Aspect-based currency of customer reviews: A novel probability-based metric to pave the way for data quality-aware decision-making

Hägele, Lukas; Klie , Ma hias; Moes ue, La s; Obe meie , And eas
A icle — Published Ve sion
Aspec -based cu ency o cus ome e iews: A no el
p obabili y-based me ic o pa e he way o da a quali y-
awa e decision-making
Elec onic Ma ke s
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Hägele, Lukas; Klie , Ma hias; Moes ue, La s; Obe meie , And eas (2025) : Aspec -
based cu ency o cus ome e iews: A no el p obabili y-based me ic o pa e he way o da a
quali y-awa e decision-making, Elec onic Ma ke s, ISSN 1422-8890, Sp inge , Be lin, Heidelbe g,
Vol. 35, Iss. 1,
h ps://doi.o g/10.1007/s12525-025-00760-4
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/318880
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
Vol.:(0123456789)
Elec onic Ma ke s (2025) 35:10
h ps://doi.o g/10.1007/s12525-025-00760-4
RESEARCH PAPER
Aspec ‑based cu ency o cus ome e iews: Ano el p obabili y‑based
me ic opa e heway o da a quali y‑awa e decision‑making
LukasHägele1· Ma hiasKlie 1 · La sMoes ue1· And easObe meie 1
Recei ed: 23 May 2024 / Accep ed: 17 Janua y 2025
© The Au ho (s) 2025
Abs ac
Cus ome e iews om digi al pla o ms a e a i al da a esou ce o ecommende and o he decision suppo sys ems. The
pe o mance o hese sys ems is highly dependen on he quali y o he unde lying da a—pa icula ly i s cu ency. Exis ing
me ics o assessing he cu ency o cus ome e iews a e o en based solely on da a age. They do no conside ha cus ome
e iews can be ou da ed wi h espec o one aspec (e.g., gues oom a e eno a ion) while s ill being up- o-da e wi h espec
o o he s (e.g., loca ion). Mo eo e , hey dis ega d ha cus ome e iews can only become ou da ed due o s a e changes o
he co esponding i em (e.g., eno a ion), which a e associa ed wi h unce ain y. We p opose a p obabili y-based me ic o
he aspec -based cu ency o cus ome e iews. The alues o he me ic ep esen he p obabili y ha in o ma ion in a se
o cus ome e iews is s ill up- o-da e. Ou e alua ion on a la ge T ipAd iso da ase shows ha he alues o he me ic a e
eliable and disc imina e well be ween up- o-da e and ou da ed da a, pa ing he way o da a quali y-awa e decision-making
based on cus ome e iews.
Keywo ds Da a quali y· Cu ency· Cus ome e iews· Da a quali y me ic
JEL Classi ica ion M10
In oduc ion
Today, a la ge and g owing numbe o cus ome e iews a e
a ailable o all kinds o p oduc s and se ices (i.e., i ems)
on many di e en digi al pla o ms (Yin e al., 2014). Fo
example, he a el-based pla o m T ipAd iso boas s o e
one billion cus ome e iews co e ing mo e han eigh mil-
lion es au an s, ho els, a ac ions, and expe iences (T i-
pAd iso , 2022). This weal h o in o ma ion, eedback
di ec ly om cus ome s who ha e expe ienced hese i ems
(B and e al., 2022; Chen, 2023), makes cus ome e iews
one o he mos impo an da a sou ces in e-comme ce
(Hung e al., 2024). This is especially ue o da a-d i en
decision-making, o example, suppo ed by decision sup-
po sys ems o da a analy ics me hods (Sun e al., 2022;
Sysko-Romańczuk e al., 2022). Howe e , all hese applica-
ions equi e high-quali y da a in o de o p o ide iable
esul s (Elgendy e al., 2022; Hein ich e al., 2021). In pa -
icula , nume ous s udies unde sco e he impo ance o da a
cu ency (i.e., whe he da a s ill co esponds o i s coun-
e pa s in he eal wo ld) o he pe o mance o decision
suppo sys ems and da a analy ics me hods (Ab aham e al.,
2023; Hägele e al., 2024; Hols ein e al., 2023; H is o a,
2014). Consequen ly, cus ome e iews need o be up- o-da e
o s eng hen decision-making quali y (Hägele e al., 2024;
Hein ich & H is o a, 2016; McKinney e al., 2002). How-
e e , he c ea ion o cus ome e iews is la gely unmoni o ed
(i.e., wi h li le o no quali y o e sigh ), allowing anyone
o con ibu e wi hou majo ba ie s (Dha & Bose, 2022).
Fu he mo e, he c ea o s o cus ome e iews o en ail
o upda e hei e iews (Fi che & Hoogendoo n, 2019).
Responsible Edi o : Judi h Gebaue .
* Ma hias Klie
[email p o ec ed]
Lukas Hägele
[email p o ec ed]
La s Moes ue
[email p o ec ed]
And eas Obe meie
[email p o ec ed]
1 Ins i u e o Business Analy ics, Uni e si y o Ulm,
Helmhol zs . 22, 89081Ulm, Ge many
Elec onic Ma ke s (2025) 35:10 10 Page 2 o 18
This may be due o a lack o mo i a ion on he pa o he
au ho s o a lack o unc ionali y o he e iew pla o m o
upda e a e iew (Jin e al., 2014). Ano he eason may be
ha he au ho s do no ecognize when hei e iews become
ou da ed (e.g., i a ho el eno a es i s ooms, a pas gues
would only become awa e o his when e isi ing he ho el)
(Yakubu & Kwong, 2021). The e o e, on he one hand,
main aining a high le el o da a quali y and especially cu -
ency in he con ex o cus ome e iews is e y challenging.
On he o he hand, a well- ounded me hod o assessing he
cu ency o cus ome e iews is needed as he basis o any
a ge ed da a quali y imp o emen e o (Hein ich & Klie ,
2015; Hein ich e al., 2018).
To add ess his challenge, he assessmen o cu ency can
d aw on he ac ha cus ome e iews can only become
ou da ed i he s a e o he co esponding i em in he eal
wo ld changes. Fo example, a ho el ha eno a es i s gues
ooms changes i s s a e, and cus ome e iews ha desc ibe
he ooms as old and un-down (which was ue when he
e iews we e c ea ed) become ou da ed. This example also
highligh s ha s a e changes can be a ibu ed o ce ain
aspec s o he i em (e.g., he aspec gues ooms in he case
o a eno a ion). In his sense, cus ome e iews may be
up- o-da e wi h espec o ce ain aspec s, while being ou -
da ed wi h espec o o he s (e.g., he eno a ion may change
he s a e o he gues ooms, while no a ec ing he s a e
o o he aspec s such as ood o loca ion). In he ollow-
ing, we e e o such changes as aspec -based s a e changes,
because hey change he s a e o he co esponding i em in
a pa icula aspec such ha he in o ma ion associa ed wi h
ha aspec in he cus ome e iews is ou da ed. Ye , exis ing
app oaches o assessing he cu ency o cus ome e iews
neglec (aspec -based) s a e changes and use only simple
ea u es such as he ime since a cus ome e iew was c e-
a ed (i.e., hei “age”). While, in gene al, he age o da a
can indeed in luence he pe o mance o da a-d i en models
and he esul ing decisions (Laza idou e al., 2021; Raza
and Ding, 2022), indings ega ding cus ome e iew-based
decision suppo sys ems, such as ecommende sys ems,
a e mixed. Some s udies indica e ha ecommende sys em
pe o mance imp o es when ained on ecen cus ome
e iews only (Ve ach e e al., 2022), while o he s udies
sugges ha e en olde e iews can enhance ecommende
sys em pe o mance (Zheng & Ip, 2013). These con adic-
o y esul s highligh ha he ime since c ea ion alone is
no a su icien indica o o he ue cu ency o cus ome
e iews. In pa icula , e iews wi h a low age migh be
assumed o be up- o-da e, ega dless o whe he he e has
been a espec i e s a e change. Vice e sa, olde e iews
migh be conside ed ou da ed e en i no such s a e change
has occu ed. In addi ion, basing he assessmen o cu ency
on age alone does no allow o a ine-g ained iden i ica ion
o ac ually ou da ed o up- o-da e passages (i.e., a whole
cus ome e iew is decla ed ou da ed e en i only indi idual
aspec s a e ou da ed). Fo example, a ho el e iew may s ill
be up- o-da e conce ning he loca ion bu ou da ed ega ding
oom quali y.
The e o e, i is impo an o assess he cu ency o cus-
ome e iews based on aspec -based s a e changes. Indeed,
i has been shown ha conside ing he cu ency o cus ome
e iews in an aspec -based manne can lead o signi ican ly
be e decision-making quali y (Hägele e al., 2024). In p ac-
ice, no assessing (aspec -based) cu ency is p oblema ic
in many espec s. Fo example, a cus ome who p io i izes
oom quali y migh no ecei e a ecommenda ion o a
ecen ly eno a ed ho el because only a ew e iews e lec
he eno a ion. As a esul , he use may book a less sui able
ho el o none a all, leading o subop imal ou comes o e e-
yone in ol ed. The ho el loses a po en ial gues , while he
cus ome misses ou on a be e expe ience and may se le
o a less a o able one—o none. This, in u n, nega i ely
impac s he cus ome ’s pe cep ion o he pla o m, lowe ing
cus ome sa is ac ion and po en ially ha ming he pla o m’s
epu a ion.
The e o e, we aim o make a Design Science Resea ch
(DSR) con ibu ion by answe ing he esea ch ques ion “how
o design a p obabili y-based me ic o assessing cu ency
o cus ome e iews ha accoun s o aspec -based s a e
changes.” Speci ically, we design and e alua e a new me hod
o assess he cu ency o cus ome e iews (on an aspec
le el). The eby, on he one hand, we p o ide a me hodologi-
cal con ibu ion o da a quali y, which “has been a cen al
IS esea ch opic o decades” (Padmanabhan e al., 2022, p.
ii) and whose impo ance is cons an ly emphasized in he IS
li e a u e (c . e.g., Ab aham e al., 2023; Peng e al., 2023).
On he o he hand, we con ibu e o he ongoing deba e on
he bene i s o cus ome e iews o digi al pla o ms as one
o he mos p ominen examples o use -gene a ed con en
and a co e componen o digi al pla o ms (c . e.g., B and
e al., 2022; W abel e al., 2022). Indeed, cus ome e iews
o en se e as a da a basis o ecommende sys ems, whe e
i has al eady been shown ha high da a quali y is impo an
o achie e high pe o mance (Hein ich e al., 2021).
Ou p oposed no el me ic o he aspec -based cu ency
o cus ome e iews based on he iden i ica ion o aspec -
based s a e changes. He eby, i is no known wi h ce ain y
i , when, and how o en aspec -based s a e changes occu ,
as explici da a in his ega d is ypically no a ailable. We
a gue ha he p inciples and he knowledge base o p ob-
abili y heo y a e adequa e and aluable, p o iding well-
ounded me hods o desc ibing and analyzing such si ua-
ions unde unce ain y. Thus, we de elop ou aspec -based,
s a e change-d i en me ic based on p obabili y heo y. The
me ic alues can be de ined unambiguously and a e in e -
p e able as p obabili ies (Klie e al., 2021). Mo eo e , hey
can suppo decision-making, o example, by in eg a ing
Elec onic Ma ke s (2025) 35:10 Page 3 o 18 10
hem in o expec ed alue calculus (Hein ich & Klie , 2015).
We demons a e he applicabili y o ou me ic and e alua e
i s alues in e ms o eliabili y and abili y o disc imina e
be ween up- o-da e and ou da ed cus ome e iews. Focusing
on he gues ooms aspec , we use a la ge eal-wo ld da ase
o o e h ee million cus ome e iews o 1500 ho els om
he 50 la ges ci ies in he USA o he pla o m T ipAd iso .
The esul s o he e alua ion show ha he me ic alues a e
eliable and allow a clea disc imina ion be ween up- o-da e
and ou da ed cus ome e iews.
The emainde o he pape is o ganized ollowing he
publica ion schema p oposed by G ego and He ne (2013).
Speci ically, a e illus a ing he p oblem con ex in he
nex sec ion, we p o ide an o e iew o he backg ound
and p io wo ks. Then, we de elop a no el aspec -based,
s a e change-d i en cu ency me ic o cus ome e iews.
We ins an ia e ou me ic using a la ge eal-wo ld da ase
om T ipAd iso and e alua e he me ic alues in e ms
o eliabili y and abili y o disc imina e be ween up- o-da e
and ou da ed ins ances. A e wa ds, we discuss implica ions
o heo y and p ac ice, e lec on limi a ions, and p o ide
an ou look o u u e esea ch. Finally, we conclude wi h a
b ie summa y.
P oblem con ex
A cus ome e iew is a ex ual and/o nume ical e alua ion
o an i em by someone who has p e iously pu chased, is-
i ed, o used he i em usually published on a digi al pla -
o m (Biswas e al., 2022). These e alua ions help mi iga e
he in o ma ion asymme y be ween i em p o ide s and
cus ome s (Hossain e al., 2022). Anyone can c ea e cus-
ome e iews wi hou majo ba ie s, and he e a e ew
es ic ions on he s yle o o ma o he ex (Mudambi &
Schu , 2010). This esul s in a la ge numbe o cus ome
e iews wi h a weal h o in o ma ion abou he co espond-
ing i em, which can se e as a aluable da a esou ce o
da a-d i en decision-making (Shen e al., 2015; Sun e al.,
2022; Sysko-Romańczuk e al., 2022). A p ime example
a e ecommende sys ems, which add ess he p oblem o
cus ome in o ma ion o e load on e-comme ce pla o ms
(Lowin e al., 2023), leading o inc eased e enues (Bawack
e al., 2022) and highe cus ome sa is ac ion (Hana izadeh
e al., 2021; Lu e al., 2015). In addi ion, pla o ms and i em
p o ide s can bene i om he in o ma ion con ained in cus-
ome e iews— o example, by using da a analy ics me h-
ods such as machine lea ning (Bawack e al., 2022). The eby,
he cu ency o cus ome e iews is o g ea impo ance,
as ou da ed da a can lead o inco ec conclusions (Bay ak-
a o e al., 2019; Sadiq & Indulska, 2017). In pa icula ,
decision suppo sys ems such as ecommende sys ems and
da a analy ics asks (e.g., applica ion o machine lea ning
algo i hms) ha use cus ome e iews as inpu da a pe o m
poo ly i he da a is no up- o-da e (Bi kbeck e al., 2022;
Fe encek & Kljajić Bo š na , 2020; Lu e al., 2015). Thus, o
le e age he bene i s o cus ome e iews o ecommende
sys ems and da a analy ics, i is c ucial ha he in o ma-
ion con ained in cus ome e iews e lec s he cu en s a e
o he i em in he eal wo ld. Howe e , i is challenging o
ensu e high quali y and especially cu ency, as a manual
assessmen is no economically easible due o he shee
numbe o cus ome e iews (Paul e al., 2017). We u he
illus a e ou p oblem con ex by in oducing an example
in he o m o a se o cus ome e iews o a ho el wi h an
exempla ily ocus on he aspec gues ooms (c . Table1).
Based on he i s h ee cus ome e iews, a ecommende
sys em would ypically no sugges his ho el o a cus ome
wi h a p e e ence o mode n gues ooms. This ollows
he in ui ion ha he a ings o he ooms a e low and he
espec i e pola i ies in he e iew ex s a e mos ly nega i e,
in o ming abou un-down ooms wi h old u ni u e. Since
Janua y, howe e , he e iews show a di e en pic u e. In
ac , hese e iews indica e a ecen eno a ion. As a esul ,
he ooms a e no longe un-down, bu s ylish and a ac i e.
This is e lec ed in he la es oom a ings and ex s only.
The e o e, he cus ome e iews c ea ed be o e he eno-
a ion may hinde in o med decision-making and lead o
educed pe o mance in da a analy ics asks ha e lec ou -
da ed ac s (i.e., hey desc ibe an ou da ed s a e o he ooms
ha is no longe up- o-da e as he eno a ion has changed
he s a e o he ooms).
The e o e, assessing he cu ency o cus ome e iews is
o pa icula impo ance. In o de o base his assessmen on
Table 1 Illus a i e example o
he p oblem con ex Re iew No Times amp Exce p om cus ome e iew Room a ing
Re iew 1 23.07.2023 “[…] bu he ooms a e ou da ed […]” 2
Re iew 2 28.07.2023 “[…] my oom was old wi h b oken u nishing […]” 2
Re iew 3 11.08.2023 “[…] ooms a e com y bu need e u bishmen .” 3
… … … …
Re iew 94 02.04.2024 “Mode n and s ylish oom design!” 4
Re iew 95 18.04.2024 “[…] awesome newly eno a ed ooms […]” 5
Re iew 96 23.04.2024 “[…] he ooms we e nice […]” 4
Elec onic Ma ke s (2025) 35:10 10 Page 4 o 18
a well- ounded de ini ion o cu ency in ou con ex , we d aw
on he gene al in e p e a ion o cu ency as one dimension
(along wi h, o example, accu acy and comple eness) o he
mul idimensional cons uc o da a quali y (Chengalu -Smi h
e al., 1999; Lee e al., 2002; Redman, 1997). Cu ency meas-
u es whe he da a in an in o ma ion sys em is up- o-da e, i.e.,
whe he he da a in he in o ma ion sys em s ill co esponds
o i s coun e pa s in he eal wo ld (Hein ich & Klie , 2015;
Nelson e al., 2005; Redman, 1997). In his ein, we de ine
cus ome e iews as up- o-da e i he in o ma ion con ained
s ill e lec s he cu en s a e o he co esponding p oduc o
se ice in he eal wo ld. Mo e p ecisely, cus ome e iews
a e ou da ed wi h espec o an aspec i he s a e o his aspec
has changed in he eal wo ld and ice e sa. In he illus a i e
example, i is ob ious ha he gi en se o cus ome e iews
is no longe up- o-da e wi h espec o he aspec gues ooms
since i con ains e iews ha do no e lec he cu en s a e
o he ooms. None heless, he se o e iews may s ill be up-
o-da e wi h espec o o he aspec s men ioned, such as he
loca ion o he ho el o he ood in he ho el’s es au an , i
no s a e changes ha e occu ed wi h espec o hese aspec s.
Thus, we aim o an aspec -based assessmen o he cu ency
o cus ome e iews, since cus ome e iews may e lec he
cu en s a e o he eal wo ld wi h espec o one aspec while
being ou da ed wi h espec o ano he aspec . Indeed, iden i-
ying he occu ence o aspec -based s a e changes is c ucial
o assessing he aspec -based cu ency o cus ome e iews.
Howe e , i is usually no known i , when, o how o en such
aspec -based s a e changes occu , since espec i e explici da a
is ypically missing. Mo eo e , such s a e changes canno be
expec ed o occu wi h p edic able egula i y. Thus, he assess-
men o cu ency is ied o he unce ain y associa ed wi h
he occu ence o aspec -based s a e changes. Si ua ions ha
in ol e unce ain y can be e ec i ely analyzed using me hods
ha ely on he p inciples and knowledge base o p obabili y
heo y (G imaldi e al., 2023). Addi ionally, da a quali y me -
ic alues ep esen ing p obabili ies o e nume ous bene i s
(Hein ich & Klie , 2015). Fo example, hey ha e a measu -
able uni , a e in e al-scaled, and can be used o calcula ing
expec ed alues. The e o e, ou goal in de eloping a me ic o
he aspec -based cu ency o cus ome e iews is o base i on
p obabili y heo y and hus p o ide an indica ion a he han
a e i ied (bina y) s a emen unde ce ain y. In pa icula , he
alues o ou me ic a e in ended o ep esen he p obabili y
ha he in o ma ion associa ed wi h an aspec in a se o cus-
ome e iews is s ill up- o-da e.
Backg ound and ela ed wo k
In his sec ion, we desc ibe p io p esc ip i e knowledge
and exis ing a i ac s in he con ex o da a quali y and cu -
ency o uns uc u ed da a and especially cus ome e iews.
Indeed, he e ha e been signi ican con ibu ions o he
assessmen o da a quali y o s uc u ed da a (Ba ini e al.,
2011; Lee e al., 2002; Pipino e al., 2002) and some ini ial
e o s o uns uc u ed da a (Immonen e al., 2015; Kie e ,
2016, 2019). Howe e , in he con ex o cus ome e iews,
he e is s ill a lack o esea ch on assessing da a quali y in
gene al and cu ency in pa icula . While app oaches o
assessing he cu ency o s uc u ed da a a e no di ec ly
applicable o cus ome e iews due o hei uns uc u ed
na u e, hey can s ill p o ide aluable s a ing poin s o
de eloping espec i e me ics. Agains his backg ound, in
his sec ion, we i s discuss wo ks ega ding he cu ency o
bo h s uc u ed and uns uc u ed da a. In pa icula , we ocus
on ideas ha can se e as s a ing poin s o de eloping a
me ic o cu ency o cus ome e iews. We hen p o ide
an o e iew o exis ing wo ks on assessing he da a quali y
o cus ome e iews in gene al and cu ency in pa icula .
Rega ding he assessmen o cu ency o s uc u ed da a,
wo o he mos no able con ibu ions ha e been made by
Ballou e al. (1998) and E en e al. (2010). They model cu -
ency ( e e ed o as imeliness by he au ho s) based on he
age (a he ins an o assessing cu ency), a gi en shel li e
o a s uc u ed da a a ibu e, and a sensi i i y pa ame e o
adap he me ic o he con ex o he applica ion. To o e -
come he weakness ha no all a ibu es ha e a p ede e -
mined shel li e, in ecen yea s, p obabili y-based me ics
ha e eme ged as a p omising a enue o measu ing cu ency.
Fo example, based on he assump ion ha he shel li e ol-
lows an exponen ial dis ibu ion (Hein ich & Klie , 2011)
o by modeling cu ency wi h Ma ko chains (Wechsle &
E en, 2012). Ano he app oach is o inco po a e condi ional
expec a ions and addi ional me ada a in o he calcula ion
o he me ic alues (Hein ich & Klie , 2015). P obabili y-
based me ics ha e he ad an age ha hei alues a e in e -
al-scaled and in e p e able (Hein ich & Klie , 2009, 2015).
Despi e hei ad an ages and hei po en ial o au oma ically
assess he cu ency o s uc u ed da a, all hese app oaches
a e de ined o s uc u ed da a wi h sepa a e a ibu es ha
a e no a ailable in uns uc u ed da a, such as cus ome
e iews. As a esul , hey canno be di ec ly applied o cus-
ome e iews. Ne e heless, he concep o an au oma ed,
p obabili y-based me ic ha p o ides in e p e able esul s
seems p omising and may be adap ed in he con ex o cus-
ome e iews as well.
The li e a u e also p o ides ini ial con ibu ions ega ding
he assessmen o he cu ency o uns uc u ed da a (Ba ini
& Scannapieco, 2016; Fi mani e al., 2016; Hao e al., 2020;
Shah e al., 2015; Zhu & Gauch, 2000). Fo example, when
assessing he cu ency o big da a en i onmen s (Fi mani
e al., 2016), knowledge bases (Shah e al., 2015), and web-
si es (Zhu & Gauch, 2000), cu ency is o en ep esen ed
by he ime since he las upda e (Ba ini & Scannapieco,
2016). The ad an age o he ime since he las upda e as an

Elec onic Ma ke s (2025) 35:10 Page 5 o 18 10
indica o o cu ency is ha i is a ailable o many applica-
ions o uns uc u ed da a in gene al and cus ome e iews
in pa icula . Howe e , in he con ex o cus ome e iews
con aining di e en aspec s, he ime since he las upda e
is p oblema ic as a sole cu ency indica o , since cus ome
e iews can s ill be up- o-da e wi h espec o ce ain aspec s
while being ou da ed wi h espec o o he s. In addi ion,
aspec -based s a e changes, such as majo eno a ions, do
no occu wi h p edic able egula i y. The e o e, a me ic
based on he ime since he las upda e canno accoun o
he unce ain y o hese aspec -based s a e changes.
In he li e a u e on cus ome e iews, da a quali y is o en
associa ed wi h he help ulness o a e iew (Almag abi e al.,
2015), e.g., by measu ing he p opo ion o help ul o es i
ecei es (Chua & Bane jee, 2016). As a esul , many s ud-
ies ocus on au oma ically de e mining he help ulness o
cus ome e iews and in es iga ing he impac o di e en
ea u es o cus ome e iews on hei help ulness (Alma-
g abi e al., 2015). To es ima e he help ulness o cus ome
e iews, some esea che s use eg ession models o calcu-
la e a help ulness sco e be ween ze o and one (Kim e al.,
2006; Lee & Choeh, 2014; Zhang & Va ada ajan, 2006),
while o he s use classi ie s o de e mine whe he a e iew is
help ul o no (Ghose & Ipei o is, 2011; Hong e al., 2012;
Malik & Hussain, 2017). Al hough hese app oaches can
accu a ely p edic he help ulness o cus ome e iews based
on ea u es such as e iew leng h, numbe o spelling e o s,
and subjec i i y sco es, hey do no accoun o da a quali y
as a mul idimensional cons uc , no do hey accoun o
cu ency in pa icula . In addi ion, he help ulness o cus-
ome e iews is no di e en ia ed o di e en aspec s o an
i em. The e o e, app oaches ha ocus on he help ulness o
cus ome e iews canno iden i y quali y issues ela ed o
speci ic dimensions, such as cu ency, no can hey iden i y
da a quali y issues ela ed o speci ic aspec s o he i em.
Despi e he impo ance o he cu ency o cus ome
e iews o da a-d i en decision-making, only e y ew
esea che s ha e add essed he au oma ic assessmen o
cu ency o cus ome e iews. They also ely on age, ei he
measu ed in days since he e iew was c ea ed (Meng e al.,
2021) o as he numbe o days be ween he cus ome e iew
and he i s cus ome e iew c ea ed, o assess he cu ency
o cus ome e iews (Chen & Tseng, 2011). Howe e , nei-
he he in o ma ion ha a e iew was c ea ed a ce ain num-
be o days a e he i s e iew no he in o ma ion ha a
cus ome e iew was c ea ed a ce ain numbe o days ago is
di ec ly ela ed o he ex en o which he espec i e e iew
is s ill up- o-da e. This is due o he ac ha he cu ency o
cus ome e iews is ied o he occu ence o s a e changes o
he e iewed i ems, such as ho el eno a ions, which occu in
un egula pa e ns ha a e no s ic ly ela ed o age o ime.
The inabili y o accoun o hese s a e changes ende s he
a o emen ioned app oaches inapp op ia e o assessing he
cu ency o cus ome e iews, as hei accu acy is comp o-
mised. In addi ion, i is necessa y o assess he cu ency o
cus ome e iews on an aspec le el, since he in o ma ion
in a cus ome e iew can be up- o-da e wi h espec o one
aspec and ou da ed wi h espec o ano he . Howe e , o he
bes o ou knowledge, he e is no wo k ha assesses he cu -
ency o cus ome e iews wi h espec o indi idual aspec s
o he i em being e iewed.
In summa y, he e has been signi ican p og ess in
assessing he cu ency o bo h s uc u ed and uns uc u ed
da a. Howe e , in he con ex o cus ome e iews, hese
app oaches ace challenges due o hei eliance on s uc-
u ed da a a ibu es o he conside a ion o ea u es ha
a e di icul o de ine o cus ome e iews, such as shel
li e. While some ini ial e o s ha e been made o assess he
cu ency o cus ome e iews, hese app oaches a e limi ed
in hei abili y o o e come he challenges o assessing he
cu ency o cus ome e iews. In pa icula , hey assess cu -
ency as a sole unc ion o he age o he cus ome e iews
and hus canno accoun o s a e changes o he co espond-
ing i em in he eal wo ld. Indeed, hey do no make use o
he ich in o ma ion (e.g., ( ex ual) eedback) con ained in
cus ome e iews ha could indica e s a e changes. Mo eo-
e , hey do no ocus on assessing he cu ency o cus ome
e iews wi h espec o di e en aspec s o he associa ed
i em. O e all, his leads o an inaccu a e and a he coa se
assessmen o he cu ency o cus ome e iews. To add ess
his esea ch gap, we p opose a p obabili y-based me ic o
assessing he aspec -based cu ency o cus ome e iews.
The me ic is based on he iden i ica ion o aspec -based
s a e changes using s a is ical ou lie es s and he ich in o -
ma ion con ained in cus ome e iews. Thus, i accoun s
o he unce ain y in he occu ence o aspec -based s a e
changes and p o ides easily in e p e able alues ha ep e-
sen he p obabili y ha he in o ma ion con ained in cus-
ome e iews is s ill up- o-da e wi h espec o an aspec
o he i em.
A me ic o aspec ‑based cu ency
o cus ome e iews
In his sec ion, we desc ibe ou p oposed a i ac : a no el,
p obabili y-based me ic o he aspec -based cu ency o
cus ome e iews. Speci ically, we i s desc ibe he gene al
se ing and he basic idea o ou me ic. On his basis, we
de elop ou me ic. Finally, we show how he me ic can be
ins an ia ed using he G ubbs ou lie es .
Gene al se ing andbasic idea
In he con ex o cus ome e iews, aspec -based cu -
ency exp esses whe he he in o ma ion associa ed wi h a
Elec onic Ma ke s (2025) 35:10 10 Page 6 o 18
pa icula aspec con ained in a se o cus ome e iews s ill
exp esses he cu en s a e o ha aspec o he co espond-
ing i em in he eal wo ld a he ins an o assessmen . As
a esul , cus ome e iews can only become ou da ed wi h
espec o an aspec i he co esponding i em in he eal
wo ld changes wi h espec o ha aspec . Such changes can
be ei he ab up o g adual o e ime. Rega ding ho el cus-
ome e iews, o example, he s a e o he aspec gues
ooms changes ab up ly in he case o a majo ho el eno a-
ion p ojec , while i changes g adually o e a longe pe iod
in he case o inadequa e main enance o when a long- e m
eno a ion p og am is conduc ed. We e e o such changes
ha a ec he cu ency o cus ome e iews as aspec -based
s a e changes. As cus ome e iews become ou da ed i and
only i such an aspec -based s a e change occu s, assessing
he cu ency o cus ome e iews is ied o iden i ying such
s a e changes. Thus, we base ou me ic o aspec -based
cu ency on he iden i ica ion o aspec -based s a e changes
as he unde lying causes o ou da ed aspec s o cus ome
e iews. Howe e , iden i ying aspec -based s a e changes
is associa ed wi h unce ain y because hey ypically do no
occu wi h p edic able egula i y. Indeed, i is no known
wi h ce ain y i , when, and how o en aspec -based s a e
changes will occu . Such si ua ions unde unce ain y can
be desc ibed and analyzed using well- ounded me hods
based on he p inciples and knowledge base o p obabili y
heo y. Thus, we aim o de elop a me ic based on p ob-
abili y heo y. In his line, he alues o ou me ic ep e-
sen he p obabili y ha no s a e change has occu ed o a
e iewed i em du ing he obse a ion pe iod, and hus ha
he associa ed cus ome e iews emained up- o-da e du ing
his pe iod. De ining he me ic alues as p obabili ies has
se e al ad an ages (Hein ich & Klie , 2015). They ha e a
conc e e uni o measu emen , a e in e al-scaled, and can
be in eg a ed in o expec ed alues calculus.
Gi en a se o cus ome e iews as e alua ions o he s a e
o di e en aspec s o an i em, i con ains e idence as o
whe he o no an aspec -based s a e change is likely o ha e
occu ed o e ime, e.g., in he ex s and/o a ings ega ding
he espec i e aspec s o he e iews (Hu & Liu, 2004; Sun
e al., 2019). Fo example, he eno a ion o a ho el’s gues
ooms may be e lec ed in a shi owa ds mo e commen s
abou mode n new ooms, esul ing in a highe p opo ion o
posi i e and ewe nega i e imp essions ega ding he ooms
compa ed o be o e he eno a ion. We base ou me ic on
indica o s de i ed om cus ome e iews and he coexis -
ence o posi i e and nega i e use imp essions ha make
such e idence ega ding he p obabili y o a s a e change
angible. Examples o such indica o s a e he ela i e e-
quency o posi i e and nega i e commen s on he espec-
i e aspec s in he e iew ex s, o posi i e and nega i e
aspec -based a ings o e ime (e.g., on a daily o mon hly
basis, o ming an indica o cu e o e ime). Changes in
he espec i e indica o cu e sugges a highe p obabili y
o an aspec -based s a e change. Howe e , since indi idual
cus ome s may ha e di e en pe cep ions depending on
hei subjec i e p e e ences and expe iences, he indica o
cu e is subjec o andom and undi ec ed noise e en in he
absence o aspec -based s a e changes (Della ocas, 2003;
Mus o & Dahanayake, 2022). The e o e, we aim o iden i y
changes in he indica o cu e ha go subs an ially beyond
his andom noise. The a ea unde he indica o cu e is
mo e esis an o ( andom) noise (Box e al., 2015; Hynd-
man & A hanasopoulos, 2021) and allows he iden i ica ion
o changes in he indica o cu e (Cha ield & Xing, 2019;
Shumway & S o e , 2017). Thus, basing he design o ou
me ic on he a ea unde he indica o cu e allows modeling
he p obabili y o an aspec -based s a e change while being
mo e esis an o ( andom) noise. Figu e1 illus a es wo
(aspec -based) indica o cu es o he wo di e en cases
ega ding (aspec -based) s a e changes: one wi hou a s a e
change (le side) and one wi h a s a e change ( igh side).
In he case o no s a e change, he a ea unde he indica-
o cu e emains app oxima ely he same o e all (equally
sized) ime s eps, wi h only small luc ua ions due o he
expec ed and una oidable noise. In con as , in he case o a
s a e change, he a ea unde he indica o cu e changes sub-
s an ially compa ed o he a ea unde he indica o cu e o
p e ious ime s eps. Such subs an ial changes cons i u e ou -
lie s o he a ea unde he indica o cu e o a ime s ep wi h
espec o p e ious ime s eps. Indeed, in a ma hema ical
sense, ou lie s ep esen obse a ions signi ican ly di e en
om o he obse a ions (G ubbs, 1969; Maddala & Lahi i,
1992). Thus, gi en a pa i ioning o he indica o cu e in o
ime s eps, he p obabili y o an aspec -based s a e change
can be iden i ied wi h he p obabili y o he exis ence o an
ou lie in he a ea unde he indica o cu e. To de e mine
hese p obabili ies, he e a e well-es ablished and sound
me hods om s a is ical hypo hesis es ing o ou lie s.
They p o ide p- alues ha can be in e p e ed as he p ob-
abili y ha no ou lie is p esen (Chandola & Kuma , 2009;
Hodge & Aus in, 2004). Based on hese p- alues, ou me ic
models he p obabili y ha a se o cus ome e iews is s ill
up- o-da e. This is achie ed by mul iplying he indi idual
p- alues o all ime s eps (whe e each p- alue ep esen s
he p obabili y ha he a ea unde he indica o cu e o
ha ime s ep is no an ou lie and hus no s a e change has
occu ed) o calcula e he me ic alue (i.e., he p obabili y
ha no s a e change has occu ed in any o he ime s eps and
hus he se o e iews is s ill up- o-da e). In his sense, ou
me ic is capable o de ec ing bo h ab up and g adual s a e
changes. While changes in he indica o cu e om g ad-
ual s a e changes may no be as dis inc as hose esul ing
om an ab up s a e change, hey s ill show a change in he
indica o cu e ha di e s subs an ially om he expec ed
and una oidable andom noise. Especially, since hey o en
Elec onic Ma ke s (2025) 35:10 Page 7 o 18 10
pe sis ac oss se e al ime s eps. Thus, when a g adual s a e
change occu s, he agg ega ed p obabili y—calcula ed as he
p oduc o he p obabili ies ac oss indi idual ime s eps—
ha he cus ome e iews being up- o-da e is low (as desi ed
o de ec he g adual s a e change).
To conclude, a se o cus ome e iews can only become
ou da ed wi h espec o an aspec due o an aspec -based
s a e change o he co esponding eal-wo ld en i y. The
p obabili y o an aspec -based s a e change can be iden-
i ied wi h he p obabili y o an ou lie (i.e., a subs an ial
change) in he a ea unde he indica o cu e. Consequen ly,
we de ine ou cu ency me ic as he p obabili y ha no ou -
lie (and hus no s a e change) has occu ed ega ding he
a ea unde he indica o cu e a any ime s ep and base i s
calcula ion on he p- alues o a s a is ical ou lie es .
Design o  hebasic model o  heme ic
To design ou me ic, we model he p obabili y ha a
se o cus ome e iews
R
o a pa icula i em wi hin
he ime pe iod
[
0
;
1]
is s ill up- o-da e wi h espec o
an aspec
a
o he co esponding i em a he ins an o
assessmen
1
. Cus ome e iews can only become ou -
da ed wi h espec o an aspec
a
i a s a e change occu s
ha changes he s a e o aspec
a
o he co esponding
i em in he eal wo ld. Thus, he p obabili y ha he se o
cus ome e iews is s ill up- o-da e a he ins an o assess-
men
1
co esponds o he p obabili y ha no espec i e
s a e change has occu ed in he ime pe iod
[
0
;
1]
. Gi en
a pa i ioning o he ime pe iod
[
0
;
1]
in o (equally-sized)
ime s eps
[
s
0
;s
1]
,…,
[
s
l−1
;s
l]
(wi h
0=s0<
⋯
<sl= 1
),
he p obabili y ha no s a e change occu ed in
[
0
;
1]
is
equi alen o he p obabili y ha no s a e change occu ed
in any o he ime s eps
[
s
0
;s
1]
,…,
[
s
l−1
;s
l]
. To a oid a pos-
sible loss o aluable in o ma ion om p e ious ime s eps
when assessing he p obabili y o a subsequen ime s ep,
we o malize he p obabili ies o subsequen ime s eps
by means o condi ional p obabili ies. This conside a ion
o mul iple ime s eps and he pa h desc ibing ou p ob-
abili y o in e es is illus a ed by he ee diag am shown
in Fig.2.
He e,
Ok
o
k=1, …,l
deno es he (p obabili y- he-
o e ic) e en ha a s a e change wi h espec o aspec 
a
occu ed in he
k
- h ime s ep
[sk−1,sk]
. In con as ,
Ok
ep-
esen s he coun e -e en ha no espec i e s a e change
occu ed in he
k
- h ime s ep. On his basis, in Eq.(1), we
de ine ou me ic (i.e., he p obabili y ha no s a e change
has occu ed in
[
0
;
1]
and hus he p obabili y ha he se
o cus ome e iews
R
is s ill up- o-da e) by mul iplying
all condi ional p obabili ies along he pa h wi h no s a e
change occu ing in any o he ime s eps:
Fig. 1 Indica o cu es wi hou s a e change (le ) and wi h s a e change ( igh )
Fig. 2 T ee diag am highligh -
ing he pa h o p obabili ies ha
no ou lie occu s
Elec onic Ma ke s (2025) 35:10 10 Page 8 o 18
To de e mine he p obabili ies in Eq.(1), we employ ha
he p obabili y ha no aspec -based s a e change occu ed
in he
k
- h ime s ep (i.e., in
[
s
k−1
;s
k]
) co esponds o he
p obabili y ha
Ak
is no an ou lie wi h espec o he a eas
unde he indica o cu e
A1,…,Ak−1
in he p e ious ime
s eps. Fo he i s ime s ep
[
s
0
;s
1]
, no p e ious a ea unde
he indica o cu e is a ailable in
[
0
;
1]
. Thus, ano he well-
ounded me hod is equi ed o es ima e
P
(O
1)
. Fo example,
i would be possible o use a quali y-assu ed e e ence ime
s ep be o e
0
.
S a is ics and he b anch o ou lie de ec ion based on
hypo hesis es ing o e a ich se o well- ounded me hods
o suppo he es ima ion o he equi ed p obabili ies (i.e.,
P(O1)
and
P(
Ok
|
Ok−1,…,O1
)
o
k=2, …,l)
such as he
G ubbs es , Dixon’s Q es , o Thompson Tau es (Chan-
dola & Kuma , 2009; Hodge & Aus in, 2004). In pa icula ,
he well-known concep o he p- alue in hypo hesis es ing
can be used o de i e a ma hema ically sound indica ion o
whe he an ou lie is p esen a a gi en ime s ep (Hodge &
Aus in, 2004). Indeed, gi en he null hypo hesis ha he
alue unde conside a ion is no an ou lie , he co espond-
ing p- alue ep esen s he highes le el o signi icance a
which his null hypo hesis canno be ejec ed. T ans e ed o
ou con ex , he p obabili y ha he a ea unde he indica o
cu e in he
k
- h ime s ep
Ak
is no an ou lie wi h espec
o
A1,…,Ak−1
can be assessed by means o he p- alue
pk
o he hypo hesis es based on he null hypo hesis ha
Ak
is
no an ou lie wi h espec o
A1,…,Ak−1
, unde he condi-
ion ha no ou lie occu ed in hese ime s eps (i.e.,
p1
=P(O
1)
and pk=P
(
Ok
|
Ok−1,…,O1
)
) o
k=2, …,l
).
Finally, he alue o ou me ic o he aspec -based cu ency
o a se o cus ome e iews
R
, which ep esen s he p oba-
bili y ha no aspec -based s a e change has occu ed in any
ime s ep in
[
0
;
1]
, and hus he in o ma ion associa ed wi h
he aspec
a
in
R
is up- o-da e a he ins an o assessmen
1
,
is gi en by:
This o maliza ion o he me ic as he p oduc o an es i-
ma ed p obabili y
pk
pe ime s ep
k=1, …,l
ha he a ea
unde he indica o cu e does no ep esen an ou lie a o s
he de ec ion o bo h ab up and g adual s a e changes. In
he case o an ab up s a e change, he o e all p obabili y
QCURR(
R,
0
,
1)
ha he cus ome e iews
R
a e up- o-da e is
es ima ed o be small because he p- alue
pk
o he ime s ep
in which he ab up s a e change occu s is es ima ed e y
low (as
Ak
cons i u es a e y clea ou lie wi h espec o
(1)
Q
CURR
(
R, 0, 1
)
=P
(
O1
)
⋅P
(
O2
|
O1
)
⋅⋯⋅P
(
Ol
|
Ol−1,…,O1
)
(2)
Q
CURR(R, 0, 1)=p1⋅p2⋅⋯⋅pl=
l
∏
k=1
p
k
A1,…,Ak−1
). G adual s a e changes also show low p- alues.
While no as dis inc as in he case o a sudden s a e change,
o g adual s a e changes, hese low alues ex end ac oss
se e al ime s eps. The e o e, when a g adual s a e change
occu s, he agg ega ed p obabili y o he cus ome e iews
being up- o-da e (as a p oduc o all p- alues) is low (as
desi ed o de ec he g adual s a e change).
Ope a ionaliza ion o  hebasic model using
heG ubbs ou lie es
Ou me ic p o ides he p obabili y ha he in o ma ion
associa ed wi h an aspec
a
in a se o cus ome e iews
R
is up- o-da e by assessing he p obabili y o occu ence o
an aspec -based s a e change in he espec i e ime pe iod
[
0
;
1]
. To be able o de e mine
p1
o he i s ime s ep
analogously o
pk
o he ollowing ime s eps
k=2, …,l
,
we addi ionally use a quali y-assu ed e e ence ime s ep
( e e ed o as ime s ep 0, wi h a ea unde he cu e
A0
)
be o e
0
. Equa ion(2) p o ides he ma hema ical de ini ion
o ou me ic based on p- alues om s a is ical hypo hesis
es s o ou lie s (i.e., he condi ional p obabili ies
pk
o
ime s eps
k=1, …,l
ha he a ea
Ak
is no an ou lie wi h
espec o he a eas
A0, ..., Ak−1
unde he condi ion ha no
s a e change has occu ed in he p e ious ime s eps). In
s a is ical hypo hesis es ing, he G ubbs es (G ubbs, 1969;
S e ansky, 1972; Thompson, 1935) is a widely used, eliable,
obus , and compu a ionally inexpensi e choice o de ec -
ing ou lie s (U oy & Au usseau, 2014). Agains his back-
g ound, o he ope a ionaliza ion o ou basic model, we use
he G ubbs es o de e mine
pk
(
k=1, …,l
). He e,
pk
is he
p- alue o he G ubbs es , which ep esen s he p obabili y
ha
Ak
is no an ou lie wi h espec o he dis ibu ion o he
a eas unde he indica o cu e in he p e ious ime s eps.
The es s a is ic
Gk
o he G ubbs es is calcula ed by ak-
ing he absolu e alue o he di e ence be ween
Ak
(i.e., he
a ea unde he indica o cu e ha we a e es ing o being
an ou lie ) and he expec ed alue
𝜇A,k
o he dis ibu ion
o he a eas unde he indica o cu e in he p e ious ime
s eps, di ided by i s s anda d de ia ion
𝜎A,k
:
To de e mine he es s a is ic, he mean
𝜇A,k
and he s and-
a d de ia ion
𝜎A,k
o he no mal dis ibu ion o he a eas unde
he indica o cu e in p e ious ime s eps a e needed. Fo his
pu pose, we exploi he ac ha he a eas unde he indica-
o cu e depend on he alues o he espec i e indica o .
Unde he condi ion ha no s a e change has occu ed in he
p e ious ime s eps, hese indica o alues show he ypical
beha io o a no mally dis ibu ed andom a iable (Shumway
& S o e , 2017), since hey a e nea ly cons an wi h small
(3)
G
k=
|
|
Ak−𝜇A,k
|
|
𝜎
A,k
Elec onic Ma ke s (2025) 35:10 Page 15 o 18 10
which is he mos c ucial ac o o assessing he cu ency
o cus ome e iews. To his end, he p oposed me ic is
based on he iden i ica ion o aspec -based s a e changes and
is o mula ed in e ms o p obabili ies o accoun o he
unce ain y in hei occu ence. We demons a e he p ac i-
cal applicabili y o ou me ic and e alua e i s alues based
on a la ge eal-wo ld da ase o ho el cus ome e iews om
he pla o m T ipAd iso . The esul s a e p omising in ha
he p o ided me ic alues a e eliable and allow o a clea
disc imina ion be ween up- o-da e and ou da ed ins ances.
Funding Open Access unding enabled and o ganized by P ojek
DEAL.
Da a A ailabili y The da a ha suppo he indings o his s udy a e
a ailable om he co esponding au ho , MK, upon easonable eques .
Open Access This a icle is licensed unde a C ea i e Commons A i-
bu ion 4.0 In e na ional License, which pe mi s use, sha ing, adap a-
ion, dis ibu ion and ep oduc ion in any medium o o ma , as long
as you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce,
p o ide a link o he C ea i e Commons licence, and indica e i changes
we e made. The images o o he hi d pa y ma e ial in his a icle a e
included in he a icle’s C ea i e Commons licence, unless indica ed
o he wise in a c edi line o he ma e ial. I ma e ial is no included in
he a icle’s C ea i e Commons licence and you in ended use is no
pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will
need o ob ain pe mission di ec ly om he copy igh holde . To iew a
copy o his licence, isi h p://c ea i ecommons.o g/licenses/by/4.0/.
Re e ences
Ab aham, R., Schneide , J., & om B ocke, J. (2023). A axon-
omy o da a go e nance decision domains in da a ma ke -
places. Elec onic Ma ke s, 33(22). h ps:// doi. o g/ 10. 1007/
s12525- 023- 00631-w
Almag abi, H., Maliba i, A., & McNaugh , J. (2015). A su ey o
quali y p edic ion o p oduc e iews. In e na ional Jou nal o
Ad anced Compu e Science and Applica ions, 6(11), 49–58.
h ps:// doi. o g/ 10. 14569/ IJACSA. 2015. 061107
A kinson, K. E. (1989). An in oduc ion o nume ical analysis (Second
Edi ion). Wiley. h p:// digi ale- objek e. hbz- n w. de/ webcl ien /
Deli e yMa nage ? pid= 16413 34& cus om_ a _2= simple_ iew e
Ballou, D., Wang, R., Paze , H., & Tayi, G. K. (1998). Modeling in o -
ma ion manu ac u ing sys ems o de e mine in o ma ion p oduc
quali y. Managemen Science, 44(4), 462–484. h ps:// doi. o g/ 10.
1287/ mnsc. 44.4. 462
Ba ini, C., Daniele, B., Fede ico, C., & Simone, G. (2011). A da a qual-
i y me hodology o he e ogeneous da a. In e na ional Jou nal
o Da abase Managemen Sys ems, 3(1), 60–79. h ps:// doi. o g/
10. 5121/ ijdms. 2011. 3105
Ba ini,C., & Scannapieco,M. (2016). Da a and in o ma ion quali y:
Dimensions, p inciples and echnique. Sp inge In e na ional
Publishing. h ps:// link. sp in ge . com/ con e n / pd / 10. 1007/ 978-
3- 319- 24106-7. pd
Bawack, R. E., Wamba, S. F., Ca illo, K. D. A., & Ak e , S. (2022).
A i icial in elligence in E-Comme ce: A bibliome ic s udy and
li e a u e e iew. Elec onic Ma ke s, 32(1), 297–338. h ps:// doi.
o g/ 10. 1007/ s12525- 022- 00537-z
Bay ak a o , E., Ehmke, G., O’Conno , J., Bu ns, E. L., Nguyen, H.
A., McRae, L., Possingham, H. P., & Lindenmaye , D. B. (2019).
Do big uns uc u ed biodi e si y da a mean mo e knowledge?
F on ie s in Ecology and E olu ion, 6, A icle 239. h ps:// doi.
o g/ 10. 3389/ e o. 2018. 00239
Bi kbeck, G., Nagle, T., & Sammon, D. (2022). Challenges in esea ch
da a managemen p ac ices: A li e a u e analysis. Jou nal o
Decision Sys ems, 31(sup1), 153–167. h ps:// doi. o g/ 10. 1080/
12460 125. 2022. 20746 53
Biswas, B., Sengup a, P., & Ganguly, B. (2022). You e iews o mine?
Explo ing he de e minan s o “pe cei ed help ulness” o online
e iews: A c oss-cul u al s udy. Elec onic Ma ke s, 32(3), 1083–
1102. h ps:// doi. o g/ 10. 1007/ s12525- 020- 00452-1
Box, G. E. P., Jenkins, G. M., Reinsel, G. C., & Ljung, G. M. (2015).
Time se ies analysis: Fo ecas ing and con ol ( i h edi ion).
Wiley se ies in p obabili y and s a is ics. Wiley. h ps:// sea ch.
ebsco hos . com/ login. aspx? di ec = ue& scope= si e& db= nlebk
& db= nlabk & AN= 10613 22
B and, B. M., Kopplin, C. S., & Rausch, T. M. (2022). Cul u al di -
e ences in p ocessing online cus ome e iews: Holis ic e sus
analy ic hinke s. Elec onic Ma ke s, 32(3), 1039–1060. h ps://
doi. o g/ 10. 1007/ s12525- 022- 00543-1
B öcke , J., & Smi h, L. A. (2007). Inc easing he eliabili y o eliabil-
i y diag ams. Wea he and Fo ecas ing, 22(3), 651–661. h ps://
doi. o g/ 10. 1175/ WAF993.1
Chandola, V., & Kuma , V. (2009). Anomaly de ec ion: A su ey. ACM
Compu ing Su eys, 41(3), 1–58. h ps:// doi. o g/ 10. 1145/ 15418
80. 15418 82
Cha ield,C., & Xing,H. (2019). The analysis o ime se ies: An in o-
duc ion wi h R (Se en h Edi ion). Chapman & Hall/CRC ex s in
s a is ical science se ies. Chapman and Hall/CRC. h ps:// ebook
cen al. p oqu es . com/ lib/ kxp/ de ail. ac ion? docID= 57608 99
Chen, C. C., & Tseng, Y.-D. (2011). Quali y e alua ion o p oduc
e iews using an in o ma ion quali y amewo k. Decision Sup-
po Sys ems, 50(4), 755–768. h ps:// doi. o g/ 10. 1016/j. dss. 2010.
08. 023
Chen,J.-S. (2023). An eceden s and ou comes o i ual p esence in
online shopping: A pe spec i e o SOR (S imulus-O ganism-
Response) pa adigm. Elec onic Ma ke s, 33(1). h ps:// doi. o g/
10. 1007/ s12525- 023- 00674-z
Chengalu -Smi h, I. N., Ballou, D. P., & Paze , H. L. (1999). The
impac o da a quali y in o ma ion on decision making: An
explo a o y analysis. IEEE T ansac ions on Knowledge and Da a
Enginee ing, 11(6), 853–864. h ps:// doi. o g/ 10. 1109/ 69. 824597
Chua,A.Y., & Bane jee,A. (2013). Reliabili y o e iews on he In e -
ne : The case o T ipAd iso . In P oceedings o he Wo ld Con-
g ess on Enginee ing and Compu e Science, San F ancisco, CA.
Chua, A. Y., & Bane jee, S. (2016). Help ulness o use -gene a ed
e iews as a unc ion o e iew sen imen , p oduc ype and in o -
ma ion quali y. Compu e s in Human Beha io , 54, 547–554.
h ps:// doi. o g/ 10. 1016/j. chb. 2015. 08. 057
Da is,P.J., & Rabinowi z,P. (1984). Me hods o nume ical in eg a ion
(Second Edi ion). Academic P ess. h p:// www. loc. go / ca di /
enhan cemen s/ y0667/ 20060 50288-d. h ml
DeG oo , M. H., & Sche ish, M. J. (2010). P obabili y and s a is ics
(Fou h Edi ion). Addison-Wesley.
Della ocas, C. (2003). The digi iza ion o wo d o mou h: P omise
and challenges o online eedback mechanisms. Managemen
Science, 49(10), 1407–1424. h ps:// doi. o g/ 10. 1287/ mnsc. 49.
10. 1407. 17308
Dha , S., & Bose, I. (2022). Walking on ai o hopping mad? Unde -
s anding he impac o emo ions, sen imen s and eac ions on
a ings in online cus ome e iews o mobile apps. Decision Sup-
po Sys ems, 162, 113769. h ps:// doi. o g/ 10. 1016/j. dss. 2022.
113769

Elec onic Ma ke s (2025) 35:10 10 Page 16 o 18
Elgendy, N., El agal, A., & Päi ä in a, T. (2022). DECAS: A mode n
da a-d i en decision heo y o big da a and analy ics. Jou nal o
Decision Sys ems, 31(4), 337–373. h ps:// doi. o g/ 10. 1080/ 12460
125. 2021. 18946 74
E ickson, N., Muelle , J., Shi ko , A., Zhang, H, La oy, P., Li, M.,
& Smola, A. (2020). Au oGluon- abula : Robus and accu a e
Au oML o s uc u ed da a. h ps:// a xi . o g/ pd / 2003. 06505. pd
E en, A., Shanka ana ayanan, G., & Be ge , P. D. (2010). E alua ing a
model o cos -e ec i e da a quali y managemen in a eal-wo ld
CRM se ing. Decision Suppo Sys ems, 50(1), 152–163. h ps://
doi. o g/ 10. 1016/j. dss. 2010. 07. 011
Fe encek, A., & Kljajić Bo š na , M. (2020). Da a quali y assessmen
in p oduc ailu e p edic ion models. Jou nal o Decision Sys-
ems, 29(sup1), 79–86. h ps:// doi. o g/ 10. 1080/ 12460 125. 2020.
17769 27
Fi mani, D., Mecella, M., Scannapieco, M., & Ba ini, C. (2016). On
he meaning ulness o “Big Da a Quali y”. Da a Science and
Enginee ing, 1(1), 6–20. h ps:// link. sp in ge . com/ con e n / pd /
10. 1007/ s41019- 015- 0004-7. pd
Fi che , J. M., & Hoogendoo n, G. (2019). Explo ing he clima e
sensi i i y o ou is s o Sou h A ica h ough T ipAd iso
e iews. Sou h A ican Geog aphical Jou nal, 101(1), 91–109.
h ps:// doi. o g/ 10. 1080/ 03736 245. 2018. 15410 22
Ghose, A., & Ipei o is, P. G. (2011). Es ima ing he help ulness and
economic impac o p oduc e iews: Mining ex and e iewe
cha ac e is ics. IEEE T ansac ions on Knowledge and Da a
Enginee ing, 23(10), 1498–1512. h ps:// doi. o g/ 10. 1109/
TKDE. 2010. 188
G ego , S., & He ne , A. (2013). Posi ioning and p esen ing design
science esea ch o maximum impac . MIS Qua e ly, 37(2),
337–355. h ps:// doi. o g/ 10. 25300/ MISQ/ 2013/ 37.2. 01
G e zel,U., & Yoo,K.H. (2008). Use and impac o online a el
e iews. In P oceedings o he In e na ional Con e ence on
In o ma ion and Communica ion Technologies in Tou ism
2008, Innsb uck, Aus ia.
G imaldi,D., Sallan,J.M., A boleda,H., & Sehgal,S. (2023).
Re hinking he ole o unce ain y and isk in Ma ke ing. Jou -
nal o Decision Sys ems, 1–22. h ps:// doi. o g/ 10. 1080/ 12460
125. 2023. 22325 70
G ubbs, F. E. (1969). P ocedu es o de ec ing ou lying obse a-
ions in samples. Technome ics, 11(1), 1–21. h ps:// doi. o g/
10. 2307/ 12667 61
G ubbs, F. E., & Beck, G. (1972). Ex ension o sample sizes and
pe cen age poin s o signi icance es s o ou lying obse a-
ions. Technome ics, 14(4), 847–854. h ps:// doi. o g/ 10. 2307/
12671 34
Hägele, L. J., Klie , M., Obe meie , A. A., & Spa n, C. (2024). Oldie
bu goodie – Cu ency bea s da a age o cus ome e iews when
i comes o ecommende sys em pe o mance. In P oceedings o
he Fo y-Fi h In e na ional Con e ence on In o ma ion Sys ems,
8, Bangkok, Thailand. h ps:// aisel. aisne . o g/ icis2 024/ da a_ soc/
da a_ soc/8
Hana izadeh, P., Ba kho da i Fi ouzabadi, M., & Vu, K. M. (2021).
Insigh mone iza ion in e media y pla o m using ecommende
sys ems. Elec onic Ma ke s, 31(2), 269–293. h ps:// doi. o g/ 10.
1007/ s12525- 020- 00449-w
Hanley, J. A., & McNeil, B. J. (1982). The meaning and use o he
a ea unde a ecei e ope a ing cha ac e is ic (ROC) cu e. Radi-
ology, 143(1), 29–36. h ps:// doi. o g/ 10. 1148/ adio logy. 143.1.
70637 47
Hao,S., Chai,C., Li,G., Tang,N., Wang,N., & Yu,X. (2020). Ou -
da ed ac de ec ion in knowledge bases. In In e na ional Con e -
ence on Da a Enginee ing, Dallas, TX.
Hein ich, B., & H is o a, D. (2016). A quan i a i e app oach o model-
ling he in luence o cu ency o in o ma ion on decision-making
unde unce ain y. Jou nal o Decision Sys ems, 25(1), 16–41.
h ps:// doi. o g/ 10. 1080/ 12460 125. 2015. 10804 94
Hein ich, B., & Klie , M. (2011). Assessing da a cu ency—A p oba-
bilis ic app oach. Jou nal o In o ma ion Science, 37(1), 86–100.
Hein ich, B., & Klie , M. (2015). Me ic-based da a quali y assess-
men — De eloping and e alua ing a p obabili y-based cu ency
me ic. Decision Suppo Sys ems, 72, 82–96. h ps:// doi. o g/ 10.
1016/j. dss. 2015. 02. 009
Hein ich, B., Hop , M., Lohninge , D., Schille , A., & Szuba owicz,
M. (2021). Da a quali y in ecommende sys ems: The impac o
comple eness o i em con en da a on p edic ion accu acy o ec-
ommende sys ems. Elec onic Ma ke s, 31(2), 389–409. h ps://
doi. o g/ 10. 1007/ s12525- 019- 00366-7
Hein ich,B., & Klie ,M. (2009). A no el da a quali y me ic o imeli-
ness conside ing supplemen al da a. In P oceedings o he Eu o-
pean Con e ence on In o ma ion Sys ems, Ve ona, I aly.
Hein ich, B., H is o a, D., Klie , M., Schille , A., & Szuba owicz, M.
(2018). Requi emen s o da a quali y me ics. Jou nal o Da a
and In o ma ion Quali y (JDIQ), 9(2), A icle 12, 1–32.h ps://
doi. o g/ 10. 1145/ 31482 38
Hein ich, B., Hube , M. F., K ap , T., & Schille , A. (2023). The cu -
ency o Wiki A icles – A language model-based app oach. In
P oceedings o he Fo y-Fou h In e na ional Con e ence on
In o ma ion Sys ems, Hyde abad, India. h ps:// aisel. aisne . o g/
icis2 023/ dab_ sc/ dab_ sc/ 14
Hodge, V., & Aus in, J. (2004). A su ey o ou lie de ec ion me hod-
ologies. A i icial In elligence Re iew, 22(2), 85–126. h ps:// doi.
o g/ 10. 1023/B: AIRE. 00000 45502. 10941. a9
Hols ein, J., Schemme , M., Jakubik, J., Vössing,M., & Sa zge ,G.
(2023). Sani izing da a o analysis: Designing sys ems o da a
unde s anding. Elec onic Ma ke s, 33(52). h ps:// doi. o g/ 10.
1007/ s12525- 023- 00677-w
Hong, Y., Lu, J [Jun], Yao, J., Zhu, Q., & Zhou, G. (2012). Wha
e iews a e sa is ac o y: No el ea u es o au oma ic help ulness
o ing. In P oceedings o he 35 h in e na ional ACM SIGIR con-
e ence on Resea ch and de elopmen in in o ma ion e ie al,
Po land, OR.
Hosme , D. W., Lemeshow, S., & S u di an , R. X. (2013). Applied
logis ic eg ession (Thi d Edi ion). Wiley se ies in p obabili y
and s a is ics. John Wiley & Sons.
Hossain, M. A., Ak e , S., & Rahman, S. (2022). Cus ome beha io
o online g oup buying: An in es iga ion using he ansac ion
cos economics heo y pe spec i e. Elec onic Ma ke s, 32(3),
1447–1461. h ps:// doi. o g/ 10. 1007/ s12525- 021- 00479-y
H is o a, D. (2014). Conside ing cu ency in decision ees in he con-
ex o big da a. In P oceedings o he Thi y-Fi h In e na ional
Con e ence on In o ma ion Sys ems, Auckland, New Zealand.
h ps:// aisel. aisne . o g/ icis2 014/ p oce edings/ Decis ionAn aly i
cs/1/
Hu,M., & Liu,B. (2004). Mining and summa izing cus ome e iews.
In P oceedings o he 10 h ACM SIGKDD in e na ional con e -
ence on Knowledge disco e y and da a mining, Sea le, WA.
Hung, H.-Y., Hu, Y., Lee, N., & Tsai, H.-T. (2024). Explo ing online
consume e iew-managemen esponse dynamics: A heu is ic-
sys ema ic pe spec i e. Decision Suppo Sys ems, 177, 114087.
h ps:// doi. o g/ 10. 1016/j. dss. 2023. 114087
Hyndman, R. J., & A hanasopoulos, G. (2021). Fo ecas ing: P inciples
and p ac ice (Thi d Edi ion). OTex s. h ps:// o ex s. com/ pp2/
Immonen, A., Paakkonen, P., & O aska, E. (2015). E alua ing he qual-
i y o social media da a in big da a a chi ec u e. IEEE Access, 3,
2028–2043. h ps:// doi. o g/ 10. 1109/ access. 2015. 24907 23
Jin, L., Hu, B., & He, Y. (2014). The ecen e sus he ou -da ed: An
expe imen al examina ion o he ime- a ian e ec s o online
consume e iews. Jou nal o Re ailing, 90(4), 552–566. h ps://
doi. o g/ 10. 1016/j. j e ai. 2014. 05. 002
Elec onic Ma ke s (2025) 35:10 Page 17 o 18 10
Kie e , C. (2016). Assessing he quali y o uns uc u ed da a: An ini-
ial o e iew. In LWDA 2016 P oceedings, Po sdam, Ge many.
h p:// ceu - ws. o g/ ol- 1670/ pape - 25. pd
Kie e , C. (2019). Quali y indica o s o ex da a. In BTW 2019 -
Wo kshopband, Ros ock, Ge many. h ps:// dl. gi. de/ handle/ 20.
500. 12116/ 21801
Kim, S.-M., Pan el, P., Chklo ski, T., & Pennacchio i, M. (2006).
au oma ically assessing e iew help ulness. In P oceedings o
he 2006 Con e ence on Empi ical Me hods in Na u al Language
P ocessing (EMNLP), Sydney, Aus alia.
Klie ,M., Moes ue,L., Obe meie ,A., & Widmann,T. (2021). E en -
d i en assessmen o cu ency o Wiki A icles: A no el p oba-
bili y-based me ic. In P oceedings o he Fo y-Second In e na-
ional Con e ence on In o ma ion Sys ems, Aus in, TX. h ps://
aisel. aisne . o g/ icis2 021/ da a_ analy ics/ da a_ analy ics/ 14
Laza idou, A., Kunco o, A., G ibo skaya, E., Ag awal, D., Liska,
A, Te i, T., de Masson d’Au ume, C., Kocisky, T., Rude , S.,
Yoga ama, D., Cao, K., Young, S., & Blunsom, P. (2021). Mind
he gap: Assessing empo al gene aliza ion in neu al language
models. In Ad ances in Neu al In o ma ion P ocessing Sys ems
34, A i ual E en .
Lee, S., & Choeh, J. Y. (2014). P edic ing he help ulness o online
e iews using mul ilaye pe cep on neu al ne wo ks. Expe
Sys ems wi h Applica ions, 41(6), 3041–3046. h ps:// doi. o g/
10. 1016/j. eswa. 2013. 10. 034
Lee, Y. W., S ong, D. M., Kahn, B. K., & Wang, R. Y. (2002). AIMQ:
A me hodology o in o ma ion quali y assessmen . In o ma ion
& Managemen , 40(2), 133–146. h ps:// doi. o g/ 10. 1016/ S0378-
7206(02) 00043-5
Li in, S. W., Goldsmi h, R. E., & Pan, B. (2018). A e ospec i e
iew o elec onic wo d-o -mou h in hospi ali y and ou ism
managemen . In e na ional Jou nal o Con empo a y Hospi-
ali y Managemen , 30(1), 313–325. h ps:// doi. o g/ 10. 1108/
IJCHM- 08- 2016- 0461
Lowin, M., Oz, D., Somech, I., Zahn, M. on, Hinz, O., & Reich-
man,S. (2023). Designing p o i -maximizing ecommende
sys ems in E-comme ce: An expe imen al analysis. SSRN Elec-
onic Jou nal. Ad ance online publica ion.h ps:// doi. o g/ 10.
2139/ ss n. 45538 76
Lu, J., Wu, D., Mao, M., Wang, W., & Zhang, G. (2015). Recommende
sys em applica ion de elopmen s: A su ey. Decision Suppo
Sys ems, 74, 12–32. h ps:// doi. o g/ 10. 1016/j. dss. 2015. 03. 008
Maddala, G. S., & Lahi i, K. (1992). In oduc ion o econome ics.
Macmillan.
Malik, M., & Hussain, A. (2017). Help ulness o p oduc e iews as a
unc ion o disc e e posi i e and nega i e emo ions. Compu e s
in Human Beha io , 73, 290–302. h ps:// doi. o g/ 10. 1016/j. chb.
2017. 03. 053
McKinney, V., Yoon, K., & Zahedi, F. (2002). The measu emen o
web-cus ome sa is ac ion: An expec a ion and discon i ma ion
app oach. In o ma ion Sys ems Resea ch, 13(3), 296–315. h ps://
doi. o g/ 10. 1287/ is e. 13.3. 296. 76
Meng, Y., Yang, N., Qian, Z., & Zhang, G. (2021). Wha makes an
online e iew mo e help ul: An in e p e a ion amewo k using
XGBoos and SHAP alues. Jou nal o Theo e ical and Applied
Elec onic Comme ce Resea ch, 16(3), 466–490. h ps:// doi. o g/
10. 3390/ j ae 16030 029
Mudambi,S.M., & Schu ,D. (2010). Wha makes a help ul online
e iew? A s udy o cus ome e iews on Amazon.com. MIS
Qua e ly, 34(1), 185–200. h ps:// doi. o g/ 10. 2307/ 20721 420
Mu phy, A. H., & Winkle , R. L. (1977). Reliabili y o subjec i e
p obabili y o ecas s o p ecipi a ion and empe a u e. Jou nal
o he Royal S a is ical Socie y Se ies c: Applied S a is ics, 26(1),
41–47. h ps:// doi. o g/ 10. 2307/ 23468 66
Mus o,J., & Dahanayake,A. (2022). Quali y cha ac e is ics o use -
gene a ed con en . In M. T opmann-F ick, B. Thalheim, H.
Jaakkola, Y. Kiyoki, & N. Yoshida (Eds.), F on ie s in a i icial
in elligence and applica ions, 343. In o ma ion modelling and
knowledge bases XXXIII (pp.244–263). IOS P ess. h ps:// doi.
o g/ 10. 3233/ FAIA2 10490
Nelson, R. R., Wixom, B. H., & Todd, P. R. (2005). An eceden s o
in o ma ion and sys em quali y: An empi ical examina ion wi hin
he con ex o da a wa ehousing. Jou nal o Managemen In o -
ma ion Sys ems, 21(4), 199–235. h ps:// doi. o g/ 10. 1080/ 07421
222. 2005. 11045 823
Nunkoo, R., Tee oo engadum, V., Ringle, C. M., & Sunnassee, V.
(2020). Se ice quali y and cus ome sa is ac ion: The mode a -
ing e ec s o ho el s a a ing. In e na ional Jou nal o Hospi-
ali y Managemen , 91, 102414. h ps:// doi. o g/ 10. 1016/j. ijhm.
2019. 102414
Padmanabhan, B., Fang, X., Sahoo, N., & Bu on-Jones, A. (2022).
Machine lea ning in in o ma ion sys ems esea ch. MIS Qua -
e ly, 46(1), iii–xix.
Paul,D., Sa ka ,S., Chelliah,M., Kalyan,C., & Sinai Nadka ni,P.P.
(2017). Recommenda ion o high quali y ep esen a i e
e iews in e-comme ce. In RecSys ‘17: Ele en h ACM Con e -
ence on Recommende Sys ems, Como, I aly.
Peng, J., Hahn, J., & Huang, K.-W. (2023). Handling missing al-
ues in in o ma ion sys ems esea ch: A e iew o me hods and
assump ions. In o ma ion Sys ems Resea ch, 34(1), iii– ii.
h ps:// doi. o g/ 10. 1287/ is e. 2022. 1104
Pipino, L. L., Lee, Y. W., & Wang, R. Y. (2002). Da a quali y assess-
men . Communica ions o he ACM, 45(4), 211–218. h ps://
doi. o g/ 10. 1145/ 505248. 506010
Rahman, Q. I., & Schmeisse , G. (1990). Cha ac e iza ion o he
speed o con e gence o he apezoidal ule. Nume ische
Ma hema ik, 57(1), 123–138. h ps:// doi. o g/ 10. 1007/ BF013
86402
Raza, S., & Ding, C. (2022). News ecommende sys em: A e iew
o ecen p og ess, challenges, and oppo uni ies. A i icial
In elligence Re iew, 55, 749–800. h ps:// doi. o g/ 10. 1007/
s10462- 021- 10043-x
Redman,T.C. (1997). Da a quali y o he in o ma ion age. A c ech
House. h ps:// dl. acm. o g/ ci a ion. c m? id= 548570
Sadiq, S., & Indulska, M. (2017). Open da a: Quali y o e quan i y.
In e na ional Jou nal o In o ma ion Managemen , 37(3), 150–
154. h ps:// doi. o g/ 10. 1016/j. ijin omg . 2017. 01. 003
Sande s, F. (1963). On subjec i e p obabili y o ecas ing. Jou nal o
Applied Me eo ology, 2(2), 191–201. h ps:// doi. o g/ 10. 1175/
1520- 0450(1963) 002% 3c0191: OSPF% 3e2.0. CO;2
Shah,A.A., Ra ana,S.D., Hamid,S., & Ismail,M.A. (2015). Web
c edibili y assessmen : A ec ing ac o s and assessmen ech-
niques. In o ma ion Resea ch, 20(1), 365–391. h p:// in o ma io
n . ne / i / 20-1/ pape 663. h ml
Shen,W., Hu, Yu, J., & Ulme ,J.R. (2015). Compe ing o a en ion:
An empi ical s udy o online e iewe s’ s a egic beha io . MIS
Qua e ly, 39(3), 683–696. h ps:// doi. o g/ 10. 25300/ MISQ/ 2015/
39.3. 08
Shumway,R.H., & S o e ,D.S. (2017). Time se ies analysis and i s
applica ions: Wi h R examples (Fou h Edi ion). Sp inge Tex s in
S a is ics. Sp inge . h ps:// doi. o g/ 10. 1007/ 978-3- 319- 52452-8
S a is a. (2023a). Global ho el and eso indus y ma ke size wo ld-
wide 2013–2023. S a is a Inc. h ps:// www. s a i s a. com/ s a i s ics/
11862 01/ ho el- and- eso - indus y- ma ke - size- global/
S a is a. (2023b). To al numbe o use e iews and a ings on T ipad-
iso wo ldwide om 2014 o 2022 (in millions). S a is a Inc.
h ps:// www. s a i s a. com/ s a i s ics/ 684862/ ipa d iso - numbe -
o - e ie ws/
S e ansky, W. (1972). Rejec ing ou lie s in ac o ial designs. Techno-
me ics, 14(2), 469–479. h ps:// doi. o g/ 10. 2307/ 12674 36
Sun, Q., Niu, J., Yao, Z., & Yan, H. (2019). Explo ing eWOM in online
cus ome e iews: Sen imen analysis a a ine-g ained le el.
Elec onic Ma ke s (2025) 35:10 10 Page 18 o 18
Enginee ing Applica ions o A i icial In elligence, 81, 68–78.
h ps:// doi. o g/ 10. 1016/j. engap pai. 2019. 02. 004
Sun, J., Song, J., Jiang, Y., Liu, Y., & Li, J. (2022). P ick he il e
bubble: A no el c oss domain ecommenda ion model wi h adap-
i e di e si y egula iza ion. Elec onic Ma ke s, 32(1), 101–121.
h ps:// doi. o g/ 10. 1007/ s12525- 021- 00492-1
Sysko-Romańczuk, S., Zabo ek, P., W óblewska, A., Dąb owski, J.,
& Tkachuk, S. (2022). Da a modali ies, consume a ibu es and
ecommenda ion pe o mance in he ashion indus y. Elec onic
Ma ke s, 32(3), 1279–1292.
Thompson, W. R. (1935). On a c i e ion o he ejec ion o obse -
a ions and he dis ibu ion o he a io o de ia ion o sample
s anda d de ia ion. The Annals o Ma hema ical S a is ics, 6(4),
214–219. h ps:// doi. o g/ 10. 1214/ aoms/ 11777 32567
T ipAd iso . (2022). T ipAd iso in es o ela ions. h ps:// i . ipa d i-
so . com/
U oy,M., & Au usseau,F. (2014). Applica ion o G ubbs es o ou -
lie s o he de ec ion o wa e ma ks. In Wo kshop on In o ma ion
Hiding and Mul imedia Secu i y, Salzbu g, Aus ia.
Ve ach e , R., Michiels, L., & Goe hals, B. (2022). A e we o ge ing
some hing? Co ec ly e alua e a ecommende sys em wi h an
op imal aining window. In Pe spec i es on he E alua ion o
Recommende Sys ems Wo kshop co-loca ed wi h he 16 h ACM
Con e ence on Recommende Sys ems, Sea le, WA.
Wechsle ,A., & E en,A. (2012). Using a Ma ko -Chain model o
assessing accu acy deg ada ion and de eloping da a main enance
policies. In Ame ican Con e ence on In o ma ion Sys ems, Sea -
le, WA.
W abel, A., Kup e , A., & Zimme mann, S. (2022). Being in o med
o ge ing he p oduc ? How he coexis ence o sca ci y cues
and online consume e iews a ec s online pu chase decisions.
Business & In o ma ion Sys ems Enginee ing, 64(5), 575–592.
h ps:// doi. o g/ 10. 1007/ s12599- 022- 00772-w
Xiang, Z., Du, Q., Ma, Y., & Fan, W. (2018). Assessing eliabili y
o social media da a: Lessons om mining T ipAd iso ho el
e iews. In o ma ion Technology & Tou ism, 18(1–4), 43–59.
h ps:// doi. o g/ 10. 1007/ s40558- 017- 0098-z
Xie, K. L., Chen, C., & Wu, S. (2016). Online consume e iew ac-
o s a ec ing o line ho el popula i y: E idence om T ipad iso .
Jou nal o T a el & Tou ism Ma ke ing, 33(2), 211–223. h ps://
doi. o g/ 10. 1080/ 10548 408. 2015. 10505 38
Yakubu, H., & Kwong, C. K. (2021). Fo ecas ing he impo ance o
p oduc a ibu es using online cus ome e iews and Google
T ends. Technological Fo ecas ing and Social Change, 171,
120983. h ps:// doi. o g/ 10. 1016/j. ech o e. 2021. 120983
Yin,D., Bond,S.D., & Zhang,H (2014). Anxious o ang y? E ec s o
disc e e emo ions on he pe cei ed help ulness o online e iews.
MIS Qua e ly, 38(2), 539–560. h ps:// doi. o g/ 10. 25300/ misq/
2014/ 38.2. 10
Zhang,Z., & Va ada ajan,B. (2006). U ili y sco ing o p oduc
e iews. In P. S. Yu (Chai ), he 15 h ACM in e na ional con e -
ence, A ling on, VA.
Zhang, Y., Lai, G., Zhang, M., Zhang, Y., Liu, Y., & Ma, S. (2014).
Explici ac o models o explainable ecommenda ion based
on ph ase-le el sen imen analysis. In P oceedings o he 37 h
In e na ional ACM SIGIR Con e ence on Resea ch and De elop-
men in In o ma ion Re ie al, 83–92. h ps:// doi. o g/ 10. 1145/
26004 28. 26095 79
Zheng, Q., & Ip, H. H. S. (2013). E ec i eness o he da a gene a ed
on di e en ime in la en ac o model. In P oceedings o he 7 h
ACM con e ence on Recommende sys ems, Hong Kong, China.
h ps:// doi. o g/ 10. 1145/ 25071 57. 25072 02
Zhu,X., & Gauch,S. (2000). Inco po a ing quali y me ics in cen al-
ized/dis ibu ed in o ma ion e ie al on he wo ld wide web. In
P oceedings o he 23 d Annual In e na ional ACM SIGIR Con-
e ence on Resea ch and De elopmen in In o ma ion Re ie al,
A hens, G eece.
Publishe 's No e Sp inge Na u e emains neu al wi h ega d o
ju isdic ional claims in published maps and ins i u ional a ilia ions.