Master thesis
Measuring brand association strength for different brand types:
Applying free association to measure the number of associations
and attached response time latencies
Author: F.J.J. van Eysinga (10674829) Faculty of Economics and Business
MSc. Business Studies Specialization: Marketing
August 20th, 2014
Under supervision of: drs. J. Labadie MBM. Second assessor: drs. R.E.W. Pruppers
Preface
Amsterdam, August 8th, 2014
By writing the master thesis for the Master Business Studies – Marketing, I had the opportunity to take my knowledge to practice. This research concerns brand association strength, which proved to be a complicated topic, which made it challenging yet interesting to work on.
In order to conduct the experiment a measurement tool needed to be created able to record associations and response times attached to them. To do this, I contacted the faculty of Psychology who suggested a software program called Inquisit. Since programming language was new to me it initially seemed very challenging to learn in such a short time. However after extensive research it proved possible to write the script within the program, which provided the measurement tool.
I would like to thank drs. J. Labadie MBM. for supervising my thesis with great enthusiasm and helping me put my ideas into practice. Also I would like to thank drs. R.E.W. Pruppers for providing additional support. Finally, I would like to thank all respondents who took the effort of participating during the pre-tests and experiment.
With kind regards,
Abstract
This research aimed at measuring the strength of brand associations by looking at the influence of brand concept (i.e. functional vs. symbolic value orientation) and brand breadth (broad vs. narrow product range orientation) on brand association strength. Together concept and breadth were referred to as brand type. To test the influences of different brand types on brand association strength, the number of associations mentioned by respondents and the response times attached to these associations were analyzed for different types of brands. Furthermore, the influence of product category was included in the analysis, since it was hypothesized that durable consumer goods (represented by Clothing) have a different influence than non-durable consumer goods (represented by Cosmetics) on brand association strength. Respondents were shown two brands of different categories for which they were asked to provide their associations. Following, these associations were coded as either functional (e.g. user friendly) or symbolic (e.g. status symbol). The main contributing aspect of this research was the measurement of response time for each association mentioned. The results revealed brand concept has an influence on the number of associations, however, not on response times. Brand breadth on the other hand has no influence on number of associations, however it does influence response times. Furthermore, an interaction effect between concept and breadth was found concerning the number of functional associations, and an interaction effect between concept and category was found concerning the number of symbolic associations. It was concluded that brand type has an influence on brand association strength, where response time and number of associations revealed different information of added value. Therefore both should be used when measuring brand association strength.
Keywords: Brand association strength, brand concept, brand breadth, response time latency, number of associations, functional associations, symbolic associations, product category.
Content
1. Introduction ... 1
1.1. Background ... 1
1.1.1. Brand association strength ... 1
1.1.2. Measures for brand association strength ... 2
1.2. Problem definition ... 3
1.2.1. Problem statement ... 3
1.2.2. Sub-questions ... 3
1.3. Delimitations of the study ... 4
1.4. Contribution of the study ... 4
1.4.1. Theoretical contributions ... 4
1.4.2. Managerial contributions ... 4
1.5. Outline... 5
2. Brand associations, association networks and strength, and methods of elicitation ... 6
2.1. Brand associations explained ... 6
2.2. Brand association networks and brand association strength ... 8
2.3. Eliciting brand association networks and brand association strength ... 8
2.3.1. Zaltman’s metaphor elicitation technique ... 9
2.3.2. Brand concept mapping ... 9
2.3.3. Brand association network value ... 11
2.3.4. Brand association strength measure ... 11
2.3.5. Response time latency ... 12
2.3.6. Free association ... 14
2.3.7. Limitations and critiques on free association ... 15
3. Methods of elicitation evaluated ... 16
3.1. ZMET evaluated ... 16
3.2. BCM evaluated ... 16
3.3. BANV evaluated ... 16
3.4. BAS evaluated ... 17
3.5. Response time evaluated ... 17
3.6. Free Association evaluated ... 18
3.7. Why free association and response time latency were chosen for the experiment ... 18
4. Different types of brands... 19
4.1. Symbolic, functional, or experiential brands ... 19
4.3. Brand types distinguished by product category ... 23
5. Hypotheses development ... 24
5.1. The influence of brand concept ... 25
5.2. The influence of brand breadth ... 26
5.3. Interactions between concept and breadth ... 27
5.4. The influence of product categories ... 29
6. Methodology ... 31
6.1. Research design ... 31
6.2. Stimuli requirements ... 32
6.3. Pre-test 1: Qualitative pre-test ... 32
6.4. Pre-test 2: Quantitative pre-test ... 32
6.5. Pre-test 3: Qualitative and Quantitative pre-test ... 33
6.5.1. Introduction ... 33
6.5.2. Questionnaires and Respondents ... 34
6.5.3. Items measuring concepts ... 34
6.5.4. Items measuring breadth ... 35
6.5.5. Items measuring brand familiarity ... 36
6.5.6. Anlysis pre-test 3 ... 37 6.5.7. Results pre-test 3 ... 37 6.5.7.1. Clothing ... 37 6.5.7.2. Cosmetics ... 39 6.5.8. Conclusion pre-test 3 ... 41 6.6. The experiment ... 42
6.6.1. Creating the experiment ... 42
6.6.2. Experiment procedure ... 43
6.6.3. Respondents ... 43
6.6.4. Coding the associations ... 44
7. Results ... 45
7.1. Manipulation check ... 45
7.1.1. Introduction ... 45
7.1.2. Reliability check of scales ... 45
7.1.3. Methods of analysis manipulation check ... 45
7.1.4. Results manipulation check ... 46
7.1.4.1. Clothing ... 46
7.1.5. Control variable gender ... 49
7.1.6. Conclusion manipulation check ... 50
7.2. Hypotheses testing ... 50
7.2.1. Analysis hypotheses ... 50
7.2.2. Results hypotheses ... 51
7.2.2.1. The influence of brand concept tested ... 52
7.2.2.2. The influence of brand breadth tested ... 54
7.2.2.3. Interaction effects between concept and breadth tested ... 55
7.2.2.4. The influence of category tested ... 56
7.2.3. Additional analysis ... 58
8. Discussion ... 60
8.1. Discussion of the results ... 60
8.1.1. Brand concept ... 60
8.1.2. Brand breadth ... 61
8.1.3. Interaction between brand concept and breadth ... 62
8.1.4. Category and concept ... 64
8.2. Implications ... 65
8.2.1. Theoretical implications ... 65
8.2.2. Managerial implications ... 68
9. Conclusion ... 70
9.1. Summary and answer problem statement ... 70
9.2. Limitations and suggestions for future research ... 71
References ... 74
Appendices ... 78
Appendix A: Overview of the different elicitation methods ... 78
Appendix B: Results pretest 1 ... 79
Appendix C: Questionnaire and results pre-test 2... 80
Appendix D: Questionnaire and results pre-test 3 ... 82
Appendix E: Dataset number of associations... 84
Appendix F: Additional explanation response time measurement ... 84
Appendix G: Questionnaire manipulation check ... 85
Appendix H: Assumption checks ... 85
List of figures and tables
Figures
Figure 2.1. Keller’s model displaying CBBE and brand association types 7
Figure 2.2. Example of a brand concept map 10
Figure 5.1. Conceptual model 24
Figure 6.1. Comparing means of variables Clothing pre-test 3 37 Figure 6.2. Comparing means of variables Cosmetics pre-test 3 39 Figure 7.1. Comparing means of variables Clothing manipulation check 46 Figure 7.2. Comparing means of variables Cosmetics manipulation check 48 Figure 7.3. Number (N) of functional/symbolic associations per brand concept 53 Figure 7.4. Response times (RT) per brand concept in milliseconds 54 Figure 7.5. Interaction concepts and breadth on number of functional associations 55
Tables
Table 6.1. The filled in brands by the second focus group 34
Table 6.2. Results pre-test 3, Paired-Samples T test Clothing 38
Table 6.3. Results pre-test 3, brand familiarity Clothing 39
Table 6.4. Results pre-test 3, Paired-Samples T test Cosmetics 40
Table 6.5. Results pre-test 3, brand familiarity Cosmetics 40
Table 7.1. Results manipulation check, Paired-Samples T test Clothing 47 Table 7.2. Results manipulation check, brand familiarity Clothing 48 Table 7.3. Results manipulation check, Paired-Samples T test Cosmetics 49 Table 7.4. Results manipulation check, brand familiarity Cosmetics 49
Table 7.5. Between subjects data for hypotheses analysis 51
Table 7.6. Within subjects data for hypotheses analysis, effect of category (Clothing/Cosmetics) 56 on number (N) of associations and response time (RT) of associations
Table 7.7. Number (N) of symbolic associations per category for symbolic brands 57 Table 7.8. Number (N) of functional associations per category 58 Table 7.9. Between subject effect of brand concept on the total number and total response time 58
of associations (both association types accumulated)
Table A.1. Summary of the different elicitation methods 78
Table B.1. Results pre-test 1 79
Table C.1. Results One-Sample T tests pre-test 2 81
Table D.1. Results per item John Frieda and Rituals 83
Table E.1. Database number of associations (37°Celsius, brand consultancy) 84
Table H.1. Results normal distribution check data 85
Tables Appendix: SPSS Output main analysis
Table I.1. Repeated Measures ANCOVA, between subjects effects on number of functional associations 86 Table I.2. Repeated Measures ANCOVA, between subjects effects on number of symbolic associations 87 Table I.3. Repeated Measures ANCOVA, between subjects effects on response time of functional associations 87 Table I.4. Repeated Measures ANCOVA, between subjects effects on response time of symbolic associations 88 Table I.5. Repeated Measures ANCOVA, between subjects effects on difference response times functional and 88
symbolic brands
Table I.6. Repeated Measures ANCOVA, within subjects contrasts on number of symbolic associations 89 Table I.7. Repeated Measures ANCOVA, within subjects contrasts on response time of symbolic associations 89 Table I.8. Repeated Measures ANCOVA, within subject contrasts on number of functional associations 90 Table I.9. Repeated Measures ANCOVA, within subjects contrasts on response time of functional associations 90 Table I.10. Repeated Measures ANCOVA, between subjects effects comparing total number of associations 91
between brand concepts
Table I.11. Repeated Measures ANCOVA, between subject effects comparing total response time of associations 91 between brand concepts
1 1. Introduction
1.1. Background
1.1.1. Brand association strength
Brands have become an inevitable part of modern day society. When you sit in your living room at home, or in your office at work, or when you walk on the streets, you will encounter brands everywhere. It concerns everything from your television, to the coffee you drink, the clothes you wear, and even the very city you live in.
However, why would consumers care what brand they use, and why would companies care about how their brand is perceived? The answer is, because a brand can enhance or decrease the value of a product (Aaker, 1996). According to Keller (1993) consumers have certain associations with specific brands which shape their attitude (favourable or unfavourable) towards the brand, and ultimately their purchase intent. Furthermore, Keller explains that brand associations can be divided into three components, namely strength, favourability, and uniqueness. In this research the emphasis will be on strength of associations, since strength of associations is an important factor to take into account before looking at favourability and uniqueness.
Strength of associations can significantly influence the degree of (un)favourability and purchase intent. Furthermore, when unique associations are beneficial for the brand, it is desired for them to be strong. An example would be Apple which is associated with being user friendly and innovative (www.interband.com). However, when unique associations are strongly unfavourable, brand managers would also want to be aware of this in order to be able to improve the situation. An example of the latter would be Nike which was strongly associated with child labour (www.businessinsider.com). Therefore, research on what brand association strength implies and on how it can be measured, is significantly relevant.
2 1.1.2. Measures for brand association strength
Measuring brand association strength concerns a topic which received a significant amount of attention in academic literature recently. John et al (2006) initiated Brand Concept Mapping (BCM), a method where respondents are asked to create a map representing their network of associations concerning a certain brand. The maps consist of nodes (associations) and links at which the links can take values (I, II, III) indicating the strength between the nodes. Furthermore, French and Smith (2013) introduced the measure of Brand Association Strength (BAS), a method based on BCM, however this method uses structural density combined with the number of associations to measure association strength. These and more methods will be explained in more detail in chapter 3.
According to Lei et al (2008) another well-established method of measuring association strength, is by using response time latency. They explain that response time measures the strength between nodes by measuring the time it takes respondents to link one node to another, supported by predetermined nodes programmed in computer software. The less time it takes to link nodes, the stronger the associations are. In their research, Lei et al measure the strength of association between brands and sub-brands. Chapter 4 will provide further information concerning response time and its applications.
Lastly, the concept of free association, a technique developed by Sigmund Freud (http://psychcentral.com), is of interest since this technique allows respondents to relate their own associations to the brand, instead of a predetermined selection of associations selected by the researcher(s).
To summarize, different methods for measuring brand association strength have been developed. Furthermore, response time can be used to measure association strength. Lastly, free association allows respondents to bring up their own ideas instead of being biased by predetermined associations.
3 1.2. Problem definition
1.2.1. Problem statement
Although brand association strength was measured by different researchers using different approaches, response time latency has not been used to measure brand association strength when applying free association. Conceptually, using the response time method combined with free association for measuring brand association strength could deliver new and interesting insights in the strength of brand associations. Furthermore, response time has not been tested amongst different types of brands. Therefore it was aspired to further test response time latency amongst four conditions, which were manipulated by different brand types. Four brand types were used in this research, distinguished by brand concept (functional vs. symbolic value positioning), and by brand breadth (broad vs. narrow product range orientation). Consequently the problem statement to be answered in this research is:
What is the influence of brand type on brand association strength, looking at response time latencies and number of associations?
1.2.2. Sub-questions
In order to acquire a clear overview of the topic of brand association strength and eventually to answer the research question, several sub-questions needed to be answered:
A. What are brand associations and associative networks, and what is brand association strength?
B. Which methods have (recently) been developed for eliciting brand associative networks and brand association strength respectively?
C. How can free association and response time be used to measure brand association strength?
4 1.3. Delimitations of the study
This study was set up to increase applicability of measures of strength of consumer associations to brands, and to provide new insights concerning strength of associations amongst different brand types. This study will focus on customer-based brand equity as defined by Keller (1993), however, not on the broader definition of brand equity. Furthermore, the study does not go into depth concerning brand concepts and brand breadth. Only two concepts (functional vs. symbolic value orientation) will be assessed and breadth refers to a broad or narrow positioning strategy. Lastly, the number of categories used in the experiment will be limited to two.
1.4. Contribution of the study
1.4.1. Theoretical contributions
This study will contribute to current research in several ways. Current research concerning brand association strength has focused on developing methods of eliciting brand association strength. However, it was argued that these methods are biased, since strength is allocated arbitrarily. Furthermore, for the creation of a mental map, already 45-50 percent of the mentioned associations are excluded from the research (e.g. John et al, 2006).
The response time method in combination with free association might be significantly relevant since with this method access can be acquired to all of the associations respondents have with the brand, without the bias of pre-selection of associations or arbitrarily allocated strengths. Furthermore, response time in combination with free association provides a new method for measuring brand association strength.
1.4.2. Managerial contributions
According to Chen (2001) brand associations have several functions, namely retrieving and processing of information, brand differentiation, create positive attitudes/feelings, providing a
5
basis for brand extension, and they provide purchase intent. Furthermore, Samu (1994) explains that it is important for brand managers to know which associations are strongly related to their brand(s) in order for them to decide on the appropriate product category (concerning brand extension) or product usage (concerning line extension). Furthermore, when brand managers know which associations consumers have with their brand, they can use this information to increase the efficiency and effectiveness of their marketing programs (Kim et al, 2003).
1.5. Outline
The following chapter (chapter 2) will provide an integrated view on what brand associations, and brand associative networks are, and what brand association strength implies, including methods of elicitation (of associations and associative networks). In chapter 3 these methods of elicitation are evaluated. Following, chapter 4 will cover the different brand types. In chapter 5, hypotheses development is covered. Then in chapter 6 the methodology will be explained and the pre-tests performed, followed by chapter 7 providing the results of the manipulation check and the hypotheses results. Then in chapter 8 the results are discussed and put into perspective by providing theoretical and managerial implications. Finally chapter 9 provides the conclusion and answer to the problem statement, limitations, and suggestions for further research.
6 2. Brand associations, association networks and strength, and methods of elicitation
2.1. Brand associations explained
In order to provide a clear research framework, it was aspired to first establish what brand associations are, and how they can be categorized. In this section, the perspective of different authors concerning brand associations will be reviewed.
Over the past decades much attention has been attributed to consumer-based brand equity (CBBE) (e.g. John et al 2006, Krishnan, 1996). CBBE can be defined as “the differential effect of brand knowledge on consumer response to the marketing of the brand” (Keller, 1993, p2). Keller (1993) explains that this knowledge exists of the components brand awareness (recall and recognition of the brand) and brand image (set of associations linked to the brand in memory), which occurs when the consumer is familiar with the brand.
Brand associations can for example include product attributes, people, places, or occasions relevant to the brand (Henderson et al, 1998). According to Aaker (1996) associations can be structured in three pillars, namely value (value for money, value over competitors), brand personality (e.g. the social setting in which a certain brandy can be served), and organizational associations (e.g. Mc Donald’s, Ronald McDonald).
Keller (1993) divides associations into only three different types, namely attributes, benefits and attitudes, however these are then sub-divided. Attributes consist of product related attributes, which are those related to the physical composition and the service function of the product, and non-product related attributes which are related to price, packaging, user imagery (which type of people use the product) and usage imagery (where and when the product or service is used). Benefits are divided into functional benefits, which are the advantages of product usage (correspond to product related attributes), experiential benefits, which satisfy sensory pleasure, variety and cognitive stimulation (correspond to product related attributes), and symbolic benefits, which satisfy social approval, self-expression and
7
outer directed self-esteem (correspond to non-product related attributes). Lastly as Keller (1993) explains, brand attitude is the combined view of all associations the consumer has with the brand and generally leads to consumer behaviour, e.g. purchasing the product. Figure 2.1 below displays CBBE and association types as explained by Keller (1993).
Figure 2.1. Keller’s model displaying CBBE and brand association types (Keller, 1993)
Biel (1992) divides the brand image into three sub-images, namely the corporate image, the user image, and the product image. However, he adds that the relative usefulness of these subcategories varies per product category and brand. Furthermore, he divides associations into hard associations, which imply functional or tangible attributes such as user-friendliness or premium price, and soft associations, which imply emotional attributes such as excitement or dullness.
To summarize, there are different ways in which brand associations can be structured. Several similarities and difference in structures were discovered. It was assumed it depends
8
on the environment and the brands which types of associations will be dominant. In the experiment of this research, associations were classified as functional or symbolic, based on Keller’s model (1993). The next section will explain how these associations are mapped in the mind of the consumer and how they can be elicited from the mind of the consumer.
2.2. Brand association networks and brand association strength
Brand associative networks identify which associations are directly or indirectly (through other associations) linked to the brand and how they are interconnected in the mind of the consumer (Schnittka et al, 2012). Keller (1993) explains that such networks incorporate memories and knowledge, modelled as a set of nodes and links. The nodes represent the information, and the links the connections between points of information, which can vary in strength. The strength of association between the activated nodes determines the extent of retrieval of memories connected to the brand. The more nodes are activated, the stronger the brand. Several different techniques to elicit brand associative networks from the mind of the consumer have been developed and will be explained next.
2.3. Eliciting brand association networks and brand association strength
Now that the concepts of CBBE and brand association networks have been introduced, it is time to look at several methods developed to identify brand association networks, and to measure the structure and strength of associations. First, a qualitative measure, Zaltman’s Metaphor Elicitation Technique (ZMET), will briefly be explained to gain insight in the complexity of brand association networks. Then Brand Concept Mapping (BCM), a more practical approach incorporating strength measurement, followed by Brand Associative Network Value (BANV) incorporating explicit measures, and lastly, Brand Association Strength (BAS), specifically developed to measure brand association strength.
9 2.3.1. Zaltman’s metaphor elicitation technique
Christensen and Olson (2002) explain that mental models imply attitudes, emotions and feelings, symbols, memories of past consumption events, actions, images, personal values, and even representations of sensory experience, e.g. smell, taste and touch (broader than purely cognitive structures). Within certain models, the content itself is more important than structure (how the content is organized in memory). The ZMET method includes Laddering to elicit associations, which means participants are continuously confronted with their choices to gain an in depth insight in their decision making (Renolds and Gutman, 1988). Furthermore, Coulter et al (2001) explain that ZMET uses metaphors and images for eliciting unconscious or tacit content, where consumers are asked to express their thoughts and feelings by selecting different images.
The ZMET technique shows that the associations within a consumers mind can be rather complex, and how these complex associations can be elicited. Christensen and Olson (2002) explain that ZMET can be used to elicit consumers’ meaning about certain topics (brands) and to map those meanings as mental models. However they explain this technique takes a significant amount of time for both researcher and participant and can only be used to explore data and/or mental maps in more varied and deeper ways. The next method, Brand Concept Mapping (BCM) is more practical in nature.
2.3.2. Brand concept mapping
Brand Concept Mapping (John et al, 2006) concerns a methodology based upon the idea of concept maps (see for example Carpenter, 1989) and filled the gap between qualitative and quantitative measures. French and Smith (2013) explain that the qualitative measures could create individual maps, however could not aggregate them, whereas quantitative measures could not accurately replicate the mental maps in consumers’ minds. According to John et al (2006) BCM is able to identify brand association networks and to show which associations
10
are directly (first order association) or indirectly (through other associations) linked to the brand, called placement level. Furthermore, BCM shows the strength between different nodes by means of the number of lines attached to them, single, double or triple, where a triple line represents the strongest connection. Figure 2.2. provides an example of a brand concept map.
Figure 2.2. Example of a brand concept map (John et al, 2006)
The individual maps are created by the participants by means of laying cards with pre-selected associations and cards with single, double or triple lines connecting the associations. The associations were previously selected by the majority of votes (i.e. frequency of mention) after a brainstorming session (i.e. free association), with a cut-off point at 45-50 percent of total votes. The number of lines attached (I, II, III) are also based upon the majority of votes of respondents, who may decide how strong an association is linked to either the brand or to another association, when creating a brand concept map.
John et al (2006) explain that the technique of BCM was developed to make brand mapping more accessible to marketing practitioners. They say that the advantage of this method is that it does not cost the researcher or the participants a significant amount of time, as opposed to for example ZMET which requires extensive into depth interviews with
11
participants. Furthermore, it does not require expert knowledge about analytical techniques. Lastly, BCM facilitates aggregation of individual maps which results in acquiring consensus maps.
The major contribution of this method is its flexibility and little time requirements (French & Smith, 2013; John et al, 2006). However, its major drawback is that it does not elicit associations which require in depth interviewing like ZMET does (John et al, 2006).
2.3.3. Brand association network value
Schnittka et al (2012) further developed the BCM (Advanced BCM) approach by adding explicit information on the favourability of single brand associations. Furthermore as explained in their article, they developed a new measure called Brand Association Network Value (BANV) which enables quantification of the overall network favourability. The BANV metric combines network structure (i.e. uniqueness and strength of associations) and favourability (evaluative judgment and importance to purchase decision) of individual associations, thereby quantifying the overall favourability of the brand associations into a single measure. The favourability of individual associations was measured using a seven point Likert scale.
Although both BCM and BANV include strength of association as a factor, they do not focus on it. The next method focuses on a specific measure of strength, and is based on the previously mentioned methods.
2.3.4. Brand association strength measure
French and Smith (2013) wrote an article which focuses on measuring brand association strength (BAS), in which a specific measure of strength was created. The authors explain that brand association strength is a function of the number of associations, the strength of links between associations, and the structure of the associative networks, where the associations
12
closest to the brand are the strongest as was established by previous researchers (e.g. John et al, 2006; Schnittka et al, 2012). According to French and Smith (2013) a measure of strength should therefore incorporate all these elements.
To structure the associative networks, the authors use the BCM approach. Associations linked directly to the brand are called first order associations, those linked through one other are called second order associations, and those linked through two other associations are called tertiary associations. Furthermore, new to the equation is the inclusion of density (D). Density implies the proportion of concept map links out of the maximum number of links possible (between associations). Furthermore the links are weighted by a single, double, or triple line (I, II, III). Consequently, values of density range from 0, i.e. no links, to 1, i.e. maximum number of links, all weighted III (a density value of 1 would be very unlikely). Furthermore, density is evaluated for each link type (brand to first order, first order to first order etc.), creating structural density (Ds). The higher Ds, the stronger the brand associative network’s connection to the brand. Lastly, Ds is multiplied by the number of associations (n) in the brand concept map. Consequently, a measure of brand association strength is developed incorporating the three key elements which are the number of associations, strength of links between associations, and structure of the associative network. Although the previously discussed methods were developed rather recently, these are not the only available methods. The next section elaborates on a different approach for measuring brand association strength, namely by means of response time latency.
2.3.5. Response time latency
Even though the previous methods for eliciting brand associative networks and strength of associations have been developed rather recently, these are not necessarily the best possible methods available to assess brand association strength. According to Bassili and Fletcher (1991) another way to test strength of associations is by means of response time latency
13
(response time). In the introduction there was already the example of Lei et al (2008) using response time to measure negative spill-over of brands. The following paragraphs will elaborate further on response time, provide additional examples, and explain its utility.
Fazio et al (1989) explain that strength of associations is related to attitude, and that attitude can predict behaviour. The stronger the associations, the more accessible the attitude, and hence the better predictor of consumer behaviour. Furthermore, Bassili (1993) explains that response time is not a pure index of attitude accessibility, since it is affected by e.g. comprehension, however, response time has proven to be a rather accurate predictor of attitude, and hence behaviour.
Bassili and Fletcher (1991) demonstrated that response time can be used to measure how well attitudes are pre-processed and integrated in memory. In their research they measure the time it takes respondents to respond to certain questions, assisted by a device called ‘Voice-key’, which activates a computer clock (measures in milliseconds) when a response is recorded.
Furthermore, Bassili (1993) explains that attitude strength can be measured by meta-attitudinal self-reported indexes such as importance, intensity or certainty. However, he suggest that these cannot accurately capture the attitude strength due to the limitations of verbal documents. In his study, Bassili uses response time to test if it can measure attitude strength which in turn is a good predictor of behaviour. The study concerns presidential elections and it turns out that response time is a more suitable and accurate measure (valid in 90 percent of the cases) both in predictive power and economic factors (time, effort) than meta-attitudinal predictors.
Another example concerns Herr et al (1996), who used response time to investigate the links between different product categories, relevant for brand extensions. They explain that dominance is the first key factor in possible brand extension, which implies the strength
14
of the directional association between parent category and the branded product. The second key factor is relatedness between the parent category and branded product. Both dominance and relatedness are measured by response time, where consumers could indicate whether or not they found there to be a relation between the product-brand pairs shown to them, by entering yes or no (keyboard response). However, the difference with previous experiments at which the yes answer is studied, here the focus is on the no answer. The faster respondents answer no, the stronger the disconformity (disconformity method).
To summarize, response time is a widely used method to measure strength of associations, and can be used in different manners, supported by different technologies. Furthermore, response time has proved to be a more suitable and accurate measure than meta-analysis for measuring attitude strength, which in turn is a predictor of behaviour. Following, although briefly put forward earlier (within the BCM approach), the next sections will explain the free association method in further detail.
2.3.6. Free association
Free association is a method originally developed by Freud to analyse symptoms, elements of dreams, or parapraxes, where subjects were asked to report everything that came to mind without the impact of any predetermined directions or influences (MacMillan, 2001). However, free association can also be used to analyse brand associations. Chen (2001) uses free association in his research to brand associations, where he asks subjects to write down whatever comes to mind when they think about a certain brand. Furthermore, John et al (2006) also used free association in their BCM approach.
Another example of using free association is provided by Nelson et al (2004) who use free association to produce the largest free association database of linked words in the United States (6,000 participants; 750,000 responses to 5,019 stimulus words). They explain that free association implies providing a cue (word) which activates other nodes (words) in memory
15
according to their relative accessibility. This accessibility depends on lexical knowledge acquired through experience, and the experience leads to associative structures/networks consisting of nodes and links between them. Furthermore, they explain that complexity concerning lexical knowledge is established in the fact that also norms concerning trends and local culture need to be taken into account, which in turn are sensitive to experience that deviates from the norm. Therefore, when making a map by means of free association, it is preferred that subjects share the same cultural heritage.
To conclude, free association concerns a widely used method for applicable to different types of research, used for eliciting associations. However, free associations also has limitations and received several critiques which will be mentioned in the next section.
2.3.7. Limitations and critiques on free association
According to Nelson et al (2004) the disadvantage of free association is that it can tell which nodes are linked, however not which ones are not linked. The reason is that one can never tell when, where, and how any link was learned, and thus how it is interconnected to other links. Furthermore, according to Koll et al (2010), free association can tap into nonverbal associations, however the focus is on the easily accessible and recordable verbal associations from semantic memory. Consequently as they suggest, the major downside of free association is that it is not ideally suited to elicit deeper, implicit associations. This was also suggested by John et al (2006) as being a downside of BCM.
To conclude, free association cannot accurately replicate a mental map in consumers’ mind and it is not ideally suited to elicit deeper, implicit associations. In the next chapter, the different methods of elicitation will be evaluated.
16 3. Methods of elicitation evaluated
As was discovered, several different methods have been developed for eliciting brand associations and brand associative networks respectively. In this section, these methods will be evaluated, and motivation will be provided why it was assumed free association combined with response time latency is the most appropriate method for researching strength of associations.
3.1. ZMET evaluated
Although it seems this method would be suitable for eliciting symbolic associations, it is highly time consuming. Therefore, it was expected this method would only provide time to interview several respondents and build mental maps based on their answers. This would make it difficult to generalize the results. Furthermore, this method does not specifically focus on measuring strength of associations, which was the aim of this research.
3.2. BCM evaluated
BCM offers a practical technique for eliciting brand association networks and brand association strength. However, the major disadvantages of BCM are that before half of the associations are not being included in the research, since they are considered not significantly important. However, there might be interesting information to be distracted from these associations, which is now lost. Furthermore, the measures of strength (I, II, III) are arbitrary, which for this research was considered inappropriate.
3.3. BANV evaluated
As the name implies this method measures the value of an associative network, however the focus is on favourability, not on strength. Furthermore, this method is based on BCM which provides the same disadvantages concerning the cut-off point at 45-50 percent of associations
17
and the arbitrary strength allocations. Therefore it was suggested this is not an effective measure when researching brand association strength.
3.4. BAS evaluated
The name of this method suggests it would be an effective method for measuring strength of associations. This method is based on BCM and provides the same advantages, however also the same disadvantages. Although this method incorporates density as a measure of strength, the arbitrary strength allocation of BCM is still included in the formula.
3.5. Response time evaluated
This method provides a rather accurate predictor of attitude according to Bassili (1993). Furthermore, relatively little time and effort is required (less than BCM) to perform research using response time, especially from the respondents’ perspective. This will conceptually increase the number of respondents, which increases the reliability of the outcomes. Another major advantage of this method is that the measurements are not arbitrary as opposed to the strength of links used in BCM and all models which are based on BCM. Furthermore, by recording response time, a clear difference in strength (amongst associations, different brands) can be measured. Hence, this method provides the possibility to prompt different brand types and categories (e.g. functional vs. symbolic, consumer electronics vs. automotive) of which the response times can be compared.
The major disadvantages of this method however, is that it is a quantitative measure, and therefore cannot accurately replicate mental maps in consumer’s minds (French and Smith, 2013). Furthermore, when using voice key, al responses are recorded, including irrelevant sounds or words. When the responses are typed in using the keyboard, the speed of typing and the length of responses will influence response times.
18 3.6. Free Association evaluated
The major advantage of free association is that there is no pre-selection of associations to be used in the research. Free association allows respondents to come up with their own thoughts without any pre-influence. Furthermore, using free association as a means of response, provides the advantage of acquiring the order, number, and type of associations which facilitates finding patterns and/or structure.
The major disadvantage of free association is that it is not ideally suited to elicit deeper, implicit associations (Koll et al, 2010). Therefore, finding the more symbolic associations might be more difficult when applying free association.
3.7. Why free association and response time latency were chosen for the experiment
The use of free association seems of major significance in order to attain associations within a relatively short time span. Furthermore, it was expected that response time would be an adequate method of measuring brand association strength, mainly since strength is then measured by recorded response time, and not by arbitrarily allocated scores. Therefore, in order to measure the influence of different brand types on brand association strength, free association was combined with response time latency. This allowed for measuring response times and number of associations. Consequently, this allowed for researching the problem statement by conducting an experiment in which this method was applied. A summary overview of all methods including their advantages and disadvantages is provided in Appendix A. The next section will elaborate on the different brand types used in the experiment.
19 4. Different types of brands
In this research it was aspired to provide insights in the strength of associations related to different types of brands. The types are distinguished by functional vs. symbolic, and broad vs. narrow. This chapter will explain these items.
4.1. Symbolic, functional, or experiential brands
According to Park et al (1986), the brand image is merely a perception created by marketers’ brand management. According to them, any product can be positioned using a functional, symbolic, or experiential image. They explain that functional brands aim to satisfy externally generated needs which are consumption-related, and symbolic brands aim to satisfy internally generated needs for self-enhancement, group membership, role position, or ego-identification. Furthermore they explain that experiential brands aim to satisfy sensory pleasure, variety, and/or cognitive stimulation. These three types of positioning correspond to the benefit associations derived from a brand according to Keller (1993).
However, Park et al (1986) discourage the idea to develop a brand with two or more positioning concepts for several reasons. The reasons include that a multiple positioning strategy creates inconsistency, is more difficult to manage since it competes against more brands, and creates confusion amongst consumers in identifying the brand’s meaning.
Bhat and Reddy (1998) explain that brand managers need to decide on positioning their brand image in a certain way, which can be functional, symbolic, or both. They based their research on Park et al (1986) amongst others, however leave out the experiential positioning as an option. Furthermore, they say that functional and symbolic positioning are not mutually exclusive and that symbolic can be divided into prestige (expensiveness and exclusivity) and personality expression. They use the Macintosh (Apple) as an example which is positioned as both functional and symbolic (creative, anti-establishment). However,
20
they add that the associations need to fit the brand, otherwise it may be considered neither functional nor symbolic.
Vigneron (2004) explains that the use or display of luxury (symbolic) products brings self-esteem to the owners, besides the functional utilities of the product. Consequently, he suggests that luxury products are able to satisfy both functional and psychological needs. Furthermore Vigneron (2004) provides five criteria to separate luxury from non-luxury brands, divided in non-personal and personal oriented perceptions. The three non-personal perceptions are perceived conspicuousness (social status), perceived uniqueness (scarcity of the product), and perceived quality (e.g. technology, engineering, craftsmanship), which are all higher at luxury brands. The two personal oriented perceptions concern extended self (own identity) and hedonism (sensory gratification and sensory pleasure), also both higher at luxury brands.
Vickers and Renand (2003) find similar results in their research, in which they demonstrate that luxury goods can be distinguished from non-luxury goods by the extent to which they mix three key dimensions, namely functionalism, experientialism, and symbolic interactionism. The difference between luxury and non-luxury goods was that within luxury goods the symbolic and experiential items are valued more compared to functional items, whereas within the non-luxury goods the functional items are valued more than symbolic or experiential items.
To conclude, it remains to be debated whether or not the positioning of a brand should be focused on only one dimension or on multiple dimensions. Furthermore, some researchers say there are two dimensions (functional, symbolic), whereas others add a third dimension (experiential). Although in both functional and symbolic products the three dimensions of symbolism, experientialism, and functionalism are valued, the focus is on different aspects. In symbolic brands, symbolic and experiential aspects are valued more, and in functional brands
21
the functional aspects are valued more. For this research, it was chosen to make a distinction between functional and symbolic brands.
4.2. Broad vs. narrow brands
As indicated in the introduction, the scope of this research does not extend to in depth brand extension analysis. However, this section aims to provide a basic overview of what brand extensions are and how they relate to different positioning strategies and brand association strength.
According to Meyvis and Janiszewski (2004) it is common for brands to extend into more than one product category. They define broad brands as brands that offer a portfolio of diverse products, as opposed to narrow brands which offer a portfolio of similar products. Aaker and Keller (1990) explain that during brand extension, existing associations with the brand can potentially be transferred to the extensions. These associations can be helpful for the extension (when they are positive), however only if they fit the new product. Furthermore as they suggest, the success of these transferred associations depends on the perceived quality of the brand. The higher the perceived quality of the original brand, the more successful an extension will be.
Meyvis and Janiszewski (2004) further suggest that extensions are successful when they include brand benefits which are valued in the extension category, and when benefit associations are accessible. The accessibility as they suggest, depends on the amount of interference by competing associations (e.g. product category associations). Additionally, they say that broader brands tend to have more accessible benefit associations than narrow brands and can therefore more successfully expand than narrow brands. However, when benefit associations are equally accessible and diagnostic, it is the similarity between the brand and its extensions that causes the success, which provides narrow brands the advantage.
22
Boush and Loken (1991) explain that brand breath refers to the variability of product types represented by the brand. They say that this variability can be limited to one single product, e.g. only Ketchup (extremely narrow), or highly differing products Ketchup and lawnmowers (extremely broad). However, they explain that brands can also be moderately broad when they represent many different products, which for example are all related to food (Heinz was provided as an example).
Concerning functional vs. symbolic brands, Park et al (1991) explain that consumers take two issues into account when evaluating brand extensions. The first issue concerns information concerning the product feature similarity between the old and the new product. The second issue concerns the concept consistency (brand unique abstract meanings) between the brand and its extension. Furthermore, they explain that for both functional and symbolic brands, the most favourable evaluations (increased purchase intensions) occur when both brand concept consistency and product feature similarity in the brand extension are high. However, as they explain, the symbolic brand has greater extendibility concerning products with low feature similarities compared to the functional brand.
To conclude, the success of brand extensions depend on the degree of accessibility of brand associations. Broad brands tend to have more accessible favourable associations, which provides an advantage in brand extensions in general. However, when accessibility of benefit associations is similar, narrow brands have an advantage in extension strategies due to higher similarity between the brand and its extension. Furthermore, both functional and symbolic brands are most successfully evaluated when brand concept consistency is high, however symbolic brands can more successfully extend into products with lower product feature similarities. The next section will briefly cover a distinction in brand types according to product category.
23 4.3. Brand types distinguished by product category
As mentioned in the previous section, Meyvis and Janiszewski (2004) explain it is common for brands to extend into more than one product category. Nevertheless, product category is rather broad term, which could imply many different products. However, according to Estelami and de Maeyer (2004) a distinction can be made between brands representing durable consumer goods (e.g. automotive) and brands representing non-durable consumer goods (e.g. fast moving consumer goods). They explain that typically, non-durable consumer goods are purchased frequently, as opposed to durable consumer goods, which are not purchased frequently.
Although it was assumed other distinctions in categories could also be made, a distinction in durable vs. non-durable consumer goods was considered sufficient given the scope and purpose of this research. In the next chapter the hypotheses are covered, which were based on the theory provided in the previous chapters.
24
IV: Breadth
1. Broad 2. Narrow
5. Hypotheses development
The aim of this study is to find out what the influence of different brand types is on brand association strength. Based on the previous theory (chapter 6) the brand types are distinguished by concept (functional vs. symbolic value positioning) and breadth (broad vs. narrow product range orientation). Furthermore brand associations are divided into two types, functional benefit associations and symbolic benefit associations (based on Keller, 1993). Brand association strength is measured by number of associations and response times for the following reasons:
o Number of associations is an important factor of strength measurement (French & Smith, 2013).
o Response time is considered an accurate measure of strength (Bassili, 1993).
o It was aspired to investigate the influence of brand type on number of associations and on response times to explore and compare the information both factors provide.
In order to provide a clear overview, a conceptual model was created which describes the variables and relations (figure 5.1). The first seven hypotheses are based on this model. Furthermore, four hypotheses related to category (H8-H11) were created.
Figure 5.1. Conceptual model
H1- 3 H6, 7 H4, 5 IV: Concept 1. Functional 2. Symbolic DV: Association strength 1. Number of functional associations 2. Number of symbolic associations
3. Response time functional associations
4. Response time symbolic associations
25 5.1. The influence of brand concept
According to Vickers and Renand (2003), within luxury goods (symbolic brands) the symbolic and experiential items were valued more compared to functional items, whereas within the non-luxury goods (functional brands) the functional items were valued more than symbolic or experiential items. Hence, it was expected that symbolic brands have more symbolic associations than functional brands, and functional brands have more functional associations than symbolic brands. Therefore the following hypotheses were developed:
H1a: The number of functional associations will be higher when a functional brand is activated than when a symbolic brand is activated.
H1b: The number of symbolic associations will be higher when a symbolic brand is activated than when a functional brand is activated.
As the former hypotheses explain, it was expected that less symbolic associations would be recalled from memory when a functional brand is activated than when a symbolic brand is activated. Meyvis and Janiszewski (2004) explain that associations are more accessible when they are not interfered with competing associations. This as was assumed would imply that the functional associations are not interfered with any many symbolic associations. This as was assumed leads to higher accessibility of the functional associations when a functional brand is activated, compared to when a symbolic brand is activated. Consequently, it was expected the response times of functional associations will be lower when a functional brand is activated than when a symbolic brand is activated.
Concerning symbolic brands, it was expected that the higher accessibility of symbolic associations (related to the higher number of accessible symbolic associations) would lead to lower response times of symbolic associations when a symbolic brand is activated than when a functional brand is activated. Therefore the following hypotheses were developed:
26
H2a: The response time of functional associations will be lower when a functional brand is activated than when a symbolic brand is activated.
H2b: The response time of symbolic associations will be lower when a symbolic brand is activated than when a functional brand is activated.
Furthermore, according to Vigneron (2004) luxury products are differentiated from non-luxury products since they bring esteem to the owner. However this is an advantage occurring besides the functional utilities of the products. Consequently, only a small difference in terms of number and response times between functional and symbolic associations is expected when a symbolic brand is activated.
Although Vickers and Renand (2003) explain consumers value functional attributes more than symbolic attributes when it concerns functional products, it remained questionable whether or not symbolic associations would be present at all besides the functional associations with every functional brand. Furthermore, free association is not ideal to elicit symbolic associations due to their limited accessibility (Koll, 2010). Therefore, a relatively large difference in response times of functional and symbolic associations is expected when a functional brand is activated. Consequently, the following hypothesis was developed:
H3: The difference in response times between symbolic associations and functional associations will be smaller when a symbolic brand is activated than when a functional brand is activated.
5.2. The influence of brand breadth
Meyvis and Janiszewski (2004) suggest that broad brands have more accessible benefit associations than narrow brands. It was assumed this is related to the number of product types these brands represent (Boush & Loken, 1991), which will lead to more product related associations. Since brand concept consistency is important for brand extensions, it is expected
27
that this remains similar during brand extension (Park et al, 1991). Hence, when a brand extends, the main increase in associations would be caused by functional associations related to the products, rather than symbolic associations. Therefore, it was expected that the number of functional associations will be higher when a broad brand is activated than when a narrow brand is activated. Hence, the following hypotheses was developed:
H4: The number of functional associations will be higher when a broad brand is activated than when a narrow brand is activated.
Similar as with brand concept, the higher accessibility of broad brands is expected to lead to lower response times since associations would be easier to recall from memory. Hence, the following hypotheses was developed:
H5: The response time will be lower when a broad brand is activated than when a narrow brand is activated.
5.3. Interactions between concept and breadth
According to Boush and Loken (1991) broad brands have more different product types. Hence, it was assumed these different product types lead to more functional associations. However, brand concept consistency of symbolic broad brands is expected not to change to a large extent (Park et al, 1991). This would mean that the symbolic associations with the brand will remain approximately the same when a brand increases its product range. Consequently, it is assumed that the increase in number of associations between a broad and a narrow brand is mainly caused by more functional associations, not by symbolic associations. Furthermore, based on Meyvis and Janiszewski (2004) it is assumed that the accessibility of functional associations is interfered by competing symbolic associations, mainly amongst symbolic brands. Consequently, it is expected that breadth will have a larger influence on functional brands than on symbolic brands in terms of number of associations. Furthermore, the
28
influence is expected to be mainly on the functional associations.
Narrow brands have limited products to relate functional associations to. Therefore, the symbolic associations rather than functional associations would cause the number of associations to increase, since they can be present besides functional associations (Bhat & Reddy, 1998). This would mainly be the case for symbolic brands, however symbolic associations are expected to be limited due to brand concept consistency. Hence, no difference in terms of number of symbolic associations between broad or narrow brands is expected. Consequently, it is expected that the influence of breadth will be mainly present on the number of functional associations, and mainly on functional brands. Hence, the following hypothesis was developed:
H6: Functional brands will have a larger number of functional associations than symbolic brands, and this difference is larger when a broad brand is activated than when a narrow brand is activated.
As explained earlier, it is expected broad brands will have more accessible functional associations than narrow brands (Boush & Loken, 1991), which would lead to lower response times. Also, functional brands are expected to have more functional associations than symbolic brands, since symbolic brands can have competing symbolic associations (Meyvis & Janiszewski, 2004). Hence, the response times of functional associations are expected to be lower when a functional brand is activated than when a symbolic brand is activated. Furthermore, this difference is expected to be larger when a brand is broad, than when a brand is narrow. Consequently, the following hypothesis was developed.
H7: Functional brands will have a lower response time for functional associations than symbolic brands, and this difference is larger when a broad brand is activated than when a narrow brand is activated.
29 5.4. The influence of product categories
In order to test the hypotheses, it was aspired to use different product categories to increase reliability of the results. However, based on the theory explained in the previous chapters it was assumed product categories have an influence on the results. This section will elaborate on the influence of product categories based on these theories, leading to four additional hypotheses.
According to Bhat and Reddy (1998) symbolism can be described by 17 items. However, it was assumed that these items are not applicable to all product categories. For example the items status symbol, exciting and unique could be applicable to the automotive category (e.g. Ferrari, Rolls Royce) however are assumed to be less or not at all applicable to e.g. Fast Moving Consumer Goods (FMCG). Nevertheless, it was assumed this does not mean these categories cannot have symbolic brands. It merely implies that brands could be considered symbolic based on less or different items.
Furthermore, in accordance with Meyvis and Janiszewski (2004) the difference in number of associations is assumed lead to higher/lower accessibility of symbolic benefit associations. It was expected that the difference in accessibility will lead to different response times concerning symbolic associations in different categories.
In accordance with Estelami and de Maeyer (2004), product categories can be divided into durable consumer goods (e.g. automotive) and non-durable consumer goods (e.g. FMCG). Based on the previous arguments, it was expected that durable goods will have more symbolic associations than non-durable consumer goods. Hence, the following hypotheses were developed:
H8: The number of symbolic associations will be higher for durable consumer goods than for non-durable consumer goods, when a symbolic brand is activated.
30
H9: The response time of symbolic associations will be lower for durable consumer goods than for non-durable consumer goods, when a symbolic brand is activated.
Concerning the functional items, it was assumed in accordance with Vigneron (2004) that all products have a certain degree of functionalism besides possible symbolism. Furthermore, according to Park et al (1986) functional brands aim to satisfy externally generated needs which are consumption-related. Since non-durable consumer goods are assumed to be more related to direct consumption than durable consumer goods, they are expected to be related to more functional associations (direct applications) than durable consumer goods. Similar to the symbolic associations, it is expected that the higher number of functional associations (higher accessibility) will lead to lower response times of functional associations. Consequently, the following hypotheses were developed:
H10: The number of functional associations will be higher for non-durable consumer goods than for durable consumer goods.
H11: The response time of functional associations will be lower for non-durable consumer goods than for durable consumer goods.
31 6. Methodology
6.1. Research design
In order to answer the hypotheses mentioned in the previous chapter an experiment needed to be conducted, measuring number of associations and response times for different brand types. This section describes the research design used for the experiment.
The experiment used a combined design of 2x2 (between subjects) x2 (within subjects) repeated measures design. The independent variables were brand concept (functionalism, symbolism), and brand breadth (narrow, broad). The within subject variables were the different categories. The dependent variables were number of associations and response times. The design was a repeated measures design since the same respondents were confronted with the same measurement repeated over two different product categories. Together, concept and breadth formed four conditions for each category which were: Functional-Narrow, Functional-Broad, Symbolic-Narrow, and Symbolic-Broad.
In order to decrease the risk of errors, two control variables were included, namely gender and brand familiarity. It was assumed gender could influence the outcomes within both categories, since women might have more knowledge of both clothing and cosmetics than men, which would make it easier for them to recall associations. Furthermore, it was assumed brand familiarity could have an influence on the outcomes in both categories, since more familiar brands would make it easier to recall associations. Besides including control variables, several pre-tests were conducted to find the right stimuli to represent the four conditions for both categories. In the next sections, the stimuli requirements and the pre-tests will be explained.
