The purpose of this paper is to investigate the relationship between the sales volume of a firm and its brand

image. Consumersself-perception and perception of brand image, with respect to congruency models, have a strong influence on their behavior in the marketplace. Therefore it is expected that the fluctuations (the authors use fluctuation and variation interchangeably) in image attributes will explain the fluctuations in sales figures. In order to test this hypothesis, consecutive surveys were carried out, on a monthly basis to collect image data. Factor analysis was performed on the image attributes over time and three main image factors were attained. To determine the net effect of image attributes on sales, multiple regression analysis was performed, using the time series data, and all three image factors were found to be significant in the model. Article Type: Research paper Keyword(s): Econometrics; Brand image; Tracking; Sales turnover; Sales forecasting; Consumer behaviour. Journal: Journal of Product & Brand Management Volume: 12 Number: 4 Year: 2003 pp: 237-250 Copyright MCB UP Ltd ISSN: 1061-0421 Introduction Behavior deals The study of marketing and consumer behavior deals with the factors that affect the purchase and use of goods and services such as availability, advertising, image, and consumer characteristics (Engel et al., 1986). In situations where consumers have a choice in how they behave, their behavior is governed by many factors. These include their needs and desires, their attitudes and expectations, their understanding of what is

available, their financial resources and their decision processes (Bromley, 1993). The relationship between the consumer and the brand consumers perception of that brand is the key to brand acceptance. The strength of the relationship between the consumer and the brand will reflect the fit between the consumers own physical and psychological needs and the brands functional attributes and symbolic values, as perceived by the consumer (Hankinson and Cowking, 1993). According to Maslows hierarchy of needs, the functional attributes of a fast moving consumer product can, at most, satisfy the safety needs of a consumer. On the other hand, consider a carbonated soft drink whose marketing strategies are image oriented. That specific product is also able to satisfy higher levels of needs besides physiological needs, if the consumer believes that it is up-to-date, gives him/her enjoyment anytime, revitalizes him/her, helps him/her to look his/her best and is a drink he/she is happy to be seen drinking in public. Brand image, which usually includes the products name, its main physical features and appearance (including the packaging and logo), and its main function(s), is the key to answer the question of how the consumer chooses among alternative brands after information-gathering processes of buyer behavior. Alternative evaluation is how the consumer uses this information to evaluate the options and arrive at a brand choice. The perspective focusing on psychological attributes of a product of consumer behavior states that consumers often choose products, services and activities over others because they associate these attributes with a certain lifestyle, a self-image or an ascribed status. They attempt to preserve or enhance their self-image by buying products that they believe are congruent with that self-image and by avoiding products that are not (Sirgy, 1982). For that reason, marketing academicians and practioners see the symbolic image of products or services as more important in their success than their physical characteristics and attributes (Aaker, 1991). Therefore, within such a framework it seems possible to mention the direct effect of image attributes on sales figures, which would also support the statement that the role of self-concept and brand image is recognized in all elements of the marketing mix, mainly in promotional strategies (Kotler, 1991). The paper is organized as follows. The next section overviews brand image and sales modeling issues followed by the image congruency models that in a way constructs the basis of the model described in the following section. The sample and the field study are also explained in detail in the same section. The fifth section reports the empirical findings and the last section concludes. Brand image and sales modeling Studies Numerous studies in the literature have been devoted to explaining brand/product preference (Hong and Zinkhan, 1995; Zinkhan and Hong, 1991; Onkvisit and Shaw, 1987; Sirgy, 1985), brand evaluation (Graeff, 1996, 1997b), purchase intention or usage (Mehta, 1999; Malhotra, 1988) in terms of image congruency models. In a study where the relation between self-concept and advertising effectiveness is investigated, Hong and Zinkhan (1995) conclude even though the brand memory is not mediated by the extent to which advertising expressions are congruent with viewers self-concept, brand preference and purchase intention are found to be influenced by the self-congruency of an ad. Sirgy (1985) aimed to test the hypothesis that the effects of selfcongruity and ideal congruity on purchase motivation are additive. The results of the study showed that ideal congruity and self-congruity significantly influence purchase motivation. Malhotra (1988) investigated the effect of congruence on preference with an integrated perspective and as a result supported the hypothesis. Graeff (1996) concluded that consumers evaluations of publicly consumed brands are more affected by the congruence between brand image and ideal self-image than actual self-image, whereas actual and ideal congruence have equal effects on consumers evaluations of privately consumed brands. From a different perspective Murphy (1990) described the life cycle of a product in three stages, namely proprietary, competitive and image. According to the author, in the first stage brand is perceived as unique, then in the second stage it gets into competition with other brands on the basis of its functional characteristics. In the maturity stage the image of the product or symbolic values plays a greater role in competition and also sales. Conclusions in the literature Although some of the above mentioned studies draw conclusions about the effect of brand image and brand/self-image congruence models on sales, none of them investigates this relation directly. A recent work of Malhotra et al. (1999) also underlines the need for linking brand image strategies to market performance of the

brand, like sales or market share. Therefore this paper tries to model a brands sales with its perceived brand image. Investigation of sales figures is generally considered as part of market response models. The effects of marketing instruments and household specific characteristics on various marketing performance measures such as sales, market shares and brand purchase are heavily studied in the literature (for a general view of marketing models see Leeflang, 1995 or Franses and Paap, 2001 and for the evolution of marketing response models see Leeflang and Wittink, 2000). These models are dependent on revealed preference data also referred to as hard data instead of soft (preference, satisfaction, etc.) data coming from stated preferences. But, as Leeflang and Wittink (2000) propose and Wedel et al. (2000) discuss, that the integration of two model types is necessary. A recent treatment of combining the two modeling approaches can be seen in Hensher et al. (1999). But this paper should be considered as a modeling approach that tries to combine revealed preference data models with primary data collected by surveys, instead of stated preference data. Modeling In the last two decades, the effort to model sales of a brand is based on UPC scanner data or household panels. The purpose of this type of modeling is mainly to support marketing decisions with the evidence from the constructed models (Bucklin and Gupta, 1999). Decisions concerning the marketing mix instruments, such as price, advertising expenditures, features, display, coupons and interactions between these instruments, are made, taking the results of these models into consideration. These models may serve as marketing decision support systems when they are predictive models. On the other hand, similar models can be constructed just for descriptive reasons, intended to describe some processes of interest (Franses and Paap, 2001). In this sense, the model constructed in this paper can be considered as a purely descriptive model, concentrating on the effect of brand image on sales. Unfortunately, the model excludes the marketing mix variables because of the reasons stated in the section describing the limitations of the study. Therefore, the authors are only dealing with the relation between brand image and sales. Image congruency models Self-image Each individual has a perceived self-image as a certain kind of person with certain traits, habits, possessions, relationships and ways of behaving. This self-concept or self-image can be defined asthe individual as perceived by that individual in a socially determined frame of reference (Loudon and Bitta, 1988). Preferences may actually develop for certain brands because the consumer perceives them as reflecting his/her own selfimage. Or some brands may be desired because the consumer views them as projecting an image that he/she presently does not possess but aspires to have (Schiffman and Kanuk, 1983). So it is possible to claim that consumers self-perception can have a strong influence on their behavior in the marketplace. Self-concept is composed of multidimensional characteristics and includes physical as well as psychological attributes and interacts with the various roles a person must take on (Mehta, 1999). This leads us to a multiple component perspective of the self. Loudon and Bitta (1988) brought together several approaches, mainly based on Sirgys (1982, 1985) work. Single component perspective depends on actual-self, the perception of oneself, as one believes he/she actually is. According to the authors the ideal-self may be defined as the perception of oneself as one would ideally like to be. Social-self (the perception of oneself as one believes others actually perceive him to be), ideal-social-self (the perception of oneself as she/he would like others to perceive him/her) and the expressive-self (the ideal-self or the social-self depending on situational and social factors) were added to the single component perspective to extend the viewpoint (see Figure 1). Consumers tend to compare their self-image with the product images in the pre-purchase stage of the decisionmaking process. This kind of comparison is known as self-image/product image congruence process. Selfimage congruence models predict that products will be chosen when their attributes match some aspect of the self (Onkvisit and Shaw, 1987). The simplest form of this model can be seen in Figure 2. Many findings support the argument that consumers prefer, intend to purchase, or actually use brands with images they see as being congruent with their actual self-concept (Malhotra, 1988). Other studies also report that consumers are more likely to prefer, intend to purchase or use brands with images that they see as being congruent with their ideal self-concept (Malhotra, 1988). Brand/social-self-image relationship is not strongly supported by studies in the literature (Sirgy, 1982). However, the relationship with brand/ideal-social-self-image is supported by empirical evidence.

Brands Surely, a product that is perceived as fitting an individuals self-image has a greater probability of being selected. That is why marketers try to create images for their brands so that they are positioned to fit a distinct market segment occupied by no other brand. They strive to create a brand image that is similar to (congruent with) the self-image of the target consumers (Aaker and Biehl, 1993; Kapferer, 1992). But it cannot always be assumed that consumers will buy products whose personality characteristics match their own characteristics. This assumption is generally satisfied for expressive, image-oriented products. This idea leads us to a wellknown vicious circle. Do people buy products because they are seen as connected to their self-image or do they assume that these products must be connected to their self-image because they have bought them? How a persons self-concept changes over time is another factor that has to be taken into consideration. This complicated process of self-concept development lasts long and hardly changes. Psychologists argue that a persons conception of self displays a high degree of consistency, particularly in the short term. But the idea of situational-ideal-self-image may cause fluctuation in image attributes. This can be supported by the concept of you are what you consume just like Belk (1988) postulated in his review of self-concept, we are what we have and possess. Therefore, it is reasonable to expect variation or fluctuation in measured image attributes over time, when observations from a heterogeneous sample, including people of different ages with different educational backgrounds and socio-economic status, are aggregated. Then one might expect fluctuations in the sales of a brand in connection with the changing image perceptions. Model Physical attributes The previous sections describe mostly how image as well as physical attributes affect preference, purchase intention and usage, and consequently sales. When self-image changes by the influence of several factors such as culture, sub-culture, family, social class, consumption situation and expectations, the result of the congruency process will be different from the previous one. What is meant by self-image change can be directly taken from a study of Graeff (1997a, p. 50): Consumers attempt to behave in ways that match the social expectations of the others within a particular situation. For instance an individual can assume a role of a father, businessman In each situation, the consumer attempts to behave and express an image that will best lead to a favorable impression of others in that social situation. Use of particular products is one way consumers are able to project the appropriate image for a given situation. For that reason, it is reasonable to expect a change in sales figures associated with the above-mentioned changes. This claim is based on the assumption that when a characteristic or image dimension is attributed to a brand, it is in a way related to a buyers actual, ideal, social or situational-ideal image and as a result it will be purchased. The assumption seems more reasonable when the product category being studied belongs to FMCG, where the products are similar in terms of their functionalities but differentiate themselves with imageoriented activities. Thus the following non-directional hypothesis can be tested. 1. H1. The fluctuation variation in image attributes is significantly correlated with the fluctuation variation in the sales figure. Observational data The authors used observational data and an econometric methodology to gather empirical evidence for the hypothesis. In order to measure the relationship between image attributes and the sales figure, the authors took an image data of 23 image attributes (Table I) and sales figures from one of the leading beverage producers. The attributes included in the survey can be divided into several sub-categories, as functional and psychological attributes. Over the survey period, some image attributes were deleted and some were added following the companys directions. Only 16 image attributes remained on the survey throughout. The attributes

were measured using a five-point Likert scale with 1 corresponding to definitely not suitable for the brand and 5 definitely suitable for the brand. No self-image data was collected. The selected product was the leader of its category in the beverage industry and the company had an enormous budget to spend on promotional activities, especially advertising. The company had used all possible media including billboards, television, radio, print media and store displays, flyers etc. to advertise the product. Campaigns lasted for a long time, a new one starting right after the previous. This is congruent with the companys general preferences regarding campaigns. For that reason, we can claim that a change in promotion strategies or the frequency of the campaigns has no influence on the hypothesis tested here. Thirtysix consecutive face-to-face surveys were carried out every month from January 1997 to December 1999 in an average of 800 households. The respondents are randomly selected from urban areas. People above the age of 12 are surveyed. Of respondents 51 per cent were male. After 36 months of field study the average frequency distributions of occupation and literacy were found as follows. Of the sample 43 per cent were salaried workers, 25 per cent self-employees, 20 per cent students and 12 per cent retired people where 52 per cent were graduated from primary school, 24 per cent were graduated from high school, 8 per cent were university graduates and 16 per cent literate without diploma. The demographic distribution of the surveys remained approximately constant over the sample period and it reflected Turkeys urban population distribution properly. Over the sample period the percentage of the respondents who saw any type of advertisement varied between 65 per cent and 75 per cent. However, on average, the respondents product usage rate was 83 per cent in the last six months, 57 per cent in the last month and 36 per cent in the last week. This implies that it is reasonable to rely on the respondents evaluations about the product. Keeping all input from previous discussions in mind, we can say that aggregating observations from different respondents reflecting a similar demographic distribution throughout the survey period offers us the advantage of capturing variation in image attributes, since it is nearly impossible to expect a change in self-perception in the short-run. So, if the movement of sales and brand image is traced in time it is possible to find a significant relationship. We formulate our model in this manner: The variation in image attributes over time is significant in explaining the variation in sales. Following is the proposed model:(see equation 1) where Y t is the monthly sale of the firm at time t; X it is the ith image attribute at timet and t is the random disturbance term. If there exists other dynamic effects that are not captured by the model they will show themselves in the ill-behaved error term. This is likely to occur when the fact that variation in sales figures is being explained by the fact that only variation in image attributes is considered. Empirical findings The dependent variable exhibits seasonal fluctuations Since monthly sales, the dependent variable, is a time series, it can exhibit seasonal fluctuations, as is the case in many sales figures in the beverage industry. The authors tested the existence of seasonal fluctuation first and the existing seasonal pattern was removed from the monthly sales by seasonal adjustment process. We also calculated the three monthly moving averages of the image variables in order to get smoother movements of image attributes. This was also done to reduce the effect of possible severe fluctuations in the image attributes measured in the consecutive survey. First, equation (1) was estimated by OLS using 16 image attributes as explanatory variables but the estimated coefficients were found insignificant and some estimates had the wrong signs, although the model as a whole was significant. Facing a high degree of multicollinearity condition index is 103.45 we performed a principal component factor analysis on the 16 image attributes. The result of the factor analysis is reported in Table II. The first three eigenvalues explain cumulatively 92.16 per cent of the total variation among the 16 image attributes. Factor analysis

Applying factor analysis on the 16 attributes also brought us the advantage of determining image factors that will lead us to a more detailed investigation in further research. The three factors were named as follows: selfimage, taste and image of product as a whole. The attributes gathered under the self-image factor included phrases related to our self-concept definition. But it was impossible to define which attribute was related to the ideal-self, the actual-self, the social-self or the ideal-social-self because this study did not include any questions examining respondentsself-perception. The second was the taste-image factor. It was possible to expect this factor to exhibit only small changes over time, when no change in production process was taken into consideration. The third factor was the one that gathered the rest of the variables. Some of these are related to the self-image and some to the taste-image factor. The depiction of the factor scores across time can be seen in Figure 3. After the application of factor analysis equation (1) could be revised as:(see equation 2)where Y t still represents monthly sales and F i t are the image factors. Variables of interest First of all we tested whether the variables of interest were stationary or not, since we are working with time series data. Well-known ADF tests and their generalizations to the case of structural breaks a structural break could be present were used to test the unit root null hypotheses. As a result, the authors concluded that all the variables were stationary at 0.10 level, in other words I(0). In order to test our hypothesis that variation in image attributes is able to explain the variation in the sales figure, we used OLS method to estimate the slope coefficients of equation (2). The first regression revealed the presence of autocorrelation (Durbin-Watson statistics is 0.483). Probably this points out a specification bias because of the excluded variables. This is discussed in the conclusion. The results are summarized inTable III. Standardized beta coefficients were given. The F-statistic was 63.885, which indicated that the estimated equation was significant at 0.01 level. All the estimated coefficients were significant at 0.05 level. The Adjusted R 2 was 0.896 which is considered to be a high coefficient of determination. Self-image factor had the highest standardized coefficient (0.693) as the first independent variable and was positively related to monthly sales. This meant that an increase in peoples acceptance of a products congruence with the stated attribute would also increase the sales of that product. The taste and image of product as whole factors were also significant in equation (2), having standardized coefficients of 0.388 and 0.218 respectively. The authors also tested the long-term effects of image factors on sales but the coefficients of the lagged independent variables appeared to be insignificant. Therefore, those results are not reported here. Conclusion In this study, self-image/product image congruency models of consumer behavior literature were examined and the hypothesis of a significant relationship between the fluctuations in image factors and the sales of the selected product was tested. Information Consumers evaluate the information gathered through out the pre-purchase decision process to decide which brand of a product to purchase. One source of this information is the comparison of perceived self-image and brand image. They tend to choose the brands whose perceived images are similar to the consumers actual, ideal, social, ideal-social and situational-ideal-social images. This means sales, from a producers point of view. The changes in the perception of brand image may cause changes in brand preference. The reality that the variation in image attributes affects the variation in sales should be kept in mind while creating new marketing strategies. If the consumers believe that the usage of that brand is fun; is best for when he/she is having fun with his/her friends; will give him/her enjoyment anytime; will always be worth what it costs and will be recognized as the best, then it will be preferred in other words will be purchased. As mentioned before, a brand perceived with several image attributes is preferred because those attributes are congruent with the consumers different image reflections. If the theoretical discussion above is summarized in terms of the research hypothesis, the agreement that the brand holds those image attributes is positively related to the sales of that brand. Regression analysis

As a result of the regression analysis of sales on image factors for testing the hypothesis, the relationship is found to be statistically significant. This finding is consistent with the expectation stating that if the perception of brand image is congruent with the consumers actual, ideal, social, ideal-social and situational-ideal-social image then the brand will be purchased. The signs of the independent variables the image factors obtained as a result of the principal component factor analysis are positive (0.693 for self-image factor, 0.388 for taste image factor and 0.218 for the factor image of product as a whole) and significant. But the appearance of such a high coefficient of determination should cause us to investigate the model much more carefully. The variation of image factors explains 89.6 per cent of the variation of sales figure and only 10.4 per cent is left to the effects of distribution channel, advertising, brand loyalty and other factors. This seems a problem for the generalization of this result. For generalization of such a claim, the consistency of this high coefficient of determination should be investigated through out other studies both in beverage industry and the remaining industries. But the lack of consequent image tracking surveys in Turkey is a limitation for a detailed study and cross-examination of the claim. Limitations Finally, the authors would like to discuss the limitations of the study. As the section on modeling brand sales overviews, marketing mix instruments should also be included in the model in order to investigate the long- and short-term effects of advertising, the possible effects of other promotional activities. The model described in the third section should be modified by adding advertising expenditure at time t, to test for the immediate effect and lagged values of the same variable to test the long run effects. But the producer of the product refused to reveal the information on marketing mix instruments for the specified sample period because of confidentiality of the data. As a consequence, the authors were forced to include only the image data in their model. Inevitably, the disturbance term of the first model represented the effect of variables not included in the model. This effect is reduced by adding a moving average term to the model but the authors believe that further research should be carried out in order to test the exact effect of brand image on sales altogether with the marketing mix instruments. Executive summary and implications for managers and executives Get more sales invest in brand image Too little is understood about the precise nature of the link between investment in the brand and real returns by way of sales. At times marketers see investment in the brand as almost an act of faith since they are unable to demonstrate (and certainly are unable to quantify) the positive effects that this investment has on the firms sales revenues. Such a situation exposes marketers to the preference for prosaic sales promotions rather than the poetry of brand development. With sales promotion or even direct marketing there is a direct and clear link between the money spent and the sales achieved as a result of that expenditure. Nevertheless, even the driest of accountants can see that brands have value not just because on occasion other businesses have been willing to pay large premiums to obtain a brand but because they observe how, in life as a consumer, people use brands as a means of making a purchase choice. Here, Ataman and lengin examine one important aspect of this issue how brand image is used by consumers in purchase decisionmaking. The authors begin with the argument that brand image is the key to answer the question of how the consumer chooses among alternative brands. This is followed by an examination of the way in which consumers actually apply brand image with the contention that consumers prefer, intend to purchase or actually use brands with images they see as being congruent with their actual self-concept. What I am, what Id like to be and what I say I am Ataman and lengin discuss the idea of self-concept, dividing it into three aspects: the social self (what I am), the ideal social self (what Id like to be) and the expressive self (what I say I am). We can see that these different aspects of self-concept assuming brand image is used in the expected manner are central to the impact of brand image on sales. The particular choice of the consumer may depend on any or all of these aspects.

Thus we can recognise that any given brand may be purchased by one consumer because it reflects selfimage, by a second because it reflects a desired image and a third by way of expressing a transient feeling about self. And we can also speculate that different brands have different distributions of buyers across these categories. Indeed, marketers recognise this distinction when they describe some brands as aspirational the expectation is that most buyers will be reflecting their ideal social self rather than actual social self. What marketers do not always appreciate is that decisions about being an aspirational or ordinary brand are, in essence, positioning decisions. And, since we also identify and target specific groups of consumers rather than the entire market, our brand positioning decisions must be consistent with brand image. Post- and pre-purchase effects of brand image An interesting observation in this article is the chicken and egg situation do people buy because the image fits their self-concept or is the link with self-image a post-purchase rationalization? The impact of this on strategy lies in the emphasis we place on two aspects brand awareness and brand loyalty. If it is awareness that makes the consumer buy (it fits the individual consumers self-concept) then the focus is on achieving higher levels of awareness within the target audience targeting buyers and non-buyers. If, however, the link with self-image happens after purchase then brand advertising and other promotion needs to reinforce the benefits that follow from purchase and on the reduction of dissonance. I suspect that researching this aspect of the link between sales and brand image will prove difficult given the need for a degree of confidence in consumer responses to any research questions. It may be that experimental approaches are needed to build a model that can be tested in a real consumer environment. Given that Ataman and lengin demonstrate a link between brand image attributes and sales it will be helpful to gain more understanding of the way in which this link develops does brand image impact on repeat purchase and loyalty or does it impact on awareness? Ataman and lengin also note that self-concept is not a constant and changes over time. Thus todays fit between brand image and self-image may not apply tomorrow. Again this challenges marketers to keep a regular check on the way in which brand image plays in target markets. In addition there is a need to renew the market as changes in self-image among current customers draws them away from our brand. Going back to the chicken and egg question, this occurrence suggests that brand awareness should receive greater strategic attention. Brand image does directly affect sales Regardless of the questions raised by Ataman and lengins research, it does show that investment in brand image and in the attributes of brand image provides a positive sales benefit. And this benefit can be quantified (within the expected limitations of all consumer research) providing additional ammunition in the argument that investing in brands is, for consumer goods businesses at least, the best way to deliver long-term profitability and real increases in business and shareholder value. We can also see that investing in research at the individual product level is a requirement if we are to appreciate the value added by brand management. It is no longer the case that directors and shareholders will accept the benefits of brands as an act of faith they require hard evidence that this investment will pay off.