Now that we've reviewed the questionnaire, let’s look at how consumers' responses to items on this questionnaire were used to segment the health club market. The first example illustrates the creation of 'benefit segments.'   By examining customers' preferences for different combinations of facilities and services, we can identify market segments that are based on differences in these needs. 

Benefit Segmentation

Before we examine the mechanics of the benefit segmentation process, let's first explore its conceptual underpinnings.  Benefit segmentation probably is the most logical way to segment markets in that it is very consistent with the marketing concept. Remember that the marketing concept says, "first identify the customer’s wants and needs, and then develop marketing programs that fill these needs." This is exactly what we are doing when we create benefit segments. We are identifying different needs that people have and then are creating segments based on these needs. The idea is that different segments will contain customers with different sets of needs.  Firms then can develop finely tuned marketing programs that are highly effective at serving the needs of one or more of the segments that have been identified.

In my opinion, benefit segmentation is the best way, by far, to segment markets.  We have already looked at several examples of benefit segmentation. The toothpaste, paint, and snack food examples all illustrate breaking up larger product-markets into segments in which customers have very different needs. In the toothpaste market, as a case and point, whether you are a sociable, worrier, or a sensory reflects very closely the needs that you have with respect to toothpaste. Sociables want fresh breath and white teeth. Worries are concerned with preventing tooth decay. Sensories are interested in pretty colors and good taste. A similar rationale holds for both the paint and the snack food market examples that we discussed.

Creating Benefit Segments

The procedure for segmenting markets based on differences in needs is really not all that complicated. The process follows the basic three-step model for segmenting markets (Exhibit 1). The first step is to identify the particular variables (questions) from the questionnaire to be used in forming or defining our market segments. In other words, we must first select the appropriate "segmentation base."  The second step is to employ this base to create our market segments.  For the current example, this second step requires the use of a fairly sophisticated statistics program called cluster analysis. Once segments are formed via this statistical process, the third step is to profile the customers in the resulting segments. Fortunately, our questionnaire has all of the information that we need to perform all three of these steps. Let’s start out with the first step -- selecting the appropriate segmentation base.

Since we want to segment the health club market based on the benefits desired by customers, we are, realistically, segmenting the market based on consumers' needs. Looking back at our questionnaire, the items used to ask subjects about the importances of various facilities and services amount to assessing the benefits that consumers want from health clubs. There were a total of 23 items covering facilities and services (Exhibit 2).   We are going to use all 23 of these items to define our 'benefit segmentation base.'   

Using these 23 questions to create our 'benefit' segments requires the use of the statistical procedure called cluster analysis.  The basic idea underlying the use of cluster analysis for creating market segments is that it simultaneously examines all 23 items representing facilities and services in an attempt to find groups of consumers that express about the same pattern of "importances" on all 23 items. Consumers with the similar patterns are 'clustered' together by the program.  We then assume that each cluster represents a different, unique market segment.

To give you a feel for how this is done, lets look at two items from our list of facilities and services: the importance of weight machines and the importance of an indoor track. The X and Y axes in Exhibit 3 represent the 7-point scales that we used for these items.  The dots on this graph represent the responses of specific individuals to both questions.  For example, respondent one indicated that weight machines were very important (i.e. the respondent rated its importance at about a "1") but the indoor track was relatively unimportant (i.e. the subject rated the track at about "6").  Other 'dots' are interpreted in the same way.  For some individuals both the weights and the track are very important, for others neither are important.

If we take a look at the density of points on this graph, there appear to  be approximately three major clusters. We can identify these clusters by drawing arrows through the low-density areas on the graph i.e. where there are relatively few dots. One cluster is in the upper left-hand corner. Another cluster appears to be more in the upper right, and the final cluster is towards the bottom of the screen.  What this says is that there are probably three market segments with respect to these two facilities. One market segment (upper left quadrant) really wants weights and the indoor track is relatively unimportant.  The second segment views neither as important (upper right quadrant), and the segment toward the bottom (the largest segment) appears to want an indoor track (the average importance ratings are more toward the 'important' end of the scale). However, importance ratings for weight machines appear to be across the board for this segment,  indicating that it is important to some but not others.

This type of "visual clustering" is very easy to do when only two dimensions are present -- as in this example. We can even add in a third dimension -- a third facility or service -- and can still create visual clusters on a three-dimensional graph. Once we move beyond three dimensions, however, it becomes virtually impossible to manually identify clusters. This is where our statistical program -- cluster analysis -- comes into play. There are 23 "dimensions" in our questionnaire that reflect different facilities and/or services that customers may want. What we did visually with two dimensions, the cluster analysis program can do with all 23 at once. It will identify groups of people that express the same basic patterns in how they respond to all 23 items on the questionnaire.  The next section examines how the clustering process works.

K-Means Cluster Analysis -- The Most Popular Clustering Program for Segmenting Markets

The clustering program most commonly employed for market segmentation applications is called 'K-Means Clustering' and is found in most commercial statistical packages, such as the Statistical Package for the Social Sciences (SPSS). SPSS offers a Windows-based version of its program that it is virtually all menu-driven and, therefore, is very user friendly. Exhibit 4 highlights the opening menu in SPSS. Let’s assume we have all the data from our survey entered into SPSS.  The data are visible in the SPSS 'spreadsheet' that takes up the bulk of the exhibit. Each column in the spreadsheet represents one of the questions from the survey questionnaire.  Each row represents all of the responses from one individual (one subject) in the survey. For example, the first row contains the responses from subject one for all of the questions on the questionnaire. Row two contains the responses from subject two, and so on. All we have done in order to create this database of responses is type each subject's responses from the questionnaire into this spreadsheet. The process is exactly like entering information into any electronic spreadsheet, like Lotus or Excel.

Once the data are in the spreadsheet, we then can run the desired statistics program -- in this case, cluster analysis. This is done simply by selecting the appropriate menu item as shown in Exhibit 4. We start by selecting Statistics from the main SPSS menu. This opens up a series of additional selections, one of which is the submenu referred to as Classify. This selection, in turn, opens up another sub-menu which includes several different 'classification' programs. The particular classification program that we want is  K-Means Cluster Analysis. As mentioned earlier, this is by far the most popular clustering program for market segmentation applications.  Of course, it is beyond the scope of this course to get into the specific mechanics of how the K-Means Clustering process works. However, simply knowing that such a tool exists is valuable for planning future courses that you may wish to take as prospective marketing managers. 

The result of applying cluster analysis to our health club data is that the program identified four 'clusters' from our sample of 575 respondents to the questionnaire (Exhibit 5).  We can assume that each cluster represents a distinct market segment.   Remember, these clusters were formed based on consumers' responses to all 23 facilities and services questions. Therefore, the clusters identify four basic groups of people with somewhat different needs with respect to health club facilities and services. Cluster One consists of 156 people. Cluster Two consists of 201 people -- it is the largest cluster. Cluster Three contains 143 people, and Cluster Four has 75 people.

We have some valuable information at our fingertips as a result of this analysis. To summarize, we know that there are four clusters or market segments.  We also know that each segment has a different set of needs. But what we don’t know is who is in each segment and specifically how these market segments differ from one another with respect to needs.  Moreover, we don't know anything about how the four segments may differ on other important customer characteristics, such as demographic traits and life style preferences. In order to obtain this additional information, we must now profile people in each cluster.

Profiling the Clusters or Market Segments

The profiling process is intended to describe or characterize the consumers in each market segment identified by the clustering process. Our questionnaire provides all of the information that we need to develop the profiles. The clustering process tells us exactly which respondents were placed in each cluster.  It even wrote this information back to the data file.  As a result, all we need to do is examine the data file for traits that individuals in each cluster appear to have in common.  This is done by 'cross-tabulating' segment membership with additional items in the questionnaire. Exhibit 6 provides an example table produced by SPSS that summarizes each cluster's preferences for the range of facilities and services desired by members of each segment. The clusters or market segments emerging from the cluster analysis are listed across the top of the table.  Instead of using cluster numbers (i.e. 1, 2, 3, and 4), I cheated a little and worked ahead to see if I could come up with some descriptive names for these clusters.   Based on my preliminary overview of a number of tables (yet to be discussed), I decided to name the clusters the pseudo-body builders, social exercisers, fitness purists, and the over-the-hill gang.  Based on my review of responses by people in each of these segments to many of the questions on the survey, these labels appear to be fairly descriptive of the average individual in each segment.

The left-hand column identifies the services and facilities desired by health club patrons in our market segments -- the existence of trainers, the availability of free weights, weight machines, indoor tracks, and so on. The numbers in the columns for each segment identify the average responses (i.e. the average importances) that individuals checked for each item scale, the standard deviation of these responses, and the number of individuals in that segment. By examining the mean scale responses for each variable, we can identify any differences between the market segments. For example, with respect to free weights, the pseudo-bodybuilders' average response was 1.69 on this scale.  This simply means that, on average, members of this segment consider free weights to be very important.  Recall that the smaller the number, the more important the service or facility.  In contrast, the over-the-hill gang rated free weights as relatively unimportant -- an average of almost 6 on the 7-point scale. These individuals tend to prefer other forms of exercise. This kind of pattern exists for virtually every question on the questionnaire. It is up to the market researcher to examine the profile of responses across all of the facilities and services questions for each market segment and summarize the major differences between segments. 

Demographic characteristics are also very important for building profiles of consumers in market segments. Gender is one demographic trait that  is nearly universally applied in segmentation studies. Simply knowing whether market segments are predominately male or female can aid in designing promotional materials, or even designing the exercise programs that health clubs may offer. Based on the information in Exhibit 7,  it appears that the over-the-hill gang is exclusively female and, in contrast, the pseudo-bodybuilders are nearly all males!

Subjects' marital status of subjects is another commonly employed demographic trait for profiling segments.  Exhibit 8 indicates the existence of some marital status differences between market segments.  The pseudo-body builder segment, in addition to being nearly all male, also appears to consist mostly of individuals that have never been married. In contrast, the over-the-hill gang consists mainly of woman who are divorced or currently married. Very few of the individuals in this latter segment indicated that they were never married.

Income is also widely used for profiling market segments because it directly reflects the segment's buying power.  Exhibit 9 suggest some distinct income differences between segments. The pseudo-body builder segment, on average, possesses a much higher average income than does the over-the-hill gang segment. The social exerciser segment also possesses a lower average income, probably reflecting the fact that most individuals in this category are college students.

Fully profiling the customers in market segments requires more than just looking at their needs and demographics, Exhibit 10 examines the differences between our health club segments across a wide range of "attitudinal" traits. Understanding consumers' attitudes about health and fitness provides insight into the factors that motivate people to exercise.   This, in turn, can suggest how we need to structure health club product offerings and promote to these people.  There were a large number of attitudinal statements included on our questionnaire. Part of the profiling process is to look at responses to these items and ask ourselves what the numbers mean.  What can we infer about basic differences or similarities from one market segment to the next in terms of people’s attitudes about health and fitness? 

I think it is easy to see that there is a lot of work involved when attempting to meaningfully segment markets.  A tremendous amount of information has to be collected and processed (generally with the help of some sophisticated statistics programs) in order to accurately identify and profile market segments. Once we have gone through all of the data and determined exactly how segments differ in terms of needs and relevant consumption-related characteristics, we then summarize what we have learned in a table similar to the one presented in Exhibit 11. By summarizing information in such a table, it is easier to pick out the major difference and similarities between market segments.  This makes it that much easier to create appropriate marketing strategies targeted to one or more of the segments. This type of table is very similar to the table employed for the snack food segments discussed earlier in this topic.

Page last modified: February 06, 2002