Professional Documents
Culture Documents
Attributes
Author(s): Gregory S. Carpenter, Rashi Glazer, Kent Nakamoto
Source: Journal of Marketing Research, Vol. 31, No. 3 (Aug., 1994), pp. 339-350
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/3152221
Accessed: 14/07/2010 10:07
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ama.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Marketing Research.
http://www.jstor.org
and KENTNAKAMOTO*
RASHIGLAZER,
GREGORYS. CARPENTER,
Conventional product differentiation strategies prescribe distinguishing
a product or brand from competitors' on the basis of an attributethat is relevant, meaningful, and valuable to consumers. However, brands also
successfully differentiate on an attribute that appears to create a meaningful product difference but on closer examination is irrelevant to creating
differentiation. The authors examine how
that benefit-"meaningless"
can
differentiation
produce a meaningfully differentiated
meaningless
brand. They argue that buyers may infer that a distinguishing but irrelevant attribute is in fact relevant and valuable under certain conditions, creating a meaningfully differentiated brand. They outline the consumer inference process and develop a set of hypotheses about when it will produce
meaningful brands from meaningless differentiation. Experimental tests in
three product categories support their analysis. They explore the implications of the results for product differentiation strategies, consumer preference formation, and the nature of competition.
Meaningful Brands
From
The
Differentiation:
Meaningless
Dependence on
Attributes
Irrelevant
I
339
Journal of MarketingResearch
Vol. XXXI (August 1994), 339-350
JOURNALOF MARKETING
RESEARCH,AUGUST1994
340
anisms producingpositive valuationof the irrelevantattribute dependingon whetherthe true value of the differentiating attribute is known to buyers and on the price of the
differentiatedbrand.We derive a set of hypotheses regarding conditions under which meaningless differentiation
could be successful and test the hypotheses in two experiments using three differentproductcategories.
Our results show that meaningless differentiationis valued by consumers in a surprisingnumberof situations.For
example, meaningless differentiationis valued even if the
differentiatedbrandis pricedabove all othersand, more surprisingly, in some cases increasing price actually can increase preferencefor the differentiatedbrand.Furthermore,
the competitiveadvantagecreatedby addingan irrelevantattribute can be sustained even if consumers acknowledge
that the differentiating attributeis irrelevant. Considering
these circumstancesenables us to identify conditions under
which meaningless differentiationmight be successful. We
explorethe implicationsof our resultsfor competitivedifferentiation strategies.
Our results have importantimplications for understanding consumer preferences. Recent research on consumer
choice demonstratesthat the context of choice as defined
by the composition of the choice set affects consumerpreferences (e.g., Glazer, Kahn, and Moore 1991; Huber,
Payne, and Puto 1982; Simonson and Tversky 1992; Tversky and Sattath1979). These studies show thatincluding an
asymmetrically dominated, or "irrelevant," alternativein
the choice set can affect preferencesfor the remainingalternatives as can normativelyirrelevantinformationaboutconditions of choice. We examine the impact that an irrelevant
attributecan have on consumerchoices. It can elevate an unimportantor even irrelevantattributeto one of great importance, affecting relative brand evaluations by creating a
choice situation in which buyers infer it is valuable. This
suggests that the attribute structure, in addition to the
choice context, may affect preferences.We explore the implications of this for a richer understandingof consumer
preferences.
Our results suggest a ratherdifferentapproachto marketing strategy.The traditionalview holds thatcompetitiveadvantage results from catering to consumer preferencesbetter than competitors. Consistent with studies of consumer
preferenceformation(e.g., Carpenterand Nakamoto 1989;
Hoch and Ha 1986), our presentanalysis suggests that marketing strategy can affect the formationof preferencesand
therebygain competitiveadvantage.Althoughwe do not address whetherthis advantageis sustainablein the long run,
we do show that competitive advantageis possible through
differentiationthat, ratherthan meeting customerneeds better than competitors, provides no meaningful value to the
consumer, though it might appear otherwise. We discuss
the ethical and policy implicationsof our analysis for using
meaningless differentiationto create a long-run sustainable
advantage.
AN IRRELEVANT
ATTRIBUTE
VALUING
AttributeIrrelevance
We focus on consumer brandevaluation in a marketfor
multiattributeproductsin which brandsdiffer on a common
341
Irrelevant Attributes
mation of the attribute'svalue, even though trial is by definition uninformative. This type of mechanism could give
rise to competitive advantage for the first brand to adopt
such a claim (Carpenterand Nakamoto 1989, 1990).
Even in the absence of experience or independentinformation, consumers may treat the irrelevantattributeas valuable because of the informationit conveys. According to
the informativeness principle of communications theory,
the purpose of a communicationis to inform, to communicate somethingnot alreadyknown (e.g., Clark1985). A communicationcan be considered to have two parts:a semantic
component (the message's literal meaning) and a pragmatic
component (the reason for the communication;e.g., Harrris
and Monaco 1978). In cases in which the literal component
is uninformative,individualsfocus on the pragmatic(Gruenfeld and Wyer 1992). For example, Sears Auto Centersadvertising that they do not perform unnecessary work on
their customers' cars is uninformative in the literal sense.
Sears would never advertise that it does perform unnecessary work on customers' cars. So, the pragmaticcomponent
becomes the focus-why are they telling us this?
In a similar sense, an irrelevant attribute attached to a
brandis semanticallyuninformative.Consequently,consumers may focus on the pragmaticcomponent, speculating as
to why the attribute is there at all. Why would P&G develop a technology to flake coffee crystals, seek and receive
a patentfor it, and then spend considerablesums promoting
it? One obvious inference is that it produces better-tasting
coffee. Why else would P&G go to such effort?The very existence of the irrelevantattribute,therefore,implies it is valuable and may lead buyers to value it. Thus, an irrelevantattribute can become relevant because it conveys pragmatic
information.
The uniquenessof the irrelevantattributecan lead to positive valuation of the differentiatedbrand.Studies of causal
inference by Einhorn and Hogarth (1986) and McGill
(1989) show thatcausalityis attributedmore often to distinctive ratherthan common attributes,suggesting that buyers
may infer that the irrelevant attribute actually causes the
productto performbetter.For example,one may sampleFolger's Instant,like it, and thus infer that the crystals caused
the satisfactoryoutcome.
Two other characteristics of unique attributes suggest
that the differentiatedbrandwill be more favorably evaluated than an undifferentiatedone. First, the irrelevantattribute is novel (all other brands present redundantinformation), and greaterweight is often given to novel information
in judgment (Kahnemann 1973; Wyer 1970; Wyer and
Carlston 1979). This suggests that if brands are similar in
all respects except for the irrelevantattribute,the branddistinguishedby it will be more favorablyevaluatedsimply because part of the informationconveyed by it is novel. Second, the irrelevantattributeis likely to be more salient in interbrandcomparisonsbecause it is unique.Competitorslacking the irrelevantattributewill suffer in comparison(Houston, Sherman,and Baker 1989). In a broadersense, the irrelevant attributemakes the differentiatedbranddistinctive in
consumers' minds-not only differentbut more salient, perhaps perceptuallydominant,and thereforepreferred(Carpenter and Nakamoto 1987, 1989).
The uniqueness, novelty, and salience of the differentiated brandalso can simplify a consumer's decision. Fiske
and Taylor(1984) arguethatdecision makerstend to be cognitive misers, taking shortcuts when possible. Numerous
studies of decision making have suggested that consumers
adopt simplifying strategieswhen faced with complex problems (e.g., Payne 1976) and may not even be aware they
are doing so (e.g., Einhor 1980; Langer 1978). An irrelevant attributecan simplify choice by offering a simple, single-attributedecision rule.
In sum, a brand differentiated on the basis of a unique
but irrelevantattributecan have advantages in evaluation.
Communicationstheory suggests consumers will infer that
the attributehas value. The uniqueness of the attributeis
likely to make it loom large in brand evaluation. From a
competitive standpoint, the differentiated brand will fare
well in interbrandcomparisons.Moreover, product experience is unlikely to attenuatethese effects because trial will
be uninformative. These arguments suggest our first
hypothesis:
H1: In the absenceof information
revealingthe irrelevanceof
thedifferentiating
a branddifferentiated
attribute,
by theirrelevantattributewill be judgedmorefavorablythanthe
samebrandwithoutthe irrelevantattribute.
Aaker (1991) suggests a similar inference process in the
case of relevant product attributesand product quality. If
consumers lack the ability to judge products, he argues,
they often rely on "seemingly trivialbut observable" attributes to judge overall productquality. Largerstereo speakers
indicate better sound, and detergentsproducing more suds
are thought to clean better. In these cases, buyers make an
overallinferenceaboutqualityfrom a relevantproductattribute that may or may not be unique. H1 suggests that a similar inference process may be occurring if the attribute is
unique but irrelevant.
Valuationwith Revelation of Irrelevance
In the absence of information about the attribute's true
value, consumersare forced to infer it, and these inferences
are likely to lead to a favorableevaluation.But what if it is
revealed to buyers that the attributeadds nothing, that it is
irrelevant? Normative theory would argue that the irrelevant attributewill be discounted completely and thus will
have no impact on preferencefor the brandoffering it. Consumers will realize that the attributecontributesno benefits
and attach no value to it.
However, another intriguing possibility exists. We have
argued so far that consumers value the differentiatedbrand
in partbecause the irrelevantattributemakes the branddistinctive. To be sure, the revelation of the unique attribute's
irrelevance is negative. But the irrelevant attribute still
makes the differentiatedbrandunique and distinctive relative to others in the choice set. Thus, as in the case without
the revelation, the differentiatedbrandstill may be favored
in interbrandcomparisons, more salient, and even dominant, leading to greaterpreferencefor it. Furthermore,in a
difficult choice environment such as the one considered
here, the differentiatedbrandcontinues to simplify choice.
Thus, consumersimplicitly(perhapsunintentionally)may attach value to the irrelevantattribute,even knowing that the
342
attributehas no value. For example,buyersmay persistin favoring Folger's Instanteven after it is revealed to them that
flaked crystals have no impact on taste.
Such a phenomenonis consistent with a varietyof cognitive biases in judgment. For example, Gilovich (1981)
found thatirrelevantinformationaffected sportswriters'predictions about the futuresuccess of graduatingcollege football players. Includingthe irrelevantdatum that the college
senior was from the same hometown as a famous professional footballplayerincreasedexpertevaluationof that senior. Gilovich argues that subjects made positive associations between the irrelevantinformationand the graduating
senior that, once formed, were difficult to ignore despite
their obvious irrelevance.The inability to ignore the irrelevant attribute is also consistent with studies of hindsight
bias in which actual outcomes bias even recollections of
prior forecasts (Fischoff 1975). In both the Gilovich study
and the mechanismwe propose, the consumeris not fooled
into disbelieving the informationthat the attributeis irrelevant. The attributeis recognized as irrelevantbut, even so,
it cannot be ignored in decision making.
A more extreme view is suggested by Fiske and Taylor
(1984) on the perseverance effect. Consider the following
scenario: A politician is accused of breaking a law; an
investigationis initiated, an indictmentreturned,and a trial
held. The verdict is not guilty, clearing the politician of all
wrongdoingin a completely fair and open hearing.Nonetheless, a perceptionof guilt may linger, and with it there may
be a negative association despite the evidence (cf. Nisbett
and Ross 1984). Here the irrelevance of the accusation is
actually disbelieved-after all, if the charges were baseless
why was an investigation initiated, an indictment sought
and received, and a trial conducted? In this case, the true
informationis ignored and an incorrectinference persists.
Therefore,a basis for positive valuationexists whetheror
not the revealed information is believed. If it is not believed, a preferencefor the irrelevantattributeof course can
persist. However, we take a more conservative position.
The attribute'strue value, once revealed, is unlikely to be
completely ignored. In that case, a basis for positive valuation of the irrelevantattributeremainsbecauseof the uniqueness and distinctivenessof the differentiatedbrand.This suggests that buyers will value the irrelevantattributeand the
brandoffering it even if the true irrelevanceof the differentiating attributeis acknowledged:
is revealed,
attribute
of thedifferentiating
H2: If theirrelevance
consumers will recognize that the attributehas no value.
However, a branddifferentiatedby it nevertheless will be
preferred to the same brand without the irrelevant
attribute.
Summary
We have argued that the addition of an irrelevantattribute to one brandchanges the structureof the consumer'sdecision and thus can lead the consumerto value it. The irrelevant attributemakesthe differentiatedbranduniqueanddistinctive. The attributeconveys pragmatic information, so
that its existence suggests it is valuable. Moreover, relying
on it simplifies choice. These factors may be difficult to
Table 1
1
ATTRIBUTES
OF DOWNJACKETSUSED INEXPERIMENT
Attribute
Levels*
Fill rating
Covermaterial
Stitching
500, 550
Cotton,synthetic
Regular,extratight
Down fill
*Attribute
of targetbranditalicized.
**Irrelevant
attribute.
eliminate,even if the irrelevanceof the uniqueattributeis revealed and acknowledgedby buyers. Thus, a unique, distinguishing, but irrelevantattributecan become relevant to a
buyer's decision, creating a meaningfully differentiated
brand.
EXPERIMENT1
Our first experimentis an initial test of our first two hypotheses. In this experiment,subjects are presentedwith an
arrayof brands,one of which is differentiatedby an irrelevant attribute.We predict that buyers will value the brand
with the irrelevantattributemore than the same brandwithout it, even if its true irrelevanceis acknowledged.
Method
A set of eight hypotheticaldown jackets was constructed
by factoriallycrossing three attributes-temperaturerating,
type of shell cover, and the type of stitching. All attributes
had two levels, shown in Table 1, creatingeight brands.To
these brandsa fourth attributewas added, manipulatingthe
presenceor absence of an irrelevantattribute.In one set, we
added a common attribute, "regular down filling," to all
brands. In another set, seven of the eight brands were described as having "regulardown filling" and to the eighth
(the "target" brand)we added a fictional, irrelevantattribute created for the experiment-"alpine class down fill."
We also manipulatedthe attributeinformation revealed to
subjects. In the subjective condition, subjects were provided with no explanationabout the meaning of attributes.
In the revealed condition, subjects were provided with a
paragraphdescribing all the attributesand their functions.
For example, subjects were told that a higher fill rating indicated greater warmth, and down fill was described as
goose down or a mix of goose and duck down. Alpine class
fill was described as goose down, but subjects were also
told that though the age of a bird matters,the type of bird it
comes from "does not make a difference." A manipulation
check, reported subsequently, indicates that, given this
manipulation, subjects indeed acknowledged that alpine
class fill was irrelevant.The two manipulations,presenceor
absence of the irrelevantattributeand presence or absence
of attributeinformation,were crossed, forming a 2 X 2 between-subjects factorial design. Ninety-nine masters' degree students were compensated for participating in the
study.
IrrelevantAttributes
343
Figure 1
Procedure
Subjectsfirst were told thatthey would be ratinghypothetical products.The productclass was describedand, in the revealed condition, attributedescriptions were provided. In
the subjectivecondition,only physical descriptionsof the attributes were provided. Next, the eight brands of jackets
were described and the subject was asked to rate the
brands.The ratingswere done by having the subjectfirst indicate his or her least preferredbrand at the left end of a
line segmentand the most preferredat the rightend. The subject then placed a labeled mark for each of the six remaining brands along the line to indicate relative preference.
The rating for each brandthen was the distance in centimeters of its markfrom the least preferredbrand(rangingfrom
0 to 10). After rating all eight brands, subjects completed
manipulation check and control measures. Subjects responded to several questions regarding experience with
down jackets and the importanceof various attributes(including the irrelevantattributeif it was present), as well as
open-ended questions about attributes.
VALUATION
OF TARGETBRANDWITHANDWITHOUT
IRRELEVANT
ATTRIBUTE
10-
642I' '
-=-
Subjective
-Oe
Revealed
Results
We first analyzed the individual attributeratings for the
subjects who were exposed to the branddifferentiatedwith
alpine class fill. Each subjectratedthe importanceof alpine
class fill versus regulardown fill on a scale of 1 to 5, with
5 being "Alpine class fill is much betterthan regulardown
fill," 3 being "Regular down and alpine class fill are about
the same," and 1 being "Regular down fill is much better
than alpine class fill." If the revelationof the equality of alpine class fill and regular down fills was effective, one
would expect a mean rating no different from 3. The mean
rating for the group with revelationwas 2.4, statisticallyindistinguishable from 3 (t < 1). For those exposed to the
same brandwithout the informationrevelation, the average
rating was 4.3, significantly larger than three (t = 1.94, p <
.05). Thus, individuals in the revealed condition acknowledged the equivalence of alpine class and regulardown fill.
We next analyzed ratings for the differentiated brand
across the four conditions using ANOVA with two factors
(the addition of the irrelevantattributeand the revelationof
attributemeanings) and their interaction.Results show that
only the addition of the irrelevant attribute is significant
(F3,93= 16.6, p < .01). To test our two hypotheses, we compared means across the experimentaland control groups in
the revealed and subjective condition, respectively.
To test H1, we comparedthe mean ratingof the differentiated brandin the subjectivecondition for those exposed to
the targetbrandincluding alpine class fill with the mean rating for those exposed to the same brandwith regulardown
fill. We predict that the mean rating will be higher among
those exposed to the differentiatedbrandthanits undifferentiated form. Analysis shows that the average rating for the
target brandwith regulardown fill in the subjective condition was 3.1; among those exposed to the same brandthatincluded alpine class fill, the average brandratingclimbed to
9.1. That difference is significant (t93= 6.27, p < .01), demonstratingthat buyers positively value the differentiated
brand if the true value of the irrelevantattributeis not re-
With
Irrelevant
Attribute
Without
Irrelevant
Attribute
344
groups that were exposed to the undifferentiatedbrandappears minimal; mean ratingof the targetbrandin those two
groups are statisticallyindistinguishable(t = 1.50, p > .10).
Discussion
These results demonstratethat an irrelevantattributecan
be valued and this effect persistseven if consumersacknowledge the attributeis irrelevant.It was surprising,however,
that the effect was virtually unaffectedby revelation.It appears that the differentiating attributeremains distinctive
and unique, making discounting it difficult. Inferences regardingthe semantic meaning of the irrelevantattributeappearedlimited in the open-endedresponses, whetheror not
the attribute's true value was revealed. Subjects preferred
the differentiated brand regardless of the information revealed to them, suggesting that the primaryimpactof the irrelevant attributewas to increase the competitive salience
of its brand.Inferences were limited, perhapsbecause subjects had only attributeinformationavailable to them.
In actual markets,of course, consumershave much more
information,making richerinferences possible. One important factor in making such inferences can be price. Higher
prices, for example, can be interpretedas indicatinghigher
value (e.g., Gerstner 1985; Peterson 1971; Rao 1971; Rao
and Monroe 1989). It is plausible, therefore,that buyer inferences regardingthe value of a differentiatedbrandmay
be affected by its price. Consideringprices also would enable us to examine more fully the inferenceprocess by adding an irrelevantattributeand examining consumerreaction
at different prices.
AN IRRELEVANT
ATTRIBUTE:
VALUING
THEROLE OF PRICE
To examine the inference process more fully, we consider the impact of price on consumer valuationof the differentiatedbrandif buyers must infer the value of the irrelevant attributeand, subsequently,if it is revealed to them.
Inferential Valuationwith Price but WithoutRevelation of
Irrelevance
Our first study shows that when price informationis not
provided,buyers value a distinguishingbut irrelevantattribute. However, how much a differentiatedbrandwill be valued may depend on its price. Considerthree cases: The differentiatedbrandis priced at the low end of all brands(we
term this a low price), the high end of all brands (a high
price), and above all others (a premiumprice).
If the differentiated brand is priced below most others,
one inference is that, even though it is differentiatedand attentiongetting, the differentiatingattributemust be worthlittle, if anything at all, because its price is low. In other
words, the pragmatic component of the information conveyed by the irrelevantattributeis offset by the pragmatic
component of the information conveyed by a low price.
("If the attributeis so valuable, why aren't they charging
more for it?") Thus, meaningless differentiationshould affect brand evaluation little. However, if the differentiated
brandis priced towardthe top end of all alternatives,one inference is that its unique attribute must be valuable. The
IrrelevantAttributes
value. A premiumprice can lead to no additionalpreference
from differentiating because of the mixed message it produces. It is importantto note that at all three prices, the differentiatedbrandremains unique. It is the only brandwith
the irrelevantattribute.Thus, uniquenessalone does not produce the anticipatedimpact from differentiating.Rather, it
is the combination of uniqueness and the inference made
from the price of the differentiatedbrand.
EXPERIMENT2
In this study, we test H3 and H4 by adding price to the
productprofiles, one of which may include an irrelevantattribute.In addition, we expand the numberof productcategories to three to begin to consider the generality of our
results.
Method
In a mannersimilar to that used in experiment 1, we created a set of brandsusing three binary attributesplus price
(fixed at one of two levels) as a fourth attribute. In this
case, we used three product categories-down jackets,
pasta, and compact disc players. For each category, 8 of the
16 possible attributecombinations were chosen so that all
pairs of attributeswere orthogonal (that is, a one-half fraction of the factorialcrossings of the four attributes).As in experiment 1, these attributesare all relevant for the categories under consideration, in that they correspond to attributes and levels found in real productsand they contributeto
the overall productvalue. Attributesand levels for all products are shown in Table 2.
In each category, a fifth attributewas added to manipulate the presence or absence of the irrelevant attribute.In
one set of brands,a fictional, irrelevantattributewas added
to a (target) brand: One brand of down jackets was described as having "alpine class fill," one pasta brand as
being "authentic Milanese style," and one compact disc
player as having a unique "studio-designed signal processing system." The irrelevant attributewas added to one of
two possible brands to prevent confounding the effects of
the attributewith a brand-specificeffect. Attributesof the target brands are shown in Table 3. To the remaining brands
we added a common attribute(e.g., regular down fill; see
Table 2). In a second set of brands, all received the common attribute(e.g., regulardown fill).
In this experiment we manipulated two factors. First,
with regardto the irrelevantattributethere were two conditions. The attributecould be absent (we term this the undifferentiated condition or the control group), which was the
case for one-third of the subjects. The irrelevant attribute
was presentin two cases for which productswere presented
in sequence. For the first productclass presented,the target
brandwith the irrelevantattributebut without additionalinformation about its true value (we term this the subjective
condition);for the second productclass, the irrelevantattribute was included with the revelation of its irrelevance(this
formed the revealed condition). This sequential presentation was used to accentuate the impact of revealing the
unique attribute's true value. These three cases-control,
subjective, and revealed-formed the attributecondition of
experiment2.
345
Table 2
2
USED INEXPERIMENT
PRODUCTSANDATTRIBUTES
ProductClass
Attribute
Levels
Tracklocationtime
.5 or 2 seconds
Compactdisc
Remotecontrolaccess
No, yes
players
Randomaccessprogramming 6 or 12 selections
Price
$195,$295,$395
Signalprocessingsystem Regular,studio-designed*
Pasta
Dried,frozen
Storageform
Saltcontent
Salted,unsalted
Richness
Regular,extrarich
Price
$1.89,$2.79,$3.69
Style
Regular,authenticMilanese*
Fill rating
Downjackets
500, 550
Covermaterial
Synthetic,cotton
Regular,extratight
Stitching
Price
$100,$140,$180
Downfill
Regular,alpineclass*
*hrrelevant
attribute.
Table 3
2
TARGETBRANDS'ATTRIBUTES
USED INEXPERIMENT
TargetBrand
Product Class
Attribute
Brand 1
Brand 2
.2
.5
Compactdisc players Tracklocationtime'
Yes
Yes
Remotecontrolaccess
6
Random
accessprogramming2 6
All
Price3
All
Studio-designed
signal
processingsystem*
Dried Frozen
Pasta
Storageform
Saltcontent
Unsalted Unsalted
Richness
Regular Regular
All
Price3
All
AuthenticMilanesestyle*
Fill rating
550
500
Downjackets
Covermaterial
SyntheticSynthetic
Stitching
Regular Regular
All
Price3
All
Alpineclassfill*
attribute.
*Irrelevant
IInseconds
2Innumberof selections.
3Pricesaregivenin Table2.
346
undergraduatesubjects participatedin the study for class
credit.
Procedure
As in experiment 1, subjects were told that they would
be ratinghypotheticalproductsand were shown an example
of the rating procedure. Then, a product class was described; the subject rated the eight brands using the same
scale as in experiment 1 (with measures shown in millimeters). If the subject was in the condition in which an irrelevant attributewas present, a second product class was described,includingthe revelationof the irrelevanceof the differentiatingattribute.3The second class was omittedfor subjects in the controlgroup.Finally, subjectscompletedmanipulation check and control measures.
Analysis
As a manipulationcheck, an open-endedquestion asking
the respondentwhat the unique attribute"meant" in product evaluation was asked. A judge, naive to the study,
coded subject responses. As expected, the number of subjects who stated explicitly that they did not know what the
added attributemeant was greaterwhen no revelation was
provided than when they were told it was irrelevant.Conversely, more subjects statedthat the addedattributewas irrelevant when they were told so (X2= 6.38, p < .05). As a
second check, we analyzed preferencesin the control group
with respect to price. As expected, preferencefor the target
brandfalls as price increases, as shown in Figure 2.
Given our hypotheses, the main results of the experiment
focus on the differencesbetween the controlgroupin which
the target brand is undifferentiated and the condition in
which it is differentiated,either with or without the revelation of the differentiatingattribute'svalue. To test for these
differences, we analyzed preference ratings for the target
brandfrom the subjective and revealed cases separatelyto
test H3 and H4. This procedureenabled us to treateach portion of the design as a completely between-subjects
experiment.
908070
60-
50:
40302010
0O-4W
LowPrice
HighPrice
Price
Premium
Figure 3
PREFERENCEDIFFERENCE
FROMADDINGAN IRRELEVANT
ATTRIBUTE
WITHOUT
REVELATION
50 SO
40-.
30-"
20-"
10
10'
-111
SubjectiveEvaluation
To test H3, we analyzepreferenceratingsfor the branddifferentiatedby the irrelevantattribute,including data from
the subjective and control conditions. We ran an ANOVA
on brandpreferenceratings with attributeconditions (presence or absence of the irrelevantattribute)and price condition (low, high, premium) as factors; we included product
class and the particularbrandto which the irrelevantattribute was attachedas covariates. The interactionof attribute
condition and price was significant (F2,115= 4.73, p < .01).
Our interestfocuses on specific aspects of the interaction
pattern. Consistent with H3, the evaluation of the target
brandis enhancedby the addition of an irrelevantattribute
when its irrelevanceis not revealedand price is high or premium (tll5 = 2.93, p < .01 and tl15 = 3.89, p < .01, respectively). When the price is low, preference rating does not
3This procedureconfounds order of presentationand revelation. However, this orderingwould tend to highlight (by contrast)the revelationand
thus would be expected to reduce or to eliminate the impact of the irrelevant attributeon preferenceat all price levels.
LowPrice
HighPrice
Premium
Price
Irrelevant
Attributes
entiated.An ANOVAwith attributecondition(presenceor
absenceof theirrelevantattribute)
andprice(low,high,premium)as factors,andincludingproductclass andbrandas
covariates,shows thatthe maineffectsfor bothfactorsare
significant(F2,117= 11.39,p < .01 for attributecondition
and for price F2,117= 16.27, p < .01); the interactionbetweenthetwo factorsis not(F < 1).Ourinterestis in thepatternof preferenceacrosspriceconditions.
H4 predictsthatwhenthe irrelevanceof the differentiating attributeis revealed,valuationwill be positiveonly at a
attribhighprice.Inotherwords,revealingthedifferentiating
ute's trueirrelevanceeliminatesthe impactof addingtheirrelevantattributeat a low or premiumpricebecauseof the
inferencesinduced.
To test these predictions,we computedthe differencein
mean rating between the attributepresent and control
groupsat thethreepricelevels.At a highprice,theincrease
in valuationfromdifferentiation
is positive,increasingpreference 33 points (t117 = 3.05, p < .01). At both low and pre-
miumprices, valuationof the targetbrandappearsto increase,but these increasesare not significant(p > .10 in
bothcases).Thus,consistentwith H4,if the differentiating
attribute's
irrelevanceis revealed,addingit increasesvaluation only at a high price.The resultsappearin Figure4.
Summary
A brandwitha distinguishing
butirrelevantattribute
creates potentialuncertaintyaboutwhetherto valueor ignore
it. Ouranalysissuggeststhatan irrelevantattributeis valuedpositivelyin a surprising
numberof cases.If no otherinformationis available,brandvaluationis extremelydifficult,andbuyerscan simplifytheirtaskby inferringthatthe
butdistinguishing
irrelevant
attribute
is valuable.Itsdistinct-
347
Figure4
PREFERENCEDIFFERENCE
FROMADDINGAN IRRELEVANT
ATTRIBUTE
WITHREVELATION
3530
2520-
15.
105u
*M-M
LowPrice
Price
HighPrice Premium
Ourfindingssuggestanimportant
linkbetweenmeaningless differentiationand brandequity. Aaker(1991) describesbrandequityas consistingof brandassetsandliabilities thatcan be groupedin five categories:brandloyalty,
nameawareness,perceivedquality,brandassociations,and
otherproprietary
brandassets.Brandloyaltyconsistsof the
abilityto attractandretaincustomers;awarenessis buyers'
familiarityandlikingof the brand;perceivedqualityvaries
with price and positioning;brandassociationsdependon
the abilityto recallthebrand,its position,andpositiveattitudesor feelingstowardthebrand;andotherproprietary
assets arethosethatareuniqueto the brand.
348
Meaningless differentiationmay influence these sources
of brand equity. Distinctiveness may increase awareness
and familiarity.Perceived quality may be higher, especially
if the differentiatedbrandprices above others. Brand associations may be positive if the irrelevantattributeis attractive and suggestive. The differentiatingattribute'suniqueness can be a proprietaryasset, creatinga competitiveadvantage. Thus, Folger's strategy may be effective by catching
buyer attentionin a crowded market,being perceived as a
high-quality brand, having a network of positive associations, creatingan exceptionallyhigh weight on its unique attribute,and doing so with an attributeothers cannot duplicate by virtue of its patent protection.
This suggests that meaningless differentiationcan operate to build valuable brands,despite the lack of a meaningful productdifference.Ratherthanexplicitly presentingbuyers with positive, valuable differences, a brand differentiated with an irrelevantattributecreates a context in which
positive associations,awareness,high value for uniqueattributes, and perceivedqualityand otherbenefits are logical outcomes of consumer inference, building brandequity in the
process. An importantdirectionfor furtherresearchis to examine this inference process and its implications for building valuable brandsmore explicitly.
ConsumerPreference and Choice
Our results are somewhatdisquietingfor the model of rational choice. Centralto this view is the notion that preferences are fixed, exogenous, and revealed by choice. In this
context, more information improves decision makingbetter informed consumers make betterjudgments. Irrelevant information in such a framework is immaterial and
shouldnot affect decisions. We show thatirrelevantinformation does indeed have an impact. A brandattributemay not
have objective value. Rather,preferencesfor it may be constructed in response to the context in which valuation is
made (Tversky, Sattath, and Slovic 1988). This suggests
that, contraryto the model of rational choice, preferences
are endogenous, that is, constructed rather than revealed,
and more informationcan bias decisions systematically.
Ourdemonstrationthatpreferencescan be influencedsubstantiallyby the attributestructure,prices, and available informationraises anotherimportantquestion:Will these preferences be reinforced or eliminated through trial and use?
Or, more generally,will these preferencespersist?One possibility is suggested by the preferenceupdatingmechanism
describedby Carpenterand Nakamoto(1989). With trial,initial preferencesare either confirmedor not. If trial is satisfactory, consumers may attributethe positive outcome to
the brand'sattributestructureand levels, includingan irrelevant attribute.Buyers may develop a naive theory of the
value of the attribute,confirmedthroughtrial, repeateduse,
and advertising (Hoch and Ha 1986). Thus, one plausible
outcome is that use will reinforceratherthan erode preferences for an irrelevant attribute. This is consistent with
Hutchinson and Alba (1991), who demonstratethat learning is inhibited when an irrelevantattributeis salient. The
continued success of Folger's and others would not appear
to be inconsistent with this.
JOURNALOF MARKETING
RESEARCH,AUGUST1994
The Nature of Competition
Our research, combined with other work demonstrating
that the informationavailable to decision makers-the context-affects preferenceand choice, suggests thatrichertheories of competitive advantagemay be constructedby formally integratingthe "behavorial" view of consumerdecision making into concepts of marketingstrategy.Developments in institutionaleconomics (Williamson 1975) and organizationaltheory (Miles and Snow 1978) have been influential in the realizationthat firms do not always behave as
predictedby classical, rationalmodels and they have had a
significant impact on developments in the area of strategy
formulation.However, theories of marketingstrategyhave
been built with a more limited role for consumer decision
making than is perhapsdesirable or appropriate.
A useful startingpoint for an integrationcould be to examine the role of informationin consumerdecision making
and its implications for competition and strategy.The marketing concept rests on the traditional assumptions about
the role of informationin choice. In turn,competitive marketing strategyis a race to meet consumerneeds best at the
lowest price. However, if judgments about brandvalue depend on the informationavailable to the consumer-that is,
preferences are context dependent-competition can take
on an entirely differentcharacter.
Rather than being a race to meet customer needs at the
lowest price,competitionmay become a battleover the structure of consumerpreferences.Firms may attemptto elevate
the importanceof one attributeover anotheror add distinctive but irrelevant attributesto shift competition. This argues, fundamentally,that the marketingconcept may be too
limiting in some markets.The implicationsfor an expanded
view of the marketingconcept are significant.A fundamental componentof being "customerdriven" may involve devoting resourcesnot just to satisfying customersbetterthan
competitorsbut to creating value for customersby shaping
the context of consumerpreferencesand thus competition.
There is, however, a disquieting aspect to our findings.
Ourresearchsuggests that it is possible to garer the advanmight aptages of productdifferentiationwith attributes'that
pear to provide functionalbenefits to the consumer but, in
fact, do not. The attribute only serves to make the brand
unique. Although there are clearly limits to such a strategy
(as experiment 2 shows), the possibility raises ethical and
policy issues as well as strategicones. From a policy standpoint, the issues are similar to those raised by deceptive advertising (e.g., Gaeth and Heath 1987). The differentiating
attributes may provide an economic benefit to the brand
throughhigher profits (from higher prices, greatersales, or
both) with little or no investment in productimprovement.
However, meaningless differentiation differs from deceptive advertisingin an importantway: The advertisingclaim
reflecting meaningless differentiationneed not be false or
deceptive in that no explicit claim regarding its value is
made. Rather, it is a bias toward causal inference on the
partof the consumerthat results in the increasedpreference
for the differentiatedbrand.For example, Perdue chickens
are differentiatedby their yellow skin, which derives from
their feed (which includes marigold petals). The color has
no impact on the flavor or texture of chickens. Perdue
IrrelevantAttributes
makes no explicit claims in that regard. However, a consumer could infer that Perdue chickens' yellow color leads
to better-tasting,more tender chicken, which may create a
competitive advantage.
On the other hand, competitive advantage created
throughmeaningless differentiationmay not be sustainable
because of competitive reaction. Our analysis considers
cases in which the differentiatingattributeis unique-that
is, not copied by competitors.If competitorsdo react, is the
competitive advantage sustainable? We have not formally
addressedthat question, but observationsuggests that competitive reaction can limit both the distinctiveness and
uniqueness of a brand'sirrelevantattribute.If, for example,
all competitorsreact by introducingtheir own unique irrelevant attribute,no one brandwill be distinctive, and the impact of meaninglessdifferentiationcould be reduced.The optimal reaction is not clear, and we leave that for furtherresearch.However, we can suggest that the dynamics of competition may make sustaining an advantagebased on an irrelevant attributedifficult if the uniquenessand distinctiveness of the attributeare reduced.
CONCLUSION
We examine how a brandcan achieve competitive advantage through differentiatingnot on an important,meaningful, or valuable attributebut throughthe additionof a distinguishing, unique,but irrelevantattribute.Contraryto the traditional view of productdifferentiation,meaningless differentiation can be positively valued by buyers. An advantage
createdthroughmeaningless differentiationmay not be sustainable in the long run if competitorsreact and reduce its
uniquenessor distinctiveness.However, if the differentiated
brandremainsdistinctive and unique, it can create a context
in which buyers infer that the additionalif irrelevantattribute is valuable. Its distinctiveness suggests its relevance,
prompting positive associations, even causal inferences
about its impact on product performance. Its uniqueness
can simplify decision making, convey novel information
that receives greater weight, or affect the comparisonprocess in such a way as to disadvantagebrandslacking this attribute. These factors operate even if buyers acknowledge
that the differentiating attribute is irrelevant, though we
have not considered whetherthese factors continue to operate in the long run. We propose a set of hypotheses about
when consumers will value this additional attribute, and
tests in two experimentssupportour explanation.Thus, we
suggest that differentiationbased on a unique, distinguishing, but irrelevantattribute,throughthe consumerinference
process, can create a relevantand valuable perceiveddifference, leading to a meaningfully differentiatedbrand.
REFERENCES
Aaker, David A. (1991), Managing Brand Equity.New York:The
FreePress.
Adweek(1986), "Silk in a Bottle"(May19), 18.
Carpenter,
GregoryS. andKentNakamoto(1987), "MarketPioneering,Learningand Preference,"in Advancesin ConsumerResearch, Vol. 15, M. Houston, ed. Provo, UT: Association for
Consumer Research.
349
and
(1989), "ConsumerPreferenceFormationandPio-
New York:RandomHouse.
Gaeth, GaryJ. and Timothy B. Heath (1987), "Cognitive Processing of MisleadingAdvertisingin Young and Old Adults:Assessment and Training,"Journal of ConsumerResearch, 14 (June),
43-54.
Gerstner, Eitan (1985), "Do Higher Prices Signal Higher Quality," Journal of MarketingResearch, 22 (May), 207-15.
Gilovich, Thomas (1981), "Seeing the Past in the Future:The Effect of Associations to FamiliarEvents on Judgmentsand Decisions," Journal of Personality and Social Psychology, 40, 797808.
Glazer, Rashi, BarbaraE. Kahn, and William L. Moore (1991),
"The Influence of ExternalConstraintson Brand Choice: The
Lone-AlternativeEffect," Journal of ConsumerResearch, 18,
119-127.
Gruenfeld,Deborah H. and Robert S. Wyer, Jr. (1992), "Semantics and Pragmaticsof Social Influence:How Affirmationsand
Denials Affect Beliefs in ReferentPropositions,"Journalof Personality and Social Psychology, 62, 38-49.
Ha, Young-Won and Stephen J. Hoch (1989), "Ambiguity, Processing Strategy,and Advertising-EvidenceInteractions,"Journal of ConsumerResearch, 16 (December), 354-60.
Harris,R. J. and G. E. Monaco (1978), "Psychology of Pragmatic
Implications:InformationProcessingBetween the Lines," Journal of ExperimentalPsychology: General, 107, 1-27.
Hoch, Stephen J. and Young-WonHa (1986), "ConsumerLearning: Advertising and the Ambiguity of Product Experience,"
Journal of ConsumerResearch, 13, 221-33.
Houston, David A., Steven J. Sherman,and SaraM. Baker(1989),
"The Influence of Unique Featuresand Direction of Comparison on Preferences," Journal of ExperimentalSocial Psychology, 25, 121-41.
Huber,Joel, John W. Payne, and ChristopherPuto (1982), "Adding AsymmetricallyDominatedAlternative:Violations of Regularityand the SimilarityHypothesis,"Journalof ConsumerResearch, 9, 90-98.
JOURNAL
OF MARKETING
AUGUST1994
RESEARCH,
350
J. Harvey, W. Ickes, and R. Kidd, eds. Hillsdale, NJ: Lawrence
EarlbaumAssociates.
McGill, Ann L. (1989), "Context Effects in Judgmentsof Causation," Journal of Personality and Social Psychology, 57, 189200.
Meyer, Robert J. (1981), "A Model of MultiattributeJudgments
Under Uncertainty and InformationalConstraint," Journal of
MarketingResearch, 18 (November), 428-41.
Miles, Raymond and Charles C. Snow (1978), Organizational
Strategy, Structureand Process. New York:McGraw-Hill.
Nisbett, Richard and Lee Ross (1980), Human Inference: Strategies and Shortcomingsof Social Judgment.Englewood Cliffs,
NJ: Prentice Hall, Inc.
Payne, John W. (1976), "Task Complexity and Contingent Processing in Decision Making," Organizational Behavior and
Human Performance, 16, 366-87.
Peterson,RobertA. (1971), "The Price-PerceivedQualityRelationship," Journal of MarketingResearch, 7, 525-38.
Porter, Michael E. (1985), Competitive Advantage. New York:
The Free Press.
Rao, Akshay R. and Kent B. Monroe (1989), "The Effect of Price,
Brand Name, and Store Name on Buyers' Perceptionsof Product Quality: An IntegrativeReview," Journal of MarketingResearch, 26, 351-57.
W U EU
~-
* WU
U
W
KU
KU
W~
Since such reproductionsare providedwith attachedcovers and preparedin responseto your order,you should be awareof the AMA's no-returnpolicy
on such custom orders. However, the Association still wants you to be completely satisfied with your reprints.For example, if pages are unreadable,
or
$15
$12
or
$22
or
$25
$50
$38
or
$85
$55
or
$98
5-14*
of
reproductions
thesamearticle
$7 each
15-49*
50-149
"
150-500
"
$6 each
$4 each
$3 each
> 500
"
$2 each
"
forInternational
AirS & H
* Please add $6 per title for articles that must be reproducedfrom bound volumes, which generally includes those more than 15 years old but
mayapplyto somemorerecentworks.
As of Jan. 1, 1994, the AMA's PublicationsGroupis no longer in the single-copyreprintbusiness. If you want more informationon orderingone to four
PERMISSIONS
A
to reproduce
To obtainpermission
materials.
of its copyrighted
forthereproduction
AMAproudlygrantspermission
multiplecopies,
f
Law,anyonemaymakea
contacttheAMAPublications
Group312-831-2751.Underthe"fairuse"provisionof theU.S. Copyright
theirown worksin quantityshould A4IE' CAN
Authorswho wantto reproduce
photocopyfor his or herown use withoutseekingpermission.
contact the AMA PublicationsGroup.
AOCMTKN
I