Projecting the right image: using projective techniques to measure brand image

Introduction

Ask any audience what Bulls, Bears, Sharks, Tigers, Eagles, Broncos and Rhinos have in common, and the response “rugby league teams” will be almost instantaneous. Similarly, Canaries, Magpies, Foxes, Owls, Rams and Wolves will be quickly identified as the images of soccer teams.

Associating animals with brands or products extends far beyond the sports marketing arena. The delightful puppy used to promote Andrex , the chimps who drink PG Tips and the tiger associated with petrol sales have become part of our popular culture.

The reason for this proliferation of associations between brands and animals or other images and symbols is simply an attempt to use secondary features to distinguish between almost identical brands and products.

This paper explores the use of projective techniques for examining the transfer of characteristics from two sets of cue cards to a sample of food brands.

In recent years there has been an explosion in the number of product categories and the number of similar brands available, which has required marketers and advertisers to look more closely at how customers relate to their brands. In the FMCG industry there is the added problem of the retailer labels: many large manufacturers of branded goods have lost clear brand leadership in a number of important markets to private label products, which is particularly problematic as price and brand name are often the only distinguishing features. This has prompted a shift away from focusing on consumers wants and needs in favour of brand equity and maximising investment in existing products with line extensions (Piirto Heath, 1996). Max Blackston, of Ogilvy and Mather, says: “There’s little that distinguishes between brands these days in consumers’ eyes. It’s often very, very small, subtle things that you have to leverage” (Piirto Heath, 1990).

The proliferation of new data sources entering the market, including scanner data and single‐source databases, is making it tempting for marketers to seek their answers through consumer response to marketing effort and to track changes in market share, brand loyalty and brand switching. However, these are objective, behavioural measures that provide little evidence of the emotional and textural aspects of consumer response. Although standard quantitative techniques can be used to extract the many explicit, concrete, rational factors, that consumers use to evaluate products not all “buying decisions are exclusively or even predominantly thoughtful, rational decisions that are easily articulated” (Raffel, 1996). Consumer products have a significance that goes beyond their utilitarian, functional and commercial value (Erickson, 1996). This accounts for the renaissance in the use of qualitative techniques over recent years; many qualitative methods provide insight into the “whys” of consumer behaviour, explaining the relationship between a consumer and a particular branded product, retailer, or service provider (Day, 1989).

Of particular interest are those methods that resemble as closely as possible the Freudian technique of Free Association, which are collectively called projective techniques (Atkinson et al., 1990).

Projective techniques involve presenting subjects with ambiguous stimuli to which they may respond as they wish. Theoretically, because the stimuli are ambiguous and do not require a specific response, the individual is able to project his or her personality on to the stimuli. In qualitative research, projective techniques allow respondents to “project their perceptions and feelings onto some other person or object ‐ or in some other way are allowed to depersonalise their responses, and thus they feel freer to express their thoughts and feelings” (Day, 1989), thus breaking several barriers to communication: lack of awareness of repressed motivations; inability to express themselves; unwillingness to disclose certain feelings; irrationality, and subjects trying to say the right things to please the interviewer. Nevertheless, projective techniques are not infallible, as each moderator brings his or her own style to discussions and subjective judgement to the interpretation of results. Despite some specialists employing content analysis and special analytical techniques to bring some measure of objectivity to the analytical process, it remains subjective. Also, standard focus group limitations apply; they require personal interviews with highly‐trained interviewers, which is incredibly time consuming and expensive, and is impossible to replicate identically over time. Therefore, some method of using projective techniques in a more objective manner is required. Recent research has suggested that one solution may be the identification or creation of a set of stimuli that produce consistent response across individuals.

The implicit model

Heylen et al. (1995) have developed an “implicit model of consumer behaviour” which attempts to provide a systematic, objective, scientific base for data collection, data analysis and data interpretation, and to provide the link between qualitative and quantitative research. The formulation of the implicit model has made it possible to develop standardised research methods that permit researchers to delve into the unconscious, instinctive dimensions of consumer behaviour. The tools involve projective and transference mechanisms for data gathering, and the use of the implicit, dynamic model for data interpretation.

The implicit model provides both a personality and an identity for an object, be it a brand, product or company. The personality of an object is the implicit, internal features that are experienced by the primal, subconscious brain, while the identity is the explicit, external features that are observed by the rational and conscious brain and easily articulated by respondents. A brand image will be a combination of these two dimensions. According to Heylen et al. (1995), however, as brands become more homogeneous, consumers are making brand choices more on the basis of brand personality than identity. Figure 1 depicts how, in our relationships with people, products, brands, and company, we are more affected by the personality than by the identity, although both play an important role in responding to our needs.

In order to develop the implicit model, three different sets of carefully validated photographs were compiled:

1. 1

   Ascription sets ‐ to identify the perceived personality of the user/non‐user of products, brands, etc. Using these projective stimuli allows the underlying motives that may be used in purchase decisions to be identified.

2. 2

   Gratification sets ‐ to identify the ultimate psychological, social, and physical gratification provided by products, brands, etc.

3. 3

   Animation sets ‐ to identify the underlying personality, image and perceptions of a brand, company, product, etc. Using the technique of personification, whereby respondents project a personality into a stimulus, a brand can be attributed a personality.

The ascription and gratification photo sets are used to define the vectors of the implicit space. It is the animation sets with which this report is concerned and how, through the development of sets of photographs to which respondents will respond similarly and with consistency, and to which a set of characteristics can be attributed; we can identify the brand image held and consequently the motivations behind brand choice.

Method

The first stage in the investigation was to compile two sets of photographs. The first set used different animals for elements and the second set used cars for stimuli.

Animals were selected as stimuli because of their long history of popularity with advertisers. Symbols like the Energizer Bunny used by the Eveready Battery Company now have a firm place in popular culture (Fowles, 1996). The use of animal symbolism in art is also well documented (Carr‐Gom, 1995).

Car cue cards were chosen because of the well‐documented fact that consumers tend to identify strongly with cars and purchase models that are in line with their own self‐image (Foxall and Goldsmith, 1994; Grubb and Hupp, 1968). Only the car style was shown in the prompt cards. No brand information was provided.

The sets were made up as shown in Table I.

The response photographs were selected from books, magazines and promotional literature, according to a number of predetermined criteria:

1. 1

   Neutrality, avoiding anything that may bias or offend the respondent. This could be anything from the sex or age of the driver of a car to an animal in the process of killing another animal. It would be difficult to bias the results in the strictest sense of the word, as every subject will be shown the same photograph.

2. 2

   The photographs needed to be different from one another on at least one crucial element.

3. 3

   The range of pictures should cover all extremes, for example, cars from each end of the market, and from wild, jungle animals to domestic pets.

To help fulfil the above criteria, and to help in the later stage of free association, content analysis was carried out on the text accompanying the photographs to gain an idea of the vocabulary used to define them.

Following selection of the eight photographs in each set, they were prepared to be of similar sizes, numbered randomly from one to eight and displayed on identical backgrounds.

To determine the characteristics that are attributed to each photograph, the repertory grid technique was used (Fransella and Bannister, 1984; Kelly, 1955). As quantitative results were sought in the long term only the first six respondents were asked to complete repertory grids using free association. The software package Flexigrid was then used to carry out cluster analysis of the six individual data matrices. Using the most frequently mentioned constructs, a new standard grid with predetermined constructs and poles was constructed. This was then administered to a further 30 subjects.

For the second part of the study, a set of 12 brand photographs were compiled to which a further 15 respondents were asked to assign the stimulus photographs.

As brand image is important in the fast moving consumer goods market, four product categories were selected from this area: coffee, sliced bread, chocolate confectionery and cereal. Within these categories, three different brands were chosen to which one would expect consumers to hold different images. For example, one would expect Mellow Birds to have a much more convenience, little thought involved, cheap image compared with Gold Blend which is more likely to have prestige, high‐status appeal. Furthermore, in the cereals category two Kellogg’s brands were selected to determine whether respondents chose elements with similar characteristics to represent them.

The food categories were selected not only because of their commercial importance but also because it was felt that the brands would be very well‐known to a wide range of respondents.

The full selection of brands was as follows:

1. 1

   Coffee

• •

  Nescafé Gold Blend;

• •

  Kenco;

• •

  Mellow Birds.

1. 2

   Sliced bread

• •

  Kingsmill White;

• •

  Hovis Granary;

• •

  Private Label.

1. 3

   Chocolate confectionery

• •

  Cadbury’s Dairy Milk;

• •

  Nestlé Kit Kat;

• •

  Snickers.

1. 4

   Breakfast cereal

• •

  Kellogg’s Cornflakes;

• •

  Kellogg’s Bran Flakes;

• •

  Weetabix.

The 16 stimulus photographs were laid out clearly in front of the respondents prior to giving them the questionnaire; they were then shown the brand photographs and asked to select first a car photograph and then an animal picture that they think best represents that brand. Brand photographs were shown in their category sets; pilot tests indicated that even if all of the photographs were shown in a random order, respondents still drew comparisons and asked to change previous answers. Next, they were asked to indicate their purchase frequency so that it could be determined whether a “favourable” response was the result of brand purchasing. Following the presentation of all of the brands, the respondents were asked to rate the difficulty of the task.

All interviews were conducted using standardised instructions.

The objective of the study did not require any form of probability sampling, particularly as only 51 subjects were used. However, it is one of the most consistent facts in gender research that females are more emotional than males (Fabes and Martin, 1991). Achievement, autonomy and striving for control are terms regularly associated with masculinity. Femininity, on the other hand, is usually associated with interpersonal communication, communality, and the awareness and active expression of one’s own feelings. Much of the research on the issue suggests that males and females are actually equally emotional, but they express emotion differently because of different norms regarding emotional expression for men and women (Ickes and Barnes, 1978). Therefore, we would expect that male subjects would respond differently to the photographs from female subjects because they have been socialised differently. Consequently, in the two repertory grid experiments there were equal quantities of females and males interviewed. In the final brand image experiment, seven men and eight women were interviewed. This not only allowed for a fuller picture to be built up but also permitted any differences between male and female responses to be investigated at a later stage.

Results

The list of words/phases used to describe the car photographs in the six unstructured grids is shown in Table II. Where a number follows a word/phrase this indicates that the term was used by more than one respondent.

The cluster analysis for the car photographs showed high and numerous correlations between the Espace and Jeep, and the Morgan and Porsche. Nevertheless, this need not be a problem as they are also highly clustered with other cars and therefore have some characteristics distinct from each other.

Table III gives the constructs elicited from the six unstructured grids for the set of animal photographs.

Cluster analysis revealed high and numerous correlations between the Bear and Lion; however, as the Bear also clusters with the Panda on two respondents results it still has some characteristics distinct from the Lion.

On the basis of the frequency counts shown in Tables II and III and with reference to the cluster dendrograms, a total of 24 constructs for the car photographs and 18 constructs for the set of animal photographs were selected for the 30 structured grids.

From a significance analysis of the means of the construct scores, a group of defining constructs were identified for each element (see Tables IV and V).

The results from the final 15 respondents who were asked to select one car and one animal photograph to represent each brand, suggests that people do use this method of predicting brand image with discrimination and consistency.

On the coffee brands, both Gold Blend and Kenco were represented by the Porsche, thus emphasising their high‐status; while Mellow Birds was thought to be like the Mini. The ascribed personality of the Mini includes standard features, cheap, non‐technical and old‐fashioned. Respondents also selected the Mondeo to represent Gold Blend, emphasising its “everydayness” and respondents moderate feelings towards it. The same animal picture was also selected for Gold Blend and Kenco, the Lion, showing that respondents are consistent in their ranking. The rabbit was chosen to represent Mellow Birds; this element has characteristics in common with the Mini such as British and small; thus respondents are consistent in the basis of their evaluation.

Respondents chose different animals and different cars to represent each of the stimulus brands of bread. For Kingsmill, the Lion and the Porsche/Morgan/Mondeo were chosen. The Porsche and Morgan have a very similar personality, however the Mondeo is very different and represents a very moderate brand. Hovis is represented by the Jeep and Squirrel; these have contrasting profiles, for example, small versus large, weak versus powerful. The economy loaf is represented by the Chimp and Fiat, which have very similar profiles and thus the evaluation sets are the same.

The chocolate confectionery brands of Dairy Milk and Snickers are both represented most often by the Jeep; however, the choice of animal pictures is different: the Panda and Bear represent Dairy Milk, whereas the Lion depicts Snickers. Lion and Jeep have a very similar style of personality, and the Bear/Panda and the Jeep have quite similar characteristics. The Kit Kat has a rather negative brand image in contrast to the other two brands as it is portrayed by the Mini and the Cat or Chimp.

There is a “British” emphasis for Cornflakes as respondents chose the Mondeo and Squirrel to represent the brand. The positioning of Cornflakes as an everyday, family cereal would suggest that the Mondeo is a close approximation to its true brand image because of its moderate image, and characteristics such as run‐of‐the‐mill, common and reliable. Owing to the size of the brand itself, it is surprising that the Squirrel was chosen; however, the defined personality does include fast‐moving, wild and active. Bran Flakes are not depicted by either of the same stimuli as Cornflakes, despite both being Kellogg’s brands, and thus respondents appear to have based their evaluation on the generic product rather than the manufacturer brand.

Although these results do suggest a general consistency in the thinking of respondents in terms of the relative frequencies of cue cards selected, we do not know on what basis the respondents actually made their judgements. For example, that Kingsmill was represented by the Lion (Lion = king of the jungle), Kit Kat by the Cat, and Hovis by the Squirrel ‐ a possible link between nuts and granary? Nevertheless, when comparing the Economy Loaf and Hovis, we saw that the former is perceived to have only standard features, whereas the latter has additional features, which is quite true given that Hovis is a granary bread. However, this aside there is definitely potential for these stimuli sets to be used as part of a technique for measuring brand image.

A summary of the most popular picture choices is given in Table VI.

When asked how difficult they found it to select car and animal photographs to represent the brands, respondents said that it was generally harder to select an animal photograph than a car.

Of the 336 possible opportunities for recording sex differences in both sets of photographs, only 27 element‐construct combinations record differences that are significant at the 5 per cent level. The small sample size and the lack of consistent significant differences suggest that, as far as this research is concerned, gender differences are unlikely to be a problem.

Conclusions and future research

Given that the current work has demonstrated a potential for using these sets of photographs as part of a technique for measuring brand image which results in consistent responses across individuals, the next practical step would be to determine whether the selected stimulus’ unique personality is congruent with the actual image that the consumer holds about the product. This could be done by simply presenting an independent set of respondents with the structured repertory grids ‐ as used when identifying the characteristics of the stimulus sets ‐ and asking them to select the characteristics that they associate with a particular brand. By comparing the response profiles of the brand in question with that of the most dominant animal or car an empirical measure of goodness‐of‐fit could be determined.

Another practical step must be to determine if this model is appropriate to use for products other than food, particularly those that require greater involvement from the customer when choosing which product to buy.

Also, the current response sets, cars and animals, need to be developed in order to provide a greater repertoire from which respondents could select. Extra response sets are also being developed so as to provide a much richer picture.

For future work at the modelling level, some attention needs to be given to self‐concept and congruity theory in order to provide sound underpinning for any empirical findings (Grubb and Grathowl, 1967).

The relationship between the product image and consumer’s self‐image is known to determine the effect of the symbolic meaning of a product (Chisnall, 1995). It is also thought that brands have a symbolic value which, in turn, influences the consumer’s self‐image (de Chernatony and McDonald, 1992).

Congruity theory suggests that the greater the brand/self‐image congruence the more a brand will be preferred, and that this congruity can exist along a number of the dimensions of the self‐concept. Ross (1971) confirmed that: “Subjects preferred brands of products that were more, rather than less similar to their own self‐concept. The magnitude of the discrepancy between self‐concept (actual and ideal) and brand image, increased as a function of a decrease in subject’s preferences for those brands”. Ross (1971) also found that “preferences in consumption were actually more closely related to actual self‐concept than to the ideal self‐concept for each of the brands in the product categories researched”.

These are clearly issues that need to be addressed in any future work.

As the world changes, new markets open up for consumer product firms. Multinational corporations want to take advantage of these changes, and thus some way of measuring brand‐image in different countries is becoming increasingly important.

By conducting research in the appropriate countries, a personality can be attributed to the photograph specific to that country. This is a particularly suitable method to transfer overseas as there is no written language in the ascription sets.