighting Cavities .702
Whitening Teeth .324
Cleaning Stains/Tartar .081
Good Taste .056
Likeable Flavor .167
Freshening Breath -.427
Brand Image -.491
Color .002
Attractive Packaging .201
Innovative
features/Ingredients -.165
(Constant) -3.041
Unstandardized coefficients
54
Using this kind of an equation, a discriminate score is calculated for each respondent and
the programme uses a cut-off point to classify respondents as users or non-users.
Now go to the data file and find two new variables added there – predicted group (dis_1)
and discriminant scores (dis1_1). Change to data view and find the discriminate scores.
To compute the cut-off point calculate the mean discriminant score for each group and
use the following formulae:
where:
N1 = No. of Users
N2 = No. of Non Users
D1 = Average Discrimanat Score for Users
D2 = Average Discriminant Score for Non Users
These values are available in the output in the Table: Functions at group Centroids.
Functions at Group Centroids
Function
User of Aqua Fresh 1
user 1.102
Non-user -.580
Unstandardized canonical discriminant functions evaluated at group means
We know from the next Table that we have 20 users and 38 non-users.
User of Cases
Aqua Used in
Fresh Analysis
Prior Unweighted Weighted
user 0.345 20 20
Non-
user 0.655 38 38
Total 1 58 58
We can calculate the cut-off point as 0.522. Respondents with discriminate scores above
0.522 are classified as users and the others are classified as non-users.
55
Confusion Matrix:
If we cross tabulate the actual Vs predicted, we can get the extent of misclassficcation.
Classification Results(a)
Predicted Group
Membership
User of
Aqua Fresh user Non-user Total
Original Count user 14 6 20
Non-user 5 33 38
% user 70.0 30.0 100.0
Non-user 13.2 86.8 100.0
a 81.0% of original grouped cases correctly classified.
From the output we can see the that the classification accuracy is 81%. The above table is
also known as confusion matrix.
Structure matrix:
Now to determine which varaiables were effective in discriminating the users from non-
users, we could look at the Table of standardized canonical Discriminant coefficients.
These coefficients can be interpreted the same way as the β values were interpreted in
multiple regression. Fighting cavities appears to be the most important in predicting
whether a respondent is an Aquafresh user or not. The least important one seems to be
color. this has to be taken with a pinch of salt as the coefficients can get distorted if the
independent variables are correlated.
Function
1
Fighting Cavities .986
Whitening Teeth .589
Cleaning Stains/Tartar .132
Good Taste .129
Likeable Flavor .402
Freshening Breath -.831
Brand Image -.903
Color .006
56
Attractive Packaging .405
Innovative
features/Ingredients -.359
When the independent variables are correlated among themselves, it is safer to look at the
Structure matrix.
Structure Matrix
Function
1
Fighting Cavities .514
Cleaning Stains/Tartar .296
Whitening Teeth .290
Brand Image -.225
Likeable Flavor .107
Attractive Packaging -.100
Innovative
features/Ingredients -.089
Good Taste .073
Color -.068
Freshening Breath -.045
Pooled within-groups correlations between discriminating variables and standardized canonical discriminant
functions Variables ordered by absolute size of correlation within function.
This values in the Structure matrix can be interpreted much the way as factor loadings in
factor analysis. For example, if you correlate `fighting cavities’ with discriminate score
you will get 0.514. The attributes are arranged in their order of importance.
The attributes that differentiate Aqua Fresh users from others are Fighting cavities and
Cleaning stains/Tartar. These are strong dental hygiene factors.
However, it is better to run the analysis with factor scores, as we already have done factor
analysis. Use the following commands.
ANALYZE
CLASSIFY
DISCRIMINANT
Drag usage to GROUPING VARIABLE
DEFINE RANGE
Type 1 MINIMUM VALUE
Type 2 MAXIMUM VALUE
CONTINUE
Drag Variables 11-13 to INDEPENDENTS
57
STATISTICS
MEANS
Check UNSTANDARDIZED COEFFICINETS
CONTINUE
CLASSIFY
Check COMPUTE FROM GROUP SIZES
Check SUMMARY TABLE
CONTINUE
SAVE
Check PREDICTED GROUP MEMBERSHIP
Check DISCRIMINANT SCORE
CONTINUE
OK
Classification Results(a)
Predicted Group
Membership
User of
Aqua Fresh user Non-user Total
Original Count user 11 9 20
Non-user 11 27 38
% user 55.0 45.0 100.0
Non-user 28.9 71.1 100.0
a 65.5% of original grouped cases correctly classified.
Structure Matrix
Function
1
dental Hygiene .864
visibility -.620
sensory benefits .148
Pooled within-groups correlations between discriminating variables and standardized canonical discriminant
functions Variables ordered by absolute size of correlation within function.
Function
1
dental Hygiene .773
visibility -.483
sensory benefits .223
58
In terms of importance of attributes, we find that `dental hygiene’ is the most discriminating attribute for
Aquafresh. Also find that there is no contradictory interpretataions arising from standardized discriminant
coefficients and the structure matrix.
Brand Drivers:
The implication for the brand is that dental hygiene is the driver attribute for Aquafresh. They should ensure
that the rating on the same attribute does not go down.
Conduct discrimant analysis for Colgate and Crest using files disc_colgate.sav and disc_crest.sav and
identify the brand drivers.
59
Chapter 8
Multi Dimensional Scaling
Objectives:
• To learn how to conduct multidimensional scaling analysis using SPSS and interpret
the results.
- Meaning of Stress
- Stress decomposition
- Interpreting the dimensions
Basic Concepts:
This is a techniques used to identify the underlying dimensions on which a set of stimuli
are differentiated based on the similarity-dissimilarity rating of the stimuli. The stimuli
could be brands or some other objects or people or even countries.
Suppose we get people to rate a set of six countries on a 0-10 scale of similarity-
dissimilarity with 0 when the countries are identical and 10 when they are diametrically
opposite to each other. The following is the matrix generated from a single respondent.
The above data has been arranged in the form required for analysis in the
mds_sample.sav file. Open the same file and study the file structure. The rows and
columns are defined by row_id and col_id. The corresponding ratings are given under the
variable rating. Run the following SPSS commands and study the output.
60
ANALYZE
SCALE
MULTIDIMENSIONAL SCALING (PROXCAL)
Check THE PROXIMITIES ARE IN A SINGLE COLUMN
DEFINE
Drag variable rating to PROXIMITIES
Drag row_id to ROWS
Drag col_id to COLUMNS
MODEL
Check INTERVAL
CONTINUE
OUTPUT
Check INPUT DATA
Check ITERATION HISTORY
CONTINUE
OK
You will notice that the input data used by us has been reproduced in the Input data
matrix.
Proximities
The next table shows the improvement in stress value. Stress in MDS refers to the extent
of model misfit. It is through an iterative process that the program arrives at the
configuration that best fits the dissimilarity rating that we have input into the program.
Iteration History
Normalized
Iteration Raw Stress Improvement
0 .09053(a)
1 .00423 .08630
2 .00397 .00025
61
3 .00388 .00009(b)
a Stress of initial configuration: simplex start.
b The iteration process has stopped because Improvement has become less than the convergence
criterion.
Stress value of less than 0.01 is considered to be very good. Stress values upto 0.03 are
acceptable. The stress value of 0.00388 for the current data shows that the model has
given a good fit.
The program has found a solution in two dimensions and the co-ordinates of the stimuli
(countries) are given in the final coordinates matrix. The co-ordinates are plotted on a
graph too.
Final Coordinates
Dimension
1 2
England .656 -.057
USA .656 .057
France -.337 -.518
Germany -.319 -.572
India -.337 .518
China -.319 .572
A cursory look at the chart shows that the countries that are similar to each other are
placed together and the ones that are different from each other are placed apart from each
other. Notice that there are three clusters – USA & England, France & Germany and
India & China.
62
Object Points
Common Space
China
.6 India
.4
.2
USA
-.0 England
-.2
Dimension 2
-.4
France
Germany
-.6
-.4 -.2 0.0 .2 .4 .6 .8
Dimension 1
In order to interpret the chart we shall take this to Powerpoint. Draw lines passing
through the origin. Now the next task is to uncover the meaning of the underlying two
dimensions.
63
.6 China
India
.4
Dimension 2
.2
US
-.0
England
-.2
-.4
France
-.6 Germany
-.4 -.2 0.0 .2 .4 .6 .8
Dimension 1
In order to get better sense out of this graph, rotate the x-axis and y-axis as shown in the
next chart.
64
.6 China
India
.4
Dimension 2
.2
US
-.0
England
-.2
-.4
France
-.6 Germany
-.4 -.2 0.0 .2 .4 .6 .8
Dimension 1
Now you can clearly see that the x-axis refers to support for Iraq war and the y-axis refers
to economic development.
It raises another question as to how many dimensions will be needed. In general MDS is
used mostly for a two dimensional solution. If the researcher has a priori knowledge that
more than two dimensions may be required then it has to be planned at the data collection
stage itself. If you need more dimensions you will need to get rating on more number of
stimuli. The rule is that for every dimension you need a minimum of three stimuli. If you
need a three dimensional solution you need to have at least 9 stimuli (in this case
countries) rated on a dissimilarity scale. However, if we have data collected from
multiple respondents we use the criteria of Stress. If the stress (measure of mis-fit) value
is more than 0.03 then go for more dimensions.
Here we had only one respondent rating the countries. In a survey typically we will have
several respondents rating the stimuli. In the case of tooth-paste survey, we had a total of
43 valid responses to the MDS question (i.e. question 8).
65
I have arranged those 43 responses in 43 columns in the data file md2.sav. Now open this
file and study the file structure.
ANALYZE
SCALE
MULTIDIMENSIONAL SCALING (PROXCAL)
Check MULTIPLE MATRIX SOURCES
Check THE PROXIMITIES ARE IN A SINGLE COLUMN
DEFINE
Drag variables resp01 to resp69 to PROXIMITIES
Drag row_id to ROWS
Drag col_id to COLUMNS
MODEL
Check INTERVAL
CONTINUE
OUTPUT
Check INPUT DATA
Check ITERATION HISTORY
Check STRESS DECOMPOSITION
CONTINUE
OK
This time we try to fit a model by taking into account the 43 responses. Note that a stress
value of 0.05555 has been achieved. The program has gone through 21 iterations to
arrive at the final configuration.
Iteration History
Normalized
Iteration Raw Stress Improvement
0 .12660(a)
1 .08510 .04150
2 .07399 .01111
3 .06833 .00566
4 .06493 .00340
5 .06273 .00220
6 .06126 .00147
7 .06024 .00102
8 .05949 .00075
9 .05891 .00059
10 .05841 .00050
11 .05797 .00044
12 .05756 .00041
13 .05719 .00037
14 .05686 .00034
66
15 .05656 .00030
16 .05630 .00026
17 .05608 .00022
18 .05591 .00018
19 .05576 .00014
20 .05565 .00012
21 .05555 .00009(b)
a Stress of initial configuration: simplex start.
b The iteration process has stopped because Improvement has become less than the convergence
criterion.
Stress value of 0.05555 is a little too high to make any meaningful interpretation.
You will notice another large table that gives the stress decomposed for every individual.
Notice that respondents 6,20,23,38,39,41,47,58 and 6y9nhave stress values above
0.07. If we leave out these respondents and run MDS analysis, we can possibly get a
solution with lowered stress values. Alternately, we may try for a solution with more
than two dimensions. Let us get a solution in three dimensions.
ANALYZE
SCALE
MULTIDIMENSIONAL SCALING (PROXCAL)
Check MULTIPLE MATRIX SOURCES
Check THE PROXIMITIES ARE IN A SINGLE COLUMN
DEFINE
Drag variables resp01 to resp69 to PROXIMITIES
Drag row_id to ROWS
Drag col_id to COLUMNS
MODEL
Check INTERVAL
Check WEIGHTED EUCLIDIAN
Change 2 to 3 MINIMUM DIMENSIONS
Change 2 to 3 MAXIMUM DIMENSIONS
CONTINUE
OUTPUT
Check INPUT DATA
Check ITERATION HISTORY
Check STRESS DECOMPOSITION
CONTINUE
OK
Note that we are now asking for 3 dimensions. We get a much lower stress value of
0.02319 which is an acceptable level of error in this type of analysis.
67
Stress and Fit Measures
Final Coordinates
Dimension
1 2 3
Aquafresh -.399 .164 -.483
Crest .486 -.189 .360
Colgate .422 -.018 -.471
Mentadent -.482 .063 .408
Arm & Hammer .147 .595 .187
Pepsodent -.175 -.614 -.002
We could plot two at a time to study the relative positioning of brands. However we ‘ll
need to interpret the dimensions. We used a subjective method of interpretation in the
case of six countries. It will be difficult in the case of toothpaste brands as we have a
three dimensional solution.
We shall use regression method to fit attributes on to the same map where the brands are
positioned. Use file mds4.sav to get the co-ordinates for the attributes. Note that the mean
ratings of brands on different attributes and three factors are going to be used for this
analysis.
Perform a multiple regression analysis with the mean rating of attributes as dependent
variable and the co-ordinates for the brands as independent variables. The resulting β
values give the three co-ordinates for the attributes concerned.
ANALLYZE
REGRESION
LINEAR
68
Drag Fighting Cavities to DEPENDENT
Highlight and drag Dim1, dim2 and dim3 to INDEPENDENT
OK
The β values given in the coefficients table give the three co-ordinates for the variable
`fighting cavities’.
Coefficients(a)
Unstandardized Standardized
Coefficients Coefficients
In the same way the co-ordinates for all the attributes are arrived at and a new data file
mds5.sav has been created. This new file contains the co-ordinates for all the brands and
attributes. Now we can plot and interpret the dimensions.
GRAPH
SCATTER
SIMPLE
DEFINE
Drag dim1 to X-AXIS
Drag dim2 to Y-AXIS
Drag attri to LABEL CASES BY
OPTIONS
Check DISPLAY CHART WITH CASE LABELS
CONTINUE
OK
You will get the scatter plot which will have the brands and attributes on the same space.
The same can be taken to a PowerPoint and draw vectors for the attributes as shown in
the chart below.
69
.8
Arm & Hammer
.6 A10
Dental Hygiene A1. Fighting cavities
.4 A2. Whitening Teeth
A A3. Cleaning Stains
A1
Aquafresh A4. Good Taste
.2 A3
A6 A5. Likeable Flavor
Mentadent
Colgate A6. Freshening Breath
.0 A7. Brand Image
A8 A8. Color
Crest
-.2 A9. Attractive Pack
Visibility A10. Innovations
A9
-.4
Pepsodent
Dimension 2
-.6 A7
A4
Sensory Benefits
-.8
A5
-1.0
-.8 -.6 -.4 -.2 0.0 .2 .4 .6
Dimension 1
Only the long vectors belong to this space. Attributes A3, A1, A6 etc. are too short and
need not be considered while interpreting the meaning of Dimensions 1 and 2. Also we
can rotate the axis as shown in the picture to get better meaning of the space.
You can see that Dimension 1 is basically attribute A7, namely Band Image and
Dimension 2 is comprised of attributes A4 and A5 (Good Taste and Likeable Flavor).
Crest and Colgate are seen to have a good brand Image. Pepsodent is high on Dimension
2, namely flavor and taste. This result is consistent with the earlier findings.
To name dimension 3, plot dim1 Vs dim3. From the plot notice that Dimension 3
comprises of A8 + A9 and A2 + A1 (Color + Attractive Packaging & Fighting cavities +
white teeth). Dimension 1 is clearly made up of A7 (brand image) only.
In the same way plot dim2 Vs dim3. Dimension 2 comes out as A4 and A5 (Taste and
flavor) and dim 3 remains as a combination of four attributes A1, A2, A8 and A9.
70
1.0
A8 A6 A2
A3 A1 A1
A9 Visibility
.8
Dental Hygiene
.6 A7 A1. Fighting cavities
A4
A2. Whitening Teeth
A3. Cleaning Stains
.4 Mentadent Crest
A5 A4. Good Taste
Sensory Benefits A5. Likeable Flavor
.2 Arm & Hammer A6. Freshening Breath
A7. Brand Image
Pepsodent A8. Color
-.0 A9. Attractive Pack
A10. Innovations
-.2
Dimension 3
-.4
Aquafresh Colgate
-.6
-.8 -.6 -.4 -.2 0.0 .2 .4 .6
Dimension 1
1.
A9 A8 A6 A3 A2
A1 A1
.8 Visibility
Dental Hygiene A1. Fighting cavities
A7 A2. Whitening Teeth
.6 A4 A3. Cleaning Stains
A4. Good Taste
Mentadent A5. Likeable Flavor
.4 A5 Sensory Benefits Crest
A6. Freshening Breath
A7. Brand Image
Arm & Hammer
.2 A8. Color
A9. Attractive Pack
Pepsodent A10. Innovations
-.0
-.2
Dimension 3
-.6
-1.0 -.8 -.6 -.4 -.2 .0 .2 .4 .6 .8
Dimension 2
71
72
QUESTIONNAIRE ON TOOTH-PASTE BRANDS
This research study is conducted purely for academic purposes only. As the interest is on the collective
opinion of the group as a whole, individual identity will not be revealed. Please have the questionnaire
filled-in and returned to the instructor.
1. Which of the tooth paste brands listed in the table below question 4 are you aware of? (Please put ☑
marks in the appropriate boxes in the Table)
2. Which of these have you tried ever?
3. What brand of tooth paste are you currently using? (If you use more than one brand, please mark `M’
by the side of the most used brand)
4. What was the brand that you used immediately prior to this brand?
Aqua-fresh ☐ ☐ ☐ ☐
Colgate ☐ ☐ ☐ ☐
Crest ☐ ☐ ☐ ☐
Mentadent ☐ ☐ ☐ ☐
Arm & Hammer ☐ ☐ ☐ ☐
Pepsodent ☐ ☐ ☐ ☐
Others1 ________________ ☐ ☐ ☐ ☐
Others2 ________________ ☐ ☐ ☐ ☐
5. Please indicate the relative importance of the following factors in terms of choosing a brand using a
seven-point rating scale. (Please circle the appropriate number to indicate the importance to you)
a. Fighting cavities 1 2 3 4 5 6 7
b. Whitening teeth 1 2 3 4 5 6 7
c. Cleaning Stains/Tartar 1 2 3 4 5 6 7
d. Good taste 1 2 3 4 5 6 7
e. Likeable Flavor 1 2 3 4 5 6 7
f. Freshening Breadth 1 2 3 4 5 6 7
g. High Prestige Brand 1 2 3 4 5 6 7
h. American Dental 1 2 3 4 5 6 7
Association Recommendation
i. Attractive Packaging 1 2 3 4 5 6 7
j. Innovative feature/new ingredient 1 2 3 4 5 6 7
73
6. Please rate Aqua-Fresh, Colgate and Crest on the following factors. If your `current brand’ is other
than these three please rate your current brand too. Use a 0-10 scale with 0 being poor and 10
being good on that attribute. Your response will be a number between 0 and 10 in the boxes below
- indicating the extent to which that attribute is present in the brand being rated. (If you have never
used any of these brands you may leave the corresponding column(s) blank.)
1.Fighting Cavities
2. Whitening teeth
3. Cleaning stains/Tartar
4. Good Taste
5. Likeable Flavor
6. Freshening Breath
7. Brand Image
8. Color
9. Attractive Packaging
10. Innovative
features/Ingredients
11. Overall rating for
brands(out of 10)
7. Kindly put check-marks in the boxes to indicate the personality traits of users that match with the
three brands of tooth pastes mentioned in the following table. You may also check more than one
brand for a personality trait.
1.Outgoing
2. Sensuous
3. Hedonist
4. Fun loving
5. Achiever
6. Romantic
7. Traditional
8. Ambitious
9. Overcautious
10. Feminine
11. Masculine
74
8. Please rate the following pairs of toothpaste brands from most similar pair (1) to most dissimilar pair
(10). Circle the appropriate number. Please leave out the pairs containing brands that you are not aware
of.
9. Please indicate how likely are you to buy the following new toothpaste concepts by using a seven point
rating scale. Please circle the appropriate number that indicates your preference.
75
10. Please rank the following 15 toothpaste concepts from most preferred (1) to least preferred (15). (Enter
your preference ranks in the little eggs in every box) One easy strategy to make the ranking task easier
is to do it in two stages. First mark concepts as H (High preference), L (Low Preference) and M
(Medium preference) and then rank the cards within each category. It is also suggested that you assign
equal number of cards to each category.
* `spicy taste’ could be due to the addition of clove or cinnamon. They also have medicinal properties
apart from being natural ingredients (as opposed to chemicals).
76
11. Please indicate the extent of your agreement with the following statements using the Agree-Disagree
scale given below. Strongly Disagree- 1, Disagree- 2, Neither Agree Nor Disagree - 3, Agree - 4,
Strongly Agree-5.
77
12. Please provide the following information about yourself.
c. Gender:
☐ Male
☐ Female
78
CODE SHEET FOR TOOTH PASTE SURVEY
79
7. Others
18 q03b Current Brand 2 1. Aquafresh
2. Colgate
3. Crest
4. Mentadent
5. Arm & Hammer
6. Pepsodent
7. Others
8. No Second Brand
19 q04 Previous Brand 1. Aquafresh
2. Colgate
3. Crest
4. Mentadent
5. Arm & Hammer
6. Pepsodent
7. Others
20 q05a Fighting Cavities 1. Not Important
To
7. Very Important
21 q05b Whitening Teeth 1. Not Important
To
7. Very Important
22 q05c Cleaning Stains/Tartar 1. Not Important
To
7. Very Important
23 q05d Good Taste 1. Not Important
To
7. Very Important
24 q05e Likeable Flavor 1. Not Important
To
7. Very Important
25 q05f Freshening Breath 1. Not Important
To
7. Very Important
26 q05g High Prestige Brand 1. Not Important
To
7. Very Important
27 q05h American Dental Association 1. Not Important
Recommendation To
7. Very Important
28 q05i Attractive Packaging 1. Not Important
To
7. Very Important
29 q05j Innovative Features/new 1. Not Important
Ingredients To
7. Very Important
80
30 q06a01 Aquafresh - Fighting Cavities 0 – Poor
To
10 – Good
31 q06a02 Aquafresh - Whitening Teeth 0 – Poor
To
10 – Good
32 q06a03 Aquafresh - Cleaning 0 – Poor
Stains/Tartar To
10 – Good
33 q06a04 Aquafresh - Good Taste 0 – Poor
To
10 – Good
34 q06a05 Aquafresh - Likeable Flavor 0 – Poor
To
10 – Good
35 q06a06 Aquafresh - Freshening Breath 0 – Poor
To
10 – Good
36 q06a07 Aquafresh - High Prestige Brand 0 – Poor
To
10 – Good
37 q06a08 Aquafresh - Color 0 – Poor
To
10 – Good
38 q06a09 Aquafresh - Attractive 0 – Poor
Packaging To
10 – Good
39 q06a10 Aquafresh - Innovative 0 – Poor
Features/new Ingredients To
10 – Good
40 q06a11 Aquafresh - Overall Rating 0 – Poor
To
10 – Good
41 q06b01 Colgate - Fighting Cavities 0 – Poor
To
10 – Good
42 q06b02 Colgate - Whitening Teeth 0 – Poor
To
10 – Good
43 q06b03 Colgate - Cleaning Stains/Tartar 0 – Poor
To
10 – Good
44 q06b04 Colgate - Good Taste 0 – Poor
To
10 – Good
45 q06b05 Colgate - Likeable Flavor 0 – Poor
81
To
10 – Good
46 q06b06 Colgate - Freshening Breath 0 – Poor
To
10 – Good
47 q06b07 Colgate - High Prestige Brand 0 – Poor
To
10 – Good
48 q06b08 Colgate - Color 0 – Poor
To
10 – Good
49 q06b09 Colgate - Attractive Packaging 0 – Poor
To
10 – Good
50 q06b10 Colgate - Innovative 0 – Poor
Features/new Ingredients To
10 – Good
51 q06b11 Colgate - Overall Rating 0 – Poor
To
10 – Good
52 q06c01 Crest - Fighting Cavities 0 – Poor
To
10 – Good
53 q06c02 Crest - Whitening Teeth 0 – Poor
To
10 – Good
54 q06c03 Crest - Cleaning Stains/Tartar 0 – Poor
To
10 – Good
55 q06c04 Crest - Good Taste 0 – Poor
To
10 – Good
56 q06c05 Crest - Likeable Flavor 0 – Poor
To
10 – Good
57 q06c06 Crest - Freshening Breath 0 – Poor
To
10 – Good
58 q06c07 Crest - High Prestige Brand 0 – Poor
To
10 – Good
59 q06c08 Crest - Color 0 – Poor
To
10 – Good
60 q06c09 Crest - Attractive Packaging 0 – Poor
To
82
10 – Good
61 q06c10 Crest - Innovative Features/new 0 – Poor
Ingredients To
10 – Good
62 q06c11 Crest - Overall Rating 0 – Poor
To
10 – Good
63 q06d01 Current Brand - Fighting 0 – Poor
Cavities To
10 – Good
64 q06d02 Current Brand - Whitening Teeth 0 – Poor
To
10 – Good
65 q06d03 Current Brand - Cleaning 0 – Poor
Stains/Tartar To
10 – Good
66 q06d04 Current Brand - Good Taste 0 – Poor
To
10 – Good
67 q06d05 Current Brand - Likeable Flavor 0 – Poor
To
10 – Good
68 q06d06 Current Brand - Freshening 0 – Poor
Breath To
10 – Good
69 q06d07 Current Brand - High Prestige 0 – Poor
Brand To
10 – Good
70 q06d08 Current Brand - Color 0 – Poor
To
10 – Good
71 q06d09 Current Brand - Attractive 0 – Poor
Packaging To
10 – Good
72 q06d10 Current Brand - Innovative 0 – Poor
Features/new Ingredients To
10 – Good
73 q06d11 Current Brand - Overall Rating 0 – Poor
To
10 – Good
74 q07a01 Aquafresh - Outgoing `1’ If Marked
`0’ Otherwise
75 q07a02 Aquafresh - Sensuous `1’ If Marked
‘0’ Otherwise
76 q07a03 Aquafresh - Hedonist `1’ If Marked
‘0’ Otherwise
83
77 q07a04 Aquafresh - Fun Loving `1’ If Marked
‘0’ Otherwise
78 q07a05 Aquafresh - Achiever `1’ If Marked
‘0’ Otherwise
79 q07a06 Aquafresh - Romantic `1’ If Marked
‘0’ Otherwise
80 q07a07 Aquafresh - Traditional `1’ If Marked
‘0’ Otherwise
81 q07a08 Aquafresh - Ambitious `1’ If Marked
‘0’ Otherwise
82 q07a09 Aquafresh - Overcautious `1’ If Marked
‘0’ Otherwise
83 q07a10 Aquafresh - Feminine `1’ If Marked
‘0’ Otherwise
84 q07a11 Aquafresh - Masculine `1’ If Marked
‘0’ Otherwise
85 q07b01 Colgate - Outgoing `1’ If Marked
‘0’ Otherwise
86 q07b02 Colgate - Sensuous `1’ If Marked
‘0’ Otherwise
87 q07b03 Colgate - Hedonist `1’ If Marked
‘0’ Otherwise
88 q07b04 Colgate - Fun Loving `1’ If Marked
‘0’ Otherwise
89 q07b05 Colgate - Achiever `1’ If Marked
‘0’ Otherwise
90 q07b06 Colgate - Romantic `1’ If Marked
‘0’ Otherwise
91 q07b07 Colgate - Traditional `1’ If Marked
‘0’ Otherwise
92 q07b08 Colgate - Ambitious `1’ If Marked
‘0’ Otherwise
93 q07b09 Colgate - Overcautious `1’ If Marked
‘0’ Otherwise
94 q07b10 Colgate - Feminine `1’ If Marked
‘0’ Otherwise
95 q07b11 Colgate - Masculine `1’ If Marked
‘0’ Otherwise
96 q07c01 Crest - Outgoing `1’ If Marked
‘0’ Otherwise
97 q07c02 Crest - Sensuous `1’ If Marked
‘0’ Otherwise
98 q07c03 Crest - Hedonist `1’ If Marked
‘0’ Otherwise
99 q07c04 Crest - Fun Loving `1’ If Marked
‘0’ Otherwise
84
100 q07c05 Crest - Achiever `1’ If Marked
‘0’ Otherwise
101 q07c06 Crest - Romantic `1’ If Marked
‘0’ Otherwise
102 q07c07 Crest - Traditional `1’ If Marked
‘0’ Otherwise
103 q07c08 Crest - Ambitious `1’ If Marked
‘0’ Otherwise
104 q07c09 Crest - Overcautious `1’ If Marked
‘0’ Otherwise
105 q07c10 Crest - Feminine `1’ If Marked
‘0’ Otherwise
106 q07c11 Crest - Masculine `1’ If Marked
‘0’ Otherwise
107 q08a Aqua Fresh - Crest 1 – Most Similar
To
10 – Most Dissimilar
108 q08b Aqua Fresh – Colgate 1 – Most Similar
To
10 – Most Dissimilar
109 q08c Aqua Fresh – Mentadent 1 – Most Similar
To
10 – Most Dissimilar
110 q08d Aqua Fresh – Arm & Hammer 1 – Most Similar
To
10 – Most Dissimilar
111 q08e Aqua Fresh – Pepsodent 1 – Most Similar
To
10 – Most Dissimilar
112 q08f Crest – Colgate 1 – Most Similar
To
10 – Most Dissimilar
113 q08g Crest – Mentadent 1 – Most Similar
To
10 – Most Dissimilar
114 q08h Crest – Arm & Hammer 1 – Most Similar
To
10 – Most Dissimilar
115 q08i Crest – Pepsodent 1 – Most Similar
To
10 – Most Dissimilar
116 q08j Colgate – Mentadent 1 – Most Similar
To
10 – Most Dissimilar
117 q08k Colgate – Arm & Hammer 1 – Most Similar
To
85
10 – Most Dissimilar
118 q08l Colgate – Pepsodent 1 – Most Similar
To
10 – Most Dissimilar
119 q08m Mentadent – Arm & Hammer 1 – Most Similar
To
10 – Most Dissimilar
120 q08n Mentadent – Pepsodent 1 – Most Similar
To
10 – Most Dissimilar
121 q08o Arm & Hammer – Pepsodent 1 – Most Similar
To
10 – Most Dissimilar
122 q09a Tooth paste in a big jar (like 1 – Definitely Not Buy
styling gel) To
2 – Will Definitely Buy
123 q09b Make your own tooth-paste kit 1 – Definitely Not Buy
(with whitener, mouth wash, To
foaming agent, baking soda etc. 2 – Will Definitely Buy
in different tubes)
124 q09c Cafinated Tooth paste for a 1 – Definitely Not Buy
refreshing feeling To
2 – Will Definitely Buy
125 q09d Toothpaste with weight-control 1 – Definitely Not Buy
formulae (make you feel full To
after brushing) 2 – Will Definitely Buy
126 q09e Toothpaste containing multi- 1 – Definitely Not Buy
vitamin To
2 – Will Definitely Buy
127 q09f Spicy toothpaste (clove and/or 1 – Definitely Not Buy
cinnamon) To
2 – Will Definitely Buy
128 q09g Transparent tube to see how 1 – Definitely Not Buy
much is left inside To
2 – Will Definitely Buy
129 q09h Toothpaste containing colored 1 – Definitely Not Buy
beads and flavor crystals To
2 – Will Definitely Buy
130 q09i Use and throw away paste-pre- 1 – Definitely Not Buy
applied tooth brushes (Better To
Hygiene) 2 – Will Definitely Buy
131 q09j Tube fitted with dispenser for 1 – Definitely Not Buy
right amount To
2 – Will Definitely Buy
132 q09k Male /Female toothpaste 1 – Definitely Not Buy
To
86
2 – Will Definitely Buy
134 q09l Toothpaste that doubles as 1 – Definitely Not Buy
shaving cream To
2 – Will Definitely Buy
135 q09m Night paste with sleep inducers 1 – Definitely Not Buy
To
2 – Will Definitely Buy
136 q10a Concept 1 Rank 1 To 15
137 q10b Concept 2 Rank 1 To 15
138 q10c Concept 3 Rank 1 To 15
139 q10d Concept 4 Rank 1 To 15
140 q10e Concept 5 Rank 1 To 15
141 q10f Concept 6 Rank 1 To 15
142 q10g Concept 7 Rank 1 To 15
143 q10h Concept 8 Rank 1 To 15
144 q10i Concept 9 Rank 1 To 15
145 q10j Concept 10 Rank 1 To 15
146 q10k Concept 11 Rank 1 To 15
147 q10l Concept 12 Rank 1 To 15
148 q10m Concept 13 Rank 1 To 15
149 q10n Concept 14 Rank 1 To 15
150 q10o Concept 15 Rank 1 To 15
151 q11a I am very concerned about my 1 – Strongly Disagree
looks To
5 – Strongly Agree
152 q11b I am very inquisitive like a child. 1 – Strongly Disagree
To
5 – Strongly Agree
153 q11c I like to be different in whatever 1 – Strongly Disagree
I do. To
5 – Strongly Agree
154 q11d I love whatever work I do. 1 – Strongly Disagree
To
5 – Strongly Agree
155 q11e I am always playful 1 – Strongly Disagree
To
5 – Strongly Agree
156 q11f I keep thinking about my future. 1 – Strongly Disagree
To
5 – Strongly Agree
157 q11g I take a lot of care about the 1 – Strongly Disagree
dress I wear To
5 – Strongly Agree
158 q11h I like to try new and different 1 – Strongly Disagree
things To
87
5 – Strongly Agree
159 q11i I am sensitive to others feelings 1 – Strongly Disagree
To
5 – Strongly Agree
160 q11j I believe in keeping myself 1 – Strongly Disagree
physically fit. To
5 – Strongly Agree
161 q11k I dislike being left alone. 1 – Strongly Disagree
To
5 – Strongly Agree
162 q11l I'd say I'm rebelling against the 1 – Strongly Disagree
way I was brought up To
5 – Strongly Agree
163 q11m I try to understand deeply about 1 – Strongly Disagree
anything that I study To
5 – Strongly Agree
164 q11n I always worry about my past 1 – Strongly Disagree
failures. To
5 – Strongly Agree
165 q11o I like to follow what others do. 1 – Strongly Disagree
To
5 – Strongly Agree
166 q11p I am very sensitive to what 1 – Strongly Disagree
others think about me. To
5 – Strongly Agree
167 q11q My objective in life is to acquire 1 – Strongly Disagree
wealth to the maximum extent To
possible. 5 – Strongly Agree
168 q11r I am very organized 1 – Strongly Disagree
To
5 – Strongly Agree
169 q11s I'm a "spender" rather than a 1 – Strongly Disagree
"saver." To
5 – Strongly Agree
170 q11t Love and sex are great 1 – Strongly Disagree
distractions to achieving ones To
objectives. 5 – Strongly Agree
171 q11u I act on my hunches 1 – Strongly Disagree
To
5 – Strongly Agree
172 q11v I am more conventional than 1 – Strongly Disagree
experimental To
5 – Strongly Agree
173 q11w I am a little fickle minded 1 – Strongly Disagree
To
5 – Strongly Agree
88
173 q11x I love to have good food every 1 – Strongly Disagree
day. To
5 – Strongly Agree
174 q12a Age 1. Less Than 25
2. 25 and Above
175 q12b Household Income 1. $30,000 or less
2. $30,001 to $80,000
3. $80,001 to $125,000
4. Above $125,000
176 q12c Gender 1. Male
2. Female
177 q12d Race/Ethnicity 1. White
2. Others
178 q12e Nickname Enter as given
179 aware Number of Brands Aware
180 trial Number of Brands Tried
181 qcl_1 Benefit Clusters 1. Dental Care
2. Sensory benefits
3. a la carte
182 fact1_1 External Locus of Control
183 fact2_1 Internal Locus of Control
184 fact3_1 Creative
185 fact4_1 Unique
186 fact5_1 Goal Oriented
187 fact6_1 Fun Loving
188 fact7_1 Rule Bound
189 fact8_1 Rebel
190 qcl_2 Psychographic Clusters 1. Middle of the Road
2. Creative/unique
3. Goal directed rebels
4. Systematic Achiever
5. Inner Directed
Individual
Variables 179 and 180 are computed by the researcher from variables defined already.
These are called transformed variables.
Variables 181-190 are SPSS System created variables and labels given by the researcher.
89
Slide 1
BY
Prof. M J Xavier
Slide 2
Contents
• Executive Summary - 3
• Profile of Respondents - 8
• Awareness, Trial and Usage - 15
• Benefits Sought and Benefit Segmentation - 28
• Competitive Positioning of Brands - 37
• Evaluation of New Product Ideas - 48
• Psycho-graphic Segmentation - 65
90
Slide 3
Executive Summary
• A study on toothpaste market was carried out among the university
students during September-November, 2003 to asses the
preferences and to evaluate new product ideas.
• The study was done in two phases. The qualitative phase used
focus groups to understand the nature of the market and to come up
with a number of new product ideas. In the quantitative phase, a
questionnaire was developed and administered to 70 students.
• Respondent Profile
– The sample comprised of 15 Graduate Students and 55 undergraduate
students from Orfalea College of Business.
– 35 were males and 35 were female students.
– 58 of them were below the age of 25
– The sample had a wide income distribution
– In terms of ethnicity 80% were whites
Slide 4
91
Slide 5
Slide 6
92
Slide 7
Slide 8
Profile of Respondents
93
Slide 9
Class
Valid Cumulative MKT-347-02
Frequency Percent Percent
45.7%
GBS-573-02 15 21.4 21.4 GBS-573-02
MKT-347-01 23 32.9 54.3 21.4%
MKT-347-02 32 45.7 100.0
Total 70 100.0
MKT-347-01
32.9%
Slide 10
Age Profile
Frequ- Valid Cumulative 30-34 years
Age ency Percent Percentage 1.4%
18-24
years 58 83 83
25-29 years
25-29
15.7%
years 11 16 99
30-34
years 1 1 100
18-24 years
Total 70 100
82.9%
10
94
Slide 11
Gender
Cumulative
Frequency Percent Valid Percent Percent
Valid Male 35 50.0 50.0 50.0
Female 35 50.0 50.0 100.0
Total 70 100.0 100.0
Male
Female 50.0%
Gender has a perfect
50:50 split 50.0%
11
Slide 12
Income Distribution
Cumulative
Frequency Percent Valid Percent Percent
Valid Less than $30,000 35 50.0 50.0 50.0
$30,000 to $80,000 10 14.3 14.3 64.3
$80,001 to $125,000 11 15.7 15.7 80.0
Above $125,000 14 20.0 20.0 100.0
Total 70 100.0 100.0
Above $125,000
20.0%
$30,000 to $80,000
14.3%
12
95
Slide 13
Race/Ethnicity
Cumulative
Frequency Percent Percent
Asian/Pacific Islander 6 8.6 8.6
Hispanic 3 4.3 12.9
White 55 78.6 91.4
Multi-racial 3 4.3 95.7
Refused 3 4.3 100.0
Total 70 100.0
Multi-racial
Close to 80% respondents 4.3% Hispanic
are whites 4.3%
White
78.6%
13
Slide 14
14
96
Slide 15
15
Slide 16
Brand Awareness
Aqua-fresh Aware
Pepsodent Colgate
7.1%
Unaware
1.4%
Unaware Aware
Unaware 25.7%
11.4%
4.3%
Unaware
Aware Aware 74.3%
95.7% 88.6%
Major Brands – Colgate, Aqua-fresh and Crest have close to 100 percent
awareness while others have low levels of awareness 16
97
Slide 17
40 37.1%
30
Percent
20
10 8.6%
5.7%
1.4%
0
2.00 3.00 4.00 5.00 6.00
AWARE
17
Slide 18
Mean
AWARE * RACE
INCOME AWARE
Less than $30,000 4.4857 Mean
$30,000 to $80,000 4.4000 RACE AWARE
$80,001 to $125,000 4.4545 White 4.4727
Above $125,000 4.4286 Others 4.4000
Total 4.4571
Total 4.4571
98
Slide 19
Usage/Trial
Aqua-fresh Colgate Crest
Never Used
Never Used 5.7% Never Used
2.9% 5.7%
Used
Used 94.3% Used
97.1%
94.3%
Pepsodent
Mentadent Arm & Hammer
Used Used
2.9%
21.4%
Never Used
44.3%
Used
55.7%
Never Used
78.6% Never Used
97.1%
The three major brands – Aqua-fresh, Colgate and Crest have close to 95% trial
19
Slide 20
50
40
Percent
30
20
10
0
1.00 2.00 3.00 4.00 5.00 6.00 7.00
TRIAL
99
Slide 21
Slide 22
Current Brand
40
35.7%
28.6%
30
Percent
20
14.3%
8.6% 8.6%
10
4.3%
Base - 70
0
Aqua-Fresh Crest Arm & Hammer
Colgate Mentadent Others
Current Brand
Crest has a larger share compared to the other two major brands Colgate and
Aqua-fresh though they match Crest in terms of awareness and trial. The niche
22
brands Mentadent and Arm&Hammer have reasonable shares.
100
Slide 23
Slide 24
No Second Brand
92.9%
24
101
Slide 25
Previous Brand
Previous Brand
50
45.7%
40
Percent
30
22.9%
21.4%
20
10 7.1%
1.4% 1.4%
0 Base - 70
Aqua-Fresh Crest Arm & Hammer
Colgate Mentadent Others
Previous Brand
The pattern is more or less the same as the current Brand shares
25
Slide 26
Very low level of Brand Loyalty is seen. Crest has the highest retention
figure of 37.5%. People seek variety. This offers great opportunity for
manufacturers to offer variety through innovative ingredients.
26
102
Slide 27
27
Slide 28
Benefits Sought
& Benefit Segmentation
28
103
Slide 29
Slide 30
30
104
Slide 31
31
Slide 32
Un-standardized Standardized
Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) .598 .349 1.716 .088
Fighting Cavities .208 .048 .215 4.368 .000
Whitening Teeth .097 .037 .130 2.606 .010
Cleaning Stains/Tartar .016 .052 .018 .307 .759
Good Taste .109 .072 .154 1.513 .132
Likeable Flavor .060 .071 .086 .844 .400
Freshening Breath .150 .045 .173 3.311 .001
Brand Image .135 .038 .195 3.582 .000
Color .154 .037 .238 4.187 .000
Attractive Packaging -.096 .041 -.132 -2.361 .019
Innovative
.120 .033 .181 3.660 .000
features/Ingredients
a. Dependent Variable: Overall rating
Though the most important and the least important attributes remain the
same, there are several changes that can be noticed in relation to the
Importance of attributes obtained using direct method. 32
105
Slide 33
Fighting Cavities
14 Brand Image
High Innovative features Freshening Breath
Good Taste
12 Delight Whitening Teeth
DERIVED IMPORTANCE
Attributes s Likeable Flavor
10 u te
ib
Attr Cleaning Stains/Tarter
8 r
n ea
6 Li
Minimum
4
Attractive Packaging Expected
Low
2
7 8 9 10 11 12 13 14
Low High
STATED IMPORTANCE
Slide 34
Benefit Segmentation
Final Cluster Centers
Cluster
Number of Cases in each Cluster
1 2 3
Fighting Cavities 6.92 4.19 6.62 Cluster 1 12.000
Whitening Teeth 6.42 6.06 5.26 2 16.000
Good Taste 3.17 5.81 5.79 3 42.000
Likeable Flavor 3.17 5.63 5.57 Valid 70.000
American Dental Missing .000
Association 6.00 2.19 4.93
Recommendation
Cluster 1 has 12 respondents and the attributes that they look for are –
fighting cavities, whitening teeth and American Dental Association
Recommendation – We call this `dental care’ cluster
Cluster 2 with 16 respondents is interested in oral sensation – good taste,
likeable flavor and white teeth – We shall name it as `sensory benefits’
cluster
Cluster 3 with 42 respondents want a bit of everything – we shall call this
a la carte cluster. 34
106
Slide 35
50% of cluster 2 members use Colgate which clearly indicates that Colgate is perceived to be high on
Oral sensory benefits
42.9% of a la carte cluster uses Crest which indicates that Crest is seen as a balanced tooth paste
offering dental care as well as sensory benefits.
Though Arm Hammer has a large proportion of cluster 1 members (caution - the numbers are too low
to give any meaningful interpretation) there is no paste that has a clear positioning for `dental
hygiene’. It offers an opportunity for launching a new product with new ingredients like clove to fight
cavities and offer total `dental care’.
Aqua-fresh and Mentadent are stuck in the middle with no clear positioning. 35
Slide 36
36
107
Slide 37
Competitive Positioning of
Brands
37
Slide 38
Crest has been consistently rated as high on almost all the attributes.
The high rating for Arm & Hammer on some of the attributes will have
to understood against the fact that only three respondents have rated
the same.
38
108
Slide 39
g
th
or
e
es
s
oo tar
th
e
vo
in
nt
st
ag
ee
ea
ol
ag
iti
ar
ie
Ta
la
C
Im
T
av
Br
/T
ed
ck
ng
d
C
e
ns
Pa
gr
ng
l
an
ab
g
n i
ai
/ In
ni
G
in
te
ive
Br
St
ke
he
ht
es
hi
Li
ct
g
ng
W
es
ur
Fi
t ra
at
ni
Fr
At
a
fe
le
e
C
tiv
va
no
In
Crest scores over the other two competitors on all attributes. Colgate scores
over Aqua Fresh on all attributes except Attractive Packaging. 39
Slide 40
109
Slide 41
Aqua Fresh
.4
Fun Loving
Colgate
Overcautious
Dimension 2
.2
Outgoing Sensuous
Romantic
0.0 Feminine
Traditional
-.2
Crest
-.4
Ambitious
-.6 Brand
Achiever
-.8 Personality Trait
-1.0 -.5 0.0 .5 1.0
Dimension 1
Slide 42
Crest
.5 Arm&Hammer Colgate
Dimension 2
Mentadent
0.0
Pepsodent
-.5
-1.0
Aqua Fresh
-1.5
-2.0 -1.5 -1.0 -.5 0.0 .5 1.0 1.5 2.0
Dimension 1
110
Slide 43
Component
1 2 3
Cleaning Stains/Tartar .878 .203 .014
Fighting Cavities .763 .051 .103
Whitening Teeth .758 .150 .126
Freshening Breath .550 .220 .484
Innovative
.482 .437 .234
features/Ingredients
Attractive Packaging .154 .867 .115
Color .013 .830 .323
Brand Image .364 .757 .077
Likeable Flavor .098 .183 .950
Good Taste .152 .195 .933
Slide 44
Crest
.2
Visibility
.1
Mentadent
-.1 Aqua-Fresh
-.2
-.6 -.4 -.2 -.0 .2 .4 .6 .8 1.0
Dental Hygiene
111
Slide 45
Mentadent
.5
Sensory Benefits
Crest
0.0
Aqua-Fresh
Colgate
-.5
Dental Hygiene
Slide 46
Mentadent
.5
Sensory Benefits
Crest
0.0
Aqua-Fresh
Colgate
-.5
Visibility
Colgate and Aqua Fresh get lower rating on both sensory and
visibility factors
46
112
Slide 47
47
Slide 48
48
113
Slide 49
Slide 50
114
Slide 51
While more males prefer Night-paste with sleep inducers more females
prefer Toothpaste with weight control formulae
51
Slide 52
Younger people are more receptive to new ideas than mature students.
52
115
Slide 53
Members of the oral sensation cluster have a higher preference for a paste
with weight-control formulae
53
Slide 54
31.3% PRICE
Preferred Levels
3.4833
4 Transparent Tube
Taste and price 3
are the two 2
-0.6167 -2.8667
Mint Taste
1
most important 0
White Color
attributes -1 Crest Brand
-2
-3 Price - $2
2 Dollars 4 Dollars 6 Dollars
116
Slide 55
0.5
Non-Transparent
0
Fight Cavities
Transparent
-0.5 Oral sensation
A la carte
-1 Semi-Transparent
-1.5
-2
-2.5
Slide 56
56
117
Slide 57
While color is immaterial for `Fight Cavities’ cluster, `Oral Sensation’ cluster
ahs a positive preference for green color. 57
Slide 58
Oral sensation cluster members have a positive preference for Colgate (this
is consistent with earlier results – see slide ) while the other two clusters
prefer Crest. 58
118
Slide 59
1
$6 Fight Cavities
0 Oral sensation
$2 A la carte
-1
$4
-2
-3
-4
$2 $4 $6 Importance
Fight Cavities 2.7833 0.4833 -3.2667 31.31
Oral sensation 3.1333 -0.4417 -2.6917 26.78
A la carte 3.1 -0.925 -2.175 25.76
59
Slide 60
Male-Female Differences in
Conjoint Utilities
Transparent Semi-Transparent Non-Transparent Importance
Male -0.2667 0.7583 -0.4917 6.16
Female 0.58333 0.4583 -1.0417 8.02
Mint Spicy Cafinated Importance
Male 2.9 -2.7 -0.2 27.6
Female 4.4167 -3.8333 -0.5833 40.7
White Red Green Importance
Male 1.0667 -0.9083 -0.1583 9.73
Female 2.0833 -2.1667 0.0833 20.99
Colgate Aqua Fresh Crest Importance
Male 0.4 -1.825 1.425 16.01
Female -0.5833 -0.3333 0.9167 7.41
$2 $4 $6 Importance
Male 4.4 -0.575 -3.8250 40.52
Female 2.25 0.125 -2.375 22.84
While brand and price are more important for males, taste and color
are more important for females.
60
119
Slide 61
61
Slide 62
120
Slide 63
While the middle income is sensitive to brand, high income is taste and price sensitive.
Though they are willing to pay up to 4 dollars per tube, the utility drops drastically when the
63
price goes up to 6 dollars
Slide 64
64
121
Slide 65
Psycho-graphic Segmentation
65
Slide 66
Scree Plot
4
Component Number
66
122
Slide 67
Slide 68
Factor Labeling
Factor – 1 (External Locus of Control) Factor – 5 (goal-oriented)
o I am very sensitive to what others think o Love and sex are great distractions to achieving
about me. ones objectives.
o I always worry about my past failures. o I dislike being left alone.
o I like to follow what others do. o I try to understand deeply about anything that I
study
o I act on my hunches
Factor - 2 (Internal Locus of Control)
o I am very concerned about my looks
o I take a lot of care about the dress I wear Factor 6 (Fun Loving)
o I believe in keeping myself physically fit. o I'm a "spender" rather than a "saver."
o I keep thinking about my future. o I am a little fickle minded
o I am more conventional than experimental o My objective in life is to acquire wealth to the
maximum extent possible.
o I love to have good food every day.
Factor – 3 (Creative)
o I am very inquisitive like a child.
Factor 7 (Rule Bound)
o I am always playful
o I am very organized
o I like to try new and different things
o I am sensitive to others feelings
Factor – 4 (Unique)
o I love whatever work I do. Factor8 (Rebel)
o I'd say I'm rebelling against the way I was
o I like to be different in whatever I do. brought up
68
123
Slide 69
Slide 70
Close to half the mature student population falls in the middle of the road
category and also have a large representation in the systematic achiever
category.
The younger people are spread across other segments.
70
124
Slide 71
Income Vs Psycho-graphic
Segments
Cluster Number of Case
Middle of Creative/ Goal Directed Systematic Inner Directed
the Road Unique Rebels Achiever Individual Total
INCOME Less than $30,000 Count 4 11 5 5 9 34
% within INCOME 11.8% 32.4% 14.7% 14.7% 26.5% 100.0%
$30,000 to $80,000 Count 5 0 1 2 1 9
% within INCOME 55.6% .0% 11.1% 22.2% 11.1% 100.0%
$80,001 to $125,000 Count 1 4 1 1 3 10
% within INCOME 10.0% 40.0% 10.0% 10.0% 30.0% 100.0%
Above $125,000 Count 2 2 4 1 3 12
% within INCOME 16.7% 16.7% 33.3% 8.3% 25.0% 100.0%
Total Count 12 17 11 9 16 65
% within INCOME 18.5% 26.2% 16.9% 13.8% 24.6% 100.0%
More middle income people fall in the Middle of the Road category while
larger proportion of high income people fall in the goal directed rebels
category.
אּ2 value of 16 at
Due to the small sample size the level of significance of
12 degrees of freedom has only significance level of 0.187 71
Slide 72
Gender Vs Psycho-graphic
segments
Cluster Number of Case
Middle of Creative/ Goal Directed Systematic Inner Directed
the Road Unique Rebels Achiever Individual Total
Gender Male Count 9 7 6 3 8 33
% within Gender 27.3% 21.2% 18.2% 9.1% 24.2% 100.0%
Female Count 3 10 5 6 8 32
% within Gender 9.4% 31.3% 15.6% 18.8% 25.0% 100.0%
Total Count 12 17 11 9 16 65
% within Gender 18.5% 26.2% 16.9% 13.8% 24.6% 100.0%
72
125
Slide 73
The relationship between brand used and the psycho graphic segments is not statistically
significant. Large sample sizes will be needed to establish firm relationships.
73
Slide 74
Psycho-graphic Segmentation -
Summary
• Factor analysis of the ratings of 24 statements yielded
eight factors.
• These eight factor scores were clustered to arrive at the
following five segments
– Cluster 1 - Middle of the Road
– Cluster 2 - Creative /Unique
– Cluster 3 - Goal Directed Rebels
– Cluster 4 - Systematic Achiever
– Cluster 5 - Inner Directed individual
• The relationship between brand used and the psycho graphic
segments is not statistically significant. Large sample sizes will be
needed to establish firm relationships.
74
126