This document discusses identifying influential individuals for online word-of-mouth marketing campaigns. It addresses two main challenges: 1) Identifying seed influencers who belong in target geographic and demographic profiles. 2) Measuring the impact of campaigns continuously and in a timely manner. Several commonly used parameters are discussed for identifying influencers, including number of readers, links, social connections, and content sharing metrics. Methods for continuous monitoring and measurement are also presented.
2. 2
ABST
Word
Hum
In th
how
analy
comp
comp
Accu
such
are c
dem
beyo
parti
Mea
meas
metr
We d
TRACT
d of mouth (
(1) Identi
(2) Post‐e
an analysts a
is paper we
a software
ysis, demog
ponents requ
ponents and
urate quantif
as inlink co
commonly us
ographics of
ond simple c
icipating nod
suring of the
surable met
rics, it must m
c
s
m
n
c
discuss each
WOM) mark
fication of se
engagement
are common
tackle both
platform is
raphic segm
uired to buil
share our le
fication of in
ount, numbe
sed for this t
f matching a
counts, e.g.,
des, and topi
e impact of
rics for thes
meet several
omprehensiv
calable data
multilingual t
natural langu
ustomizable
of these attr
keters face tw
eed influenc
continuous
nly employed
problems ut
engineered
mentation, t
d a unified p
earnings from
nfluence is ce
r of commen
ask. The pro
author profi
while analy
c of the cont
a campaign
e tasks whic
l requiremen
ve coverage
mining
text analytics
age processi
reporting ca
ributes in de
wo main chal
ers who belo
monitoring o
d for these ta
tilizing inform
to assist m
ext analytic
platform to s
m building th
entral in sele
nts on a blog
oblem is furth
les. We exp
yzing inlink c
tent.
is important
ch we build
nts:
from multip
s
ing
apability
tail.
llenges:
ong in target
of impact of t
asks, making
mation techn
arketers. We
cs, linguistic
serve WOM m
e Sysomos' s
ection of see
g post, frien
her complica
lore all thes
count taking
t to establis
upon. For a
le sources
geographic
the campaig
WOM camp
nology and a
e recognize
s technolog
marketers. In
software plat
ed influence
d count in s
ated by inclu
se alternativ
g in account
h the ROI. T
software pl
and demogr
n.
aigns expens
lgorithmic te
influence q
gy, and mea
n this paper
tform.
rs for a cam
ocial networ
sion of restr
ves in detail
the networ
The commun
atform to b
aphic profile
sive and time
echniques de
uantification
asurable me
we discuss e
mpaign. Sever
rks, and num
ictions on ge
and discuss
k structure,
ity has prop
e able to m
e
e consuming
emonstrating
n, geographic
etrics as key
each of these
ral attribute
mber of digg
eography and
s ways to go
authority o
posed severa
easure these
g.
g
c
y
e
s
s
d
o
f
al
e
4. 4
IDEN
Ident
step
prob
onlin
lists c
In the
infor
The n
availa
from
blogs
what
tradi
(each
frequ
prop
"goo
The F
prese
NTIFYING IN
tification of in
for online WO
ably there is
ne influence. S
commonly us
e case of blog
mative factor
number of inl
able. While c
the structure
s and one wh
t Google does
tional web w
h representin
uency of post
agation mech
d" blogs high
Figure below
ence of a link
Blog
• Number
readers
• Number
inlinks
• Structure
inlink net
• Bookmar
on Delicio
• Total Digg
• Post freq
• Average n
of comm
post
NFLUENCERS
ndividuals wit
OM marketin
no single 'bes
Such paramet
sed measures
gs, the numbe
r. This numbe
links to a blog
ount of inlink
e of the inlink
ich is linked b
s with PageRa
here links are
g a single blo
ting by the blo
hanism in this
her than one s
shows a simp
between two
gs
of
of
e of
twork
rk count
ous
g count
uency
number
ents per
S
th influential
ng. What cons
st' definition,
ters also dep
s of influence
er of individu
er however is
g by other blo
ks provides so
k network to
by a few high
ank. This netw
e the only crit
og), are also a
ogger, bookm
s network mu
surrounded b
plified examp
o blogs. The c
Social N
• Number
friends
• Friend n
structur
• Profile p
count
online prese
stitutes online
there are a m
end on the pa
across multip
uals reading th
s not always p
ogs is the mos
ome informat
a blog. For ex
authority blo
work analysis
terion for aut
ssociated wit
mark counts,
ust take in acc
by many "aver
Figure 2 Differ
ple network w
color of circle
Networks
r of
network
re
page view
nce (possibly
e influence is
multitude of p
articular insta
ple Web 2.0 p
he posts of a
publicly disclo
st commonly
tion about inf
xample, ranki
ogs requires s
process how
thority and in
th additional
digg counts, a
count all thes
rage" blogs.
ent criterions of in
whereas each
e represents it
Micro
• Follow
• Update
freque
around the e
s an issue of o
parameters th
ance of online
platforms.
candidate inf
osed, and oth
used parame
fluence, there
ing a blog wh
sophisticated
wever is much
fluence. In th
information r
and writing st
se attributes t
nfluence.
circle represe
ts quality bas
oblogging
wer count
e
ency
entity of inter
ongoing discu
hat need to b
e forum cons
fluencer on a
er metrics are
eter since this
e is a lot more
ich is linked b
network ana
more compl
he blog world
regarding num
tyle of the blo
to rank a blog
ents a blog an
sed on all add
Vide
• View
• Ratin
• Num
comm
• Favor
rest) is an imp
ssion. Althou
be considered
idered. The t
daily basis is
e therefore u
s information
e that can be
by many low a
alysis similar i
icated than th
, nodes in the
mber of comm
ogger. The in
g surrounded
nd each edge
ditional attrib
eo Sharing
count
ng
ber of
ments
rite count
portant first
ugh most
d in assessing
able below
s the most
used instead.
n is easily
inferred
authority
n spirit to
hat on
e network
ments,
fluence
by few
e represents
utes, darker