Segmentation Analysis -- a Tool For Measuring Life Styles

ABSTRACT - In this paper psychographic segmentation, as a tool for measuring life styles, is considered. A specific segmentation study done at Needham, Harper and Steers is cited; and the assets and liabilities of psychographic segmentation in general, and in relation to "profiling" is discussed.


Sunil Mehrotra (1976) ,"Segmentation Analysis -- a Tool For Measuring Life Styles", in NA - Advances in Consumer Research Volume 03, eds. Beverlee B. Anderson, Cincinnati, OH : Association for Consumer Research, Pages: 504-505.

Advances in Consumer Research Volume 3, 1976      Pages 504-505


Sunil Mehrotra, Needham, Harper and Steers Advertising, Inc.


In this paper psychographic segmentation, as a tool for measuring life styles, is considered. A specific segmentation study done at Needham, Harper and Steers is cited; and the assets and liabilities of psychographic segmentation in general, and in relation to "profiling" is discussed.

The necessity for segmentation arises because of the limited capacity of human comprehension. Segmentation for humans serves the same function as organizing data into bytes does for a computer. It facilitates storage and retrieval of large amounts of data. Cognitive psychologists call this chunking.

We, in our daily lives, categorize information naturally. We classify people by sex, color, nationality and what have you; we think of cars as luxury, compact and subcompacts. A recent letter to Ann Landers illustrates an imaginative use of segmentation. A high school girl wrote saying that the boys in her class had rated all the girls from 1 to 5 on various criteria -- which we shall let you imagine. Her complaint was she had been rated 5 and she felt it was unjustified. (An example of misclassification, a problem I shall touch upon a little later).

This preoccupation with classifying women is best explained by James Thurber in his book "Is Sex Necessary" published in 1929. Here is what he had to say and I quote" .... successfully to deal with a woman a man must know what type she is. There have been several methods of classification none of which I hold thoroughly satisfactory, neither the glandural categories -- the gonoids, thyroids, etc. -- nor the astrological -- Saigttarious, Virgo, Leo and so on. One must be pretty expert to tell a good gonoid when he sees one.

.... of much greater importance is a classification of females by actions. It comes out finally, the nature of women in what she does - her little bag of tricks as one might say."

Thurber, in his wisdom classified women into five groups. He categorized them as (1) clinging vine type, (2) "don't dear" type, (3) outdoors type, (4) button hole twister type and (5) the quiet type.

We at NH&S looked at her little bag of tricks and came up with a battery of Activity, Interest and Opinion measures. This formed the questionnaire that was sent to a national sample of 2,000 women. We used the responses to this questionnaire to segment the female data.

Here is what we did -- for each pair of women we computed a correlation coefficient indicating the similarity of their responses to the 250 AI&O questions. We applied principal components analysis with varimax rotation to this person by person correlation matrix. The factors obtained are regarded as hypothetical "woman types"; the factor loadings represent the correlation of each woman with each type. The women are then grouped by seeing which "type" each woman is most highly correlated with.

We ended up by classifying all the women into five groups. Each group was then cross-tabulated with the entire questionnaire and five profiles obtained. The five groups, on the basis of their central tendency, were identified as - 1) Cynthia, the chic city dweller; 2) Ursula, the urbane urbanite; 3) Mildred, the militant mother; 4) Cathy, the contented country girl and 5) Thelma, the old-fashioned traditionalist.

As an example of the descriptive value of segmentation by psychographics a summary of the life style profile of Ursula the Urbane Urbanite appears in Tables 1 and 2.






I shall now focus on some of the problems encountered in segmenting data in this manner. I may add these problems are inherent in most clustering algorithms.

The major problem is one of reproducibility; that is, do the segments really exist or are they merely artifacts of the method of analysis?

We have handled this problem by doing a split half analysis. The data were randomly split into two halves, and the clustering algorithm run separately on the two halves. When the solutions obtained for the two halves were compared, it was found that solutions with more than five groups did not reproduce.

However, this still did not guarantee reproducibility with a totally different set of AI&O measures or a totally different sample.

Of course, we had the reassuring knowledge that our five group solution was replicated by the high school kids and James Thurber.

A related problem is the problem of groups sizes. Are the sizes unique or would a different clustering algorithm or different data set produce totally different numbers per group?

To get a handle on this problem we looked at the Chi-square statistics for the distribution of group sizes in the two halves. We found the group sizes in two halves were not significantly dissimilar.

Lastly, we have the problem of interpretation. Are the group central tendencies representative of the group? Are the groups homogenous enough to infer that most people in the group look like the group average? This is unlikely. Given attitudinal data, which are fuzzy, there are likely to be more people at the fringes of each group than around the average.

Despite its problems, we feel segmentation is a worthwhile way of looking at data.

It reduces considerably the dimensionality of the data space. This is important both for the researcher and the user of research, for it helps bring out the information in the data in all its richness without inflicting information overload.

Another advantage of segmentation is that it is objective. We let the data speak for itself. Therefore, objective instruments can be designed to validate the findings of segmentation.

Once the segments have been identified it provides a means for keeping track of consumer markets through trend analysis. One can determine which segments are growing and which shrinking. This would help in repositioning existing products and in identifying new opportunities.

Finally, segmentation gets around the problem often encountered in "profiling"; that is, several disparate groups get mishmashed into one profile and the profile emerges looking something like a film that has had multiple exposures.



Sunil Mehrotra, Needham, Harper and Steers Advertising, Inc.


NA - Advances in Consumer Research Volume 03 | 1976

Share Proceeding

Featured papers

See More


Q4. The notion of self-optimization in context of self-tracking and beyond

Agnieszka Krzeminska, Leuphana University Lüneburg
Dorthe Brogård Kristensen, University of Southern Denmark, Denmark

Read More


Two-By-Two: Categorical Thinking About Continuous Bivariate Data

Bart de Langhe, ESADE Business School, Spain
Philip M. Fernbach, University of Colorado, USA
Julie Schiro, University College Dublin

Read More


Stigma at Every Turn: Exploring Bi+ Consumer Experiences

Abigail Jean Nappier Cherup, University of Nebraska-Lincoln
Andre F. Maciel, University of Nebraska-Lincoln

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.