<%@ Language=VBScript %> <% CheckState() CheckSub() %> Methods and Measures that Profile Heavy Users
Journal of Advertising Research

Journal of Advertising Research


Advertising Research Foundation
641 Lexington Ave., New York, NY 10022; (212) 751-5656

August/September 2000


Methods and Measures That Profile Heavy Users

Brian Wansink
and
Sea Bum Park
University of Illinois

The 'heavy half' concept generally suggests that 80% of the volume of a product is consumed by 20% of its consumers (Twedt, 1964). Ever since the introduction of the '80/20 Rule,' many managers of packaged goods have attempted to target consumers based on the volume they purchased (Haley, 1968; Wansink, Kent, and Hoch, 1998) or consumed (Wansink, 1994, 1997a, 1998a). Unfortunately, many efforts to identify characteristics of heavy users have been plagued with methodological or measurement problems that have limited the depth and usefulness of their conclusions. Many of these studies incorrectly concluded that heavy users are not especially different from light users (Clancy and Schulman, 1993). This paper identifies methods and measures that can be most effective in differentiating heavy users.

Two methods � the simple mean comparison method and the basic cluster method � have been used to examine heavy users. Both have their shortcomings. The commonly used simple mean comparison method can lead to overgeneralized, nondiagnostic profiles. The basic cluster method can lead to overly specific profiles, which are mathematically valid but empirically nonexistent. What deserves to be examined is a hybrid technique that combines the best features of a mean comparison method with those of two-stage clustering.

The question then turns to which measures of characteristics will most effectively differentiate heavy users from light users. Past marketing research has used demographics, psychographics, benefits, or behavioral variables to profile heavy-user segments (Assael, 1973; Assael and Poltrack, 1994; Assael and Roscoe, 1976; Bass et al., 1968; Goldsmith et al., 1994). Demographic data, however, have their limits in helping to generate insights about customers (Haley, 1984), and psycho-graphic data are costly to assemble and difficult to objectively interpret (Wells, 1974). Likewise, the benefits a person seeks in a product are also limited in their ability to provide reliable profiles that are stable over time. While most of these variables seem to be indeterminant, by integrating them we can better examine the common factors they might represent (Sheth, 1974). In summary, this research examines two questions:

  1. Which segmentation method generates the most diagnostic heavy-user profiles?
  2. Which characteristics best differentiate heavy users?

To answer these questions, we first review previous research on the methods and measures used to profile heavy users. Second, we report results from a national survey to show how different segmentation methods can profile the same user groups quite differently. The methods and measures we test show that a hybrid cluster method provides distinctly different profiles than the more commonly used mean comparison model. We also show that personality variables are more differentiating than lifestyle variables. Following this, the implications for identifying heavy users are discussed along with how this information can be used to investigate whether media preferences and shopping baskets ('affinity marketing') can be used to profile heavy users of a particular product.

DIFFERENTIATING HEAVY USERS

What segmentation method generates the most diagnostic heavy-user profiles?

The performance of most marketing programs is determined by its effectiveness and efficiency. To this end, identifying and profiling heavy users has been a 'Holy Grail' to some brand managers. Unfortunately, most of these studies have generated only modest descriptive and explanatory power (Goldsmith et al., 1994; Singh, 1990). More disconcerting, however, is that other studies have generated inconsistent findings. For example, some studies found that heavy users are more deal prone, but other studies found quite oppositely that they are innovators or early adopters (Hackleman and Duker, 1980; Morgan, 1970). Why the inconsistencies? They may come from variations in measurement and methodologies. While some efforts captured differences between the segments, others found only noise.

Two basic techniques that have been used to cluster heavy users have been the simple mean-comparison method and the basic clustering method. The commonly used mean-comparison method simply divides 'heavy half' consumers from the 'light half' and evaluates the mean differences across these two groups. While a simple mean-comparison method of profiling heavy users has often been seen as too crude, the basic clustering method has also met with skepticism (Frank and Green, 1968; Wells, 1975). From a methodological viewpoint, cluster analysis groups consumers together based on a set of relevant variables, such as personality and lifestyle, without any prior assumptions about important differences that might differentiate usage. Unless usage segments are very different, however, homogeneous groups identified from the cluster analysis can appear to be heterogeneous because researchers can arbitrarily choose the number of clusters.

To solve the ad hoc nature of the basic clustering method, Punj and Stewart (1983) suggested a two-stage cluster analysis that helps enhance external validity as well as internal reliability. Figure 1 describes how this hybrid two-stage cluster analysis can be used to profile heavy users.

Considering that both mean comparisons and the two-stage cluster analysis have their own advantages and disadvantages we believe that combining the two methods can generate more diagnostic segment profiles based on the same population (Punj and Stewart, 1983) as described in in Table 1, we expect that the hybrid two-stage cluster analysis may resolve some of the fundamental problems found with both other methods.

TABLE 1: THREE METHODS TO PROFILE HEAVY USERS

 

Simple Mean Comparisons 

Basic Cluster Analysis

 Hybrid Two-Stage Cluster Analysis

Description and Procedures

A priori segmentation basis is decided based on a single variable (for example product usage, loyalty, customer type).
Segment�s size, sociodemographic, psychographic, and other relevant characteristics are compared.




No prior assumptions about differences within a population
Cluster respondents  are based on a set  of relevant multiple variables (for example personality, lifestyle, preference, behavior).
Cluster�s size, sociodemographic, psychographic, characteristics are compared.





Prior assumptions are made about differences within a population.

Utilizes both hierarchical and nonhierarchical cluster analyses to group respondents.

Cluster�s size,  sociodemographic, and   psychographic, and other relevant other relevant characteristics are compared.

Precise



Advantage

Simple and easier to do and understand

Familiar and widely available variables

Low cost and time efficient

More scientific and  empirical 

Generates graphical structure of data (for example, dendogram, overlapping clusters) 

Hidden dimensions can be found

Considers outlying
characteristics of  population

Identifies overlaps within and across segments

Identifies subsegments of heavy users

Disadvantage

Overemphasizes an a priori segmentation basis

Choosing number of clusters can sometimes be arbitrary.

Time consuming

 

Ignores possible overlaps between segments

The best clustering algorithm for one may not be the best for another.

May lead to oversegmentation

 

May lead to overgeneralizations

Can generate unneeded clusters that are nonexistent within the population.

 

 

 

 

 

 

 

 

 

Despite its robustness, any method will be ineffective if it collapses too many distinct segments of consumers together. Not all heavy users are created equal, and aggregating across two different groups of heavy users can give a blurry profile, because the result is a hybrid customer � perhaps one that does not even exist. For example, while one might be a heavy user for convenience, the other might be for price. Or one might be a heavy user of a store brand while the other a heavy user of that same category�s premium brands.

A two-stage cluster analysis enables different profiles to emerge as two- and three-cluster solutions are used. As has been found, when multiple clusters are examined, the first cluster is maximally different from subsequent clusters. That is, the second and third clusters are often more likely to show overlaps in their profiles. A recent study looking at soup preferences (Wansink and Park, 2000) indicated that while the first cluster profiles very clearly differentiate soup lovers of various soup flavors (tomato, vegetable, chili, and so on), the second clusters were more similar than different (see Figure 2). This means that while multiple cluster solutions are important for differentiating heavy users, their real value comes more in their ability to provide distinct, highly differentiated first clusters than to provide highly differentiated second and third clusters.

Which characteristics best differentiate heavy users?

While some researchers believe demographic profiles of heavy users provide safe surrogates for psychographic profiles (Assael and Poltrack, 1994; Grenhaug and Zaltman, 1981), several researchers have expressed their skepticism about the use of demographic and psychographic data as a basis for market segmentation (Frank et al., 1972; Kassarjian, 1971). Indeed, demographic variables have been shown to be poor predictors of brand-choice behavior, partly because of narrowing differences in income, education, and occupational status in an affluent, mass consumption society (Sheth, 1974). Even when demographics discriminate heavy users of products, they still have their limits in helping generate insights about customer segments (Clancy and Schulman, 1994).

However, lifestyle and personality characteristics that are specific to certain consumers and product categories must be defined and measured in order to be useful to marketers. For instance, it would be relevant for a food company to identify a health-conscious segment or for a clothing company to identify a fashion-conscious segment. In other words, personality and lifestyle characteristics, in contrast to demographics, need to be defined by researcher�s objectives. Even if marketers get managerial insights from the psycho-graphics of consumers that are consistent over time, they are still costly and lack objectivity. For example, the Value and Lifestyle (VALS) typology provides rich description about a unique way of life and values (Mitchell, 1981), but the pieces may or may not have any relevance for a certain brand. This also holds true for benefit segmentation, which has been considered the most preferred segmentation methodology for almost two decades (Haley, 1968, 1984). If two competing brands provide an identical benefit that consumers want, benefit segmentation may not fully discriminate the target consumer segment for one brand from another.

When focusing on heavy users, however, the universe of potentially discriminating variables decreases. Heavy-user studies are typically focused on food and fast-moving consumer packaged goods (Wansink and Gilmore, 1999), and issues of fashion leadership and aesthetic taste become less important.

To better begin developing a standard for personality and lifestyle characteristics that differentiate heavy users of foods and packaged goods, a collection of personality and lifestyle characteristics were collected from past studies (see Wansink and Sudman, forthcoming 2001). These variables were subsequently modified and combined through factor analysis to arrive at 14 lifestyle and 20 personality variables that have been effective in differentiating heavy users of various foods and packaged goods. A factor analysis of these variables was completed and the dimensions shown in Table 2 were found to be significant in differentiating heavy users.

The factor analyses of these variables showed that there were six lifestyle factors that differentiated heavy users (active lifestyle, family-spirited, homebody, intellectually stimulating pastimes, TV lover, and pet lover) and five personality characteristics that tended to differentiate heavy users (mentally alert, social, athletic, carefree, and stubborn).

Of particular interest is that there were a number of variables that have been used in commercial studies but not in academic studies. These revolved around lifestyles and interests and include the extent to which a person is a pet lover, workaholic, good cook, churchgoer, TV addict, book lover, technology whiz, or world traveler.

METHOD

One key objective of this study is to determine how accurately the segments obtained through a mean-comparison method and a hybrid two-stage cluster analysis differentiates heavy users. American adults from 18 to 72 years old (mean of 47) were surveyed by telephone in June of 1999. The survey collected the key personality and lifestyle traits outlined in Table 2, along with demographic information and behaviors and preferences related to soup and soup consumption. A total of 1,003 interviews were completed among 602 women and 401 men.

To create user profiles, two basic segmentation methods were used: a simple mean-comparison and a hybrid two-stage cluster analysis. In the mean-comparison method, consumers were categorized into three user groups, heavy, light, and non-users, and comparisons of each user groups average responses were made. (This was conducted on lifestyle characteristics and on personality characteristics.) In the hybrid two-stage cluster analysis, we modified the two-stage cluster analysis method proposed by Punj and Stewart (1983) by conducting three separate two-stage cluster analyses for heavy, light, and nonusers.

 RESULTS

Which segmentation method generates more diagnostic heavy-user profiles?

Consumers were categorized into three groups: heavy users (38% ate soup more than once a week), light users (47% ate any soup more than once a month), and nonusers (15%). Our results indicate that a simple mean-comparison method provided somewhat distinctive but limited profiles of heavy users. Heavy users are socially active, creative, optimistic, witty, and less stubborn than light- and nonusers. Yet even though heavy and light users were compared across 34 characteristics, only 3 characteristics differentiated the two groups.

To allow more discrimination between profiles, hybrid two-stage cluster analyses were conducted across lifestyle and personality characteristics. As indicated in Figure 3, both personality and lifestyle hybrid two-stage clustering methods generated more diagnostic heavy-user profiles than the mean-comparison method. While the mean-comparison method provided the least differentiation across light and heavy users (only 9% of characteristics differentiated the user groups), the clustering methods differentiate among 36% of the lifestyle characteristics and 85% of the personality characteristics. Figure 3 illustrates these levels of differentiation and Appendices 1 and 2 provide additional detail.

In contrast to the mean-comparison method, both the lifestyle clustering method and the personality clustering method provide richer descriptions of heavy users. Both demonstrate that heavy users are not all the same. Furthermore, both show, without further separating heavy users on the basis of an a priori variable (such as benefits sought, health consciousness, and so on), the simple mean-comparison method inaccurately combines distinct types of heavy users together into one generic profile.

TABLE 2: LIFESTYLE AND PERSONALITY VARIABLES THAT DIFFERENTIATE USERS AND PREFERENCE

Lifestyle

Personality

Active Lifestyle

Mentally Alert

I am outdoorsy.

I am intellectual.

I am physically fit.

I am sophisticated.

I am a workaholic. 

I am creative.

I am socially active.

I am detail-oriented.

Family Spirited

I am witty.

I am family-oriented.

I am nutrition conscious.

I am a churchgoer.

Social

Homebody

I am fun at parties.

I enjoy spending time alone.

 I am outgoing.

I am a homebody.

I am not shy.

 I am a good cook. 

I am spontaneous.

Intellectually Stimulating Pastimes

I am a trendsetter.

 I am a technology whiz.

Athletic

I am a world traveler. 

I am athletic.

I am a book lover.

I am competitive.

TV Lover

I am adventurous.

I am addicted to TV.

Carefree

Pet Lover

I am down-to-earth.

I am a pet lover. 

I am affectionate.

I am a fun-lover.

Stubborn

I am optimistic.

I am stubborn.

 

I am sarcastic.

Note. Principal component analysis with varimax rotation was used. Suggested factors have Eigen values greater than 3 and explain 55% (lifestyle) and 50% (personality) of variances in total (see Wansink and Park, 2000 for details). Variables were measured on a 5-point scale (1 = strongly disagree; 5 = strongly agree).

Which characteristics best differentiate heavy users?

Do all heavy users fit the same profile? As discussed earlier, our major concern is that not all heavy users are alike. One heavy user of soup might be a heavy user for convenience, while another might be one for price. These two might have very different profiles. For example, even if our results indicate that heavy users are socially active, creative, optimistic, and witty, we believe there may be significant differences among heavy users that we have overlooked.

To examine which variables personality characteristics or lifestyle characteristics better profiled heavy users, we compared the numbers of variables that showed statistically significant differences in means between soup users in segment (cluster) 1 and segment (cluster) 2. As indicated in Table 4, personality segment number 1 showed that light users were 100% different from nonusers and heavy users were 85% different from light users in terms of their personality, while the two lifestyle segments showed 71% and 36%, respectively.

Consider the two personality segments of heavy users (Table 3). One segment might be classified as 'traditionalists' while the other group might be classified as dynamos.' The traditionalists enjoy home and family magazines but are not affectionate, sophisticated, competitive, trend-setting, intellectual, nutrition conscious, fun loving, sarcastic, or stubborn. This segment is in dramatic contrast to the second segment. The 'dynamo' segment likes reading sports magazines and is adventurous, creative, outgoing, athletic, optimistic, fun at parties, witty, spontaneous, detail-oriented, and down-to-earth. While it is interesting that the heavy user of soup segments could be so dramatically different, this is because both groups eat soup for different reasons. The 'traditionalist' segment eats soup because it is an 'inexpensive meal solution' that 'goes with other foods.' The 'dynamo' segment eats soup because it is 'quick and convenient' and perceived as 'healthy.'

All heavy users may not be uniformly considered the same customer segment, and there may be overlapping clusters among heavy, light, and nonuser segments (recall Figure 2). The hybrid two-stage cluster analyses for personality and for lifestyle characteristics supported the notion that the heavy users could be differentiated from each other because they were not all alike. While the distinctiveness between Segment1 and Segment2 depends on whether we use lifestyle or personality variables, Segment1 appears to be more differentiated. This is consistent with the pattern shown in Figure 2. Interestingly, the new lifestyle variables that were tested - pet lover, TV addict, book lover, and churchgoer differentiated between heavy and light users and between the two segments of heavy users.

SUMMARY AND DISCUSSION

Implications for profiling heavy users

This research showed how two general segmentation methods, a simple mean comparison and a hybrid two-stage cluster analysis, can lead to very different lifestyle and personality profiles of heavy users. When the results of the two methods are compared, the hybrid two-stage cluster analysis which combines the simple mean-comparison method and basic cluster analysisis the more diagnostic method for building heavy-user profiles.

TABLE 3: LIFESTYLE AND PERSONALITY PROFILES OF SOUP USERS: MEAN COMPARISONS VS. CLUSTER ANALYSES

 

Mean Comparison Method

Cluster Comparison Method

 

Lifestyle
Characteristics

Personality
Characteristics

Segment 1

Segment 2

Segment 1

Segment 2

Nonuser

Bad cooks

Stubborn

Not family oriented

Socially active and outdoorsy

Not competitive

Adventurous, creative, outgoing athletic

 

Introvert

Not witty

No pets

 

Shy

Very optimistic

 

 

 

 

 

Not affectionate or sophisticated

Fun at parties and witty

 

 

 

 

 

Not detail oriented or intellectual

Fun lovers

 

 

 

 

 

Not down-to-earth or sarcastic

Trendsetters

 

 

 

 

 

 

Spontaneous

 

 

 

 

 

 

Nutrition conscious

Light user

Family orientated

Not creative

Enjoy reading, hobbies, magazines

Enjoy reading business magazines

Like reading,
health and fitness, human interest, and entertainment magazines

Enjoy reading home and family magazines

 

Book lovers

Not spontaneous

Owns a pet

Not a book lover

Adventurous, creative, outgoing, athletic

Shy

 

 

 

 

 

Very optimistic but sarcastic and stubborn

Not affectionate
or sophisticated

 

 

 

 

 

Fun at parties, love fun and witty

Not competitive

 

 

 

 

 

 

Not spontaneous

 

 

 

 

 

 

Not nutrition conscious

 

 

 

 

 

 

Not a trendsetter
or intellectual

Heavy user

Socially active

Nutrition conscious

Not book lovers

Outdoorsy

Enjoy reading home and family magazines

Like reading
sports magazines

 

Technology whiz

Optimistic

Not church goers

Technology whiz

Not affectionate
or sophisticated

Adventurous and creative

 

Good cooks

Down-to-earth

Not pet owners

Not addicted to TV

Not competitive

Outgoing and athletic

 

 

 

 

 

Not a trendsetter or intellectual

Very optimistic

 

 

 

 

 

Not nutrition conscious

Fun at parties and witty

 

 

 

 

 

Not a fun lover

Spontaneous but detail orientated

 

 

 

 

 

Not sarcastic or stubborn

Down-to-earth

TABLE 4: PERSONALITY PROFILES DIFFERENTIATE BETTER THAN LIFESTYLE  PROFILES

 

Segment (Cluster) 1

Segment (Cluster) 2

 

Non- and Light Users

Light and Heavy Users

Non- and Light Users

Light and Heavy Users

Lifestyle Cluster 

71%
(10/14)

36%
(5/14)

36%
(5/14)

29%
(4/14)

Personality

100%

85%

95%

95%

Cluster

(20/20)

(17/20)

(19/20)

(19/20)

Note: Percentages represent the percentage of characteristics that differentiated user groups.

Second, these results provide an important benchmark by specifying the characteristics that differentiate heavy from light users. Although personality variables generate more distinctive and unique heavy-user profiles than the lifestyle and demographic variables, all help build rich heavy-user profiles. In particular, it was found that there were six lifestyle factors that differentiated heavy users (active lifestyle, family-spirited, homebody, intellectually stimulating pastimes, TV lover, and pet lover) and five personality characteristics that differentiated heavy users (mentally alert, social, athletic, carefree, and stubborn). Of particular interest is that there were a number of variables that have been used in commercial studies but have not been used in academic studies until now. These include the extent to which a person is a pet lover, workaholic, good cook, churchgoer, TV addict, book lover, technology whiz, or world traveler.

Last, this study provides useful guide-lines and methodological implications for future research on heavy users as well as customer database management. Future research on heavy users needs to adopt a rigorous segmentation approach in order to provide accurate profiles of heavy users. One way to do this is by using consumer profiles generated from one method that need to be supported by another method, such as customer prototyping techniques (Wansink, 1999). Managers of customer databases need to take a holistic attitude in building and managing their customer database because any single variable alone, such as personal lifestyle, demographics, or media preference, cannot provide full descriptions of customers.

TABLE 5

 

User group

 

Magazine preference

Nonusers

Light users

Heavy users

Chi-square

Hobby

23%

13%

18%

3543

Health & Fitness

5%

5%

2%

1.775

Business

2%

2%

2%

0.297

Sports

3%

9%

2%

7.928**

Human Interest

8%

20%

15%

5.658*

Home & Family

15%

14% 

23%

5.708*

News

8%

13%

14%

1.588

Entertainment 

8%

11%

5%

4.528

Womens

6%

7%

8%

0.541

*p<0.1. **p<0.05

Can media preferences differentiate heavy users?

Media preferences have often been used as a segmentation device. While broadcast programming preferences have not always proved to be diagnostic, it may be that magazine subscriptions are a stronger indicator of preference since they necessitate action (subscribing) and payment. Can media preference as measured by magazine subscribership differentiate heavy, light, and nonusers? The results in Table 5 suggest so. Compared to light users, heavy users are less likely to read sports and entertainment magazines and more likely to read home and family magazines.

The result of these differences in magazine readership suggest that consumers preferences across other products might also be differentiated across heavy, light, and nonusers. 'Affinity marketing' is of significant interest to e-commerce and to database marketing, because it suggests a bundle or a metaphorical 'shopping basket' of products that an ideal target market might be attracted to (Wansink and Ray, 1992, 1996). This provides new justification for marketers to try building their customer prototypes based on their preference for certain products or brands along with building customer prototypes based on personality and lifestyle data.

CONCLUSION

We identified the methods and measures that can be used to effectively profile heavy users. The most effective market segmentation generates the most accurate, detailed, diagnostic, and in-depth profiles of heavy users. This article shows the shortcomings of the commonly used mean-comparison method of heavy-user segmentation, and it outlines a clustering method that effectively differentiates different types of heavy users from light users. The various characteristics that differentiated heavy users from light users are shown to relate to five basic lifestyle factors and six personality factors. While providing a key starting point for studying heavy users, they also show the dominant role that personality characteristics (versus lifestyle or demographic characteristics) play in differentiating heavy users from light users.

APPENDIX 1: THE MEAN COMPARISON METHOD OF PROFILING SOUP USERS

 

 

Nonuser

Light User 

Heavy User

 

 

 

(n = 147)

(n= 471)

(n = 383)

F-values

Lifestyle

Homebody 

 2.6

2.8

2.8

2.828*

 

Socially Active

3.0

3.1

3.2

3.144**

 

workaholic

2.6

2.5

2.5

0.600

 

Physically Fit

2.9

2.9

3.0

2.119

 

Pet Lover

2.6

2.6

2.7

0.797

 

Outdoorsy

3.2

3.1

3.2 

1.420

 

TV addict 

2.4

2.3

2.3

0.087

 

Book Lover 

2.7

3.0

3.0 

4.830***

 

Family-oriented

3.4

3.6

3.6

3.704**

 

World Traveler

1.9

1.9

2.0

2.477*

 

Technology whiz

2.0

2.2

2.2

3.728**

 

Churchgoer

2.4

2.6

2.6

1.259

 

Good Cook

2.8 

3.1

3.1

4.376**

 

Spending Time Alone 

2.9

2.8

2.9

0.629

Personality

Adventurous

3.0

 2.9

3.0

1.709

 

Affectionate

3.4

3.5

3.5

0.833

 

Creative

3.1

3.0

3.2

4.438**

 

Outgoing

3.1

3.2

3.3

2.993*

 

Athletic

2.5

2.6

2.7

1.501

 

Down-to-Earth 

3.5

3.6

3.6

4.096**

 

Fun loving

3.5

3.5

3.6

0.791

 

Shy

2.1

2.0

1.9 

1.851

 

Sophisticated

 2.3

2.3

2.4

1.263

 

Trendsetter

1.9

1.9

2.1

2.992*

 

Nutrition Conscious

2.6

2.9

2.9

5.898***

 

Optimistic

3.1

3.2

3.3

3.149**

 

Detail-oriented

3.0

3.1

3.1

0.695

 

Competitive

2.7

2.7

2.8

0.365

 

intellectual

2.8

3.0

2.9

1.541

 

Spontaneous 

2.8

3.0

3.1 

3.152**

 

Sarcastic

2.7

2.5

2.6 

2.733*

 

Stubborn

3.1 

2.9

2.8

3.892**

 

Fun at Parties

2.9

2.9

2.9

0.079

 

Witty 

2.8

2.9 

3.0

4.158**

Demographicsa

Gender

Female

Female

Female

 

 

Age

35 to 44

35 to 44

35 to 44

 

 

Education Level

Voc./Tech - Some College

Some College

Some College

 

 

Income Level

$15K to $34.9K

$35K to $50K

 R35K to $50K

 

 

Primary Grocery Shopper

Yes

Yes

Yes

 

 

Primary Meal Preparer 

Yes

Yes

Yes

 

 

No. of Children under 17

None

None 

None

 

*p < .01**p < .05;***p < 0.01
aCentral tendencies of demographics indicate median values.

   APPENDIX 2: CLUSTERING METHODS AND MEASURES THAT DIFFERENTIATE SOUP USERS

 

Soup Nonuser

F-values/

Soup Light User

 

Soup Heavy User

 

 

Cluster 1

Cluster 2

Chi.square*

cluster 1

cluster 2

F-values

cluster 1

cluster 2

F-values

Lifestyle Clusters

 

 

 

 

 

 

 

 

 

Homebody 

2.5

2.6

0.110

2.8

2.7

2.114

2.8

2.8

0.007

Socially Active

2.7

3.1

7.775***

3.1

3.1

0.017

3.2

3.2

0.073

Workaholic

2.5

2.7

1.440

2.6

2.5

1.231

2.4

2.6

1.342

Physically fit

2.6

3.0

4.490

2.9

3.0

1.289

3.0

3.1

0.683

Pet Lover

1.4

3.4

133.228***

3.9

1.2

6544.462***

1.1

3.9

5358.208***

Outdoorsy

2.8

3.4

13.920***

3.2

3.1

3.018*

3.0

3.4

20.649***

TV Addict

2.4

2.3

0.591

2.4

2.2

3.012*

2.5

2.2

5.077**

Book Lover

2.5

2.9

3.094**

3.1

2.8

5.454**

2.9

3.1

4.267**

Family-oriented

3.1

3.7

16.243***

3.6

3.6

1.270

3.6

3.7

2.908*

World Traveler 

1.8

2.0

0.591

1.9

1.9

0.282

2.0

2.0

0.005

Technology Whiz

1.6

2.2

14.756***

2.2

2.2

0.294

2.1

2.3

3.526*

Churchgoer 

2.1

2.7

9.273***

2.5

2.7

3.213*

2.8

2.5

5.064**

Good Cook

2.1

3.2

43.781***

3.1

3.0

0.944

3.1

3.1

0.029

Spending Time Alone

2.7

3.0

3.284*

2.8

2.8

0.325

2.8

2.9

2.215

Number of Cases in Each Cluster 

57

89

 

254

209

 

152

224

 

Personality Clusters

 

 

 

 

 

 

 

 

 

Adventurous

2.7

3.3

22.788***

3.3

2.4

138.859***

2.6

3.4

78.573***

Affectionate

3.1

3.7

21.331***

3.6

3.4

15.460***

3.4

3.6

3.928**

Creative

2.8

3.4

14.339***

3.2

2.7

39.113***

2.9

3.5

53.584***

Outgoing

2.6

3.6

54.170***

3.5

2.7

147.654***

2.9

3.6

71.011***

Athletic

2.1

3.0

33.759***

3.0

2.0

119.137***

2.1

3.2

143.345***

Down-to-earth

3.3

3.6

5.522**

3.6

3.6

0.118

3.7

3.6

4.523**

Fun Loving

3.2

3.8

17.167***

3.8

3.3

70.760***

3.5

3.7

17.384***

Shy

2.4

1.8

10.726***

1.9

2.2

18.861***

2.0

1.9

2.115

Sophisticated

2.0

2.7

22.930***

2.6

1.9

90.105***

1.9

2.9

138.997***

Trendsetter

1.5

2.3

33.363***

2.3

1.4

145.701***

1.6

2.6

112.778***

Nutrition Conscious

2.2

3.1

32.515***

3.0

2.6

23.667***

2.9

3.1

4.115**

Optimistic

2.7

3.4

21.128***

3.4

3.1

14.152***

3.1

3.4

10.222***

Detail-oriented

2.7

3.4

16.166***

3.2

3.1

0.760

2.8

3.5

49.002***

Competitive

2.2

3.2

31.577***

3.2

2.1

166.767***

2.3

3.3

103.711***

Intellectual

2.5

3.2

26.523***

3.2

2.7

39.084***

2.5

3.3

109.243***

Spontaneous

2.4

3.3

46.884***

3.3

2.6

90.568***

2.7

3.4

59.231***

Sarcastic

2.3

3.1

15.115***

2.8

2.1

60.406***

2.2

2.9

45.236***

Stubborn

2.9

3.2

3.689*

3.2

2.6

34.880***

2.5

3.1

37.800***

Fun at Parties

2.4

3.4

43.433***

3.3

2.3

177.426***

2.5

3.3

76.220***

Witty

2.5

3.3

34.788***

3.2

2.4

105.797***

2.7

3.3

58.190***

Number of Cases in Each Cluster 

73

69

 

262

197

 

176

191

 

 

Soup nonuser

Soup light user

Soup heavy user

 

Cluster 1

Cluster 2

Cluster 1

Cluster 2

Cluster 1

Cluster 2

Demographic clusters a

 

 

 

 

 

 

Gender

Male

Female

Female

Female

Female

Female

Age

25 to 34

35 to 44

35 to 44

35 to 44

45 to 54

35 to 44

Education level

HS grad

some college

some college

some college

some college

some college

Income level

$15K to $34.9K

$35K to $50K

$35K to $50K

$35K to $50K

$35K to $50K

$35K to $50K

Primary grocery shopper

yes

yes

yes

yes

yes

yes

Primary meal preparer

yes

yes

yes

yes

yes

yes

No. of children under 17

None

5 or more

None

None

None

None

*p<0.1; **p<0.05; ***p<0.01
a Central tendencies of demographics indicate median values

REFERENCES

ASSAEL, HENRY, and A. MARVIN Roscoe, JR. 'Approaches to Market Segmentation Analysis.' Journal of Marketing 40, 4 (1976): 6776.

_____ and DAVID F. POLTRACK. 'Can Demographic Profiles of Heavy Users Serve as Surrogate for Purchase Behavior in Selecting TV Programs?' Journal of Advertising Research 34, 1(1994): 1117.

BASS, FRANK M., DOUGLAS J. TIGERT, and RONALD, T. LONSDALE. 'Market Segmentation: Group versus Individual Behavior.' Journal of Marketing Research 5, 3 (1968): 26470.

CLANCY, KEVIN J., and ROBERT S. SHULMAN. Marketing Myths That Are Killing Business: The Cure for Death Wish Marketing. NY: McGraw-Hill, Inc., 1994.

FRANK, RONALD F., and PAUL E. GREEN. 'Numerical Taxonomy in Marketing Analysis: A Review Article.' Journal of Marketing Research 5, 1 (1968): 8398.

WILLIAM F. MASSY, and YORAM WIND. Market Segmentation. NJ: Prentice-Hall, Inc., 1972.

GOLDSMITH, RONALD E., and LEISA REINECKE FLYNN. 'An Empirical Study of Heavy Users of Travel Agencies.' Journal of Travel Research 33, 1 (1994): 3843.

GRONHAUG, KJELL, and GERALD ZALTMAN. 'Complainers and Non-complainers Revisited: Another Look at the Data.' In Advances in Consumer Research, 8, Kent Monroe, ed. Ann Arbor: Association for Consumer Research,1981.

HACKLEMAN, EDWIN C., and JACOB M. DUKER. 'Deal Proneness and Heavy Usage; Merging Two Market Segmentation Criteria.' Journal of the Academy of Marketing Science 8, 4 (1980): 33244.

HALEY, RUSSELL I. 'Benefit Segmentation: A Decision-Oriented Research Tool.' Journal of Marketing 32, 3 (1968): 3035.

______ 'Benefit Segments: Backwards and Forwards.' Journal of Advertising Research 28, 3 (1984): 1925.

KASSARJIAN, HAROLD H. 'Personality and Consumer Behavior: A Review.' Journal of Marketing Research 8, 4 (1971): 40918.

MITCHELL, ARNOLD. Changing Values and Lifestyles. CA: SRI International, 1981.

MORGAN, FRED W., JR. 'Are Early Triers Heavy Users?' Journal of Business 52, 3 (1979): 42934.

PUNJ, GIRISH, and DAVID W. STEWART. 'Cluster Analysis in Marketing Research: Review and Suggestions for Application.' Journal of Marketing Research 20, 2 (1983): 13448.

SHETH, JAGDISH N. 'Role of Demographics in Consumer Behavior.' Faculty working paper #218, University of Illinois, UrbanaChampaign, Illinois, 1974.

SINGH, JAGDIP. 'A Typology of Consumer Dissatisfaction Response Styles.' Journal of Retailing 66, 1 (1990): 5799.

TWEDT, DIR WARREN. 'How Important to Marketing Strategy Is the Heavy User?' Journal of Marketing 28, 1(1964): 7172.

WANSINK, BRIAN.'Advertisings Impact on Category Substitution.' Journal of Marketing Research 31, 4 (1994): 50515.

'Making Old Brands New.' American Demographics, December 1997.

_____ 'Developing Accurate Customer Usage Profiles.' In Values, Lifestyles, and Psycho-graphics, Lynn Kahle, ed. Cambridge, MA: Lexington, 1999.

_____ Expansion Advertising.' In Why Advertising Works. John Philip Jones, ed. Thousand Oaks, CA: Sage Publishing, 1998.

______ ROBERT J. KENT, and STEPHEN J. HOCH 'An Anchoring and Adjustment Model of Purchase Quantity Decisions' Journal of Marketing Research 35, 1 (1998b): 7181.

____and SEA BUM PARK. 'Accounting for Bum Tastes: Building Consumer Preference Prototypes.' Journal of Database Marketing, 7, 4 (2000): 30820.

______ and MICHAEL L. R. 'Advertising Strategies to Increase Usage Frequency.' Journal of Marketing 60, 1 (1996): 3146.

______ and . 'Estimating an Advertisements Impact on Ones Consumption of a Brand.' Journal of Advertising Research 32, 3 (1992): 916.

______ and JENNIFER GILMORE. 'New Uses That Revitalize Old Brands.' Journal of Advertising Research 39, 2 (1999): 9099.

_____,and SEYMOUR SUDMAN. Consumer Panels. Chicago: American Marketing Association, forthcoming 2001.

WELLS, WILLIAM U. 'Life Style and Psycho-graphics: Definitions, Uses, and Problems.' In Life Style and Psychographics, William D. Wells, ed. Chicago: American Marketing Association, 1974.

'Psychographics: A Critical Review.' Journal of Marketing Research 12, 2 (1975):196213.

WIND, YORAM. 'Issues and Advances in Segmentation Research.' Journal of Marketing Research 15, 3 (1978): 31737.

ENDNOTES

1As shown in Figure 1, this clustering algorithm used a priori user group segment basis. We then obtained statistically significant numbers of clusters for each of three user groups by examining the percentage changes in the agglomeration coefficients from a hierarchical cluster analysis. For example, in lifestyle cluster analysis we found a two-cluster solution most appropriate after noticing the highest percentage changes of the agglomeration coefficient from two- to one-cluster solution in nonuser (8.9%), light user (10.9%), and heavy user (9.6%). Also, in personality clusters analysis a two-cluster solution generated the highest changes of agglomeration coefficients in nonuser (14.4%), light user (72.3%), and heavy user (12.0%). From these two-cluster solutions we obtained cluster cent roids that were used as initial seed points for K-means cluster analysis. We then obtained final cluster cent raids from the K-means cluster analysis and conducted a series of significance tests.



http://www.warc.com
Advertising Research Foundation