Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Perceptual Mapping
Perceptual mapping has been used to satisfy marketing and advertising information
needs related to product positioning,1 competitive market structure,2 consumer preferences and
brand perceptions.3 Perceptual maps satisfy these types of information needs by analyzing and
then translating consumers' numeric ratings, brand similarity data and brand preference data
into a visual representation of how those consumers view the set of brands and products.4
There are two approaches to perceptual mapping: attribute based and nonattribute based.
Attribute based approaches require a respondent to evaluate a set of brands on a large number
of specific attributes, typically those attributes felt to influence how consumers perceive,
evaluate and distinguish among brands and products. Attribute based perceptual maps can be
created through the use of one of three mathematical techniques: factor analysis, discriminant
analysis and correspondence analysis. These approaches to attribute based perceptual mapping
are discussed in the next section. Nonattribute based approaches require a respondent to rate
brands in terms of similarities or preferences rather than attributes. A discussion of nonattribute
based perceptual mapping is presented later. While attribute and nonattribute based approaches
to perceptual mapping differ in terms of the types of data collected, both approaches share the
fundamental assumption of perceptual maps: that consumers use broad dimensions to evaluate
brands and products.
Research indicates that consumers in most product categories try to evaluate brands
using the least amount of time and energy. Consumers accomplish this goal by first identifying
and then using a relatively few broad dimensions to compare brands or products. Consumers
evaluate brands and products in terms of broad dimensions because, in doing so, they save time
and energy. They can refer to and use a small number of broad dimensions instead of having to
perform much more complex evaluations based on a larger set of narrower individual attributes.
It is clearly easier to remember and act upon two or three broader dimensions rather than twenty
or thirty individual attributes. All perceptual mapping attempts to make these dimensions
explicit.
The approach to dimension creation differs according to the type of perceptual map and
the corresponding differences in underlying data. Attribute based perceptual mapping creates
dimensions from an analysis of the underlying brand and product attributes. For example, in
the context of "gasoline," two dimensions that consumers might use to evaluate alternative
brands of gasoline are "performance" and "convenience." Each dimension, in turn, is made up
of a number of individual brand or product attributes. The dimension of gasoline "perfor-
mance," for example, might contain the attributes: no knock, no run-on, smooth acceleration and
quick acceleration . The dimension for "convenience" might contain the attributes: many
locations; locations have many pumps; can pay by cash, charge or ATM; easy to pull in and
out. Nonattribute based perceptual maps create dimensions from an analysis of consumers'
evaluations of brand and product similarities or preferences. Here, dimensions reflect the
implicit criteria consumers use to determine similarities and differences across brands or overall
brand preference.
The types of insights provided by perceptual mapping are illustrated in the hypothetical
perceptual map of the beer category shown in the figure below.5 First, the perceptual map
indicates the two primary dimensions used by consumers to evaluate brands of beer. The
horizontal dimension relates to quality while the vertical dimension relates to strength and taste.
Second, the map identifies the specific characteristics of each dimension. The horizontal
dimension is anchored by expensive, fine beers consumed when dining out and low quality
beers drunk when alone. The second dimension is anchored by pale, sweet beers and full-
bodied, malty beers. Third, it appears that consumers use these dimensions to separate beer
brands into six distinct groups.
* Brands A, B and D are perceived similarly. These brands are all seen as very
low in price and quality and are appropriate only for drinking alone. Their
taste falls in the center of the continuum, neither too sweet nor too malty.
* Brands G and H are also seen as very low in price and quality and are
appropriate only for drinking alone. These brands, however, are felt to have a
very full bodied, malty taste.
* Brands C and E are felt to be of average quality and expense and are
appropriate for drinking alone or dining out. These brands are felt to be very
pale and sweet.
* Brands K and L are also felt to be of average quality and expense and are
appropriate for drinking alone or dining out. These brands are also felt to
have a taste that falls in the center of the continuum, neither too sweet nor
too malty. The perceived characteristics of these brands places them closest
to the "ideal" beer.
* Brands F and I are also seen as being high quality, expensive and are only
appropriate for dining out. These brands, however, are felt to have a taste that
falls in the center of the continuum, neither too sweet nor too malty.
Fourth, the groupings of beer brands along the two dimensions provide important insights into
brand positioning, both with regard to the "ideal" brand and competitive products. Brand G, for
example, is seen as an inexpensive, low quality, full bodied, malty beer that is primarily
consumed when drinking alone. It is a beer that most directly competes with Brand H and is
considered to be far from the "ideal beer." Finally, the perceptual map provides a starting point
for discussion of advertising strategy. Brand G, or any other beer brand, can examine the
perceptual map and then decide if their brand's position and competitive set is acceptable (and
thus should be supported in the advertising) or unacceptable (and thus should be addressed by
advertising designed to alter brand perceptions).
Attribute-Based Perceptual Maps
Attribute based perceptual maps make explicit the broader dimensions consumers use to
evaluate and distinguish among brands and products. Attribute based perceptual maps begin
with the creation of a list of specific product category attributes. Because dimensions are
constructed from individual attributes, it is very important that the list of brand and product
attributes contain all attributes that are known to be (or that judgment or research indicates
could potentially be) important in consumers' evaluation of target brands or products. Important
dimensions cannot be discovered in the absence of their component attributes.
Next, semantic differential or Likert rating scales are developed. These scales enable
respondents to rate each brand or product on each attribute. A consumer might, for example, be
asked the following question to assess perceptions of brands on the attribute of "expense":
Rate each brand of beer shown below on the basis of expense. Place a number after each
brand of beer to indicate how expensive or inexpensive you feel that brand is. You can
use any number between '1' (to represent "not at all expensive") and '10' (to represent "extremely
expensive").
The rating of all brands on all attributes, and the subsequent mathematical combination of
attributes into broader dimensions, reflects the underlying assumption of attribute-based
perceptual maps, that is, that a respondent's "rating or judgments about specific attributes are
manifestations of the underlying or latent dimensions that [they] use to distinguish between
brands."6
The next step in the development of an attribute based perceptual map is the selection of
a mapping technique. Three approaches may be used: factor analysis, discriminant analysis and
correspondence analysis.
Factor analysis
A positive correlation coefficient indicates that two measures move together in the same
direction, for example, both measures tend to receive either a high or a low rating. A negative
correlation coefficient indicates that two measures move in opposite directions, for example, one
measures tends to receive high ratings at the same time the second measure tends to receive low
scores.
Examining and interpreting the correlation of one pair of measures is simple. You
simply note the direction and magnitude of the correlation coefficient. Examining and
determining the meaning of the pattern of correlation coefficients for multiple measures is more
difficult because of the large number of correlation coefficients that need to be examined. The
intercorrelations of fifteen measures, for example, result in 105 pairs of correlations. Factor
analysis responds to and solves this problem. Factor analysis is a statistical technique that
identifies the relatively small number of factors or dimensions which represent the relationships
among a large number of interrelated variables."7
The process of factor analysis begins after consumers rate each of the target brands or
products on each individual attribute. A factor analysis computer program examines the set of
ratings data and calculates the correlation coefficient for each pair of variables. These
correlations are the basis of the factor analysis. "The basic assumption of factor analysis is that
underlying dimensions, or factors, can be used to explain complex phenomena. Observed
correlations between variables result from their sharing these factors."8 After correlation
coefficients have been calculated for each pair of variables, factor analysis moves through the
following steps:
2. The program calculates three important pieces of data for each factor. This
data is illustrated in the table on the next page which displays fourteen
factors formed from the hypothetical ratings of eleven brands of big
screen/projection television sets on fourteen attributes.9
Factor Eigenvalue Percent of Variance
1 6.71 47.9
2 4.51 32.2
3 .64 4.6
4 .54 3.9
5 .46 3.3
6 .42 3.0
7 .21 1.5
8 .17 1.2
9 .11 .8
10 .07 .7
11 .06 .7
12 .06 .7
13 .02 .1
14 .02 .1
100.0%
The first column reports the factor number. The second column reports
eigenvalue, the total variance explained by each factor. Eigenvalues greater
than one typically indicate important factors while eigenvalues less than one
typically indicte factors that are less important. Factor importance therefore
increases as eigenvalues increase. The third column translates eigenvalues
into percentages. The total of the eigenvalue column is equal to the number
of variables. Thus, factor one, with an eigenvalue of 6.71, accounts for
47.9% of total sample variance (calculated as 6.71 ÷ 14). The percent of
variance is a very important calculation and indicates a factor's contribution
to an understanding of the underlying pattern of response. Larger
percentages indicate a greater contribution.
4. The factor analysis computer program reanalyzes the data restricting the
number of factors to that specified in the prior step.
The factor analysis program then examines the patterns of factor loadings
and generates a table in which variables are ordered to reflect their factor
loadings, as shown in the table on the next page. This reordering permits the
underlying pattern of association between measures and factors to be clearly
seen.
Attribute Factor 1 Factor 2
6. The researcher examines the factors represented in the map and the
attributes comprising each factor and then creates a name for each factor.
Remember, the factor analysis computer program merely uses mathematical
computations to identify the factors. The researcher must determine what the
factors represent. Factor 1 contains attributes that directly relate to the
viewing experience while Factor 2 contains attributes related to set up and
usage. (It is interesting to note that Factor 1 contains attributes that relate to
both audio and video. The factor would be quite different and would be
given a different name if audio and video attributes were associated with
different factors.)
7. The factor analysis program calculates an average factor score for each
brand. This score represents the average rating of each brand across the
measures comprising an individual factor. The average factor scores for the
eleven brands of big screen/projection televisions are shown in the table
below.
A -1.3 -1.8
B -1.7 +1.1
C -1.6 -1.6
D +1.4 -1.5
E -1.6 -1.2
F +1.4 -1.4
G +1.6 -.7
H -1.5 +1.5
I -1.3 +1.7
J -.3 +.3
K .1 -.1
8. The average factor score is used to plot the brands on the perceptual map
(see next page). Note how the brands tend to cluster into four groups and
the absence of brands from the area that indicates the brand is both easy to
set up and use and provides an excellent viewing experience. The absence of
brands from this important area, which contains the ideal brand, clearly
indicates an unmet niche and a marketing and advertising opportunity.
The prior example provided the simplest example of perceptual mapping using factor
analysis. However, there are often times when more than two factors are identified or when
more detail on the perceptual map is required for a true understanding of consumers'
perceptions. Each of these situations is handled in the following manner.
The presence of three or more factors requires that a perceptual map be formed for each
pair of factors. Thus, a case in which there are three factors (Factor A, Factor B and Factor C)
requires that three perceptual maps be created (Map 1: Factors A and B; Map 2: Factors A and
C; Map 3: Factors B and C). Examining brand positions and competitive sets across multiple,
related perceptual maps provides a comprehensive view of how consumers use important
dimensions to evaluate and distinguish among brands and products.
An Example: Using Attribute-Based Perceptual Mapping
to Measure Advertising Impact
A study conducted by the advertising agency D'Arcy Masius Benton & Bowles10
demonstrated how perceptual mapping can help to determine:
* Second, it is this new dimension, "family car - exciting powerful fun car,"
that most differentiates car manufacturers. Manufacturers on the pre-
exposure perceptual map were dispersed along both dimensions with many
manufacturers at the extremes. The post-exposure perceptual map shows
less differentiation along the horizontal axis (more brands now appear near
the center) and more differentiation on the vertical axis.