Taste, social benefits, product

quality and convenience emerged as the

factors by which Nestle used . . .

Perceptual Mapping to Reposition Brands

Henry Assael

One commonly accepted axiom of the position existing products. To date,

marketing concept is that marketing most of the emphasis has been on The Nestle Study
managers must target their brands to new product development {Business
the perceived needs of defined con- Week, 1970). Here, perceptual map- This paper describes a metbod of
sumer groups. Assuming that the sa- ping is valuable in defining "gaps" in positioning existing products based on
lient consumer perceptions are meas- the market representing unfilled needs ratings of perceived product charac-
ured, this infers the ability to define for a significant group of consumers. teristics. The study was conducted for
homogeneous consumer groups by The problem, then, is to "build" a the Nestle Company in 1968 to de-
perceived needs and to position brands brand that is perceived as fulfilling termine if regional dififerences in prod-
by the perceptions of specified con- unmet needs by a sufficient number of uet perceptions could be exploited by
sumer groups. consumers to insure profitability. repositioning certain brands through
The family of programs referred to Perceptual mapping can also be selective advertising changes.
as clustering and classification tech- used to determine how brands should Five types of generically related
niques (Young, 1968; Kruskal, 1968; be repositioned once they are located beverages plus the broader generic
Carroll and Chang, 1967; Lingoes and on a set of perceived product charac- category were to be positioned by con-
Guttman, 1966) have been applied in teristics. For example, a brand may be sumer perceptions of product charac-
marketing to position consumers and close to another in the company's line teristics. Product criteria were previ-
products based on perceptions of simi- on certain product dimensions, caus- ously established by focused group in-
larities and preferences. ing an unacceptable level of cannibal- terviews and by inclusion based on
Increasing preoccupation with the ization. Perceptual mapping could sug- the company's advertising themes.
techniques and their application (Bar- gest changes in perceptions required Since sharp regional differences in the
nett, 1969; Green, Canmone, and to reposition the brand to appeal to consumption of product types were
Robinson, 1968; Greenberg, 1969; other segments. known (Table 1), product configura-
Johnson, 1969) refiects the value of tions were to be obtained regionally
locating brands in a consumer's "per- and nationally.
ceptual map" of the market.
The author is indebted to the Nestle A national sample of 2,100 bouse-
Marketing managers have used Company and to Henry Jendras. Nestle's
Director of Marketing Research, for per- wives was asked questions concerning
such analyses to develop new and re- mission to use the data in this article. the nature and frequency of consump-
39
Journal of Advertising Research
tion, brand and advertising recall, rea- The problem in factor analysis is to the 88 cities were positioned accord-
sons for use, etc. The configuration of create a set of independent factors ingly on the first 2 factors.
product types was based on a question which account for the maximum The problem in the Nestle study was
asking housewives to rate the degree amount of common variance across all that there were five product types be-
to which 15 criteria described the variables. In the Nestle study, a typi- ing rated by consumers on 15 product
product types. cal application of factor analysis attributes. Thus, three elements were
Ratings were on an 11-point scale, would derive a uniquely independent operating simultaneously; respond-
ten being labelled "describes very set of perceptual dimensions (factors) ents, attributes, and objects. In the
well" and zero "does not describe at from the original set of perceived Green, Frank, and Robinson study,
all." Product criteria, such as "con- product characteristics. only the last two elements applied.
venient," "provides lift and pickup," Factor analysis can also be used to Factor analysis will operate on any
"nutritious," etc., were representative. position objects or consumers in n- two elements—an intercorrelation ma-
dimensional factor space. Given n trix of product attributes and assign-
factors for a product, each respond- ment of factor scores to respondents;
Positioning by Perceptions ent would be assigned a factor score or an intercorreiation matrix of city
on each factor reflecting the degree to characteristics and assignment of fac-
which the factor was representative tor scores to test cities—but not on
Principal components (factor) anal-
of his perceptions of the product. Av- all three.
ysis was used to position the product
types on the preselected product cri- erage scores could be computed by One alternative to examining all
teria. Factor analysis is ordinarily ap- predefined consumer groups. Individ- three components simultaneously
plied in marketing as a reductive tech- ual consumers or consumer groups would be to factor analyze each prod-
nique to derive a more parsimonious could then be positioned in an n-di- uct type separately, based on an inter-
set of factors (f) from a group of mensional perceptual map based on correlation matrix of consumer rat-
original variables (r). A factor matrix their scores. ings for each product attribute.
(f X r) is derived from an intercor- In a similar application of princi- If the factor structures are signifi-
relation matrix of the orignal varables pal components analysis, Green, cantly different across product types,
(rxr). Frank, and Robinson (1967) clus- then the products could not be posi-
The factor matrix is composed of tered 88 test market cities by factor tioned on the same perceptual map.
factor loadings on each factor for analyzing 14 city characteristics. In This would suggest that the original
every variable. The loading is com- this case, the basic input was not re- attributes had different meanings for
parable to a correlation coefficient: spondents but cities. The first 2 fac- the product types within the category.
The square of the factor loading rep- tors to emerge accounted for 74 per On the other hand, if as one would
resents the proportion of variance in cent of the variance. Each city (i.e., suspect, similar factors emerged across
the original variable explained by that observation) was assigned two factor product types due to their generic re-
factor. scores based on its characteristics, and lationships, there would be no prob-

TABLE I FIGURE 1 FIGURE 2

CONSUMPTION INDICES FOR PERCEPTUAL MAP OF TASTE AND PERCEPTUAL MAP OF TASTE AND
THREE BEVERAGES SOCIAL BENEFIT DIMENSIONS: SOCIAL BENEFITS DIMENSIONS:
(Total Sample = 100) TOTAL SAMPLE SOUTH
Cen- Social Benefit Factor Social Benefits Factor
East tral South West
Product B* (Positive) (Positive)
(Lbs/ A A
Household 58 120 132 66 Ideal Ideal
Product C B D
(Dollars/ Taste F^actor
(Negative)— -(Positive) Taste Factor B
Household) 156 56 97 61
D (Negative)— -(Positive)
Product D
(Lbs/
Household) 141 63 111 64
* Data for Products A and E not avail-
able (Negative) (Negative)

40
 Volume / /, Number I. February 1971
matrix of 15 product attributes was
FIGURE 4 FIGURE 5
broken down into 15 new compon-
I'F.RCEPTUAL MAP OF TASTE AND PERCEPTUAL MAP OF TASTE AND
ents based on the correlations be- SOCIAL BENEFITS DIMENSIONS:
SOCIAL BENEFITS DIMENSIONS: EAST
tween attributes. The adjusted matrix
WEST was then rotated by the Varitnax cri- Social Benefits Factor
Social Benefits Factor terion, assuring a unique solution. Fac- (Positive)
(Positive) tor derivatioti was stopped at an eigen-
value equal to 1.0. Ideal
A Ideal
B Four factors emerged in all but one
D
Tiiste Faclor case. A fifth factor emerged for the Taste Factor
(Negative) — -(Positive) product type that was the least gen- (Negative)— -(Positive)
D
erically related.
It was apparent that the first four (Negative)
(Negative)
factors were close to identical across
all products. The factors were la- teria could be regarded as constant
lem in dealing with objects, attributes, belled "taste," "social benefits," "prod- across the product types to be posi-
and respondents simultaneously. One uct benefits," and "convenience." The tioned. This meant that objects and
set of factors could represent all six factor loadings for the four comtnon respondents could be positioned on
objects, permitting the positioning of factors were intercorrelated for all the same attitudinai dimensions iti a
the product on the same perceptual pairs of products. Correlations ranged spatial configuration.
maps. There would be little variation between .79 and .96. Thus, the latent
The next step was to position the
between product types by the latent structure of the evaluative product cri-
objects in such a spatial configuration
structure of the perceptual criteria,
with the axes representing specific per-
but significant variation in rating prod-
ceptual dimensions (factors). Factor
ucts on these critera.
scores on each of the four dimensions
To test the assumption of homo- were required for each product type
geneity of dimensions across product to develop the spatial configurations.
types, separate principal components Since there were minimum variations
analyses were performed for each of between factor loadings for each of
the five product types and the generic the five product types and the generic
category. category, it was necessary to select a
The factors were derived from inter- constant factor structure across prod-
correlations of the 15 attitudinal cri- uct types. The factor loadings for the
teria for each product type based on generic category were used for this
Ihe original attitudinal ratings by the purpose.
2,100 respondents. The correlation
The factor scores for each product
were computed as follows: A re-
; I
FIGURE 3
PERCEPTUAL MAP OF TASTE AND Henry Assael is an associate professor
FIGURE 6
SOCIAL BENEFITS DIMENSIONS: of marketing at the Graduate School
of Business Administration, New York PERCEPTUAL MAP OF TASTE AND
CENTRAL REGION University. A graduate of Harvard CONVENIENCE DIMENSIONS FOR
PRODUCT C BY REGION
Social Benefits Factor University, he received his M.B.A.
Convenience
(Positive) from the Wharton School and Ph.D. Factor
from Columbia. An expert in the ap-
A Ideal (Positive)
plication of multivariate analyses to
Tiiste Factor B marketing problems, Dr. Assael bas West East
(Negative) — -(Positive) published ailicles on market segmen- Taste Factor
D tation, sampling tecbniques and rela- (Negative) -(Positive)
C tionsbips of consumer attitudes to be- South
E bavior. He consults for a number of Central
(Negative) leading marketing firms in tbe New
York area. (Negative)

41
Journal oj Advertising Research
spondent's original rating of a prod- ing from 58 in the East to 132 in the Perceptions must, of course, be re-
uct on a specific attribute was multi- South) suggests that taste and social lated to behavior if strategic implica-
plied by the attribute's factor loading benefits do not discriminate by re- tions are to be drawn from consumer
for the generic category. The result gional behavior. or product positions in perceptual
was divided by the standard deviation More pronounced differences space. Given such associations, man-
of the product attribute, creating a emerge by region for Products C and agement can then test hypotheses re-
standardized score with standard de- D. Product C is perceived much more garding product or advertising strate-
viation of one. Summing this result favorably by Easterners on taste and gies to introduce new products or re-
across the 15 criteria produced the social benefits. Dollar expenditures position existing ones.
respondent's factor score on the di- per bousehold for Product C in the Tbe key element is the need for
miension for the given product. The East are 56 per cent above the na- continual measurements of perceptions
factor scores were then averaged for tional average, a logical association to and behavior to trace shifts in prod-
the total sample and by region. This positive taste perceptions. The prod- uct perceptions and evaluate the suc-
meant that every product type was as- uct was positioned similarly by re- cess of repositioning strategies. The
signed a "taste," "social benefits," spondents in the South and Central Nestle data were obtained from such a
"product benefits," and "convenience" regions, yet was rated more favorably tracking study. Similar data systems
score by region. by Westerners on social benefits, de- would be required to apply perceptual
spite a consumption index of only 61. mapping effectively.
Results Taste is the prime discriminator of
regional behavior for Produet C. So- REFERENCES
Given the four factors, 6 two-di- cial benefits rather than taste is the
primary behavioral discriminator for BARNLTT, NORMAN L . Developing Effec-
mensional maps were produced for tive Advertising for New Products. Jour-
each region and nationally. Maps rep- Product D. Eastern and Southern re- nal oj Advertising Research, Vol. 8, No.
resenting perceptions of the taste and spondents position the brand most fa- 4, pp. 13-18.
vorably on social benefits, paralleling Business Week. An Outside Job Fills the
social benefits dimensions for the total Product Gap. May 16. 1970, pp. 54-56,
sample and respondents in each of the higher consumptioti indices. Except CARROLL, J. D.. AND J. J. CHANG. Relal- i
four regions are in Figures 1-5. Table for the South, moderately negative ing Preferences Data to Multidimensional
taste perceptions remained constant. Scaling Solutions. Murray Hill, N.J.:
1 presents consumption indices for Bell Telephone Labs, 1967 {mimeograph).
three of the products based on na- Sharp differences in perceptions of so- GREEN, PAUL E., FRANK J. CARMONE,
tional averages. cial benefits were not anticipated, sug- AND PATRtCK J. RoBiNSON. Nonmetric
Scaling Methods: An Exposition and
The generic category was in the gesting that the product could be re- Overview. Wharton Quarterly. Vol. 2,
most ideal position on all maps. This positioned regionally by exploiting this No. 2. pp. 27-40.
was true for the other dimensions as attribute. GREEN, PAUL E., RONALD E . FRANK, AND
PATRICK J. ROBINSON. Cluster Analysis in
well. Apparently, consumers con- Maps were also developed by re- Test Market Selection. Management Sci-
ence, Vol. 13, April 1967, pp. 387-400.
sidered the generic category as a posi- gion for a given product type. Figure
GREENBERG, MARSHALL G . A Variety of
tive yardstick and rated the other five 6 presents regional differences in per- Approaches to Nonmetric Multidimen-
product types accordingly. Product A cepts of Product C by the taste and sional Scaling. Proceedings, 16th Inter-
national Meeting. Institute of Manage-
is consistently closest to the ideal on convenience dimensions. Sucb a map ment Sciences, March 1969.
both the taste and social benefits di- serves to summarize differences by JOHNSON, RICHARD M . Techniques of
mensions. Product E appears in the consumer group, holding brand or Market Segmentation: Cluster Analysis
vs. Q. Analysis. Proceedings. Interna-
negative quadrant across all regions, product type constant. tional Marketing Congress, American
although it is in a slightly more favor- Marketing Association. June 1969.
able position on both dimensions in KRUSKAL, J. B. How to use M-D-SCAL,
Conclusion a Program to do Multidimensional Scal-
the South. ing and Multidimensional Unfolding.
Product B always appears in the Perceptual mapping is a valuable Murray Hill, N.J.: Bell Telephone Labs.
1969 {mimeograpb).
positive quadrant, yet it forms a clus- method of data reduction. At a time LiNGons, J. C. AND LOUIS GUTTMAN. An
ter with D among Southern respond- when marketing management's data IBM Program for Guttman-Lingoes
Smallest Space Analysis. Behavioral Sci-
ents and with C among Eastern re- inputs are increasingly complex', it ence. Vol. 11, 1966. pp. 322-323.
spondents. The lack of variation provides a relatively simple snapshot YOUNG. F . W . T O R S C A . An IBM Pro-
across regions despite substantial dif- of the market by product types or con- gram for Nonmetric Multidimensional
Scaling. Journal of Marketing Research.
ferences in consumption indices (rang- sumer groups. Vol. 5, No. 3, pp. 319-321.
42