predictive segmentation – getting more from your research

Predictive Segmentation

Most segmentation methods usually fail to provide the full segmentation picture.  Segmentations that are based purely on needs necessary for determining process or product improvements that will promote loyalty generally show little differentiation in terms of targetable or objective business demographic characteristics. Conversely, segmentations based purely on demographics rarely differentiate customers in terms of the service needs that can be critical to retention efforts. Our clients receive actionable benefits from a comprehensive segmentation approach that identifies more meaningful analytic groups taking into consideration all the issues of segmentation. In this way, we are better able to assess each segment from the point of view of the loyalty or retention likelihood and the profitability of those customers for your organization.

To address these concerns, we have developed a segmentation methodology that overcomes the shortcomings of other techniques by segmenting respondents based on those demographics that are empirically related to needsThus, the segmentation can differentiate groups of customers based on several sets of variables. This model is called Predictive Segmentation.


In this model, an approach is used that differentiates segments based on their need for product offerings and service support (ratings of product line breadth, billing process, price, “criterion variables,” etc.). At the same time, the segments are differentiated on the customer and demographic characteristics (“predictor variables”) that are most important in differentiating each segment’s particular configuration of importance drivers. In addition, each segment is profiled in terms of value, loyalty, profitability, and any other factors important in the overall analysis.

This model is “predictive” in that it only segments the market based on those characteristics that relate to needs. Characteristics that do not relate to needs (and vice versa) play only a weak role in deriving the segments. The segmentation, then, will not only provide actionable and maximally differentiated market segments, it will also provide an understanding of what factors, among those measured, actually affect decision making.

Segment Classifier

As an outcome of the Predictive Segmentation, we develop a prediction formula that classifies each customer as to his or her segment. The formula makes use of information that is currently available in the survey file, including database variables that you provide as part of the sample frame. We validate the classification formula by testing its ability to correctly classify respondents from the study whose segment classification is already known.

We will deliver the results of the predictive model in the form of a user-friendly classifier. Our clients then simply enter a few characteristics of the customer, and have the classifier predict the customer’s segment. The classifier can be programmed to describe the segment and suggest appropriate offerings or provide other information that you wish to include. The classifier can also be used to screen and recruit respondents who belong to segments of interest and conduct follow-up qualitative or quantitative research.