Decision tree modelling of e-consumers’ preferences for internet marketing communication tools during browsing
Abstract
The successful development of internet marketing is based on scientifically proven decisions designed for the comprehensive analysis and evaluation of internet marketing communication tool selection. Different layers of internet marketing phenomena, such as communication tool profiles and characteristics of customers and strategies for different stages of purchase models, are widely analysed. However, it has been noted that modern management theories lack scientific research on the comprehensive analysis and evaluation of internet marketing communication tools, including the relevant characteristics of electronic consumers profiles based on their generational aspects and their life cycle stages. It is therefore necessary to analyse the stages of an electronic consumer’s journey and define the most relevant communication tools and application uses during every stage by aiming to improve customer satisfaction and marketing performance. The goal of this research is to determine the most significant internet marketing communication elements in the purchase phase of the electronic consumer journey cycle using the mathematical decision tree approach for different types of customers, using the generation theory as a segmentation tool. The literature analysis on electronic consumer’s behaviour, generation theory application possibilities in marketing and internet marketing communication tools was carried out. The research methodology includes eye-tracking and descriptive and comparative statistical analysis methods (decision tree models), which create the preconditions for the evaluation of electronic consumers’ explicit and tacit reactions to the use of internet marketing communication tools during the purchase phase of an electronic consumer´s journey. It was established that comparable statistically significant different preferences for internet marketing communication tools, at the purchase phase during a browsing task, exist for baby boomers, X, Y and Z generations.
Full text
206 2019, XXII, 1
In o ma ion Managemen
DOI: 10.15240/ ul/001/2019-1-014
In oduc ion
The apid de elopmen o in o ma ion
communica ion echnologies (ICT) has expanded
he possibili ies o ma ke ing communica ion.
In o de o inc ease business compe i i eness
and ca y ou e ec i e ma ke ing ac i i ies, i
has he e o e become impo an o acqui e
knowledge abou e-consume s and o iden i y
signi i can elemen s ha shape hei i ual
beha iou and in l uence hei decision o buy.
An analysis o scien i i c li e a u e e ealed
ha he e is a gap in knowledge wi h ega ds
o he e-consume beha iou o di e en
gene a ions, as cus ome segmen s, and
hei p e e ences in he pu chase phase. The
pu chase phase is cha ac e ised by di e en
in e ne ma ke ing communica ion elemen s,
which in l uence he pe o mance o b owsing
and sea ching asks. The goal o he esea ch
p esen ed in his a icle was o de e mine
he mos signi i can in e ne ma ke ing
communica ion elemen s du ing he pu chase
phase o he e-consume jou ney by pe o ming
a b owsing ask and using he ma hema ical
decision ee app oach.
The esea ch me hodology consis ed o
se e al s ages. Fi s , da a was collec ed using
eye- acking echnology in o de o iden i y he
in e ne ma ke ing communica ion ools ha a e
needed in o de o pe o m he b owsing ask
on a selec ed e-comme ce websi e. Second,
he mos impo an elemen s we e selec ed
and hei signi i cance es ablished using he
CHAID (Chi-squa ed Au oma ic In e ac ion
De ec o ) Decision T ee model. The model hen
c ea ed he p econdi ions o he e alua ion
o e-consume s’ explici and aci eac ions o
he use o in e ne ma ke ing communica ion
ools (IMCT) in he pu chase phase o he
e-consume jou ney.
The ollowing esea ch me hods we e
applied: ques ionnai e, in e iew, obse a ion
and expe imen (using biome ic echnologies,
namely, he eye- acking sys em (ETS)). ETS
is de i ned as a echnology ha allows he
measu emen o indi idual eye mo emen s o
ecei e in o ma ion abou he isual objec s
which a ac he indi idual o emo ions caused
by isual objec s; his is done by ollowing he
subjec ’s glance and de e mining he sequence
o i s mo emen om one loca ion o ano he .
1. Li e a u e Re iew
Consume beha iou in i ual spaces has been
he subjec o scien i i c esea ch o se e al
decades. Howe e , due o he dynamics and
apid ans o ma ion o he main s akeholde s
in ol ed in his i eld, such as consume s and
businesses, his opic equi es con inuous and
complex i eld esea ch (Ahmed e al., 2018;
Chung & Pa k, 2018; Dabija e al., 2018;
Sabai y e & Da ida ičiene, 2018; Raudeliūnienė
e al., 2018; D agos & D agos, 2017; Gup a e
al., 2018; Noba & Ros amzadeh, 2018; Vila &
Kus e , 2012; Voj odic e al., 2018).
Scien is s analyse he beha iou o
consume s in i ual spaces in di e en
ways: use -cen ic aspec s, such as issues
ega ding po en ial buye con e sions (Chen
& Cheng, 2013; Teo & Liu, 2007; Vo e al.,
2017; Palamido ska-S e jado ska & Ciuno a-
Shuleska, 2017); analyses o he ac o s
in l uencing consume beha iou in i ual
spaces (Dennis e al., 2009; Dębkowska,
2017; Kho akian & Jahangi , 2018); consume
pe cep ions abou he quali y o he e-comme ce
expe ience (C is obal e al., 2007); psychological
(Choi & Kwon, 2018; Fuchs e al., 2010; Yada
e al., 2013; S epaniuk, 2017) and demog aphic
cha ac e is ics (Sabai y e & Da ida ičius, 2017;
DECISION TREE MODELLING
OF E-CONSUMERS’ PREFERENCES
FOR INTERNET MARKETING
COMMUNICATION TOOLS DURING BROWSING
Jolan a Sabai y ė, Vida Da ida ičienė, Ja mila S ako á,
Ju gi a Raudeliūnienė
EM_1_2019.indd 206EM_1_2019.indd 206 8.3.2019 9:16:328.3.2019 9:16:32
207
1, XXII, 2019
In o ma ion Managemen
B own e al., 2003; Pa k & Yang 2017; Ma u o,
2018); isk and bene i pe cep ion (Huang e
al., 2004; Xue e al., 2017). O he esea che s
in es iga e echnological cha ac e is ics,
including he echnical speci i ca ions o
e-comme ce websi es (Huang & Benyouce ,
2013; Wang e al., 2016; Da ida ičienė &
Sabai y ė, 2014; Da ida iciene, Pabedinskai e,
& Da ida icius, 2017), such as paymen (Ming-
Yen Teoh e al., 2013; Pla eaux e al., 2014) and
usabili y aspec s (Diaz e al., 2017; Fu ne e al.,
2015; Wong e al., 2014). Howe e , he use , as
he consume , emains he essen ial esea ch
subjec . The consume ’s compu e skills and
willingness o use he oppo uni ies p o ided
o hem by i ual spaces depends on ac o s
such as he social, economic, echnological
and cul u al en i onmen (Yang & Jolly, 2008;
Cab e a To es, 2013; Fe nández-Du án, 2015;
Kho akian & Jahangi , 2018). The e-consume
should he e o e be analysed h ough he lens o
he heo y o gene a ions. This allows consume s
o be g ouped acco ding o hei yea o bi h and
dis inguishes he main cha ac e is ic ha de i nes
he use , namely age (Robe s & Manolis, 2000).
The concep o gene a ions has been
used in cul u es wo ldwide o many cen u ies
(Ke ze , 1983); in he mode n con ex , he social
aspec o his heo y is becoming inc easingly
impo an . In his esea ch, a gene a ion is
ea ed as a g oup o people bo n in a simila
yea and li ing in he same his o ical dimension
o he social p ocess. A i ual ma ke p o i le
can be de e mined o di e en gene a ions,
whe eby each gene a ion is dis inguished
by hose cha ac e is ics ha o m a dis inc
expe ience and lead o di e en expec a ions
and eac ions o ma ke ing communica ion ools
(Cab e a To es, 2013; Hi am & de Run, 2013;
Reisenwi z & Iye , 2007; Ša ánko á & Šikýř,
2017; Yang & Jolly, 2008). In his esea ch, he
heo y o gene a ions is used as a consume
segmen a ion ool.
In scien i i c li e a u e, he e a e di e en
app oaches o he classi i ca ion o gene a ions.
Howe and S auss (2007) make a dis inc ion
be ween 6 gene a ions o people in he Uni ed
S a es (The G ea es , Silen , Baby Boom, X,
Y and Z gene a ions), while o he scien is s
ocus on 4 gene a ions (Li e al., 2013):
silen , baby boom, and gene a ions X and Y.
Le ickai ė (2010) makes a dis inc ion be ween
3 gene a ions based on poli ical, social and
echnological changes (X, Y and Z). Malaysian
sociologis s, analysing he peculia i ies o
dis ance lea ning in ela ion o di e ences in
gene a ional beha iou s, make a dis inc ion
be ween adi ional X and Y gene a ions
(Ahmad & Ta mudi, 2012). Chi e al. (2013)
and Gu soy e al. (2013) analyse he beha iou
o 3 main gene a ions: baby boom, X and Y.
Based on an analysis o he esul s o scien i i c
esea ch, i was decided o use he ollowing
classi i ca ion o gene a ions: baby boome s
(bo n 1941-1961); gene a ion X (bo n 1962-
1982); gene a ion Y (bo n 1983-1997); and
gene a ion Z (bo n 1998-2018).
Following consume segmen a ion on he
basis o gene a ional heo y, i is impo an o
de e mine he cha ac e is ics o he pu chase
phase on an e-comme ce websi e. On an
e-comme ce websi e, use s pe o m wo main
asks, namely b owsing and sea ching (Ca mel
e al., 1992; Hong e al., 2004; Nielsen & Pe nice,
2013; Ma u o & Di Ba is a, 2018), in o de
o pu chase. Wi hin he con ex o consume
beha iou , he di e en pa ame e s o he
elemen s o in e ne ma ke ing communica ion
a e impo an o hese asks. An e-comme ce
websi e, as he main ool in he pu chase
phase, p esen s an abundance o elemen s
suppo ing he impo an pu chase p ocess.
These a e elemen s ha gi e consume s
a sense o secu i y (gua an ee o secu i y, a e -
sales se ice, he exis ence o a physical shop
o home add ess, ce i i ca es o quali y, e u n
condi ions, he oppo uni y o pay a e eceip
o he goods), an o e iew o he cha ac e is ics
o he p oduc s (p ice; abundan asso men ;
a ich ange o de ailed in o ma ion including
ex , images, audio, ideo, mixed media;
p oduc deli e y cos s; jus i i ca ion o physical
p oduc exis ence e.g. showing he quan i y
in s ock), and unc ional possibili ies ( a ious
paymen me hods; egis a ion; isibili y o he
shopping ca ; compa ison o goods; na iga ion
elemen s such as uppe , side menu, and
sea ch engine; and con ac elemen s such
as commen s, e-help, con ac o m, phone
numbe , and mobile suppo solu ions).
2. Resea ch Design and Me hods
To ob ain he mos objec i e in o ma ion
abou he beha iou o di e en gene a ions
in i ual spaces, and aking in o conside a ion
he speci i cs o his esea ch subjec , a mixed
esea ch me hod was applied, whe eby eye-
acking echnology was used in combina ion
EM_1_2019.indd 207EM_1_2019.indd 207 8.3.2019 9:16:328.3.2019 9:16:32
208 2019, XXII, 1
In o ma ion Managemen
wi h a compa a i e s a is ical s udy using
a decision ee model.
The pu pose o his esea ch was o
de e mine which s a is ically signi i can
in e ne ma ke ing communica ion elemen s
cus ome s p e e , as de i ned by di e en
gene a ional coho s, du ing he elec onic
pu chase phase using he decision ee model.
The esea ch asks we e: (1) o de e mine
he ac ual esponse o he baby boom, X, Y
and Z gene a ions o he mos app op ia e
communica ion elemen s du ing he elec onic
pu chase phase while pe o ming a b owsing
ask; (2) o e alua e he non-exp essed ac ual
eac ion o he di e en gene a ions o he
IMCT elemen s du ing he elec onic pu chase
phase while pe o ming a b owsing ask and o
de e mine which communica ion elemen s a e
s a is ically signi i can .
The pe iod o esea ch was No embe
2015 – May 2016. The expe imen was
conduc ed anonymously; all he pa icipan s
we e olun ee s who ag eed o being es ed,
namely h ough eye- acking o hei beha iou .
Consen om all pa icipan s was ecei ed
e bally.
To de e mine he numbe o pa icipan s,
he ecommenda ions o in e na ional expe s,
he Nielsen No man G oup and Tobii we e
used. S and all (2009), om he eye- acking
company Tobii, ecommends analysing
he beha iou o 20 o 50 indi iduals o
quan i a i e esea ch ega ding indi iduals’
beha iou , pa icula ly when he esul s a e
analysed alongside hose o o he esea ch
esul s. A simila posi ion is held by Pe nice and
Nielsen (2009). They ecommend analysing
he beha iou o 20 consume s o quali a i e
consume esea ch. Since he quan i a i e
esea ch o de e mine popula ion beha iou is
an addi ional in es iga ion o deepening he
unde s anding o indi iduals´ beha iou du ing
he pu chase p ocess, 27 indi iduals om each
gene a ion we e in i ed o ake pa , he eby
bea ing in mind he isk ha he da a om e e y
indi idual may no be sui able. A e p ocessing
he esea ch da a (108 indi iduals pa icipa ed
in o al) and ejec ing esponden da a ha
was unsui able, he beha iou o 76 indi iduals
on he selec ed e-comme ce websi e was
analysed (19 om he Z gene a ion, 22 om he
Y gene a ion, 17 om he X gene a ion, and 18
om he baby boom gene a ion).
Du ing he expe imen , da a desc ibing he
ajec o y o he subjec s’ gaze ( he loca ion o
he elemen s and he o de o choice) and he
in ensi y o usage du ing he pu chase phase
(gaze leng h in ela ion o he pa icula elemen )
we e eco ded. A emo e eye- acking de ice,
Mi ame ix S2, was used o he expe imen .
To conduc his s udy, a b owsing ask
was chosen o analysis. The beha iou s o
he use s in he di e en age g oups we e
analysed h oughou he expe imen , in
pa icula wi h ega ds o he elemen s o he
e-comme ce websi e esponden s chose o use
o pe o m he ask. The chosen websi e o his
esea ch was local and had he highes a ing
o an in e ne shop ope a ing in Li huania
(Alexa.com). In o de o analyse he pu chase
p ocess beha iou o he di e en age g oups
while b owsing, he selec ed consume p oduc
had o be well-known o all he age g oups.
The chosen p oduc was he Samsung Galaxy
S6 mobile phone, which was he bes - a ed
high-end phone in Li huania a he ime o his
esea ch (Mu in, 2015; Ma u o, 2017).
When de i ning he IMCT p e e ences o
use s o di e en gene a ions, i is easonable
o de i ne he a eas o in e es o he
e-comme ce websi es unde analysis based on
he ecommenda ions o Pu ucke e al. (2013),
Chandon e al. (2009), and B asel and Gips
(2008) (see Fig. 1). Because di e en goals a e
held o he di e en pages o he e-comme ce
websi e, he pages ha e di e en unc ionali y
and a eas o in e es . The a eas o in e es a e
he e o e dis inc o he h ee e-comme ce
pages: he homepage, sea ch esul s and
p oduc .
Du ing he s udy, he esea che main ained
hei dis ance om he esponden so as no o
encou age unnecessa y alk du ing he es .
Responden s we e asked o no hink aloud,
as his could nega i ely a ec he esul s o he
s udy and encou age ac ions o be pe o med
as e han he esponden would ha e done
o he wise. Du ing he s udy, hose esponden s
ha mo ed in ensi ely we e calib a ed se e al
imes and we e old ha his was ou ine p ac ice
so as o minimise addi ional emo ions ha could
nega i ely a ec he esea ch. All he asks
we e s a ed aloud; he pu pose o he ask was
clea ly iden i i ed, and checks we e p e o med
o ensu e he esponden unde s ood wha o
do. Responden s we e in o med when hey
could begin each ask and wha ac ion signalled
he end o each ask (when he da a eco ding
EM_1_2019.indd 208EM_1_2019.indd 208 8.3.2019 9:16:328.3.2019 9:16:32
209
1, XXII, 2019
In o ma ion Managemen
was i nished). A e pe o ming all o he asks, i
he gaze o he esponden was being ollowed,
he esponden was in o med abou his ac
acco dingly. In be ween esea ch sessions,
da a was clea ed and he da a ha was le on
he in e ne b owse was dele ed.
Du ing his esea ch, he main eye- acking
me ics ( i xa ion and saccades) we e eco ded
along wi h o he me ics: he scanned a ea,
blinking a e and he size o he pupil. Each o
he me ics has i s own subca ego y: i xa ion
me ics ( o al i xa ion numbe ), i xa ion numbe
in he egion o in e es (he eina e ROI),
numbe o i xa ions in he ROI depending on
he leng h o he ex , du a ion o he i xa ion,
gaze ( o al i xa ion du a ion), he spa ial densi y
o he i xa ion, epea ed i xa ions, exac ime o
he i s i xa ion, pe cen age o he pa icipan s,
measu ed ROI indica o (all a ge i xa ions),
saccade me ics (numbe , ampli ude o he
saccades, eg essing saccades, saccades
ha e eal ma ked pu pose ul shi s), me ics
o he scanned pa s (du a ion o he scanned
pa , leng h o he scanned pa , densi y o he
scanned pa ), ansi ion ma ix, egula i y o
he scanned pa , olume o space calcula ed
No.
in lis
Sub-ca ego y
(elemen s)
Ca ego y
(In acco dance wi h In e ne
Ma ke ing Communica ion
Tools)
Ac onym
7 Uppe sea ch engine box Web sea ch engine IPS1
23 Lowe sea ch engine box IPS2
8 Con ac us o m Con ac o m SF
9 Phone numbe Ins an Messaging MZ
14a P oduc pho o
E-comme ce websi e
TEKS1
14b Discoun TEKS2
14c P ice TEKS3
14d Physical e idence elemen TEKS4
14e The name o he p oduc TEKS5
14 Call o ac ion - buy bu on TEKS6
14g Tex cha ac e is ics TEKS7
14h Compa ison elemen TEKS8
14i Video in o ma ion TEKS9
14j Link o he manu ac u e ’s page TEKS10
14k Wish lis TEKS11
22a Pho o (le p oduc elemen ) TEKS12
22b P ice (le p oduc elemen ) TEKS13
22c Call o ac ion - buy bu on (le p oduc
elemen ) TEKS14
22d Ti le o he p oduc (le p oduc elemen ) TEKS15
16 Newsle e E-mail EP
18 Re iews Commen s KM
20 FAQ FAQ DUK
21 Blog Blog TNK
25 Social ne wo ks Social ne wo ks SM
Sou ce: own
Tab. 1: Ac onyms o e-comme ce websi e elemen s
EM_1_2019.indd 209EM_1_2019.indd 209 8.3.2019 9:16:328.3.2019 9:16:32
210 2019, XXII, 1
In o ma ion Managemen
Fig. 1: A eas o in e es on he selec ed local e-comme ce websi e
Sou ce: au ho s
EM_1_2019.indd 210EM_1_2019.indd 210 8.3.2019 9:16:328.3.2019 9:16:32
211
1, XXII, 2019
In o ma ion Managemen
on he basis o he con ex hull wid h, di ec ion
o he scanned pa , equency o he saccades
and i xa ions.
The da a gene a ed du ing he eye- acking
esea ch we e sa ed as da a ables (.xls) and
ideo ma e ial (.a i). In o de o achie e he
pu pose o he s udy, he in o ma ion on sigh
i xa ion was analysed and linked o he a ea o
in e es . The in o ma ion on sigh i xa ion and
du a ion was p esen ed in da a ables (.xls),
and he de e mina ion o he a ea o in e es
was done by analysing he eye mo emen
ajec o y and sigh i xa ion in o ma ion om he
ideo ma e ial ( he pe o mance sequence o
he asks was eco ded du ing he expe imen ).
A se o elemen s was c ea ed o which he
i xa ion o sigh was no less han 180 ms. The
elemen s om he gene al lis we e, due o hei
speci i c ea u es, subsequen ly assigned as
means o in e ne ma ke ing communica ion.
Ac onyms we e gi en o hese elemen s (see
Tab. 1) and ano he da a se up p epa ed wi h
SPSS IBM S a is ics so wa e o u he wo k.
Conside ing he pu pose o his s udy is
o iden i y he p e e ed elemen s o online
ma ke ing a he pu chase phase, decision
ee models (DTMs) we e applied. DTMs
we e no only chosen because o hei abili y
o classi y elemen s by g oup, bu also in
his pa icula case o explain he consume
choices wi h espec o IMCT p e e ences.
The e a e a numbe o known DTM me hods
(CTR, QUEST, CHAID (Exhaus i e CHAID)
(Kass, 1980)), om which he DTM CHAID
(Chi-squa ed Au oma ic In e ac ion De ec ion)
was chosen. The CHAID me hod was selec ed
on he basis o he speci i cs o he dependen
a iable (gene a ion – ca ego ical a iable). In
addi ion, he choice was based on he abili y
o he model o classi y he a ailable da a by
g oup and i s abili y o o ecas he dependen
a iables (wi hin he con ex o his s udy, he
beha iou o di e en gene a ions) acco ding o
known independen a iables (in e ms o IMCT
p e e ences).
The CHAID me hod is based on a pa icula
algo i hm ha includes 3 s eps: g ouping,
sepa a ion o he elemen s and suspension.
The decision ee g ows by epea edly applying
he algo i hm, s a ing wi h he oo node and
i nishing wi h he analysis o all he da a ha
is being used. A classic s a is ical c i e ion,
Chi-squa e (x2), is used in he CHAID DTM.
The algo i hm de e mines he mos app op ia e
sepa a ion o each independen a iable (wi hin
he con ex o his s udy – IMCT) and con inues
o selec he dependen a iable (gene a ion),
he sepa a ion o which has a s a is ically
signi i can di e ence ( he lowes alue o p in
he chi-squa e signi i cance de e mina ion es ),
which allows he de e mina ion o s a is ically
signi i can IMCT p e e ences du ing he s udy.
Unde his me hod, each s ep iden i i es he
s onges in e ac ion be ween he ele an
gene a ion and he ele an online ma ke ing
communica ion ools. Whe e he in l uence on
he gene a ions is only sligh o he di e en
IMCT elemen s, hey a e combined.
The DTM CHAID me hod, based on Chi-
squa e, enabled he classi i ca ion o he
elemen s and he p edic ion o he IMCT
p e e ences; he isk o an inco ec classi i ca ion
was also e alua ed. G aphically displaying
he esul s o he analysis as a decision ee
enables he hie a chical dependency o he
a iables o be se .
The me hod ha was applied made i
possible o dis inguish be ween hose IMCT
elemen s ha a e s a is ically signi i can
o di e en asks and hose elemen s
which de e mine an inc ease in consume
engagemen in he pu chase p ocess. Since
a la ge amoun o da a was collec ed du ing
he s udy, he model allowed he selec ion o
a la ge g oup o independen a iables o ha e
only a s a is ically signi i can in l uence on he
p edic ion o gene a ional beha iou du ing he
pu chase phase in a di e en ask en i onmen
( his esea ch ocuses on a local shop). In
addi ion, he model enabled he iden i i ca ion o
in e ac ions be ween di e en use s o di e en
ages. The decision ee model o IMCT
p e e ences was compiled using IBM SPSS
S a is ics so wa e e sion 20.
The model was cons uc ed using a da a se
ob ained om he eye- acking s udy. The da a
was ca ego ised and eo ganised acco ding
o 24 sub-ca ego ies (independen a iables –
see Tab. 1). The ca ego y “Gene a ion” [“Ka a”]
(gene a ion Z, gene a ion Y, gene a ion
X and BB gene a ion) was chosen as he
dependen a iable o he cons uc ed model.
24 independen a iables we e di ided in o 9
la ge ca ego ies. The independen a iables
we e measu ed on a nominal scale (1 – he
a e age elemen i xa ion is o e 180 ms, 0 – he
a e age elemen i xa ion is less han 180 ms).
The echnical condi ions o he applica ion
EM_1_2019.indd 211EM_1_2019.indd 211 8.3.2019 9:16:338.3.2019 9:16:33
212 2019, XXII, 1
In o ma ion Managemen
o he model included a minimum p alue o 0.05
o he indi idual nodes and o he combined
ca ego ies. The signi i cance alues we e
de e mined in acco dance wi h he Bon e oni
me hod, which speci i es he p alue by
sepa a ing and combining he ca ego ies ( i s
by checking he alue in Chi-squa e and hen
adjus ing i acco ding o Bon e oni’s alue):
(1)
whe e: B – Bon e oni indica o o a ca ego ical
a iable; I – o al numbe o independen
a iable ca ego ies; – numbe o s a is ically
signi i can independen a iables a e me ging
he ca ego ies; – numbe o ca ego ies.
The cha ac e is ics o c ea ing a decision
ee we e subsequen ly iden i i ed. The DTM
g ow h limi s we e chosen: he maximum DTM
dep h was 3 le els, which indica es he numbe
o DTM le els below he oo node; he minimum
numbe o cases o oo nodes was 10 cases,
wi h 5 cases o b anch nodes. Cases we e
selec ed based on a ela i ely small numbe
o da a i le obse a ions. The con i dence
in e al, 2x s a is ics and o he pa ame e s
emained se (as ecommended by Pukėnas
(2009)). As he decision ee g ow h algo i hm
includes all a iables (ca ego ised o anked),
s anda disa ion was no equi ed. Acco ding o
he CHAID me hod, independen in e al scale
a iables a e di ided in o disc e e g oups. The
speci i ed numbe o such g oups is 10. In se ing
he pa ame e s, he misclassi i ca ion cos s, i.e.
he ela i e indica ed cos o he anking scale o
he dependen a iable o he misclassi i ca ion,
is equal ac oss all ca ego ies. The a iables
chosen o he decision ee ou pu a e Gain
and Index. The nominal and ank scales o
he dependen a iable Gain is de i ned as he
pe cen age o he selec ed ca ego y da a in
he decision ee node in ela ion o all he da a
in he selec ed ca ego ies. The nominal and
ank scales o he dependen a iable Index
is de i ned as he a io o he pe cen age o
selec ed ca ego y esponses o he solu ion in
Aspec s Cha ac e is ic Model Summa y
Speci i ca ions G owing Me hod CHAID
Dependen Va iables Gene a ion
Independen Va iables ( he explana ion
o he ac onyms o independen a iables
is gi en in Table 1)
IPS1, IPS2, SF, MZ, TEKS1, TEKS2,
TEKS3, TEKS4, TEKS5, TEKS6, TEKS7,
TEKS8, TEKS9, TEKS10, TEKS11,
TEKS12, TEKS13, TEKS14, TEKS15, EP,
KM, DUK, TNK, SM
Valida ion C oss Valida ion
Maximum T ee Dep h 3
Minimum Cases in Pa en Node 10
Minimum Cases in Child Node 5
Resul s Independen Va iables Included TEKS4, MZ, IPS1, DUK
Numbe o Nodes 9
Numbe o Te minal Nodes 5
Dep h 3
Sou ce: au ho s
Tab. 2:
Summa y o CHAID model speci i ca ions / Resul s o di e en coho s’ p e e-
ences o IMCT elemen s while pe o ming a b owsing ask on he selec ed
local e-comme ce websi e
EM_1_2019.indd 212EM_1_2019.indd 212 8.3.2019 9:16:338.3.2019 9:16:33
213
1, XXII, 2019
In o ma ion Managemen
he decision ee node o he pe cen age o he
esponses o he whole sample.
3. Explo a o y Analysis o Da a and
Discussion
Tab. 2 p o ides gene al in o ma ion on he
compila ion o he DTM and consis s o wo
pa s: p o isions o he o ma ion o he DTM
(in o ma ion on he da a used o cons uc he
model) and he esul s ob ained (in o ma ion
abou he DTM). The cha ac e is ics column
p o ides in o ma ion abou he dependen
a iable Gene a ion [Ka a] and independen
a iables ha we e used o compile he
speci i c DTM. The independen a iables
ha a e s a is ically signi i can o he DTM
a e highligh ed in bold: TEKS4 (wi hin he
con ex o his s udy – physical e idence), MZ
(ins an messages), IPS1 (uppe sea ch ba ),
DUK ( equen ly asked ques ions). The model
consis s o 9 nodes (wi hin he con ex o his
s udy hey ep esen he IMCT p e e ences), o
which 5 a e endpoin s. The decision ee dep h
is 3 nodes below he oo node, which enables
he p e e ence o IMCT elemen s o be se
om a hie a chical poin o iew.
A g aphical p esen a ion o he DTM o he
b owsing analysis on he local e-comme ce
websi e is p esen ed in Fig. 2. The esul s e eal
ha he independen a iable TEKS4 (wi hin
he con ex o his s udy – physical e idence)
has he mos in l uence on he dependen
a iable Gene a ion [Ka a]. The igh side
o he compila ion model makes i possible
o dis inguish he impo ance o his elemen
o he ep esen a i es o he Z, Y and baby
boom gene a ions. The impo ance o physical
e idence in he pu chasing phase demons a es
he awa eness o consume s and he need o
secu i y, which applies o bo h e-comme ce
websi es and he adi ional en i onmen and
is exp essed h ough he physical e idence
elemen s on he websi e. Ano he independen
a iable wi h a s a is ically signi i can in l uence
on he dependen a iable Gene a ion, is IPS1
(uppe sea ch ba ). The s a is ical signi i cance
o his elemen shows ha i is impo an , once
again, o he ep esen a i es o he Z, Y and
baby boom gene a ions. The impo ance o he
elemen is explained h ough he peculia i ies
o na iga ion – consume s ha belong o hese
gene a ions end o make su e o he physical
p esence o he e-supplie , and hen con inue
hei p oduc sea ch by en e ing keywo ds in
he sea ch ba . Fo 20 pe cen o gene a ion
Y and 80 pe cen o baby boome s, o whom
he elemen o physical e idence and he uppe
sea ch ba a e impo an , he e is ano he
elemen ha is impo an , namely F equen ly
Asked Ques ions (independen a iable DUK).
The indica ed p e e ences make i possible
o no ice he inclina ion o he baby boom
gene a ion owa ds web 1.0 ea u es (FAQ as
a s a ic in o ma ion p esen a ion ool).
The le side o he analysed DTM enables
he iden i i ca ion o ano he s a is ically
signi i can independen a iable MZ (wi hin he
con ex o his esea ch – ins an messaging).
When pe o ming a b owsing ask, he
signi i cance o his elemen was no ed by 84.6%
o gene a ion Z, 7.7% o gene a ion X and 7.7%
o he baby boom gene a ion. The independen
a iable in he end o DTM node indica es he
s ong inclina ion o he Z, X and baby boom
gene a ions o suppo he communica ion
p ocess h ough his mobile channel.
The c ea ed DTM o he pu chase phase on
he selec ed local e-comme ce websi e enabled
he iden i i ca ion o he s a is ically signi i can
p e e ences o online ma ke ing communica ion
ools o he di e en age g oups. Howe e , in
o de o de e mine he possibili ies o applying
a decision ee model solu ion, i is necessa y
o assess how in o ma i e and app op ia e he
applica ion o a pa icula DTM is. In assessing
he applicabili y o he model o he p edic ion o
each gene a ion’s beha iou in i ual spaces,
he amoun o da a used o he co esponding
gene a ion o each o he iden i i ed s a is ically
signi i can p e e ences was e alua ed
(p e e en ial models a e exp essed h ough
nodes in he g aphical ep esen a ion o he
model) and he eligibili y o he p e e ences
jus i i ed on he basis o he pe cen ile g aphs
o he dependence o he Gain and Index
a iables. The analysis o he isk assessmen
o he applica ion o he cons uc ed DTM was
pe o med by assessing he abili y o p edic he
beha iou o he gene a ions in he i ual space
h ough he co ec alloca ion o p e e en ial
IMCT elemen s o each gene a ion.
Jus i i ca ion o he sui abili y o he DTM o
he beha iou al o ecas o gene a ion Z o
he comple ed pu chase phase ollows. When
analysing he DTM o he consume eligibili y
o IMCT p e e ences o gene a ion Z, he da a
ep esen ing he beha iou o his gene a ion
and hei dis ibu ion was analysed o he
EM_1_2019.indd 213EM_1_2019.indd 213 8.3.2019 9:16:338.3.2019 9:16:33
214 2019, XXII, 1
In o ma ion Managemen
Fig. 2: DTM o di e en coho s’ p e e ences o IMCT while pe o ming a b owsing
ask on he selec ed local e-comme ce websi e
Sou ce: au ho s
EM_1_2019.indd 214EM_1_2019.indd 214 8.3.2019 9:16:338.3.2019 9:16:33
221
1, XXII, 2019
In o ma ion Managemen
Abs ac
DECISION TREE MODELLING OF E-CONSUMERS’ PREFERENCES
FOR INTERNET MARKETING COMMUNICATION TOOLS DURING BROWSING
Jolan a Sabai y ė, Vida Da ida ičienė, Ja mila S ako á,
Ju gi a Raudeliūnienė
The success ul de elopmen o in e ne ma ke ing is based on scien i i cally p o en decisions
designed o he comp ehensi e analysis and e alua ion o in e ne ma ke ing communica ion
ool selec ion. Di e en laye s o in e ne ma ke ing phenomena, such as communica ion ool
p o i les and cha ac e is ics o cus ome s and s a egies o di e en s ages o pu chase models,
a e widely analysed. Howe e , i has been no ed ha mode n managemen heo ies lack scien i i c
esea ch on he comp ehensi e analysis and e alua ion o in e ne ma ke ing communica ion ools,
including he ele an cha ac e is ics o elec onic consume s p o i les based on hei gene a ional
aspec s and hei li e cycle s ages. I is he e o e necessa y o analyse he s ages o an elec onic
consume ’s jou ney and de i ne he mos ele an communica ion ools and applica ion uses du ing
e e y s age by aiming o imp o e cus ome sa is ac ion and ma ke ing pe o mance. The goal o
his esea ch is o de e mine he mos signi i can in e ne ma ke ing communica ion elemen s in
he pu chase phase o he elec onic consume jou ney cycle using he ma hema ical decision ee
app oach o di e en ypes o cus ome s, using he gene a ion heo y as a segmen a ion ool. The
li e a u e analysis on elec onic consume ’s beha iou , gene a ion heo y applica ion possibili ies in
ma ke ing and in e ne ma ke ing communica ion ools was ca ied ou . The esea ch me hodology
includes eye- acking and desc ip i e and compa a i e s a is ical analysis me hods (decision ee
models), which c ea e he p econdi ions o he e alua ion o elec onic consume s’ explici and aci
eac ions o he use o in e ne ma ke ing communica ion ools du ing he pu chase phase o an
elec onic consume ´s jou ney. I was es ablished ha compa able s a is ically signi i can di e en
p e e ences o in e ne ma ke ing communica ion ools, a he pu chase phase du ing a b owsing
ask, exis o baby boome s, X, Y and Z gene a ions.
Key Wo ds: In e ne ma ke ing, communica ion, cus ome beha iou , in e ne ma ke ing
communica ion ool, e-comme ce.
JEL Classi i ca ion: M15, M31.
DOI: 10.15240/ ul/001/2019-1-014
EM_1_2019.indd 221EM_1_2019.indd 221 8.3.2019 9:16:358.3.2019 9:16:35