Data Reduction

Cluster analysis is used to identify segments; that is, groups of respondents with similar patterns of responses across a set of variables. Several alternative cluster solutions are examined. Our goal is to find segments that are both identifiable and actionable.

Cluster analysis is one of the most subtly complex multivariate procedures. Much of the complexity results from selecting the subset of variables you use to form your clusters and how you transform those variables prior to running the clustering algorithm.

Another data reduction technique that is very useful in segmentation studies is perceptual mapping. Perceptual maps show a two-dimensional picture of the relationship between segment membership and descriptive variables (i.e. attitudinal and behavioral measures). The perceptual map “reduces" the data to the key underlying dimensions. We might find that one segment separates from the others in a way not obvious from the cross-tabulation data. We might also find that certain descriptive variables are related, thus improving our understanding of the segments and their characteristics.

Data Mining

CART analysis (Classification and Regression Trees) has many applications, but one of the best is to find the key variables and variable interactions that are associated with segment membership. CART is a multivariate procedure that searches for complex variable interactions in a dataset. We would first look for differences among the segments by analyzing the cross-tabulations. This shows the relationship of each individual variable with the segments. Once we’ve taken this step, the goal is to uncover data interactions that really “tell the story” on what separates the segments. That’s where the CART procedure comes into play.

Linking Attitudes and Behavior

One of the key challenges in segmentation studies is developing attitudinal segments that are actionable.
How is this issue addressed?

  • Questionnaire design
  • Data analysis
    • Analytical methods linking attitudes and behavior

Linking attitudes and behavior results in actionable segments.