Tuesday, January 5, 2010

Analysing market research data

At a recent Xmas party I attended, a friend of mine, who considered himself quite numerate, asked me how I went about analysing market research data. Did I use the same techniques and steps for each piece of work? And if not, how did it differ?

Now, I would not describe myself as a statistician, but over the years I have gained an awareness of the methods and techniques that may be employed to tease out insights from data. And so, after a few moments to think, I tried to give an answer in simple, clear terms. It went something like this…


There are, to my mind, essentially two basic groups of techniques used to analyse research data:


On the one hand there are structural techniques. These identify the relationship among variables, for instance, when a researcher wants to know which product variables are related to one another or how consumers group into homogeneous clusters. Factor analysis, cluster analysis, etc., belong to this class of techniques.


And then there are functional techniques, that concentrate on how a set of variables influence a variable we are interested in, for example, identifying purchase drivers, and what attributes distinguish users and non-users of a certain brand.


I went on to say that historically the general approach to data analysis is sequential: first obtain the cross-tabs, then use structural and functional techniques to sharpen our understanding of the data. This approach has been, and will continue to be, useful in analysing data.


However, at AMR Interactive (a market research agency I worked for some years ago), I came across a new approach that combined the 2 classes of techniques into one single analysis procedure. In doing so, it provides considerable insight into how brands and attributes are related to one another and which attributes (or demographics) are crucial in distinguishing brands.


In addition, this Correspondence Analysis also provides Perceptual Maps that can be used to strengthen a current brand position or find opportunities for a new or existing brand. It puts important patterns in the data into bold relief by visually depicting the prominent relationships. Richer interpretation of data and more relevant cross-tabs may be generated once we identify the important patterns.

No comments: