Perceptual Mapping: MDPREF

1
Perceptual Mapping MDPREF

• Scott Smith
• Brigham Young University

2
Positioning Using Perceptual Preference Maps

• Segmentation Identification and targeting groups
  of customers most likely to purchase the products
  and services being offered.
• Differentiation Product differentiation
  typically occurs in the growth stage of the
  product life-cycle. The objective of product
  differentiation is to design and offer a product
  that has meaningful differences (tangible or
  intangible) on one or two key dimensions that
  distinguish it from the competitors.
• Positioning Product positioning is an even more
  aggressive marketing strategy that occurs after
  product differentiation has been implemented.
  Product positioning typically occurs in the
  maturity stage of the product lives cycle and
  focuses on targeting a specific product segment
  with a product having the features and benefits
  that they most desire. Positioning strategies
  are implemented to emphasize key differences
  between your product and your competitors so
  that you can gain a more competitive or distinct
  position in the minds of customers.
• Perceptual MappingStatistical Techniques that
  allow the display of objects in multivariate
  space. In marketing, managers develop
  differentiation and positioning strategies by
  visualizing the competitive structure of their
  markets as perceived by their customers.

3
The Essence of Positioning Studies

• There are multiple attributes and benefits that
  are of value to customers.
• A single product has difficulty providing
  outstanding performance across all attributes and
  benefits.For example, high gas mileage is
  incompatible with the towing power of a 9
  passenger 4x4 Chevrolet Suburban.
• Markets are segmented to find similar groups of
  consumers who seek common bundles of attributes
  and benefits. New benefit bundles can be used to
  form new concepts.
• There are differences in products that supply
  those attributes and benefits.

4
Conventional MappingTwo Products on a Snake
Chart

• 1. Company provides adequate insurance coverage
  for my car.
• 2. Company will not cancel policy because of
  age, accident experience, or health problems.
• 3. Friendly and considerate.
• 4. Settles claims fairly.
• 5. Inefficient, hard to deal with.
• 6. Provides good advice about types and amounts
  of coverage to buy.
• 7. Too big to care about individual customers.
• 8. Explains things clearly.
• 9. Premium rates are lower than most companies.
• 10. Has personnel available for questions all
  over the country.
• 11. Will raise premiums because of age.
• 12. Takes a long time to settle a claim.
• 13. Very professional/modern.
• 14. Specialists in serving my local area.
• 15. Quick, reliable service, easily accessible.
• 16. A good citizen in community.
• 17. Has complete line of insurance products
  available.
• 18. Is widely known name company.
• 19. Is very aggressive, rapidly growing company.

Company A Company B
5
Example Plot of Attributes of Laptops on a 2D
Perceptual Map
(Plain)
Common
Toshiba 1960CT
Easy setup
Slow
Performance
Light
GoodValue
IBM 701 CButterfly
Elegant
Looks
6
How to Develop A Perceptual Map Using Attribute
Ratings

• Generate an Brand by Attribute matrix of inputs
  consisting of each consumers ratings of each
  brand on each of the attributes (A1, A2, A3,....)
• A1 A2 A3 A4 ............... A15
• Dell 710 6 3 7 2 2
• Compaq 8100 4 3 4 1 5
• Toshiba Construct 3 6 2 7 7
• Note that in this matrix the columns become the
  vectors in the map and rows are the points in the
  map
• Compute average ratings of each brand on each
  attribute. Submit data to a suitable perceptual
  mapping technique (e.g., MDPREF or Factor
  Analysis).
• Interpret the underlying key dimensions of the
  map using the directions of the individual
  attributes.
• Explore the implications of how consumers view
  the competing products.

7
How to Develop A Perceptual Map of Market
Segments Using Attribute Ratings (INDSCAL,
PREFMAP)

• Generate an Brand by Attribute matrix of inputs
  consisting of each consumer segments (S1,
  S2,...) Ratings of each brand on each of the
  attributes (A1, A2, A3,....) for each brand
  (named).
• A1 A2 A3 A4 ............... A15
• Dell 710XN 6 3 7 2 2
• S1 Compaq 8100 4 3 4 1 5
• Toshiba Construct 3 6 2 7 7
• Dell 710XN
• S2 Compaq 8100
• Toshiba ConstructNote that in this matrix the
  columns become the vectors in the map and rows
  are the points in the map
• Compute average ratings of each brand on each
  attribute. Submit data to a suitable perceptual
  mapping technique (e.g., MDPREF or Factor
  Analysis).
• Interpret the underlying key dimensions of the
  map using the directions of the individual
  attributes.
• Explore the implications of how consumers view
  the competing products.

8
Interpreting MDPREF Perceptual Maps

• MDPREF is a Point-Vector Model. The points
  represent the rows of the preference matrix
  (brands) and the vectors represent the attributes
  that are evaluated
• The vector is of length 1.0 and the perceptual
  map should show an arrow indicating the direction
  toward which the attribute is increasing (The
  attribute is decreasing in the direction opposite
  to the arrow).
• The length of the line from the origin to the
  arrow is an indicator of the variance of that
  attribute explained by the attribute vector in
  that dimensional space (more dimensions add more
  variance explanation). The longer this line, the
  greater is the importance of that attribute in
  this space.

9
Interpreting Perceptual andPreference Maps

• Technical adequacy
• What percentage of variance in the raw data is
  captured in the map?
• What percentage of the variance of each attribute
  is captured in the map?
• Managerial interpretation
• What underlying dimensions characterize how
  consumers view the products?
• What is the competitive set associated with the
  new concept?
• How well is the new concept positioned with
  respect to the existing brands?
• Which attributes are related to each other?
• Which attributes influence customer preferences
  positively? Negatively?
• What improvements will enhance the value of the
  new concept?
• Which customer segments have positive perceptions
  and high preference for the new concept?

10
Example Input Data forMDPREF Vector Model
Input matrix has attributes on rows and objects
on columns B1 B2 B3 B3 B4 B5 B6 B7 B8 New Attract
ive 5.1 3.6 3.5 5.4 3.9 4.8 5.2 4.0 5.2 4.0Light
6.0 3.5 5.0 3.9 3.3 5.3 5.0 2.5 5.5 2.5Unreliable
3.4 4.1 4.5 2.1 4.5 2.7 4.5 3.7 2.5 3.8Plain 1.5
4.1 2.9 2.3 4.5 2.7 3.5 4.3 2.2 5.2Battery
life 3.3 4.9 4.3 4.1 3.9 3.0 3.5 6.2 3.5 4.0Scree
n 3.5 5.3 3.4 6.4 5.4 5.2 3.3 6.0 3.3 4.8Keyboard
2.6 3.5 2.5 3.4 3.8 3.3 2.8 5.0 4.3 4.7Roomy 5.5
4.3 5.4 3.1 3.4 3.3 4.7 3.5 4.3 4.2Easy
service 4.5 4.9 3.3 5.0 4.4 4.5 3.3 4.7 3.8 4.5Ex
pandability 5.5 4.3 5.4 3.1 3.4 3.3 4.7 3.5 4.3 4.
2Setup 5.6 3.5 5.6 5.4 2.5 4.2 5.2 3.3 5.8 2.5Co
mmon 4.1 3.5 3.3 2.9 4.0 4.3 2.2 4.2 3.3 4.2Value
3.5 4.8 4.4 3.6 3.6 2.7 3.2 4.7 3.5 4.0Preferenc
e 7.4 3.4 4.8 6.6 4.4 7.4 7.1 3.8 6.9 3.3
11
Preference Map Using MDPREF Vector Model
Low battery life
Keyboard
Expandability

• Toshiba

Elegant
Distinct

• New Concept

Unsuccessful
Avant-Garde
Heavy
Fast operation

• IBM
• Compaq

Reliable

• Sanyo

• Good design

• TI

• AST Exec

Difficult to use

• Dell

• NEC

Value Graphics
Poor setup

• Samsung

Screen quality

• Preference

12
Approaches To Creating Perceptual Maps
Perceptual map
Attribute data
Nonattribute data
Preference
Similarity
Correspondence analysis
Discriminant analysis
Factor analysis
MDS
13
Attribute Based Approaches

• Assumption
• The attributes on which the individuals'
  perceptions of objects are based, can be
  identified
• Methods Used to Reduce the Attributes to a
  Smaller Number of Dimensions
• Factor Analysis
• Discriminant Analysis
• Correspondence Analysis
• Multidimensional Preference Mapping

14
Distinctions Between Methods

• Factor Analysis
• Combines variables to create new factors and
  accesses the underlying constructs based on the
  commonality within each of the variables
  (variable common unique error)
• Factors New non-correlated variables Factor
  Scores Scores of respondents on factors
• Discriminant Analysis
• Used to classify the objects or people into
  predefined groups (user-non user) based on their
  attributes
• Discriminant function based on independent
  variables is used to predict the category
• Correspondence Analysis
• Used for convenience of collection of data in
  binary form (frequency counts)
• Selection of some attributes or listing of user
  perceived attributes to reduce attributes
• Multidimensional Preference Analysis
• Used to map in joint attribute and brand space,
  the preference for brands based on the attributes
  used to evaluate them
• Underlying dimensions, positioning of brands and
  positioning of brands with respect to dimensions
  and attributes

15
Factor Analysis Why do we look at dimensions

• We study phenomena that can not be directly
  observed
• (ego, personality, intelligence, perceptions of
  attributes and products)
• We have too many attributes or variables
• need to reduce them to a smaller set of factors
• We identify attribute-variable Items that
  describe an underlying set of latent factors.
• We want to know what these factors are.
• We have an idea of the phenomena that a set of
  items represent (construct validity).
• We find underlying latent constructs
• As manifested in individual items
• We assess the association between these factors
• We produce usable scores that reflect critical
  aspects of any complex phenomenon
• (e.g. Attributes, life style, personality,
  intelligence, etc.)
• This is an end in itself in terms of defining
  structure. Factor analysis is also a major step
  toward creating measures of the dimensions or
  constructs that define the behavior or activity
  of interest.

16
The Basic Idea of Factor Analysis

• If two items are highly correlated
• They must represent the same phenomenon
• If they tell us about the same underlying
  variance, combining them to form a single measure
  is reasonable for two reasons
• Parsimony
• Reduction in Error
• Correlated variables can be represented by a
  regression line
• BUT suppose a whole group of variables provide
  information that represents this underlying
  phenomena.
• FACTOR ANALYSIS looks for the phenomena
  underlying the observed variance and covariance
  in a set of variables.
• These phenomena are called factors or
  principal components.

17
Factor Analysis
Two variable situation
Three variable situation
Refers to common variance or extracted factor
Refers to common variance or extracted factor
18
What Happens - Start With Correlation Matrix
19
FACTOR ANALYSIS - communalities

• A measure of how much variance is or can be
  accounted for by the observed factors
• Uniqueness is 1-communality
• With Principal Components Analysis with all
  factors, Communality always 1 (100 of variance
  is explained by common unique factors
• With FA, the initial value is the maximum
  multiple R2 for the association between a item
  and any of the other items in the model

PCA
FA
20
What HappensComputing Factors
Eigenvalue/N of items
An Eigenvalue is an index of the strength of the
factor. An eigenvalue reports the amount of
variance accounted for by the factor. It is the
sum of the squared loadings (correlations between
the variables and the factor).
21
What happensFactor Loadings
Eigenvalue of factor 1 .612 .612 .592 .732
.772 .762 2.802
22
Rotation to Make the Solution more Interpretable

• Makes solution more interpretable
• Orthogonal (Uncorrelated, or geometrically,
  having spatial representations at 90 degree
  angles) or Oblique (correlated factors, or
  geometrically having spatial representations at
  angles other than 90 degrees).
• ROTATION IS FOR INTERPRETATION PURPOSES
• Varimax is the most commonly used.

23
Real Example

• To further examine attendance success at offshore
  US and domestic Japanese trade shows, and to
  explore the role of prior trade show attendance
  success in generating interest in that show in
  the future, an orthogonal principal components
  factor analysis was conducted.
• All factors with Eigenvalues of 1.0 or greater
  were retained in the final solutions, yielding
  four-factor solutions for both the US and
  Japanese shows, explaining 74 percent and 78
  percent of each set of variables common
  variance, respectively.
• With factors identified, factor scores were then
  regressed upon a measure of future interest for
  each respective trade show1. Therefore, the
  results from factor analyses presented in Tables
  IV and V are useful in two predominant ways
• (1) to identify underlying dimensions of Japanese
  attendance objective success at a US and domestic
  show and
• (2) to address effectively issues of
  multicollinearity between independent variables
  (trade show success ratings) when exploring the
  impact of success of prior trade show attendance
  on interest in future attendance.
• Because a factor is a qualitative dimension, the
  researcher is required to name each factor based
  on an interpretation of the variables loading
  most heavily on it.

24
Success of Japanese Attendees at a US Show
25
Success of Japanese Attendees at a Japanese Show
26
How Do We Evaluate the Quality of the Solution?

• Does it make sense?
• Are the values in the reproduced matrix small?
• Are the final communalities large when the number
  of factors is small?
• Is it useful?
• Think of why you did the analysis to begin with.
  OR
• 1. Dont evaluate, just accept it as reality.
• Try using theory to validate

27
Basic Concepts of Multidimensional Scaling(MDS)

• MDS uses proximities among different objects as
  input
• Proximity value which denotes how similar or
  how different two objects, are perceived to be
• MDS uses this proximities data to produce a
  geometric configuration of points (objects), in a
  two-dimensional space as output

28
Advantages of Attribute-based MDS

• Attributes can have diagnostic and operational
  value
• Attribute data is easier for the respondents to
  use
• Dimensions based on attribute data predicted
  preference better as compared to non-attribute
  data

29
Disadvantages of Attribute-based MDS (MDPREF)

• If the list of attributes is not accurate and
  complete, the study will suffer accordingly
• Respondents may not perceive or evaluate objects
  in terms of underlying attributes
• May require more dimensions to represent them
  than the use of flexible models

30
Application of MDS With Nonattribute Data (KYST)

• Similarity Data
• Reflect the perceived similarity of two objects
  from the respondents' perspective
• Perceptual map is obtained from the average
  similarity ratings
• The power of the technique lies in the ability to
  find the smallest number of dimensions for which
  there is a reasonably good fit between the input
  similarity rankings and the rankings of the
  distance between objects in the resulting space
• Preference Data
• An ideal object is the combination of all
  customers' preferred attribute levels
• Location of ideal objects is to identify segments
  of customers who have similar ideal objects,
  since customer preferences are always
  heterogeneous

31
Evaluating the MDS Solution

• The fit between the derived distances and the two
  proximities in each dimension is evaluated
  through a measure called stress
• The appropriate number of dimensions required to
  locate the objects can be obtained plotting the
  stress values against the number of dimensions

32
General Issues in MDS

• Perceptual mapping has not been shown to be
  reliable across different methods, but methods
  like KYST (for similarities or dissimilarities
  data) are pretty good.
• The effect of market events on the perceptual
  maps cannot be ascertained
• The interpretation of dimensions is difficult
• When more than two or three dimensions are
  needed, the usefulness is reduced