Académique Documents
Professionnel Documents
Culture Documents
Journal of
Marketing
33,1/2
38
Received August 1996
Revised August 1997
Eva M. Thelen
Department of Marketing, University of Innsbruck, Austria, and
Rik Pieters
Department of Business Administration, T ilburg University,
T he Netherlands
Keywords Benefit segmentation, Consumer behaviour, Image, Market segmentation,
Product attributes, Services marketing
Abstract Although the basic idea of benefit segmentation lies in using causal, as opposed to
descriptive, factors as segmentation criteria, most of the empirical studies do not differentiate
between product attributes and the benefit sought by consumers. T he objectives of this article are
to clarify the distinction between attributes and benefits sought, and to apply a modified laddering
technique, based on means-end theory to use the elicited benefits to form benefit segments. A
compar ison with attr ibute-based segments demonstrates that means-end chains provide a
powerful tool for true benefit segmentation.
Introduction
Haley (1968) and Wind (1973) proposed the segmentation of markets on the
basis of benefits sought by identifiable groups of consumers. While
psychographic and general attitudinal approaches to segmentation may work
well statistically they are less helpful when it comes to deriving effective
marketing strategies (Young et al., 1978). Therefore, benefit segmentation has
become the preferred technique for successful product positioning, new product
introduction, pricing, advertising, and distribution (Wind, 1978).
Most of the published segmentation studies follow a similar approach: after
analyzing secondary data and/or conducting one-on-one in-depth or focus
group interviews to identify relevant attributes and benefits a measure of
importance of attributes/benefits is developed (Haley, 1968; Moriatory and
Reibstein, 1986; Mhlbacher and Botschen, 1988). This instrument is pre-tested
and the data collection starts. Generally, responses are given on a scale
representing low to high importance and/or variability. Data analysis begins
with data treatment (e.g. normalization, standardization) and frequently
continues with factor and cluster analysis to identify benefit segments.
Means-end
structures
39
European
Journal of
Marketing
33,1/2
40
Gutman (1982) defines functional and practical benefits and emotional pay-off
as types of consequence. If we consider the two approaches and follow
Gutmans (1982) definition of benefits and consequences we propose the
following distinction between attributes and benefits sought by consumers.
Attributes are characteristics of products, services or behavior which may be
preferred or sought by consumers. These attributes do not explain per se for
what reasons the product, service or behavior is or might be bought. According
to means-end theory we can distinguish between two levels of attributes:
concrete and abstract attributes, e.g. price and good quality. In both cases
we receive additional information about the product, the service or the behavior
itself but we do not discover any underlying reasons why the product is chosen
and/or bought. Most of the companies prefer to work with attribute-based
segmentation which they consider more operational and more easy to act on
than on benefits sought.
Benefits which people are seeking in consuming a given product, service or
behavior offering various attributes explain why people are looking for certain
attributes. Again, according to means-end theory we can differentiate between
consequences on the functional level, e.g. I can easily handle it, and
consequences on the psychosocial level, e.g. others regard me as being special
(Olson, 1989; Olson and Reynolds, 1983; Peter and Olson, 1987). Both are linked
to attributes sought and personal values, which are divided into instrumental
and terminal values. From the customers point of view it is not the products
attributes which count, but the problem solution the benefit sought which
they derive from a certain combination of attributes.
We will use this definition of benefits when doing benefit segmentation. We
believe that much of the conducted empirical studies to determine benefit
segments do not focus on benefits as described above. Instead, they focus on
concrete and abstract attributes and sometimes on a combination of attributes
and benefits or consequences, respectively.
Let us give some examples. If during the purchasing process of data
terminals reliability (abstract attribute) is the sixth important attribute
(Moriatory and Reibstein, 1986) this does not equal the benefit sought by the
customer. Underlying benefits may be: no disturbance of data processing,
our investment pays back better, optimal customer support, or employees
are seldom annoyed. If customers of analgesic brands consider package size
(concrete attribute) important (OConnor and Sullivan, 1995), the benefits
derived from different package size can range from fits perfectly in my pocket
to easy to carry with me, and lasts for more than one month. If customers of
holiday destinations perceive sunny weather (concrete attribute) as an
important attribute (Mhlbacher and Botschen, 1988) they may desire sunny
weather because they are seeking benefits like get a wonderful suntan,
allows us to do open air sports, or children may play games the whole day
outside. Even concerning the example in Haleys (1968) landmark article
Benefit segmentation: a decision-oriented research tool, we would claim that
Means-end
structures
41
European
Journal of
Marketing
33,1/2
42
Levels of
MeansEnd
Chains
Example
Motor-Cycle
Triumph
Speed Triple
900
Concrete
Attributes
Abstract
Attributes
Functional
Consequences
Psycho-social
Consequences
Instrumental
Values
Terminal
Values
Price
Good
Quality
Can easily
handle it
Others see me
as special
Being center
of attention
Selfesteem
98 H.P.
Powerful
engine
Can perfectly
accelerate
Feel very
strong
Attribute based
Figure 1.
Segmentation types
based on means-end
chains
Type
of
Segmentation
Value based
Linkages
based
Means-end
structures
43
European
Journal of
Marketing
33,1/2
44
Sample
The population for the study was the 250 graduate students who specialized in
marketing or retailing. Students were asked to complete a questionnaire in
class.
Questionnaire
First, respondents were asked to describe in detail one of their recent clothing
purchases. They indicated when they had bought this clothing item, what the
clothing item was, why, for whom, and where they bought it, how often they
previously purchased at this shop, how much money and time they spent on it
and if they had clear expectations concerning the type of clothing item.
Respondents were asked to keep this clothing purchase in mind when
answering the specific laddering and rating questions.
Means-end
structures
45
European
Journal of
Marketing
33,1/2
46
Figure 2.
Identified attributes,
benefits/consequences
and values
Benefits/consequences
Friendly
Competent
Honest
Overview
Fun in shopping
Sufficient time
Advising
Distant
Helpful
Polite
Empathetic
Own decision
Buying urge
Customer loyalty
Quality of the
product
Control contact
moment
Values
Respect
Efficient shopping
Reduces uncertainty
Feeling good
Right clothing
Helps
decision
Feeling good
Customer
loyalty
Efficient
shopping
Quality of
the product
Own
decision
Respect
Buying
urge
Fun in
shopping
Right
clothing
Sufficient
time
Control
contact
moment
Helps
decision
Reduces
uncertainty
Overview
Advising
Empathetic
Polite
Figure 3.
HVM of the whole
sample
Helpful
Honest
Friendly
Competent
Distant
Hierarchical cluster analysis was used to identify market segments. It was first
applied to the attribute rating scale. Squared Euclidean distances were used as
proximity measure and Wards procedure for calculation of the similarity
between objects and clusters. Bergs study (1981) showed that compared to
other methods Wards procedure finds good partitions and assigns cases
correctly. Two clusters described in Table I could be found.
From both clusters all attributes were rated as more important or less
important. Cluster 2 rates most of the items significantly lower than cluster 1,
which means cluster 2 has higher expectations concerning characteristics and
behavior of sales-personnel than cluster 1. These results could be due to
individual answering tendencies, therefore, the analysis was redone with
standardized data and the two clusters could be found again. To prove the
selectivity of the identified clusters discriminant analysis was conducted. The
highest standardized canonical discriminant function coefficients were 0.68 for
the attribute gives expert advice, 0.55 for the attribute endeavoured and
0.50 for the attribute empathetic.
The sample was randomly split in half. One half of the sample was used to
calculate the discriminant function, and the other half was only used for
classification.
Attributes
Friendly even when store is crowded
Friendly although customer is indecisive
Friendly although customer is leaving
without buying
Friendly regardless of customers appearance
Endeavoured
Service available when needed
Reliable
Expert advice
Customer-oriented
Respectful
Honest
Knows sizes
Knows material
Fashionable outfit
Distant
Regards customers taste
Empathetic
Notes: * = < 0.1; ** = < 0.05; *** = < 0.01
Cluster 1
69 cases
Means
Cluster 2
154 cases
Means
1.90
1.51
1.32
1.55
1.24
1.25
***
1.42
2.68
2.72
1.77
3.09
1.55
1.74
2.43
2.87
2.23
2.67
1.58
1.87
2.07
1.16
1.51
2.05
1.50
1.88
1.21
1.43
1.44
2.80
2.30
2.73
1.65
1.53
1.42
***
Means-end
structures
47
Significance
***
***
***
***
***
***
***
***
***
***
Table I.
Market segments based
on attribute ratings
European
Journal of
Marketing
33,1/2
48
Attributes
Cluster 1
82 cases
Percentage of
cluster
members
Cluster 2
Cluster 3
Cluster 4
43 cases
59 cases
47 cases
Percentage of Percentage of Percentage of
cluster
cluster
cluster
members
members
members
Means-end
structures
Significance
Friendly
92
90
66
***
Competent
27
26
51
38
**
Expert advice
56
67
73
45
**
Distant
100
84
21
***
Helpful
28
12
96
***
Empathetic
15
21
27
17
Polite
28
12
40
***
Honest
27
33
48
13
***
= > 50 percent;
49
Table II.
Market segments
identified by clustering
attributes derived from
laddering analysis
European
Journal of
Marketing
33,1/2
Consequences
Cluster 1
58 cases
Percentage of
cluster
members
Sufficient time
53
36
28
25
***
Right clothing
47
68
47
27
***
Reduces uncertainty
24
64
28
12
***
Overview
21
55
19
12
***
10
29
23
**
41
91
12
***
38
27
12
12
***
Helps decision
10
46
26
***
Own decision
67
48
12
38
***
Buying urge
22
21
30
50
***
Efficient shopping
16
13
33
***
Respect
66
2
Table III.
17
25
Benefit segments based Customer loyalty
**
***
on meanings derived
Notes: * = < 0.1; = < 0.05;
= < 0.01;
from laddering analysis
= > 75 percent
20
***
12
***
50
Cluster 2
Cluster 3
Cluster 4
56 cases
57 cases
60 cases
Percentage of Percentage of Percentage of
cluster
cluster
cluster
members
members
members
Significance
Control contact
moment
Fun in shopping
Quality of the
product
= > 50 percent;
Cluster 1
Row total
Benefits segment
Cluster 1
Count
Row percentage
Col. percentage
24
41
29
11
19
26
12
21
20
11
19
23
58
25
Cluster 2
Count
Row percentage
Col. percentage
20
36
24
11
20
26
11
19
19
14
25
30
56
24
Cluster 3
Count
Row percentage
Col. percentage
14
25
17
8
14
18
22
39
37
13
22
28
57
25
Cluster 4
Count
Row percentage
Col. percentage
24
40
30
13
22
30
14
23
24
9
15
19
60
26
Column
Total
82
36
43
19
59
25
47
20
231
100
For all benefit segments there are two or three attributes that are desired by
most of their members.
These results support that the same attribute can lead to different benefits
sought and that a single benefit can be based on multiple attributes.
Benefit segments should be accessible to be useful for management. The
more variability segmentation shows in demographic or behavioral variables
the easier the segments can be accessed. To test the variability of true
benefit segments compared to attribute segments derived by laddering and
by attribute ratings, we related the three cluster solutions to two demographic
variables (sex and age) and five behavioral variables (the amount of former
purchases in the same shop, the amount of different shops visited for the
purchase, the amount of time and money spent for the purchase, existence of
special ideas before purchase). In Tables VII, VIII and IX significant
differences within the three cluster solutions are extracted. The true benefit
segments show more variability in some important variables than the
others.
The comparison of the two attribute clusters based on rating data revealed
only one significant difference in respect of the behavioral variable amount of
time spent for the purchase.
Means-end
structures
51
Table IV.
Comparison of segments
based on attributes
derived from laddering
and benefit segments
European
Journal of
Marketing
33,1/2
52
Cluster 1
Benefits segment
Cluster 1
Count
Row percentage
Col. percentage
20
36
29
35
64
23
55
25
Cluster 2
Count
Row percentage
Col. percentage
15
27
22
40
73
26
55
25
Cluster 3
Count
Row percentage
Col. percentage
19
35
28
36
66
23
55
25
15
26
22
43
74
28
58
26
69
31
154
69
223
100
Cluster 4
Count
Table V.
Row percentage
Comparison of segments
Col. percentage
based on attribute
Column
ratings and benefit
segments
Total
Attributes
Table VI.
Market segments
identified by clustering
attributes derived from
laddering analysis
Cluster 1
58 cases
Percentage of
cluster
members
Benefits segment
Cluster 2
Cluster 3
Cluster 4
56 cases
57 cases
60 cases
Percentage of Percentage of Percentage of
cluster
cluster
cluster
members
members
members
Significance
Friendly
74
64
74
63
Competent
40
38
44
20
**
Expert advice
64
64
70
43
**
Distant
62
63
40
65
**
Helpful
36
39
35
23
Empathetic
21
23
18
17
Polite
24
19
25
Honest
31
32
30
23
= > 50 percent;
**
= > 75 percent
The four-cluster solution at the attribute level derived from laddering data
shows two significant differences in the behavioral variables existence of
special ideas before purchase and money spent for the purchase.
At the true benefit level the variability in important variables increases.
The two-behavioral variables the amount of former purchases in the same
shop, the amount of time spent for the purchase and the demographic
variable age significantly differ between the four clusters.
Variables
Cluster 1
Basis: 64 cases
Less demanding
customers
Mean
Cluster 2
Basis: 150 cases
Demanding
customers
Mean
Significance
2.60
2.97
***
Time spent
Note:
***
= < 0.1
Variables
Special
ideas
Money
spent
Cluster 1
Basis: 74 cases
Mean
Cluster 2
Basis: 41 cases
Mean
2.27
2.23
1.79
1,358
1,454
1,873
1,339
(1 <> 4)***
(2 <> 4)*
(3 <> 4)*
(1 <> 3)**
Variables
Former
purchases
Time spent
Age
Cluster 2
Correct clothing
Basis: 52 cases
Mean
3.18
2.89
22.09
3.67
3.09
22.95
53
Table VII.
Comparison of behavior
and demographic
variables between
attribute clusters derived
from ratings
Cluster 3
Cluster 4
Basis: 53 cases Basis: 46 cases
Mean
Mean
Significance
2.48
Cluster 1
Control/relax
Basis: 54 cases
Mean
Means-end
structures
Table VIII.
Comparison of behavior
and demographic
variables between
attribute clusters derived
from laddering
Cluster 3
Cluster 4
Fun
Purchasing
Basis: 54 cases Basis: 54 cases
Mean
Mean
Significance
3.31
2.82
23.26
3.56
2.65
22.52
(1 <> 2)**
(2 <> 3)**
(1 <> 4)***
Table IX.
Comparison of behavior
and demographic
variables between
benefit clusters derived
from laddering
European
Journal of
Marketing
33,1/2
54
Managers can use this information when training and recruiting their
personnel. A basic tenet of any sales training seminar is understanding the
customer. This is because salespeople must manage dynamic communication
processes. Rather than reading a script that remains constant for all, a
salesperson must tailor an approach to each customer, listening to those who
need to be heard and explaining details to customers who need information
(Gengler et al., 1995; Spiro and Weitz, 1990). Indeed, Weitz (1978) found that a
salespersons perceptual ability, the ability to understand a customer and the
customers decision making process, is related to sales performance. The usage
of information of identified true benefit segments in training could improve
efficiency and effectiveness. Under an innovative perspective, sales employees
might identify additional characteristics and kinds of behavior to satisfy
identified benefits sought by different segments.
However, the proposed benefit segmentation approach could be applied to
many other aspects of services industry, e.g. how do customers differ according
to their benefits sought in respect to atmosphere, the core service, and
additional services?
Limitations and future research
One limitation of our study concerns the paper-and-pencil laddering (hard
laddering). The primary advantage of this method is the relative efficiency with
which data can be collected. However, little is known about the validity and
reliability of the procedure and the comparability of results obtained from
traditional laddering interview (soft laddering) and paper-and-pencil laddering.
For example, the many empty boxes that appear in the questionnaire may put
strong demands on subjects to list consequences and goals that are of little
importance to them (Pieters et al., 1994). It would be interesting to see research
comparing the results of hard and soft laddering. Should a test of convergent
validity establish that both techniques lead to similar results, one could safely
conclude that hard laddering is a preferable technique, since it is easier to
administer and less costly. Differing results would call for an investigation of
predictive validity in a larger context (Grunert et al., 1995).
Another limitation concerns the free elicitation technique. So far we do not
know if the application of other techniques, e.g. triadic sorting, free sorting or
attribute selection, would produce similar results. This raises two questions:
(1) Will the set of attributes finally selected as the starting point of the
ladders differ depending on which elicitation method is used?
(2) If yes: which set of attributes is the right one?
Although in this paper we tried to further elaborate the distinction between
attributes and benefits sought in practice, many borderline cases may still turn
up. Categorization in a uniform way is heavily dependent on the availability of
context information. Hard laddering usually provides very little context
information.
Means-end
structures
55
European
Journal of
Marketing
33,1/2
56
Grunert, K.G., Grunert, S.C. and Sorensen, E. (1995), Means-End Chains and Laddering: A n
Inventory of Problems and an Agenda for Research, Working paper No. 34, November, Center
for Market Surveillance, Aarhus, Denmark.
Gutman, J. (1982), A means-end model based on consumer categorization processes, Journal of
Marketing, Vol. 46, pp. 60-72.
Gutman, J. (1991), Exploring the nature of linkages between consequences and values, Journal
of Business Research, Vol. 22, pp. 143-8.
Haley, R.I. (1968), Benefit segmentation. A decision-oriented research tool, Journal of Marketing,
Vol. 32, pp. 30-5.
Hensel, J.S. (1990), Service improvement and control. A customer-based approach, Journal of
Business Research, Vol. 21, pp. 22, 43-54.
Heskett, J.L., Sasser, W.E. and Hart, C.W.L. (1990), Service Breakthroughs Changing the Rules of
the Game, The Free Press, New York, NY.
Johnson, M.D. (1989), On the nature of product attributes and attribute relationship, in Srull,
T.K. (Ed.), Advances in Consumer Research, Vol. 16, pp. 598-604.
Kotler, P. (1991), Marketing Management, 7th ed., Prentice-Hall, Englewood Cliffs, NJ.
Mairamhof, G. (1996), Handling customer disconfirmations in Kleinaltenkamp, M., Engelhardt,
W.H., Meyer, A., Mhlbacher, H. and Stauss, B. (Eds), Focus Dienstleistungsmarketing, Gabler
Edition Wissenschaft, DUV, Wiesbaden.
Mathews, B.P. and Watt, A.W. (1986), Benefits segmentation for the National Girobank,
International Journal of Bank Marketing, Vol. 4 No. 4, pp. 39-51.
Meyer, A. and. Mattmller, R. (1987), Qualitt von dienstleistungen. Entwurf eines
praxisorientierten qualittsmodells, Marketing-Zeitschrift fr Forschung und Praxis, Vol. 9,
August, pp. 187-95.
Moriatory, R.T. and Reibstein, D.J. (1986), Benefit segmentation in industrial markets, Journal of
Business Research, Vol. 14, pp. 463-86.
Mhlbacher, H. and Botschen, G. (1988), Trade-off analysis for holiday packages, Journal of
Business Research, Vol. 17, pp. 117-31.
Mhlbacher, H. and Botschen, G. (1990), Benefit-segmentierung in Dienstleistungs-mrkten,
Marketing-Zeitschrift fr Forschung und Praxis, Vol. 3, pp. 159-68.
Myers, J.H. and Alpert, M.I. (1968), Determinant buying attitudes: meaning and measurement,
Journal of Marketing, Vol. 32, pp. 13-20.
OConnor, P.J. and Sullivan, G.L. (1995), Market segmentation: a comparison of benefits/
attributes desired and brand preference, Psychology and Marketing, Vol. 12, pp. 613-35.
Olson, J.C. and Reynolds, T.J. (1983), Understanding consumers cognitive structures:
implications for advertising strategy, in Percy, L. and Woodside, A. (Eds), Advertising and
Consumer Psychology, Lexington Books, Lexington, MA.
Olson, P. (1989), Theoretical foundations of means-end chains, Werbeforschung & Prax is,
Folge 5, pp. 174-8.
Payne, A. (1993), T he Essence of Services in Marketing, Prentice-Hall, Hemel Hempstead.
Peter, P. and Olson, J. (1987), Consumer Behaviour. Marketing S trategy Perspectives, Richard
Irwin, Inc.
Pieters, R. (1993), A Control View on the Behavior of Consumers: Turning the Triangle, in Van
Raaij, W.F. and Bamossy, G.J. (Eds), E uropean A dvances in Consumer Research , Vol. 1,
pp. 507-12.
Pieters, R., Baumgartner, H. and Allen, D. (1995), A means-end chain approach to consumer goal
structures, International Journal of Research in Marketing, Vol. 12, pp. 227-44.
Pieters, R., Baumgartner, H. and Stad, G.J. (1994), Diagnosing means-end structures: the
perception of wordprocessing software and the adaptive-innovative personality of managers,
Means-end
structures
57
European
Journal of
Marketing
33,1/2
58
in Bloemer, J., Lemmink, J. and Kaspar, H. (Eds), Proceedings 23rd E mac conference,
Maastricht, pp. 749-63.
Reynolds, T.J. and Gutman, J. (1988), Laddering theory, method, analysis, and interpretation,
Journal of Advertising Research , Vol. 28, pp. 11-31.
Roehrich, G. and Valette-Florence, P. (Ed.) (1991), A weighted cluster-based analysis of direct and
indirect connections in means-end chains: an application to lingerie retail, in Workshop on
Values and Lifestyle Research in Marketing, EIASM, Brussels.
Rust, T.R. and Cooil, B. (1994), Reliability measures for qualitative data: theory and
implications, Journal of Marketing Research, Vol. 31, February, pp. 1-14.
Spiro, R.L. and Weitz, B.A. (1990), Adaptive selling: conceptualization, measurement, and
nomological validity, Journal of Marketing, Vol. 27, pp. 61-9.
Walker, B.A. and Olson, J.C. (1991), Means-end chains: connecting products with self, Journal of
Business Research, Vol. 22, pp. 111-18.
Weitz, B.A. (1978), The relationship between sales performance and understanding of customer
decision making, Journal of Marketing Research, Vol. 15, pp. 501-16.
Wilkie, W.L. and Weinreich, R.P. (1972), Effects on the Number and Type of Attributes Included in
an A ttitude Model: More Is Not Better , Working Paper No. 385, Purdue University, West
Lafayette, IN.
Wind, Y. (1973), A new procedure for concept evaluation, Journal of Marketing, Vol. 37, pp. 2-11.
Wind, Y. (1978), Issues and advances in segmentation research, Journal of Marketing Research,
Vol. 15, pp. 317-37.
Young, S. and Feigin, B. (1975), Using the benefit chain for improved strategy formulation,
Journal of Marketing, Vol. 39, July, pp. 72-4.
Young, S., Ott, L. and Feigin, B. (1978), Some practical considerations in market segmentation,
Journal of Marketing Research, Vol. 15, pp. 105-412.