8.mv - correspondence analysis
TRANSCRIPT
-
8/13/2019 8.MV - Correspondence Analysis
1/17
December 25, 2013
Application of Multivariate Statistical Methods inMarketing Research
Industrial Statistics
MS3001
Advanced Marketing Research
Faculty of ScienceUniversity of Colombo
Session 5Correspondence Analysis
-
8/13/2019 8.MV - Correspondence Analysis
2/17
December 25, 2013
Illustration
I am keen on understanding how consumers perceive retailchains (banners) on key attributes.
I want to discover positioning gaps, if any, in the oral care
category.
What are the prominent imagery parameters in my category?
What are the parameters where my brand is strong, and which
my brand owns?
Correspondence Analysis
-
8/13/2019 8.MV - Correspondence Analysis
3/17
December 25, 2013
Correspondence Analysis
An attribute-based positioning exercise.
Respondents associate objects (brands) on a set of attributes. No rating or ranking involved, only yes and no responses.
The resulting frequency data are used to generate a
correspondence map, often called a brand map.
Too many objects (brands) and too many attributes could resulted in
maps that are useless.
Typically 2D outputs are forced. Maps with more dimensions haveproblems of interpretation.
-
8/13/2019 8.MV - Correspondence Analysis
4/17
-
8/13/2019 8.MV - Correspondence Analysis
5/17
December 25, 2013
What is Correspondence Analysis (CA)?
Correspondence analysis is a technique for summarizinglarge tables of data in terms of a visual map CA analyses respondents perceptions of the similarity or
dissimilarity of certain brands, products and services across arange of attributes
Maps present simple graphical summaries of a market: forexample: Brand positioning: the relationship between brands and attributes The relationship between current brand positioning and the ideal
positioning Image ratings by brand users, segments, etc
Research Questions Answered?What attributes do consumers associate my brand with?What are my brands / competitors strengths and
weaknesses maps present results of cross-tabs or count data visually need to consider the absolute scores and relative scores in
explaining the research findings
-
8/13/2019 8.MV - Correspondence Analysis
6/17
December 25, 2013
Data Table: Cereal Brands Image DataCOCO
POPS FRUITYBIXKELLOGG
S CORN
FLAKESKELLOGG
S RICE
BUBBLES NUTRI-GRAIN VITABRITS WEET-BIXWEETBI
X
CRUNCH MILOHigh in fibre 3% 11% 11% 4% 19% 41% 73% 2% 1%Good source of
energy 11% 12% 26% 12% 46% 34% 63% 2% 6%Most nutritious
breakfast 2% 8% 17% 5% 19% 29% 65% 1% 1%Meets my familys
needs 14% 8% 36% 17% 25% 21% 54% 1% 3%Australian owned &
made 6% 5% 19% 10% 11% 18% 53% 2% 2%Children like the
taste 69% 12% 23% 34% 32% 8% 22% 1% 8%Good for kids
10%
12%
31%
19%
22%
33%
66%
2%
3%
Good value for
money 8% 4% 35% 16% 12% 23% 60% 1% 1%Like the taste 37% 12% 43% 26% 38% 19% 47% 2% 5%Meets my needs 11% 7% 31% 13% 22% 20% 53% 1% 3%Low in sugar 2% 4% 24% 13% 8% 37% 70% 1% 0%Convenient 33% 17% 49% 33% 36% 29% 61% 2% 6%My kids want it 45% 5% 14% 21% 23% 4% 17% 1% 5%Everyone eats it 18% 3% 48% 19% 20% 11% 46% 1% 3%A brand I trust 22% 10% 53% 31% 29% 26% 64% 2% 4%Number 1 cereal
brand 6% 1% 33% 7% 9% 4% 28% 1% 1%
-
8/13/2019 8.MV - Correspondence Analysis
7/17
December 25, 2013
Input for Correspondence Maps
Attitudinal data are most common
Brand association grids are a typical type of input Anything with absent / present type scores is appropriate (eg. Yes
associate that brand with that attribute, or no dont associate it) Tables of either percentages or raw numbers are acceptable
Means can be used
Whether based on means, or percentages, correspondence maps
usually provide similar results. Often maps are just based onpercentage data
-
8/13/2019 8.MV - Correspondence Analysis
8/17
December 25, 2013
Points to consider withCorrespondence analysis
What is the minimum number of attributes?This is subjective, but a map of data with fewer than four brands
(columns) or 8 attributes (rows) may be relatively uninformative
Sample size issues are less critical than insegmentation studies, as analysis has a qualitative feel
about itBut a sample size of between 200-400 would be a minimum
threshold
Care is needed with interpretationOverplaying weak relationships
Underplaying strong relationshipsUsing overly precise language in describing the map
-
8/13/2019 8.MV - Correspondence Analysis
9/17
December 25, 2013
COCO
POPS FRUITY BIX
KELLOGGS
CORN
FLAKES
KELLOGGS
RICE
BUBBLES
NUTRI-
GRAIN VITA BRITS WEET-BIX
WEETBIX
CRUNCH MILO Average
High in fibre 3% 11% 11% 4% 19% 41% 73% 2% 1% 18%
Good source of energy 11% 12% 26% 12% 46% 34% 63% 2% 6% 24%
Most nutritious breakfast 2% 8% 17% 5% 19% 29% 65% 1% 1% 16%
Meets my familys needs 14% 8% 36% 17% 25% 21% 54% 1% 3% 20%
Australian owned & made 6% 5% 19% 10% 11% 18% 53% 2% 2% 14%
Children like the taste 69% 12% 23% 34% 32% 8% 22% 1% 8% 23%
Good for kids 10% 12% 31% 19% 22% 33% 66% 2% 3% 22%
Good value for money 8% 4% 35% 16% 12% 23% 60% 1% 1% 18%
Like the taste 37% 12% 43% 26% 38% 19% 47% 2% 5% 25%
Meets my needs 11% 7% 31% 13% 22% 20% 53% 1% 3% 18%
Low in sugar 2% 4% 24% 13% 8% 37% 70% 1% 0% 18%
Convenient 33% 17% 49% 33% 36% 29% 61% 2% 6% 30%
My kids want it 45% 5% 14% 21% 23% 4% 17% 1% 5% 15%
Everyone eats it 18% 3% 48% 19% 20% 11% 46% 1% 3% 19%
A brand I trust 22% 10% 53% 31% 29% 26% 64% 2% 4% 27%
Number 1 cereal brand 6% 1% 33% 7% 9% 4% 28% 1% 1% 10%
Average 18% 8% 31% 17% 23% 22% 53% 1% 3%
Data Table: Cereal Brands Image Data
-
8/13/2019 8.MV - Correspondence Analysis
10/17
December 25, 2013
Correspondence Map: Example 1
High in fibre
Good source of energy
t nutritious breakfast
Meets my familys needs
Australian owned & made
Children like the taste
Good for kids
Good value for money
Like the taste
Meets my needs
ugar
Convenient
My kids want it
Everyone eats it
A brand I trust
Number 1 cereal brand
COCO POPS
FRUITY BIX
KELLOGGS CORN FLAKES
KELLOGGS RICE BUBBLES
NUTRI-GRAIN
VITA BRITS
WEET-BIX
WEETBIX CRUNCH MILO
Axis 2 16.8%
Axis 1
73.7%
Cereal Brand Ima e Data
= Correlation < 0.50
-
8/13/2019 8.MV - Correspondence Analysis
11/17
December 25, 2013
Interpreting the Map
Brands that are closeto each other are seen to have
similarprofiles in the eyes of the consumer
Brands are located next to attributes which are their
greatest relative strength(I.e. consumers feel most
characterise the brand)
Attributes that differentiatethe brands are close to theedges.
Attributes that do not discriminate (i.e. could be
considered are generic to the category) are located near
the centre of the map
The axes also have meaning (but not in terms of positivev/s negative)the horizontal is more important than the
vertical. Thus, the position of a brand relative to the
horizontal axis is more important than its location vertically
-
8/13/2019 8.MV - Correspondence Analysis
12/17
December 25, 2013
Interpreting Correspondence Maps
Angle in d Correlation Level of Correlation/Association
0 1 Perfect +ve
15 0.97 +ve Correlation
30 0.87 +ve Correlation
45 0.71 +ve Correlation
60 0.5 Some +ve
75 0.26 Small +ve
90 0 No association105 -0.26 Small oppostite -ve
120 -0.5 Some oppostite -ve
135 -0.71 -ve Correlation
150 -0.87 -ve Correlation
165 -0.97 -ve Correlation
180 -1 Perfect opposite ve
Distance from the origin to the brand or attribute:Brands furthest
from the origin, particularly horizontally (east or west), are more
distinct than brands nearer the middle of the map. Similarly forattributes.
Relationships between brands and attributes:The smaller the angle
between a brand and an attribute the more that attribute applies to
that brand. Brands that are 180 degrees apart have the opposite
positioning to each other. Brands at right angles are simply different
or uncorrelated. Attributes that are at right angles to a brand have
no association with that brand.
-
8/13/2019 8.MV - Correspondence Analysis
13/17
December 25, 2013
Correspondence Map: UK ST Data Example 2
ASDA SAINSBURY MORRISONS SOMERFIELD WAITROSE LIDL ALDI COOP KWIK SAVE NETTO
MARKS &
SPENCER TESCO
Convenient to get to 49 58 50 48 37 41 42 57 19 29 35 69
Staff provide good service 60 66 54 34 67 28 37 40 21 18 73 65
Food and Groceries are VFM 75 55 67 34 40 71 70 31 47 54 36 75
Everything I need in the one shop 70 59 52 19 41 17 19 15 14 11 22 74
Well presented display of products 64 71 56 30 65 20 28 26 18 15 69 69
Always have what I want in stock 53 49 45 18 46 18 19 15 14 12 33 54
Better selection of high quality products 52 66 46 18 72 15 17 17 10 13 72 62
Wide product range and variety 70 71 60 24 59 25 29 19 17 15 45 74
Is close to home 44 52 46 46 35 40 37 56 21 32 37 60
Modern, comfortable store 63 69 53 28 62 25 33 24 15 19 65 69
Good quality instant foods 59 65 54 26 65 21 26 22 15 13 76 65
Good range of fresh products 66 75 66 32 72 30 32 24 19 18 72 74
Attractive and interesting promotions 56 54 60 33 33 48 49 25 22 37 35 66
It's easy to find what I need 52 56 53 33 47 34 41 39 27 26 45 64
Low prices for most items 74 38 62 30 17 73 74 23 38 58 13 67
Clean and hygienic store 67 74 64 38 70 45 55 43 26 31 81 76
Ease of parking 67 66 64 36 50 62 65 33 26 46 28 72
High quality fresh food 65 75 65 33 76 32 36 31 19 21 82 72
Spacious 68 66 55 27 59 28 34 22 21 19 51 71
Efficient checkout counters 61 63 53 28 63 28 36 26 20 17 61 67
Provides good own brand alternatives 67 72 57 36 58 34 44 36 27 18 72 78
Long opening hours 74 66 47 23 33 26 22 42 15 22 24 86
Has programs that reward regular purch 25 69 27 13 20 8 8 28 7 4 18 83
-
8/13/2019 8.MV - Correspondence Analysis
14/17
December 25, 2013
Convenient to get to
Staff provide good service
Food and Groceries are VFM
Everything I need in the one shop
Well presented display of products
Always have what I want in stock
Better selection of high quality products
Wide product range and variety
Is close to home
Modern, comfortable store
Good quality instant foodsGood range of fresh
Good promotions
It's easy to find what I need
Low prices for most items
Clean and hygienic store
Ease of parking
High quality fresh food
Spacious
Efficient checkout counters
Provides good own brand alternatives
Long opening hours
Has programs that reward regular purchases
ASDA
SAINSBURY
MORRISONS
SOMERFIELD
WAITROSE
LIDL
ALDI
COOP
KWIK SAVE NETTO
MARKS & SPENCER
TESCO
Axis 2 19%
Axis 1
61.4%
2008 ACNielsen, 2006= Correlation < 0.50
Correspondence Map: UK ST Data Example 2
-
8/13/2019 8.MV - Correspondence Analysis
15/17
December 25, 2013
Data Table: Cereal Brands Image Data
COCO
POPS FRUITYBIXKELLOGGS
CORN
FLAKESKELLOGGS
RICE
BUBBLES NUTRI-GRAIN VITA BRITS WEET-BIX WEETBIXCRUNCH MILOHigh in fibre 3% 11% 11% 4% 19% 41% 73% 2% 1%Good source of energy 11% 12% 26% 12% 46% 34% 63% 2% 6%Most nutritious breakfast 2% 8% 17% 5% 19% 29% 65% 1% 1%Meets my familys needs 14% 8% 36% 17% 25% 21% 54% 1% 3%
Australian owned & made 6% 5% 19% 10% 11% 18% 53% 2% 2%Children like the taste 69% 12% 23% 34% 32% 8% 22% 1% 8%Good for kids 10% 12% 31% 19% 22% 33% 66% 2% 3%Good value for money 8% 4% 35% 16% 12% 23% 60% 1% 1%Like the taste 37% 12% 43% 26% 38% 19% 47% 2% 5%Meets my needs 11% 7% 31% 13% 22% 20% 53% 1% 3%
Low in sugar 2% 4% 24% 13% 8% 37% 70% 1% 0%Convenient 33% 17% 49% 33% 36% 29% 61% 2% 6%My kids want it 45% 5% 14% 21% 23% 4% 17% 1% 5%Everyone eats it 18% 3% 48% 19% 20% 11% 46% 1% 3%A brand I trust 22% 10% 53% 31% 29% 26% 64% 2% 4%Number 1 cereal brand 6% 1% 33% 7% 9% 4% 28% 1% 1%
-
8/13/2019 8.MV - Correspondence Analysis
16/17
December 25, 2013
Biplot: Example Cereals
Biplots use
absolutedata values
High in fibre
Good source of energy
Most nutritious breakfast
Meets my familys needs
Australian owned & made
Children like the taste
Good for kids
Good value for money
Like the taste
Meets my needs
Low in sugar
Convenient
My kids want it
Everyone eats it
A brand I trust
Number 1 cereal brand
COCO POPS
FRUITY BIX
KELLOGGS CORN FLAKES
KELLOGGS RICE BUBBLES NUTRI-GRAIN
VITA BRITS
WEET-BIX
WEETBIX CRUNCH
MILO
Axis 2 16%
Axis 1
77.1%
Cereal Brand Ima e Data Projector
Projector II
-
8/13/2019 8.MV - Correspondence Analysis
17/17
December 25, 2013
Correspondence Analysis -Summary
CA.... Summarises large amount of
information from tables
succinctly and visually
Identifies relationships
between statements, betweenbrands & between statements
and brands
Removes halo effects of
brands as it is a relative
analysis Probably need to show
absolute scores as well
But....
CA can... Be misinterpreted - map
presented visually, highlights
relative strengths of brands
mean numbers from analysis
difficult to interpret Be hard to compare different
different maps - how different
they are?
Should be described in
qualitative, or passivelanguage...eg brands tends to
be or near to