Analytic Services

In marketing research, cluster analysis is often used to determine whether there are distinct groups of customers with different needs, preferences, perceptions, product usage, and purchasing behavior.  By understanding how customers differ, management can develop products and/or marketing strategies that are tailored to each group’s individual needs.

K-means analysis is a clustering technique that can be used to create customer segments.  This type of cluster analysis uses a procedure in which individuals are assigned and reassigned to “clusters” or “segments” repeatedly until each individual is assigned to a final segment. Each final segment is comprised of individuals who are more similar to other people within that segment than those in other segments.  This method implicitly minimizes the diversity within each segment — and thus, in this case, produces distinct segments with homogenous needs, preferences, etc.

It should be mentioned that other techniques such as regression and discriminant analysis can also be used to predict a dichotomous dependent variable. However, these techniques depend on certain assumptions (i.e., multivariate normality of the independent variables and equal variance-covariance matrices for the two groups). When these assumptions are violated (which they typically are in studies of this type), the results of the analysis may be less than accurate. Logistic regression does not depend on these assumptions.

By examining the association among variables, factor analytic techniques produce a smaller set of variables or factors that represent the underlying dimensions of the original set of variables.  Each factor is not a single, directly measurable entity, but rather a construct that is derived from the measurement of the original set of variables.  This technique is often used for the purpose of data reduction —  that is, reducing a large number of variables to a smaller set of factors greatly simplifies the description and understanding of large sets of data.  

Consider an example of a data set that includes preference ratings for a large number of specific foods such as bread, apples, blueberry pie, cereal, spare ribs, green beans, watermelon, chocolate mousse, pasta, ham, asparagus, artichoke, and roast beef.  A factor analysis of this data set would distill this large amount of information by identifying the dimensions that underlay the individual ratings.  

More specifically, dimensions are identified for groups of variables that are highly intercorrelated with each other, but are not highly correlated with variables outside of that group.  In this data set, for example, we might expect a fairly high association among preferences for cereal, pasta and bread — that is, in general, the higher an individual’s preference for bread is, the higher we might expect his preference would be for pasta and cereal.  However, knowing an individual’s preference rating for cereal, pasta, and bread would probably not be useful in predicting that individual’s preference rating for any other food on the list.  If this were the case, these three variables would be represented by a factor that might be called the “starch” factor.  Cereal, bread, and pasta would then be said to “load” on the “starch” factor. Besides the “starch” factor, other factors likely to emerge in this example include “fresh fruits,” “deserts,” “green vegetables,” and “beef.”  

When a small number of factors account for most of the variance in the original set of data, we can “explain” the original set of variables in terms of a smaller set of factors without losing important information.  Simplifying large sets of market research data in this way provides a conceptual clarity that facilitates one’s comprehension of the data and its strategic marketing implications.

For example, it might be useful to identify those variables that determine whether or not, for a given product, an individual would be most likely to choose to deal with a (1) bank, (2) finance company, or (3) credit union. Variables to be used in such an analysis would depend on your particular interests, but might include demographics, past loan/credit experience, likely reasons for loans, and needs/benefits.

If desired, discriminant analysis can also be used to develop a procedure for predicting group membership (bank vs. finance company vs. credit union) for individuals not included in the analysis. This type of procedure might be useful in selecting potential customers for a targeted marketing campaign.