This module introduces two perceptual mapping methodologies attribute rating and overall similarity Authors Ron Wilcox and Stu James 2013 Ron Wilcox Stu James and Management by the Numbers Inc ID: 145198
Download Presentation The PPT/PDF document "Perceptual Mapping" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Perceptual Mapping
This module introduces two perceptual mapping methodologies: attribute rating and overall similarity.
Authors: Ron Wilcox and Stu James© 2013 Ron Wilcox, Stu James and Management by the Numbers, Inc.Slide2
Perceptual Mapping2
Perceptual Mapping
MBTN | Management by the NumbersThere are two primary methods of constructing perceptual maps from consumer-level data:Attribute Rating Method
Overall Similarity Method
Insight
“Simple Graphics are often the most powerful way to communicate complicated statistical information.”
- Edward R.
Tufte
, The Visual Display of Quantitative InformationSlide3
Sample Perceptual Map (Attribute Rating Method)
3Sample Perceptual Map (Attribute Rating)
MBTN | Management by the Numbers
Prestigious
Not
Prestigious
Cadillac Escalade
Chevy Tahoe
Jeep Grand Cherokee
Nissan Pathfinder
Hummer
H2
Toyota
4Runner
Ford
Explorer
Jeep
Cherokee
Not Rugged
Rugged
The Market for Sports Utility Vehicles (circa 2005)Slide4
I believe
[insert product name] is an excellent [insert product category].
StronglyAgreeAgreeNeither Agree Nor DisagreeDisagree
Strongly
Disagree
How is this Perceptual Map Created?
4
How is
this
Perceptual Map Created?
MBTN | Management by the Numbers
Strongly
Agree
Agree
Neither Agree
Nor
Disagree
Disagree
Strongly
Disagree
[Insert product
name] is a [insert attribute]
[insert product category].
One question for each attribute for each product
Direct measurement of consumer perceptions of attributes and products based on the following question format:
One question for
each
productSlide5
How is this Perceptual Map Created?
5How is this Perceptual Map Created?
MBTN | Management by the NumbersNext, calculate the average attribute rating for each vehicle based on the survey population (summarize 1st question)
Product 1
Product 2
………..
Product
N
Attribute 1
Attribute 2
[average rating]
………….
Attribute
MSlide6
How is this Perceptual Map Created?
6How is this Perceptual Map Created?
MBTN | Management by the NumbersThen, calculate the Avgerage Preference Rating score for each product based on the survey population (summarize 2nd question)Slide7
How is this Perceptual Map Created?
7How is this Perceptual Map Created?
MBTN | Management by the NumbersNext, run a regression analysis of the data such that the attribute ratings for each product are the independent variables and the product preference is the dependent variable.Overall Preferencein = α + β
1
Attrib1
in
+ β
2
Attrib2
in
+ …+
β
M
AttribM
in
+
ε
in
Overall Preference
= -2.7 +
1.25 *
Prestige
+
2.5 *
Ruggedness
+...
For example, the results of the regression analysis for the vehicles in this study might be:
Insight
The coefficients that are the greatest in absolute value will form the axis of your perceptual map.
Note: This
presumes
all of the survey questions are on the same 1-5 scale. If the scale
is
not the same for all measures, this would not necessarily be true.Slide8
Sample Perceptual Map (Attribute Rating Method)
8Sample Perceptual Map (Attribute Rating)
MBTN | Management by the NumbersSports Utility Vehicles with Average Ratings
Prestigious
Not Prestigious
Cadillac
Escalade (1.2, 4.7)
Chevy
Tahoe (1.9, 3.8)
Jeep Grand Cherokee
(2.5, 3.8)
Nissan
Pathfinder (3.6, 4)
Hummer
H2 (4.5, 4.6)
Toyota
4Runner (3.5, 3.3)
Ford
Explorer (2.1, 2)
Jeep
Cherokee (4, 1.7)
Not Rugged
Rugged
Insight
Now we also know that Prestigious / Not Prestigious and Rugged / Not Rugged had the highest absolute value coefficients in the regression.Slide9
Constructing an Ideal Vector
9Constructing the Ideal Vector
MBTN | Management by the NumbersCan we get anymore information from this regression?Take the ratio of the coefficient of the second-most important perceptual attribute to the most important. In this case we would have 1.25 / 2.5 = ½. Plot
the ideal
vector with slope defined by this ratio and whose beginning point is at the origin of the graph (assumes most important attribute is on the X axis
) as shown on the following slide.
Overall
Preference
in
= α + β
1
Attrib1
in
+ β
2
Attrib2
in
+ …+
β
M
AttribM
in
+
ε
in
Overall Preference
= -2.7 +
1.25 *
Prestige
+
2.5 *
Ruggedness
+…
YES!Slide10
Constructing an Ideal Vector10
Constructing the Ideal Vector
MBTN | Management by the Numbers
Prestigious
Not
Prestigious
Cadillac Escalade
Chevy Tahoe
Jeep Grand Cherokee
Nissan Pathfinder
Hummer
H2
Toyota
4Runner
Ford
Explorer
Jeep
Cherokee
Not Rugged
Rugged
The Market for Sports Utility Vehicles (circa 2005)
From less preferred
To more preferred
(slope = ½)
Insight
By drawing lines perpendicular to the ideal vector to each product we can “order” the vehicles from least preferred to most preferred.Slide11
Multi-dimensional Scaling (MDS)
11Multi-Dimensional Scaling (MDS)
MBTN | Management by the NumbersVery Different (1)(2)
(3)
(4)
Very Similar
(5)
On a scale from 1 (very different) to 5 (very similar), please compare [Product
Name
A
] with [Product
Name
B
].
One question for
each product pair
Now let’s move to the
Overall
S
imilarity
M
ethod
of
creating
perceptual
maps.
Here, rather than comparing products on particular attributes, we instead measure their overall similarity
using a
scaled method or by ranking products from most similar to least similar. While the methodologies are different, they offer similar interpretive challenges.
Insight
Notice how the question does not care
why
the responder rates the products as similar or different only the degree to which the responder perceives them to be.Slide12
Multi-dimensional Scaling
12Multi-Dimensional Scaling
MBTN | Management by the NumbersMovie Similarity Matrix
About Schmidt
Lord of Rings
Gangs of NY
Maid in Manhattan
A Guy Thing
Bowling for Columbine
About Schmidt
5.0
Lord of Rings
4.2
5.0
Gangs of NY
3.0
4.0
5.0
Maid in Manhattan
2.0
1.5
1.7
5.0
A Guy Thing
1.0
2.0
3.4
2.1
5.0
Bowling for Columbine
3.5
2.5
2.2
1.9
1.2
5.0
Here is a sample similarity matrix for a set of movies that corresponds to the average similarity ratings for a population based on the question from the previous slide.
Using this data,
we
can use a MDS program to plot these products in a two-dimensional space that best maintains their relative similarities as shown on the next slide.Slide13
Multi-dimensional Scaling
13Multi-Dimensional Scaling
MBTN | Management by the NumbersNotice that the axes are not defined on this map due to the nature of how the data is collected. Also, recognize that we are attempting to describe products that have an unknown number of perceived attributes in a two dimensional space. Does this impact our ability to use the map? Possibly. Let’s explore this further.Slide14
Overall Similarity with Perceptual Attributes
14Overall Similarity with Perceptual Attributes
MBTN | Management by the NumbersOne way to aid our interpretation of the map is to include some additional questions in the survey which can be used to enhance the
MDS analysis. In effect,
we are
adding additional “products” that are pure perceptions.
Very
Different
Very Similar
Nissan Pathfinder and Rugged
1
[ ]
2
[ ]
3
[ ]
4
[ ]
5
[ ]
Very
Different
Very Similar
Reliable and
Hummer
1
[ ]
2
[ ]
3
[ ]
4
[ ]
5
[ ]
Very
Different
Very Similar
Rugged and
Reliable
1
[ ]
2
[ ]
3
[ ]
4
[ ]
5
[ ]Slide15
Sample Perceptual Map (Attribute Rating Method)
15MDS with Non-Product Perceptions
MBTN | Management by the Numbers
Insight
Notice that the position of the brands, though similar to the attribute rating method, is different. This is due to
different methodology used in MDS which captures
overall perception of the
brands rather
than the direct measurement of
an attribute such as ruggedness.
There
are no axes
for guidance, only relative positioning to the perceptions.
Cadillac
Escalade
Chevy
Tahoe
Jeep Grand Cherokee
Nissan
Pathfinder
Hummer
H2
Toyota
4Runner
Ford
Explorer
Jeep
Cherokee
Luxurious
Reliable
Rugged
Good ValueSlide16
Multi-dimensional Scaling (MDS)
16Multi-Dimensional Scaling (MDS)
MBTN | Management by the NumbersAnother approach for creating perceptual maps using the overall similarity method is to have study participants rank each pair of products from most similar to least similar. One could also use the same data collected in the prior example to create this rank order. In addition to ranking each pair, one could also collect rank order or ratings on various perceptions (such as ruggedness, good value, etc.) and preference or sales data to aid in interpretation of the perceptual map as we’ll see in the following example.
The following
screen
shows
a perceptual map created from a survey from a particular target market segment in an automobile simulation. In this example, sales data was also collected for this segment, and brands were rated on three dimensions: Price, size, and dealer service. Finally, data was also collected about the characteristics of their ideal brand. Let’s take a look at this perceptual map and attempt to interpret the information provided. Slide17
MDS Example with Vectors and Stress
17MDS Example with Vectors and Stress
MBTN | Management by the NumbersBefore considering the vector and ideal brand information, let’s look at the relative positioning of the brands themselves. We could say that brands A and B are perceived as being fairly similar (and brands E, G and I as well). We could also say that brand F is perceived as very different from all other brands in the study. Take a moment to consider what underlying factors might be driving the positioning of the vehicles.Slide18
MDS Example with Vectors and Stress
18MDS Example with Vectors and Stress
MBTN | Management by the Numbers
Estimated preferred,
expected or ideal position for the customer is
marked by the “*”
Top 10
brands
for customer are listed in order of sales to
the target customer segment
r
2
measures
how well the position of the vector on the map reflects the ratings data collected
(1.0 means
perfectly correlated).
Stress is a measure of how well the map
captures the
brand relationships
in two dimensions. Lower is better
. Under .20 is good, .20 - .40 is acceptable, over .40 means that the map is struggling to capture the relationships in 2 dimensions.
With this additional information, what can we say about the map?Slide19
Multi-dimensional Scaling (MDS)
19Multi-Dimensional Scaling (MDS)
MBTN | Management by the NumbersFirst we can say that this map does a good job capturing the relative positioning of the brands in a two dimensional space (stress = .17). Next we can say that as we move from left to right on the map, we’re generally going from lower priced brands to higher priced brands and that it appears that this customer prefers a lower priced brand. We can also say that as we move from the bottom left to the top right brands are going from small to large (in size). Since the r2 is fairly high on these two dimensions (and the stress is low), these relationships are fairly accurate.
As we might expect, since brands A and B are closest to the ideal, their sales are the highest. We could also say that despite the relatively good positioning of brand J, something is keeping it from achieving higher sales that is not captured on the map (distribution, advertising, etc.)Slide20
Further
Reference20
Further ReferenceMBTN | Management by the NumbersDolan, Robert J. "Perceptual Mapping: A Manager's Guide." Harvard Business School Background Note 590-121, July 1990.Michael Deighan, Stuart W. James, and Thomas C. Kinnear, StratSimMarketing: The Marketing Strategy Simulation, Interpretive Software, 2011Sawtooth Technologies http://www.sawtooth.com