Segmentation success

Editor’s note: Michael Sosnowski is an account executive with The Response Center, a Fort Washington, Pa., research company.

Segmentation research can provide your organization with a unique and competitively advantageous perspective on the marketplace. In practice, however, such studies can involve significant time and effort, and what constitutes a winning solution is not always clear. The benefits can be great, but the risks can be too.

So how do you avoid the common pitfalls of segmentation research and produce a useful, lasting outcome for your clients? No one approach is foolproof, but you can greatly enhance your odds of success by understanding and implementing a process-based approach to segmentation - one that begins with client needs and ends with an actionable, objectives-focused set of segments.

This article is meant to explain the basic building blocks of the segmentation process, and to provide you with a roadmap for implementation that will be of help regardless of industry or subject matter. These fundamental steps, more than any one technique or methodology, will determine the success of your research efforts and ultimately will influence management’s perspective on the value of market research.

From the start, think about the finish

Successful segmentation research is heavily dependent upon buy-in from your clients, so it makes sense to seek their input right from the start. Work to clarify the objectives they have for this research, and take the time to understand the resources at their disposal. Brainstorm with them before writing a questionnaire or defining variables for use in your analysis, in an effort to answer two important questions.

  • What is the ideal way to divide consumers for marketing purposes? Demographic groupings, attitudinal questions, and behavioral data all have advantages and disadvantages depending on what your clients see as “ideal” segments.
  • What information is needed to best enable marketing action? Specifically, what data are needed to most help your clients target prospects, assemble offers, and communicate benefits to consumers within the various segments that are developed.

There is no guarantee, of course, that the final analysis of results will perfectly reproduce this view of the marketplace. Visualizing the end game, however, allows you and your clients to create a set of guiding principles that will inform the remainder of the project, and this in turn will ensure results that are as actionable as possible.

Focus the questionnaire design process

Segmentation research can be performed solely using variables available within a company’s customer database. Frequently, though, client objectives require the collection of information not already in hand, and in these cases a questionnaire is needed.

It is likely to be lengthy, since for practical reasons segmentation studies commonly perform double duty as comprehensive market profiling exercises. The important thing to remember is that your ability to identify actionable groups of consumers will be closely tied to the nature of variables available, not just the number used. Ideally every question should somehow serve the goals of your segmentation effort. As the person guiding the research, ask yourself repeatedly: would this be a useful way to think about how consumers may differ in the marketplace? And, will this information help us market to different customer segments? If the answer to both of these is no, then that variable probably should not be included in the final questionnaire.

Also remember that the structure of questions can have great impact on your final solution. For example, asking consumers “How likely are you to purchase this service?” will produce different results than asking “Are you likely to purchase this service?” In the end, the principles that define the segmentation study should be referenced to decide upon wording that is most appropriate.

Carefully select variables for the analysis

It is not wise to include all possible survey and database variables in the segmentation analysis, regardless of technique used. This “kitchen sink” approach will only muddy the waters of analysis, and needlessly complicate your primary task: finding segments that are meaningful, reachable and strategically advantageous.

At the same time, excluding a potentially important variable can also hinder the discovery of useful segments. It is therefore important to have a structured process for choosing which variables will be used to define segments and which will serve to profile segments. The following rules of thumb may be of help.

  • In general, it can be counterproductive to mix demographic and attitudinal variables in the same segmentation.
  • You should take steps to limit the number of variables included in the analysis. Commonly a handful of critical questions will form the backbone of your most useful solution.
  • It is wise to consider questions where variation exists in the data. On their face, these highlight points of differentiation in the marketplace.
  • Try and avoid including two or more questions that appear to serve the same analytic purpose.

Test the utility of the segmentation

Segmentation analysis will yield one or more possible solutions to evaluate and you will be called upon to decide which is most useful. Surprise! There is no simple formula for determining the ideal segmentation scheme. The key, rather, is to systematically examine the profile of each segment and build a business case in support of the ideal solution. In doing so, consider the following important questions.

  • Can members of attractive segments be easily identified in the marketplace, or are they based only on information that is easily collected via survey research but not readily available otherwise?
  • Are one or more segments particularly attractive and/or lucrative? Conversely, are there one or two consumer groups that can be eliminated as attractive marketing prospects, or at least relegated to a low priority tier of attention?
  • Are these segments based on variables readily available in the company database, so that we can (if desired) assign current customers to one or the other grouping?

As a rule of thumb, useful segments tend to incorporate 10 percent or more of the population. It is also possible, however, to find lucrative segments that are very small. In the end, your winning solution should focus attention on groupings that are easily reached, effective at identifying the most and least attractive targets, and based on information that can be used in the future to group consumers based on segment membership.

Apply your solution to the marketplace

If the overarching goal of segmentation research is to sort consumers into distinct and meaningful “types” for marketing purposes, then it follows that such studies should provide you with the means of finding these types moving forward. Classification models can be created that use a short list of variables to effectively determine segment membership. Armed with this tool, you can bridge the gap between marketing research and marketing action and - in the process - strengthen the argument for conducting such projects in the future.

You may use a classification model to place new customers into segments at sign-up, to sharpen customer acquisition efforts, or to track segment membership over time. Or, you may identify different applications for this model based on your clients’ needs. Regardless, developing a classification model will help promote the segmentation study within your organization, provide staying power for the hard-found results, and help demonstrate the practical benefits of market research to others in your organization.

Educate yourself to the various techniques available for finding segments

A number of segmentation techniques have developed over time, each with a unique way of finding segments. None of them can be thought of as the “best” approach, and you may find it useful to compare and contrast solutions from multiple methods. It is therefore helpful to have a broad understanding of the options available. These are described briefly below.

  • Clustering techniques array respondents in multi-dimensional space, and then group them based on proximity. This family of techniques is clearly the most traditional form of segmentation analysis, and can be broadly divided into two types: hierarchical clustering (which provides a range of segment solutions from which to choose) and non-hierarchical or k-means clustering (where you specify the number of segments to be created).
  • Latent class modeling (LCM) is the only technique that offers a strong statistical basis for segmenting data because, unlike other methods, it is model-based. Therefore its fit can be evaluated like other statistical methods such as regression. In simple terms, LCM works from the assumption that data sets, as a whole, consist of different distributions mixed together. It then proceeds to unmix them, in the process uncovering unique consumer segments.
  • Self-organizing maps (SOMs) represent a neural network-based technique. In it, respondents are introduced one at a time to an initially blank map. As each record is presented to the map, the dominant patterns in the data are learned and segments are created.
  • Tree-based algorithms such as CHAID are particularly useful when a target or dependent variable is available in the data. If, for example, the acceptance of a direct mail offer represents a desired outcome, a segmentation scheme can be derived that first identifies what distinguishes acceptors from non-acceptors. The data set can then be further split into segments within these umbrella categories - a process that can continue until an optimal number of segments are obtained.

There is much more detail available in the literature on each of these techniques. It is important to remember, however, that the success or failure of your segmentation effort will hinge primarily on the processes you have in place, not the technique selected. It is smart to know your analytic options, but even the most advanced technique will not prove effective without careful planning, questionnaire design, and variable selection.

Lead to success

Segmentation research involves many steps, and a host of choices along the way. Managing this process is never easy, but there are ways to ensure that your hard work will lead to success.

Involve your clients from the beginning, and clearly understand what they hope to gain from the research. Work with them to understand critical data needs, and craft an efficient questionnaire where all information gathered directly serves the purposes of asegmentation. Use profiling as a tool to find attractive targets in the marketplace, and to enable marketing action in a way that is consistent with the resources available to your organization. Give users a tool for classifying consumers into segments in a way that demonstrates the business impact of research. All the while remember that managing a successful segmentation project means providing your clients with a solution they can understand and act upon.