Sunteți pe pagina 1din 15

Understanding Sources of Brand Equity: A New Method to Represent Unbiased Perceptions ABSTRACT In this study we propose a new method

to represent unbiased perceptions of brands. The method is a likelihood-based model that simultaneously disentangles a major class of psychological bias affecting attribute based perceptions, and represent the common structure across multiple variable batteries, reducing the dimensionality of the problem. Our results indicate that it is feasible to deepen brand equitys sources through a cognitive representation that depicts the actual brand performance on attributes. KEYWORDS: Brand ratings, perceptual mapping, dimensionality reduction, brand equity. EMAC TRACK: Marketing Research and Research Methodology

INTRODUCTION In this study we propose a new method to represent unbiased perceptions of brands. More specifically we propose a new MDS procedure to jointly represent genuine attribute-based and attribute-free perceptions, coming from an empirical study of competitive brand positioning. As behavioural processes underlying the formation of consumers brand ratings indicate they may contain biases, our procedure implements a decompositional model that isolates brand-specific associations from other effects (Dillon et al. 2001). At the same time the procedure uses proximity judgements between brands to obtain a dimensionality reduction of the problem, in order to represent perceptions on a two-dimensional space. The model offers a parsimonious representation of multiple batteries of brand related data, implementing suggestions coming from psychological and classification literature. An application is given using Aakers (1997) brand personality scale. The aim of the study is twofold, it being rooted both in methodological and conceptual domains. The former refers to the development of a new technique that deals with the collection of multiple batteries of measurements from the same set of respondents (DeSarbo and Wu, 2001), simultaneously decomposing brand rating into its main components, namely functional and image. The latter is related to the contribution to the brand equity literature. Any attempt to analyze and to map brand ratings is connected to the measurement of the equity associated to a brand: in this way we hope to shed further light on the understanding of equitys sources. BACKGROUND Understanding how consumers ratings are formed relies on cognitive processes aimed to capture brand associations that are stored in consumers memory (Keller, 1993). Consumers usually associate a brand name with a particular attribute, some benefits, a specific usage situation, a logo, and features or characteristics (Roedder-John et al., 2005). The analysis of brand associations represents the basis of brand equity measurement, as models of brand equity usually give emphasis to brand knowledge and brand associations (Aaker, 1991; Keller, 1993; 1998). On the other hand brand ratings capture not just specific attribute information, or brand-specific associations (i.e. the knowledge of brand features, attributes or benefits), but they may also contain non-attribute information like the overall brand image (i.e. the general brand impressions), which are based on a more holistic view of the brand (Dillon et al., 2001). Besides that, disentangling sources of bias in brand ratings can provide managerial and theoretical insights for brand equity. In Keller (1993) and in Aaker (1991) views of brand equity, brand associations are captured by attribute ratings whereas brand awareness, uniqueness and strength for instance - could be assessed through the analysis of whats actually in brand ratings. Perceptual mapping has been historically regarded as one of the most important analytical tools in marketing research (Green et al., 1988; Shocker and Srinivasan, 1979; Steenkamp et al., 1994) and it is especially suited for the analysis of brand equitys sources. Perceptual mapping, in fact, is mainly used in the process of brand building and brand positioning. According to Keller and Lehmann (2005), Brand positioning sets the direction of marketing activities and programs what the brand should and should not do with its marketing. It involves establishing key brand associations in the minds of customers and other important constituents to differentiate the brand and establish competitive superiority (p. 11).

In brand positioning literature, researchers often deal with the necessity to collect multiple batteries of measurements from the same set of respondents (DeSarbo and Wu, 2001). In this line, as previously pointed out in related literature, a promising research stream is the one that jointly represents different types of data in a common dimensional space, that is a single geometric representation (Carroll and Green, 1997). Following DeSarbo and Wu (2001), our study proposes a MDS procedure that endeavours to analyze jointly various different variable batteries, such as brand-by-brand proximities (dissimilarities) and brand-byattribute ratings (attribute ratings), sorting out at the same time sources of bias in ratings. The scale we chose is the brand personality measure, which has been developed in the context of consumer behaviour research (Aaker, 1997). Previous research on this topic was mainly focused - on the symbolic use of brands, since consumers often fill brands with human personality traits (e.g. Gilmore, 1919; Aaker, 1997). The general tenet of this literature is that the greater the correspondence between human characteristics and those that describe a brand, the greater the preference for the brand (Malhotra, 1988; Sirgy, 1982). Relatedly, the brand personality dimensions can be seen as a key way to differentiate a brand in a product category (Halliday, 1996) and therefore to compare brands. A specific contribution to the study is given by psychological literature that identifies different classes of bias affecting consumers evaluation of brands along attributes. As Dillon et al. (2001) discuss, classical perceptual mapping may be not informative and difficult to interpret in situations in which interattributes correlations are considerably high. See figures 1a, b (Appendix B) for a clear depiction of this effect within the two categories considered in our study. This fact makes it impossible to evaluate the effect of a given attribute on the purchasing behaviour for each brand. Marketing literature on brand equity (Aaker, 1991; Keller, 1998; Dillon et al., 2001) recognizes that a brand rating contains something more than a mere performance on specific attributes. The process through which consumers obtain information likely depends on several factors, such as a specific brand related experience, the level of brand awareness, context effects that may affect information salience, earlier attribute ratings that affect later attribute ratings, and so on. The general brand impressions bias is relevant when information on a particular attribute are unavailable for a brand. In terms of brand rating variation this effect could be not due to an attribute specific evaluation of the brand. In this study we do not account for response style biases (Clemans, 1956; Rossi et al., 2000) while we drive our attention to the possibility to disentangle brand holistic preference from the joint set of brand ratings on specific attributes. THE STUDY The design of the study reflects our need to control for brands relevance for the target group as well as for intellectual and emotional situations. After identifying the product categories to be used1, we built a questionnaire using the scale developed in the literature on the FCB matrix (McWilliam, 1997) to determine products positioning in the FCB matrixs quadrants. We decided to assess brand ratings in high involvement conditions, where the analytic valuation of the brand performance on one or more attributes is supposed to be high, controlling for emotional and intellectual situations in which the overall brand impressions (that is non-attribute information) may, or not, play a crucial role. We chose the 2 product categories resulted as polarities among the factors generated as outcome of the factor analysis, i.e.: notebook computer (high involvement, intellectual learning), cellular phone (high

This information was collected through a questionnaire in which we asked to rate, on a 7 point Likert scale, the level of familiarity and relevance for each of 12 product categories we found in the literature on the FCB matrix (Weinberger and Spotts, 1989; Lambin, 2002). This procedure is aimed to obtain internal validity for the study.

involvement, emotional learning). Then we assigned brand names to each product category using subjects evoked set2. As already mentioned in the above section, the measure we selected to map brand perceptions is the brand personality scale (Aaker, 1997) which is composed of 15 items and measured with 7-point Likert scales. To this scale an attribute-free section was added in which subjects were asked to express dissimilarity judgments among brands on a 7-point Likert scale. Appendix A shows the questionnaire structure. Acknowledging the cognitive effort required of respondents to make multiple brand-by-attribute ratings, we wanted to avoid any possible biases that may arise in responding to such a complex and large battery of questions, exceeding human capabilities to remain concentrated (e.g. Brown and Melamed, 1990). Along these lines, we collected data from two samples of subjects, one for the notebook category and the other for the cellular category. Afterwards, as we will describe in detail in the next section, we develop a method which allows to ascertain the homogeneity of perceptions through the two groups. The questionnaires were distributed to university students. We gathered 130 questionnaires for each product category, however a total of 211 questionnaires were completed and considered usable (109 cellular phone questionnaires, 102 notebook questionnaires). THE MODEL Starting with subjects perceived brand ratings, the model implemented in the study decomposes the holistic dimension, defined as general brand impressions, and the unbiased perception of a brand on attributes, which represents the brand specific associations. This measure is then projected on T-dimensional space (where T = 2) using a joint representation of both attribute-free and attribute-based perceptions. 1 Let j= ... J be the brands within a given product category. Brands evaluated by I i consumers i =1... I along the attributes m =1... M . Let x j , m be the evaluation given by i individual i of brand j along attribute m . For each couple of brands let j , k be the perceived dissimilarity between brand j and brand k rated by individual i . The aim of the paper is to develop a statistical procedure to simultaneously reduce the original dimensionality of the problem, and disentangle brand preference effects and attribute specific evaluations. To reduce the dimensionality of the problem we use an approach that finds a projection matrix A = (m,t ) that, given the M -dimensions representation of brand j (namely the vector ( x j ,1 ... x j , m ) of judgments along attributes), outputs a T -dimensional (usually for perceptual mapping purposes) representation T =2 m m X j A = xk , mm ,1 ,..., x k , mm ,T . The matrix A will be the projection that ensures m =1 m =1 the best correspondence between empirical non attribute-based judgments , j and distances k d k , j between projected brands in the final T space. Namely distances in the projected spaces are defined by Euclidean norm:

i j ,k

m m i = x ij , m m ,t xk , m m ,t t =1 m =1 m =1 T

(1)

The notebook category is composed by the followings 6 competing brands: Acer, Apple, Asus, HewlettPackard, Sony, Toshiba. The cellular phone category is composed by the following 6 competing brands: LG, Motorola, Nokia, Samsung, Siemens, Sony-Ericsson.

i i As in DeSarbo and Wu (2003) we can suppose that d k , j = k , j +i , k , j with i , k , j N (0, ) . This allows us to solve the problem of finding A maximizing the

P associated likelihood L = Li where: i =1

(2) 2 2 is the individual likelihood for the projection problem. The second problem requires a different approach. We suppose (see Dillon et al., 2001) that attribute specific judgements x j , m for brand j can be decomposed into a brand preference effect j and an attribute specific evaluation j , m . We can suppose that for each observed i individual i , it holds x j , m = j , m + j +i , j , m where i , j , m N (0, ) . This allows us a likelihood representation of the problem, with individual likelihood given by:
j k

LP ( A, | X i , i ) = i

( ij ,k d ij ,k ) 2

2 2

LD (j , m , m , B j , | X i ) = i
j =1 m =1
I

1 2 2

( x ij , m ( j , m m + j )) 2

e
(3)

2 2

D Maximization of L = Li gives the best choice of matrix = ( j , m ) of genuine attribute i =1

based representation of brands, and vector B = ( j ) of holistic evaluation of brands. Note that LP and LD are defined by different psychological underlying processes. This i i (Ramsay, 1980; MacKay et al., 1995) allows to consider (i , j , m ) and (i , k , j ) independent and to evaluate a joint likelihood LC . The individual likelihood LC is formally the product i LP = LP LD but the definition of distances d k , j between projected brands must discount the i i i holistic terms. The new definition of d k , j is obtained modifying equation (1) and gives:
m m i d = x ij , m j m ,t xk , m k m ,t t =1 m =1 m =1 Consequently the complete likelihood of the problem is: i j,k T

(4)

i i i where X = ( x j , m ) is the I J M matrix of attribute based data, X = ( x j , m ) is the i th

L( A, , B, , | X , ) = LD (, B, | X i ) LP ( A, | X i , i , B) i i
i =1

(5)

i = ( ij , m ) store the I J ( J 1) / 2 dissimilarities judgements.

J M

i block pertaining individual i , and with the same notation = ( j , m ) and

RESULTS Figure 1a and Figure 1b (Appendix B) are the PCA maps obtained from attribute-based ratings respectively in the cellular phone and notebook product categories. The first PCA component accounts for about 90% of variance in the case of cellular phones and for more than 95% in the case of notebooks. These results are consistent with Dillon et al. (2001) even with a different measurement of perceptions. The two situations are non informative in terms of attribute specific performance for each brand. As the maps show the attributes arrows have all the same orientation. Therefore brand ratings dont indicate why some brands are most liked, as the first dimension provides only a connotative meaning. Conversely, Figure 2a and 5

2b (Appendix B) show a clear and informative configuration of brand positions along attributes. In fact disentangling general brand impressions (i.e. brand image) the configuration of brands and attributes spreads out to explain the specific impact of attributes on each brand. The following two examples may clarify the implications of our result. Acer is a leader in the national notebook market used in this study, however the PCA map (Figure 1b) fails to assign a set of favourable brand associations. Figure 2b explains the origin of this lack of brand personality, as remarked by the holistic rating calculated in Table 2b. In the same line, Nokia seems to be the favourite cellular phone brand in the PCA map (Figure 1a). However, table 2a fails to show the highest holistic rating for this brand, and map 2a fails to show a well defined positioning on most attributes. CONCLUSIONS The objective of the study was to offer a perceptual mapping approach that depicts the real perceived brand positioning along a set of attributes. In fact, perceived brand performance on one or more attributes may lack revealing diagnostic power given the spurious intertwinement among the actual brand-specific associations and the overall global brand impressions. The main contribution of the study is the simultaneous decomposition of a brand rating into its two main components, and the projection of the unbiased rating in a twodimensional space, in which distances between objects reflect the observed dissimilarities, as the map is obtained through the joint representation of two batteries of variables (dissimilarities and attribute ratings). The model is a likelihood-based procedure which uses a parsimonious set of parameters, as compared to other similar models in literature. The implemented model has important implications for the analysis of brand equity. By showing the actual brand ratings, our method provides a means to assess the uniqueness and strength of brand associations (Dillon et al., 2001). The cognitive representation offered by the map helps us to reach a better comprehension of both the specific impact of each attribute on the brands, and the role played by the semantic structure underlying the measurement scale used in this study. LIMITATIONS AND FUTURE DIRECTIONS The model presented in this paper focuses on the purification of perceptions with respect to general brand impressions bias. Other sources of biases can be analyzed and their effect could be disentangled as well. A major bias to take into account is given by the idiosyncratic perceived importance of attributes assigned by consumers. This might be a strategic issue to i classify heterogeneity, operatively changing the expression x j , m = j , m + j +i , j , m into the i i expression x j , m = j , m m + j +i , j , m . This formulation could be convenient but the estimation problem could be difficult to address. Another direction suggested by the study is the use of other measurement scales. Our study focuses only Aakers brand personality, but some other theoretical frameworks of brand equity measurement should be taken into account. Finally, this study focuses on high-involvement product categories while could be of interest to consider routine or impulse products.

APPENDIX A
QUESTIONNAIRES STRUCTURE

BRAND PERSONALITY QUESTIONNAIRE


Strongly disagree Strongly agree

X is a brand:

Down-to-earth. Honest..... Wholesome.. Cheerful Daring . Spirited. .. Imaginative. Up-to-date....... Reliable...

Successful.. Upper class..... Charming... Outdoorsy.. Tough.


Intelligent...

ATTRIBUTE-FREE QUESTIONNAIRES a. CELLULAR PHONES


Very dissimilar Very similar

1. Nokia and Motorola ............... 2. Sony-Ericsson and Samsung .. 3. Siemens and Nokia .

4. LG and Motorola..... 5. Siemens and Sony-Ericsson 6. Samsung and Asus .. 7. Siemens and LG .......... 8. Apple and Nokia...................... 9. Toshiba and Sony-Ericsson .... 10. Nokia and LG.................................. 11. Sony-Ericsson and Motorola ... 12. Motorola and Siemens ..... 13. Samsung and Motorola 14. Samsung and LG . 15. Nokia and Sony-Ericsson ............

b. NOTEBOOK COMPUTERS
Very dissimilar Very similar

16. Acer and Sony ....................... 17. Hewlett-Packard and Apple .. 18. Asus and Acer ... 19. Toshiba and Sony... 20. Asus and Hewlett-Packard 21. Apple and Asus . 22. Asus and Toshiba ...... 23. Apple and Acer...................... 24. Toshiba and Hewlett-Packard .. 25. Acer and Toshiba .......................... 26. Hewlett-Packard and Sony 27. Sony and Asus ............... 28. Apple and Sony . 29. Apple and Toshiba . 30. Acer and Hewlett-Packard .........

APPENDIX B
THE MAPS Figure 1a (cellular phones): Principal Component Analysis

Table 1a: Importance of components:


Standard deviation PC1 2.414 PC2 0.6893 PC3 0.3416 PC4 0.14513 PC5 0.14189 PC6 8.34e-16

Proportion of Variance Cumulative Proportion

0.902 0.902

0.0735 0.9756

0.0181 0.9936

0.00326 0.99688

0.00312 1.00000

0.00e+00 1.00e+00

10

Figure 2a (cellular phones): Model map (joint unbiased perception map)

Table 2a: Brand image (holistic) ratings


Nokia 4.552316 Motorola 3.324714 Sony-Ericsson 4.485089 Samsung 5.182024 Siemens 2.929133 LG 5.749445

11

Figure 1b (computer notebook): Principal Component Analysis

Table 1b: Importance of components


Standard deviation Proportion of Variance Cumulative Proportion PC1 2.355 0.954 0.954 PC2 0.451 3 0.035 0 0.988 8 PC3 0.18073 0.00562 0.99438 PC4 0.14834 0.00379 0.99817 PC5 0.10317 0.00183 1.00000 PC6 8.15e-16 0.00e+00 1.00e+00

12

Figure 2b (computer notebook): Model map (joint unbiased perception map)

Table 2b: Brand image (holistic) ratings


Acer 3.487614 Sony 4.908848 HewlettPackard 4.100208 Toshiba 3.749849 Asus 3.678156 Apple 5.377646

13

REFERENCES
Aaker, David A. (1991), Managing Brand Equity. New York: Free Press. Aaker, Jennifer L. (1997), Dimensions of Brand Personality, Journal of Marketing Research, 34 (August), 347-356. Brown, Steven R. and Lawrence E. Melamed (1990), Experimental design and analysis. Sage University Papers. Carroll, J. Douglas and Paul E. Green (1997), Psychometric Methods in Marketing Research: Part II, Multidimensional Scaling, Journal of Marketing Research, 34 (May), 193-204. Church, Timothy A. and Peter J. Burke (1994), Exploratory and Confirmatory Tests of The Big Five and Tellegens Three and Four-Dimensional Models, Journal of Personality and Social Psychology, 66 (1), 93-114. Clemans, William V.(1956), An Analytic and Experimental Examination of Some Properties of Ipsative Measures, doctoral dissertation, University of Washington, Seattle. DeSarbo, Wayne S. and Jianan Wu (2001), The Joint Spatial Representation of Multiple Variable Batteries Collected in Marketing Research, Journal of Marketing Research, 38 (May), 244-253. Dillon, William R., Thomas J. Madden, Amna Kirmani, and Soumen Mukherjee (2001), Understanding Whats in a Brand Rating: A Model for Assessing Brand and Attribute Effects and Their Relationship to Brand Equity, Journal of Marketing Research, 38 (November), 415-429. Gilmore, George W. (1919), Animism. Boston: Marshall Jones Company. Green, Paul E., Frank J. Carnone, and Scott M. Smith (1988), Multidimensional Scaling: Concepts and Applications. Boston: Allyn and Bacon. Halliday, Jean (1996), Chrysler Brings Out Brand Personalities with 97 Ads, Advertising Age, (September 30), 3. Keller, Kevin Lane (1993), Conceptualizing, Measuring and Managing Customer-Based Brand Equity Journal of Marketing, 57 (January), 1-22. Keller, Kevin Lane (1998), Strategic Brand Management: Building, Measuring and Managing Brand Equity. Upper Saddle River, NJ: Prentice Hall. Keller, Kevin Lane and Donald R. Lehmann (2005), Brands and Branding: Research Findings and Future Priorities, Marketing Science Institute Special Report, n. 05-200. MacKay, David B., Robert F. Easley, and Joseph L. Zinnes (1995), A Single Ideal Point Model for Market Structure Analysis, Journal of Marketing Research, 32 (November), 433-443. Malhotra, Naresh K. (1988), Self Concept and Product Choice: An Integrated Perspective, Journal of Economic Psychology, 9, 1-28. McWilliam, Gil (1997), Low Involvement Brands: Is the Manager to Blame?, Marketing Intelligence and Planning, 15/12, 60-70 Ramsay, J.O. (1980), The Joint Analysis of Direct Ratings, Pairwise Preferences, and Dissimilarities, Psychometrika, 45 (June), 149-165. Roedder-John, Deborah, Barbara Loken, Kyeong-Heui Kim, and Alokparna Basu Monga (2005), Brand Concept Maps: A Methodology for Identifying Brand Association Networks, Marketing Science Institute Reports, Issue 3, 05-112, 41-65. Rossi, Peter E., Zvi Gilula, and Greg M. Allenby (2000), Overcoming Scale Usage Heterogeneity: A Bayesian Hierarchical Approach Journal of American Statistical Association.

14

Shocker, Allan D. and V. Srinivas (1979), Multiattribute Approaches for Product Concept Evaluation and Generation. A Critical Review, Journal of Marketing Research, 16 (May), 159-180. Sirgy, Joseph (1982), Self-Concept in Consumer Behavior: A Critical Review, Journal of Consumer Research, 9 (December), 287-300. Steenkamp, Jan-Benedict E.M., Hans C.M. Van Trijp, and Jos M.F. Ten Berge (1994), Perceptual Mapping Based on Idiosyncratic Sets of Attributes, Journal of Marketing Research, 31 (February), 15-27. Weinberger, Marc G. and Harlan E. Spotts (1989), A Situational View of Information Content in TV Advertising in the U.S. and U.K., Journal of Marketing, 53 (January), 89-94.

15

S-ar putea să vă placă și