Learning Multi-Dimensional Scaling (MDS)

Today, i.e. on 30^th August, 2011 we had our lecture number 7, 8 &9 on business analytics. In this we learnt, understood, applied & discussed the technique of Multi-Dimensional Scaling (MDS). We learnt the design & interpretation of the map & making a strategy based on that using relevant data, first-level analysis & above all innovative ideas & solutions.

Also known as PERCEPTUAL MAPPING, it is a popular method to analyse & understand consumer perceptions of products, brands, etc. It produces a graphic representation of consumer’s mind, which is visually appealing & readily discernible. Often it is more valuable than heaps of data or complex tables, full of numbers. Hence, it is no surprise it is commonly used across various fields, ranging from business to academics & henceforth.

In marketing, it is commonly used for product or brand positioning. For eg- in a two dimensional X-Y plane, where X axis measures price & Y axis measures quality, we can map consumer perceptions of say, cars. Thus, Mercedes will fall in the quadrant of high-price & high quality, whereas a brand like Maruti maybe perceptually low price & medium quality. Hence, it helps to identify gaps in positioning & helps businesses & managers to design suitable strategies for the existing product or brand or completely reposition it.

It can be based on two methods-

→ Attribute based

→ Overall similarity/ dissimilarity

The tools required for perceptual mapping, which we also used in the class are- SPSS, Permap software, MS Excel & Notepad. In class we solved few problems on the perceptual mapping & made some suggestions. For eg- we studied the perceptions of 6 different beer brands like Budweiser, Heineken, Fosters, Carslberg, Kingfisher, etc. First, we have pairs of the brands & then rated them together as a pair & found out the mean. Then, using SPSS functions & commands, we generated a proximity matrix showing the relative distance amongst brands. Eg- Budweiser-Heineken pair had a distance of 0.856 while, Kingfisher-Carlsberg had a distance of 0.343 in the matrix. Thus, this showed a stronger association between Budweiser- Heineken as opposed to Kingfisher-Carlsberg pair. This proximity matrix data was then transferred to the notepad, from where it was extracted on the permap software. Here, we did the actual mapping & formed clusters on the basis of overall similarity. We tried to minimize the error value as much as possible, to ensure reliability of data. Thus, by adjusting the data to suit the requirements & objectives of the study, we produce a valid perceptual map, which can then be processed, interpreted & used for strategy formulation or providing solutions.

We also did attribute based scaling amongst cellphone users studying relationship between users of sms, alarm, camera, etc. Our group also did an analysis on retail store satisfaction & age groups. The key learning here was, closer the variable group is to the arrow-head, the stronger the relationship. Thus, in our case, the students (18-24) were the farthest from 4 out the 6 variables & hence, using perceptual mapping & first level analysis, we were able to identify the gaps & problems & suggest solutions.

Overall, it was a rich learning experience in the class today & it was good to learn such a useful business technique, (through different data sets & examples) which has diverse applications & is being used by companies on a wide scale.

- Author: Abhinav Jain, 13121, Marketing Group-1