segmentation and market structure when both consumer and situational characteristics are explanatory

14
Segmentation and Market Structure When Both Consumer and Situational Characteristics Are Explanatory Dwayne Ball University of Nebraska Charles Lamb Lincoln University Roderick Brodie University of Auckland ABSTRACT This article addresses the problem of exploring data for useful segments (and subsequently, market structure) in markets in which demand is a function of both consumer and situational characteristics. Current exploratory segmentation methods use either persons or situations as the unit of analysis, but not both. Yet choice and needs are frequently an unknown function of both personal characteristics and situational characteristics. Under these conditions of unknown strength of sources of influence, it may be advantageous to explore data. To allow the fruitful use of data exploration, we propose the use of a unit of analysis we call the person-in-situation, sampling of these units from their population, segmentation of these units with respect to choices, and the post-hoe exploration of such segments for regularities in individual and situational characteristics. 0 1992 John Wiley & Sons, Inc. Psychology & Marketing Vol. 9(5): 395-408 (September/October 1992) 0 1992 John Wiley & Sons, Inc. CCC 0742-6046/92/050395-14$04.00 395

Upload: dwayne-ball

Post on 09-Aug-2016

218 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Segmentation and market structure when both consumer and situational characteristics are explanatory

Segmentation and Market Structure When Both Consumer and Situational Characteristics Are Explanatory Dwayne Ball University of Nebraska

Charles Lamb Lincoln University

Roderick Brodie University of Auckland

ABSTRACT

This article addresses the problem of exploring data for useful segments (and subsequently, market structure) in markets in which demand is a function of both consumer and situational characteristics. Current exploratory segmentation methods use either persons or situations as the unit of analysis, but not both. Yet choice and needs are frequently an unknown function of both personal characteristics and situational characteristics. Under these conditions of unknown strength of sources of influence, it may be advantageous to explore data. To allow the fruitful use of data exploration, we propose the use of a unit of analysis we call the person-in-situation, sampling of these units from their population, segmentation of these units with respect to choices, and the post-hoe exploration of such segments for regularities in individual and situational characteristics. 0 1992 John Wiley & Sons, Inc.

Psychology & Marketing Vol. 9(5): 395-408 (September/October 1992) 0 1992 John Wiley & Sons, Inc. CCC 0742-6046/92/050395-14$04.00

395

Page 2: Segmentation and market structure when both consumer and situational characteristics are explanatory

Interest in situational effects in consumer behavior has been prominent since a seminal article by Belk (1975); for example, see Miller and Ginter (1979). In markets such as restaurants and entertainment, for example, consumer choice is obviously affected by situational variables huch as mood, presence of other persons, situation-specific goal states, etc. Yet such markets would virtually never exhibit variation in choice that is entirely a function of situation. There will be variation in choice due to consumer individual differences as well.

This article presents a method of segmentation (and, based on the segmentation, a market structure analysis) that allows for a data-based balance of situational and consumer-specific effects. The method is based on the use of a unit of analysis called a person-in-situation, which seems appropriate when situational effects may be substantial.

For example, lunchtime in a downtown business area contains people who are driven by the constraints of a time-pressured situation to seek fast food. Another group of consumers may be driven more by individual characteristics rather than situational ones. These latter may prefer less expensive establishments generally, largely including fast food, across a very large variety of situations. They may differ from other consumers quite markedly in terms of individual characteristics such as socioeconomic status or attitudes. Within each group, the bases for competition will vary; the time-pressured group may rank alternative restaurants quite differently than the group that prefers less expensive restaurants, even though the choice set may be quite similar. To further complicate the issue, neither group is influenced by either situational or personal variables alone; the choices of both are influenced to some degree by both kinds of effects. Competition has a different character in each group.

Segmentation by situation alone ignores consumers, and by con- sumers alone ignores situation. Useful segments may quite reasonably be composed of certain kinds of people, or certain kinds of situations, or certain kinds of people in certain kinds of situations. Hence some method of segmentation which accounts for both sources of variation in choice is desirable.

It would further be desirable in many cases to allow these certain kinds of people and situations to be found by a data-based, post-hoc segmentation procedure, rather than to propose segments of consumers or situations a priori. A priori segmentation designs (Wind, 1978) are those in which management decides upon a basis for segmentation prior to research, and determines which segments shall be investigated. Re- search then estimates the sizes of segments and describes their char- acteristics. Post-hoc designs allow management to choose bases for seg- mentation, but use an analytic method to derive segments. Given the complexity of combining situational and individual consumer influ- ences on purchase, a post-hoc design may well be preferred.

Introducing situational effects into segmentation and market struc- ture analysis has been a topic of interest among methodologists (Dick-

396 BALL ET AL.

Page 3: Segmentation and market structure when both consumer and situational characteristics are explanatory

son, 1982; Hustad, Mayer, & Whipple, 1975; Srivastiva, Alpert, & Shocker, 1984; Stefflre, 1971). The approach presented here is meth- odologically differentiated from previously presented approaches in the definition of the unit of analysis combined with the methodological approach.

INTRODUCING SITUATION INTO SEGMENTATION AND MARKET STRUCTURE

Methods have been introduced to deal with situational effects in seg- mentation. Stefflre (19711, for example, grouped snack options by their relative presence in usage situations, aggregating across individuals. Hustad et al. (1975) grouped beverages by similarity ratings across users in each of eight common usage situations. Dickson (1982) seg- mented people and situations for sun lotions separately by judgement, and then created a person-type-by-situation-type matrix, suggesting appropriate products for each cell in the matrix. Srivastiva et al. (1984) created an empirical taxonomy of usage situations for forms of money and credit, based on a usage situation-by-product appropriateness ma- trix provided by consumers. Based on this taxonomy, product options could be grouped with respect to their similarity on the usage taxonomy dimensions.

We agree with Srivastiva et al. (1984) that market definition is best understood when utilizing the idea of a product market: ". . . the set of products judged to be substitutes within those usage situations in which similar patterns of benefits are sought by groups of customers." The product market concept, then, explicitly includes the ideas of segmen- tation based on both usage situations and customers, if not their in- teraction. The interaction of the two with product perceptions and choice may in fact be the most useful effects in understanding the market.

As an example, consider some segments that might be defined in the coffee market. In a situation defined by the characteristics of tight time constraints, a feeling of sleepiness, an anticipated need for alertness, no social constraints, and being at home (breakfast, students studying for finals, preparing for late-night intercity driving, etc.), caffeinated instant coffee would be the choice of many coffee drinkers. However, individual tastes might still dictate a caffeinated brewed coffee for some, who might be attitudinally distinct from those choosing instant. The former might refuse to drink instant coffee regardless of the situational characteristics, due to high involvement in coffee flavor and a discrim- inating taste self-image.

In another situation, defined by relaxed time constraints, a social atmosphere at home, and a desire to please and impress others, some consumers would choose mass-market brewed coffee, and others grind- them-yourself expensive beans from a specialty shop. The differences

SEGMENTATION AND MARKET STRUCTURE 391

Page 4: Segmentation and market structure when both consumer and situational characteristics are explanatory

between the consumers may have to do with involvement, taste, dis- posable income, or any of a number of consumer characteristics.

In the cases cited above, it would be necessary to first segment by customer group, in order to find homogeneous choices or preferences within the usage occasion. Or, one could segment by occasion first, then search for consumer groups within each occasion segment. Or, one can segment both consumers and occasions separately, into, say, five seg- ments each, then cross them into 25 possible segments. In any of these cases, a very large number of segments, many of questionable utility, are likely.

An alternative, which we propose here, is to find a way to segment the market that allows the data to determine if a segment shall be discriminated from other segments on the basis of characteristics of the occasions, consumers, or both. Perhaps the most useful segmentations of a market may be those in which some segments are person-based (instant decaf drinkers almost regardless of situation, for example), some situation-based (time constraints, alertness required), and some a mixture (upper-middle class, status-conscious consumers in an im- portant-guests-at-home-for-dinner situation).

“Kotler (1980) suggests that the usefulness of a particular segmen- tation approach be assessed in terms of ease of measurement, accessi- bility, and substantiality,” notes Dickson (1982). As a set of criteria for the segmentation approach proper, this seems reasonable. We hope to demonstrate that the first of these criteria is satisfied by the proposed method, and that the proposed method has an advantage with respect to the third. The second criterion, accessibility, may or may not be satisfied by this method, or any of the others proposed. However, use of situation-based descriptors of a segment increases the probability that a segmentation scheme will make the segments accessible, or allow choice of the most accessible segments.

Four criteria specific to a comprehensive method of situation-based market structure analysis are proposed by Shocker, Zahoric, and Stew- art (1984). These are (a) a means of specifying the appropriate set of products, (b) a means of dealing with the diverse requirements of dif- ferent usage situations, (c) a means of dealing with the different degrees of appropriateness in use for the same product, and (d) validity and reliability. After presenting the method, we will discuss each criterion in turn.

PROPOSED METHOD

Choice of the unit of analysis in segmentation is critical (Wind, 1978). We propose sampling from the universe of people-in-situations. An ele- ment of this universe consists of an occasion in which an individual uses one or more of the products in the product set. The definition can

~~

398 BALL ET AL.

Page 5: Segmentation and market structure when both consumer and situational characteristics are explanatory

be expanded to include occasions in which an individual would have preferred an element from the product set but could not obtain one. The products in the product set are determined to be those of managerial interest, although a consumer-based procedure may very well be used to aid in identifying the set. These products may be conceptual products as well as real ones, assuming consumers can accurately assess their own likelihood of purchase of a conceptual product.

Choice of measurement of the relationship between people, products, and occasions is critical as well. Some authors have used appropriate- ness. Others have used substitutability (Fraser and Bradford 1983) or switching (Vanhonacker 1980). We suggest measurement of preference. Appropriateness is only a necessary, but not sufficient condition for choice. Structure based on appropriateness may be misleading. Indeed, the lack of person-by-situation effects found by Srivastiva et al. may be a function of their use of the appropriateness measure-people in general may be quite in agreement on the appropriateness of products to a context, but in substantial disagreement on preference. Preference is generally the issue that motivates marketers in any case.

Choice of taxonomic methodology requires choosing bases on which the units of analysis are to be aggregated or disaggregated, and requires choosing an algorithm to perform the task. The basis, as mentioned above, is preference. The algorithmic approach could take a number of forms. Because this is an exploratory procedure, cluster analysis seems suitable. The particular brand of cluster analysis used in the study presented here is discussed below. The approach of asking for preference across some set of options, then clustering people into segments based on the similarity of their preferences, was mentioned by Pessemier (19821, although not in the context of situational effects.

METHODOLOGY OF THE EXAMPLE STUDY

The Christchurch, New Zealand radio market was investigated. There are seven radio stations broadcasting in this market, of which three are government sponsored and four are private. At the time this study was performed, all were AM stations, although one of the AM stations was broadcasting in both AM and FM as a test. Radio listening and the sorts of alternative behaviors with which it competes are clearly oc- casion-based phenomena (this was confirmed by exploratory research). In addition, we may expect substantial effects on behavior due to mus- ical or informational tastes, and needs arising from consumer charac- teristics such as life style.

Questionnaire Development and Administration The method of developing and administering the final questionnaire had five steps. These were as follows.

SEGMENTATION AND MARKET STRUCTURE 399

Page 6: Segmentation and market structure when both consumer and situational characteristics are explanatory

(1) Identification of the Competitive Set. Based upon managerial judge- ment and prior qualitative work, a competitive set consisting of 16 types of radio program format, 9 types of TV program format, 2 types of recreationallinformational reading, selecting one’s own music for the record or tape player, and chatting informally with a friend was de- veloped. The radio and TV program formats were expressed in the final questionnaire in generic terms, with one or more familiar examples appended if necessary to clarify them (e.g., “News, views, information with a slower pace, such as the 3YA national programme or Good Morn- ing New Zealand”). Some of these program types were not broadcast in the Christchurch area; most were available at one time of day only. Henceforth, in this article the 29 program types and competing activ- ities will be termed options.

A more refined procedure, such as that employed by Srivastiva et al. (1984) to suggest the competitive set, might have yielded a slightly more complete set. In their approach, samples of customers were given one product known to be of interest to management, and asked to sug- gest uses for it. Then they were asked to suggest other products that might be used in those usage contexts. Third, the customers were asked to suggest usage contests for the expanded product list. Managerial input was used to suggest other products not suggested by customers; some of these products were unavailable to the sample and were thus conceptual among the target group. This Srivastiva et al. procedure seems quite complete; however, the competitive set obtained for the present study seems sufficient for practical purposes and for illustration of methodology.

f2) Generation of Descriptors for the Options, Consumers, and Situa- tions. Focus groups were conducted regarding listening behavior. Par- ticipants were asked to describe radio options they preferred, to suggest the benefits (positive and negative) of various options, to describe lis- tening situations and what they were doing and how they were feeling, and to suggest the characteristics of people who listened to various op- tions. Focus group notes were used by the investigators as an aid in generating hypotheses regarding explanations for option choice. The characteristics of options, consumers, and situations fell into these ar- bitrary categories:

Characteristics of options: positive and negative benefits (calmed me, bored me, provided me with a feeling of belonging, gave me something to talk about, etc.).

Characteristics of consumers: demo- and psychographics (age, gender, income, education, etc., checklist-type life-style items primarily relat- ing to leisure activities).

Characteristics of situations: various activities (attention intensive versus nonattention intensive and goal oriented versus non-goal ori-

400 BALL ET AL.

Page 7: Segmentation and market structure when both consumer and situational characteristics are explanatory

ented), types of social interactions, and types of internal feeling states (tense, alert) and goals directed at the achievement of feeling states (such as relaxing or trying to concentrate). The Baumeister and Tice (1985) scheme of categories of situation attributes served to stimulate hypotheses here, based on the focus group notes.

(3) Reduction of the List of Characteristics. A pretest of the items was conducted among 300 consumers intercepted on shopping malls. Each was asked to focus on the most recent option of which he or she had availed him or herself. The large battery of characteristics was reduced into three groups (options, consumers, occasions) by three factor anal- yses, with the exception that the demographic items were not included in the factor analysis of the consumer characteristics. Several items were chosen to represent each factor and the remainder were discarded.

The characteristics of the occasions were reduced to nine factors: feeling active and alert, feeling lonely and frustrated, feeling tired and slow, feeling relaxed and busy, feeling tense and nervous, feeling busy and harassed, waking up or eating meals or listening to radio alone, moving around the house and taking care of family while feeling busy and harassed, and concentrating on paperwork or reading or conver- sation or TV.

The seven factors representing the characteristics of the options were involves and challenges, relaxes and calms, annoys and irritates, re- quires too much attention, keeps me on schedule, provides a feeling of belonging, and amuses and cheers.

The four life-style/leisure factors were low-activity entertainments (reading, watching TV, dining out, work in garden), hobbies and low- activity sports, high-activity sports (such as skiing and surfing), and night life (pubs, parties, concerts, and listening to music). Other items descriptive of the person, but not included in the lifestyle factors, were demographics: age of youngest child living at home, education level of subject, sex, age, marital status, employment status outside the home, and total family income.

(4) Devising a Sampling Scheme fur Consumers- Within-Occasions. The waking day was divided into five parts of approximately equal length: early morning, late morning, early afternoon, late afternoon, and eve- ning. In the administration of the final questionnaire, each participant would be asked if he/she had availed him or herself of radio, television, or recreational reading for at least 30 minutes during each of the five day parts on the previous day. Of those day parts for which the answer was yes, one day part was chosen at random about which to be asked in detail. The day part thereby formed an occasion.

Ideally, a smaller unit of time would have been chosen. It would be desirable to have all characteristics of the occasion, such as moods, social conditions, and so forth, constant over the occasion, and it could be

SEGMENTATION AND MARKET STRUCTURE 401

Page 8: Segmentation and market structure when both consumer and situational characteristics are explanatory

argued this is less likely as the investigated time interval increases. However, the difficulty of remembering such characteristics after the passing of a day presented a problem, which would have become more acute as the time interval was shortened. Also, for managerial reasons, radio formats are often of several hours in duration, and correspond roughly to the day parts in this study. The questions regarding such characteristics were designed to elicit the dominant mood, social con- straints, etc., which in most cases should have been related to the en- tertainment option chosen, The study was performed over an entire week as a mall intercept, excluding Sunday, when no malls were open. This procedure, then, constitutes an approximation to a random sam- pling of people-within-situations during which options are chosen.

(5) Administration of the Final Questionnaire. Each of the 1000 indi- viduals intercepted on the malls mentioned in (4) was asked to describe the chosen occasion in detail, with respect to the characteristics-of-the- occasion items developed above. This task served not only to provide the necessary occasion descriptors, but also to forcefully remind the subject of his or her activities, social milieu, mood, etc., on the occasion. Second, each subject was asked to rate each of the 29 options on a 7- point preference scale with respect to that occasion. Third, two of the options were selected at random and the subject was asked to rate each of the benefit scales developed above. Finally, the subject was asked to rate him or herself on the demographic and psychographic items.

Data Analysis

The data analysis had three steps.

( I ) Establishing Segments. 829 subjects-within-situations had perfectly complete data and could be used in the analyses. An agglomerative hierarchical cluster analysis, average linkage type, from the SAS sta- tistical package, was performed on the consumers-within-situations, using the 29 preference ratings to form similarities between them. This first clustering was done to identify outliers, as suggested by Punj and Stewart (1983). Three outlying groups of sizes 1,1, and 5 were identified and eliminated from the data base.

Again following suggestions by Punj and Stewart, a K-means algo- rithm (from SAS) was applied to the remaining data from 822 subjects. Again, the data clustered were the preference ratings made of the 29 entertainment options, rated on a 1-7 scale. Clusters formed on this basis contain people who would have chosen similar options in the situations they rated. Nine clusters, or segments, of sizes 42-160 were identified, using criteria of variance reduction and interpretability. The nine segments, especially in light of their differences with respect to the desired benefits, situational characteristics, and personal charac-

402 BALL ET AL.

Page 9: Segmentation and market structure when both consumer and situational characteristics are explanatory

teristics, as discussed below, were highly interpretable and manage- rially meaningful.

(2) Differentiating the Segments. Composite scores (one per factor) for the occasion characteristics, person characteristics, and option char- acteristics factors were computed. These composite scores represented the proportion of the items loading on a factor that the person-in-sit- uation checked. Each score represents the strength of the situational, option, or personal factor for that person-in-situation. These composite scores were used as dependent variables in ANOVA and discriminant analysis techniques to suggest the factors that differentiated the seg- ments. The time-of-day distribution was also examined for each cluster; as expected, most clusters were concentrated in one or two day parts (Figure 1).

It should be recapitulated here that the broad approach described in this article is an exploratory and not a confirmatory technique. There- fore, although some other authors (e.g. Srivastiva et al., 1984) have focused on confirming the existence and magnitude of person and sit- uation main effects and interactions, our purpose here is to raise data- based hypotheses regarding which particular effects may explain a use- ful segmentation.

(3) Perceptual Mapping. Means for the seven benefit factors were cal- culated for each option across the entire data set, producing a seven-

DA

CLUSTER (SEGMENT)

Figure 1. Cluster (segment) size by day part.

SEGMENTATION AND MARKET STRUCTURE 403

Page 10: Segmentation and market structure when both consumer and situational characteristics are explanatory

dimensional aggregate perceptual map. In this map, each of the 29 options is represented as a point in the seven-dimensional space cor- responding to its means on the seven factors. A vector of mean pref- erences for each of the 29 options was calculated for each cluster, and regressed against the seven benefit factors (i.e., 29 units of analysis, seven independent variables, one dependent variable-preference). As expected, preference was related to very different benefits for different clusters; an example is presented below. This procedure required the assumption of perceptual homogeneity across clusters. The r-squared values for the nine regressions were generally acceptable to quite high, between 0.26 and 0.70.

RESULTS OF THE EXAMPLE STUDY

Both situational and personal characteristics distinguished the seg- ments. For example, segment 8 is primarily distinguished by a high level of the occasion factor that represents waking up/eating a meal/ listening to the radio alone, the life-style factor that represents high- activity sports, higher-than-average incomes, and younger ages. Knowl- edge of each of these distinguishing characteristics would be useful in devising radio formats aimed at this segment.

Another example of one of the segments would be No. 9, called in- tellectuals seeking relaxation. The differentiating characteristics of this segment are both also person-based and situation-based. In terms of consumer characteristics they are somewhat more likely to be female, and have far higher education than average, higher-than-average in- comes, higher ages, and a greater likelihood of having older children. They are lower than average on the life-style factors that reflect low- activity entertainments and high-activity sports.

In terms of the situation, which occurs most frequently in the evening, the individual is often trying to concentrate or is conversing with an- other adult, or perhaps accomplishing paperwork (situation factor 9). At the time, they are often feeling more lonely and frustrated than average (situation factor 2), and more busy and harrassed (situation factor 8).

Not surprisingly, they prefer relaxing and calming benefits of en- tertainment options that rate highly on option factor 2, and to a lesser extent wish to avoid the lighter (uplifting, amusing) options which rate highly on option factor 7. These preferences can be seen in Figure 2, which maps the 29 entertainment options with respect to option factors 2 and 7. The positions of the options reflect the mean ratings across subjects on those factors. The vector plotted through the space is a best fit of the mean option positions with their mean preferences for this segment.

404 BALL ET AL.

Page 11: Segmentation and market structure when both consumer and situational characteristics are explanatory

Upli

fts

and

amuses 1

0.50

-

0.45.

0.40-

I

0.35-

0.30

-

0.25.

0, 55

{

0.2

0.

RADI

O *T

V AC

TION

SH

OW (

2.5

)

TALK

BACK

EA

LANI

/DIC

K (3.31'1

SPORTS

'RIC

HARD

SON(

3.0)

0.15-

0.10

-

0.05.

0.00-

HARRISON* I

(1.8

) 'T

V FILM(3.9)

(4.2)

*TV SPORTS

MINE

HAN(

4.2)

' 12.4) 1

TV QUIZ.

=RADIO INF

O PROC 'S

INCL

AIR(

3.6)

TV

IN

FO PROC'

(2.5)

RADIO'

(5.3)

(4.9)

FEATURE

'TV

PLAY(4.1)

(4.1)

'YA(

AMl(

4.9)

'HORSE

'TV

SOAP(2.4)

RACINC(l.4)

*YA (

PM

I 'TV

NEWS

(4

.0

)

(5

.5)

*TV

SITC

OM

(3.9)

'CHAT

(5.6)

'RADIO SI

TCOM

(3.8)

'ELLIS/DANIELS

(2.9)

'OWN

TALK

EACK

(4.7)

'COREETT

(3.2)

OWN

MUSI

C*

(5.7)

'NOVELS

& ST

ORIE

S (5.1)

'MAC'S

& PA

PERS

(5.3)

Fig

ure

2. P

erce

ptua

l map

for

clus

ter

9: o

ptio

ns w

ith

maj

or b

enef

its d

esir

ed. (

Ave

rage

pr

efer

ence

for

optio

n in

par

enth

eses

.)

Page 12: Segmentation and market structure when both consumer and situational characteristics are explanatory

MANAGERIAL IMPLEMENTATION

The substantive conclusions drawn from the study were concerned with identifying targetable segments (i.e., sufficiently large groups at each day part) and providing the benefits that best met the needs arising from their situations and personal characteristics.

Management at the client firm implemented a new strategy based substantially upon the results of this study. Results of an independent media-market audit six months after implementation showed an in- crease of 46% in listenership among the groups targeted. This reversed a trend of decline in market share, and management attributed the results primarily to the new strategy.

DISCUSSION

With respect to Kotler’s criteria of ease of measurement, substantiality, and accessibility for the segmentation aspects of the method, the pro- posed method seems sufficient. Measurement is relatively easy and follows well-established procedures. The segments found by the method are almost certain to be more substantial than those found by the four methods just mentioned. The reason for this is that the methods above, to incorporate both situational effects and consumer effects, must sep- arately segment occasions and consumers, or one within the other. This results in a profusion of segments, many of which are small and of dubious utility. If they are not used, the information contained in them is discarded.

The method proposed in this article, on the other hand, utilizes all the data except for outliers. It recovers just a few segments, usually each of substantial size, which are homogeneous with respect to pref- erence. Accessibility of the segments is not guaranteed, but one of the arguments in favor of occasion-based segmentation is that the segments thus found should be more reachable.

The proposed method seems to stack up well against the four criteria mentioned by Shocker et al. (1984) for a comprehensive method of mar- ket structure analysis. The first criterion is that the method have a means for specifying the appropriate set of products. Our preferred method is borrowed from Srivastiva et al.

The second criterion is a “means of dealing with the diverse require- ments of different usage occasions and contexts.” The motivation for the method was, in fact, to provide such a means. The method allows the data to suggest the extent to which usage occasion is determinative of choice, alone or in combination with consumer characteristics.

The third criterion is a “means for dealing with different degrees of appropriateness in use for the same product.” Recognizing that we have made an argument for the use of preference as opposed to appropriate-

406 BALL ET AL.

Page 13: Segmentation and market structure when both consumer and situational characteristics are explanatory

ness, our measure is interval scaled and allows for degrees of preference. Shocker et al. point out that ". . . a dishwashing liquid, an abrasive cleaner, and a laundry detergent are all cleaners. However, an abrasive cleaner may be regarded as a less appropriate substitute for laundry detergent than a dishwashing liquid would be." We argue that substi- tutability is as much a matter of preference as appropriateness, if not more. Two individuals may perceive two options to be equally appro- priate for a given context; however, one of them may have a strong preference for option A over option B, while the other has an equally strong preference in reverse. The use of an appropriateness measure would result in the two subjects-in-occasion being classified together, and an interpretation of consumer homogeneity with respect to occa- sions. The use of a preference measure would indicate that the two consumers do not belong in the same segment, and that the two products are not equally substitutable in their eyes.

The final criterion of Shocker et al. is validity and reliability, which we cannot measure directly, as previously mentioned. However, the management response and market results provide some indirect evi- dence of external validity.

CONCLUSION

The exploratory method presented here for investigating market struc- ture in the presence of situational effects on choice is relatively simple and relies on existing statistical tools. The unit of analysis used-the person-in-situation-has not before been proposed to address this prob- lem. The person-in-situation as the unit of analysis allows understand- ing of the situational constraints and goals affecting the individual, as well as his or her personal enduring characteristics that may affect choice. Combined with this unit of analysis is the use of measured preferences for a wide variety of options as the bases of segmentation. These preferences provide imputed information on needs, and are prox- imal to marketplace behavior. Persons-in-situations grouped together into a segment prefer the same sorts of options, regardless of the char- acteristics of their situations or their personal characteristics. Postseg- mentation analyses of the commonalities among the persons-in-situa- tions in a segment, and the differences among the segments, provide insight into the dynamics of choice in a segment, and suggest strategic actions. Unlike other proposed methods for dealing with occasion-based market structures, the method presented here permits segments to be defined as a mixture of consumer and occasion characteristics. No in- dependent taxonomy of situations or people is required, merely good measurement of the characteristics of consumers and occasions. The method produces occasion- and person-based segments that are easy to measure, generally more substantial than those found by other meth-

SEGMENTATION AND MARKET STRUCTURE 407

Page 14: Segmentation and market structure when both consumer and situational characteristics are explanatory

ods, and, at least in the example study, accessible. An example study using the method was performed and appears to have yielded actionable results from a managerial perspective.

REFERENCES

Baumeister, R. F., & Tice, D. M. (1985). Toward a theory of situational struc- ture. Environment and Behavior, 17(2), 147-192.

Belk, R. W. (1975). Situational variables and consumer behavior. Journal of Consumer Research, 2, 157-164.

Dickson, P. R. (1982). Person-situation: Segmentation’s missing link. Journal of Marketing, 46, 56-64.

Fraser, C., & Bradford, J. W. (1983). Competitive market structure analysis: Principal partitioning of revealed substitutibilities. JournaZ of Consumer Research, 10, 15-30.

Hustad, T. P., Mayer, C. S., & Whipple, T. W. (1975). Consideration of context differences in product evaluation and market segmentation. Journal of the Academy of Marketing Science, 3(1), 34-47.

Kotler, P. (1980). Marketing management. Englewood Cliffs, NJ: Prentice Hall. Miller, K. E., & Ginter, J. L. (1979). An investigation of situational variation

in brand choice behavior and attitude. Journal of Marketing Research, 16, 111-123.

Pessemier, E. A. (1982). Product management. New York: Wiley. Punj, G. and Stewart, D. (1983). Cluster analysis in Marketing research: review

and suggestions for application. Journal of Marketing Research, 20, 134- 148.

Shocker, A. D., Zahorik, A. J., & Stewart, D. W. (1984). Competitive market structure analysis: A comment on problems, Journal of Consumer Research, 11, 836-841.

Srivastiva, R. K., Alpert, M. I., & Shocker, A. D. (1984). A customer-oriented approach for determining market structures. Journal of Marketing, 48,32- 45.

Stefflre, V. (1971). New products and new enterprises: A report on an experiment in an applied social science, Irvine, CA: University of California.

Vanhonacker, W. R. (1980). An empirical model of hierarchical choice based on switching dynamics. In R. Bagozzi et al. (Eds.), A M A Educators Confer- ence Proceedings (pp. 398-401). Chicago: American Marketing Association.

Wind, Y. (1978). Issue and advances in segmentation research. Journal of Marketing Research, 15, 317-337.

Dwayne Ball is Assistant Professor of Marketing at the University of Ne- braska, 303 CBA, Lincoln, NE 68588-0492. Charles Lamb is Senior Lecturer, Lincoln University, Private Bag, Christchurch, New Zealand. Roderick Brodie is Chair, Department of Marketing and International Business, University of Auckland, Auckland, New Zealand.

408 BALL ET AL.