conceptualizing and measuring prospect wants: understanding the source of brand preference

Cust. Need. and Solut. (2014) 1:23–39DOI 10.1007/s40547-013-0001-9

ORIGINAL PAPER

Conceptualizing and Measuring Prospect Wants:Understanding the Source of Brand Preference

Geraldine Fennell · Greg M. Allenby

Published online: 12 December 2013© Springer Science+Business Media New York 2013

Abstract Prospect wants originate upstream from themarketplace, in the context of everyday life and work.Researchers in marketing attempt to read wants by measur-ing and decomposing consumer preferences for marketplaceofferings. In this paper, we show that consumer prefer-ence for offerings reflects an interaction between motivatingconditions that prompt users to action, and capability ofa brand’s attributes to address the source of the motiva-tion. A hierarchical Bayes conjoint model is proposed formeasuring motivating wants that exist upstream from themarketplace and instrumental wants that are expressed asreactions to marketplace offerings. The model is illustratedwith data from a national survey of the concerns and inter-ests that prompt prospects to brush their teeth and theirpreference for toothpaste attributes .

Keywords Conjoint analysis · Motivation · Part-worth ·Hierarchical Bayes

1 Introduction

The concept of wants is central to the discipline of mar-keting and its role in guiding management to make goods

Electronic supplementary material The online version of thisarticle (doi:10.1007/s40547-013-0001-9) contains supplementarymaterial, which is available to authorized users.

G. Fennell · G. M. Allenby (�)Fisher College of Business, Ohio State University,Columbus, USAe-mail: [email protected]

G. Fennelle-mail: [email protected]

and services that people will want to buy. Wants are typ-ically associated with actual or hypothetical marketplaceofferings (e.g., wanting a brand of toothpaste, soda, a pet)and associated attributes (e.g., good breath freshening, cit-rus flavored, easy care). The importance of marketplace, orinstrumental, wants is measured with data that reflect con-sumer preferences for real and hypothetical offerings, oftenusing statistical models (e.g., conjoint analysis) that decom-pose the preference for an offering into utility part-worthsassociated with features and attributes.

Researchers in marketing have a long history of studyingwants and drivers of brand preference. Analysis has histor-ically focused on instruments used in achieving a desiredgoal and, more recently, the goal itself. While there iswide acceptance for a view of motivation as arising fromdisparity between an individual’s current state and theirimagined, desired state [1, 2], theory and research havefavored studying the latter state to the virtual neglect ofthe former state. For example, the analysis of benefits [19],goals [7, 16, 17, 20], and means-end chains [26] describesthe objects, attributes, or the activities that are instrumen-tal for achieving desired imagined states or the imaginedstates themselves. Such analysis does not investigate themotivating conditions that allocate and direct an individual’sresources in the first place, which describe the current stateof the individual. The individual is simply assumed to bemotivated toward the imagined state. While the distinctionbetween motivation and goals is recognized, the implicationof the distinction for understanding prospective user wantshas not been developed.

In this paper, we report an approach to measuring moti-vating wants that describes the current state of the individualand compare it to a traditional analysis of instrumentalwants associated with the imagined state of the individual.Our analysis separately examines both where one is coming

http://dx.doi.org/10.1007/s40547-013-0001-9

mailto:[email protected]

mailto:[email protected]

24 Cust. Need. and Solut. (2014) 1:23–39

from and where one is headed, providing insight into condi-tions for which a brand is preferred. When no variable thatdescribes the current (motivating) state is included, analysisof consumer preferences leaves much ambiguity regardingthe nature of the motivating conditions, which are the con-ditions that valued goods and services must address. Thereexists much literature in psychology and marketing on whatis being pursued, but not how the current state contributesor gives rise to the pursuit.

For example, consumers may report that they want to“look good” and “feel good” in relation to the goal of los-ing body weight. Such items are end points and do not statethe conditions that lead to wanting to “look good.” Is theperson overweight over all their body or just in particu-lar places? Which places? Does their shortness/height entertheir sense of not looking good? Do nonapparent musclesenter their concern? Do they have concerns about hang-ing skin, if they lose weight? Is their sense of not feelinggood due to their having failed to take care of their appear-ance? Do they feel bad because they cannot move easily dueto being overweight? Has their present condition happenedslowly or rapidly? Knowing where one is coming from pro-vides guidance to manufacturers for brand (re)formulationand the creation of media content that is often not availablefrom knowing only the imagined state.

As a secondary issue, we provide evidence that the mea-surement of attribute and benefit importance is confoundedwith brand beliefs. More specifically, we find that the part-worth of an attribute-level is low when the attribute-level isjudged to not describe the current array of brands. This isnot true, however, for the presence of the motivating con-ditions, which is shown to be unrelated to brand beliefs.Studying motivating conditions therefore provides a plat-form for assessing the extent to which the current rangeof brand offerings and their attributes are responsive to themotivating conditions present.

Finally, we demonstrate that motivating conditions canbe combined with information on desired attributes and ben-efits to yield improved predictions of brand preference. Ouranalysis extends the work of Yang et al. [30] who demon-strate the existence of diverse motivating conditions acrossindividuals within, as well as intraindividual variation inmotivations across, objectively specified environments.

The remainder of this paper is organized as follows:In the next section, we lay out the conceptual differencesbetween motivating wants that describe the current stateof the individual and instrumental wants associated withthe attributes and benefits of marketplace offerings and theimagined state. By marketplace or instrumental wants, werefer to wants inferred from reactions to goods and servicesoffered at the retailer, the box office, or on the Internet.In Section 3, we describe a method of measuring motivat-ing wants, illustrating the method with an analysis of the

conditions present in the context for brushing one’s teeth.Data and parameter estimates from our measurement modelare described in Section 4, and in Section 5, we present find-ings from our motivational analysis, along with those from atraditional conjoint analysis of toothpaste. In Section 6, weoffer a conceptual discussion of the information containedin motivating wants and marketplace preferences.

2 Conceptualizing and Measuring Prospect Wants

Figure 1 displays an abbreviated model of action thatfocuses attention on key aspects of our analysis. Personaland environmental systems intersect to produce motivatingconditions that lead to desired benefits and attributes andeventually to marketplace action including brand choice.Motivating conditions allocate an individual’s resources toa domain of action and prompt them to adjust their relation-ship with the environment within that domain. For example,an individual may feel cold because of a drop in the ambienttemperature and become motivated to ease their discom-fort. The individual may look to remedies at hand (e.g.,close the window) and/or marketplace offerings (e.g., asweater) to improve their condition or, weighing resourcesrequired against discomfort, may decide that adjustmentis not cost-worthy, and action is not forthcoming. Finally,ex post and ex ante analysis is relative to the productoffering.

Our model of motivation and behavior is consistent withLewin’s [23] formulation of behavior that comprises per-son (P), environment (E), situation (S), and behavior (B).Person and environment jointly contribute to a situation asperceived (i.e., S = f(P,E)), and behavior is assumed to arisefrom within the situation (i.e., B = f(S)). Other authors (e.g.,[8, 11]) have used person–situation models of the form ofB = f(P,S), which describes variation in behavior but failsto provide access to how the person perceives the environ-ment; such a formulation results in ambiguity regarding thecurrent state of the individual.

For example, Dickson [11] identifies benefits and fea-tures of suntan lotion that arise from various person (e.g.,young children, teenagers, women, men)–situation (e.g.,beach/boating sunbathing, home-poolside sunbathing, sun-lamp sunbathing, snow skiing) settings. Situation bene-fits and features include items like “windburn protection,”“large pump dispenser,” “won’t stain wood/concrete,” and“antifreeze formula” that describe the imagined state andinstrumental attributes, but do not describe the current stateof the individual. Examples of motivating conditions thatdescribe the current state of the individual include itemssuch as “I’m concerned that my pale skin makes me lookunhealthy,” “If I don’t look tanned, I’ll feel out of it incompany with friends who’ve been on vacation,” and “The

Cust. Need. and Solut. (2014) 1:23–39 25

Fig. 1 Model of action andbrand use

season changes, and it’s time to look tanned.” Such itemsdescribe the motivating features of the individual’s currentstate (i.e., where the individual is coming from), which spec-ify features of the desired state (i.e., where the individual isgoing to).

Person and environment are viewed as comprising mul-tiple systems, allowing for a small subset of each inter-secting to produce motivating conditions by instating adesired state, i.e., comparing the present with an imaginedstate, the individual is ready to allocate resources to bringabout the imagined state, expecting or hoping to improvetheir state of being. Viewed from left to right, the modeldisplayed in Fig. 1 represents a behavioral process thatallocates an individual’s resources to a substantive domain(e.g., feeling lonely) and desired state (e.g., reconnectingwith friends) and directs how the individual deploys thoseresources within that domain, favoring actions and objects(e.g., attending a picnic, making a phone call, writing a let-ter) likely to bring about an improved state of being. InFig. 1, motivating wants correspond to motivating condi-tions, and the instrumental wants they specify correspond todesired benefits and attributes. The terms ex ante and ex postsuperimposed in Fig. 1 refer respectively to two conceptsof demand. Ex-post represents a view of demand where theoffering is given; ex-ante is a view of demand based onconditions that pre-exist the offering [13].

Our model of behavior is intended to describe a singleoccasion of an activity. Motivation is operationalized as theconcerns and interests relevant to an activity, in contrast tothe term “motive,” which psychologists have used to referto a trait-like variable intended to apply across activity and

over time (e.g., achievement motive [25]). Moreover, it dif-

fers from approaches to studying goals (e.g., [6, 20, 29])

where interest focuses on identifying a range of goals, from

high levels of abstraction to specific features. Our model is

intended to help investigate concrete concerns and interests

that allocate human resources, not higher-order constructs

that are sometimes of focus of interest.

Consider, for example, the use of means-end theory

[26] to understand drivers of brand preference. The the-

ory assumes that people choose product offerings that can

be instrumental to achieving desired consequences. Using a

procedure of iteratively asking the respondent to state why

each answer is important, they obtain high-order interpreta-

tions of what people want. Regarding alcoholic beverages,

for example, Reynolds and Gutman report that respondents

reply with reasons such as “to socialize,” “avoid getting

drunk,” and “thirst quenching,” often arriving ultimately at a

value statement expressed at a high level of abstraction. The

focus is not on describing the present state in concrete terms.

For the respondent interested in not getting drunk, does he

have a medical condition where alcohol is problematic? Is

he a problem drinker or a designated driver on that occa-

sion? A producer bent on responding to the conditions that

prospect face requires greater guidance than that offered by

simply knowing that the goal is to “avoid getting drunk.” In

this paper, we describe a method for studying the concrete

wants of an individual’s present state and compare it to the

current approach of describing wants in terms of preference

for product attributes and benefits.


2.1 Generating Candidate Items

We employ Fennell’s [14] motivational formulation to gen-erate items for study. This formulation comprises sevenqualitatively distinct classes as described in Table 1, i.e., fivesimple classes and two complex classes that involve mul-tiple conditions. The class structure is used as a guide forgenerating candidate items expressed as concerns and inter-ests and is not used to impose any structural relationshipamong the items in our analysis. It describes qualitativelydistinct kinds of motivating condition that may be present inthe context for an individual’s action. Compared to the usualpurely empirical approach, often guided only by the existingbrand array, the availability of such a set of classes facili-tates the researcher/analyst checking if current offerings areresponsive to the range of qualitatively distinct motivatingconditions. Also included in Table 1 are examples of con-cerns and interests for selected activities. Such examplesillustrate that the motivating classes provide the structure ofdifferent kinds of condition that may allocate resources inany domain of action.

The classes originate in the settings that researchers useto instigate behavior for the experimental study of learningin lower animals and are adapted for use in studying humanbehavior. The first three classes in Table 1 are about mov-ing away from an undesirable state of affairs that is presentfor the individual, whether currently experienced (class 1),imagined to occur at some future time (class 2), or broughtto focal attention only by default (class 3). For example,an individual may engage in oral hygiene activities becauseof concern about bad breath, dull teeth, or to deal with thecurrent conditions that lead to cavities (class 1); because ofconcerns about what their peers, or the actor, themselves,may think if they did not brush (class 2); or simply asrelatively mindless routine (class 3).

Where in the case of the first three classes, the individualmoves away from the source of the motivation, in the case ofthe next two classes, the individual moves toward the sourceof the motivation. Class 4 describes interests that involvemental exploration as, for example, in a hobbiest orientationto the focal activity. Class 5 deals with the pursuit of sensoryenjoyment. An example of a class 4 motivation for tooth-brushing would be interest in knowing about the science oforal hygiene, and an example of a class 5 motivation wouldbe enjoying sensory experiences from brushing.

The final two classes in Table 1 describe complex con-ditions in which the individual is motivated to act but isdeterred from doing so either because of expected harm—excessive cost in the broadest sense (class 6)—or expecteddissatisfaction (class 7). These two classes combine moti-vations to act that occur outside the marketplace, and theexpected outcome of using some version of the product. Anexample of a class 6 item written for toothbrushing would

be agreement with the statement that toothpastes taste toostrong or cost too much. An example of a class 7 item writ-ten for toothbrushing would be toothpastes are not strongenough to prevent cavities.

Candidate items are generated from focus group tran-scripts and the analysis of brand claims in broadcast com-mercials and product packages. Prior to conducting quali-tative research, the analyst is well advised to consider howthe various kinds of motivation may be manifested in thecontext for the focal activity. Only prospects, i.e., respon-dents who qualify as predisposed to buying/using the focalproduct category, are included in qualitative and quantitativephases.

• In the case of class 1, for example, with regard tothe focal activity, the analyst will generate examplesof grave, unpleasant circumstances, or unusual specialcases, whose occurrence is outside the actor’s control inthe short run. Among others, “grave” may refer to inten-sity, speed of onset, or frequency of some condition anindividual dislikes. It is useful to remember that, wherecommon usage invokes the verb, “prevent,” e.g., preventtooth decay, prevent engine wear-out, the motivatingelement that must be dealt with is, in fact, somethingthat is occurring at the present time, for example, sub-stances present in the mouth that are harmful to the teethand gums or wear and tear due to moving metal parts.Although many examples reflect conditions in the rele-vant environment as perceived, personal elements, suchas values that the individual believes are being thwarted,may also contribute examples.

• As regards class 2, at issue are examples of individu-als experiencing discomfort while anticipating how theywill judge themselves, or how they imagine others willjudge them, in the event that they fail to act appropri-ately. Examples comprise imagined censure or failureto gain praise from self or others. Reflecting on exam-ples of psychology’s major constructs, e.g., traits, roles,and self-concepts, as they may be experienced in regardto the focal activity, is a useful source of ideas.

• As regards class 3, at issue is the believed presenceof a state of affairs that requires only minimal mainte-nance for normal functioning. Deterioration is outsidethe actor’s control in the short run, who can do nomore than periodically make good whatever deficit hasoccurred.

• As regards class 4, at issue are examples where the actorbecomes aware that insufficient, too much contradic-tory, unexpected, incongruent information engages theircognitive skills until they resolve matters.

• As regards class 5, the actor believes that actual orimagined sensory pleasure is available, making themfeel deficient until they engage with the experience.


Tabl

e1

Mot

ivat

ions

for

sele

cted

activ

ities

Mot

ivat

iona

lFe

edin

gth

eca

tB

rush

ing

teet

hE

atin

ggr

apef

ruit

Atte

ndin

gliv

eth

eate

r

clas

sT

heca

t-fe

eder

may

be...

The

indi

vidu

alm

aybe

...E

atin

ggr

apef

ruite

may

be...

Ape

rson

may

be...

1.Pr

oble

mT

roub

led

byth

eca

t’ssl

uggi

shm

ovem

ents

,dry

Esc

apin

gfr

omth

eun

plea

sant

proc

ess

ofA

mea

nsof

over

com

ing

leth

argy

,thi

rst,

Seek

ing

rest

orat

ion

for

aw

eary

body

and

solv

ing

skin

,ove

rwei

ghtb

ody,

orla

ckof

appe

tite

bact

eria

inth

em

outh

crea

ting

bad

brea

thhu

nger

,or

nutr

ient

depl

etio

nov

erta

xed

min

d;re

lief

from

bore

dom

,

orda

mag

ing

teet

hor

from

the

uglin

ess

drud

gery

,ban

ality

,and

stul

tifyi

ng

ofte

eth

disc

olor

edor

stai

ned

from

rout

ine

orfr

omab

sorp

tion

with

the

smok

ing

ciga

rette

s/dr

inki

ngco

ffee

conc

erns

ofyo

ung

orai

ling

char

ges;

or

leav

ing

anen

viro

nmen

ttha

tis

oppr

essi

ve

ordi

stra

ctin

gor

lack

ing

inpr

ivac

y

2.Pr

oble

mC

ater

ing

toa

spoi

led

child

,nur

turi

nga

loya

lPr

even

ting

imag

ined

criti

cism

sfr

omO

fsy

mbo

licsi

gnif

ican

cean

dim

plic

ate

Con

side

ring

the

impl

icat

ions

ofat

tend

ing

prev

entio

nfr

iend

,or

tend

ing

anex

pens

ive

stat

ussy

mbo

lon

esel

f/si

gnif

ican

toth

ers

ongr

ound

sth

atre

spon

sibl

eha

bits

orbe

ing

good

toth

epe

rfor

man

cefo

rhi

sor

her

self

-

one

isla

zy,c

arel

ess

ofpe

rson

alhy

gien

e,on

esel

fco

ncep

tas

a(d

isce

rnin

g)cu

ltiva

tor

ofth

e

orla

ckin

gin

cons

ider

atio

ngo

odlif

e,a

gene

rous

prov

ider

/hos

t,or

a

thou

ghtf

ullo

ver/

spou

se/p

aren

t/chi

ld

3.R

outin

eM

indl

essl

ype

rfor

min

ga

rout

ine

chor

eM

aint

aini

nga

syst

emth

atne

eds

only

Be

perf

orm

edm

indl

essl

you

tof

pure

Eng

agin

gin

aro

utin

em

atte

r

mai

nten

ance

rout

ine

atte

ntio

nro

utin

ew

ithm

inim

alin

vest

men

tof

thou

ghta

nd

inte

rest

4.E

xplo

rato

ry“I

nto”

catn

utri

tion,

find

ing

inte

rest

inE

xplo

ring

anin

tere

stin

gqu

estio

nre

late

dB

ea

sour

ceof

intr

insi

cin

tere

stfo

ra

Intr

insi

cally

inte

rest

edin

thea

ter

asa

oppo

rtun

ityle

arni

ngev

erm

ore

and

mor

eab

outt

heto

brus

hing

tech

niqu

esgr

apef

ruit

buff

tow

hom

text

ure,

skin

stud

ento

fhu

man

cond

ition

orth

e

func

tions

ofva

riou

sin

gred

ient

sin

the

cat’s

diet

thic

knes

s,co

lor,

and

smel

lare

fille

daf

icio

nado

fasc

inat

edby

the

com

plex

ities

sign

ific

ance

and

fine

rpo

ints

ofth

eth

eate

rar

ts

5.Se

nsor

yE

mpa

thiz

ing

with

the

cat,

Les

liem

ayta

keE

njoy

ing

the

sens

ory

expe

rien

ces

Be

view

edas

aso

urce

ofpu

rese

nsor

yC

onsi

deri

ngth

eth

eate

ras

anop

port

unity

oppo

rtun

itypl

easu

rein

pres

entin

gan

arra

yof

dele

ctab

leas

soci

ated

with

bris

tleon

gum

s,ta

ste,

plea

sure

tofe

astt

hese

nses

mea

lsto

plea

seth

eca

t’spa

late

and

tingl

eof

dent

ifri

cean

dth

esi

ghto

f

glis

teni

ngpe

arly

teet

h

6.Pr

oduc

t-D

oing

any

ofth

epr

eced

ing

whi

lew

orri

edIn

addi

tion

toon

eor

mor

eof

the

Be

any

ofth

ese

and

also

enta

ilso

me

Perc

eive

das

enta

iling

som

etr

oubl

ing

caus

edab

outc

ost,

trou

ble,

was

te,s

mel

l,an

dot

her

prec

edin

gor

ient

atio

ns,w

orry

ing

abou

tel

emen

tof

unpl

easa

ntne

sssu

chas

disl

ike

elem

ents

,suc

has

expe

nse,

prob

lem

cons

ider

atio

nspo

ssib

leda

mag

eto

skin

from

wat

er,o

rof

pits

,bitt

erne

ss,s

wee

tnes

s,to

ughn

ess,

inco

nven

ienc

eor

poss

ibili

ties

for

soap

,or

abou

twas

ting

time

onpe

rson

alor

size

(“to

ola

rge

for

me”

)em

barr

assm

ento

rfo

rfe

elin

gm

ore

“out

of

care

it”th

anif

one

stay

edho

me

7.Fr

ustr

atio

n“M

akin

gdo

”w

ithfo

odde

liver

ysy

stem

sW

ithon

eor

mor

eof

the

prec

edin

gB

ea

sour

ceof

frus

trat

ion

inth

atth

eFi

ndin

gav

aila

ble

thea

ter

less

enjo

yabl

e

that

are

defi

cien

tin

som

ere

spec

tor

ient

atio

ns,f

rust

rate

dth

atno

pers

on’s

desi

res

are

notb

eing

satis

fied

than

one

wou

ldw

ish

satis

fact

ory

way

ofcl

eans

ing

the

face

is

avai

labl

e


• As regards classes 6 and 7, the individual is alreadymotivated and realizes that taking the indicated actionwill be unduly costly in any of a variety of ways,e.g., time, effort, physical, or psychological side effects,money (class 6), or futile in that the available actionswill not be adequate to the present condition (class 7).

2.2 Measuring the Presence of Motivating Wants

Measuring the presence of an individual’s concerns andinterests that describe their current state is different frommeasuring the importance of the imagined state associatedwith owning and using attributes that are available in goodsand services. Many product attributes are defining, in thesense that all versions of a product must possess somelevel of the attribute. For example, all apartments have floorspace, all computers have CPUs, and all credit cards haveinterest rates. In contrast, not all apartments have balconies,so a balcony is not a defining attribute. It is not possibleto measure the importance of a defining attribute, only theimportance of changing the level of an attribute. As a result,measuring the importance of product features employs aninterval scale because the presence of defining attributesrules out the presence of a natural “zero” point.

In contrast, we view motivating conditions as being eitherpresent or not present. An individual may not be moti-vated by any particular want and can certainly have multiplesources of motivation. For example, an individual brushingtheir teeth can simultaneously be concerned about cavities,bad breath, and social impact. Therefore, a natural zeroexists when studying motivating conditions, and there doesnot exist the concept of a defining attribute. When measur-ing a motivating condition, the researcher must allow for itspossible absence.

The objects of analysis when studying motivating wantsare conditions that people experience in the context of anactivity. Writing the questionnaire and toothbrushing occa-sions are described in terms of the concerns and interestspresent, and the similarity of the description to the respon-dent’s own toothbrushing concerns and interests is used asa basis for inferring importances. In contrast, the objectsof analysis in a traditional conjoint study are attributes ofproduct offerings, and the importance of attributes and theirlevels is derived using preference data.

We use a conjoint-like technique for measuring moti-vating conditions. Hypothetical toothbrushing occasions aredescribed by the concerns and interests present, and thedependent variable is the similarity of the description tothe respondent’s own concerns and interests as reported.Respondents are instructed to reflect on a specific occasionof an activity (e.g., the last time you brushed your teeth) andindicate the perceived similarity of the descriptions to theirown motivations. Since the concerns and interests that lead

individuals to their actions are ratio-scaled, it is important toinclude a “null” description in which none of a set of moti-vating conditions is present. Figure 2 provides an exampleset of stimuli for toothbrushing.

Other aspects of the design of the stimuli and analysisof the responses are identical to traditional conjoint analy-sis [18]. The stimuli can be constructed using methods ofexperimental design, including the use of fractional facto-rial designs [4, 22]. The dependent variable can be choices,ranks or ratings, and likelihood specified as a linear or latentlinear model (see [24]). Moreover, respondent heterogeneitycan be incorporated into the analysis using continuous [5],finite mixture densities [21], and continuous mixture modelswith covariates [10, 27].

3 Method

We investigate differences between motivating and instru-mental wants by comparing the concerns and interests thatlead individuals to brush their teeth with the importance oftoothpaste attributes and benefits. Concerns and interestsfor toothbrushing were obtained from qualitative studiesthat included focus groups among prospects (see [28]) inwhich the moderator used the motivation classes to guidediscussion. Table 2 displays the 31 candidate concerns andinterests used in our analysis.

The attributes and benefits (a/b) of toothpaste are dis-played in Table 3. These items are matched to correspond tothe concerns and interest (c/i) items in Table 2. For example,the toothpaste benefit “helps remove stains” corresponds toconcern “my teeth stain easily”; “helps take away morningbreath” corresponds to the concern “I wake up with a badtaste/feeling in my mouth.”

Table 4 lists the toothbrushing c/i items in Table 2 nextto the corresponding toothpaste a/b items in Table 3. Thematch between the c/i and a/b items is intended to be close,with the difference only reflecting the change in the word-ing needed to move from the c/i object (i.e., motivatingconditions that hypothetical people experience) to the a/bobject (i.e., attributes of hypothetical product offerings) ofanalysis. By an oversight, we did not write an a/b corre-sponding to the c/i “toothpastes claim more than they candeliver.” Overall, 30 out of the 31 c/i items were matchedto a/b items.

The presence and importance of the c/i and a/b itemswas measured using a conjoint model based on rank data.For the c/i items, ten sets of stimuli (see Table 2) wereprovided to respondents with each set comprising fourtriplets. Ranks were obtained for each of the four tooth-brushing occasions described. Each triplet comprises threec/i items, and respondents indicated the agreement betweenthese statements and their own c/i’s during the last time


Fig. 2 Example stimuli formeasuring the importance ofmotivating conditions

QUESTION 1

Person BJ Person AW Person MC Person JDStains, bad

taste/feeling in my mouth and gums aren't

a problem for me

My teeth stain easily. I wake up with bad taste/ feeling in my

mouth.

I am concerned about the condition of my

gums.

Sensitive teeth, tartar, plaque and bad breath aren't a problem for

me.

I am predisposed to having sensitive teeth.

I am concerned about tartar and plaque

build-up on my teeth.

I am concerned about bad breath.

Regularly brushing my

my self image, or the impression I want to

create.

letting myself down if I

regularly.

I believe that people expect me to brush

regularly.

I believe that people expect me to brush

regularly.

they brushed their teeth. We varied the c/i items comprisingthe toothbrushing occasions across the ten sets of stimuli,

and dummy variable coding was used to parameterize thehypothetical occasions. The rank data were modeled using

Table 2 Concerns andinterests for toothbrushing Problem solving: Exploratory opportunity:

A1: My teeth stain easily. D1: I like to try different oral brushing

A2: I wake up with a bad taste/feeling techniques/routines just for a

in my mouth. change of pace.

A3: I am concerned about the condition D2: I’m interested in knowing about the

of my gums science of oral hygiene—including different

A4: I am predisposed to having kinds of brushes and toothpastes.

sensitive teeth.

A5: I am concerned about tartar and

plaque build-up on my teeth. Sensory opportunity:

A6: I am concerned about bad breath. E1: I like the tingle I feel in my mouth after

A7: My teeth are dull/not white enough. I brush.

A8: I am predisposed to having cavities. E2: I enjoy the fresh taste I get from brushing.

A9: I have trouble getting my kids to E3: I love to see my teeth gleam like pearls.

brush. E4: Bubbling action adds to the sensory

A10: I am concerned there are cavity pleasure of brushing.

prone places on my teeth.

A11: I am concerned about germs and Product-caused problems:

mouth infections. F1: Toothpastes are too strong tasting.

A12: I am concerned about not getting F2: Toothpastes scratch the enamel on

to hard to reach places. my teeth.

F3: Toothpastes irritate my mouth.

Problem prevention F3: Toothpastes irritate my mouth.

B1: I would feel I’m letting myself down F4: Toothpastes cost too much.

if I didn’t brush regularly F5: Toothpastes contain artificial ingredients.

B2: I believe that people expect me to F6: Toothpaste packaging can be harmful to

brush regularly. the environment.

Routine maintenance: Frustration:

C1: I don’t have problems, worries or G1: Toothpastes aren’t strong enough to

interests regarding my teeth. I just prevent cavities.

brush my teeth regularly. G2: Toothpaste breath-freshening doesn’t

C2: For me, brushing my teeth is just last long enough.

something I do with little thought or G3: Toothpastes claim more than they can

interest. deliver.


Table 3 Attributes and benefits of toothpaste

Medical benefits: Price:

A1: Helps prevent cavities. F1: Regular price.a

A2: Delivers protection in hard to reach F2: 20 % less.

places.

A3: Helps remove tartar and plaque. Ingredients:

A4: Helps promote healthy gums. G1: 80 % natural / 20 % artificial ingredientsa

A5: Penetrates to strengthen your teeth G2: 100 % natural ingredients.

against cavities.

A6: Helps fight germs and infections Packaging:

in your mouth. H1: 80 % recyclable packaging.a

H2: 100 % recyclable packaging.

Taste:

B1: Mild tasting. Interests:

B2: Fresh tasting. I1: An interesting way to clean your teeth.

B3: Gives your mouth a tingle I2: Provides a change of pace.

B4: A taste kid’s love.

B5: Great bubbling action. Social:

J1: Shows others you care about your teeth.

Abrasiveness: J2: Help you feel good about yourself for

C1: Doesn’t irritate my mouth. brushing regularly.

C2: For sensitive teeth.

C3: Safe for tooth enamel (non- Maintenance:

scratching) K1: For everyday brushing.

K2: For routine maintenance.

Resulting appearance:

D1: Helps clean teeth.

D2: Helps remove stains.

D3: Whitens your teeth.

D4: Makes your teeth gleam like pearls.

Resulting breath:

E1: Fights bad breath.

E2: Freshens breath for 12 hours.

E3: Helps take away morning mouth.

aNull conditions

a logit model, in which the probability of observing a par-ticular rank ordering for the four triplets presented togetheris equal to:

Pr (U1j > U2j > U3j > U4j )h =3∏

i = 1

exp(zij′γh)

4∑k= i

exp(zkj′γh)

(1)

where U1j is assumed to be the triplet with highest rankin the j th stimulus set, U2j has the second highest rank,etc., zij is the dummy variable coding of the c/i’s for theith-ranked triplet in the j th set, and γh is the vector of c/iimportance weights for the respondent (h).

One triplet of c/i items in each set of four comprises items

describing the absence of the motivating conditions present

in the other three triplets (see Fig. 2, left column). As noted

earlier, the c/i items can each either be present or absent

on an occasion for the focal activity, and it is necessary to

measure the absence of a motivating condition as well as

its presence. We represent the hypothetical “null” condition

by a vector z with elements all equal to zero. This coding

scheme leads to estimates of the elements of γ which, if

positive, indicate the presence of the corresponding moti-

vating condition, and, if negative, correspond to the absence

of the condition. The magnitude of the coefficient indicates


Table 4 Concerns/interestsand matched attributes/benefits Concerns and interests Attributes and benefits

A1: My teeth stain easily. D2: Helps remove stains.

A2: I wake up with a bad taste/feeling in my mouth. E3: Helps take away morning breath.

A3: I am concerned about the condition of my gums A4: Helps promote healthy gums.

A4: I am predisposed to having sensitive teeth. C2: For sensitive teeth.

A5: I am concerned about tartar and plaque build- A3: Helps remove tartar and plaque.

up on my teeth.

A6: I am concerned about bad breath. E1: Fights bad breath.

A7: My teeth are dull/not white enough. D3: Whitens your teeth.

A8: I am predisposed to having cavities. A1: Helps prevent cavities.

A9: I have trouble getting my kids to brush. B4: A taste kid’s love.

A10: I am concerned there are cavity prone places A5: Penetrates to strengthen your teeth

on my teeth. against cavities.

A11: I am concerned about germs and mouth A6: Helps fight germs and infections

infections. in your mouth.

A12: I am concerned about not getting to hard to A2: Delivers protection in hard to reach places.

reach places.

B1: I would feel I’m letting myself down if I didn’t J2: Helps you feel good about yourself for

brush regularly. brushing regularly.

B2: I believe that people expect me to brush J1: Shows others you care about your teeth.

regularly.

C1: I don’t have problems, worries or interests K2: For routine maintenance.

regarding my teeth. I just brush my teeth regularly.

C2: For me, brushing my teeth is just K1: For everyday brushing.

something I do with little thought or interest.

D1: I like to try different oral brushing techniques/ I2: Provides a change of pace.

routines just for a change of pace

D2: I’m interested in knowing about the science of I1: An interesting way to clean teeth.

oral hygiene – including different kinds of brushes

and toothpastes.

E1: I like the tingle I feel in my mouth after I brush. B3: Gives your mouth a tingle.

E2: I enjoy the fresh taste I get from brushing. B2: Fresh tasting.

E3: I love to see my teeth gleam like pearls. D4: Makes your teeth gleam like pearls.

E4: Bubbling action adds to the sensory pleasure B5: Great bubbling action.

of brushing.

F1: Toothpastes are too strong tasting. B1: Mild tasting.

F2: Toothpastes scratch the enamel on my teeth. C3: Safe for tooth enamel (non-scratching).

F3: Toothpastes irritate my mouth. C1: Doesn’t irritate your mouth.

F4: Toothpastes cost too much. F2: 20 % less than regular price.

F5: Toothpastes contain artificial ingredients. G2: 100 % natural ingredients.

F6: Toothpaste packaging can be harmful to the H2: 100 % recyclable packaging.

environment.

G1: Toothpastes aren’t strong enough to prevent D1: Help clean teeth.

cavities.

G2: Toothpaste breath-freshening doesn’t last long E2: Freshens breath for 12 hours.

enough.

G3: Toothpastes claim more than they can deliver.


likelihood of absence (−) or presence (+) of the item to therespondent.

The importance of the a/b items is measured in a simi-lar fashion. Hypothetical product offerings described by thea/b items were presented to the respondent, who was askedto provide a rank ordering of the objects in terms of theirpreference. Ten sets of stimuli were presented, with eachcomprising four hypothetical product offerings describedby three attribute-levels. Respondents were told that the a/bitems not listed in the description were the same for theofferings. The likelihood of the rank ordering for the fourtriplets in one of the sets is equal to:

Pr (V1m > V2m > V3m > V4m)h =3∏

i = 1

exp(xim′βh)

4∑j = i

exp(xjm′βh)

(2)

where V1m is assumed to be the utility of the hypotheticalproduct offering with the highest rank in the mth set, xi isthe dummy variable coding of the a/b’s for the ith rankedoffering in set m, and βh is the vector of a/b importanceweights (part-worths) for respondent h.

In contrast to the coding for the c/i analysis, we donot include a null alternative in each of the four tripletproduct offering sets. The null attribute-levels for the a/banalysis are indicated by footnote a in Table 3: F1 (reg-ular price); G1 (80 % natural/20 % artificial ingredients);and H1 (80 % recyclable packaging). We view the attributesof price, ingredients and packaging as defining, and theremaining attributes as optional for toothpaste. For example,toothpastes can exist that do not provide any medical bene-fit, or have any taste, or any breath-freshening properties. Itis possible to describe toothpaste without reference to theseattributes.

Despite the lack of a null offering in each of ten triplets,the model for estimating the a/b part-worths from the prod-uct rank data is statistically identified. The likelihood for anindividual’s ranks is defined across all ten triplets:

�(γh |Data) =10∏

j = 1

Pr (U1j > U2j > U3j > U4j )h

=10∏

j = 1

3∏

i = 1

exp(zij′γh)

4∑k= i

exp(zkj′γh)

(3)

�(βh |Data) =10∏

m= 1

Pr (V1m > V2m > V3m > V4m)h

=10∏

m= 1

3∏

i = 1

exp(xim′βh)

4∑j = i

exp(xjm′βh)

(4)

Therefore, the identifying restrictions for the model extendbeyond the specific a/b items present in any one of thetriplets. It is not possible to arbitrarily increase the value ofany or all of the elements of βh without changing the valueof the likelihood for the entire set of ranks.

Heterogeneity is incorporated into the model specifi-cation by assuming a random-effects distribution for theparameters:

θh = (γh′, βh

′)′ ∼ Normal(μ, �) (5)

Markov chain Monte Carlo (MCMC) methods are used toestimate the model parameters. The chain was run for atotal of 50,000 iterations, with parameter estimates basedon the last 10,000 iterations. We investigate multiple startpoints and found the chain to converge to a common poste-rior distribution. The estimation algorithm is provided in theOnline Appendix.

4 Data and Parameter Estimates

Data were obtained from a nationally representative panelin mailed questionnaires administered by a leading market-ing research firm. 863 completed surveys were availablefor analysis. The data in the survey included ten sets ofstimuli each comprising four triplets of c/i descriptions oftoothbrushing occasions, and ten sets of stimuli each com-prising four triplets of a/b descriptions of toothpaste. Brandbelief ratings for Aquafresh, Colgate, Crest and Mentadentwere also obtained by asking respondents to rate each brandon each of the 30 a/b items using a five-point scale where“5” means “describes completely” and “1” means “does notdescribe at all.” For example, respondents were asked toindicate the degree to which attribute A1: “Helps preventcavities” describes each brand, and so on. Finally, actualbrand use information was obtained by asking respondentsto identify whether they usually buy a particular brand oftoothpaste, and if so, which brand. Estimates of the meanof the random-effect distribution are reported in Table 5.Estimates of the covariance matrix � are not reported butare available from the authors upon request. The fit of themodel described by Eqs. 3–5 is good, with an average in-sample hit probability of 0.60. We find that the responses tothe c/i and a/b items were of equal consistency as measuredby in-sample fit.

Estimates of the mean of the random-effects distribu-tion for the c/i items are different than those for the a/bitems. The most prevalent c/i item is B1 “I would feel I’mletting myself down if I didn’t brush regularly” with an aver-age importance of 3.211 and a posterior standard deviationof 0.132. The most important a/b item is B3 “Gives yourmouth a tingle” with an average importance of 1.658 and


Tabl

e5

Est

imat

esof

the

mea

n(μ

)of

the

rand

om-e

ffec

tsdi

stri

butio

n

Con

cern

san

din

tere

sts

Post

erio

rm

ean

Attr

ibut

esan

dbe

nefi

tsPo

ster

ior

mea

n

A1:

My

teet

hst

ain

easi

ly.

0.56

7A

1:H

elps

prev

entc

aviti

es.

−0.2

18A

2:I

wak

eup

with

aba

dta

ste/

feel

ing

inm

ym

outh

.1.

464

A2:

Del

iver

spr

otec

tion

inha

rdto

reac

hpl

aces

.−1

.025

A3:

Iam

conc

erne

dab

outt

heco

nditi

onof

my

gum

s1.

845

A3:

Hel

psre

mov

eta

rtar

and

plaq

ue.

−0.2

36

A4:

Iam

pred

ispo

sed

toha

ving

sens

itive

teet

h.−1

.105

A4:

Hel

pspr

omot

ehe

alth

ygu

ms.

−0.3

46

A5:

Iam

conc

erne

dab

outt

arta

ran

dpl

aque

build

-up

onm

yte

eth.

0.64

5A

5:Pe

netr

ates

tost

reng

then

your

teet

hag

ains

tcav

ities

.0.

504

A6:

Iam

conc

erne

dab

outb

adbr

eath

.0.

373

A6:

Hel

psfi

ghtg

erm

san

din

fect

ions

inyo

urm

outh

−0.8

30

A7:

My

teet

har

edu

ll/no

twhi

teen

ough

.0.

660

B1:

Mild

tast

ing.

1.05

3

A8:

Iam

pred

ispo

sed

toha

ving

cavi

ties.

−0.5

79B

2:Fr

esh

tast

ing.

1.15

8

A9:

Iha

vetr

oubl

ege

tting

my

kids

tobr

ush.

−1.0

20B

3:G

ives

your

mou

tha

tingl

e.1.

658

A10

:Iam

conc

erne

dth

ere

are

cavi

typr

one

plac

eson

my

teet

h.−0

.167

B4:

Ata

ste

kid’

slo

ve−0

.861

A11

:Iam

conc

erne

dab

outg

erm

san

dm

outh

infe

ctio

ns.

0.26

2B

5:G

reat

bubb

ling

actio

n.−0

.950

A12

:Iam

conc

erne

dab

outn

otge

tting

toha

rdto

reac

hpl

aces

.0.

003

C1:

Doe

sn’t

irri

tate

my

mou

th.

−0.6

91

B1:

Iw

ould

feel

I’m

letti

ngm

ysel

fdo

wn

ifI

didn

’tbr

ush

regu

larl

y.3.

211

C2:

For

sens

itive

teet

h.−1

.208

B2:

Ibe

lieve

that

peop

leex

pect

me

tobr

ush

regu

larl

y2.

105

C3:

Safe

for

toot

hen

amel

(non

-scr

atch

ing)

.0.

662

C1:

Ido

n’th

ave

prob

lem

s,w

orri

esor

inte

rest

sre

gard

ing

my

teet

h.1.

018

D1:

Hel

pscl

ean

teet

h.0.

300

Iju

stbr

ush

my

teet

hre

gula

rly.

C2:

For

me,

brus

hing

my

teet

his

just

som

ethi

ngI

dow

ith0.

962

D2:

Hel

psre

mov

est

ains

.−0

.941

little

thou

ghto

rin

tere

st.

D1:

Ilik

eto

try

diff

eren

ttee

thbr

ushi

ngte

chni

ques

/1.

158

D3:

Whi

tens

your

teet

h.0.

161

rout

ines

just

for

ach

ange

ofpa

ce.

D2:

I’m

inte

rest

edin

know

ing

abou

tthe

scie

nce

oror

al0.

920

D4:

Mak

esyo

urte

eth

glea

mlik

epe

arls

.−1

.339

hygi

ene—

incl

udin

gdi

ffer

entt

ypes

ofbr

ushe

san

dto

othp

aste

s

E1:

Ilik

eth

etin

gle

Ife

elin

my

mou

thaf

ter

Ibr

ush.

0.12

6E

1:Fi

ghts

bad

brea

th.

0.22

7

E2:

Ien

joy

the

fres

hta

ste

Ige

tfro

mbr

ushi

ng.

0.71

7E

2:Fr

eshe

nsbr

eath

for

12ho

urs.

0.88

6

E3:

Ilo

veto

see

my

teet

hgl

eam

like

pear

ls.

0.17

2E

3:H

elps

take

away

mor

ning

mou

th.

0.25

7

E4:

Bub

blin

gac

tion

adds

toth

ese

nsor

ypl

easu

reof

brus

hing

−0.4

87F2

:Pri

ce20

%le

ss0.

386

F1:T

ooth

past

esar

eto

ost

rong

tast

ing.

−0.6

47G

2:10

0%

natu

rali

ngre

dien

ts.

0.99

3

F2:T

ooth

past

essc

ratc

hth

een

amel

onm

yte

eth.

−0.7

17H

2:10

0%

recy

clab

lepa

ckag

ing.

0.81

1

F3:T

ooth

past

esir

rita

tem

ym

outh

.−0

.684

I1:A

nin

tere

stin

gw

ayto

clea

nyo

urte

eth.

−0.9

00

F4:T

ooth

past

esco

stto

om

uch.

−1.3

80I2

:Pro

vide

sa

chan

geof

pace

.−0

.493

F5:T

ooth

past

esco

ntai

nar

tific

iali

ngre

dien

ts.

−1.3

63J1

:Sho

ws

othe

rsyo

uca

reab

outy

our

teet

h.−0

.590

F6:T

ooth

past

epa

ckag

ing

can

beha

rmfu

lto

the

envi

ronm

ent.

−2.4

02J2

:Hel

psyo

ufe

elgo

odab

outy

ours

elf

for

brus

hing

regu

larl

y0.

670

G1:

Toot

hpas

tes

aren

’tst

rong

enou

ghto

prev

entc

aviti

es.

0.39

5K

1:Fo

rev

eryd

aybr

ushi

ng.

0.40

2

G2:

Toot

hpas

tebr

eath

-fre

shen

ing

does

n’tl

astl

ong

enou

gh0.

581

K2:

For

rout

ine

mai

nten

ance

.−0

.830

G3:

Toot

hpas

tes

clai

mm

ore

than

they

can

deliv

er.

0.37

1xx

x

Est

imat

esth

atar

em

ore

than

two

post

erio

rst

anda

rdde

viat

ions

from

zero

are

italic

ized


a posterior standard deviation of 0.070. Thus, a differentview of wants emerges from studying c/i’s and a/b’s respec-tively. The two most present c/i items are B1 “I would feelI’m letting myself down if I didn’t brush regularly” and B2“I believe that people expect me to brush regularly.” Theranks of the corresponding a/b items are 7 and 20. The mostimportant a/b items deal with taste attributes: B3 ”Givesyour mouth a tingle,” B2 ”Fresh tasting,” and B1 ”Mildtasting.” The associated c/i ranks are 20, 9, and 24.

In the next section we explore the source of the differ-ences between the c/i and a/b analyses. Respondents appearto interpret statements such as “I believe that people expectme to brush regularly,” very differently from statementssuch as “Shows others you care about your teeth.” The twosets of items are clearly providing different implications forproduct policy, despite our having closely matched the a/band c/i items.

5 Findings

In this section we document three limitations of usingattribute preferences to measure prospective user wants.First, we find that attribute-level part-worth estimates (β)are confounded with the capabilities of current offerings,while the presence of the concerns and interests (γ ) areunrelated to these capabilities. Attribute preferences there-fore do not offer an independent assessment of marketdemand. Second, we find evidence of a complex relation-ship between (γ ) and the attribute-level part-worths (β).Hence, knowledge of attribute importance is not sufficientfor understanding motivating conditions. Finally, we showthat the concerns and interests (γ ) can be used in con-junction with the attribute-level part-worths (β) to improvebrand choice predictions. This indicates that the currentarray of attributes and levels do not completely respond tothe existing motivating wants of individuals, providing indi-cation of unmet demand in the marketplace. The analysisreported below employs the parameter estimates reportedabove, in conjunction with the analysis of data on brandbelief ratings and brand use that were collected during thesurvey.

5.1 The Confounding Influence of Current Capabilities

Figure 3 provides a plot of the average importance of the c/iitems versus the average brand belief ratings for Aquafresh,Colgate, Crest and Mentadent on each of the a/b items.The importance of the c/i items is measured in terms ofthe mean of the random-effects distribution of γ reportedon the left side of Table 6. For each of the 30 c/i itemsthat have a matched attribute and benefit (see Table 5),brand belief ratings, measured on the five-point “describes

completely”/“not at all” scale, were averaged over the fourbrands. This procedure results in the respondents’ averagebrand belief ratings of leading brands in the product cate-gory on items corresponding to the c/i items. Data labels(e.g., H2) refer to toothpaste attributes and benefits listed inTable 2.

Figure 3 contains three outliers defined as observationsthat are away from the bulk of the data. The three outlierscorrespond to extreme levels of the attributes: uses 100 %recyclable packaging and 100 % natural ingredients andfreshens breath for 12 h. For the remaining c/i items, thecorrelation between the average importance and the averagebrand rating is near zero (r = 0.02, p = 0.91).

Figure 4 provides a corresponding plot for the a/b items.The three outliers are more pronounced in this figure,clearly separated from the rest of the data. Respondentsplace a high value on the attribute-levels of the outliers,but feel that the leading brands are not well described bythe extreme levels of these attributes. The remaining pointsin the plot exhibit a fairly strong association (r = 0.52,p = 0.01) between the average brand belief rating and thea/b coefficients, suggesting that there may be a confound-ing influence between the importance of an attribute-leveland the ability of the currently available brands to offer cor-responding value. Data labels (e.g., H2) refer to toothpasteattributes and benefits listed in Table 3.

5.2 Motivational Ambiguity

The presence of a confounding influence of the currentbrands is just one reason that analysis based on c/i itemswould differ from analysis based on a/b items. Anotherreason is the presence of motivational ambiguity, whereconsumers may construe a particular attribute or benefitto be responsive to their c/i, which is unobserved to theresearcher. We argue that the relationship between the a/band c/i items is not as simple as the paired relationship dis-played in Table 4 due to cross-sectional heterogeneity. If therelationship were in fact paired uniformly among respon-dents, then the correlation structure of the β and γ itemswould be similar, with any associations among the β itemsalso present among the γ items.

Our estimate of the covariance matrix (reported in theonline appendix) has 380 of the 930 covariances betweenthe γ and β elements with posterior mass away from zero,indicating the likely presence of cross-sectional ambigu-ity where different respondents see different a/b items as aresponse to a particular c/i. We find, for example, that thecross-sectional association among the c/i items (B1, B2) and(E1, E2, E3) is uniformly positive, while the correspondingassociations in terms of the a/b items (J1, J2) and (B1, B2,B3) are both positive and negative. Despite the close matchbetween them as shown in Table 4, the presence of both


Fig. 3 Presence of C/I vs.average brand ratings

432

3

2

1

0

-1

Avg. Brand Rating

c/i C

oeffi

cien

t

K2 K1

J2

J1

I2I1

H2

G2 F2 E3

E2E1

D4

D3 D2

D1C3C2

C1B5

B4

B3

B2

B1

A6

A5

A4

A3

A2

A1

Contains 100 % natural ingredients

Freshens breath for 12 hours

Uses 100 %recyclable packaging

r = 0.02 for rest

positive and negative covariations among c/i and a/b items

indicates that prospects may mistakenly read into claimed

attributes of offerings (a/b) as a response to the conditions

they experience (c/i). Individuals who are concerned that “I

would feel I’m letting myself down if I didn’t brush regu-

larly” prefer the a/b “Gives your mouth a tingle” and may

construe a tingling sensation as being responsive to this

social concern [15].

5.3 Predictive Performance

The degree to which an analysis of attributes and benefits

fully reflects the range of motivating wants, or is located in

an optimal region of demand, can be partially assessed by

its predictive relationship to actual brand use. We caution,

however, in using brand predictions as the only standard

for assessing the information content of attributes and ben-

efits. As illustrated in Fig. 1, the movement from attribute

preferences to actual brand purchase involves a number of

constructs that are not studied in our analysis, including

consideration sets and shelf prices. While we view the pre-

diction of brand use as informative about the information

contained in a/b and c/i items, we note that our analysis is

limited to a subset of important variables.

We assess the predictive performance of the c/i and a/b

items by using them to weigh the brand belief ratings data

collected for the Aquafresh, Colgate, Crest, and Mentadent

brands to arrive at an overall score for each brand. The

brand with the highest score is predicted to be the brand that

the respondent will use, and this prediction is compared to

the actual brand used by the respondent as reported in the

questionnaire.

We employ a Bayesian approach to measuring predic-

tive performance using likelihood ratios. The predicted (P)

and true (T) choices for each brand are related using Bayes’theorem:

Pr(T = + |P = +)

Pr(T = − |P = +)= Pr(T = +)

Pr(T = −)× Pr(P = + |T = +)

Pr(P = + |T = −)

(6)

or

Posterior Odds = Prior Odds × Likelihood Ratio

where “+” indicates that the brand is actually chosen orpredicted to be chosen and “−” otherwise.

The “Prior Odds” is equal to the odds of brand prefer-ence without knowledge of the c/i or a/b information. Theposterior odds is the prediction of brand preference giveninformation from the c/i or a/b (or both), plus informationabout the prior odds. The likelihood ratio summarizes thepredictive information from the c/i and a/b about the truepreferences. When the likelihood ratio is greater than 1, theposterior odds are greater than the prior odds, and whenthe ratio is less than 1, the posterior odds are less than theprior odds. We would expect a LR > 1 when the predic-tion is + and a LR < 1 when the prediction is −. Thelikelihood ratios were computed for 578 of the 863 respon-dents. The 285 respondents not included in the predictiveanalysis either did not provide brand ratings informationor used a brand that was different from the four brands inour analysis.

Table 7 displays the likelihood ratios for each brand,using the a/b and c/i importances to weigh the brand ratingsto obtain an overall measure of brand value. As expected,the likelihood ratios are greater than 1 when the brand ispredicted to be the favorite brand and less than 1 when thebrand is predicted to not be the favorite brand. This indicatesthat the a/b and c/i importance measures have predictivevalidity. In addition, we find that the c/i’s lead to likelihoodratios that are more predictively accurate than the a/b’s, with


Tabl

e6

Ran

ked

conc

erns

and

inte

rest

san

das

soci

ated

attr

ibut

ean

dbe

nefi

ts

c/iR

ank

Con

cern

and

Inte

rest

Item

sC

orre

spon

ding

Attr

ibut

ean

dB

enef

itIt

ems

a/b

Ran

k

1B

1:I

wou

ldfe

elI’

mle

tting

mys

elf

dow

nif

Idi

dn’t

brus

hre

gula

rly.

J2:H

elps

you

feel

good

abou

tyou

rsel

ffo

rbr

ushi

ngre

gula

rly

7

2B

2:I

belie

veth

atpe

ople

expe

ctm

eto

brus

hre

gula

rly

J1:S

how

sot

hers

you

care

abou

tyou

rte

eth.

20

3A

3:I

amco

ncer

ned

abou

tthe

cond

ition

ofm

ygu

ms

A4:

Hel

pspr

omot

ehe

alth

ygu

ms.

18

4A

2:I

wak

eup

with

aba

dta

ste/

feel

ing

inm

ym

outh

.E

3:H

elps

take

away

mor

ning

mou

th.

13

5D

1:I

like

totr

ydi

ffer

entt

eeth

brus

hing

tech

niqu

es/

I2:P

rovi

des

ach

ange

ofpa

ce.

19

rout

ines

just

for

ach

ange

ofpa

ce.

6C

1:I

don’

thav

epr

oble

ms,

wor

ries

orin

tere

sts

rega

rdin

gm

yte

eth

K2:

For

rout

ine

mai

nten

ance

.22

Iju

stbr

ush

my

teet

hre

gula

rly.

7C

2:Fo

rm

e,br

ushi

ngm

yte

eth

isju

stso

met

hing

Ido

with

K1:

For

ever

yday

brus

hing

.10

little

thou

ghto

rin

tere

st.

8D

2:I’

min

tere

sted

inkn

owin

gab

outt

hesc

ienc

eor

oral

I1:A

nin

tere

stin

gw

ayto

clea

nyo

urte

eth.

25

hygi

ene

–in

clud

ing

diff

eren

ttyp

esof

brus

hes

and

toot

hpas

tes

9E

2:I

enjo

yth

efr

esh

tast

eI

getf

rom

brus

hing

.B

2:Fr

esh

tast

ing.

2

10A

7:M

yte

eth

are

dull/

notw

hite

enou

gh.

D3:

Whi

tens

your

teet

h.15

11A

5:I

amco

ncer

ned

abou

ttar

tar

and

plaq

uebu

ild-u

pon

my

teet

h.A

3:H

elps

rem

ove

tart

aran

dpl

aque

.17

12G

2:To

othp

aste

brea

th-f

resh

enin

gdo

esn’

tlas

tlon

gen

ough

E2:

Fres

hens

brea

thfo

r12

hour

s.5

13A

1:M

yte

eth

stai

nea

sily

.D

2:H

elps

rem

ove

stai

ns.

26

14G

1:To

othp

aste

sar

en’t

stro

ngen

ough

topr

even

tcav

ities

.D

1:H

elps

clea

nte

eth.

12

15A

6:I

amco

ncer

ned

abou

tbad

brea

th.

E1:

Figh

tsba

dbr

eath

.14

16G

3:To

othp

aste

scl

aim

mor

eth

anth

eyca

nde

liver

.xx

x

17A

11:I

amco

ncer

ned

abou

tger

ms

and

mou

thin

fect

ions

.A

6:H

elps

figh

tger

ms

and

infe

ctio

nsin

your

mou

th23

18E

3:I

love

tose

em

yte

eth

glea

mlik

epe

arls

.D

4:M

akes

your

teet

hgl

eam

like

pear

ls.

30

19A

10:I

amco

ncer

ned

ther

ear

eca

vity

pron

epl

aces

onm

yte

eth.

A5:

Pene

trat

esto

stre

ngth

enyo

ute

eth

agai

nstc

aviti

es.

9

20E

1:I

like

the

tingl

eI

feel

inm

ym

outh

afte

rI

brus

h.B

3:G

ives

your

mou

tha

tingl

e.1

21A

12:I

amco

ncer

ned

abou

tnot

getti

ngto

hard

tore

ach

plac

es.

A2:

Del

iver

spr

otec

tion

inha

rdto

reac

hpl

aces

.28

22E

4:B

ubbl

ing

actio

nad

dsto

the

sens

ory

plea

sure

ofbr

ushi

ngB

5:G

reat

bubb

ling

actio

n.27

23A

8:I

ampr

edis

pose

dto

havi

ngca

vitie

s.A

1:H

elps

prev

entc

aviti

es.

16

24F1

:Too

thpa

stes

are

too

stro

ngta

stin

g.B

1:M

ildta

stin

g.3

25F3

:Too

thpa

stes

irri

tate

my

mou

th.

C1:

Doe

sn’t

irri

tate

my

mou

th.

21

26F2

:Too

thpa

stes

scra

tch

the

enam

elon

my

teet

h.C

3:Sa

fefo

rto

oth

enam

el(n

on-s

crat

chin

g).

8

27A

9:I

have

trou

ble

getti

ngm

yki

dsto

brus

h.B

4:A

tast

eki

d’s

love

24

28A

4:I

ampr

edis

pose

dto

havi

ngse

nsiti

vete

eth.

C2:

For

sens

itive

teet

h.25

29F5

:Too

thpa

stes

cont

ain

artif

icia

ling

redi

ents

.G

2:10

0%

natu

rali

ngre

dien

ts.

4

30F4

:Too

thpa

stes

cost

too

muc

h.F2

:Pri

ce20

%le

ss11

31F6

:Too

thpa

ste

pack

agin

gca

nbe

harm

fult

oth

een

viro

nmen

t.H

2:10

0%

recy

clab

lepa

ckag

ing.

6


Fig. 4 Importance of a/b vs.average brand ratings

432

2

1

0

-1

Avg. Brand Rating

a/b

Coe

ffici

ent

K2

K1J2

J1I2

I1

H2G2

F2E3

E2

E1

D4

D3

D2

D1

C3

C2

C1B5B4

B3

B2B1

A6

A5

A4 A3

A2

A1

Contains 100% natural ingredients

Freshens breath for 12 hours

Uses 100% recyclable packaging

r = 0.52 for rest

larger ratios when the prediction is positive and smallerratios (closer to zero) when the prediction is negative.

The bottom portion of Table 7 reports likelihood ratiosbased on both the c/i’s and the a/b’s. When the weightedratings are positive for both the c/i’s and a/b’s, the likelihoodratios are approximately equal to four, indicating that theprior odds are increased by a factor of 4 to yield posteriorodds that a brand is preferred by a respondent. The fact thatthe combined likelihood ratios are greater than either of theindividual likelihood ratios indicates that the c/i’s and a/b’sreflect different aspects of demand, with the c/i’s capturingmotivating conditions that are upstream and independent ofcurrent offerings.

Table 7 Likelihood ratios

Attribute/benefit Positive (+) Negative (−)

Aquafresh 1.86 0.74

Colgate 1.75 0.79

Crest 2.04 0.78

Mentadent 2.46 0.65

Concerns/interests Positive (+) Negative (−)

Aquafresh 2.25 0.69

Colgate 2.53 0.70

Crest 3.00 0.69

Mentadent 2.17 0.62

Both positive One positive Both negative

Both (+,+) (+,− or −,+) (−,−)

Aquafresh 3.87 1.29 0.58

Colgate 3.73 1.48 0.59

Crest 7.82 1.55 0.58

Mentadent 4.21 1.58 0.40

6 Discussion

Wants originate upstream from the marketplace, in the con-text of everyday life. Individuals find value in marketplaceofferings that are responsive to the concerns and intereststhat lead them to engage in the activities of their lives. Pref-erences for product attributes therefore result from peoplesearching for correspondence between upstream conditionsthat lead them to action and the capability of marketplaceofferings to deliver utility within the activity. In this paper,we introduce a method of augmenting the standard analysisbased on reactions to marketplace offerings by identifyingthe concerns and interests that specify valued attributes.

Research on wants in marketing has focused on the imag-ined state of an individual, either in terms of the end stateitself (e.g., goals, benefits) or instruments helpful in gettingthere (e.g., the part-worths of product attribute-levels). Inthis paper, we report on the research that includes measuresboth of the present and the desired state. The present stateis described in terms of concrete concerns and interests rel-evant for brand purchase; the desired state is described interms of the usual product benefits/attributes. The additionalmeasure differs from product benefits/attributes in a numberof respects:

1. It measures motivating features of the context in whichprospects engage in the activity for which the prod-uct benefits/attributes should be relevant. In otherwords, it describes the conditions to which productattributes/benefits should be responsive, if prospects areto value them.

2. In place of the purely empirical approach to generat-ing product attributes and benefits associated with theimagined state, our measure is derived from a set ofseven qualitatively distinct classes of motivating con-ditions. These classes cover a range of qualitatively


diverse condition, whereas no such platform has beenavailable from which to judge the comprehensiveness orpossible redundancy of product attributes and benefits.

3. A generally accepted motivational formulation speaksof motivation as arising when, comparing a current statewith an imagined/desired state, an individual allocatesresources to trying to bring about the desired state. Inlight of such a formulation, wants up to now have beenstudied in ways that are relevant to the second of thesetwo states, i.e., the imagined/desired state. Conceptu-ally, our new measure operationalizes the motivatingfeatures of the current state, i.e., present features whoseabsence, or absent features whose presence, defines thedesired state.

The concerns and interests that lead individuals to thepursuits in their lives exist whether or not managementsrespond with appropriate goods and services. Concern aboutbad breath or dull teeth may not be satisfied within thecurrent array of toothpaste offerings, leaving the individ-ual wanting or deprived. The presence of a motivating wantwithout a corresponding marketplace offering (e.g., socialexpectations about toothbrushing and the absence of a tooth-paste that shows others you care about your teeth) can beregarded as unmet demand. A limitation of using market-place preferences as a guide to prospect wants is that thereis no guarantee that the analysis fully reflects the range ofmotivating wants among potential users.

If there is no guarantee that an analysis of marketplacepreference fully reflects the range of motivating conditions,then the analysis of marketplace offerings must consider itslocation in the demand space. However, the offerings usedin the analysis of preferences often arise from the array ofcurrently available goods and services, whose existence isnonsystematic. In a conjoint study, for example, hypotheti-cal offerings are constructed with feature combinations thatare typically within, or close to, the convex hull of existingofferings. The evolution of actual offerings builds incre-mentally on past actions and is dependent on current andfeasible technologies. Analysis based on product offeringsmay therefore be located in a portion of the demand spacethat is suboptimal for marketing’s role in guiding productpolicy.

The analysis of preference also leads to unclear directionfor product policy because of motivational ambiguity [12].Consider, for example, the reasons that consumers prefertoothpaste that “gives your mouth a tingle.” Preference fora sharp taste may be due to sensory enjoyment of the tingleor because consumers construe this attribute as responsiveto concerns about letting oneself down, the expectations ofothers, or the possibility of tartar and plaque buildup. Sim-ply knowing which features are preferred does not provideaccess to the nature of conditions that lead people to act and

find value in the offering. Such information is often criti-cal so that management can optimally generate and chooseamong options for brand formulation and communicationstrategy (see [3]).

Such limitations of using preferences to guide productpolicy are present because wants are conceptualized andmeasured in terms of an array of real or hypothetical offer-ings. We offer a conceptualization of wants in terms ofmotivating conditions that are independent of marketplaceofferings and develop a conjoint-based approach to measur-ing their importance. These motivating wants are the sourceof brand preference and can provide substantive guidancefor product formulation and communication efforts. Ouranalysis illustrates that attribute-level part-worths can becorrelated with brand belief ratings (Fig. 4), that there existsmotivational ambiguity in terms of the complex mappingamong c/i and a/b items, and that the c/i items can be usedto improve brand preference.

Both within marketing and economics, authors havesought improved methods of understanding the consumerand, specifically, of investigating what users want fromgoods and services. Research and analysis, however, ismainly focused on instrumental wants and marketplaceofferings. While recent work on extending the scope ofthese models has included a range of psychological vari-ables (e.g., [9]), it has not offered an explicit operational-ization of motivation. In this paper, an explicit opera-tionalization of motivation is offered, and its systematicrole in a model of action and brand use is demonstrated.Accordingly, in addition to the economist’s ex post viewof demand (e.g., instrumental wants expressed as reactionsto good/services), marketers can use our approach to studyan ex ante view of demand in which motivating wants areexpressed as concerns and interests in the context for theeveryday activities, for which goods/services are offeredand used.

Acknowledgments The authors thank Derek Rucker for the helpfulfeedback on the paper.

References

1. Ajzen I (1991) The theory of planned behavior. Organ Behav HumDecis Process 50(2):179–211

2. Ajzen I, Sheikh S (2013) Action versus inaction: anticipated affectin the theory of planned behavior. J Appl Soc Psychol 43(1):155–162

3. Allenby G, Fennell G, Bemmaor A, Bhargava V, Christen F,Dawley J, Dickson P, Edwards Y, Garratt M, Ginter J et al (2002)Market segmentation research: beyond within and across groupdifferences. Mark Lett 13(3):233–243

4. Allenby GM, Ginter JL (1995) Using extremes to design productsand segment markets. J Mark Res 32:392–403


5. Allenby GM, Rossi PE (1998) Marketing models of consumerheterogeneity. J Econ 89(1):57–78

6. Arndt J (1976) Reflections on research in consumer behavior. AdvConsum Res 3:213–221

7. Bagozzi RP, Dholakia U (1999) Goal setting and goal striving inconsumer behavior. J Mark 63:19–32

8. Belk RW (1974) An exploratory assessment of situational effectsin buyer behavior. J Mark Res 11:156–163

9. Ben-Akiva M, McFadden D, Train K, Walker J, Bhat C, BierlaireM, Bolduc D, Boersch-Supan A, Brownstone D, Bunch DS et al(2002) Hybrid choice models: progress and challenges. Mark Lett13(3):163–175

10. Chandukala SR, Edwards YD, Allenby GM (2011) Identifyingunmet demand. Mark Sci 30(1):61–73

11. Dickson PR (1982) Person-situation: segmentation’s missing link.J Mark 46:56–64

12. Fennell G (1978) Consumers’ perceptions of the product usesituation. J Mark 42:38–47

13. Fennell G (1987) A radical agenda for marketing science: rep-resent the marketing concept. In: Dholakia N, Furat F, BaggozziR (eds) Philosophical and radical thought in marketing. D.C.Health, Lexington, pp 289–306

14. Fennell G (1997) Value and values: relevance to advertising. Val-ues, lifestyles, and psychographics. Earlbaum, Hillsdale, pp 83–110

15. Fennell G (2012) Marketing insights from a model of actionand empirical findings. In: Posavac S (ed) Cracking the code:leveraging consumer psychology to drive profits. M.E. Sharpe,Armonk

16. Fishbach A, Ferguson MJ (2007) The goal construct in socialpsychology. Soc Psychol Handb Basic Princ 2:490–515

17. Fishbach A, Henderson MD, Koo M (2011) Pursuing goals withothers: group identification and motivation resulting from thingsdone versus things left undone. J Exp Psychol Gen 140(3):520

18. Green PE, Srinivasan V (1990) Conjoint analysis in marketing:new developments with implications for research and practice. JMark 54(4):3–19

19. Haley RI (1968) Benefit segmentation: a decision-orientedresearch tool. J Mark 32:30–35

20. Huffman C, Ratneshwar S, Mick DG (2000) Consumer goalstructures and goal-determination processes: an integrativeframework. In: The why of consumption: contemporary perspec-tives on consumer motives, goals and desires. Routledge

21. Kamakura W, Russell G (1989) A probabilistic choice modelfor marketing segmentation and elasticity structure. J Mark Res23:89–100

22. Lenk P, DeSarbo W, Green PE, Young M (1996) Hierarchi-cal Bayes conjoint analysis: recovery of part-worth heterogeneityfrom reduced experimental designs. Mark Sci 15:173–191

23. Lewin K (1936) Principles of topological psychology. McGraw-Hill, New York

24. Marshall P, Bradlow ET (2002) A unified approach to conjointanalysis models. J Am Stat Assoc 97(459):674–682

25. McClelland DC, Atkinson JW, Clark RA, Lowell EL (1976) Theachievement motive. Irvington

26. Reynolds TJ, Gutman J (1988) Laddering theory, method, analy-sis, and interpretation. J Advert Res 28:11–31

27. Rossi PE, Allenby GM, McCulloch R (2005) Bayesian statisticsand marketing. Wiley, New York

28. Saegert J, Hoover RJ, Landeck M (1993) Using focus groupsto explore domains of action. In: Proceedings of the societyfor consumer psychology. American Psychological Association,Washington, pp 36–42

29. Wells WD (1993) Discovery-oriented consumer research. J Con-sum Res 19:489–504

30. Yang S, Allenby GM, Fennell G (2002) Modeling variation inbrand preference: the roles of objective environment and motivat-ing conditions. Mark Sci 21(1):14–31

conceptualizing and measuring prospect wants: understanding the source of brand preference

Documents