You are on page 1of 28

Section

Section A
A -- Group-1
Group-1
Ankush
Ankush Mittal
Mittal |Akhil
|Akhil Bhambri
Bhambri || Pooja
Pooja Sharma
Sharma ||
Sriram
Sriram Venkatakrishnan
Venkatakrishnan || Viraj
Viraj Hede
Hede

Backgrou
nd

Research
Design

Objective
s

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Background
Smirnoffs R&D team has prepared two new blends; claim to be superior
than the one already in market
Blind test of the two new blends vs the one in the market among
regular consumers of vodka needs to be done
Blind test to be done between Smirnoff drinkers and important
competition brand like Fuel and Magic Moments drinkers
Any of the two new blends to be considered for a change, if it comes out
to be significantly (at 95%) better than the current blend (with max 5%
sampling error)
Markets chosen for study: Delhi, Bangalore, Mumbai, Chennai and Kolkata

Backgrou
nd

Research
Design

Objective
s

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

RESEARCH DESIGN
The research will be conducted with
sequential monadic exposure with 3
blends placed for consumption one
after the other and feedback will be
taken after each consumption

Males/Female
s in the age
group of 25
35 yrs

Consuming
vodka at
least twice a
week

Regular
consumer of
any one of
the three
brands
Smirnoff, Fuel
,Magic
Moments

Target Group

Neutralizers given after each


consumption

Current blend(control blend) is blend 2


in data. The two new brands are blend
1 and 3 in data

Sample
Size
760, each
person being
given 3 blends
for feedback
Thus there are
760 X 3 = 2280
data points as
feedback on the
vodka blends

Backgrou
nd

Research
Design

Objective
s

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Primary Objective of the


Analysis

Conclusio
n

Action
Item

Can the Current Product (Control) be replaced by any of the two Test Products?
ACTION STANDARD: The Test Product has to be significantly better (at 90% or
95% Confidence level) in Overall Likeability and two of the other most important
attributes
Additionally, it should do better than Control among important sub-groups (at 90%
Confidence level)

Backgrou
nd

Research
Design

Objective
s

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Secondary Objectives of the


Analysis

Action
Item

Which are the attributes that drive overall preference of vodka?


What is the extent to which they drive? (Meaning, which are the
more important and which are the less important drivers?)

Can the set of attributes be reduced to a smaller set of Factors, and


later, it could be found, how the broad Factors drive the overall
preference of vodka?

If the purchase intention is taken as a categorical variable (YES /


NO), can it be predicted in future, by reading some of the ratings of
the attributes only? How?

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Information - Attributes
Dependent:
5A_Attr1 : Overall Likeability of the vodka blend
Independent:
5A_Attr2 : Likeability of Aroma
5A_Attr3 : Likeability of Taste
5A_Attr4 : Likeability of Smoothness
5A_Attr5 : Likeability of Flavour
5A_Attr6 : Likeability of Throat-feel when the vodka goes down
5A_Attr7 : Likeability of After-taste
5A_Attr8 : Likeability of Mouth-feel when the vodka is sipped

Conclusio
n

Action
Item

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

CROSS TAB ANALYSIS:

Done over the entire sample on :


Overall
Likeability

Taste

Mouth Feel

1) Overall Likeability:
) Product 2: Control Product
Product 1 and Product 3: Test Products
) We consider the summation of the 9th and
10th rating of all the three products and add
their respective errors.
Hence we get
percentage ranges for the products. If
overlap found with 95% confidence level,
we try with 90%.
) If there is no overlap between the ranges of
product 1 and 2, Product 1 can be

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Overall Likeability, Confidence


Level = 95%
For Product 2:

For Product 1:
23.8% + 12.0% = 35.8%
Error at Confidence level
95% with N = 760
1.96 * sqrt (0.36 * 0.64 / 760) =
3.4%
Range : 35.8% + 3.4% = 39.2%
35.8% 3.4% = 32.4%
Range = 32.4% to 39.2%
For Product 3:
17.9% + 11.2% = 29.1%
Error at Confidence level 95%
with N = 760
1.96 * sqrt (0.29 * 0.71 / 760) =
3.2%
Range : 29.1% + 3.2% = 32.3%
29.1% 3.2% = 25.9%
Range = 25.9% to 32.3%

18.4% + 11.8% = 30.2%


Error at Confidence level
95% with N = 760
1.96 * sqrt (0.30 * 0.70 / 760)
= 3.2%
Range : 30.2% + 3.2% =
33.4%
30.2% 3.2% = 27%
Range = 27% to 33.4%

The test products and the


control have overlaps
between their ranges. We
will not be able to launch
the new products with
95% confidence level
based on overall
likeability

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Overall Likeability, Confidence


For Product 2:
Level = 90%

For Product 1:
23.8% + 12.0% = 35.8%
Error at Confidence level
90% with N = 760
1.645 * sqrt (0.36 * 0.64 / 760)
= 2.8%
Range : 35.8% + 2.8% = 38.6%
35.8% 2.8% = 33%
Range = 33% to 38.6%
For Product 3:
17.9% + 11.2% = 29.1%
Error at Confidence level 90%
with N = 760
1.645 * sqrt (0.29 * 0.71 / 760) =
2.7%
Range : 29.1% + 2.7% = 31.8%
29.1% 2.7% = 26.4%
Range = 26.4% to 31.8%

18.4% + 11.8% = 30.2%


Error at Confidence level
90% with N = 760
1.645 * sqrt (0.30 * 0.70 /
760) = 2.7%
Range : 30.2% + 2.7% =
32.94%
30.2% 2.7% =
27.5%
Range = 27.5% to 32.94%

The range of Product 1


does not overlap with
that of Product 2. We
should launch Product 1 if
considered with 90%
confidence level based on
Overall Likeability

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Taste at Confidence Level = 95%


For Product 1:
22.5% + 12.0% = 34.5%
Error at Confidence level
95% with N = 760
1.96 * sqrt (0.35 * 0.65 / 760) =
3.4%
Range : 34.5% + 3.4% = 37.9%
34.5% 3.4% = 31.1%
Range = 31.1% to 37.9%
For Product 3:
17.1% + 11.3% = 28.4%
Error at Confidence level 95%
with N = 760
1.96 * sqrt (0.29 * 0.71 / 760) =
3.2%
Range : 29.1% + 3.2% = 32.3%
29.1% 3.2% = 25.9%
Range = 25.9% to 32.3%

For Product 2:
17.4% + 13.3% = 30.7%
Error at Confidence level
95% with N = 760
1.96 * sqrt (0.31 * 0.69 / 760)
= 3.2%
Range : 30.7% + 3.2% =
33.9%
30.7% 3.2% =
27.5%
Range = 27.5% to 33.9%

Since the test products


and the control have
overlaps between their
ranges, we will not be
able to launch the new
products with 95%
confidence level based on
taste

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Taste at Confidence Level = 90%


For Product 1:
22.5% + 12.0% = 34.5%
Error at Confidence level
90% with N = 760
=1.645 * sqrt (0.35 * 0.65 /
760) = 2.8%
Range : 34.5% + 2.8% = 37.3%
34.5% 2.8% = 31.7%
Range = 31.7% to 37.3%
For Product 3:
17.1% + 11.3% = 28.4%
Error at Confidence level 90%
with N = 760
=1.645 * sqrt (0.29 * 0.71 / 760)
= 2.7%
Range : 29.1% + 2.7% = 31.8%
29.1% 2.7% = 26.4%
Range = 26.4% to 31.8%

For Product 2:
17.4% + 13.3% = 30.7%
Error at Confidence level
90% with N = 760
=1.645 * sqrt (0.31 * 0.69 /
760) = 2.8%
Range : 30.7% + 2.8% =
33.5%
30.7% 2.8% =
27.9%
Range = 27.9% to 33.5%

Since the test products


and the control have
overlaps between their
ranges, we will not be
able to launch the new
products even with 90%
confidence level based on
taste.

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Mouth-Feel at Confidence Level =


For Product 2:
95%

For Product 1:
20.9% + 9.6% = 30.5%
Error at Confidence level
95% with N = 760
1.96 * sqrt (0.31 * 0.69 / 760) =
3.2%
Range : 30.5% + 3.2% = 33.7%
30.5% 3.2% = 27.3%
Range = 27.3% to 33.7%
For Product 3:
16.7% + 11.4% = 28.1%
Error at Confidence level 95%
with N = 760
1.96 * sqrt (0.28 * 0.72 / 760) =
3.2%
Range : 28.1% + 3.2% = 31.3%
28.1% 3.2% = 24.9%
Range = 24.9% to 31.3%

15.9% + 12.4% = 28.3%


Error at Confidence level
95% with N = 760
1.96 * sqrt (0.28 * 0.72 / 760)
= 3.2%
Range : 28.3% + 3.2% =
31.5%
28.3% 3.2% =
25.1%
Range = 25.1% to 31.5%

Since the test products


and the control have
overlaps between their
ranges, we will not be
able to launch the new
products with 95%
confidence level based on
mouth-feel.

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Mouth-Feel at Confidence Level =


For Product 2:
90%

For Product 1:
20.9% + 9.6% = 30.5%
Error at Confidence level
90% with N = 760
1.645 * sqrt (0.31 * 0.69 / 760)
= 2.7%
Range : 30.5% + 2.7% = 33.2%
30.5% 2.7% = 27.8%
Range = 27.8% to 33.2%
For Product 3:
16.7% + 11.4% = 28.1%
Error at Confidence level 90%
with N = 760
1.645 * sqrt (0.28 * 0.72 / 760) =
2.6%
Range : 28.1% + 2.6% = 30.7%
28.1% 2.6% = 25.5%
Range = 25.5% to 30.7%

15.9% + 12.4% = 28.3%


Error at Confidence level
90% with N = 760
1.645 * sqrt (0.28 * 0.72 /
760) = 2.6%
Range : 28.3% + 2.6% =
30.9%
28.3% 2.6% =
25.7%
Range = 25.7% to 30.9%

Since the test products


and the control have
overlaps between their
ranges, we will not be
able to launch the new
products even with 90%
confidence level based on
mouth-feel.

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Cross Tab Analysis is also done on the less important attributes, i.e. Aroma, Smoothness, Flavour
and Throat-Feel
Aroma, Confidence Level =
95%

Smoothness, Confidence
Level = 95%

Range For Product 1: 28.8% to


35.4%
Range For Product 2: 23.9% to
30.3%
Range For Product 3: 23.2% to
29.4%

Range For Product 1: 30.3% to


37.1%
Range For Product 2: 27% to
33.6%
Range For Product 3: 23.8% to
30.2%

Overlap, hence should not


launch based on Aroma at

Overlap, hence should not


launch based on Smoothness

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

Cross Tab Analysis is also done on the less important attributes, i.e. Aroma, Smoothness, Flavour
and Throat-Feel
Flavour, Confidence Level =
95%

Throat-feel, Confidence Level


= 95%

Range for Product 1: 26.8 to 33.2


Range for Product 2: 22.9 to 29.1
Range for Product 3: 22.9 to 29.1

Range for Product 1:27.8 to 34.2


Range for Product 2: 24.9 to 31.1
Range for Product 3:24.9 to 31.1

Overlap, hence should not


launch based on Flavour at

Overlap, hence should not


launch based on Throat-feel

Which are the attributes that drive overall preference of vodka?

From R value we could interpret that there is a strong positive corelation between dependent & independent variable.
R-Square value I high, indicating that independent variable is able
to explain larger portion of dependent variable.

Results shows that there is no significant


difference between the means of
dependent & independent variables

Observing Sig. value we could conclude that all attributes except attribute 7 (Likeability of After-taste) drive the
overall performance of Vodka.

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

What is the extent to which they


drive?

We removed the insignificant attribute att7


Observing Sig. value we could conclude that all the 6 attributes drive the overall performance of Vodka.

% Impact
9%
10%

30%

12%

18%

Q5A_att3

Q5A_att8

Q5A_att2

21%
Q5A_att4

Q5A_att6

Q5A_att5

Action
Item

an the set of attributes be reduced to a smaller set of


actors

Sig. (0.000) < 0.05, there is no correlation between variables and thus we can do factor analysis
KMO Sampling adequacy is .945 ( ie. >0.5), we can conclude that sample size is adequate.

We can interpret that by having single factor, 78.3 % variance of dependent variable could be explained by
independent variables

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Can the set of attributes be reduced to a smaller set of


Factors

Secondar
y
Objective
Analysis

Summary

Conclusio
n

In order to improve the total variance explained, we have considered number of factors as 3.

KMO Sampling adequacy is .945 ( ie. >0.5), we can conclude that sample size is adequate.

We can interpret that by having three factor, 88.4 % variance of dependent variable could be explained by
independent variables

Action
Item

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Can the set of attributes be reduced to a smaller


set of Factors

Note :- In the above matrix, we have supressed absolute


value below .55.
From the above Matrix, we could conclude that attributes
could be divided based on the below three factors: Experience (Att3, Att5, Att7, Att8)
Senses (Att4, Att6)
Aroma (Att2)

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Action
Item

LikeabilityofTaste
LikeabilityofFlavour
LikeabilityofAfter-taste
LikeabilityofMouth-feel

Experie
nce

Senses

LikeabilityofSmoothness
LikeabilityofThroat-feel
when the vodka goes down

Aroma

LikeabilityofAroma

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Factor Analysis

Regression EquationY= 0.569 * Experience + 0.499 * Senses + 0.440 * Aroma


Factor 1 Experience has the highest beta value and therefore it has
the highest significance on the dependent variable.

Action
Item

Background

Objective

Primary
Objective
Analysis

Attribute

Secondary
Objective
Analysis

Summary

Conclusion

Action Item

If the purchase intention is taken as a categorical variable (YES / NO), can it be predicted
in future, by reading some of the ratings of the attributes only? How?
Possible to predict the
purchase intention for a case
if information/ratings of the
attributes are available for
that case.
Predict group membership of
a case.
Purchase intention
A predictive model built using
ratings of the attributes will
classify and put the case in
appropriate group.
Two categories - Yes or No
Discriminant function
analysis.
Discriminant Analysis used
continuous variables as input
and gives the output of

Variables that appear in Discriminant Analysis.


Discriminating Variables
Q2A
Q3A
5A_Attr2
5A_Attr3
5A_Attr4
5A_Attr5
5A_Attr6
5A_Attr7
5A_Attr8
Grouping

Variable

Likeability of Aroma, after pouring the


neat vodka
Likeability of Aroma, after adding relevant
mixer
LikeabilityofAroma
LikeabilityofTaste
LikeabilityofSmoothness
LikeabilityofFlavour
LikeabilityofThroat-feel when the vodka goes
down
LikeabilityofAfter-taste
LikeabilityofMouth-feel when the vodka is
sipped

Discriminates between the case based on the ratings given to the


attributes.
Dependent variable in the predictive model derived from the
discriminant analysis.
Variables used for discrimination is Q6.
Intention to buy the vodka blend.
Yes and No categories.

Background

Objective

Primary
Objective
Analysis

Attribute

Secondary
Objective
Analysis

Summary

Conclusion

Action Item

Test in SPSS

Test in SPSS

Tests of Equality of Group Means


Wilks'
F
df1
df2
Lambda
.828
472.729
1
2278

Sig.

Q2A
Q3A
Q5A_att2
Q5A_att3
Q5A_att4
Q5A_att5
Q5A_att6
Q5A_att7
Q5A_att8

.797
579.505
1
.727
856.331
1
.683
1057.337
1
.726
860.431
1
.733
829.545
1
.720
883.702
1
.742
791.519
1
.718
896.497
1
Eigenvalues
Funct Eigenv % of
Cumula Canonic
ion
alue Varianc tive %
al
e
Correlati
on
a
1
.592
100.0
100.0
.610

Wilks' Lambda
Wilks'
Lambda

Chi-square

df

Sig.

.628

1057.567

.000

2278
2278
2278
2278
2278
2278
2278
2278

.000
.000
.000
.000
.000
.000
.000
.000
.000

Functio
n
1
.089
.110
.226

Q2A
Q3A
Q5A_att2
Q5A_att3
Q5A_att4
Q5A_att5
Q5A_att6
Q5A_att7
Q5A_att8

.365
.139
.037
.201
-.056
.137

Classification of cases
Once we have computed the classification scores for a case, it is easy
to decide how to classify the case: in general we classify the case as
belonging to the group for which it has the highest classification
score. Thus, if we are to find if a person is going to buy a vodka
brand, we could put in the values of variables in the classification
functions to predict what each is most likely to do with a vodka blend
show intention to buy or not buy.

Reliability of the predictive Model


85.5% of original grouped cases correctly classified.

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Can the Current Product (Control) be replaced by any of the two Test Products?
We considered Overall Likeability as 1 attribute & compared to all three products, we
found
We could not launch the new blend with 95% confidence BUT
We could launch the product with 90% confidence
We considered taste as 1 attribute & compared to all three products, we found
We could not launch the new blend under both the confidence level ( 90% and 95%).
We considered mouth feel as 1 attribute & compared to all three products, we found
We could not launch the new blend under both the confidence level ( 90% and 95%).

Action
Item

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Which are the attributes that drive overall preference of vodka?


The Most influencing attributes that drive overall preference we found from the study
are:
Likeability of taste with 30.03% impact on overall preference
Mouth-feel with 20.62% impact
Aroma with 17.43% impact
We found that likeability after taste is insignificant on overall preference of vodka
Hence we have removed that parameter.

Action
Item

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

Can the set of attributes be reduced to a smaller set of Factors.


We found that it can be divided into only 1 factor which defined 78.3 % variance of dependent variable
which could be explained by independent variables.
But In order to improve the total variance explained, we have considered number of factors as 3
We divided the attributes into 3 set of factors which are:
Experience
Senses
Aroma

If the purchase intention is taken as a categorical variable (YES / NO), can it be predicted in
future, by reading some of the ratings of the attributes only? How?
We have done discriminant analysis and divided into grouping variables
We have found that it is Possible to predict the purchase intention for a case if ratings of the attributes are
available

Action
Item

Backgrou
nd

Research
Design

Objective

Attribute

Primary
Objective
Analysis

Secondar
y
Objective
Analysis

Summary

Conclusio
n

We wouldnt be launching the product on 95% confidence level as the cross tab analysis show
none of the two products when tested for all attributes are significantly better in Overall
Likeability than the existing product.
We could launch the product on 90% confidence level as the cross tab analysis show no
overall with the two product in case of overall likeability.
We have only considered rating 9 & 10 as they are generally acceptable by the customer.
Even when Eigen value is less than 1 for other 2 factors, in order to improve the total
variance explained ,we have taken 3 factors
The data given is for region wise but we have not concentrated on that as we wanted to have
a study of the market on a broader sense rather than changing the blend of the product
region-wise

Action
Item

You might also like