Day 5 - Team F - Unmesh Kulkarni

The lecture started with the introduction of PERMAP (Perceptual mapping). Perceptual mapping is also called as MDS (Multi-Dimensional scaling). In PERMAPs, you can map anything, not necessary that you can map only perceptions.

There are two methods of mapping in PERMAP:

1) Overall similarity

2) Attribute based

In Overall similarity method, for each pair of objects we get the similarity between pairs. Here, attributes are not given/known. You have to discover the attributes looking at the diagram. So, the method is best used when you have thorough product knowledge so that you discover the attributes or the reasons why the particular pair of objects is similar.

Advantage: It provides lot of flexibility. It may reveal what people may not directly tell; throwing up latent/hidden attributes

Disadvantage: It cannot be used when we do not have thorough product knowledge.

In attribute based method, you provide the attributes to the respondents and ask them to rate the attributes. The rating can be done by using a 5 –point rating scale etc.

Advantage: It is simple to apply and respond

Disadvantage: It may not bring up the hidden/latent attributes as in the overall similarity method.\

Example of application of PERMAP:

Suppose you have to test the similarity between 6 brands of cold-drinks (i.e. 6C2 =15 pairs of cold-drinks since we take the unique pairs because Euclidean distance between two brands is same irrespective of the order considered), we note down all the 15 pairs of the cold-drinks and ask respondents to rate them on scale 0-9. (0 for least similar brands and 9 for most similar brands).

Then we take an average rating/10 for these brands (because score of perfect 10 is reserved when you compare a brand with itself).

Then we make a 6*6 matrix (as there are 6 brands) noting down their average rating/10 score and giving a score of 1 when a brand is compared with itself. This type of matrix is called ‘Similarity matrix’.

When you give a score of 0 when an object is compared with itself, it is called as ‘dissimilarity matrix’. Then you copy the whole matrix or the bottom half of the diagonal matrix in a notepad. In a notepad, following should be written above the matrix:

Title (here, Perceptions of soft-drinks)

nobjects=.. (here, nobjects=6)

similaritylist  (as it is a similarity matrix. If the matrix is dissimilar, we write “dissimilaritylist”).

Then we load this notepad file in the PERMAP software and click on start and and again click anywhere in the map. Proximity of the objects now indicates the similarity or dissimilarity between them.

For getting the lowest error (Objective function value), we check the ‘auto repeat’ and ‘auto stop’ buttons, click on start and when the lower objective function value stabilizes we click the ‘stop’ button and uncheck the ‘auto repeat’ and ‘auto stop’ buttons.

We can remove any object and store it in ‘parked objects’ and observe what happens to the similarity/dissimilarity between the existing ones. This feature has the application in dynamically changing markets where you do not have to survey the markets again and again even when the products go out of the markets.

We can also view the map in co-ordinate system by going to View, and then clicking on ‘show co-ordinate system’. This helps us to view objects category wise, if there is any.

We can also see objects in different dimensions by clicking on ‘Dimensions’ box on left side of the screen and again pressing the start button. If significant difference in error exists for example in two or three dimensions, then the map is concluded to be viewed in 3 –dimensions.

We can also use the featured like, ‘zoom and rotate’ etc.  by right clicking on screen. The features are easy to use and quite self-explanatory.

- BY Unmesh Kulkarni 

Team F