2. Agenda
g
Why use Multidimensional Scaling (MDS)
and Overall Similarity (OS) Perceptual Maps
What is MDS?
Background and History
Tips for Implementation
Things to Consider
Example Application
3. Challenges with AR Perceptual
Maps
Maps
Users have difficulty scoring attributes, even
when they are aware of them
Purchase d i i
P h decisions are sometimes made using
ti d i
implicit attributes, that are not easily identified
Absence of phantom attributes as map
dimensions distort analysis
AR views products as bundles of attributes,
attributes
which need to be complete in order to be
effective
4. What is Multidimensional Scaling?
g
Goal – to determine GAPs in the market based on
consumer / purchaser’s perception.
Data
D t analysis methodology f mapping
l i th d l for i
similarities and dissimilarities among items
Used in Market Research to map consumer
perceptions of products to draw out attributes
not otherwise easily communicated
Analysis often graphically represented in order
to identify gaps in the market
5. Background and History
g y
1938 - Pioneered by Young & Householder
y g
1958 – Psychometrician Torgerson revived with the
earliest application in market research
C
Consumer perceptions of silverware patterns
ti f il tt
1969 – Stefflre was the first to use systematically, focused
on visual representation of consumer’s perceptions of
consumer s
brand similarities –developed “Tinkertoys”
1975 – Green discussed the issues surrounding new
product development
1983 – present : Numerous developments in methodology
and applications in market research
7. Steps for Implementation
p p
What brands and how many? y
Formulate the Problem 8<x<20 ideal
What is the purpose of the
analysis?
l i ?
Obtain Input Data Perception data : direct
approach
Run MDS Q = N (N - 1) / 2
Statistical Program Where # of questions depends on #
of brands
Metric (interval) or Nonmetric
Map the results and (ordinal) MDS programs
define dimensions 2 D Map, often subjective
8. Example Application –
OS Perceptual Map of Search Engines
Overall Si il it P
O ll Similarity Perceptual M of Search
t l Map f S h
Engines using Multidimensional Scaling
technique
14 Subjects
5B Brands/Products
d /P d t
Yahoo, Google, Copernic, AOL, MSN.com
2 Di
Dimensions
i
Usability and Overall Value
5 point Ordinal Measurement Scale
9. Tools used
Data Gathering
Online Survey Tool : QuestionPro
www.questionpro.com
Data Analysis and Visual Representation
Excel Statistics Plugin : XLStat
g
www.xlstat.com
10. MDS Proximity Map
y p
Configuration (Kruskal's stress (1) = 3.952E-5)
1
0.8
Go o gle
0.6
Co pernicus
0.4
0.2
02
Dim2
0
-1.2 -1 -0.8 -0.6 -0.4 -0.2 0 M SN.co m
0.2 0.4 0.6 0.8 1 1.2
-0.2
-0.4
Yaho o
-0.6 A OL
-0.8
Dim 1
11. OS Perceptual Map
p p
Go o gle
Gap 1
p
Co pernicus
Usability
M SN.co m
Value
Yaho o
A OL
12. Things to consider [3]
“Comparison of MDS Methods for Perceptual Mapping”,
7f t
factors varied over 4 diff
i d different MDS algorithms
t l ith
Large samples are best –
Analysis of <10 subjects should be interpreted cautiously
The same holds true for # of brands, to lesser extent
Greater than 6, less than 12 ideal
Dissimilarity judgements should be collected on
interval scales or on ordinal scales with large
number of scale values
Most frequent criticism of MDS is that assumptions
must be made on the error components
components.
13. References
1) Crawford,
Crawford Merle; DiBenedetto Anthony: New
DiBenedetto,
Products Management, eight edition
2)
) Carroll, J. Douglas; Green, Paul R.,
, g ; , ,
“Psychometric Methods in Marketing Research:
Part II, Multidimensional Scaling”, Journal of
Marketing Research, (May 1997), pp. 193-204
193 204
3) Bijmolt, Tammo H.A.; Wedel, Michel: “A
Comparison of Multidimensional Scaling
Methods for Perceptual Mapping”, Journal of
Mapping
Marketing Research, (May 1999), pp. 277-283
4) “How do I run a Multidimensional Scaling (
g (MDS)
)
with XLSTAT?”, http://www.xlstat.com/tutorials