Multi Dimension Scaling a Dimension Reduction Technique for Brand Positioning

Sangamesh K S

November 9, 2017

Introduction

In my previous article I discussed about 2 popular scaling i.e PCA and Factor Analysis, its applicability in Marketing Domain and how they are different from each other.

In this article, I will be discussing about Multi-dimensional scaling which is a dimension reduction technique used to select the factors which influence the models the most in a metric data.

Wherein in PCA the data was transformed in such a way that the correlation is “0” (Orthogonal) and based on variance we have modeled the data in such a way that we can find each features direction/objective and how they are different from one other. I also gave an overview how this data can be used to understand best performers, brand rating, area of improvement etc..

To View PCA Visit:http://experimentswithdatascience.blogspot.in/2017/11/brand-rating-analysis-using-pca.html

In Factor Analysis, out of CFA and EFA. we looked in EFA and understood latent variables and manifest variables. How a latent variable stand as a proxy for manifest variable. Which can be applied on to marketing domains like customer satisfaction, product popularity etc with a latent variable which is kept as proxy i.e customers rebuying the product is a latent for customer loyalty(which is a manifest variable).

To view Factor Analysis Visit: http://experimentswithdatascience.blogspot.in/2017/11/factor-analysis-dimension-reduction.html

Similarly we will apply muti-dimension scaling on the same data. But the technique is different compared to PCA and Exploratory Factor Analysis. MDS can be applied for brand positioning, price analysis etc

Now here in this article we will cover 2 types of MDS

1)MDS on Metric Data(Numerical)

2)MDS on Non-MEtric Data(catagorical)

Now we will look at the sample data to perform out Metric data

##   perform leader latest fun serious bargain value trendy rebuy brand
## 1       2      4      8   8       2       9     7      4     6     a
## 2       1      1      4   7       1       1     1      2     2     a
## 3       2      3      5   9       2       9     5      1     6     a
## 4       1      6     10   8       3       4     5      2     1     a
## 5       1      1      5   8       1       9     9      1     1     a
## 6       2      8      9   5       3       8     7      1     2     a

Now I would like to transform the data and take out mean of all the brands. After transformation the data looks something like this

##   brand     perform     leader     latest        fun     serious
## 1     a -0.88591874 -0.5279035  0.4109732  0.6566458 -0.91894067
## 2     b  0.93087022  1.0707584  0.7261069 -0.9722147  1.18314061
## 3     c  0.64992347  1.1627677 -0.1023372 -0.8446753  1.22273461
## 4     d -0.67989112 -0.5930767  0.3524948  0.1865719 -0.69217505
## 5     e -0.56439079  0.1928362  0.4564564  0.2958914  0.04211361
## 6     f -0.05868665  0.2695106 -1.2621589 -0.2179102  0.58923066
##       bargain      value      trendy       rebuy
## 1  0.21409609  0.1846926 -0.52514473 -0.59616642
## 2  0.04161938  0.1513396  0.74030819  0.23697320
## 3 -0.60704302 -0.4406775  0.02552787 -0.13243776
## 4 -0.88075605 -0.9326353  0.73666135 -0.49398892
## 5  0.55155051  0.4181641  0.13857986  0.03654811
## 6  0.87400696  1.0226886 -0.81324496  1.35699580

Multidimensional scaling (MDS), a popular traditional technique for marketing researchers, graphically maps how people view and differentiate brands. Similar brands are represented as points close in space, and dissimilar brands are placed further apart. Associative networks represent consumer knowledge as links of associations among “nodes,” or units of information such as brands, attributes, advertisements, etc.

Now in this pot you can see various brands which are plotted are grouped and based on this plot we can take decision based on brand positioning, assisting where we stand etc..