Quantifying consumers' motivational structures for

food products
Wilma den Hoed and Gerda I. J. Feunekes
Source: ESOMAR, Business in Asia Pacific, Bangkok, November 2000
Downloaded from WARC

This paper describes methodological issues concerning the Association Pattern Technique, a method
to quantify means-end chain structures. The method provides understanding in the way consumers are
motivated to choose specific products. The main focus is on the experience and difficulties
encountered, in using this method in Europe and China. The topics addressed include development of
the questionnaire; analysis of the data; application for the business; and challenges ahead.

The Association Pattern Technique

Gerda I. J. Feunekes
Unilever Research Vlaardingen
and
Wilma den Hoed
Unilever Research Vlaardingen

INTRODUCTION
In order to get a quantitative description of consumers' motivational structures for specific foods, the newly
developed method Association Pattern Technique (APT) has been applied. This method provides understanding
in the way that consumers are motivated to choose specific products. It can be applied to give direction to both,
product development and marketing. This method has been developed and used for foods in Europe (Ter
Hofstede et al. 1998, Ter Hofstede et al. 1999).

We have used this method to obtain an understanding of specific foods, both in Europe and China. The process
followed consisted of two steps:
 1. Qualitative phase. Laddering interviews were carried out to generate product attributes, benefits, and
values. A selection of these formed the basis of the quantitative study.
2. Quantitative phase. The strength of all possible associations between attributes, benefits and values were
assessed and cross-country consumer segments were identified.

In this paper, the focus is on the methodological aspects of this technique. The emphasis is on the knowledge
gained, the choices made, and the challenges ahead in working with the Association Pattern Technique.

MEANS-END CHAIN THEORY

The APT studies are based on the Means-End Chain (MEC) theory. In the MEC theory, products are seen as a
means through which consumers obtain certain valued ends. It recognises that consumers do not want products
for their own sake, but for what the product can do for, or provide to them. MEC theory is concerned with the
content and structure of consumer knowledge, and the motivating effects of this knowledge. Gutman (1982)
introduced the MEC theory to marketing and consumer research. He positioned MEC theory as a model to
understand consumption relevant cognitive structures of consumers, i.e. the way consumption-relevant
knowledge is stored and organised in human memory.

The conceptual model of MEC theory specifies that the consumers' subjective knowledge about product
categories is stored in associative networks that are organised as means-end chains. The key constructs in
these networks are attributes (A), benefits (B) and values (V). Attributes are properties of products that can be
sensed, i.e., products have a particular form, colour, smell, taste, sound, and so forth. Benefits refer to what the
product is perceived as doing for or providing to the consumer. Values are intangible, higher-order outcomes or
end-states of the consumer.

According to Gutman (1982), a means-end chain is a concept that seeks to explain how a product facilitates the
achievement of desired end states. He assumes that consumers choose products with attributes that can
produce desired benefits any direct or indirect positive result for the consumer and minimise undesired
consequences. Similarly for benefits, they should help the consumer to achieve desired end states (values).
There are two types of links in means-end chains:

links between attributes and benefits;

links between benefits and values.

Each of the key constructs may be split in two levels to allow a more detailed analysis of consumer knowledge
structures: concrete and abstract attributes, functional and psychosocial benefits, and instrumental and terminal
values (Olson and Reynolds, 1983). Concrete attributes represent tangible, physical characteristics of a product,
while abstract attributes represent intangible, subjective characteristics. Functional benefits are directly
experienced, tangible outcomes of product use, while psychosocial benefits are more personal and less tangible
outcomes. Instrumental values are the cognitive representations of preferred modes of conduct or behaviour.
Terminal values, on the other hand, represent preferred end-states of being (Peter and Olson, 1993).

As indicated above, attributes, benefits and values are hierarchically structured because attributes lead to
benefits which, in turn, produce value satisfaction. In moving from the attributes upwards in the attribute-benefit-
value chain (means-end chain), the level of abstraction of the constructs increases. Walker and Olson (1991)
have suggested that in the six-level means-end chain, the three lower levels (concrete attributes, abstract
attributes and functional benefits) comprise the product-knowledge of consumers, while the three higher levels
(psychosocial benefits, instrumental values and terminal values) comprise the self-knowledge of consumers.
This suggested distinction between the product and consumer is shown in figure 1.

An example of a means-end chain expressed by a certain consumer, talking about a specific type of sauce is:
'This sauce has a dark colour. This means that this sauce will have a strong taste. This will make the dish taste
better so that everyone will enjoy the meal. It will make me feel valued as a mother and I will feel happy'. This
example represents one individual means-end chain of a consumer. This individual means-end sequence is
called a 'ladder' as shown in figure 2.

In this sequence, one moves farther away from the food product and closer to the consumer, thus linking the
food product to the consumer. The theory thus supposes that this particular consumer will choose sauce with a
dark colour for the reasons he/she mentioned. A particular consumer may have several of these means-end
chains which will work together to drive this consumers' choice. The individual ladders can be interrelated
through analysis of the laddering results, and jointly they will form an associative network. Such a network of
interrelated means-end sequences based on laddering with several consumers is called a Hierarchical Value
Map (HVM). The HVM comprises common sequences, stemming from the combination of individual ladders. As
such, the HVM consists of multiple means-end 'chains' based on multiple 'ladders' as shown in figure 3.

If we want to understand how consumers perceive our products, how products fit in their lives and how they are
motivated (de-motivated) to buy products, it is important to understand the MEC structure consumers have for a
particular product.

Two methods have been proposed in the literature to reveal the MEC structure, i.e. Laddering (Reynolds and
Gutman 1988) and the Association Pattern Technique (APT) (Ter Hofstede et al. 1998).

MEC METHODS: LADDERING AND APT

Laddering

Laddering helps us in the process of understanding the needs and aspirations of the consumers towards a
specific product. This qualitative technique is based on individual interviews, linked to a semi-quantitative
analysis to uncover consumers' means-end chain structures. The technique involves an interviewing format
which is based on a series of probes, typified by the 'why is that important to you?' question to determine sets of
links between important product attributes, consequences (or benefits) and values.

Laddering is not suitable as an instrument in large representative samples. It is rather time consuming and
requires well-trained and experienced interviewers.

APT
The Association Pattern Technique has been designed as a quantitatively oriented technique to assess means-
end chains. APT is inspired by Gutman (1982). Gutman proposed that, for measurement purposes, the means-
end chain could be conceived as a series of connected matrices. In APT, an AB matrix (attribute-benefit matrix)
and a BV matrix (benefit-value matrix) are distinguished. In the AB matrix, the a priori defined attributes and
benefits are listed in the columns and rows respectively, resulting in a table of all combinations of attributes and
benefits. Similarly, the BV matrix includes all possible combinations of the benefits and values. For each column
in the AB matrix (BV-matrix), respondents indicate to which benefits (values) that attribute (benefit) is perceived
to lead. Figure 4 shows an example of a part of an Attribute/Benefit matrix with the specific question as it was
used in the questionnaire.

Consumers were instructed to read the question and to indicate whether or not a specific attribute is associated
with a specific benefit by ticking the box.

Examples of how the question was asked:

Using a sauce, which 'has a fruity aroma', means to me 'cooking an appetising dish'
Using a sauce which 'is hygienically processed' means to me 'keeping the family healthy'.

Every respondent will be considering every combination of AB and BV. This results in a data set of binary
observations. Note that in contrast to laddering, the attributes, benefits and values are to be provided by the
researcher.

A laddering study for a small number of consumers (e.g. 20-30) in a particular country could be the basis for
building APT matrices.

APT was validated by Ter Hofstede et al. (1998) and proved a solid method to measure means-end chains; it
yielded similar motivational structures.

Figure 5 shows an overview of characteristics of laddering and APT.

EXPERIENCES WITH THE QUANTIFICATION OF MEANS-END CHAINS

BY APT

Laddering

Although the focus of this paper is on the quantification, we would like to give a few examples of what was
encountered in the qualitative phase, the laddering.

Asking personal questions in different cultures. We have kept the interview procedure as much as possible
the same in all countries. No major difficulties were encountered with asking this 'why' type of questions all
the time. It is crucial though that, due to the direct character of the questions, the atmosphere has to be
informal and relaxed. In practice, performing laddering in both Europe and Asia (Thailand and China) taught
us that the introduction phase needed to be longer in Asia, as the Asian respondents required a more
elaborate familiarisation with the interviewer.
Laddering on products that are used as ingredients versus product that are consumed as such. Some of
the products of our focus were mostly used as ingredients in dishes, which made it more difficult for the
consumer to tell about the product per se. Focussing on the product in context of the preparation and
consumption of the dishes in which it was used, to take it closer to the real motivations.
Grouping of attributes, benefits and values. In the analysis phase of laddering, different expressions of a
similar concept are combined to come to the Hierarchical Value Map. For example, benefits like 'better for
my health', 'healthy', and 'good for my body' were combined as 'healthy'. This grouping can give serious
interviewer/analyser bias. Different grouping of attributes, consequences and values will affect the strength
of the links found, therefore a similar level of detail should be applied for attributes, benefits, and values.

APT

APT - Development of the questionnaire

Selecting attributes, benefits, and values. The core of an APT questionnaire consists of two matrices:
the Attribute-Benefit and the Benefit-Value matrix. With these matrices consumers indicate which AB
and which BV links they recognise for themselves. Since in the laddering a great amount of attributes,
benefits and values are generated a selection has to be made for the quantification the links. Not all
attributes, benefits, and values from the qualitative phase can be included, due to restrictions of the
amount of questions that can be handled by consumers.

In our studies a list of 25 attributes, 15 benefits and 10 values resulted in one matrix of 25*15 and one
of 15*10, giving a total of 525 associations to be checked for a consumer.

We selected on forehand what type of information would be most relevant. For our objectives the
elements of main ladders of the qualitative study were a good starting point. Other existing consumer
data can be used as well.
Wording of values. As values are generic across different domains, the values that were derived from
laddering were compared with the cross-culturally validated 'List of Values (LOV)' from Kahle et al.
(1986) and Schwartz and Bilsky's value inventory (Schwartz and Bilsky, 1990). This check served two
purposes: firstly using the same concepts as used in/validated by the academic world and secondly, to
be able to add values that did not come out of the laddering, but that could be relevant for the
development of our business.
Laddering input reflects current status. The attributes, benefits and values derived from laddering are
based on the current perception of consumers and do not provide information on what could be
important in the future. Therefore it can be considered to include attributes, benefits, and values that
are expected to become important in the future. For example, we have gathered additional information
by including a value like 'Growing with the times'.
All attributes, benefits, and values were formulated positively. The positive formulation was chosen in
order to avoid that consumers had to switch in their thinking all the time and therefore would make
mistakes. So, the benefit or consequence 'makes you fat' which was generated in the laddering, was
formulated in the APT questionnaire as 'better for your figure'.
We experienced that the danger of turning negative attributes, benefits and values into positive statements
is that concepts that were converted appeared to be associated less often than expected. For example, in
the qualitative phase the benefit 'makes me fat' was associated with 'feeling guilty'. In the quantitative phase
the benefit 'better for my figure' was not associated with 'no feelings of guilt'. We therefore recommend
using a separate matrix for negative associations, to be able to stay as close as possible to the original
consumer language.
Cross-country issues

One questionnaire for all countries. Combining qualitative data from different countries to work with one
APT questionnaire in these countries required a lot of effort. To come to one list of attributes, benefits, and
values, we have carried out consecutively: an assessment of overlap of concepts across countries; double-
check whether attributes, benefits, values really meant the same in different countries; and finally, a
prioritisation and selection.

A questionnaire developed on the basis of qualitative work in several countries broadened the scope of the
study; attributes, benefits, and values that were new to a country were used to indicate opportunities.

Translation. Translation of the attributes, benefits, and values had to be performed very carefully because
the quality of the results depends on the immediate recognition of the meaning of these concepts by the
consumers. For the European study we used one Dutch 'mother list' which had more elaborate descriptions
of what each attribute, benefit, and value meant than the few words that could be used in the
questionnaire. From these lists translations started. Back translation was used to pick up remaining issues.

Importance of attributes, benefits, and values. As associations between attributes, benefits, and values,
might not give the full quantitative picture, the importance of attributes, benefits, and values was assessed
separately. Ranking according to importance was preferred above rating; however, above 10 to 12 items
ranking becomes difficult for the consumer. Therefore, the importance of the 10 values was measured by
asking respondents to rank the values by importance, whereas the importance of the attributes and
benefits was rated on a category scale.

APT - ANALYSIS
Getting a clear overview of the associations made by consumers, and incorporating the importance that
they attached to a concept was rather complicated. We have developed a structural approach that currently
consists of the following steps:

Select strong associations (absolute and relative). The ratio of the respondents scoring an association
was assumed to represent the strength of an association. We learned that taking into account all links
scored by more than 40% of the respondents would usually result in a comprehensive overview. To
create an overview of the most important links, taking into account links scored by more than 60% of
the respondents is more appropriate.

The links that are strong in an absolute sense are clearly worthwhile, but a lot of insights were derived
from links that were only relatively strong. A link made by 40% for a specific group (country, or
segment) has a totally different meaning when a similar percentage is found for other groups, or, when
this link is non-existent in other groups. A Hierarchical Value Map can be used to give the overview of
associations and (relative) strengths in one view. Figure 6 shows an example of a Hierarchical Value
Map after quantification. Due to the confidentiality of the results the specifications of attributes,
benefits, and values, have been left out.
 Look for concepts that are linked to a large number of concepts. Sometimes an attribute, benefit, or
value, has associations with many other concepts, but none of them is very strong, and none of them
extremely important. Those concepts appeared to be very valuable to understand the core of the
product field.
Check importance scores (absolute and relative). Combining the importance scores with the strength
of links resulted in additional information. The attributes, benefits and values with low scores can
become of interest when they do have strong linkages. Similarly, a very important value with weak
links may give an interesting opportunity to build this link in the mind of the consumers.
Assess Consumer Segmentation. Latent class segmentation on the data set was carried out with the
use of the segmentation model (MECSEG - Means End Chain SEGmentation) developed by Ter
Hofstede. The input for the segmentation are the Attribute-Benefit and Benefit-Value links made by
individual respondents, while the output consists of groups of consumers with similar association
patterns.

Groups of consumers with similar patterns of means-end chains with respect to sauces could be
identified. The central association, which differentiated this group from other consumers, was the
strong association between 'product is low in fat' and 'better for my figure'. Figure 7 shows an example
of a discriminating link in a specific pattern of one segment.

In general, the consumers of the segments were only slightly different in behaviour. This might either
indicate weak predictability, but could be interpreted also as a lack of products that fulfil consumer
needs adequately. Prediction of behaviour from these structures needs further study.

APPLICATIONS FOR THE BUSINESS

Understanding of segments. Motivational structures of consumer segments enable us to define target
groups and optimise product development and advertising for this group by building or strengthening
ladders.
Position of competitors by using the HVM. The HVM is a mind map related to a specific product or
brand. In some of our studies we included associations between brands, benefits and values.
Alternatively, the structure can be used to compare brand positioning. Figure 8 shows an example of
a Hierarchical Value Map in which the 'claimed' areas are shown, which results in a clear overview of
the current positioning and the possibilities for positioning.
Initiating product development. The results of an APT study can be initiating product development and
improvement. Attributes or benefits that appear to have potential, can be studied in more depth, to
understand how they can be improved.

CHALLENGES AHEAD FOR APT

Data collection and respondent burden

In Europe the APT data were collected by means of a self-completion postal questionnaire. Consumer
panels were used, which yielded good response rates.
Filling in the matrices can be a difficult task for some consumers. Several questionnaires were not filled in
correctly, and had to be removed; consumers who had linked almost everything (> 80%) in the matrix;
those who had linked almost nothing (< 2%); those who had skipped full pages. Altogether, this resulted in
a 10-15% decreased amount of acceptable questionnaires.

In China an individual interview was suggested by the agency. In this interview, filling out the matrices was
made easier for the respondent by using cards. A consumer was guided through a list of attributes, placed
as cards on the table, and could pick for each attribute the cards of benefits that were associated. After
that, benefits were associated to values in the same way.

Using computer-aided questionnaires could further simplify the task for the respondent.

Reduction in levels of information from qualitative to quantitative phase

In contrast to laddering, APT does not discriminate between concrete and abstract attributes, between
functional of psychosocial consequences, or between instrumental values and terminal values. The issue is
whether AA, BB, and VV links are crucial for a good understanding of the motivational structure.

By checking the European and Chinese data it appeared that there are a considerable amount of linkages
between concrete attributes and psychosocial benefits and between functional benefits and terminal
values, which are the most distant concepts that can be linked. So in general, consumers had no severe
problems with linking these concepts. On the other hand, in the laddering study in China several values
appeared to lead to one 'end value', which was 'Peace of Mind'. In the quantification, consumers did not link
benefits to the end value 'Peace of Mind' often; this value was probably considered to be to very far from
product benefits. Including AA, BB, and VV matrices can form an alternative, but this means a hugely
increase of respondent burden.

Learning for us was here that a good qualitative study is of utmost importance, to improve the interpretation
of the quantitative results. Another learning is that the choice of the level of the Attributes, Benefits and
Values is crucial: especially, values should not become too abstract to be linked to benefits.

Assessment of importance

'Claimed' importance in a questionnaire is likely to differ from real importance e.g., during cooking or eating.
From other studies we learned that benefits like 'hygienic' yielded higher scores in questionnaires than in
product tests. Better measures to assess importance of these types of concepts are needed.

CONCLUSIONS
Association Pattern Technique is an interesting method that gave us insight into motivational structures.
The results can be used in several ways: to understand current motivations in a product category; to
generate hypotheses for new products; and to guide advertising. There is space for further development in
the quantification of Means-End Chains, especially with regard to: coping with the limited number of levels
that can be included; combining importance and association data; and making data collection easier.
ACKNOWLEDGEMENTS
We would like to thank Anke Audenaert for her contribution to the work described in this paper.

REFERENCES

Gutman, Jonathan (1982). A Means-End Chain Model based on Consumer Categorization Processes.
Journal of Marketing, 46 (2), 60-72.

Kahle, Lynn R., Beatty, Sharon E., Homer Pamela (1986). Alternative measurement approaches to
consumer values: The List of Values (LOV) and Values and Life Style (VALS). Journal of Consumer
Research, Vol. 13 405-409.

Olson, Jerry C. and Thomas J. Reynolds (1983). Understanding Consumers' Cognitive Structures:
Implications for Marketing Strategy. In: Larry Percy and Arch G. Woodside (Eds.), Advertising and
Consumer Psychology. Lexington (MA): Lexington Books, 77-90.

Peter, J. Paul and Jerry C. Olson (1993). Consumer Behaviour: Marketing Strategy Perspectives.
Homewood (IL): Richard Irwin, 3d ed.

Reynolds, Thomas J. and Jonathan Gutman (1988). Laddering Theory, Method, Analysis, and
Interpretation. Journal of Advertising Research, 28:11-31.

Schwartz, Shalom H., Bilsky, Wolfgang. (1990). Toward a theory of the universal content and structure of
values: Extensions and cross-cultural replications. Journal of Personality and Social Psychology, Vol.58,
no.5, 878-892.

Ter Hofstede, F., Audenaert, A., Steenkamp, J.B.E.M. and Wedel, M. (1998). An Investigation into the
Association Pattern Technique as a Quantitative Approach to Measuring Means-End Chains. International
Journal of Research in Marketing, 15: 37-50.

Ter Hofstede, F., Steenkamp, J.B.E.M. and Wedel, M. (1999), 'International Market Segmentation Based
on Consumer-Product Relations', Journal of Marketing Research, 36: 1-17.

Walker, Beth A. and Olson, Jerry C. (1991). Means-End Chains: Connecting Products with Self. Journal of
Business Research, 22, 111-118.

NOTES & EXHIBITS

© Copyright ESOMAR 2000
ESOMAR
Eurocenter 2, 11th floor, Barbara Strozzilaan 384, 1083 HN Amsterdam, The Netherlands
Tel: +31 20 664 2141, Fax: +31 20 664 2922

www.warc.com

All rights reserved including database rights. This electronic file is for the personal use of authorised users based at the subscribing
company's office location. It may not be reproduced, posted on intranets, extranets or the internet, e-mailed, archived or shared electronically
either within the purchaser's organisation or externally without express written permission from Warc.