The Impact of Different Touchpoints on Brand Consideration

Abstract

Marketers face the challenge of resource allocation across a range of touchpoints. Hence understanding their relative impact is important, but previous research tends to examine brand advertising, retailer touchpoints, word-of-mouth, and traditional earned touchpoints separately. This article presents an approach to understanding the relative impact of multiple touchpoints. It exemplifies this approach with six touchpoint types: brand advertising, retailer advertising, in-store communications, word-of-mouth, peer observation (seeing other customers), and traditional earned media such as editorial. Using the real-time experience tracking (RET) method by which respondents report on touchpoints by contemporaneous text message, the impact of touchpoints on change in brand consideration is studied in four consumer categories: electrical goods, technology products, mobile handsets, and soft drinks. Both touchpoint frequency and touchpoint positivity, the valence of the customer's affective response to the touchpoint, are modeled. While relative touchpoint effects vary somewhat by category, a pooled model suggests the positivity of in-store communication is in general more influential than that of other touchpoints including brand advertising. An almost entirely neglected touchpoint, peer observation, is consistently significant. Overall, findings evidence the relative impact of retailers, social effects and third party endorsement in addition to brand advertising. Touchpoint positivity adds explanatory power to the prediction of change in consideration as compared with touchpoint frequency alone. This suggests the importance of methods that track touchpoint perceptual response as well as frequency, to complement current analytic approaches such as media mix modeling based on media spend or exposure alone.

Full text

Journal
of
Retailing
91
(2,
2015)
235–253
The
Impact
of
Different
Touchpoints
on
Brand
Consideration
Shane
Baxendale a,
Emma
K.
Macdonald a,b,∗,
Hugh
N.
Wilson a
aCranfield
School
of
Management,
Cranfield
University,
United
Kingdom
bSchool
of
Marketing,
University
of
South
Australia,
Australia
Abstract
Marketers
face
the
challenge
of
resource
allocation
across
a
range
of
touchpoints.
Hence
understanding
their
relative
impact
is
important,
but
previous
research
tends
to
examine
brand
advertising,
retailer
touchpoints,
word-of-mouth,
and
traditional
earned
touchpoints
separately.
This
article
presents
an
approach
to
understanding
the
relative
impact
of
multiple
touchpoints.
It
exemplifies
this
approach
with
six
touchpoint
types:
brand
advertising,
retailer
advertising,
in-store
communications,
word-of-mouth,
peer
observation
(seeing
other
customers),
and
traditional
earned
media
such
as
editorial.
Using
the
real-time
experience
tracking
(RET)
method
by
which
respondents
report
on
touchpoints
by
contemporaneous
text
message,
the
impact
of
touchpoints
on
change
in
brand
consideration
is
studied
in
four
consumer
categories:
electrical
goods,
technology
products,
mobile
handsets,
and
soft
drinks.
Both
touchpoint
frequency
and
touchpoint
positivity,
the
valence
of
the
customer’s
affective
response
to
the
touchpoint,
are
modeled.
While
relative
touchpoint
effects
vary
somewhat
by
category,
a
pooled
model
suggests
the
positivity
of
in-store
communication
is
in
general
more
influential
than
that
of
other
touchpoints
including
brand
advertising.
An
almost
entirely
neglected
touchpoint,
peer
observation,
is
consistently
significant.
Overall,
findings
evidence
the
relative
impact
of
retailers,
social
effects
and
third
party
endorsement
in
addition
to
brand
advertising.
Touchpoint
positivity
adds
explanatory
power
to
the
prediction
of
change
in
consideration
as
compared
with
touchpoint
frequency
alone.
This
suggests
the
importance
of
methods
that
track
touchpoint
perceptual
response
as
well
as
frequency,
to
complement
current
analytic
approaches
such
as
media
mix
modeling
based
on
media
spend
or
exposure
alone.
©
2015
New
York
University.
Published
by
Elsevier
Inc.
All
rights
reserved.
Keywords:
Retailing;
Advertising;
Integrated
marketing
communications;
In-store
communications;
Word-of-mouth
(WOM)
Introduction
There
is
a
stream
of
research
comparing
the
impact
of
vari-
ous
paid-for
media,
which
is
helpful
to
marketers
in
determining
their
overall
media
spend
and
its
allocation
across
media
(Naik
and
Peters
2009).
Brand
owners
have
a
bigger
challenge,
how-
ever:
how
to
allocate
budgets
and
management
time
across
the
wider
range
of
touchpoints
that
occur
in
the
customer
deci-
sion
journey
(Court
et
al.
2009).
These
broader
touchpoints
go
beyond
the
brand
advertising
which
is
generally
referred
to
as
∗Corresponding
author
at:
Cranfield
School
of
Management,
Cranfield
Uni-
versity,
Cranfield,
Bedofrdshire
MK43-0AL,
United
Kingdom.
Tel.:
+44
07788
543
905.
E-mail
addresses:
S.Baxendale@Cranfield.ac.uk
(S.
Baxendale),
Emma.Macdonald@Cranfield.ac.uk
(E.K.
Macdonald),
Hugh.Wilson@Cranfield.ac.uk (H.N.
Wilson).
paid
media
(or
owned
media
where
the
firm
does
not
have
to
pay
directly),
to
include
for
example
traditional
earned
media
such
as
editorial
coverage.
Peer-to-peer
encounters
with
the
brand
such
as
word-of-mouth
(WOM)
conversation
can
also
be
regarded
as
earned
touchpoints
(Stephen
and
Galak
2012).
In
the
case
of
consumer
goods
sold
through
retailers,
the
focus
of
this
article,
the
retailer
may
also
pay
for
advertising
that
mentions
the
brand.
Furthermore,
the
store
itself
is
far
more
than
a
fulfillment
channel
to
convert
pre-existing
intentions
to
purchases.
In-store
commu-
nications
can
also
bring
new
brands
into
active
consideration
(Court
et
al.
2009;
Goodman
et
al.
2013)
and
influence
imme-
diate
or
subsequent
purchase
irrespective
of
channel
(Verhoef,
Neslin,
and
Vroomen
2007).
Of
these
touchpoints,
the
brand
owner
only
directly
controls
brand
advertising,
but
all
are
poten-
tially
within
the
brand
owner’s
influence.
The
resulting
resource
allocation
challenge
in
turn
leads
to
a
measurement
challenge:
assessing
the
relative
importance
of
these
diverse
touchpoints
in
evolving
the
customer’s
brand
attitudes
and
hence
behaviors.
http://dx.doi.org/10.1016/j.jretai.2014.12.008
0022-4359/©
2015
New
York
University.
Published
by
Elsevier
Inc.
All
rights
reserved.

236
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Despite
widespread
agreement
that
the
customer
decision
journey
needs
to
be
understood
across
all
touchpoints
(Wiesel,
Pauwels,
and
Arts
2010),
most
research
focuses
on
parts
of
this
journey
in
isolation,
such
as
brand
advertising,
in-store
com-
munications,
or
WOM.
Such
focused
studies
are
undoubtedly
necessary,
providing
granular
insight
into
these
parts
of
the
jour-
ney.
However,
managers
also
have
an
interest
in
understanding
comparative
effects
of
diverse
touchpoints
in
an
equivalent
man-
ner
in
order
to
inform
the
complete
marketing
plan.
Multiple
touchpoints
in
the
consumer
search
process,
including
customer
interactions
with
‘sales’
channels,
can
be
viewed
symmetrically
until
final
choice
occurs,
as
the
search
process
may
iterate
indef-
initely
while
consumers
revise
brand/channel
utilities
(Neslin
et
al.
2014).
Such
a
holistic
view
of
touchpoints
is
particularly
important
as
media
fragmentation
sees
brand
managers
increas-
ingly
allocate
their
budgets
to
what
are
still
often
described
as
“unmeasured
media”
such
as
news
media
coverage
and
in-store
communications
(Ailawadi
et
al.
2009,
p.
50).
We
speculate
that
the
paucity
of
empirical
studies
across
mul-
tiple
touchpoints
is
in
large
part
due
to
data
availability.
In
Table
1
we
show
representative
examples
of
research
that
does
assess
the
impact
of
multiple
touchpoints.
While
rich
individual-level
data
are
available
for
retail
transactions
and
promotions
from
loyalty-
card
holders
and
consumer
panels
(Ngobo
2011),
these
data
sources
do
not
reach
other
parts
of
the
journey
such
as
WOM.
Aggregate-level
data
such
as
media
spend
can
be
used
to
model
the
relative
impact
of
some
market
mix
variables
on
consumer
response
(Naik
and
Peters
2009),
but
again
there
are
parts
of
the
journey
such
as
peer-to-peer
touchpoints
that
this
method
can-
not
reach.
In
the
online
context,
automatically
captured
data
can
allow
a
rich
picture
of
the
customer
journey
(Trusov,
Bucklin,
and
Pauwels
2009),
but
there
is
no
ready
equivalent
for
offline
brand
encounters.
Surveys
can
in
theory
ask
about
touchpoints
holistically,
but
respondents
find
it
difficult
to
remember
touch-
points
accurately
(Wind
and
Lerner
1979);
in
particular,
affective
response
decays
rapidly
and
is
recalled
poorly
(Aaker,
Drolet,
and
Griffen
2008).
Marketing
practitioners
tend,
therefore,
to
use
brand
tracking
surveys
only
for
a
few
frequent
and
memorable
touchpoints
such
as
television
advertisements.
In
this
article,
we
therefore
apply
the
emerging
real-time
experience
tracking
(RET)
method
to
understand
how
a
range
of
touchpoints
impacts
on
brand
consideration.
Adopted
by
a
number
of
companies
such
as
BSkyB,
Energizer,
Microsoft
and
Intercontinental
Hotels
(Macdonald,
Wilson,
and
Konus¸
2012),
the
RET
method
involves
asking
a
panel
of
consumers
to
send
a
structured
text
(SMS)
message
by
mobile
phone
whenever
they
encounter
one
of
a
set
of
competitive
brands
within
a
cat-
egory
for
a
period
of
a
week.
This
has
the
benefit
of
allowing
a
wide
range
of
touchpoints
to
be
reported,
including
those
such
as
offline
WOM
that
leave
no
behavioral
trace.
It
also
allows
touch-
point
positivity,
the
valence
of
the
customer’s
affective
response
to
the
touchpoint
(Kahn
and
Isen
1993),
to
be
captured.
By
pooling
multiple
RET
samples,
we
study
four
categories:
elec-
trical
goods,
technology
products,
mobile
phone
handsets,
and
soft
drinks.
These
categories
provide
a
spread
of
high
involve-
ment,
extended
decision
journeys
in
mobile
handsets,
and
in
technology
products
such
as
laptops,
cameras,
and
televisions;
somewhat
lower
involvement
journeys
in
electrical
goods,
such
as
blenders
and
dishwashers;
and
repertoire
brands
in
the
case
of
soft
drinks
Through
these
data,
we
hence
address
two
objectives.
First,
we
examine
the
impact
on
change
in
brand
consideration
of
six
broad
touchpoints:
brand
advertising;
retailer
advertising;
in-store
communications;
peer-to-peer
conversation;
traditional
earned
media;
and
peer
observation
(observing
other
customers).
Second,
we
examine
the
roles
of
both
touchpoint
frequency
and
touchpoint
positivity
in
forming
this
impact.
This
study
thereby
makes
three
contributions
to
multichan-
nel
and
brand
choice
literature.
First,
we
evidence
the
relative
role
of
multiple
touchpoints
in
evolving
brand
consideration.
All
six
touchpoints
are
significant
in
at
least
three
categories.
While
relative
touchpoint
effects
vary
somewhat
by
category,
a
pooled
model
suggests
the
positivity
of
in-store
communication
is
in
general
more
influential
than
that
of
other
touchpoints
includ-
ing
brand
advertising.
Furthermore,
an
almost
entirely
neglected
touchpoint,
peer
observation,
is
both
pervasive
and
persuasive.
Overall,
our
findings
evidence
the
relative
impact
of
retailers,
social
effects
and
third
party
endorsement
in
addition
to
brand
advertising.
Second,
we
highlight
the
roles
of
both
touchpoint
positivity
and
frequency
across
this
wide
range
of
touchpoints.
In
particular,
we
find
that
positivity
adds
to
the
explanatory
power
of
a
model
predicting
consideration
change
based
on
frequency
alone.
This
suggests
a
limitation
of
media
mix
modeling
based
on
media
spend
as
a
proxy
for
frequency.
Third,
we
propose
and
exemplify
a
RET-based
approach
by
which
both
the
positivity
and
the
frequency
of
multiple
touchpoints
can
be
assessed
in
further
categories
and
with
further
touchpoints.
In
the
following
sections,
we
develop
a
conceptual
frame-
work,
describe
the
data
collection
and
data
analysis
in
more
detail,
present
findings,
and
discuss
implications
for
practice
as
well
as
research
directions.
Conceptual
Framework
We
view
the
customer
search
process
as
consisting
of
a
num-
ber
of
discrete
encounters
with
varying
touchpoints,
such
as
advertisements,
WOM,
and
so
on.
See
Fig.
1.
Drawing
on
Court
et
al.
(2009),
we
define
a
touchpoint
as
an
episode
of
direct
or
indirect
contact
with
the
brand.
Thus
touchpoints
include
but
are
not
limited
to
channels
as
defined
by
Neslin
et
al.
(2006,
p.
96)
as:
“a
customer
contact
point,
or
a
medium
through
which
the
firm
and
the
customer
interact”.
We
suggest
an
expansion
of
this
definition
is
required,
as
the
emphasis
here
on
interaction
commonly
excludes
one-way
communications
such
as
televi-
sion
advertising,
while
the
emphasis
on
the
firm
may
exclude
brand
encounters
such
as
WOM
in
which
the
firm
is
not
directly
involved.
Our
choice
of
touchpoints
emphasizes
breadth
in
the
stake-
holder
who
the
customer
touches,
from
the
brand
owner
(brand
advertising)
and
the
retailer
(retailer
advertising
and
in-store
communications)
to
peers
(WOM
and
peer
observation)
and
independent
third
parties
such
as
editorial
and
expert
reviews
(traditional
earned
media).
In
the
interests
of
parsimony
we
combine
subtypes
within
each
of
these
touchpoints:
online
and

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
237
Table
1
Illustrative
studies
on
the
impact
of
multiple
touchpoint
types.
Context
Data
collection
Main
dependent
variable(s)
Touchpoints
Real-time
encounter
recording
Perceptual
response
Brand
advertising
Retailer
advertising
In-store
comms.
WOM
Peer
observation
Traditional
earned
media
Stephen
and
Galak
(2012)
Lending
Search,
media
scanning
Sales
*
*
*
Ngobo
(2011)
Grocery
Panel
data
Preference,
purchase
intention
*
*
Stammerjohan
et
al.
(2005)
Credit
cards
Experiment
Attitude
to
brand
*
*
*
Trusov,
Bucklin,
and
Pauwels
(2009)
Social
network
Transaction
data
Member
growth
*
*
*
van
der
Lans
et
al.
(2010)
Viral
marketing
Online
form
Participation
in
the
campaign
*
*
O’Cass
(2002)
Politics
Survey
Attitude
to
brand
*
*
*
Ataman,
van
Heerde,
and
Mela
(2010)
Multiple
Panel
data
Sales
*
*
This
paper
Multiple
consumer
goods
Real-time
experience
tracking
Consideration
*
*
*
*
*
*
*
*

238
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Change in brand
con
sid
eraon
(T1 –
T0)
Brand con
sid
eraon (T0)
Retailer
adversin
g
In-store commun
icaon
s
Retail touchpoints
Brand adversin
g
Brand owner touchpoints
•
Tou
chpoi
nt frequency
•
Tou
chpoi
nt posivity
Word-of
-mouth
received
Pee
r obse
rvaon
Third party touchpoi
nts
Tradional
earned media
Brand tou
chpoints •
Focal brand
•
Com
petor
brand
s
•Touchpoint frequency
•
Tou
chpoi
nt posi
vity
•
Tou
chpoi
nt frequency
•
Tou
chpoi
nt posi
vity
Fig.
1.
Conceptual
framework.
offline
WOM,
for
example.
We
model
the
impact
of
these
touchpoints
on
change
in
consideration,
taking
account
of
prior
consideration.
Touchpoint
Frequency
and
Positivity
Unlike
many
time-series
media
mix
studies
(Thomas
and
Sullivan
2005),
our
study
allows
for
customer
heterogeneity
in
touchpoint
frequency.
Frequency
may
impact
brand
atti-
tudes
by
increasing
brand
awareness
(Yaveroglu
and
Donthu
2008).
Repetition
can
also
improve
learning
(Goh,
Hui,
and
Png
2011).
In
addition,
we
consider
perceptual
response
to
touch-
points.
Despite
experimental
findings
that
perceptual
response
to
advertisements
impacts
attitudes
(Bri˜
nol,
Petty,
and
Tormala
2004),
many
models
of
field
data,
particularly
in
the
case
of
paid
media,
focus
purely
on
frequency
or
media
spend,
presumably
because
perceptual
response
data
are
frequently
unavailable.
This
makes
it
difficult
to
untangle
the
effect
of
the
medium
from
that
of
the
message.
Inspired
by
WOM
research,
we
model
perceptual
response
with
touchpoint
posi-
tivity,
which
we
define
as
the
valence
of
the
customer’s
affective
response
to
the
encounter
(Kahn
and
Isen
1993).
Affective
response
has
been
shown
to
impact
on
spending
and
repeat
purchase
intentions
(Arnold
and
Reynolds
2009;
Liu
2006).
While
affective
response
can
be
viewed
multidimensionally
(Chitturi,
Raghunathan,
and
Mahajan,
2008),
qualitatively
dif-
ferent
emotions
can
be
related
to
the
unidimensional
construct
of
affective
valence
or
positivity
(Kahneman
and
Krueger
2006;
Westbrook
and
Oliver
1991).
Positivity
is
associated
with
out-
comes
including
satisfaction
(Westbrook
and
Oliver
1991),
commitment
(Ahluwalia,
Burnkrant,
and
Unnava
2000),
vari-
ety
seeking
(Kahn
and
Isen
1993),
and
consideration
(Desai
and
Raju
2007).
We
adopt
positivity
here
in
the
interests
of
model
parsimony.
Post-touchpoint
affect
forms
part
of
the
customer’s
evaluative
response
as
affective
markers
remain
in
episodic
memory
thereafter
(Westbrook
and
Oliver
1991),
influencing
future
brand-related
cognitions
(Baumeister
et
al.
2007).
After
a
period
of
time,
however,
affective
response
may
be
not
just
imperfectly
recalled
but
also
reconstructed
for
reasons
such
as
self-justification
(Cowley
2008).
This
suggests
that
touchpoint
positivity
should
be
assessed
immediately
after
the
encounter,
rather
than
retrospectively
in
surveys.
Brand
Consideration
We
focus
for
parsimony
on
one
brand
attitude
construct:
brand
consideration.
Following
Roberts
and
Lattin
(1997),
we
define
consideration
as
the
extent
to
which
the
customer
would
consider
buying
the
brand
in
the
near
future.
It
is
hence
closely
related
to
purchase
intention,
but
allows
for
the
observation
that
customers
evoke
a
set
of
brands,
which
may
evolve
over
time,
between
which
they
then
choose
based
on
a
comparison
of
utility
(Neslin
et
al.
2014).
Priester
et
al.
(2004)
provide
experimental
support
for
the
mediating
role
of
consideration
between
evolv-
ing
attitudes
to
the
brand
on
the
one
hand
and
purchase
on
the
other.
Brand
consideration
is
hence
useful
as
an
interme-
diate
outcome
variable
when
purchase
data
are
not
available.
Court
et
al.
(2009),
in
particular,
conceive
of
the
consumer
deci-
sion
journey
as
an
interplay
between
multiple
touchpoints
and
the
consumer’s
evolving
brand
consideration.
We
add
to
this
work
more
granularity
of
method
description,
real-time
data
collection,
the
distinction
between
touchpoint
frequency
and
positivity,
and
further
touchpoints
such
as
peer
observation.
Another
reason
for
adopting
consideration
is
that
it
is
in
com-
mon
use
among
practitioners
for
evaluating
consumer
response,
as
it
is
readily
studied
through
brand
tracker
surveys.

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
239
Touchpoints
See
Fig.
1
for
the
touchpoints
captured
in
this
study.
First,
we
examine
separately
advertisements
by
the
brand
owner
and
the
retailer.
Media
spend
models
do
not
necessarily
pick
up
the
latter
(Naik
and
Peters
2009).
Next,
we
examine
in-store
communi-
cations,
including
touchpoint
subtypes
such
as
viewing
in-store
posters
and
seeing
prominent
display
of
the
product
on
the
shelf
(Ailawadi
et
al.
2009).
In
a
bar
or
restaurant,
subtypes
include
posters,
beer
mats,
and
seeing
display
of
the
product
behind
the
bar.
The
first
of
two
peer-to-peer
touchpoints
is
peer
observation.
The
impact
of
other
customers
in
the
retail
or
consumption
envi-
ronment
has
been
explored
relatively
sparsely
as
compared
with
customer-firm
interactions
(Verhoef
et
al.
2009).
Nonetheless,
both
qualitative
(Borghini
et
al.
2009)
and
a
few
quantitative
(Sweeney
and
Soutar
2001)
studies
suggest
that
other
customers
can
impact
brand
attitudes
through
observation
alone
without
the
explicit
recommendation
or
criticism
of
WOM.
Observing
peers
may
impact
service
satisfaction
(Grove
and
Fisk
1997);
the
sim-
ilarity
of
others
may
increase
purchase
intentions
(Thakor,
Suri,
and
Saleh
2008);
and
consumers
who
purchase
products
with
the
support
of
others
may
form
more
enduring
brand
relationships
(McAlexander,
Schouten,
and
Koenig
2002).
The
influence
of
others
is
higher
in
environments
where
consumption
is
public
(Bearden
and
Etzel
1982);
this
is
the
case
to
differing
extents
in
our
four
categories.
The
second
peer-to-peer
touchpoint
is
WOM,
defined
as
any
conversation
(whether
online
or
offline)
with
other
individuals
in
which
the
brand
is
mentioned.
The
impact
of
WOM
has
often
been
examined
in
isolation
from
other
touchpoints
(East,
Hammond,
and
Lomax
2008).
Excep-
tions
largely
concern
WOM
in
social
media
which
has
been
the
focus
of
much
recent
attention
(Archak,
Ghose,
and
Ipeirotis
2011;
Liu
2006).
Finally,
earned
media
such
as
editorial
and
news
coverage
has
been
recently
rebranded
as
traditional
earned
media
to
dis-
tinguish
it
from
social
media
(Stephen
and
Galak
2012).
Such
earned
communications
have
been
the
subject
of
some
dedicated
time
series
studies
(Goh,
Hui,
and
Png
2011),
though
as
Stephen
and
Galak
(2012,
p.
626)
document
in
their
extensive
literature
review
on
earned
media,
“often
only
one
source
of
publicity
is
examined,
precluding
comparisons
between
different
types
of
channels”.
Overall,
these
authors
observe,
“The
effects
of
paid
media
on
sales
have
been
extensively
covered
in
the
marketing
literature.
The
effects
of
earned
media,
however,
have
received
limited
attention”.
Method
Data
Collection
Approach
and
Sample
See
Fig.
2
for
our
operationalization
of
the
RET
method.
Data
were
collected
by
MESH,
a
market
research
firm
which
pioneered
the
method,
on
behalf
of
multiple
sponsoring
brand
owners
over
the
four
categories.
Data
were
collected
in
Northern
America
and
Europe.
First,
an
online
survey
was
used
to
collect
demographics
and
brand
consideration
for
a
set
of
competitive
brands
at
time
T0;
consideration
was
collected
again
at
the
end
of
the
week
(time
T1).
Second,
participants
were
asked
to
send
a
text
message
whenever
they
encountered
one
of
the
brands
during
the
seven
days
of
the
study.
Each
participant
was
sent
an
initial
text
message
which
documented
the
code
frame
in
Fig.
2
so
they
always
had
the
required
information
to
hand.
This
enabled
the
capture
of
touchpoints
as
they
occurred
as
well
as
participants’
real-time
affective
response
in
a
positivity
measure.
Within
each
category,
a
sample
of
consumers
looking
to
pur-
chase
within
the
next
three
to
twelve
months
(depending
on
the
category)
was
recruited
via
an
online
panel
(Table
2).
In
the
case
of
soft
drinks
participants
were
regular
drinkers
of
carbonated
drinks.
The
data
were
collected
over
a
period
of
several
months
(dependent
on
the
category)
through
weekly
samples
in
each
category,
with
a
new
set
of
participants
recruited
each
week.
This
approach
was
adopted
in
order
to
expand
the
sample
and
to
allow
sponsoring
firms
to
track
trends
over
time.
Each
SMS
message
recorded
the
brand,
the
touchpoint,
and
the
participant’s
real-time
assessment
of
touchpoint
positivity.
Participants
were
briefed
with
a
coding
scheme
for
the
mes-
sage,
with
a
letter
for
each
brand,
a
letter
for
each
touchpoint,
and
a
Likert-scale
number
for
positivity;
so,
for
example,
“BA5”
might
represent
a
brand
named
“Quench”
(name
amended
for
confidentiality);
a
TV
advertisement;
and
a
positivity
rating
of
5
(very
positive)
on
a
5-point
scale
(measures
are
described
below).
The
conciseness
of
the
message
had
the
aim
of
mini-
mizing
the
disruption
to
the
participant’s
life.
While
touchpoints
were
collected
in
detail
(such
as
television,
radio,
billboards
and
so
on),
they
were
aggregated
into
the
broad
touchpoints
(such
as
brand
advertising)
shown
in
Fig.
1,
for
analysis
pur-
poses.
To
enhance
validity
in
this
coding,
participants
were
asked
to
visit
an
on-line
diary
at
their
convenience
(typically
in
the
evening)
every
two
days,
where
the
texts
they
had
sent
were
displayed.
In
the
diary,
they
were
asked
to
provide
fur-
ther
details
about
each
touchpoint
through
a
pull-down
menu
containing
touchpoint
sub-types.
This
allowed
checking,
for
example,
whether
a
magazine
touchpoint
was
an
advertisement
from
the
brand,
an
advertisement
from
a
retailer,
editorial
mate-
rial,
and
so
on.
We
excluded
from
analysis
any
participants
where
pre-
consideration
or
post-consideration
was
missing.
We
also
excluded
those
who
did
not
report
any
brand
encounters
at
all,
as
these
participants
either
did
not
engage
with
the
process
and
hence
constitute
missing
data,
or
genuinely
had
no
encoun-
ters
which
is
of
limited
interest
to
our
research
objectives.
We
also
cleaned
the
data
to
ensure
validity
of
entries;
if
any
touch-
point
was
recorded
with
invalid
codes
then
the
participant
was
removed.
We
used
listwise
deletion
as
imputation
methods
can
lead
to
bias
in
coefficients
and
as
the
sample
size
was
regarded
as
sufficient
to
allow
a
slight
loss
of
power.
265
(6.0%)
electri-
cal
goods
participants
were
excluded
from
the
final
dataset,
260
(4.4%)
technology
products
participants,
204
(10.7%)
mobile
handset
participants,
and
62
(2.5%
of
sample)
soft
drinks
par-
ticipants.
Table
2
shows
the
base
sizes
after
excluding
these
participants,
ranging
from
1709
for
mobile
phones
to
5632
for
technology
products.

240
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Online survey
(Time T0):
Brand
con
sid
eraon
Demographics
Real me experience tracking (RET)
via mobile handset (over 1 week)
Example: Respondent texts
“BA5
”
Online survey
(Time
T0
plus 1 week)
:
Brand
con
sid
eraon
Brand
A = Brand A
B = Brand B
C = Brand C
D = Brand D
E = Brand F
Touchpoi
nt
A = TV
B = Billboard
C = Radio
D = Me
drin
kin
g/
buyin
g
E = Conversaon
F = Cinema
G = Ne
wsp
aper
H = Magazin
e
I = Sponsorship
J = In
store
K = At an
event
L = Someone
els
e
drin
kin
g
M = Onlin
e
N = Leaflet
O = In
bar/
rest
aurant
P = Other (please
specify)
Posi
vity
How di
d it
make yo
u feel?
1 = very
negave
5 = very
posive
Fig.
2.
Method.
Measures
Brand
consideration
was
measured
using
a
6-point
scale,
anchored
by:
‘This
is
the
only
brand
that
I
would
consider
pur-
chasing’
and
‘I
would
definitely
not
consider
purchasing
it’.
This
is
similar
to
Bian
and
Moutinho
(2009).
Positivity
was
mea-
sured
with
a
single
Likert-scale
item
“How
did
it
make
you
feel
about
the
brand?”
on
a
5-point
scale
anchored
by
‘very
positive’
and
‘very
negative’,
similarly
to
McFarland
and
Buehler
(1998)
amongst
others.
Touchpoint
frequency
was
calculated
by
count-
ing
the
touchpoints
of
a
given
type:
so,
if
a
respondent
sees
two
advertisements
for
a
brand
over
the
week,
the
touchpoint
fre-
quency
is
2.
Positivity
was
re-centered
around
0,
such
that
0
represented
neutral
encounters,
+2
very
positive
and
−2
very
negative
encounters.
This
was
then
averaged
for
each
respon-
dent
and
touchpoint
type:
so,
if
the
respondent
rates
one
brand
advertisement
as
4
and
another
as
5,
the
average
positivity
(after
re-centering)
is
1.5.
If
the
participant
did
not
report
a
touch-
point
(i.e.,
frequency
is
zero),
average
positivity
was
coded
as
zero.
Hence
in
a
regression
the
impact
of
neutral
touchpoints
(or
if
the
average
positivity
is
zero)
is
equivalent
to
the
impact
of
frequency.
Hence
the
impact
of
positivity
can
be
interpreted
as
the
impact
above
the
neutral
baseline
of
frequency,
aiding
interpretation.
We
return
later
to
some
robustness
checks
on
this
approach
to
modeling
frequency,
positivity,
and
our
decision
to
code
positivity
as
zero
where
a
touchpoint
did
not
occur
during
the
week.
Models
We
combine
the
data
from
our
four
categories
in
a
pooled
model
to
further
increase
the
sample
size
and
deliver
generalized
results.
We
weight
the
data
such
that
each
brand
is
represented
in
the
dataset
equally
to
prevent
any
bias
toward
those
categories
Table
2
Sample
definition.
Electrical
goods
Technology
products
Mobile
handsets
Soft
drinks
Sample
definitiona
Age
18–64
18–64
18–64
16–44
Number
of
encounters
Brand
advertising
4446
4227
3033
4198
Retailer
advertising
7254
7003
1096
736
In-store
communications
4202
7002
1890
5402
WOM
1132
1706
1403
659
Peer
observation
2201
2689
2550
2693
Traditional
earned
media
795
2462
299
104
Respondents
4176
5632
1709
2445
aEither
a
current
user
or
purchasing
within
the
next
few
months,
depending
on
the
study.

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
241
with
a
greater
sample
size.
We
model
the
change
in
consideration
(T1
consideration
minus
T0
consideration)
at
the
customer
level
for
each
brand
by
using
prior
(T0)
brand
consideration,
demo-
graphics,
brand
dummies,
and
time
of
year
as
control
variables.
We
then
explain
additional
variability
through
incorporating
touchpoint
frequency
and
positivity
variables
for
both
the
focal
brand
and
competitor
brands.
As
we
observe
multiple
responses
per
customer
(one
response
for
each
brand
in
their
study),
there
is
likely
to
be
unmodeled
heterogeneity
across
each
set
of
customer
responses
caused,
for
example,
by
unobserved
covariates
at
the
customer
level.
To
account
for
this,
we
include
a
respondent-
level
random
intercept
via
a
linear
mixed-effects
model.
The
correlation
matrix
in
Table
3
indicates
no
severe
multi-
collinearity
problems;
however,
we
do
notice
high
correlations
between
the
frequency
and
positivity
of
each
touchpoint,
which
we
return
to
in
an
exploratory
analysis
below.
As
a
further
check
we
calculated
the
variance
inflation
factors
(VIF)
for
the
explanatory
variables
in
each
model.
All
VIF
values
(sum-
marized
in
Table
4)
fall
below
the
recommended
cut
off
of
5
(O’Brien,
2007),
suggesting
multi-collinearity
is
not
of
concern.
Our
model
formulation
is
as
follows:
Consid
Posti,k −
ConsidPrei,k =
α
+
bi+
βpreConsidPrei, k
+
βbrand Brandk+
βtime
1Quarteri+
βtime
2Yeari+
βdem
1Agei
+βdem
2Sexi+
J

j=1
{βfreq
jln(Freqi,k,j +
1)
+
βpos
jAvgPosi,k,j
+γfreq
jln(Freqi,−k,j +
1)
+
γpos
jAbgPosi,−k,j }
+
∈i,k
where
ConsidPosti,kand
ConsidPrei,kare
the
consideration
scores
of
individual
i
for
brand
k
after
and
before
the
week
of
texting,
respectively,
Brandkis
a
dummy
variable
accounting
for
heterogeneity
across
brands,
Quarteriand
Yeariare
dummy
variables
identifying
when
individual
i
was
tracked,
Ageiand
Sexiare
variables
for
the
age
and
sex
of
individual
i.
Age
is
treated
as
a
continuous
variable
and
Sex
is
a
dummy
variable
taking
1
for
male
and
0
for
female,
Freq i,k,jand
AvgPosi,k,jare
the
frequency
and
average
positivity
of
encounters
individual
i
has
through
touchpoint
j
for
brand
k,
and
J
is
the
total
number
of
touchpoints
in
the
model,
Freq i,−k,jand
AvgPosi,−k,jare
the
frequency
and
average
positivity
of
encounters
individual
i
has
through
touchpoint
j
for
all
brands
other
than
k
(i.e.,
competitors
to
the
focal
brand).
We
build
this
model
sequentially
and
summarize
the
model
fit
for
each
in
Table
4:
Model
1:
Only
the
control
variables
are
included.
This
is
to
identify
how
respondent-level
data
can
measure
consideration
shifts
and
to
provide
a
baseline
for
future
models.1
1We
also
tested
alternatives
to
Models
1
to
3
in
which
the
dependent
variable
was
post-study
consideration
and
not
change
in
consideration.
Naturally,
the
pre-consideration
coefficient
was
substantial
and
positive
(β
ranging
from
0.52
in
electrical
goods
to
0.71
in
soft
drinks
in
Model
3)
as
pre-consideration
acts
as
an
initial
estimate
for
post-consideration.
However,
substantive
results
regarding
the
role
of
touchpoint
frequency
and
positivity,
including
which
variables
were
significant
and
coefficient
magnitudes,
were
very
similar
to
those
reported
here,
Model
2:
As
we
anticipate
that
changes
in
consideration
will
also
be
a
function
of
brand
touchpoints,
Model
2
builds
on
the
previous
model
by
adding
touchpoint
frequency
with
a
natural
logarithmic
decay.
Model
3:
We
then
add
touchpoint
positivity
to
distinguish
touchpoint
frequency
from
touchpoint
perceptual
response.
Model
4:
While
Model
3
only
looks
at
same-brand
effects,
such
as
brand
A’s
touchpoints
impacting
on
brand
A
consider-
ation,
in
Model
4
we
add
competitor
touchpoint
frequency
and
positivity;
we
would
expect
these
to
have
a
negative
effect
on
the
focal
brand’s
consideration.
Model
Selection
We
compared
and
selected
models
on
the
basis
of
their
AIC
(Akaike’s
Information
Criterion)
and
BIC
(Bayesian
Infor-
mation
Criterion),
with
BIC
preferring
simpler
models
(fewer
parameters)
than
AIC.
Improved
model
fit
is
evidenced
by
decreases
in
information
criterion
between
models;
however,
neither
AIC
or
BIC
gives
an
absolute
indication
of
fit
(Burnham
and
Anderson
2004).
We
also
therefore
use
the
marginal
and
conditional
r2values
for
mixed-effects
models
(Nakagawa
and
Schielzeth,
2013).
Marginal
r2demonstrates
the
amount
of
vari-
ability
explained
by
only
the
fixed
effects
in
our
models,
and
conditional
r2demonstrates
the
variability
explained
by
both
fixed
and
random
effects.
Further,
we
calculated
model
fit
statistics
for
each
category
in
isolation,
to
understand
which
model
best
fits
individual
cat-
egory
data.
If
we
were
to
take
the
full
data
for
each
category
then
we
might
expect
the
categories
with
a
higher
sample
size
to
prefer
more
complex
models
due
to
the
formula
used
to
cal-
culate
AIC
and
BIC.
To
avoid
this
bias
we
restricted
the
sample
to
1,500
respondents
per
category
when
calculating
fit
statistics.
Using
a
bootstrapping
technique,
we
took
a
random
sample
with
replacement
of
1,500
respondents
from
each
category
and
calcu-
lated
AIC,
BIC,
and
r2values
for
Models
1–4
using
that
sample.
We
performed
5,000
iterations
of
this
procedure
and
took
the
average
of
the
model
statistics.
See
Table
4.
In
the
case
of
the
pooled
data,
according
to
both
AIC
and
BIC
the
full
Model
4
is
preferred.
Fixed
effects
explain
19.6%
of
the
variability
in
a
respondent’s
change
in
consideration,
with
unob-
served
individual-level
covariates
(random
intercept)
accounting
for
a
further
12.2%.
By
contrast,
AIC
indicates
Model
3
is
preferred
for
the
individual
categories,
likely
due
to
the
lower
sample
size
when
compared
to
the
pooled
data.
BIC
also
favors
Model
3,
except
in
soft
drinks
where
Model
1
is
preferred.
This
could
be
due
to
the
higher
price-tag
and
extended
purchase
journey
for
electrical
goods,
technology
products,
and
mobile
handsets
when
compared
to
soft
drinks,
more
factors
hence
influ-
encing
consideration.
Given
that
r2continues
to
rise
until
Model
4
in
soft
drinks,
for
simplicity
we
will
only
consider
Model
3
when
reporting
individual
category
results.
suggesting
robustness
of
the
model
with
respect
to
this
choice
of
dependent
variable.
We
therefore
do
not
report
these
in
full.

242
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Table
3
Correlation
matrix:
pooled
data.
Mean
Standard
deviation
Conside-
ration
(pre)
Age
Sex
Frequency
Average
positivity
Traditional
earned
Brand
advertising
WOM
Peer
observation
In-store
communica-
tions
Retailer
advertising
Traditional
earned
Brand
advertising
WOM
Peer
observation
In-store
communica-
tions
Retailer
advertising
Consideration
(post–pre)
−0.02
1.12
−0.38** −0.02** −0.01*0.02** 0.06** 0.02** 0.03** 0.07** 0.03** 0.05** 0.03** 0.05** 0.06** 0.08** 0.06**
Consideration
(pre)
3.79 1.17 1.00
Age
37.11 11.41
0.04** 1.00
Sex
(male)
43%
0.01
−0.01*1.00
Frequency
Traditional
earned
0.04 0.23 0.04** 0.00 0.00 1.00
Brand
advertising
0.17
0.56
0.02** −0.03** 0.03** 0.05** 1.00
WOM 0.05
0.30
0.03** −0.04** 0.02** 0.06** 0.1** 1.00
Peer
observation
0.10
0.43
0.03** −0.05** 0.00
0.02** 0.05** 0.08** 1.00
In-store
com-
munications
0.19
0.57
0.05** 0.01*0.00
0.01** 0.03** 0.05** 0.08** 1.00
Retailer
advertising
0.16
0.56
0.07** 0.11** −0.03** 0.02** 0.03** 0.02** 0.00
0.04** 1.00
Average
positivity
Traditional
earned
0.02 0.20 0.06** 0.00
−0.01*0.54** 0.00
0.02** 0.01** 0.01** 0.01*1.00
Brand
advertising
0.10
0.39
0.07** −0.01*0.01** 0.01
0.55** 0.03** 0.02** 0.02** 0.02** 0.01** 1.00
WOM
0.03
0.25
0.06** −0.01** 0.01
0.02** 0.04** 0.44** 0.04** 0.04** 0.02** 0.03** 0.05** 1.00
Peer
observation
0.05
0.32
0.09** −0.02** −0.01** 0.03** 0.02** 0.04** 0.46** 0.04** 0.00
0.04** 0.02** 0.05** 1.00
In-store
com-
munications
0.12
0.44
0.12** 0.04** −0.01** 0.02** 0.01** 0.04** 0.02** 0.52** 0.02** 0.04** 0.03** 0.06** 0.06** 1.00
Retailer
advertising
0.08
0.34
0.09** 0.07** −0.03** 0.01** 0.02** 0.01** 0.00
0.03** 0.48** 0.02** 0.05** 0.03** 0.03** 0.06** 1.00
Significant
parameters:
** p
<
.01.
*p
<
.05.

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
243
Table
4
Model
statistics.
Model
AIC
BIC
r2marginal
r2conditional
Average
VIF
Maximum
VIF
Pooled
data
Model
0:
Null
248,999
249,027
0.0%
16.7%
NA
NA
Model
1:
Baseline 235,861 236,186 14.6% 26.6% 2.02
2.49
Model
2:
Frequency 233,931 234,330 16.7%
29.4%
1.91
2.60
Model
3:
Positivity
231,244
231,717
19.4%
32.0%
1.92
2.62
Model
4:
Competitor
effects
230,905a231,527a19.6%
31.8%
1.84
2.67
Electrical
goodsb
Model
0:
Null
26,029
26,051
0.0%
18.6%
NA
NA
Model
1:
Baseline
24,584
24,705
17.1%
28.5%
1.44
1.82
Model
2:
Frequency 24,337 24,513 19.8% 32.2% 1.37
1.86
Model
3:
Positivity
24,036a24,269a22.7%
35.1%
1.55
2.04
Model
4:
Competitor
effects
24,068
24,414
23.3%
34.9%
1.55
2.08
Technology
productsb
Model
0:
Null 19,241
19,262
0.0%
17.0%
NA
NA
Model
1:
Baseline
18,152
18,253
17.8%
25.9%
1.30
1.62
Model
2:
Frequency
17,962
18,117
20.6%
30.5%
1.30
1.81
Model
3:
Positivity
17,734a17,943a23.8%
34.3%
1.61
2.18
Model
4:
Competitor
effects
17,737
18,055
24.8%
34.1%
1.62
2.22
Mobile
handsetsb
Model
0:
Null 20,060
20,081
0.0%
15.4%
NA
NA
Model
1:
Baseline
18,840
18,947
18.8%
30.1%
1.53
1.80
Model
2:
Frequency
18,710
18,871
21.2%
33.3%
1.38
1.84
Model
3:
Positivity
18,499a18,715a24.1%
36.2%
1.60
2.19
Model
4:
Competitor
effects
18,549
18,872
24.5%
36.0%
1.64
2.82
Soft
drinksb
Model
0:
Null 16,869
16,890
0.0%
19.6%
NA
NA
Model
1:
Baseline
16,332
16,420a9.6%
25.4%
1.39
1.71
Model
2:
Frequency
16,300
16,442
10.7%
26.7%
1.40
1.90
Model
3:
Positivity
16,243a16,438
12.0%
27.6%
1.42
1.93
Model
4:
Competitor
effects
16,318
16,621
12.6%
27.7%
1.38
1.99
aPreferred
model.
b1500
bootstrap
sample.
Robustness
Checks
To
check
robustness
we
tested
a
number
of
competing
models
and
reformulations
of
frequency
and
positivity
variables,
and
also
checked
our
decision
to
code
the
positivity
of
non-occurring
touchpoints
as
0.
We
discuss
these
in
turn.
Frequency
The
models
above
assume
that
frequency
has
a
natural
log
relationship
with
change
in
consideration.
This
is
to
account
for
communication
wearout
through
over-exposure
which
results
in
diminishing
returns
(Bass
et
al.
2007).
To
check
this
transfor-
mation
of
frequency
we
try
four
competing
models,
each
with
a
different
formulation
of
frequency:
Model
Freq1:
With
dichotomous
variable
(where
at
least
one
instance
of
the
touchpoint
occurs):
βfreq
jI[Freqi,k,j>0] .
Model
Freq2:
With
a
linear
term:
βfreq
jFreqi,k,j.
Model
Freq3:
With
a
quadratic
decay
term:
βfreq
1,j Freqi,k,j +
βfreq
2,j Freq2
i,k,j.
Model
Freq4:
With
a
natural
log
decay
term:
βfreq
jln(Freqi,k,j +
1).
The
fit
statistics
in
Appendix
show
that
the
log
decay
term
(Model
Freq4)
provides
the
best
fit.
Positivity
We
investigated
different
ways
of
incorporating
positivity
by
devising
several
competing
models:
again,
see
Appendix.
The
inclusion
of
average
positivity
(Model
Pos1)
leads
to
a
poten-
tial
loss
of
information.
For
example,
it
treats
an
individual
who
has
a
very
negative,
a
neutral,
and
a
very
positive
(−2,
0,
2)
encounter
the
same
as
an
individual
who
has
three
neutral
(0,
0,
0)
encounters
because
both
average
to
0.
To
check
the
robust-
ness
of
this
approach,
we
introduced
a
term
for
the
variance
of
touchpoint
encounters
(Model
Pos2)
following
Archak,
Ghose,
and
Ipeirotis
(2011).
We
alternatively
separated
the
frequency
of
negative,
neutral,
and
positive
encounters
(Model
Pos3)
follow-
ing
Liu
(2006).
We
also
investigated
a
term
for
the
positivity
of
the
last
touchpoint
instead
of
(and
as
well
as)
average
positivity
(Models
Pos4
and
Pos5).
We
conclude
from
the
fit
statistics
that

244
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Table
5
Touchpoint
impacts
on
consideration
change
(pooled
data).
Model
1
Model
2
Model
3
Model
4
β
SE
β
SE
β
SE
β
SE
(Constant)
0.07** 0.02
−0.14** 0.02
−0.10** 0.02
0.01
0.02
Pre-considerationa−0.39** 0.00 −0.40** 0.00 −0.43** 0.00
−0.43** 0.00
Frequency
Traditional
earned
0.14** 0.03
−0.01
0.03
0.02
0.03
Brand
advertising
0.26** 0.01
0.08** 0.02
0.09** 0.02
WOM
0.18** 0.02
−0.03
0.02
−0.01
0.02
Peer
observation
0.24** 0.02
0.05** 0.02
0.07** 0.02
In-store
communications
0.29** 0.01
0.06** 0.02
0.10** 0.02
Retailer
advertising
0.19** 0.01
0.06** 0.02
0.08** 0.02
Positivitya
Traditional
earned
0.04** 0.00
0.04** 0.00
Brand
advertising
0.07** 0.00
0.07** 0.00
WOM
0.06** 0.00
0.06** 0.00
Peer
observation
0.08** 0.00
0.08** 0.00
In-store
communications
0.10** 0.00
0.10** 0.00
Retailer
advertising
0.06** 0.00
0.06** 0.00
Competitor
frequency
Traditional
earned
−0.02
0.02
Brand
advertising
−0.04** 0.01
WOM
−0.06** 0.01
Peer
observation
−0.05** 0.01
In-store
communications
−0.08** 0.01
Retailer
advertising
−0.04** 0.01
Competitor
positivitya
Traditional
earned −0.01*0.00
Brand
advertising
−0.01
0.00
WOM
0.00
0.00
Peer
observation
−0.01** 0.00
In-store
communications
−0.02** 0.00
Retailer
advertising
−0.01
0.00
Significant
parameters
(p
<
0.05)
are
bolded.
*p
<
.05.
** p
<
.01.
aStandardized
coefficients.
the
most
effective
way
to
include
positivity
is
indeed
to
use
a
simple
average.
Positivity
When
no
Touchpoint
Occurs
When
a
respondent
does
not
encounter
a
particular
touch-
point
with
a
brand
during
the
week,
its
frequency
is
zero.
In
the
main
Models
1–4
we
coded
positivity
as
zero
in
this
case;
how-
ever,
an
alternate
approach
would
be
mean
imputation.
We
tested
both
approaches
on
Model
4.
Both
AIC
and
BIC
indicate
that
zero-coding
gives
the
best
model
fit
(Appendix).
Further,
while
zero-coding
gives
VIFs
below
the
recommended
cut-off
of
5,
mean
imputation
gives
six
VIF
scores
above
this
cut-off
with
the
largest
being
18.2.
Hence
using
zero
coding
seems
the
most
appropriate
approach
to
reduce
multi-collinearity
and
improve
model
fit.
Findings
and
Discussion
Results
for
the
pooled
data,
using
Models
1–4,
are
shown
in
Table
5.
In
Table
6
we
show
Model
3
estimated
for
each
category.
We
report
standardized
coefficients
for
positivity
to
aid
comparison
of
relative
impact
across
touchpoints,
but
leave
dummy
and
frequency
(count)
variables
unstandardized
for
ease
of
interpretation.
We
begin
with
these
main
results,
focusing
primarily
on
Model
4
in
the
case
of
the
pooled
data,
before
turning
to
the
exploratory
analyses.
Initially,
we
briefly
discuss
non-touchpoint
terms.
First
we
note
that
prior
consideration
is
negatively
associated
with
shift
in
consideration
(p
<
0.01,
standardized
β
=
−0.43
for
the
pooled
model
and
ranging
from
−0.33
to
−0.46
for
category
models).
This
is
presumably
an
expected
regression
to
the
mean
effect,
as
the
higher
a
respondent’s
pre-consideration,
the
more
likely
it
is
that
any
shift
will
be
down
rather
than
up.
While
the
study
focus
is
primarily
brand
neutral,
some
addi-
tional
explanatory
power
is
obtained
through
consideration
of
individual
brands.
The
coefficients
of
these
dummy
variables
correlate
highly
with
prior
consideration
(r
=
0.84).
One
possi-
ble
explanation
is
that
higher
levels
of
consideration
represent
not
just
a
more
positive
attitude
but
also
higher
attitude
strength,
which
provides
resistance
against
change
to
attitude
(Priester
et
al.
2004).

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
245
Table
6
Touchpoint
impacts
on
consideration
change
by
category
(Model
3).
Electrical
goods
Technology
products
Mobile
handsets
Soft
drinks
β
SE
β
SE
β
SE
β
SE
(Constant)
−0.31** 0.02
−0.12** 0.02
0.02
0.04
0.06
0.04
Pre-considerationa−0.45** 0.01 −0.45** 0.01 −0.46** 0.01
−0.33** 0.01
Frequency
Traditional
earned
0.02
0.06
−0.02
0.04
−0.06
0.11
−0.02
0.14
Brand
advertising
0.14** 0.03
0.14** 0.03
0.05
0.04
0.02
0.03
WOM
−0.13*0.05
−0.04
0.05
−0.06
0.06
0.08
0.06
Peer
observation
−0.07
0.04
−0.03
0.05
0.04
0.04
0.10** 0.03
In-store
communications
0.04
0.03
−0.01
0.03
0.06
0.05
0.08** 0.02
Retailer
advertising
0.06** 0.02
0.06*0.03
0.10
0.07
−0.05
0.07
Positivitya
Traditional
earned
0.03** 0.01
0.06** 0.01
0.02*0.01
0.02*0.01
Brand
advertising
0.08** 0.01
0.06** 0.01
0.09** 0.01
0.08** 0.01
WOM
0.05** 0.01
0.05** 0.01
0.09** 0.01
0.04** 0.01
Peer
observation
0.09** 0.01
0.09** 0.01
0.10** 0.01
0.04** 0.01
In-store
communications
0.12** 0.01
0.15** 0.01
0.09** 0.01
0.06** 0.01
Retailer
advertising
0.08** 0.01
0.08** 0.01
0.02** 0.01
0.02
0.01
Significant
parameters
(p
<
0.05)
are
bolded.
*p
<
0.05.
** p
<
0.01.
aStandardized
coefficients.
With
regard
to
the
temporal
dummy
variables,
we
find
that
respondents
are
likely
to
report
a
higher
shift
in
consideration
during
Quarters
2–4
compared
to
Quarter
1.
We
conjecture
that
this
may
be
due
to
a
post-Christmas
dip,
with
fewer
people
able
to
make
discretionary
expenditure
and
hence
lower
brand
attention
levels.
We
also
see
that
years
2011
and
2012
lead
to
signifi-
cantly
higher
shift
than
2010
(β
=
0.08
and
0.10,
respectively),
which
could
coincide
with
an
increase
in
consumer
confidence
following
the
recession.
There
are
also
some
demographic
predictors,
which
are
not
our
focus
here.
Touchpoint
Frequency
and
Positivity
The
pooled
analysis
suggests
that
touchpoint
frequency
and
positivity
both
play
a
role
in
shaping
consideration.
While
we
cannot
compare
these
coefficients
directly
(as
the
scale
of
data
is
radically
different),
we
do
see
that
touchpoint
positivity
adds
substantial
explanatory
power
(Model
2
vs.
Model
3).
We
also
see
the
coefficients
for
touchpoint
frequency
change
substan-
tially
between
Model
2
and
Model
3.
It
appears
that
as
frequency
is
naturally
somewhat
correlated
with
positivity
(due,
for
exam-
ple,
to
the
liking
effect),
its
separate
effect
(due,
for
example,
to
awareness
increases)
is
over-estimated
if
positivity
is
not
also
considered.
This
supports
work
on
advertising
affect
that
sug-
gests
that
emotional
appeals
may
have
a
strong
effect
despite
low
recall
(Bülbül
and
Menon
2010).
It
suggests
the
need
to
supplement
existing
methods
of
measurement
that
rely
purely
on
touchpoint
frequency,
such
as
the
respondent-level
frequency
approach
(Havlena,
Cardarelli,
and
De
Montigny
2007)
and
media
spend
modeling
(Naik
and
Peters
2009).
These
meth-
ods
for
assessing
touchpoint
impact
struggle
to
tease
out
the
difference
between
an
encounter
that
does
not
work
because
of
the
touchpoint
choice
and
one
where
the
execution
is
flawed.
Our
findings
show
that
this
difference
matters.
A
practical
implication
is
that
measurement
techniques
focusing
purely
on
touchpoint
frequency,
even
putting
aside
the
well-known
valid-
ity
problems
associated
with
recall
(Wind
and
Lerner
1979),
will
not
provide
the
specificity
of
insight
provided
by
techniques
that
track
positivity.
Relative
Touchpoint
Impacts
We
next
consider
the
relative
impacts
of
different
touch-
points,
both
by
examining
which
terms
are
significant
and
by
comparing
coefficients.
To
check
for
significance
in
the
lat-
ter
case,
we
use
the
method
proposed
by
Wooldridge
(2009,
pp.
140–143).
We
define
a
new
coefficient
βpq (=
βp−
βq),
representing
the
difference
in
the
positivity
coefficients
of
touch-
points
p
and
q.
Our
null
hypothesis
is
that
βpq =
0,
that
is,
that
there
is
no
difference
in
the
coefficients,
against
the
alternate
βpq /=
0.
We
reparameterize
the
model
to
ensure
that
β
is
estimated
as
a
coefficient
(by
simple
algebraic
manipulation),
enabling
us
to
calculate
the
standard
error
associated
with
the
difference
and
hence
the
p-value
for
the
hypothesis
test.
We
summarize
the
resulting
coefficient
comparisons
in
Table
7.
The
table
shows
detailed
results
for
the
pooled
analysis,
and
summa-
rized
results
for
the
category-specific
analysis.
The
touchpoints
are
ranked
by
the
impact
of
their
positivity
on
consideration
change.
While
we
followed
a
similar
process
to
examine
the
relative
impact
of
touchpoint
frequency,
examination
of
Tables
5
and
6
shows
that
only
some
touchpoints
have
significant
frequency
coefficients
in
any
case,
and
the
coefficient
comparison
showed
few
significant
differences
amongst
these.
Hence,
we
suppress
these
results
for
brevity
(except
occasionally
in
the
text)
and
refer

246
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Table
7
Comparative
impacts
of
touchpoint
positivity.
Touchpoints Pooled
data:
rank
Pooled
data:
coefficient
differences
(Model
4)
Category-specific:
rank
(Model
3)
In-store
communica-
tions
Peer
obser-
vation
Brand
advertising
WOM Retailer
advertising
Traditional
earned
Electrical
goods
Technology
products
Mobile
handsets
Soft
drinks
β
SE
β
SE
β
SE
β
SE
β
SE
β
SE
In-store
com-
munications
1
0.099** 0.004
−0.024** 0.006
−0.025** 0.006
−0.039** 0.005
−0.042** 0.006
−0.063** 0.006
1
1
1=
1=
Peer
observation
2=0.075** 0.004
−0.002
0.005
−0.016** 0.005
−0.018** 0.006
−0.040** 0.006
2=
2=
1=
3=
Brand
advertising
2=0.074** 0.004 −0.014** 0.005
−0.016** 0.006
−0.038** 0.006
2=
5=
1=
1=
WOM 4=0.060** 0.004
−0.002
0.005
−0.024** 0.006
5
5=
1=
3=
Retailer
advertising
4=
0.057** 0.004
−0.022** 0.006
2=
2=
5=
3=
Traditional
earned
6
0.036** 0.004
6
2=
5=
3=
Significant
parameters
(p
<
0.05)
are
bolded.
** p
<
.01.
Off-diagonal
elements
show
the
difference
in
positivity
coefficients
and
their
associated
standard
error.
On-diagonal
elements
(in
italics)
show
the
touchpoint
positivity
coefficients
from
Model
4.
Touchpoints
are
ranked
by
the
relative
impact
of
their
touchpoint
positivity
on
consideration
change.
Rankings
are
derived
from
significant
differences
between
touchpoints’
positivity
coefficients.

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
247
the
reader
instead
to
the
frequency
coefficients
and
significance
levels
in
Tables
5
and
6.2
We
begin
with
the
pooled
model
and
consider
the
touchpoints
in
turn,
in
order
of
decreasing
positivity
impact,
as
summarized
in
the
ranking
of
Table
7.
Highest-ranked
is
in-store
com-
munications,
for
which
frequency
is
also
significant.
In-store
communications
such
as
shelf
and
display
make
the
brand
more
salient
at
the
point
of
purchase
(Van
Nierop
et
al.
2010),
poten-
tially
leading
to
unplanned
purchases
(Cobb
and
Hoyer
1986).
They
are
aided
by
their
multi-sensory
nature,
as
well
as
by
high
attention
levels
in
a
store
environment
(Peck
and
Wiggins
2006).
However,
this
effect
on
sales
is
not
direct
but
via
considera-
tion
(Van
Nierop
et
al.
2010;
Zhang
2006)
and
is
the
case
not
just
for
such
in-store
communications,
such
as
feature
ads
and
display
but
also
for
price-based
promotions,
which
also
play
a
role
in
consideration
set
evolution.
This
is
in
addition
to
the
role
of
discounted
price
in
the
customer’s
judgment
of
utility
at
the
moment
of
final
choice
(Van
Nierop
et
al.
2010).
The
empirical
importance
of
in-store
communications
in
our
data
is
consistent
with
recent
arguments
that
in-store
touchpoints
are
important
in
influencing
consideration
irrespective
of
where
and
when
the
purchase
is
made
(Court
et
al.
2009;
Verhoef,
Neslin,
and
Vroomen
2007).
Second-ranked
are
two
touchpoints,
brand
advertising
and
peer
observation.
It
is
notable
that
while
brand
advertising
is
influential
in
determining
consideration
through
both
fre-
quency
and
positivity
effects,
it
is
not
the
most
influential
touchpoint
in
terms
of
positivity.
This
supports
the
wider
agenda
for
a
touchpoint-neutral
view
of
the
customer
decision
jour-
ney
(Neslin
et
al.
2014),
and
in
particular
a
touchpoint-neutral
approach
to
customer
insight
(Macdonald,
Wilson,
and
Konus¸
2012).
While
WOM
positivity
is
significant,
in
line
with
the
con-
temporary
emphasis
on
social
effects,
it
is
notable
that
the
positivity
of
the
rarely
studied
peer
observation
touchpoint
is
significantly
more
influential.
Furthermore,
its
frequency
coef-
ficient
is
significantly
higher
than
that
for
WOM
(β
=
0.07,
SE
=
0.03,
p
<
0.01).
Seeing
someone
else
drinking
a
branded
drink
was
a
common
case
in
point
in
the
soft
drinks
category.
This
observation
led
to
marketing
strategies
in
a
sponsoring
firm
to
increase
the
frequency
and
positivity
of
such
touchpoints,
for
example
through
the
prominence
and
positioning
of
the
brand
on
the
product.
Retailer
advertising
also
has
a
significant
role
in
complement-
ing
advertising
by
the
brand
owner,
impacting
consideration
via
both
frequency
and
positivity.
Its
impact
via
frequency
is
not
sig-
nificantly
different
to
brand
advertising
(β
=
0.01,
SE
=
0.02,
ns),
but
the
impact
of
its
positivity
is
somewhat
lower.
Retailer
advertising
is
frequently
missing
from
practitioners’
media
mix
models
due
to
the
lack
of
available
data
(Macdonald,
Wilson,
and
Konus¸
2012),
but
this
result
shows
that
it
has
an
important
role
and
should
be
tracked.
2Equivalents
of
Table
7
for
frequency
and
for
competitor
effects
are
available
from
the
authors
on
request.
Finally,
traditional
earned
media
plays
a
significant
role
via
touchpoint
positivity,
though
we
could
not
detect
an
effect
via
frequency.
In
this
respect,
traditional
earned
media
are
similar
to
WOM.
The
absence
of
frequency
effect
may
be
related
to
the
low
mean
positivity
of
these
two
touchpoints,
and
in
Model
2
where
positivity
is
not
considered,
both
terms
become
significant.
This
suggests
that
careful
attention
to
both
frequency
and
positivity
is
required
in
earned
media
evaluation
too,
in
order
to
diagnose
how
the
impact
of
earned
media
can
be
increased,
or
whether
efforts
should
be
focused
elsewhere.
Competitor
Effects
Competitor
touchpoint
effects
are
accounted
for
in
Model
4.
Competitor
frequency
and
positivity
variables
test
for
any
direct
competitor
influence
on
consideration
for
the
focal
brand.
We
find
that
the
effect
of
several
competitor
touchpoints
is
significant
(and
in
the
expected,
negative,
direction
on
consid-
eration
change
for
the
focal
brand).
However,
in
comparison
to
focal
brand
effects,
the
effect
size
is
moderate,
as
indicated
by
somewhat
modest
coefficients
and
a
modest
increment
to
r2.
Again,
in-store
communication
is
important,
ranking
as
the
most
influential
competitor
touchpoint
via
both
frequency
and
positivity.
The
ability
for
the
consumer
to
compare
multiple
brands
simultaneously
in
a
store
may
contribute
to
this
as
com-
pared
with
touchpoints
where
brands
are
seen
in
isolation.
Also
as
with
the
focal
brand,
peer
observation
is
significant,
its
pos-
itivity
being
significantly
more
influential
than
that
of
WOM.
Again,
this
highlights
the
need
to
track
and,
where
feasible,
optimize
peer
observation.
The
frequency
of
competitor
advertising
(from
either
the
brand
or
retailer)
is
significant
but
its
positivity
is
not,
imply-
ing
that
mere
exposure
rather
than
perceptual
response
may
decrease
focal
brand
consideration.
However,
these
are
ranked
4th
and
5th
in
terms
of
the
impact
of
competitor
touchpoint
frequency,
behind
peer
influence
and
in-store
communica-
tions.
Comparing
Touchpoint
Impacts
by
Category
Next
we
consider
briefly
similarities
and
differences
to
the
pooled
model
in
the
category-specific
analysis:
see
Table
6
and
the
category-specific
ranks
in
Table
7
for
details.
In-store
com-
munications
is
consistently
the
most
important
touchpoint
across
categories
in
terms
of
positivity.
Its
frequency
is
also
signifi-
cant
in
soft
drinks,
a
sector
with
rich
opportunities
for
brand
encounters
out
of
the
home.
Peer
observation
positivity
is
also
significant
in
each
category,
and
while
it
is
less
so
than
brand
advertising
in
the
case
of
soft
drinks,
peer
observation
frequency
is
nonetheless
significant
in
this
category
in
which
consumption
is
readily
observed.
Overall,
then,
peer
observation
retains
its
importance
across
categories.
The
relative
impact
of
brand
advertising
is
fairly
consistent
across
categories,
being
ranked
the
equal
most
influential
touch-
point
via
positivity
in
two
categories
(mobile
handsets
and
soft
drinks),
and
the
most
influential
via
frequency
in
the
others.
Its

248
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
importance
relative
to
retailer
advertising
varies,
however,
in
the
positivity
analysis.
Whereas
in
soft
drinks
and
mobile
hand-
sets
brand
advertising
has
a
higher
coefficient,
consistent
with
the
pooled
analysis,
the
reverse
is
true
is
technology
products,
an
area
where
high
margins
lead
to
intense
competition
among
retailers.
Exploratory
Analyses
We
investigate
extensions
to
Model
4
via
three
exploratory
analyses.
The
first
considers
the
possible
interaction
between
touchpoint
frequency
and
positivity,
the
second
examines
the
impact
of
pre-consideration
on
touchpoint
impact,
and
the
third
investigates
the
impact
of
competitor
touchpoints
on
brand
touchpoint
performance.
Each
analysis
is
now
briefly
discussed.
Frequency/Positivity
Interaction
In
Exploratory
Analysis
1,
we
consider
the
possibility
that
touchpoint
frequency
and
positivity
may
interact.
For
example,
while
attitude
to
a
single
message
can
influence
brand
attitude,
attitude
strength
may
be
boosted
by
repeated
positive
(or
nega-
tive)
messages
(Erdem
and
Keane
1996).
Interacting
the
frequency
and
positivity
of
touchpoints,
whether
for
the
focal
brand
or
for
both
the
focal
and
competitor
brands,
does
not
lead
to
an
increase
in
model
fit
as
calculated
by
AIC
or
BIC
(Appendix).
Furthermore,
VIF
scores
substantially
increase,
most
likely
due
to
the
collinearity
we
are
introduc-
ing
through
interaction
terms.
With
this
warning,
we
briefly
highlight
preliminary
results
without
reporting
them
in
full
for
the
sake
of
brevity.3Future
research
may
better
isolate
these
interaction
effects,
if
they
exist.
First,
interaction
effects
are
all
in
the
expected
direction
(positive
for
focal
brand
and
nega-
tive
for
competitors).
Second,
the
competitor
interactions
which
are
significant
are
WOM,
in-store
communications,
and
retailer
advertising.
These
are
the
three
environments
where
multiple
brands
are
perhaps
most
likely
to
be
experienced
in
close
prox-
imity,
which
may
invoke
a
more
complex
relationship
between
these
touchpoints
and
consideration.
Finally,
the
significant
focal
brand
interactions
are
precisely
those
which
have
significant
frequency-only
effects,
namely
peer
observation,
retailer
adver-
tising,
in-store
communications,
and
brand
advertising,
again
suggesting
that
there
may
be
a
more
complex
relationship
at
play
between
frequency
and
positivity.
This
finding
is
consis-
tent
with
work
on
attitude
strength
(Erdem
and
Keane
1996),
and
shows
another
respect
in
which
taking
account
of
positivity
and
not
just
frequency
may
be
important.
Touchpoint
Interaction
With
Pre-Consideration
In
Exploratory
Analysis
2
(models
Exp2a/b
in
Appendix),
we
suggest
that
an
individual’s
pre-disposition
to
the
brand
may
affect
how
touchpoints
influence
his/her
shift
in
consideration.
3Results
tables
for
exploratory
analyses
are
available
from
the
authors
on
request.
Hence,
we
allow
ConsidPre
to
interact
with
the
touchpoint
vari-
ables
by
reformulating
the
touchpoint
coefficients,
such
that:
βfreq
i,k,j =
β(freq×pre)
1,j +
β(freq×pre)
2,j ConsidPrei,k
And
similar
for
βpos
−,
γfreq
−,
and
γpos
−.
Model
Exp2b
provides
an
improvement
over
Model
4
(Appendix).
This
model
includes
the
interaction
of
initial
focal
brand
consideration
with
touchpoint
variables
(for
both
focal
and
competitor
brands).
However,
due
to
the
large
number
of
interactions,
VIF
scores
are
high
(average
8.79).
While
our
data
suggest
that
this
interaction
exists,
further
investigation
is
therefore
needed
to
establish
its
exact
strength
and
significance.
Hence
again
we
do
not
report
results
in
detail
but
instead
pro-
vide
an
overview.
In
general,
as
an
individual’s
pre-consideration
increases,
the
impact
of
touchpoint
frequency
and
positivity
on
their
change
in
consideration
decreases.
This
suggests
that
con-
sumers
who
have
a
more
favorable
predisposition
to
the
brand
are
less
impacted
by
brand
encounters.
This
could
be
a
straight-
forward
case
of
regression
to
the
mean,
where
consumers
who
already
hold
a
very
positive
opinion
are
more
likely
to
move
down
the
scale
or
stay
where
they
are
rather
than
further
increase
their
opinion.
This
is
managerially
interesting
when
deciding
targets
for
touchpoints
such
as
addressable
media,
particularly
where
the
aim
of
the
communication
is
attitudinal
rather
than
directly
behavioral.
We
also
see
that
as
an
individual’s
pre-consideration
increases,
the
impact
of
competitor
frequency
and
positivity
increases:
that
is,
the
pulling
power
of
competitor
touchpoints
is
greater
for
those
who
have
a
favorable
predisposition
to
the
focal
brand.
Again,
we
conjecture
that
this
is
a
regression
to
the
mean
effect.
Competitor
Effects
on
Consideration
In
Exploratory
Analysis
3
(Exp3a/b/c/d),
we
attempt
to
mea-
sure
the
indirect
effect
of
competitor
touchpoints
on
focal
brand
consideration
via
an
interaction
with
focal
brand
touchpoints.
We
investigate
the
impact
of
competitor
clutter
on
focal
brand
touchpoint
performance
(Danaher,
Bronfer,
and
Dhar
2008).
We
include
an
interaction
term
between
focal
and
competitor
touch-
point
frequency
and,
as
proposed
by
Danaher,
Bonfrer,
and
Dhar
(2008),
attempt
to
moderate
this
by
the
proportion
of
competi-
tors
experienced.
We
do
this
using
the
reparameterization
of:
βfreq
i,k,j =
βcomp
1,j +
βcomp
2,j ρ
/=
k
ρI[Freqi,p,j>0]
Bi−
1Freqi,−k,j
where
I[f(x)] =
1
if
the
statement
f(x)
is
true,
that
is,
if
respondent
i
has
an
experience
with
brand
ρ
through
touchpoint
j,
and
zero
otherwise;
and
Biis
the
total
number
of
brands
which
individual
i
was
asked
to
report
on
–
that
is,
we
are
calculating
the
propor-
tion
of
competitor
brands
which
respondent
i
has
experienced.
We
also
investigate
whether
competitor
positivity
(AvgPosi,−k,j)
moderates
focal
touchpoint
frequency,
and
further,
the
moderat-
ing
effect
on
focal
touchpoint
positivity
(a
reparameterization
of
βpos
−).
Appendix
shows
model
fit
for
each
of
these
explorations.

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
249
Whilst
none
of
these
models
decreases
BIC,
Model
Exp3b
is
preferred
over
Model
4
by
AIC
although
there
is
no
real
increase
in
the
r2.
With
this
warning,
we
briefly
report
preliminary
results
to
aid
future
research.
In
each
model,
the
significant
interactions
are
all
in
the
expected
negative
direction:
an
improvement
in
competitor
touchpoints
(whether
frequency
or
positivity)
results
in
a
lower
impact
from
focal
brand
touchpoints.
In
the
preferred
Model
Exp3b,
competitor
positivity
reduces
the
impact
of
focal
brand
frequency
for
four
touchpoints:
brand
advertising,
peer
observation,
in-store
communications,
and
retailer
advertising.
This
is
consistent
with
Danaher
et
al.
(2008)
who
found
that
when
competitors
and
focal
brands
advertise
concurrently
the
elasticity
of
the
focal
brand’s
advertising
reduces.
Our
results
show
that
this
could
also
extend
into
retailer
advertising
and
into
positivity.
Conclusion
In
this
study,
we
tracked
the
impact
of
contemporaneously
reported
touchpoints
on
brand
consideration
across
four
con-
sumer
goods
categories.
We
examined
the
impact
on
brand
consideration
change
of
six
touchpoints.
In
our
main,
pooled
Model
4
(Table
5),
we
found
that
touchpoint
positivity
signifi-
cantly
impacts
consideration
change
for
all
six
touchpoints,
and
touchpoint
frequency
does
so
for
all
but
WOM
and
traditional
earned
media.
We
further
rank
the
touchpoints
by
the
touchpoint
positivity
coefficients
(Table
7)
and
find
that
in-store
communi-
cations
are
most
influential,
followed
by
peer
observation
and
brand
advertising,
then
WOM
and
retailer
advertising.
Finally,
traditional
earned
media
are
the
least
influential.
The
impact
of
competitor
touchpoints
on
a
focal
brand
was
also
examined
(Table
5).
Again,
in-store
communications
are
most
influen-
tial
(via
both
frequency
and
positivity),
and
as
with
the
focal
brand,
peer
observation
has
a
significant
effect,
its
positivity
being
significantly
more
influential
than
that
of
WOM.
We
hence
make
three
contributions.
First,
the
study
is
to
our
knowledge
one
of
the
first,
if
not
the
first,
on
the
relative
impact
of
brand,
retailer,
peer
and
earned
touchpoints
on
the
customer’s
brand
relationship.
Notably,
peer
observation,
predominantly
the
focus
until
now
of
qualitative
research
(Grove
and
Fisk
1997),
is
both
frequent
and
influential,
suggesting
that
this
touchpoint
requires
far
more
attention
from
both
scholars
and
practition-
ers.
A
recent
line
of
research
(Nitzan
and
Libai
2011;
Risselada,
Verhoef,
and
Bijmolt
2014)
shows
the
importance
of
social
con-
nections
on
consumer
behavior.
Our
research
sheds
light
on
the
mechanisms
underpinning
these
social
effects
by
empirically
distinguishing
WOM
(recommendation
or
criticism)
from
sim-
ply
observing
peers.
Earned
media
are
somewhat
less
influential
but
are
nonetheless
significant.
While
the
role
of
retailer
advertis-
ing
is
somewhat
category
contingent,
in-store
communications
are
consistently
impactful.
Our
second
contribution
is
to
propose
and
demonstrate
that
the
assessment
of
touchpoint
impact
needs
to
take
into
account
touchpoint
positivity
and
not
just
frequency.
We
find
that
pos-
itivity
adds
explanatory
power
as
compared
with
frequency
alone
when
predicting
brand
consideration.
This
generalizes
findings
from
long-standing
experimental
advertising
research
(MacKenzie,
Lutz,
and
Belch
1986)
to
a
multi-touchpoint
con-
text.
Positivity
by
definition
is
a
real-time
affective
response
which
can
only
be
recalled
imperfectly
and
with
significant
known
biases
(Aaker,
Drolet,
and
Griffen
2008;
Cowley
2008).
This
makes
the
survey
problematic
for
such
research,
while
behavioral
measures
mostly
fail
to
capture
positivity
entirely.
We
have
illustrated
one
method
for
addressing
this,
through
the
RET
texting
approach;
alternative
methods
may
be
possi-
ble.
Real-time
reporting
takes
the
logic
of
mall
intercepts
(and
variants
such
as
exit
surveys
as
customers
leave
a
website)
and
generalizes
it
to
the
challenge
that
decision
journeys
play
out
in
real
time
across
diverse
touchpoints.
This
brings
us
to
our
third
contribution,
which
is
to
propose
and
exemplify
an
RET-based
approach
by
which
the
impact
of
multiple
touchpoints
can
be
assessed.
This
approach
treats
symmetrically
touchpoints
with
the
brand
owner,
the
retailer,
peers
and
the
media.
We
hence
respond
to
calls
for
research
which
acknowledges
that
the
consumer
decision
journey
extends
beyond
firm-owned
media
and
channel
contacts
(Ailawadi
et
al.
2009;
Court
et
al.
2009).
Customers
integrate
learning
from
mul-
tiple
sources
in
order
to
achieve
their
objectives
(Neslin
et
al.
2014).
In
our
study,
touchpoints
significantly
associated
with
brand
consideration
included
those
from
four
stakeholders:
the
brand
owner,
retailers,
peers,
and
the
public
media.
Yet
there
are
other
stakeholders
who
the
customer
may
touch,
and
whose
touchpoints
could
be
included
within
further
applications
of
this
approach,
such
as
sponsors
(Court
et
al.
2009)
and
service
personnel
(Grove
and
Fisk
1997).
Practitioner
Implications
As
classic
market
research
is
increasingly
complemented
by
database
analytics,
managers
are
hardly
short
of
customer
data.
But
these
data
are
fragmented,
hiding
key
insights
on
the
customer’s
holistic
relationship
with
the
brand.
They
are
also
fre-
quently
incomplete,
as
empowered
customers
take
less
notice
of
company-driven
communication,
choosing
instead
to
learn
from
the
experience
of
other
customers
and
doing
their
own
research
online.
Marketers
need
to
know
which
parts
of
the
customer
journey
have
most
impact
on
attitudes
and
behaviors,
and
which
of
these
crucial
encounters
are
not
working
well.
Methods
such
as
real-time
experience
tracking
may
prove
a
useful
addition
to
the
methodological
armory
to
complement
both
ethnographic
approaches
on
the
one
hand
and,
on
the
other,
focused
quanti-
tative
work
within
subsets
of
the
touchpoint
mix.
Whether
or
not
data
collection
follows
the
SMS-based
approach
we
have
described,
we
tentatively
suggest
three
guidelines
to
practition-
ers
for
providing
holistic
customer
insight.
First,
we
suggest
widening
the
scope
of
insight
to
all
direct
and
indirect
touchpoints,
as
an
input
into
the
overall
marketing
plan.
For
instance,
should
a
company
invest
in
advertising
or
in
improving
call
center
standards,
in
product
design
improve-
ment
or
online
advice,
in
supporting
customer
communication
through
channel
partners
or
in
social
media?
While
a
company’s
overall
positioning
and
competencies
will
inform
such
decisions,
we
suggest
that
holistic
insight
across
multiple
touchpoints
can
help.

250
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
Second,
we
suggest
tracking
the
customer’s
perceptual
response
to
touchpoints
contemporaneously.
Even
if
objective
data
were
available
on
all
touchpoints,
it
would
not
include
this
important
information.
To
get
closer
to
customers,
one
might
ide-
ally
walk
along
with
them,
asking
how
they
feel
at
the
moment
when
they
encounter
the
brand.
Asking
this
at
the
end
of
the
month
in
a
tracker
survey
may
be
too
late
to
capture
the
problem
or
opportunity.
As
mobile
handsets
tend
to
travel
with
the
cus-
tomer,
they
seem
a
natural
place
to
start
in
seeking
this
real-time
feedback.
Third,
we
suggest
assessing
the
impact
of
encounters
on
key
outcomes.
These
may
be
attitudinal,
as
in
this
study,
or
behavioral,
as
we
discuss
further
below.
A
bank
might
wish
to
know,
for
example,
whether
it
should
invest
further
in
mar-
keting
communications,
or
whether
improvement
in
service
levels
would
have
a
higher
impact
on
consideration
and
pur-
chase.
Limitations
and
Research
Directions
While
we
have
employed
some
robustness
checks,
future
studies
might
usefully
further
explore
the
strengths
and
weak-
nesses
of
real-time
experience
tracking
in
focused
research
efforts,
analogous
to
the
methodological
studies
of
survey
meth-
ods
(Chandon,
Morwitz,
and
Reinartz
2006).
First,
for
some
touchpoint
types,
self-reports
could
be
checked
against
objec-
tive
sources
such
as
CRM
data.
Second,
a
comparison
against
retrospective
surveys
might
allocate
respondents
randomly
to
one
method.
We
might
expect
real-time
reporting
to
be
fuller
and
more
accurate
–
given
Wind
and
Lerner’s
(1979)
findings
when
comparing
surveys
with
purchase
diaries,
of
which
RET
can
be
thought
of
as
a
variant
–
as
well
as
more
differentiating
in
perceptual
response.
These
conjectures
could
be
tested
using
a
field
experiment.
Such
pairwise
comparisons
of
methods
might
also
examine
the
relative
explanatory
power
of
different
meth-
ods
on
an
attitudinal
or
behavioral
outcome,
to
test
the
extent
to
which
real-time
experience
tracking
captures
encounters
that
prove
to
be
significant.
Third,
touchpoints
mentioned
in
post-
study
interviews
could
be
compared
against
data
from
real-time
tracking.
Such
methodological
studies
would
amongst
other
things
enable
the
estimation
of
mere
measurement
effects.
As
with
sur-
vey
methods,
the
act
of
asking
respondents
to
respond
is
itself
an
intervention
which
may
influence
brand
attitudes
(Chandon,
Morwitz,
and
Reinartz
2006).
Unlike
some
company
surveys,
however,
our
respondents
were
not
aware
of
any
particular
brand
sponsoring
the
study.
We
conjecture,
therefore,
that
study
partic-
ipants
may
be
to
some
extent
hot-housed,
paying
more
attention
to
the
whole
category
than
they
might
otherwise,
and
perhaps
thereby
exhibiting
greater
shifts
in
brand
attitudes
than
non-
participants.
Any
such
effect
might
be
expected,
though,
to
be
equal
across
brands.
An
experimental
design
in
which
a
con-
trol
group
fills
in
only
pre-study
and
post-study
surveys
without
SMS
messaging
in-between
could
perhaps
check
this
conjec-
ture.
Hot-housing
might
also
cause
respondents
to
notice
and
hence
report
greater
touchpoint
frequencies
than
a
control
group.
Conversely,
the
agency
problem
may
lead
to
respondents
not
reporting
all
touchpoints
due
to
laziness.
Again,
experiments
are
needed
to
check
any
downward
or
upward
bias
in
report-
ing.
Another
research
opportunity
concerns
the
tracking
period.
We
found
that
even
with
around
1700–5600
respondents,
the
sheer
breadth
of
touchpoint
types
led
to
some
touchpoints
being
relatively
sparsely
represented
for
some
brands
within
the
study
period
of
one
week.
While
a
greater
number
of
respondents
might
help,
a
powerful
option
would
be
longitudinal
studies
cov-
ering
a
longer
tracking
period
of
perhaps
one
month.
In
addition
to
raising
the
statistical
power
for
relatively
infrequent
touch-
points,
this
might
also
increase
the
statistical
power
for
further
exploration
of
interactions
(Naik
and
Peters
2009).
Furthermore,
longitudinal
data
structured
in
panel
data
format
could
allow
the
examination
of
the
time-variant
dynamic
effects
of
touchpoints,
such
as
the
recency,
frequency
and
sequential
order
of
encoun-
ters.
Such
longitudinal
data
might
also
be
the
key
to
bringing
customer
initiated
touchpoints
into
the
analysis,
such
as
product
use,
product
purchase,
or
visiting
a
brand
website.
These
might
be
modeled
as
resulting
from
the
impact
of
prior
encounters
as
well
as
pre-study
attitudes.
A
further
limitation
and
research
direction
concerns
the
possi-
bility
of
touchpoint
endogeneity.
In
common
with
most
research
on
the
impact
of
touchpoints
from
advertising
to
WOM
(Archak,
Ghose,
and
Ipeirotis
2011;
Bass
et
al.
2007;
Goh,
Hui,
and
Png
2011;
Liu
2006),
we
have
treated
touchpoints
as
independent.
However,
this
simplification
may
bias
coefficients.
For
example,
those
individuals
who
are
more
likely
to
increase
their
consider-
ation
for
a
brand
may
also
be
more
likely
to
notice
touchpoints
for
that
brand
or
perceive
them
as
positive.
Therefore
their
shift
in
consideration
is
not
wholly
due
to
their
experience
but
also
a
result
of
some
unobserved
engagement
with
the
brand.
Or
there
may
be
psychographic
or
lifestyle
variables
that
impact
touchpoint
frequency
or
positivity.
Hence
there
may
be
omit-
ted
variable
bias
affecting
coefficient
estimates.
Related,
firm
actions
are
tacit
within
our
model:
while
our
analysis
is
primar-
ily
brand
neutral,
brand
strategies
may
target
a
segment
who
are
naturally
more
likely
to
increase
their
consideration
for
the
brand,
in
which
case
a
participant’s
segment
membership
is
cor-
related
with
both
their
frequency
of
exposure
and
their
change
in
consideration.
By
omitting
any
relevant
segment
variables
we
may
be
introducing
bias
into
the
estimate
of
frequency,
as
frequency
is
correlated
with
an
omitted
variable.
We
do
not
have
available
suitable
instrumental
variables
to
adequately
identify
whether
and
to
what
extent
this
endogene-
ity
issue
exists,
and
we
prefer
not
to
use
weak
or
ill-defined
instrumental
variables
as
they
are
likely
to
introduce
further
bias
rather
than
remove
it
(Larcker
and
Rusticus
2010;
Wooldridge
2009).
This
issue
deserves
focused
attention
in
future
research.
Again,
one-month
datasets
may
help,
where
psychographic
and
socio-demographic
variables
with
potential
conceptual
links
to
touchpoints
would
need
to
be
included.
While
we
have
reported
an
exploratory
analysis
of
the
interaction
between
prior
consid-
eration
and
touchpoint
impact,
conceptually
the
best
measure
of
prior
brand
relationship
in
predicting
touchpoints
might
be
the
recently
clarified
construct
of
brand
engagement
(Brodie
et
al.
2011).
Another
potential
predictor
of
touchpoints
might
be
the

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
251
respondent’s
involvement
in
the
study,
as
this
may
impact
on
the
level
of
the
potential
biases
we
have
discussed
in
touchpoint
recording.
A
final
limitation
concerns
unobserved
customer
hetero-
geneity.
The
gap
between
marginal
r2and
conditional
r2for
the
preferred
models
suggests
that
around
8–14%
of
variabil-
ity
in
consideration
change
could
be
explained
by
unobserved,
individual
level
data.
Again,
careful
consideration
of
relevant
psychographic,
socio-demographic
or
brand
health
variables
may
shed
further
light
and
be
managerially
useful.
Concluding
Remarks
There
was
perhaps
a
time
when
customers
learned
about
products
and
services
through
what
the
brand
owner
told
them.
If
this
time
ever
existed,
it
is
certainly
not
the
case
now,
as
our
data
make
plain.
A
focus
purely
on
optimizing
the
spend
within
the
brand-owner’s
control
would
be
myopic.
Instead,
we
suggest
lis-
tening
to
customers
in
real
time
to
understand
how
they
construe
their
customer
journey.
The
range
of
touchpoints
they
encounter
in
this
journey
is
undoubtedly
broad
but
perhaps
not
intractably
so.
Managers
take
decisions
every
day
based
on
their
working
assumptions
about
their
relative
importance
and
efficacy.
The
research
challenge
is
to
support
these
holistic
decisions
with
holistic
insight.
Acknowledgements
The
authors
gratefully
acknowledge
the
financial
support
pro-
vided
by
the
Technology
Strategy
Board
(UK)
and
the
Economic
and
Social
Research
Council
(UK)
via
a
Knowledge
Transfer
Partnership
grant.
The
authors
also
gratefully
acknowledge
the
support
and
data
access
provided
by
research
agency
MESH
Experience.
Appendix.
Fit
statistics
for
robustness
checks
and
exploratory
models
(based
on
pooled
data
Model
4)
Model
AIC
BIC
r2marginal
r2conditional
Average
VIF
Maximum
VIF
Robustness
checks
Model
Freq1:
Frequency
–
Dichotomous
Incidence
Variable
231,091
231,713
19.5%
31.7%
1.90
2.66
Model
Freq2:
Frequency
–
Linear
Frequency
231,010
231,633
19.5%
31.7%
1.75
2.64
Model
Freq3:
Frequency
–
Quadratic
Decay
231,150
231,921
19.6%
31.8%
2.10
3.10
Model
Freq4:
Frequency
–
Natural
Log
(ln)
decay
230,905a231,527a19.6%
31.8%
1.84
2.67
Model
Pos1:
Positivity
–
Arithmetic
Mean
230,905a231,527a19.6%
31.8%
1.84
2.67
Model
Pos2:
Positivity
–
Mean
and
Variance
230,926
231,697
19.7%
31.8%
1.77
2.67
Model
Pos3:
Positivity
–
Frequency
of
positive,
negative,
and
neutral
experiences
231,958
232,729
18.6%
30.6%
2.17
3.45
Model
Pos4:
Positivity
–
Last
experience
positivity
233,445
234,067
17.2%
29.1%
1.72
2.63
Model
Pos5:
Positivity
–
Mean
and
last
experience
positivity
230,988
231,758
19.7% 31.8%
2.28
5.71
Model
Imp1:
Positivity
imputation
–
using
zero-coding
230,905a231,527a19.6%
31.8%
1.84
2.67
Model
Imp2:
Positivity
imputation
–
using
mean
231,022
231,644
19.4%
31.5%
2.92
18.21
Exploratory
models
Model
4:
Preferred
model
for
pooled
data
230,905
231,527
19.6%
31.8%
1.84
2.67
Model
Exp1a:
Focal
brand
frequency
and
positivity
interaction
230,916
231,613
19.7%
31.8%
4.31
19.53
Model
Exp1b:
Focal
brand
and
competitor
frequency
and
positivity
interaction
230,936
231,707
19.7%
31.8%
5.19
19.91
Model
Exp2a:
Focal
initial
consideration
interaction
with
focal
brand
touchpoints
230,321
231,092
20.2%
32.2%
6.75
21.02
Model
Exp2b:
Focal
initial
consideration
interaction
with
focal
and
competitor
touchpoints
230,119a231,038a20.8%
32.7%
8.98
21.19
Model
Exp3a:
Competitor
frequency
interaction
with
focal
brand
frequency
230,969
231,665
19.7%
31.8%
1.82
2.67
Model
Exp3b:
Competitor
positivity
interaction
with
focal
brand
frequency
230,865a231,561
19.7%
31.8%
1.85
2.67
Model
Exp3c:
Competitor
frequency
interaction
with
focal
brand
positivity
230,974
231,670
19.7%
31.8%
1.79
2.67
Model
Exp3d:
Competitor
positivity
interaction
with
focal
brand
positivity
230,925
231,622
19.7%
31.8%
1.85
2.67
aPreferred
model.

252
S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
References
Aaker,
Jennifer,
Aimee
Drolet
and
Dale
Griffen
(2008),
“Recalling
Mixed
Emo-
tions,”
Journal
of
Consumer
Research,
35
(2),
268–78.
Ahluwalia,
Rohini,
Robert
E.
Burnkrant
and
H.
Rao
Unnava
(2000),
“Consumer
Response
to
Negative
Publicity:
The
Moderat-
ing
Role
of
Commitment,”
Journal
of
Marketing
Research,
37,
203–14.
Ailawadi,
Kusum
L.,
J.P.
Beauchamp,
Naveen
Donthu,
Dinesh
K.
Gauri
and
Venkatesh
Shankar
(2009),
“Communication
and
Promotion
Decisions
in
Retailing:
A
Review
and
Directions
for
Future
Research,”
Journal
of
Retail-
ing,
85
(1),
42–55.
Archak,
Nicolay,
Anindya
Ghose
and
Panagiotis
G.
Ipeirotis
(2011),
“Deriv-
ing
the
Pricing
Power
of
Product
Features
by
Mining
Consumer
Reviews,”
Management
Science,
57
(8),
1485–509.
Arnold,
Mark
and
Kristy
Reynolds
(2009),
“Affect
and
Retail
Shopping
Behav-
ior:
Understanding
the
Role
of
Mood
Regulation
and
Regulatory
Focus,”
Journal
of
Retailing,
85
(3),
308–20.
Ataman,
M.
Berk,
Harald
J.
van
Heerde
and
Carl
F.
Mela
(2010),
“The
Long-
term
Effect
of
Marketing
Strategy
on
Brand
Sales,”
Journal
of
Marketing
Research,
47
(1),
866–82.
Bass,
Frank
M.,
Norris
Bruce,
Sumit
Majumdar
and
B.P.S.
Murthi
(2007),
“Wearout
Effects
of
Different
Advertising
Themes:
A
Dynamic
Bayesian
Model
of
the
Advertising-Sales
Relationship,”
Marketing
Science,
26
(2),
179–95.
Baumeister,
Roy
F.,
Kathleen
D.
Vohs,
C.
Nathan
DeWall
and
Liqing
Zhang
(2007),
“How
Emotion
Shapes
Behavior:
Feedback,
Anticipation,
and
Reflection,
Rather
Than
Direct
Causation,”
Personality
and
Social
Psychol-
ogy
Review,
11,
167–203.
Bearden,
William
O.
and
Michael
J.
Etzel
(1982),
“Reference
Group
Influence
on
Product
and
Brand
Purchase
Decisions,”
Journal
of
Consumer
Research,
9
(2),
183–94.
Bian,
Xuemei
and
Luiz
Moutinho
(2009),
“An
Investigation
of
Determinants
of
Counterfeit
Purchase
Consideration,”
Journal
of
Business
Research,
62
(3),
368–78.
Borghini,
Stefania,
Nina
Diamond,
Robert
V.
Kozinets,
Mary
Ann
McGrath,
Albert
M.
Mu˜
niz
Jr.
and
John
F.
Sherry
Jr.
(2009),
“Why
Are
Themed
Brand-
stores
So
Powerful?
Retail
Brand
Ideology
at
American
Girl
Place,”
Journal
of
Retailing,
85
(3),
363–75.
Bri˜
nol,
Pablo,
Richard
E.
Petty
and
Zakary
L.
Tormala
(2004),
“Self-Validation
of
Cognitive
Responses
to
Advertisements,”
Journal
of
Consumer
Research,
30
(4),
559–72.
Brodie,
Roderick
J.,
Linda
D.
Hollebeek,
Biljana
Juric
and
Ana
Ilic
(2011),
“Customer
Engagement:
Conceptual
Domain,
Fundamental
Propositions,
and
Implications
for
Research,”
Journal
of
Service
Research,
14
(3),
252–71.
Bülbül,
Cenk
and
Geeta
Menon
(2010),
“The
Power
of
Emotional
Appeals
in
Advertising:
The
Influence
of
Concrete
Versus
Abstract
Affect
on
Time-Dependent
Decisions,”
Journal
of
Advertising
Research,
50
(2),
169–80.
Burnham,
Kenneth
P.
and
David
R.
Anderson
(2004),
“Multimodel
Inference:
Understanding
AIC
and
BIC
in
Model
Selection,”
Sociological
Methods
Research,
33
(2),
261–304.
Chandon,
Pierre,
Vicki
G.
Morwitz
and
Werner
Reinartz
(2006),
“Do
Inten-
tions
Really
Predict
Behavior?
Self-Generated
Validity
Effects
in
Survey
Research,”
Journal
of
Marketing,
69
(2),
1–14.
Chitturi,
Favindra,
Rajagopal
Raghunathan
and
Vijay
Mahajan
(2008),
“Delight
by
Design:
The
Role
of
Hedonic
Versus
Utilitarian
Benefits,”
Journal
of
Marketing,
72
(May),
48–63.
Cobb,
Cathy
J.
and
Wayne
D.
Hoyer
(1986),
“Planned
Versus
Impulse
Purchase
Behavior,”
Journal
of
Retailing,
62
(4),
384–409.
Court,
David,
Dave
Elzinga,
Susan
Mulder
and
Ole
J.
Vetvik
(2009),
“The
Consumer
Decision
Journey,”
McKinsey
Quarterly,
3,
96–107.
Cowley,
Elizabeth
(2008),
“Looking
Back
at
an
Experience
Through
Rose-
Colored
Glasses,”
Journal
of
Business
Research,
61,
1046–52.
Danaher,
Peter
J.,
André
Bonfrer
and
Sanjay
Dhar
(2008),
“The
Effect
of
Com-
petitive
Advertising
Interference
on
Sales
for
Packages
Goods,”
Journal
of
Marketing
Research,
45
(April),
211–25.
Desai,
Kalpesh
K.
and
Sekar
Raju
(2007),
“Adverse
Influence
of
Brand
Commitment
on
Consideration
of
and
Preference
for
Competing
Brands,”
Psychology
&
Marketing,
24
(7),
595–614.
East,
Robert,
Kathy
Hammond
and
Wendy
Lomax
(2008),
“Measuring
the
Impact
of
Positive
and
Negative
Word
of
Mouth
on
Brand
Pur-
chase
Probability,”
International
Journal
of
Research
in
Marketing,
25,
215–24.
Erdem,
Tülin
and
Michael
P.
Keane
(1996),
“Decision-Making
Under
Uncer-
tainty:
Capturing
Dynamic
Brand
Choice
Processes
in
Turbulent
Consumer
Goods
Markets,”
Marketing
Science,
15
(1),
1–20.
Goh,
Khim-Yong,
Kai-Lung
Hui
and
Ivan
Png
(2011),
“Newspaper
Reports
and
Consumer
Choice:
Evidence
from
the
Do
Not
Call
Registry,”
Management
Science,
57
(9),
1640–54.
Goodman,
Joseph
K.,
Susan
M.
Broniarczyk,
Jill
G.
Griffin
and
Leigh
McAlister
(2013),
“Help
or
Hinder?
When
Recommendation
Signage
Expands
Con-
sideration
Sets
and
Heightens
Decision
Difficulty,”
Journal
of
Consumer
Psychology,
23
(2),
165–74.
Grove,
Stephen
J.
and
Raymond
P.
Fisk
(1997),
“The
Impact
of
Other
Cus-
tomers
on
Service
Experiences:
A
Critical
Incident
Examination
of
“Getting
Along”,”
Journal
of
Retailing,
73,
63–85.
Havlena,
William,
Robert
Cardarelli
and
Michelle
De
Montigny
(2007),
“Quan-
tifying
the
Isolated
and
Synergistic
Effects
of
Exposure
Frequency
for
TV,
Print,
and
Internet
Advertising,”
Journal
of
Advertising
Research,
45
(2),
255–68.
Kahn,
Barbara
E.
and
Alice
M.
Isen
(1993),
“The
Influence
of
Positive
Affect
on
Variety
Seeking
among
Safe,
Enjoyable
Products,”
Journal
of
Consumer
Research,
20,
257–70.
Kahneman,
Daniel
and
Alan
B.
Krueger
(2006),
“Developments
in
the
Mea-
surement
of
Subjective
Well-Being,”
Journal
of
Economic
Perspectives,
20
(1),
3–24.
Larcker,
David
F.
and
Tjomme
O.
Rusticus
(2010),
“On
the
Use
of
Instrumental
Variables
in
Accounting
Research,”
Journal
of
Accounting
and
Economics,
49,
186–205.
Liu,
Yong
(2006),
“Word
of
Mouth
for
Movies:
Its
Dynamics
and
Impact
on
Box
Office
Revenue,”
Journal
of
Marketing,
70
(July),
74–89.
Macdonald,
Emma
K.,
Hugh
N.
Wilson
and
Umut
Konus¸
(2012),
“Better
Customer
Insight
–
In
Real
Time,”
Harvard
Business
Review,
90
(9),
102–8.
MacKenzie,
Scott
B.,
Richard
J.
Lutz
and
Beorge
E.
Belch
(1986),
“The
Role
of
Attitude
Toward
the
Ad
as
a
Mediator
of
Advertising
Effectiveness:
A
Test
of
Competing
Explanations,”
Journal
of
Marketing
Research,
23,
130–43.
McAlexander,
James
H.,
John
W.
Schouten
and
Harold
F.
Koenig
(2002),
“Building
Brand
Community,”
Journal
of
Marketing,
66
(January),
38–54.
McFarland,
Cathy
and
Roger
Buehler
(1998),
“The
Impact
of
Negative
Affect
on
Autobiographical
Memory:
The
Role
of
Self-Focused
Atten-
tion
to
Moods,”
Journal
of
Personality
and
Social
Psychology,
75
(6),
1424–40.
Nakagawa,
Shinichi
and
Holger
Schielzeth
(2013),
“A
General
and
Simple
Method
for
Obtaining
R2From
Generalized
Linear
Mixed-Effects
Models,”
Methods
in
Ecology
and
Evolution,
4
(2),
133–42.
Naik,
Prasad
A.
and
Kay
Peters
(2009),
“A
Hierarchical
Marketing
Communica-
tions
Model
of
Online
and
Offline
Media
Synergies,”
Journal
of
Interactive
Marketing,
23,
288–99.
Neslin,
Scott
A.,
Dhruv
Grewal,
Robert
Leghorn,
Venkatesh
Shankar,
Marije
L.
Teerling,
Jacquelyn
S.
Thomas
and
Peter
C.
Verhoef
(2006),
“Challenges
and
Opportunities
in
Multichannel
Customer
Management,”
Journal
of
Service
Research,
9
(2),
95–112.
Neslin,
Scott
A.,
Kinshuk
Jerath,
Bodapati
Anand,
Eric
T.
Bradlow,
John
Deighton,
Sonja
Gensler,
Leonard
Lee,
Elisa
Montaguti,
Rahul
Telang,
Raj
Venkatesan,
Peter
C.
Verhoef
and
Z.
John
Zhang
(2014),
“The
Interrela-
tionships
Between
Brand
and
Channel
Choice,”
Marketing
Letters,
25
(3),
319–30.
Nitzan,
Irit
and
Barak
Libai
(2011),
“Social
Effects
on
Customer
Retention,”
Journal
of
Marketing,
75
(November),
24–38.
Ngobo,
Paul
V.
(2011),
“What
Drives
Household
Choice
of
Organic
Products
in
Grocery
Stores?,”
Journal
of
Retailing,
87
(1),
90–100.

S.
Baxendale
et
al.
/
Journal
of
Retailing
91
(2,
2015)
235–253
253
O’Brien,
Robert
M.
(2007),
“A
Caution
Regarding
Rules
of
Thumb
for
Variance
Inflation
Factors,”
Quality
and
Quantity,
41
(5),
673–90.
O’Cass,
Aron
(2002),
“Political
Advertising
Believability
and
Informa-
tion
Source
Value
During
Elections,”
Journal
of
Advertising,
31
(1),
63–74.
Peck,
Joanne
and
Jennifer
Wiggins
(2006),
“It
Just
Feels
Good:
Customers’
Affective
Response
to
Touch
and
its
Influence
on
Persuasion,”
Journal
of
Marketing,
70
(4),
56–69.
Priester,
Joseph,
Dhananjay
Nayakankuppam,
Monique
A.
Fleming
and
John
Godek
(2004),
“The
A2SC2Model:
The
Influence
of
Attitudes
and
Attitude
Strength
on
Consideration
and
Choice,”
Journal
of
Consumer
Research,
30
(4),
574–85.
Risselada,
Hans,
Peter
C.
Verhoef
and
Tammo
H.A.
Bijmolt
(2014),
“Dynamic
Effects
of
Social
Influence
and
Direct
Marketing
on
the
Adoption
of
High-
Technology
Products,”
Journal
of
Marketing,
78
(March),
52–68.
Roberts,
John
H.
and
James
M.
Lattin
(1997),
“Consideration:
Review
of
Research
and
Prospects
for
Future
Insights,”
Journal
of
Marketing
Research,
34,
406–10.
Stammerjohan,
Claire,
Charles
M.
Wood,
Yuhmiin
Change
and
Esther
Thorson
(2005),
“An
Empirical
Investigation
of
the
Interaction
Between
Publicity,
Advertising,
and
Previous
Brand
Attitudes
and
Knowledge,”
Journal
of
Advertising,
34
(4),
55–67.
Stephen,
Andrew
T.
and
Jeff
Galak
(2012),
“The
Effects
of
Traditional
and
Social
Earned
Media
on
Sales:
A
Study
of
a
Microlending
Marketplace,”
Journal
of
Marketing
Research,
49
(5),
624–39.
Sweeney,
Jillian
C.
and
Geoffrey
N.
Soutar
(2001),
“Consumer
Perceived
Value:
The
Development
of
a
Multiple
Item
Scale,”
Journal
of
Retailing,
77
(2),
203–20.
Thakor,
Mrugank
V. ,
Rajneesh
Suri
and
Katayoun
Saleh
(2008),
“Effects
of
Service
Setting
and
Other
Consumers’
Age
on
the
Service
Perceptions
of
Young
Consumers,”
Journal
of
Retailing,
84
(2),
137–49.
Thomas,
Jacquelyn
S.
and
Ursula
Y.
Sullivan
(2005),
“Managing
Marketing
Communications
with
Multichannel
Customers,”
Journal
of
Marketing,
69
(4),
239–51.
Trusov,
Michael,
Randolph
Bucklin
and
Koen
Pauwels
(2009),
“Effects
of
Word-
of-Mouth
Versus
Traditional
Marketing:
Findings
from
an
Internet
Social
Networking
Site,”
Journal
of
Marketing,
73
(September),90–102.
van
der
Lans,
Ralf,
Gerrit
Van
Bruggen,
Jehoshua
Eliashberg
and
Berend
Wierenga
(2010),
“A
Viral
Branching
Model
for
Predicting
the
Spread
of
Electronic
Word
of
Mouth,”
Marketing
Science,
29
(2),
348–65.
Van
Nierop,
Erjen,
Bart
Bronnenberg,
Richard
Paap,
Michel
Wedel
and
Philip
Hans
Franses
(2010),
“Retrieving
Unobserved
Consideration
Sets
From
Household
Panel
Data,”
Journal
of
Marketing
Research,
47
(1),
63–74.
Verhoef,
Peter
C.,
Scott
A.
Neslin
and
Bjorn
Vroomen
(2007),
“Multi-
channel
Customer
Management:
Understanding
the
Research
Shopper
Phenomenon,”
International
Journal
of
Research
in
Marketing,
24
(2),
129–48.
Verhoef,
Peter
C.,
Katherine
N.
Lemon,
A.
Parasuraman,
Anne
Roggeveen,
Michael
Tsiros
and
Leonard
A.
Schlesinger
(2009),
“Customer
Experience
Creation:
Determinants,
Dynamics
and
Management
Strategies,”
Journal
of
Retailing,
85
(1),
31–41.
Westbrook,
Robert
A.
and
Richard
L.
Oliver
(1991),
“The
Dimensionality
of
Consumption
Emotion
Patterns
and
Consumer
Satisfaction,”
Journal
of
Consumer
Research,
18,
84–91.
Wiesel,
Thorsten,
Koen
Pauwels
and
Joep
Arts
(2010),
“Marketing’s
Profit
Impact:
Quantifying
Online
and
Off-line
Funnel
Progression,”
Marketing
Science,
30
(4),
604–11.
Wind,
Yoram
and
David
Lerner
(1979),
“On
the
Measurement
of
Purchase
Data:
Surveys
Versus
Purchase
Diaries,”
Journal
of
Marketing
Research,
16
(February),
39–47.
Wooldridge,
Jeffrey
M.
(2009),
Introductory
Econometrics:
A
Modern
Approach,
4th
ed.
Boston:
South
Western
College.
Yaveroglu,
Idil
and
Naveen
Donthu
(2008),
“Advertising
Repetition
and
Place-
ment
Issues
in
On-line
Environments,”
Journal
of
Advertising,
37
(2),
31–44.
Zhang,
Jie
(2006),
“An
Integrated
Choice
Model
Incorporating
Alternative
Mechanisms
for
Consumers’
Reactions
to
In-Store
Display
and
Feature
Advertising,”
Marketing
Science,
25
(3),
278–90.