Você está na página 1de 10

RESEARCH NOTE

COMPARING METHODS OF MEASURING BRAND PERSONALITY TRAITS


Jenni Romaniuk
Brand personality development has become an important marketing objective. Accompanying this objective is increased interest in brand personality measurement. In split sample tests across three categories,
the personality traits generated using the ve-point scale in Aaker (1997) were compared to those generated
by a quicker to administer, free choice association method commonly used by practitioners. The results
show both methods ranked brands similarly for each trait and obtained higher responses from brand users.
However, the free choice method discriminated more between brands and generated a greater variety of trait
associations. Therefore, if the objective is to understand the whole market, including competing brands and
nonbrand users, the ndings suggest a free choice method is a more appropriate than the ve-point scale.

An important facet of consumer-based brand equity (CBBE)


is the knowledge that consumers hold about the brand
(Keller 2003). This knowledge can be utilized by the consumer during a choice occasion. This circumstance makes
the understanding and measurement of brand knowledge
a key marketing activity. Brand knowledge is multidimensional and can include features, benets, attitudes, and
personality traits associated with the brand (Keller 2003).
This research concentrates on the measurement of personality traits, which are brand attributes that reect human
qualities, such as intelligent or fun. These traits combine to
represent the brands personality. Aaker (1997) presented
a framework that included 42 different traits, measured
using a ve-point scale. However, if a marketing or brand
manager wanted to compare the personalities of even ve
brands, a consumer would make 210 different assessments.
This method would be onerous on the consumer and costly
for the marketer. Therefore, any approach that improves
the efciency of collecting brand personality information
would be very useful for practitioners.
In the market research industry, one of the most common
methods to measure brand image is via a free choice approach where brands are not rated, but simply associated (or
not) with provided attributes (Brown 1985). This method is
attractive because it is quicker to administer, as the consumer
Jenni Romaniuk (Ph.D., University of South Australia), Head
of Brand Equity Research Program, Senior Research Fellow,
Ehrenberg-Bass Institute for Marketing Science, University of
South Australia, Adelaide, SA, jenni.romaniuk@marketingscience
.com.

does not separately address every brandattribute pair. However, although a free choice approach has strengths, it also
has weaknesses. These weaknesses are the (potential) lack of
sensitivity of a yes/no measure compared to a more detailed
scale, and the possible underreporting of beliefs. Both of
these issues can compromise the quality of the brand personality data collected. For example, if a free choice method
leads to fewer people linking a brand to a trait, a manager
may inappropriately conclude that a brands personality is
weak and take (unnecessary) action to remedy this.
Empirical testing of this alternative, commonly used
method in the context of this research is important as one
of the barriers to industry adoption of academic research
is the lack of attention to the practical implementation
(Katsikeas, Robson, and Hubert 2004). In brand personality
research, whereas scales have been validated as an appropriate method, a free choice association is more attractive
to practitioners to implement. This research makes an
important contribution by bridging that gap between ideal
academia and desirable practice. This research compares the
trait associations generated by a free choice approach with
those generated by the scale in Aaker (1997).

BRAND PERSONALITY TRAITS


The idea of people endowing brands with human characteristics or a personality has been put forward in the consumer
behavior literature for a number of decades (Belk 1988;
The author thanks her colleagues and the Institutes corporate
sponsors for their support of this research.
Journal of Marketing Theory and Practice, vol. 16, no. 2 (spring 2008), pp. 153161.
2008 M.E. Sharpe, Inc. All rights reserved.
ISSN 1069-6679 / 2008 $9.50 + 0.00.
DOI 10.2753/MTP1069-6679160205

154

Journal of Marketing Theory and Practice

Sirgy 1982). Aaker (1997) argued that the absence of a consistent operationalization has held back the contribution of
brand personality to consumer preference. She developed
a measure of brand personality that consisted of 42 traits,
which underpin the dimensions of sincerity, excitement,
competence, sophistication, and ruggedness. The value and
popularity of this operationalization can been seen in the
200+ citations that Aaker (1997) has recorded by Google
Scholar since publication.
In Aaker (1997), the relationship between these personality traits and each brand was measured on a ve-point,
verbally anchored scale where 1 = not at all descriptive and
5 = extremely descriptive. Most of the subsequent research
in the brand personality area has focused on applying this
measure to different contexts, or testing the traits and
dimensions for validity and reliability (e.g., Azoulay and
Kapferer 2003; Wee 2004). No studies have examined the
use of possible alternatives to the ve-point scale.
In commercial market research, a common method
for measuring brand associations is a binary free choice
approach. Here, consumers are asked to indicate which
brands in a market they associate with a series of attributes.
For example, the attribute of good value is provided and
consumers are asked to indicate which brands (typically
from a provided list) they associate with this quality. This
approach differs from a rating method in that each brand
is only classied as associated or not. Being easier for consumers to complete, it takes about half the time to administer (Driesener and Romaniuk 2006). This makes it more
cost-effective. Further, those conducting the interviews also
prefer to use free choice over other approaches (Joyce 1963).
These ndings indicate why the binary free choice method
is an attractive option in commercial research.
To increase the efciency and simplify the data collection
process, personality traits are often added to free choice
attribute batteries with more traditional image attributes
(such as those related to quality or price perceptions). In
this research, the two methods (free choice and Aakers
ve-point scale) are compared on four dimensions that inuence the number and variety of traits that are generated
across brands and key customer groups. These are (1) the
order of brands, relative to competitors, which is used to
assess which brands are performing best on an attribute/
trait; (2) the difference between brand user and nonuser
responses, which is used to assess who is being reached
by marketing activity; (3) the degree of discrimination
between brands, which is an indication of method sensitivity; and (4) the extent of brand associations generated,
which captures the degree to which the method captures

the wider range of brand associations. Therefore, the results


produced by each dimension affect the interpretation of a
brands personality relative to competitors.

Brand-Level Correlations
When results are analyzed, the level of association with
an attribute (or trait) compared to other brands is a key
area of interest, as this provides insight into performance
relative to competitor brands. Therefore, any differences
in brand rankings between the methods would be of concern. When rating, consumers discriminate between just
agreeing and strongly agreeing that the brand has a quality.
This condition could lead to differences in brand ranks as
stronger brands can be identied using a ratings method.
However, prior research has found high correlations when
the brand-level results from the two methods are compared.
Joyce (1963) found the brand rank-order correlations to
average 0.93 and 0.87, and Barnard and Ehrenberg (1990)
reported an average correlation across eight consumer
packaged goods categories of 0.86. More recently, Driesener
and Romaniuk (2006) found an average correlation of 0.90
in the automobile category. The consistency of these past
results leads to the hypothesis that the same patterns will
be evident for brand personality traits:
Hypothesis 1: The rank order of brands level of association with personality trait will be highly similar across
methods (correlation > 0.85) for each personality trait.

Capturing Brand User and Nonuser Associations


For any brand, there are two broad segments: those who
currently buy/use the brand (users) and those who do not
(nonusers). Both customer segments make important contributions to future brand protability. Understanding the
brand personality held by current customers contributes
to maintaining current sales or increasing loyalty. Insights
into the brand personality held by nonusers are important
if the marketing objective is to increase the size of its customer base or to replace customer attrition. It is therefore
essential that any method of measuring brand personality
traits is able to effectively capture trait associations from
both consumer groups.
Brand users have a greater level of direct experience
with the brand than nonusers. This experience means that
the quantity and strength of all brand knowledge (including brand personality traits) should be greater for brand
users than for nonusers. Each experience with the brand
reinforces existing links in memory, which increases their

Spring 2008 155


accessibility until time passes or there is competitor interference (Anderson and Bower 1979; Kent and Allen 1994).
Those who do not buy or consume the brand do not have
this opportunity to build and refresh memory structures.
Even the encoding of advertising, which is one of the primary means that nonusers can learn about the brand, is
biased toward brand users giving a higher response than
nonusers (Hofmeyr and Rice 2000). As brand personalities
are formed by consumers experience with the brand as well
as its advertising, the hypotheses are
Hypothesis 2a: Brand users will give more positive ratings
than nonusers for traits.
Hypothesis 2b: Brand users will be more likely to associate
the brand with a trait than nonusers.
However, in a rating method, respondents are prompted
for all brands. This condition should uncover some nonuser
associations that would be neglected if a free choice approach were used. Therefore, it is hypothesized that
Hypothesis 2c: Brand personality measurement using
rating scales will result in less brand user bias compared
to a free choice method.

Brand-Level Discrimination
Brand personality is considered one of the most promising options for differentiating a brand from competitors
(De Chernatony 2001). Therefore, it is important that any
method of collecting brand personality data can identify
differences between brands. With any attribute measurement, there is a category halo effect that can make it
difcult to distinguish between associations for a specic
brand, and associations the brand gets simply because of
its membership to a category (Dillon et al. 2001; Romaniuk
and Sharp 2000). When there is a large category effect, all
brands are rated similarly, and brand differentiation is more
difcult to detect.
Haley and Case (1979) found that measures that allow
multiple brands to be rated positively tended to result in
brands receiving similar scores. Although both methods
allow for multiple brands to be rated the same, a free
choice method has only two categories (yes/no), whereas
the rating scale has ve. Therefore, when using a ratings
method, a person who feels strongly that a brand has a
specic quality can be distinguished from someone who
sees the association as only slight. A free choice approach
also truncates negative and neutral responses. When rating,
these responses are separated and can be weighted accord-

ingly. Therefore a rating approach is more likely to identify


differences between brands.
The limited prior research in the area has found that
for traditional attributes, the greater consumer sensitivity captured when rating brands does not translate into
greater discrimination between brands at the aggregate
level (Barnard and Ehrenberg 1990). This result may be due
to the presence of category effects overwhelming smaller
brand differences. Brand personality traits may differ from
traditional attributes, as the traits are not derived from
category experiences. Therefore, they may be less subject
to category halo effects, which would mean more variation
in the nature of associations across brands. Rating scales
should be better able to capture this variation. Therefore,
the third hypothesis is that
Hypothesis 3: Brands will vary more in their brand personality when measurement is via ratings compared to
via a free choice approach.

Generation of Brand Associations


One of the concerns about using a free response method
is that it may lead to an underreporting of consumer beliefs (Barnard and Ehrenberg 1990; Haley 1985). Without
specic direction to respond for every brand, consumers
could totally neglect some brands in the set. Therefore, this
method would fail to capture some associations. Moreover,
when consumers fatigue, they may just respond none of
these to quickly nish the questionnaire. This effect could
then lead to an underestimation of the brands strength of
personality.
Although this underreporting is proposed in past literature, there has been no explicit testing for it in brand
attribute research. To examine this effect, there are two tests
that can be used. The rst is the proportion of people who
do not give any brands as associated with each trait. The
second is that nonusers of brands, who are users of other
brands, may have a lower propensity to provide associations with traits for brands not used. In both cases, forcing
people to respond for all brands on all attributes as is done
in a rating method should result in more personality traits
being elicited. This leads to the following hypotheses:
Hypothesis 4: The rating method will result in fewer
consumers reporting that none of the brands are linked
to a trait.
Hypothesis 5: The rating method will result in more
nonusers linking brands to traits than with the free
choice method.

156

Journal of Marketing Theory and Practice

RESEARCH METHOD
The research method was a split sample approach, similar to
Barnard and Ehrenberg (1990). Participants were randomly
recruited from two large Australian cities with the sampling
frame drawn from the electronic white pages. The recruitment process used conventional market research recruitment techniques and trained interviewers to replicate the
commercial market research data collection environment.
All interviews were conducted by telephone to standardize
the data collection approach. This standardized approach
is important to ensure that any differences detected are
due to the different methods, and not a function of how
the data were collected. Once agreeing to complete the
interview, participants were randomly allocated to one of
two groups. The rst group answered two categories in a
free choice style, and the third category using the rating
method. The second group answered two categories using
the rating method and the third in a free choice manner.
The sample size within each group was 144. Comparisons
between samples revealed no signicant differences in
gender (37 percent versus 33 percent male), age (mean ages
of 51.5 and 52.8), or interest in each category as measured
on an 11-point scale (4.51 versus 4.41 for automobiles; 5.18
versus 5.36 for nancial services; and 4.32 versus 4.81 for
fast food). Therefore, the two different samples were considered comparable for the purposes of this research.
To enhance the generalizability of this research, three categoriesfast food, nancial services, and automobileswere
included. This selection covers a variety of category types
and allows the testing of the consistency of results. Fast food
is a low-involvement repeat purchase repertoire market,
automobiles represented durables, and nancial services
represented the services sector. Each category included a
variety of large and small share brands, with eight brands
from fast food, seven brands from nancial services, and
nine brands from automobiles. The use of three categories
in each survey and different attributes within each category
would also have reduced respondent boredom.
Ten personality traits were included for each category. The
traits were selected from the most common responses of 15
marketing academics given the full list of traits from Aaker
(1997), and participants were asked to identify which traits
they thought were appropriate for each category context.
A total of 28 traits were tested across all three categories,
with one trait (successful) common to all three categories.
It should be noted that the purpose of this research was
not to compare the actual brand personalities generated
by each method, but to compare consumers responses to
personality trait questions across methods. Therefore it was
not necessary to include the full set of 42 traits. A smaller,

more relevant, list of traits will result in higher overall ratings and response levels for all brands, but this result will
not detract from addressing the research objectives.
Participants were asked to respond to personality traits
about each individual brand within each category one trait
at a time. For example, in the free choice condition for fast
food, participants were asked to write down a predetermined
list of fast food brands. They were then given the rst personality trait and asked which brands they associated with
that personality trait. This process was then repeated for
the remaining traits. For the rating condition, participants
were read out the scale, and then read out the rst trait and
were prompted to rate each brand from a randomized list.
This process was repeated for all traits. Even though each
respondent was required to make 210 brandtrait assessments for this research across both rating and free choice
methods, this assessment is still less than the effort that
would be required to rate ve brands on 42 traits.
For both methods, the order of trait and brand presentation was randomized within the category, but the presentation order of the categories was the same (automobiles, fast
food, followed by nancial services). In addition, participants were asked questions about their current usage of
brands, which was used to classify consumers into brand
users and nonusers.

RESULTS
Brand-Level Correlations
The rst hypothesis addressed how consistently brands
were ranked across methods for each trait. Brand-level
results for rating were correlated with brand-level results
for free choice within each category for each trait. This
process gave 10 correlations in each market (one for each
trait). The Spearman correlations (shown in Table 1) were
high and statistically signicant, averaging 0.85 in all categories, and 87 percent of correlations were 0.80 or higher.
This result suggests that both methods placed brands in
the same order. This consistency was further veried by
the high correlations (> 0.90) for the trait of successful in
all three categories. The one obviously low correlation, for
intelligent in the nancial services category, was due to a
lack of variation across brands in the rating method (three
brands scoring 3.0), rather than differences in brand rank
orders. Therefore, H1 is supported.

Capturing Brand User and Nonuser Associations


To test H2a and H2b, the responses of users and nonusers
of each brand were compared for each personality trait.

Spring 2008 157


Some of the smaller share brands were omitted from this
analysis due to insufcient numbers of users in either
sample. This left ve of the seven nancial services brands,
ve of the nine automobile brands, and seven of the eight
fast food brands. Analysis for each trait involved subtracting nonusers scores from brand user scores. For rating, it
was the difference in the means for each segment. For the
free choice scenario, it was the difference in the proportion
that associated the brand with a trait. The differences were
always positive (see Table 2), which means both methods
get a higher response rate from brand users. Therefore, both
H2a and H2b are supported. To compare the magnitude of
results (for H2c), the results were standardized by dividing
the difference between users and nonusers by the number
of scale points. The results show the size of the usage bias
is higher for free choice than for rating in all categories,
although with differing degrees across categories: from
automobiles (11 percent for free choice and 10 percent for
rating), to the fast food category (14 percent for free choice
and 9 percent for rating), to nancial services (20 percent
versus 10 percent). Therefore, H2c is supported.
Further investigation into why the methods differed
revealed that the higher mean rating from nonusers was
an artifact of the weighting of scale points. In a free choice
method, neutral and negative responses were weighted zero.
In a rating method, a neutral response is given a value of
three, and negative responses are weighted as two or one.
Given that nonusers are more likely to have no opinion,
nonuser neutral responses in the rating method are given
a higher value than in the free choice method. This categorization increases the nonuser response levels and results
in the lower difference between brand users and nonusers
when mean ratings are calculated.

Brand Discrimination
An important measure of any methods ability to discriminate between brands is the level of variation in brand scores.
To examine this variance, the standard deviation across
brand responses for each trait was calculated. This deviation
was then converted to a percentage of the average score for
each trait across all brands to standardize across methods.
The results (see Table 3) show that the free choice method
had consistently higher variation, with up to three or four
times greater percentage variance than the rating method.
Therefore, H3 is unsupported.

Generation of Brand Associations


To test H4, the proportion of consumers who did not associate any brand with a trait was calculated for each trait.

Table 1
Brand-Level Correlations Across Three Categories
Trait

Correlation

Fast Food (eight brands)


Wholesome
Western
Successful
Family Orientated
Real
Honest
Cheerful
Young
Cool
Fun
Average

0.95*
0.92*
0.92*
0.86*
0.85*
0.84*
0.81**
0.81**
0.81**
0.80**
0.86

Financial Services (seven brands)


Secure
Corporate
Sincere
Condent
Friendly
Leader
Successful
Technical
Up to Date
Intelligent
Average

0.98*
0.98*
0.96*
0.93*
0.92*
0.91*
0.90*
0.83**
0.70**
0.33
0.84

Automobiles (nine brands)


Successful
Tough
Reliable
Family Oriented
Sentimental
Daring
Glamorous
Good Looking
Spirited
Exciting
Average

0.95*
0.94*
0.91*
0.90*
0.86*
0.82*
0.81*
0.80*
0.79**
0.76**
0.85

* p < 0.01; ** p < 0.05.

In the free choice method, this group was the proportion


of consumers that stated none of these when asked which
brands in the list provided they associated with each trait.
For the rating method, participants were coded as associating a brand with a trait if they responded somewhat
descriptive or extremely descriptive. All other responses
were coded as zero. The responses across all brands for each
trait were added up, and consumers who scored zero (which
means the trait was not considered somewhat descriptive
or extremely descriptive for any brand) were classied
as giving a none of these response. This gure for each
trait was then compared across methods.

158

Journal of Marketing Theory and Practice


Table 2
User Bias in Personality Traits (Average Across All Traits)
Free Choice
User Percent

Free Choice
Nonuser Percent

Percent
Differencea

Rating
User Percent

Rating
Nonuser Percent

Percent
Differenceb

ANZ
CBA
NAB
St. George
Westpac
Average

55
58
55
58
56
56

35
42
36
32
33
36

6
11
5
10
11
10

3.4
3.6
3.3
3.5
3.6
3.5

3.1
3.0
3.0
3.0
3.0
3.0

19
16
19
26
23
20

Holden Commodore
Ford Falcon
Toyota Camry
Mitsubishi Magna
Toyota Corolla
Average

43
31
24
22
21
28

28
25
10
11
15
18

10
7
17
9
10
10

3.7
3.3
3.5
3.1
3.3
3.4

3.2
3.0
2.6
2.6
2.8
2.9

16
7
14
11
6
11

McDonalds
Subway
Red Rooster
Pizza Hut
KFC
HJ
Dominos
Average

52
39
37
29
26
25
23
33

34
23
15
16
18
17
12
19

14
5
7
2
14
5
16
9

3.2
3.1
2.8
2.6
3.2
2.8
2.9
2.9

2.4
2.8
2.4
2.5
2.4
2.5
2.1
2.5

19
16
23
14
8
8
11
14

Notes: a Percent User Percent Nonuser/100; b User Mean Nonuser Mean/5.

The results show that the none of these gures were


typically either higher for the rating method or there were
no signicant differences between methods (see Table 4).
Therefore, H4 is not supported. To test the nal hypothesis,
the trait associations given by nonusers were compared.
The results (see Table 5) show that in nancial services, all
brands gained more responses from nonusers when a free
choice approach is used. For fast food brands, the results
were either higher for nonusers when free choice is used
or there was little difference between methods. Finally, for
automobiles, there were mixed results across brands, with
some showing results in favor of rating scales generating
more user responses, and others having very little difference
in the nonuser response across methods. Underlying these
averages for automobiles is considerable variation for each
brand, with no clear pattern evident. Therefore, H5, that
the rating method would generate more responses from
nonusers, is generally not supported.

DISCUSSION
The objective of this research was to reconcile a gap between
academic recommendations for measuring brand personality traits and practitioner implementation. It involved com-

paring the ve-point scale validated in academic research


with a free choice association method. The latter approach
has signicant commercial advantages due to ease of administration, shorter time to collect data, and associated cost
savings. Three categories were chosen to represent packaged
goods, durables, and services. Both methods consistently
ranked brands in the same order, and so are interchangeable on this dimension. This nding suggests that brands
that have more consumers agreeing the brand has that trait
will also get more consumers strongly agreeing the brand
has that trait. Therefore, there is little additional value in
measuring both unless the intention is to undertake further
analyses that specically require a scaled response.
Both methods captured the effect of brand users having
a higher propensity to respond than nonusers; however, the
difference between brand users and nonusers was typically
lower in the rating method. This result is due to a large
proportion of nonusers in these categories being neutrally
disposed. However, in the rating method, this neutral score
still receives a relatively high weighting as the midpoint is
given a score of three out of ve. This weighting inates
nonuser responses for rating methods compared to free
choice methods. This high value given to neutral responses
also reduced the level of discrimination between brands.

Spring 2008 159


Table 3
Proportional Variance Across Brands for Each Attribute
Free Choice*
Percent

Scales
Percent

3
40
14
72
32
39
24
39
30
81
40

6
9
5
11
5
10
7
6
4
11
7

49
47
73
112
70
43
65
43
87
37
63

6
10
20
16
8
8
8
12
20
10
12

25
74
78
118
165
74
90
50
69
53
80

18
22
17
20
28
16
19
19
21
22
20

Table 4
Proportion of Consumers Indicating None of the
Listed Brands Were Associated with Each Trait
Free Choice
Percent

Financial Services
Intelligent
Corporate
Technical
Leader
Condent
Friendly
Up to Date
Successful
Secure
Sincere
Average
Automobiles
Glamorous
Good Looking
Family Oriented
Sentimental
Exciting
Spirited
Daring
Successful
Tough
Reliable
Average
Fast Food
Western
Family Oriented
Cheerful
Fun
Wholesome
Young
Cool
Honest
Real
Successful
Average

* Standard deviation/average across all brands.

Accompanying this effect is that some of the strong positive


responses (ve out of ve) are canceled out by the strongly
disagree responses (one out of ve). This reduces all brand
ratings to being close to the midpoint. In a free choice
method, neutral and negative responses are equally not
included, leaving only all the positive associations.
One advantage of the scale approach was that it was able
to discriminate between neutral and negative opinions
about brands. In particular, this research highlights that

Scales
Percent

Financial Services
Intelligent
Corporate
Technical
Leader
Condent
Friendly
Up to Date
Successful
Secure
Sincere
Average

44a*
17*
35*
24*
17*
28*
20*
6*
16*
48*
26

55b
39
54
30
32
38
40
17
31
58
39

53
23*
5*
39*
53*
43*
57*
19*
21*
7*
32

63
47
29
60
63
56
65
38
41
33
50

49
23
51*
61*
48*
33
58
48
54
19
44

52
16
39
42
38
31
63
49
46
16
39

Automobiles
Glamorous
Good Looking
Family Oriented
Sentimental
Exciting
Spirited
Daring
Successful
Tough
Reliable
Average
Fast Food
Western
Family Oriented
Cheerful
Fun
Wholesome
Young
Cool
Honest
Real
Successful
Average

Notes: a Interpreted as 44 percent did not associate any of the brands


with the trait intelligent in the free choice approach. b Interpreted
as 55 percent did not agree or strongly agree that any of the banking
brands were intelligent on the ve-point scale. * p < 0.05.

most nonusers had neutral opinions. This nding suggests that a nonresponse on a free choice method is best
described as a no opinion/dont know rather than a no,
which means that marketing communications to build
brand personality among nonusers might largely be about

160

Journal of Marketing Theory and Practice


Table 5
Nonuser Associations Given by Each Method
(Brand Averages Across 10 Traits)
Rating
Percent

Free Choice
Percent

7
25
30
15
17
22
8
16
22
18

11
25
28
12
11
14
8
11
15
15

9
12
12
15
6
5
9
11
10

7
11
11
23
11
2
10
13
11

23
22
27
23
18
13
22
21

35
33
42
36
32
25
33
34

Automobiles
Ford Focus
Ford Falcon
Holden Commodore
Holden Astra
Mitsubishi Pajero
Mitsubishi Magna
Nissan Pulsar
Toyota Corolla
Toyota Camry
Average
Fast Food
Dominos
HJ
KFC
McDonalds
Pizza Hut
Quiznos
Red Rooster
Subway
Average
Financial Services
ANZ
Bendigo
CBA
NAB
St. George
Suncorp
Westpac
Average

establishing associations, rather than changing negative


perceptions. This is an area for future research.
Finally, the concern that a free choice method would
result in the underreporting of beliefs appears to be unfounded. Free choice methods resulted in more consumers
associating at least one brand with a trait and typically
generate equal or more responses from nonusers. This research is one of the few published tests of this dimension
of free choice methods.

sonality strategy, and more condence in the assessments of


marketing performance. It bridges the gap between Aakers
(1997) identifying the dimensions of brand personality and
how to assess brand performance on such facets.
The results of this research will be reassuring to practitioners who have already incorporated brand personality
traits into their traditional attribute batteries, using a free
choice method. It will have a positive effect on those who
were considering such a change, but were concerned about
compromising the quality of information. Improvements in
efciency of data collection may also encourage those with
insufcient resources to properly measure brand personality traits to incorporate this element of brand knowledge
into their brand health measures. This improves the efciency of market research and facilitates greater adoption of
brand personality research into marketing practice.
Understanding the perceptions of nonusers has signicant managerial importance as nonusers are one of the
major sources of future brand growth. Therefore, if managers are collecting brand personality information to try
to determine the future sales of the brand, then they need
the method that best captures the perceptions of nonusers. Based on this research, this would be the free choice
approach. Another key objective of most brand perception
tracking is to assess the effectiveness of marketing activities
such as advertising, which can reach and inuence nonusers. A free choice method would be more effective at
assessing overall advertising effectiveness.
More generally, the findings here suggest caution
when weighting neutral responses. If a large segment of
respondents are neutrally disposed, then a high weighting on neutral can give a misleading impression of a
good score. An alternative is to classify people into three
groupspositive, negative, and neutralfor each of the
user/nonuser segments. This classication may provide a
clearer picture of brand performance than was provided
by the mean scores.
An important outcome of this research is that it shows
that assumptions about which will be the better measure
should be tested. It is easy to assume that a method with
more scale points will be more sensitive. This research
shows that this is not always the case. Therefore, further
testing of alternative options for other widely accepted
batteries of scales may help improve the effectiveness and
efciency of research in general.

MANAGERIAL IMPLICATIONS
This research identies the most effective method to measure and identify brand personality traits for managers. This
provides a more solid foundation for developing brand per-

LIMITATIONS AND FUTURE RESEARCH


The main limitation of this research is that is used a split
sample approach. The implication is that although this

Spring 2008 161


research provides insights at the aggregate level, it cannot
necessarily be inferred that individuals will use the different measurement approaches in the same way. This is
an area for future research. Further, only two methods of
measuring brand personality traits were included in this
research. Future research should compare other methods,
such as ranking measures, which have been suggested as
alternatives to rating or free choice approaches.
A further limitation is that only a subset of traits was
examined, and further extensions comparing the nal assessments of brand personalities for competitive brands
from the 42 traits across both methods would be a useful
next stage of method comparison. Another area of research
would be to compare the results of higher-level analysis,
such as factor analysis or perceptual mapping. It would
also be useful to see if the same ve dimensions found by
Aaker (1997) are also found using the different method of
measurement.
This research would suggest that a free choice is the most
effective method of measuring brand personality traits, but
there may be some circumstances where scales are more appropriate. For example, if the intention is to conduct certain
types of analyses (e.g., factor analysis, regression), scales
may be more appropriate. However, this method would only
be recommended if only a subset of the 42 traits proposed
by Aaker (1997) is included, otherwise respondent fatigue
is highly likely. This research was also conducted in one
cultural context, that of Australia. It may be that in other
(particularly non-Western) cultures, scales are more effective than the free choice approach. Therefore, replication
of this research in other cultures is recommended.
This research was also limited to data collection via
telephone. It is unclear if different methods would be better suited to different data collection methods. It may be
that scales are more appropriate when the questionnaire is
visually in front of a person (as in mail or online surveys),
whereas free choice is more appropriate over the telephone.
Given the move toward online surveys, this is an important
area of future research.
One of the interesting ndings of this research was the
inuence of the neutral point in scales on results. This
inuence leads to questions about whether nonassociation
of a brand with a trait is a neutral or a negative response.
This is an important area for future research as it provides
insight into the degree to which those that manage brand
personalities need to build from neutral personality traits,
versus overcome negative associations. It might be that
smaller share brands, which are not known by many, need
to build brand personalities while for larger share brands,
the challenge is to change brand personality to the desired

traits. These approaches would result in very different


tactics.

REFERENCES
Aaker, Jennifer L. (1997), Dimensions of Brand Personality,
Journal of Marketing Research, 34 (August), 347356.
Anderson, John R., and Gordon H. Bower (1979), Human Associative Memory, Hillsdale, NJ: Lawrence Erlbaum.
Azoulay, Audrey, and Jean-Nol Kapferer (2003), Do Brand Personality Scales Really Measure Brand Personality? Journal
of Brand Management, 11 (2), 143155.
Barnard, Neil R., and Andrew S.C. Ehrenberg (1990), Robust
Measures of Consumer Brand Beliefs, Journal of Marketing
Research, 27 (November), 477484.
Belk, Russell W. (1988), Possessions and the Extended Self. Journal
of Consumer Research, 7 (August), 139168.
Brown, Gordon (1985), Tracking Studies and Sales Effects: A UK
Perspective, Journal of Advertising Research, 25 (1), 5264.
De Chernatony, Leslie (2001), From Brand Vision to Brand Evaluation: Strategically Building and Sustaining Brands, Oxford:
Butterworth-Heinemann.
Dillon, William R., Thomas J. Maddern, 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 (4), 415429.
Driesener, Carl, and Jenni Romaniuk (2006), Comparing Methods
of Brand Image Measurement, International Journal of Market
Research, 48 (6), 681698.
Haley, Russell I. (1985), Developing Effective Communication Strategy, New York: Wiley/Ronald Press.
, and Peter B. Case (1979), Testing Thirteen Attitude Scales
for Agreement and Brand Discrimination, Journal of Marketing, 43 (Fall), 2032.
Hofmeyr, Jan, and Butch Rice (2000), Commitment-Led Marketing,
Chichester, UK: John Wiley & Sons.
Joyce, Timothy (1963), Techniques of Brand Image Measurement, in New Developments in Research, London: Market
Research Society.
Katsikeas, Constantine, Matthew J. Robson, and James M. Hubert (2004), In Search of Relevance and Rigor in Marketing Research, Marketing Intelligence and Planning, 22 (5),
568578.
Keller, Kevin Lane (2003), Brand Synthesis: The Multidimensionality of Brand Knowledge, Journal of Consumer Research, 29
(March), 595601.
Kent, Robert J., and Chris T. Allen (1994), Competitive Interference Effects in Consumer Memory for Advertising: The Role
of Brand Familiarity, Journal of Marketing, 58 (3), 97105.
Romaniuk, Jenni, and Byron Sharp (2000), Using Known Patterns
in Image Data to Determine Brand Positioning, International Journal of Market Research, 42, 2 (SpringSummer),
219230.
Sirgy, M. Joseph (1982), Self-Concept in Consumer Behavior: A
Critical Review, Journal of Consumer Research, 9 (December),
287300.
Wee, Thomas Tan Tsu (2004), Extending Human Personality to
Brands: The Stability Factor, Brand Management, 11 (4),
317330.

Você também pode gostar