You are on page 1of 6

A Two-Dimensional Concept

Of Brand Loyalty
GEORGE S. DAY
Stanford Umversity

There is more to brand loyalty than just consistent


buying of the same brandattitudes, for instance.

TTEMPTS TO DESCRIBE biand-Ioyal buyers a p p e a r

^. to have reached a temporary impasse. For example, Frank (1967) recently concluded, "Brand
loyal customers almost completely lack identifiability in terms of either socioeconomic or personality
characteristics . . . (they) do not appear to have
different average demand levels , . . (and they) do
not appear to have economically important differences in their sensitivity to either the short-run
effects of pricing, dealing, and retail advertising or
to the introduction of new brands."
The studies cited by Frank were of product classes
where brand loyalty is both measurable and highly
reliable (Massy, 1966). The principal variables used
to identify brand-loyal buyers also are widely used
to guide specialized advertising and promotion
efforts toward definable market segments. The net
These data have been described in greater detail in Day
(1967). They were eollected as part of the Columbia University Research Project on Buyer Behavior, and the author
is grateful to Professor John Howard for the opportunity to
use them in this study and to the Ford Foundation and the
University of Chicago for supporting the collection and
preparation of the data.

RESEARCH, VOL. 9, NO. 3

result is that very little can be said about brand


loyalty as a phenomenon or as a basis for market
segmentation despite the fact that nurturance o
brand loyalty is one of the chief objectives of packaged goods brand advertising and promotion. However, many suggestions have been made for overcoming shortcomings in both brand loyalty measures
and descriptive variables. The purpose of this paper
is to empirically explore several of these suggested
improvements. !
GEORGE S. DAY is an assistant
professor of marketing in tbe
Graduate School of Business at
Stanford Uaiveniity. He is a
graduate of the University of
British Columbia, University of
Western Ontario and Columbia
University. Before getting a Ph.D.
from Columbia in 1967 he was a
lecturer in tbe Scboot of Business
Administration at tbe University
of Western Ontario. He has published several articles on attitude
researcb methodology and its application to marketing problems
and a book, Buyer Attitudes and Brand Choice Behavior.

29

Realistic Measures of Brand Loyalty

The usual loyalty measures are derived in one


way or another from Lhe proportion of total product
puiclKises devoted to the brand most often purchased. Although these measures are based on rejx)rted purchase behavior, there has been a tendency
to regard them as evidence of lelative brand preference (Carman and Stromberg, 1967).
The equating of behavior patterns with preferences to infer loyalty has been challenged directly
(Newman, 1966). And a closely related criticism is
that loyalty measures, based on reports of purchase
decisions, do not distinguish between true or "intentional" loyalty and "spurious" loyalty associated
with consistent purchasing of one brand because
there are no others readily available or because a
brand offers a long series of deals, had a better shelf
or display location, etc. (Moulson, 1965).
The key point is that these spuriously loyal buyers lack any attachment to brand attributes, and
they can be immediately captured by another brand
that offers a better deal, a coupon, or enhanced
point-of-purchase visibility through displays and
other devices.
One immediate implication of this view is that
loyalty should be evaluated with both attitudinal
and behavioral criteria. However, once attitudinal
criteria are imposed, loyalty becomes a brand-specific concept and not a general concept describing
the over-all behavior pattern in the product class.
In practical terms this means a buyer has a brand
loyalty score for each brand purchased in a given
period based on share of total purchases and attitude toward the brand.
Better Descriptive Variables

Socioeconomic and demographic variables provide relatively standardized information which can
be applied to many different problems. But for the
specific problem of describing brand loyalty within
a product class, this degree of generality becomes a
significant weakness. These descriptive variables at
best are remote proxies for important individual
differences in buying styles, decision processes, or
sensitivity to promotional influences. This problem
is most severe when single choice decisions are relatively unimportant to the buyer (a situation that
encompasses many o the products where loyalty is
anticipated and has been studied).
Despite these limitations, there has been little
experimentation using variables which are specific
to tlie product class. Exceptions are measures of
30

deal buying, store switching activity within the


product class, etc. Among reasons for the lack of
attention are cost and difficulty of collecting information so it can be related to a loyalty measure
based on purdiase patterns. Even if expense was not
a consideration, there always is uncertainty about
which variables should be collected.
Incorporating the Improvements

The equation below satisfies the requirements


that the brand loyalty score be brand specific and
positively related to the degree of favorableness toward the brand:
k A"!

(1)

where:
Li

P[B]i

k, n

= the brand loyalty score for the i**"


buyer of brand m,
= proportion of total purdiases of
the product that buyers devoted to
brand m over the period of the
study,
the attitude toward brand m at the
beginning of the study (scaled so
that a low value represents a favorable attitude),
r= descriptive variables to be fitted to
Ll by least squares,
= constants whose values are varied
by trial and error to maximize the
fit between Lj and X^, X^, . . . X,.

This approach is not without problems. First, it


is not obvious what weights should be given to the
relative influence of the attitude and behavior components in the loyalty score. For lack of better criteria, it is assumed that the weights are defined by
the best fit between the loyalty score (see equation
1) and the descriptive variables used in the multiple
regression equation. There is a further problem that
is a consequence of combining a one-time estimate
of brand attitude (with a certain degree of response
unreliability and some likelihood of true change
subsequently) with an interval estimate of purchase
probability. Thus, there is a possibility that the attitude component of the brand loyalty score may not
be accurate during some of the time period required
to estimate the purchase probability.
This paper gives empirical backing to the suggestions for improvement. In confirming the reasonableness of a two-dimensional brand loyalty concept
JOURNAL OF ADVERTISING

compared to a purely behavioral definition, the first


concern is with the extent of spurious loyalty and
the degree of improvement in identifiability when
using standard socioeconomic and demographic variables. The second concern is to appraise the usefulness of descriptive variables which relate to actual
respondent purchasing activity, exposure to personal and impersonal sources of information, and
style of buying. The buying style variables are
essential to understanding individual differences in
response to deals, price changes, point-of-purchase
promotions, etc. (Day, 1967). The usefulness of
these descriptors will be their ability to explain a
meaningful proportion of the variation in household loyalty scores (Lj), over and above that explained by socioeconomic and demographic variables.
RESULTS
An empirical test of the brand loyalty model was
conducted by collecting a large number of descriptive and attitudinal variables at the beginning of a
five-month diary purchase panel of 955 households
(See Table I).
Because of the nature of the design, there was
considerable opportunity for bias due to sensitization. The results of a comparison of the manipulated panel with a parallel control panel indicated
TABLE 1

A.

B.

C.

D.

INDEPENDENT VARIABLE DESCRIPTIONS


Socioeconomic and demographic variables
1. Size of city
2. Size of household
3. Age of housewife
4. Presence of children
5. Education of housewife
6. Occupation of head of household
7. Household income
8. Hours housewife employed
Demand, price and store response variables
9. Total number of units purchased
10. Average price paid per unit
11. Range from highest to lowest price paid
12. Dealing dummy variable
(0 no purchases on deal)
13. Store activity dummy variable
(0 =: all purchases of product made in one store)
Exposure to information
14. Number of visitors to house
15. Number of invitations to visit friends
16. Television viewing (hours per week by respondent)
Determinants of "buying style," or reaction to the purchasing environment
17. Perceived impulsiveness in buying
18. Economy consciousness
19. Perceived time pressure
20. Interest in the differences among brands in the product
class
21. Confidence in judgments about brands in the product
class.

RESEARCH, VOL. 9, NO. 3

a significant upward bias in the number of product


users but a lesser bias in the proportion of households buying the analysis brand. These biases had
no apparent effect on key relationships (such as
between attitude and behavior) but reduce the
projectibility of the results.
The analysis focuses on one of two major
brands of a frequently purchased convenience food
product. There were 225 buyers who made at least
five purchases of the product during the period of
the panel, and 148 of these buyers made at least
one purchase of the brand under study.
The product class lends itself to an exploratory
study of brand-loyal behavior for several reasons.
Both major brands were reported to have obtained
between 90 and 95 per cent all-commodity distribution in the area sampled for the panel. Thus, buyers
were usually confronted with a genuine choice
situation. At the same time, each brand emphasized
different product attributes so there were valid
grounds on which a buyer could make a choice. The
product class also oflEered several methodological
advantages: (1) There were no new brands introduced during the period of the test nor was there a
significant increase in primary demand. (2) The
housewife, who was the source of the attitude ratings and subsequent purchase reports, was invariably a product user. (3) The typical purchase cycle
was two weeks, so during the test period there was
an average of about eleven distinct brand choice
decisions and 21 units purchased on which to base
reliable estimates of brand choice probability. (4) A
study of potentially related products showed that
there were no direct substitutes for the product to
bias estimates of choice probabilities.
Evidence of Spurious Loyalty

Loyal segment is often defined as a buyer who


devotes at least 50 per cent of his product purchase
to a single brand. In this study, 108 of 148 buyers
would be classified as brand loyal on that basis.
However, if loyal buyers were required to have an
extremely favorable or a very favorable initial attitude toward the brand as well as buying the brand
on a majority of occasions, then the brand-loyal
segment is reduced to 76 buyers. By this arbitrary
definition, spuriously loyal buyers comprised almost
30 per cent of the total loyal group. However, the
relative size of the spuriously loyal segment is to
some extent a function of timethat is, a longer
time period would provide spuriously loyal buyers
a greater opportunity to switch. Defining segments
in this way also conceals a great deal of individual
31

variability, which means a continuum is more practical for pinpointing true brand-loyal buyers.

Apparently less influenced by day-to-day price


fluctuations, as evidenced by the fact that while
the price paid per unit is close to average, the
range of prices is quite narrow, and fewer purchases tended to be made on deal.

True Brand-Loyal Buyers

Data in Table 2 summarize a series of multiple


regression equations in which various weights for
the continuous brand loyalty score (Lj)^ were fitted
to two different sets of descriptive variables.
Key results are found in the contrast between
equations 1(A) and 1(B) in which the additional
descriptive variables raised the R^ from .035 to .158
when the usual measure of brand loyalty was used,
and between equations 1(B) and 2(B) in which the
two-dimensional measure of brand loyalty raised
the R2 from .158 to .270 when the full set of descriptive variables was used.
The next step identified specific variables that
contributed most to the description. The relative
contribution of each variable, expressed as a Beta
coefiicient in equation 2(B), is summarized in Table
3.
DISCUSSION
Improvements in brand loyalty and descriptive
measures, taken singly and in combination, have
contributed significantly to improving the explanation of individual variability in brand loyalty. In
fact, without the improvements, the proportion of
variability explained is insignificant. The results
in Table 3 also provide a reasonably consistent picture of the true brand-loyal buyer as:
Very conscious of the need to economize when
buying
Confident of her brand judg^nents
A heavy buyer of the product
An older housewife in a smaller than average
sized household (who needs to satisfy the preferences of fewer family members)

Other less significant findings which help to focus


the picture of the true brand loyal buyer indicate
that buyers are less impulsive and patronize fewer
stores. There is some evidence that these buyers
also have less exposure to personal sources of inforTABLES

TW'O DIMENSIONAL BRANI> LOYALTY SCORES


(Coefficients for Equation 2B in Table 2)
Variable

Ranked Beta Coefficients

1. Economy consciousness
1.00'
2. Confidence in brand judgments
_ .95'
3. Household size
_ .86^
4. Total number of units purchased
.80'
5. Range of prices paid
_ .75*
6. Presence of children
_ .64
7. Dealing
_ .63'
8. Age of housewife
_ ,59"
9. Invitations to visit
,59'
10. Impulsiveness
.54"
11. Store activity
,42
Notes:
(a) The largest Beta coefficient was set equal to 1.00 and the
rest of the coefficients were scaled proportionately.
(b) The stepwise addition of the remaining ten variables lead
to an increase in the standard error of estimate.
a = Significant < .01 level;
b Significant < .05 level

mation (at least through visits to friends). Nothing


conclusive can be said about media exposure because of the gross nature of the variable used.
A plausible interpretation is that true brand loyal
buyers are committed to the value and price appeal
of the brand by being confident that they have
judged the brand correctly, coupled with a perceived need to economize. This interpretation is
supported by an independent correlation of .331
(r2 = .110) between loyalty measure (Li) and a

TABLE 2

EFFECT OF CHANGES IN BRAND LOYALTY MEASURE


AND DESCRIPTIVE VARIABLES ON REGRESSION RESULTS
Descriptive Variables
(A) Socioeconomic and
Demographic only

Weighting Scheme
Equation
Number
I.

2.
3.
4.
5.

Ri

Standard Error
of Estimate

(ADJ.)

Standard Error
of Estimate

R'

k=I:
n=0
[Usual measure
of brand loyalty]

.035

28.70

.158

28.05

.068
k=i/2;

n=l

JQ06

n=l
n = 1/4

^8
.052

3S.65
67.23
18.09
30.01

.27O'
.270'
.221"
.232"

30.60
61.20
17.08
28.50

k = 2;
k = 1;
a = Significant at < .01 level

34

(B) All Descriptive


Variables

b z= Significant < .05 level

JOURNAL OF ADVERTISING

semantic differential scale with opposite poles reading "good buy for the moneynot a good buy for
the money."
Clearly this interpretation is subject to qualification because of the large unexplained proportion
of variability in the loyalty scores. Some error certainly can be attributed to inadequacies in the variables and the statistical limitations of the regression
model as well as reporting and coding mistakes. Of
more serious consequence are the known shortcomings in the brand loyalty measure, particularly the
inability of the measure to cope with attitude
change over time. Finally, there is evidence that the
most important basis for commitment to the brand
was not represented in the set of descriptive variables. This was uncovered when an independent
scale representing food value (the actual attribute
is disguised) had a simple correlation of .502
(r^ = .252) with the loyalty measure (Lj). Had this
scale been adequately represented by a descriptive
variable in the same way that the "good buy for the
money" scale was represented by the economy conscious variable, the over-all correlation would have
been much higher and the information about the
basis of loyalty to the brand being studied would
have been more complete.
IMPLICATIONS
There are several immediate implications from
this study, notably that brand loyalty can be profitably studied as a brand-specific concept and that
measures of brand loyalty based solely on purchase
data conceal considerable spurious loyalty. In a
general sense, the results of this study help to clarify
various unrelated or competitive hypotheses about
brand loyalty. Any comments must necessarily be
tentative because the study focused only on one
brand (which means that switchers cannot be unambiguously identified), and one particular product
class which is not necessarily similar to other frequently purchased products.
According to this study, any one or all of the following hypotheses about the nature of brand loyalty
merit serious consideration; (I) Loyalty is based on
a rational decision made after an evaluation of the
benefits of competing brands. This decision is, in
effect, a commitment to the brand. Such decisions
likely are made on an infrequent basis, and once
made the buyer either, (2) feels that such an explicit
decision is no longer necessary before each purchase, in which case the process becomes habitual,
or (3) his strong affective orientation toward the
brand narrows his perceptual judgment, and he is
RESEARCH, VOL. 9, NO. 3

less likely to see competitive promotional activity


(Bieri, 1967). Obviously, the commitment is never
total, and the decision is reviewed when competitive
or other circumstances change.
There is nothing in the descriptive findings of
this study that contradict any of the above hypotheses. Evidence for an initial rational decision
is particularly compelling, given the fact that true
brand-loyal buyers have the greatest degree of confidence in their ability to judge between brands.
Because loyal buyers are heavier than average buyers of the product, we also can infer that they have
more experience on which to base an explicit decision. Heavy buyers also have more to gain by
adopting an habitual pattern which will permit
them to economize on the effort of repeated decisions. The habit hypothesis is further supported
by the finding that true brand-loyal buyers tend to
be older and presumably have developed greater
rigidity in their preferences. Of course, it was easier
for older buyers of this product to maintain an habitual pattern since they typically represented small
households with fewer competing preferences to
consider.
The evidence for a perceptual bias against competing stimuli (whether packaging, promotion, or
advertising) is tentative but very suggestive because
there would be a direct contradiction in true brandloyal buyers being economy conscious while at the
same time paying prices that are close to average.
This hypothesis is supported from a different direction in a recent study (Bogart, 1967) that found
the highest level of ad readership for a product
occurred among users who were brand loyal, "but
not among the most brand loyal."
'
REFERENCES
BlERi, JAMES. Attitudes and Arousal: Affect and Cognition
in Personality Eunctioning, in Carolyn and Muzafer
Sherif (eds,). Attitude, Ego-Involvement and Change,
New York: John Wiley, 1967.
BOGART, LEO. Strategy in Advertising. New York: Harcourt,
Brace, and World, 1967.
CARMAN, JAMES M. AND JOHN L. STROMBERG. A Comparison

of Some Measures of Brand Loyalty. Unpublished working paper. Institute of Business and Economic Research,
University of California, Berkeley, July 1967.
DAY, GEORGE S. Influences on the PredictixK Value of Brand
Attitudes. Unpublished working paper. Graduate School
of Business, Stanford University, 1967.
FRANK, RONALD E. IS Brand Loyalty a Useful Basis for
Market Segmentation? Journal of Advertising Research,
Vol. 7, No. 2, June 1967, pp. 27-33.
MASSY, WILUAM F. Brand and Store Loyalty as Bases for
Market Segmentation, in Joseph W. Newtnan (ed.). On
Knowing the Consumer. New York: John Wiley, 1966.
MOULSON, T . J. Danger Signals: How to Spot Erosion in
Brand Loyalty. Printer's Ink, March 12, 1965, pp. 55-61.
NEWMAN, JTOFPH W . An Overview, in Joseph W. Newman
(ed.). On Knowing the Consumer. New York: John
WUey, 1966.

You might also like