The Geeks Guide to Reverse Segmentation – Let’s Get Technical

How segmentation is usually done

Market segmentation research usually depends upon cluster analysis, a statistical technique for combining cases based on similarity on specified dimensions.  Typically, segmentation research will attempt to construct segments by clustering cases based on similarities in demographic/ firmographic attributes, or else cluster cases based on similarities in attitudes and behaviors.  Each of these approaches has inherent limitations that can severely curtail the usability of the segmentation analyses.

The problems with demographic/firmographic clustering

Regardless of the clustering technique that you use (e.g., Ward’s method, K-means, etc.), there are limitations inherent in demographic clusters that reduce your ability to craft a message based on similarities in attitudes and behaviors.  If you cluster on demographic attributes, the analysis will combine cases that are demographically similar, but will completely ignore information about the case’s attitudes and behaviors.

Consider a situation in which you are attempting to sell a product to employees of a large corporation, and that you can cluster employees based on multiple demographic attributes, such as regional location, years of employment, and division.  You have reason to think that each of these characteristics has some relationship with buyers’ attitudes and behavior, and you want to identify clusters that have distinctive behaviors and attitudes so you can deliver a well-targeted message.  Now suppose that the majority of the employees in the Pacific Coast region are in the research and development division, and that these employees have higher levels of seniority relative to other divisions and regions.  Based on this association, demographic clustering will create a segment of employees with these characteristics.  However, this segment may or may not have distinctive attitudes or behaviors relative to other demographic clusters.   Even if they do, a different combination of demographic attributes might have more distinctive pattern of attitudes and behaviors.  The nature of demographic clustering is to build segments that summarize common combinations of demographic characteristics (which allows targeting), not to construct a profile of people that need to hear a distinctive message that matches their unique profile of attitudes and behaviors.

The problems with attitudinal/behavioral clustering

Similarly, regardless of the clustering technique that is used, there are limitations inherent in building clusters based on attitudes or behaviors that reduce your ability to reach groups based on demographic targeting.  If you cluster based on attitudes and behaviors, the analysis will combine cases that are similar in psychological or behavioral characteristics, and which can be clearly distinguished from other clusters on these characteristics.  Usually, attitudinal/behavioral clustering provides the most precise segments possible.  However, this clustering process completely ignores information about the degree to which cases in the same cluster share similar demographic attributes.  The results of these analyses may yield clusters that are clearly in need of distinctive messages, but which are very hard to locate and reach based on their demographic characteristics.

Reverse segmentation solves these problems

Reverse segmentation pursues a strategy that helps the market researcher to identify market segments with highly differentiated attitudes and behaviors, while at the same time considering the demographic/firmographics, media usage, or channel usage information that is needed to reach people and deliver a targeted message.

Reverse segmentation starts by grouping cases into every possible combination of demographic characteristics. Each of these demographic units forms the basis for further analysis.  Attitudinal or behavioral measures can be examined across each demographic variable to gains clues as to which measures will help best form clusters.  Next, we find the average score of each of these demographic units on the measures of attitudes and behaviors. The next stage of reverse segmentation uses clustering techniques to combine demographic units based on similarity in attitudes and behaviors.  The resulting segments have distinctive psychological and behavioral profiles, which is necessary for constructing a targeted message, while at the same time having clear-cut demographic attributes, which is necessary for delivering the message.

To provide an illustration of how reverse segmentation works, let us return to our earlier example in which we had demographic information about employees’ region of employment, division, and seniority.  We would classify employees into every possible combination of these characteristics to form the basic units of our analysis.  For the sake of example, let us suppose that there are four regions, three divisions (research and development, sales, support), and three levels of seniority (high, medium, and low).   This would yield thirty-six combinations of demographic attributes.  For each of these combinations, we compute the average score on the attitudinal and behavioral variables (e.g., the average scores for the senior employees in sales in the Northeast, etc.).  We then combine demographic units into segments based on similarity in their behavioral and attitudinal profile.  In this way, the combination of demographic groups is informed and shaped by psychological information, and not merely by examining the statistical occurrence of demographic characteristics in isolation.  At the conclusion of the reverse segmentation analysis, when units have been combined into clusters with distinctive psychological or behavioral profiles, you can see which demographic units were included in each cluster, and thereby see how to reach the segments that have distinctive psychological profiles.

Maximizing the utility of reverse segmentation

To maximize action that can be taken from the results of reverse segmentation, it is common to focus on demographics/ firmographics, media usage, or channel usage variables that are currently in your database.  This approach allows you to flag people in your database as matching a particular behavioral or attitudinal type, and there is NO misclassification, because the demographic/ firmographic, etc. segmentation was built on the “units” derived from your data.  If database variables are not chosen, the targetable attributes should be those thought to be actionable by your marketing team.

As with any segmentation output, the results from reverse segmentation should be evaluated as follows:

  • Are the segments meaningful? Do they make sense?
  • Are the segments large enough to justify targeting them?
  • Are the segments reachable (e.g., via ads or direct sales)?
  • Are the segments uniquely responsive to marketing efforts? (This characteristic is evaluated over time).
  • Is the overall segmentation plan actionable?
  • Note that another quality sought in segments is that they are identifiable – reverse segmentation segments are always identifiable, given the way they are formed


Reverse segmentation is a promising approach to solve the long-standing problem of segmentation projects producing interesting, but hard to act on, results.  When segments are not only different in terms of attitudes and behaviors, but also identifiable in the market, unique groups can be reached with appropriate messages.

The output from reverse segmentation looks similar to a traditional segmentation approach; there are simply different statistics going on in the background.  For more information about how this approach could help your company, contact us at