No.of Pages21 ARTICLE IN PRESS

R

A

hmosccpIcar©

J

K

aIsetia

rl

h0

ARTICLE IN PRESS+ModelETAIL-804; No. of Pages 21

Journal of Retailing xxx (xxx, xxxx) xxx–xxx

Managing A Global Retail Brand in Different Markets: Meta-Analyses ofCustomer Responses to Service Encounters

Ruth N. Bolton b,1, Anders Gustafsson c,1, Crina O. Tarasi a,∗,1, Lars Witell d,e,1

a College of Business Administration, Central Michigan University, Mount Pleasant, MI 48859, United Statesb W. P. Carey School of Business, Arizona State University, Tempe, AZ 85287-4106, United States

c BI Norwegian Business School, Nydalsveien 37, 0484 Oslo, Norwayd Department of Management and Engineering, Linköping University, 582 31 Linköping, Sweden

e CTF, Service Research Center, Karlstad University, 651 88 Karlstad, Sweden

bstract

This study investigates how retailers can leverage their brand to shape customers’ satisfaction with service encounters. It develops and testsypotheses about how brand, store, and consumer factors moderate customer responses to experience clues during retail service encounters. Sixeta-regression analyses synthesize and compare results from 842 satisfaction equations describing customers’ encounters with a global retailer

perating 400 stores in 32 countries. The results show how customers weigh their perceptions of service encounters differently depending on brand,tore, and consumer factors. In markets where customers believe the retailer has high holistic brand quality, they place less weight on experiencelues within the store. In markets where customers believe the retailer’s service brand promise, they place more weight on in-store experiencelues. In markets where the retailer promises utilitarian value, customers weigh functional experience clues more heavily. In markets with an onlineurchasing channel, the effect of experience clues common to offline and online store environments is magnified, and unique clues are diminished.n addition, customers heavily weigh experience clues that fit their goals. In general, retail success factors include high brand quality (which makesustomers more forgiving), a service brand promise that is mirrored in the store image (which makes customers attend to the experience cluesligned with them), and the careful monitoring and managing of retail touchpoints (to customize experience clues to each market). In this way,etailers can use customer-based strategies to effectively design and manage their global retail brand in different markets.

2021 New York University. Published by Elsevier Inc. All rights reserved.

EL classification: C55; C81; C93; D91; M16; M31; R20

age

a“aCe

eywords: Brand; Experience; Global; Customer satisfaction; Service; Store im

This paper investigates how retailers can effectively man-ge a global retail brand (e.g., Walmart, Amazon, Aldi, andkea), defined as a brand offered in multiple countries usingimilar and coordinated marketing strategies (Yip 1995). Retail-rs build brands to make their offerings salient to customers,
o differentiate their offerings, to create relevance and mean-ng, and to build brand preference and loyalty. Brakus, Schmitt,nd Zarantonello (2009, p. 53) conceptualized brand experience
∗ Corresponding author.E-mail addresses: [email protected] (C.O. Tarasi),

[email protected] (R.N. Bolton), [email protected] (A. Gustafsson),[email protected] (L. Witell).

1 All authors contributed equally to this work.

rp(

ttsif

ttps://doi.org/10.1016/j.jretai.2021.03.004022-4359/© 2021 New York University. Published by Elsevier Inc. All rights reserv

Please cite this article as: Bolton, Ruth N., et al, Managing A Global Retailto Service Encounters, Journal of Retailing, https://doi.org/10.1016/j.jreta

s the sensations, feelings, cognitions, and behavioral responsesevoked by brand-related stimuli that are part of a brand’s designnd identity, packaging, communications, and environments.”ustomers are influenced by sensory information via experi-nce clues (Berry, Carbone, and Haeckel 2002, p. 85), whichelate to store access, atmospherics, merchandise assortments,rice and promotion, communications, and branding activitiesAilawadi and Keller 2004; Grewal, Levy, and Kumar 2009).

Customer evaluations are based on direct encounters withhe retail brand (Ailawadi and Keller 2004) and on contex-ual factors beyond the retailer’s control (Verhoef et al. 2009),
uch as the customer’s shopping goals and market character-stics. A successful retail strategy recognizes how contextualactors moderate the customer’s response to experience clues
ed.

Brand in Different Markets: Meta-Analyses of Customer Responsesi.2021.03.004

https://doi.org/10.1016/j.jretai.2021.03.004

mailto:[email protected]






ARTICLE IN PRESS+ModelRETAIL-804; No. of Pages 21

R

(rts2seJeomae

wmhabpac(arkq

1

2

3

reassbfZhst

t2

ibdftag2etb

haessFdmaritm(

atssiahtvv

idtfdtatras(is

.N. Bolton et al.

Seiders et al. 2005). We distinguish between functional expe-ience clues—offering utilitarian value, such as a brief waitingime—and emotional experience clues—offering hedonic value,uch as fun and playfulness (Voss, Spangenberg, and Grohmann003). A deep understanding of the moderating effects of brand,tore, and consumer factors can guide retailers in managingxperience clues (Grewal, Levy, and Kumar 2009; Homburg,ozic, and Kuehnl 2017). Moderating factors can magnify theffects of favorable experience clues and diminish the effectsf unfavorable clues. However, little is known about how theseoderators operate, their importance, or how firms can lever-

ge them. Prior research is primarily conceptual, and the fewxisting empirical studies focus on single moderators.

Our study is novel because, rather than studying main effects,e explore how customer responses to experience clues areoderated by brand, store, and consumer factors. It developsypotheses about the moderating effects of customers’ beliefsbout the retail brand (holistic brand quality and its servicerand promise), store factors (store image, availability of onlineurchasing), consumer factors (shopper goals), and control vari-bles (market factors). We investigate how retailers can improveustomer satisfaction with the service encounter by designing1) experience clues that characterize retail brand encounters,nd (2) brand and store image factors that moderate customers’esponses as well as by, (3) adapting the retail brand to mar-et and consumer contexts. Specifically, we ask the followinguestions:

. How do retail brand factors (such as customers’ beliefs orexpectations about holistic brand quality and the servicebrand promise) magnify or diminish the effects of emotionaland functional clues on customer satisfaction with the serviceencounter?

. How do store factors (store image and availability of onlinepurchasing channel) magnify or diminish the effects ofemotional and functional experience clues on customer sat-isfaction with the service encounter?

. How do consumer factors beyond the retailer’s control (shop-per goals such as buying, browsing, and searching) influencethe effects of experience clues on customer satisfaction withthe service encounter?

Our study makes several contributions to retail brandingesearch. First, it addresses calls for research on how retail-rs can manage the branded customer experience (Ailawadind Keller 2004, p. 338). Consistent with conceptual work onervice strategy (Bharadwaj, Varadarajan and Fahy 1993), thistudy identifies new mechanisms of retail brand differentiationased on contextual factors. It highlights the role of retail brandactors in customer service experiences (Brakus, Schmitt, andarantonello 2009; Verhoef et al. 2009). In particular, it showsow the service brand’s holistic quality and brand promise cantrengthen and weaken the effects of experience clues on cus-
omer satisfaction.
Second, our study responds to calls for a complex adap-ive system perspective (e.g., Tax, McCutcheon, and Wilkinson013). Ostrom et al. (2015, p. 142) observe that “[a] service a

2


s designed to be delivered in a particular service ecosystem,ut the ecosystems in other countries or regions may be veryifferent—for example, differences in the availability of trainedrontline employees, the financial and regulatory context, theechnological infrastructure, the business models, and the culturessociated with the service.” Responding to calls for research onlobal retail branding strategies (Grewal, Levy, and Lehmann004), the present study investigates how differences in retailcosystems (e.g., availability of online purchase, economic fac-ors) across countries affect the design and delivery of retailrand experiences.

Third, retailer decisions to foster customer beliefs aboutolistic brand quality, its service brand promise, store image,nd the existence of online purchasing channels moderate theffect of experience clues on customers’ assessments of theirervice encounters. We provide detailed knowledge on howuch strategic decisions influence the consumer inside the store.or example, a store’s image might promise that its outlets areesigned to make it easy to find the products. This perception willagnify experience clues such as friendly frontline employees

nd short waiting times, so managers must allocate sufficientesources to these aspects of the in-store experience. Thesensights clarify how strategic decisions that are implementedhrough retail technology, visual display decisions, and engage-ent strategies play a role in designing retail brand experiences

Grewal, Roggeveen, and Nordfält 2017).Fourth, our study demonstrates how a meta-analytic

pproach can help managers to better understand the customiza-ion and localization of global retail brands. Our meta-analysesynthesize four external data sources with 1.5 million customerurveys from a global retailer operating more than 400 storesn 32 countries in North American, Europe, and Asia. Thesenalyses use hierarchical linear models (HLMs) to control forow customers are nested within stores and stores within coun-ries. This approach can help managers reconcile conflictingiews—each based on a single market—about how customersiew retail experiences.

The following section summarizes prior work on retail brand-ng and its relevance to in-store service encounters. We thenevelop hypotheses about the moderating effects of contex-ual factors on customer responses to experience clues whenorming satisfaction judgments. Our empirical work merges fiveata sources to create a comprehensive database describing cus-omers’ experiences with a single retailer. Six meta-regressionnalyses synthesize and compare the results from 842 satisfac-ion equations describing customers’ encounters with a globaletailer operating 400 stores in 32 countries. We use meta-nalysis to understand how the regression coefficients (effectizes) vary depending on differences in retail strategy executionbrand, store, and consumer factors). To test the external valid-ty and the robustness of our results, we replicated the consumeratisfaction survey for retailers in the same industry in the USA.

Theoretical Background and Conceptual Framework

A strong retail brand is a promise of future satisfaction. It is blend of what the retailer says the brand is, what others say,


R

att2ccw

M

tGuseerppe

F

catwt“pa(

CE

uaipHtaOrscwd

bSthe

tcvigemc

H

tctce2oTserdvw

ipCb(uc1eatipPV

E

crbtbet

.N. Bolton et al.

nd how the retailer delivers on the brand promise from the cus-omer’s viewpoint (Berry 2000). Customers’ beliefs are based onheir brand experiences and service encounters (Bitner and Wang014). A service encounter is the dyadic interaction between austomer and the retailer. Customers use in-store experiencelues to assess how the retailer delivers on its brand promise,here each encounter contributes to their overall satisfaction.

ain Effects of Experience Clues

Marketers have emphasized the multi-dimensional nature ofhe customer experience (Berry, Carbone, and Haeckel 2002;rewal, Levy, and Kumar 2009), including how brand stim-li influence the holistic experience. Every service encounter ishaped by the customer’s internal and subjective responses toxperience clues (Schmitt, Brakus, and Zarantonell 2015). Anxperience clue is “anything that can be perceived or sensed—orecognized by its absence” (Berry, Carbone, and Haeckel 2002,. 86). A clue might be provided by the sensory appeal of thehysical (e.g., pleasant and relaxing) or social (e.g., friendlymployees) environment.

unctional versus emotional experience cluesWe distinguish between functional and emotional experience

lues. Functional clues signal information about the utilitarianspects of service; they are interpreted by the logical part ofhe brain (cf. Nyffenegger et al. 2015), such as evaluations ofaiting times. Emotional clues arise from smells, sounds, sights,

astes, and textures of the product and environment—includingmechanics” (emitted by things) and “humanics” (emitted byeople). Retailers must manage emotional clues as rigorouslys functional clues to provide a superior customer experienceBerry, Carbone, and Haeckel 2002).

onceptual Model of Contextual Moderating Effects onxperience Clues in Service Encounters

To study service encounters in the retailing ecosystem, ourltimate dependent variable is customer satisfaction modeleds a function of experience clues (Fig. 1). More than 50 stud-es have modeled how customer satisfaction levels depend onroduct attributes and (some) experience clues (Szymanski andenard 2001). They typically focus on the antecedents of cus-

omer satisfaction with goods (rather than retail experiences)nd assess main effects, not moderating effects (Taylor 1997).ur focal dependent variables are the effect sizes of retail expe-

ience clues that influence customer satisfaction (Fig. 1, leftide). We study the effect sizes of key experience clues for theooperating retailer, including ease-of-use, frontline employees,aiting time, frustration, ideas and inspiration, and expectancy-isconfirmation.

In our meta-analyses, the moderating contextual factors arerand, store, consumer, and market factors (control variables).
ee the top of Fig. 1. First, the retail ecosystem varies in regard
o brand factors, such as customers’ beliefs about the brand’solistic brand quality and the service brand promise. The coop-rating retailer’s managers believe these brand factors—gained

ctWt

3


hrough experience and from marketing communications—areritical to the success of its business. Second, the ecosystemaries in regard to store factors, such as the attractiveness ofts outlets and products, and consumer factors, such as shopperoals (e.g., browsing); they lead customers to attend to differ-nt experience clues. Third, the retail ecosystem varies due toarket (cultural and socio-economic) factors, which we treat as

ontrol variables.

ow the Retail Brand Context Moderates Experience Clues

This section develops theory-based hypotheses about howhe retail context moderates the effects of experience clues onustomer satisfaction, thereby explaining variation across cus-omers and retail ecosystems. Retailing thought leaders havealled for more research on contextual moderators in customervaluations of the retail experience (Grewal, Levy, and Kumar009; Verhoef et al. 2009). However, there have been few studiesf moderator variables (as shown in Web Appendix Table A1).o our knowledge, there are no systematic and comprehensivetudies of how context variables jointly moderate customers’xperience clues, thereby influencing their evaluations of theiretail brand experiences. We address this knowledge gap byeveloping and testing hypotheses about how retail ecosystemariables moderate the effects of experience clues on satisfactionith the service encounter.We draw on theoretical work in judgment and decision mak-

ng (Weber and Johnson 2009), focusing on the psychologicalrocesses of attention, information integration, and learning.ustomers engage in constructive processing, relying on specificeliefs to interpret their experiences during service encountersPayne, Bettman, and Johnson 1992). Our hypotheses buildpon research suggesting that customer preferences are oftenonstructed and context-dependent (cf. Slovic and Lichtenstein983). For example, compromise and attraction effects arexamples of shifting preferences based on different consider-tions in a choice situation (Dhar and Simonson 2003). Thus,he customer’s interpretation of their retail service encounter isnference-based; their judgments are constructed by drawing onrior beliefs and experience with the brand and store (Kardes,osavac, and Cronley 2004; Payne, Bettman, and Johnson 1992;erhoef et al. 2009).

ffect Sizes of Experience Clues as Dependent Variables

Our dependent variables are the effect sizes measuringustomer responses to two emotional and three functional expe-ience clues. We focus on these five clues because they haveeen identified by prior research as important in retail ecosys-ems, actionable by managers, and directly related to retailrand perceptions (Seiders et al. 2007). They heavily influ-nce consumers’ behavior, and the cooperating retailer useshem as key performance indicators. The functional experience
lues are frontline employees, waiting time, and ease-of-use;he emotional clues are frustration and ideas and inspiration.
e also study expectancy-disconfirmation (same/better/worsehan expected) because it is central to customer satisfaction.


R.N. Bolton et al. Journal of Retailing xxx (xxx, xxxx) xxx–xxx

ceptu

Heorap

B

cnbaratBbcsacstiWmLw(ht

H

q2

qecaepappdttnHtv

SwimTarcevahc

Hc

Fig. 1. Con

owever, the emotional/functional distinction does not apply toxpectancy-disconfirmation because the (cognitive) recognitionf a discrepancy is considered to have a (affective) fulfillmentesponse (Oliver 2014). Hence, we do not develop hypothesesbout expectancy-disconfirmation, but we include it and offer aost hoc analysis.

rand Factors

Brands are universal signals that operate across countries andultures. Marketing communications help build brand aware-ess, knowledge, image, and attachment (Keller 2003). Servicerand equity is created when consumers respond more favor-bly to its marketing actions than they do to those of competingetailers; it encompasses holistic brand quality and specific brandssociations (Keller 2003). The services literature emphasizeshis twofold distinction (Berry 2000; Brodie, Whittome andrush 2009). Customers’ beliefs and expectations about holisticrand quality are created by external marketing that concerns theompany’s reputation rather than specific characteristics of theervice offer. In contrast, customers’ beliefs and expectationsbout the service brand promise arise from brand associationsreated by external marketing about what the service brandtands for and through the service experience associated withhe delivery of the brand promise, thereby creating a distinctivemage (Bitner 1995). To illustrate, a customer might perceive that

almart supermarkets are low in holistic brand quality yet theyay believe its distinctive service brand promise: “Save money.ive better.” As well as these two dimensions, brand share ofallet is often considered a source of competitive advantage

Bharadwaj, Varadarajan, and Fahy 1993). Hence, we proposeypotheses for customer beliefs about holistic brand quality andhe service brand promise and control for brand share of wallet.

olistic brand qualityThe main effect of customer beliefs about holistic brand

uality are well understood (e.g., Erdem, Zhao, and Valenzuela004). In addition, customers’ beliefs about holistic brand

wcm

4

al model.

uality contribute to their expectations prior to the servicencounter, thereby influencing their perceptions of experiencelues. Boulding, Kalra, and Staelin (1999) developed and tested

model in which a customer’s assessment of a retail servicencounter is a blend of their perceptions of experience clues andrior beliefs about holistic brand quality—where the weightsre consistent with a Bayesian updating process. In their model,rior beliefs about holistic brand quality influence the customer’serceptions of experience clues. This process leads to a “qualityouble-whammy”—whereby customers see what they expecto see—which diminishes the weight placed on new informa-ion obtained during the service encounter. Consistent with thisotion, we believe that markets characterized by beliefs of higholistic Brand Quality create a perceptual lens that diminishes

he importance of experience clues characterizing a specific ser-ice encounter—a negative moderating effect.

This prediction is consistent with Voss, Godfrey, andeiders’s (2010) model of the satisfaction-repurchase link inhich moderating effects depend on whether a service attribute

s a complement or substitute—which arises from the relativeagnitudes of satiation and inertia effects in a purchase category.heir model predicts that a high-quality relationship creates

substitute effect that diminishes the effect of satisfaction onepurchase for durable purchase categories in which customersan become satiated. We extend this notion about substituteffects to our model of customers’ evaluations of retail ser-ice encounters. If beliefs of holistic brand quality moderate thettribute-satisfaction link, then a substitute effect implies thatigh holistic brand quality diminishes the effects of experiencelues on customer satisfaction.

1a. When the retailer has created favorable customer per-eptions of Holistic Brand Quality in a market, customers
eigh functional and emotional experience clues less heavily
ompared to markets with less favorable perceptions (negativeoderating effect).


R

S

btibeSsCtclctppwasnbSbsHth

HSel

S

tdp2SaFvia

S

camFqcl

gwswVc(iri(pfo

ebecptiaiaV“i2bwC(hcvva

H(wte

H(ue

O

or

.N. Bolton et al.

ervice brand promiseIn contrast to holistic brand quality (a summary judgment),

eliefs about concrete brand attributes are related to its func-ion (Keller 2003; Snelders and Schoormans 2004)—that is,ts service brand promise. For example, a customer mightelieve—and expect—the holistic brand quality of the Starbucksxperience to be high, whereas they have a concrete belief intarbucks’s brand promise: “To inspire and nurture the humanpirit: one person, one cup and one neighborhood at a time.”onstrual level theory states that people’s mental representa-

ions of stimuli that are psychologically near are low level andoncrete, while stimuli that are psychologically distant are highevel and abstract (Dhar and Kim 2007). Thus, customers’ con-rete belief in the service brand promise can increase attentiono and consideration of experience clues. Continuing the exam-le, their beliefs and expectations about the Starbucks brandromise might increase attention to the actions of the baristaho serves the customer. The Service Brand Promise oper-

tes through multiple mechanisms, including attention, learning,ignaling, inference, and affordance. Through conscious andon-conscious processes, concrete beliefs about the servicerand promise magnify the effect of brand stimuli (Brakus,chmitt, and Zarantonello 2009; Erdem et al. 1999). Customerrand beliefs and experiences lead to more concrete mental con-trual, influencing preferences (Hamilton and Thompson 2007).ence, we believe that the effect of experience clues on cus-

omer satisfaction will be larger in markets where customersold concrete beliefs about the service brand promise.

1b. When customers in a given market believe the retailer’service Brand Promise, they weigh functional and emotionalxperience clues more heavily compared to markets with lowevels of belief (positive moderating effect).

tore Factors

Store image is conceptually distinct from brand image. Cus-omers may have different perceptions of each store in a chainue to differences in accessibility, store atmosphere, store priceerceptions, and merchandise assortment (Ailawadi and Keller004). For example, a customer’s perceptions of a particulartarbucks outlet can be different from his/her perceptions ofnother outlet and from his/her perceptions of the service brand.or example, the atmospherics of a suburban outlet might beery different from the atmospherics of an airport outlet. Wenvestigate two store factors—store image and the existence ofn online purchasing channel.

tore imageFollowing Hartman and Spiro (2005, p. 1115), we con-

eptualize store image as “the gestalt of perceptions andttributes linked to a store as reflected in associations held inemory”—that is, the overall attitude toward the specific store.
or example, although two stores in the same chain sell the sameuality of branded goods, one store might be more (or less) tidy,lean, and well-stocked. It might also have different staffingevels, leading to different wait times and service levels. The
cmii

5


estalt of these perceptions will be considered in conjunctionith other associations held in memory. In-store experiences

hould be designed to be engaging, connecting the customerith the firm in a personal and memorable way (Zomerdijk andoss 2010). According to cognitive fit theory, the retail brandontext and experience clues must be congruent to be effectiveHong, Thong, and Tam 2004). Congruent sensory experiencesnclude smells, sights, sounds, tastes, and social interactions thateinforce the store image. The term “branded service encounter”s used to describe encounters where in-store experience cluese.g., employee behavior) are congruent with the service brandromise (Sirianni et al. 2013), where congruent clues createavorable perceptions of the store image that are key in anmni-channel context.

The customer learns new information from the servicencounter and integrates it into their perceptions of the retailrand (Erdem et al. 1999; Hartman and Spiro 2005). Newxperiences during service encounters are evaluated againstomparison standards that are stored in memory and com-ared to the brand promise. The brand promise embodied inhe retailer’s store image has the potential to magnify or dimin-sh the effects of the retailer’s actions (Bharadwaj, Varadarajan,nd Fahy 1993, p. 85). Addressing customers’ hedonic and util-tarian motives enhances satisfaction (Chitturi, Raghunathan,nd Mahajan 2007). The cooperating retailer offers Hedonicalue by promising a pleasant and relaxing environment andproducts-I-like”; it offers Utilitarian Value by promising thatnformation and products will be easy-to-find (Seiders et al.005; Voss, Spangenberg, and Grohmann 2003). Customers whoelieve and expect that the retailer will provide hedonic valueill attend to congruent emotional clues, such as inspiring ideas.ustomers who believe the retailer’s promise of utilitarian value

easy-to-find) will attend to congruent functional clues, such aselpful employees. In sum, customers will attend to experiencelues that are congruent with the retail brand promise aboutalue. In this study, two store image clues relate to hedonicalue, a pleasant and relaxing environment and products-I-like,nd one store image clue relates to utilitarian value, easy-to-find.

2a. When the brand store image promises Hedonic Valuepleasant and relaxing environment, products-I-like), customersill weigh emotional clues (e.g., frustration, ideas and inspira-

ion) more heavily and functional clues (ease-of-use, frontlinemployees, waiting time) less heavily.

2b. When the brand store image promises Utilitarian Valueeasy-to-find), customers will weigh functional clues (ease-of-se, frontline employees, and waiting time) more heavily andmotional clues (frustration, ideas and inspiration) less heavily.

nline purchasing channelRetail websites provide opportunities for the retailer to

ffer information and engage with customers. The participatingetailer had websites in every country but had not introduced e-
ommerce (i.e., online purchasing) in all countries. This featureade it possible to assess how the availability of online purchas-
ng changed customers’ responses to experience clues. Researchn domain-specific reasoning has shown that customers do not


R

npahir

awwfiwiTcplcswrttcrci

Hcicw

C

tcv2gRosbtiwa

f1dFDt

iroGtst

twtbi(

Hft

Heh

C

tmmtiss

rcHivmtnthamoate

.N. Bolton et al.

ecessarily use knowledge from one domain, such as an onlineurchasing channel, when reasoning about another domain, suchs a store (Lichtenstein, Netemeyer, and Burton 1995). Researchas shown that customers weigh in-store experiences more heav-ly than online experiences when forming expectations aboutetail service (Verhagen and Van Dolen 2009).

Bhatnagar, Lurie, and Zeithaml (2003) developed and tested mathematical model of cross-domain expectations transfer inhich experiences that are more prominent or prototypical areeighed more heavily. Their results from two experiments con-rmed that customers’ beliefs about retail service attributes areeighed more heavily when the focus of the firm’s operations

s primarily offline, and the retailer started as a traditional store.hey suggested that retailers can influence the extent to whichustomers use online experiences in forming offline beliefs byositioning their website and store as being more (or less) simi-ar or as one channel being more prominent than the other. Theooperating retailer began as a traditional chain of stores andubsequently added online channels in some markets. Hence,e believe that customers will weigh the effects of shared expe-

ience clues (i.e., common to both channels) more heavily whenhe retailer makes an Online Purchase Available to reinforcehem. This prediction should hold for experience clues that areomparable across the store and website, such as ideas and inspi-ation and waiting time. However, the presence of the onlinehannel should diminish the effects of any experience clue thats unique to the store, such as frontline employees.

3. Customers in markets where the retailer makes Online Pur-hase Available will weigh shared experience clues (ideas andnspiration, waiting time) more heavily and unique experiencelues (frontline employees) less heavily than those in marketsithout it.

onsumer Factors: Shopper Goals

Experience clues can be viewed as means or mechanismshat enable customers to achieve their goals. Depending on austomer’s goals, two identical service encounters may produceery different outcomes and feelings (Grewal, Levy, and Kumar009; Puccinelli et al. 2009). The retailing literature distin-uishes between utilitarian and hedonic motives (e.g., Chitturi,aghunathan, and Mahajan 2007). Focal shopping goals areften assigned to three categories of buying, browsing, andearching. Browsing is a hedonic goal dominated by exploratoryehavior (Bloch and Richins 1983); it occurs when the cus-omer has no immediate intention of making a purchase. Buyings primarily a utilitarian goal. It is different from searching,hich includes information acquisition, knowledge building,

nd deliberation.When they are shopping, customers retrieve information

rom memory in response to specific clues (Lynch and Srull982). They use highly selective information processing that
epends on their goals, construal level, and task conditions.or example, in a multi-channel banking study, Van Birgelen,e Jong, and De Ruyter (2006) found that a close fit between
he customer’s goal (routine/non-routine) and experience clues

lUne

6


nfluenced the importance of satisfaction as an antecedent ofepeat purchase intentions. Their study drew on cognitive fit the-ry from the decision sciences (Hong, Thong, and Tam 2004).illespie, Muehling, and Kareklas (2018) showed that affec-

ive fit—whereby clues align with an individual’s emotionaltate—is important. Appraisals of goal relevance and fit con-ribute to evaluations (Nyer 1997).

In sum, theoretical and empirical work suggests that cus-omers pay more attention to experience clues that are congruentith their goals—and weigh them more heavily in forming

heir judgments. Based on cognitive and affective fit theory, weelieve that functional clues are more congruent with (utilitar-an) buying, whereas emotional clues are more congruent withhedonic) browsing—thereby influencing their effect sizes.

4a. Customers who are Buying (a utilitarian goal) weighunctional clues (e.g., ease-of-use, frontline employees, waitingime) more heavily than customers pursuing a hedonic goal.

4b. Customers who are Browsing (a hedonic goal) weighmotional clues (e.g., frustration, ideas and inspiration) moreeavily than customers pursuing a utilitarian goal.

ontrol Variables: Market Factors

Our framework has focused on moderator effects that areheoretically important for retail strategy and actionable byanagers. However, our meta-analyses incorporate additionaloderators as control variables—that is, factors that are beyond

he retailer’s control and to which it must adapt. Market factors,ncluding economic and cultural variables, influence customeratisfaction in retail markets (Grewal, Levy, and Kumar 2009),o our meta-analyses will control for them.

Meta-Analysis: Methodology and Data

Our study is the first to use meta-analysis to investigate howetail brand context moderates the effect of experience clues onustomer satisfaction with a service encounter. Szymanski andenard (2001) performed a meta-analysis of 50 studies of sat-

sfaction to investigate how comparison standards (expectationss. performance), types of offering (services vs. goods), andethod factors moderated satisfaction antecedents of expecta-

ions, disconfirmation, equity, and performance, but they didot investigate the brand or the context factors. They foundhat choice of comparison standard and type of offering (botheld constant in our study) moderated the relationship betweenffect and satisfaction. Studies suggest that satisfaction judg-ents are context-dependent but that they typically investigate

ne or two product markets and focus on interactions amongttributes within a market. Instead, our study investigates sys-ematic differences in the effects of experience clues across retailcosystems.

Meta-analysis is best suited for our study because we have
arge amounts of data collected across many stores and countries.sing meta-analysis to compare effect sizes is a parsimo-ious and straightforward way to understand multiple moderatorffects on customer satisfaction. Researchers have begun con-


R Journal of Retailing xxx (xxx, xxxx) xxx–xxx

daBmeaem

f(cvitIcsnsr

wfiH(sstettddmdeutsem

S

adaasrsksct

bmqttbd

SS

gg1

.N. Bolton et al.

ucting single-paper meta-analysis for similar reasons as wells to obtain precise estimates of effect sizes (e.g., McShane andöckenholt 2017). A meta-analytic approach is superior to esti-ating satisfaction equations with cross-equation moderating

ffects because there is insufficient theory to completely specifyll moderating effects for all antecedents. This study consid-red approximately 100 moderating effects (6 effect sizes × 17oderators).Meta-analysis has three main objectives: (1) synthesizing dif-

erent studies’ effect size values to obtain a weighted mean,2) assessing the consistency of the results, and (3), in thease of heterogeneity, using moderator variables to explain theariability (Johnson and Huedo-Medina 2013). Marketers havencreasingly used meta-analysis to investigate how variables sys-ematically moderate the relationship between two constructs.t can also uncover systematic patterns that reflect methodologi-al differences across studies, including research context, modelpecification, measurement, and estimation, but our study doesot have these differences. Hence, our primary purpose is toynthesize results and investigate the moderating effects of theetail brand context.

In our study, customers with different goals are nestedithin stores, and stores are nested within countries—includingxed and random effects. Advanced meta-analysis methods useLMs that can take into account the nested nature of our data

Pastor and Lazowski 2018). We begin by estimating separateatisfaction equations for customers with different goals in eachtore. This stage has two consequences. First, it is not necessaryo consider country/market variability within each satisfactionquation because it is estimated within one country. Second,he effect sizes from these equations will be more conserva-ive (less statistically efficient) than those obtained from pooledata. Both features are addressed by our second stage: con-ucting a meta-analysis on the effect sizes to identify main andoderating effects. In the second stage, we control for metric

ifferences across country by specifying a country-level randomffect (Antweiler 2001). This random effect captures multiplenobserved country-level differences, including metric and cul-ural differences. Moreover, for each customer, we know thetore that they visited, so we incorporated store-specific randomffects to capture unique clientele effects. See a depiction of theethodology divided into steps in Fig. 2.

tudy Context and Customer Satisfaction Database

The first step of a meta-analysis is assembling studies thatddress the same research question using comparable researchesigns. We obtained multiple data sources from a cooper-ting retailer that operates over 400 stores in 47 countriescross North America, Europe, and Asia. The retailer sells onlytore brands—that is, products manufactured exclusively for theetailer and bearing its name. It is well established as a valuetore brand (i.e., good quality for low prices) in the global mar-
etplace. Between 2010 and 2014, the retailer administered theame survey to a sample of customers from each store in eachountry. The stores are widely separated within countries, sohere is no overlap of retail ecosystems. Customers were eligi-
C

7

Fig. 2. Steps in HLM meta-analysis methodology.

le to participate if they had visited one of the retailer’s stores andade at least one purchase in the past. The retailer used an online

uestionnaire to elicit self-reports of customers’ experiences onheir most recent store visit. Customers’ goals were used to iden-ify the following market segments: (1) buying or preparing touy, (2) browsing, or (3) searching. Table 1 summarizes theescriptive statistics.

tep 1: Development of Equations for Customeratisfaction with the Service Encounter

We estimated separate equations for each customeroal/segment within each store, enabling us to represent shopperoals by dummy variables in the meta-analyses (see Fig. 2, Step). Each model was of the following forms:

ustomer Satisfactionsg

= f sg(Experience Cluessgj,Covariatessgk), (1)



Table 1Satisfaction model variables, measures and descriptive statistics.

Construct Measure Mean Std. dev.

Dependent variable:Customer satisfaction

Rating scale 1–5 (1 = ; 5 = ) 4.134 .848

Experience cluesBrand satisfaction Rating scale 1–5 (1 = ; 5 = ) 4.062 .812Frustratinga Average of five dichotomous variables indicating emotions checked on list:

complicated, stressful, frustrating, tiring, annoying.072 .150

Ease-of-use Rating scale 1–5 (1 = ; 5 = ) 3.598 1.033Ideas and inspiration Rating scale 1–5 (1 = ; 5 = ) 3.995 .880Frontline employees Rating scale 1–5 (1 = ; 5 = ) 3.441 .971Waiting time Rating scale 1–5 (1 = ; 5 = ) 3.018 1.020Disconfirmation Rating scale 1–5 (1 = much worse; 5 = much better) 3.121 .706

CovariatesInviting and attractive Rating scale 1–5 (1 = ; 5 = ) 3.667 1.013Price fairness Rating scale 1–5 (1 = ; 5 = ) 3.493 .981Products in stock Rating scale 1–5 (1 = ; 5 = ) 3.642 1.134Products-I-like Rating scale 1–5 (1 = ; 5 = ) 4.074 .833Functionala Average of three dichotomous variables indicating characteristics checked

on list: informative, useful, functional.332 .274

Boringa Average of two dichotomous variables indicating emotions checked on list:boring, dull

.011 .077

Excitinga Average of four dichotomous variables indicating emotions checked onlist: exciting, fun, inspiring, entertaining

.270 .269

Control variablesUsed customer service 1 = Yes, 0 = No .154 .361Use catalog 1 = Yes, 0 = No .826 .379Bought previously 1 = Yes, 0 = No .942 .233Shop only this store 1 = Yes, 0 = No 4.244 .839Loyalty program 1 = Yes, 0 = No .713 .453House – apartment 1 = Yes, 0 = No .298 .457House – studio 1 = Yes, 0 = No .022 .146Living – single 1 = Yes, 0 = No .112 .315Living – with children 1 = Yes, 0 = No .441 .497

a The survey included an emotions inventory, using 14 items drawn from Izard (1991), Richins (1997) and Oliver (2014). Emotion indices were based on ap n usini

woe(d

btepf

C

weiamv

ls

l

CE

vssette

l

rincipal components analysis that identified four orthogonal factors. Rather thanterpreted.

here s denotes the store (s = 1, . . . 400), g denotes the goalf the market segment (g = 1, 2, or 3), j denotes the differentxperience clues (j = 1, . . . 15), and k denotes the covariatesk = 1,. . .9). The 24 predictor variables and their measures andescriptive statistics are shown in Table 1.

Many studies show support for a non-linear relationshipetween satisfaction and its antecedents (Oliver 2014). Consis-ent with prior research, preliminary analyses indicated that thexponential functional form fit better than the linear or multi-licative functional form for all equations. Hence, the functionalorm of Eq. (1) can be written as follows:

ustomer Satisfactionsg = exp(ΣβsgXsg), (2)

here X denotes a vector of variables representing Experi-nce Clues and Covariates. To avoid omitted variable bias, we
ncluded 24 predictor variables. This functional form has twottractive features. First, as it is inherently interactive, it is parsi-onious in capturing any interaction effects among antecedent
ariables. Second, taking the natural logarithm, we can obtain aws

8

g factor scores, indices were formulated to ensure that measures could be easily

inear additive model that can be estimated with ordinary leastquares (OLS). Transforming Eq. (2),

n(Customer Satisfactionsg) = ΣβsgXsg (3)

ustomer Service Encounter versus Overall Brandvaluation

The consumer has formed an overall evaluation of the ser-ice brand (Berry 2000), but we are not interested in overallatisfaction with the brand. Our dependent variable is a rating ofatisfaction with the service encounter. We are interested in howxperience clues contribute to the consumer’s holistic evalua-ion of the service encounter beyond their overall evaluation ofhe retail service brand. So, we explicitly control for the brandvaluation in the satisfaction equation, as follows:

n(Customer Satisfaction with Service Encountersg)

= αsgln(Brand) + ΣβsgXsg (4)

Conveniently, as consumers’ rating of Customer Satisfactionith the Service Encounter and of the overall Brand are mea-

ured on identical scales, this additional term will also capture


R

af

SE

mFTg[ui2ooeS9rem

S

fwaSn(cwfts

S

tmierseetg

wh

smtfscQtWi1cwrw

S

tsSfiactTcmalcoe(foSrw

M

c4is

.N. Bolton et al.

ny idiosyncratic effects of scale usage. Eq. (4) is the functionalorm that we estimate.

teps 2 and 3: Estimation of Customer Satisfactionquations

We used OLS to estimate Eq. (4) for each combination ofarket segment/goal (buy, browse, or search) and store.1 Seeig. 2, Step 2. Measures of all variables are shown in Table 1.he survey elicited information about the customer’s primaryoal: “What was your main reason for visiting Store X today?Select only one.]” Most store/goal equations were estimatedsing 3,000 or more observations, but sample sizes were smallern countries where data collection was difficult. Since there are4 predictor variables, we aimed to ensure a minimum levelf statistical power for all equations; this required a minimumf 100 observations for each equation. Hence, we (ultimately)stimated 930 equations, with average R-squared values of 50%.ee Fig. 2, Step 3. We do not show detailed results from the30 satisfaction equations due to space limitations. The OLSesults indicate that the experience clues are significant and in thexpected direction for the vast majority of models. A correlationatrix is shown in the Web Appendix, Table A2.

tep 4: Deriving Effect Sizes for the Meta-Analyses

We calculated effect sizes for customer responses to threeunctional experience clues (ease-of-use, frontline employees,aiting time), two emotional experience clues (frustration, ideas

nd inspiration), and expectancy-disconfirmation. See Fig. 2,tep 4. We chose these six variables given their consistent sig-ificance in the satisfaction models, their theoretical relevanceas discussed earlier in the paper), and their importance to theooperating retailer. The effect sizes were derived from Eq. (4),hich describes customer satisfaction with the service encounter

or each combination of goal and store, yielding 930 observa-ions for each effect size. The descriptive statistics for the effectizes are summarized in the first six rows of Table 2a.

tep 5: Assembling Observations for the Meta-Analyses

Using the effect sizes from the 930 equations, we preparedo perform six meta-analyses to test our hypotheses about theoderating effects of brand, store, and consumer factors on the

mportance of experience clues for satisfaction with the servicencounter. Each meta-analysis treated the effect size of an expe-ience clue as an outcome variable. The predictors are brand,tore, consumer, and market factors, which were obtained fromxternal sources (Fig. 2, Step 5). The meta-analysis weighs the
stimates of effect sizes (i.e., corrected correlations derived fromhe model coefficients) by the inverse of their variance to lendreater weight to more precise estimates (Borenstein 2009). A
1 We use OLS for two reasons. First, due to the large number of observations,e do not need the gains in efficiency from system estimation. Second, OLS isighly robust with good statistical properties.

adWo(

fw

9


trength of this approach is that we test and control for manyoderators to capture patterns—avoiding spurious relationships

hat might otherwise arise. Table 2b shows a correlation matrixor the variables in the meta-analysis. Three variables of interesthowed relatively high collinearity, close to or above the usualut-off of .70. Service Brand Promise is correlated with Branduality at .69 and with Pleasant and Relaxing at .83. The correla-

ion between Pleasant and Relaxing and Brand Quality is at .70.e tested the effect of multicollinearity on the models by tak-

ng turns eliminating one of the variables (Mason and Perreault991), and the results remained consistent, some models nothanging at all (i.e., Ease-of-Use and Frontline Employees),hile for other models, by removing a variable, some of the

esults became somewhat stronger yet consistent in directionith the models presented in the paper.

tep 6: Model Specification for the Meta-Analyses

The moderators are measured separately for each store’srade area; they are obtained from the following five dataources: Global Brand Survey Report, Customer Satisfactionurvey, Euromonitor, Hofstede’s cultural indices, and internalrm records. Measures for brand factors were obtained from

separate cross-national survey (not the satisfaction survey),alled the Global Brand Survey Report, which the retailer useso assess brand and customer equity in each national market.he retailer’s brand communications promised an “inspiringompany,” “full of surprises.” and “warm and human,” so theseeasures assessed customers’ concrete beliefs or expectations

bout the service brand promise. If the customer reports highevels, they have high expectations related to the retailer’slaims about the brand. Measures for Store factors werebtained from the same brand survey (pleasant and attractivenvironment, easy and convenient) or internal firm recordsstore size, internet purchasing capability). A few store imageactors were measured by average values (across all customers)f survey items from store-level data from the Customeratisfaction Survey. For Consumer factors, shopper goals wereepresented by dummy variables for searching and browsing,ith buying subsumed within the constant.

easures of brand, store, and consumer factorsWe were able to obtain matching measures of brand, store,

onsumer factors, and control variables for 32 of the retailer’s7 markets. These measures are all store level, so there are 331ndependent observations. See Tables 2a and 2b for descriptivetatistics; they exhibit considerable variation because the tradereas are different, as shown in Table 2a (right column). Theseata were combined with data for each market and country.
hen analyzed (separately) by store and goals, there were 842
bservations without missing data for buying (28%), browsing35%), and searching (36%). See Fig. 2, Step 6.

Our hypotheses predicted that brand, store, and consumeractors moderate the effect of experience clues on satisfactionith the service encounter (Fig. 2, Step 7). Algebraically:



Table 2aDescriptive statistics for meta-analysis variables.†

Variable Mean Standarddeviation

Outcome variables: customer responses to experience clues (n = 842)Frustration −.123 .105Ideas and inspiration .008 .017Expectancy-disconfirmation .007 .015Ease-of-use .002 .013Frontline employees .004 .014Waiting time −.004 .011

Predictor variablesBrand factorsa (n = 331)

Average value for the store calculated from Global Brand Survey dataHolistic brand quality (% respondents rating 4 or 5 ) 38.969 9.726Service brand promise (% respondents rating 4 or 5 ) 58.147 10.197Index of three survey measures: inspiring company, full of surprises, warm and human; reliability (alpha) = .90;Brand share of wallet (% share of category purchases, customer-declared) 28.935 9.703

Store factorsb (n = 331)Pleasant/relaxing store image (% respondents rating 4 or 5 ) 58.268 11.126Average value for the store calculated from Global Brand Survey dataProducts-I-like store image (Mean of 5-Point Scale) 4.095 .162Average value for store calculated from store level survey dataEasy-to-find store image (Mean of 5-Point Scale) 3.555 .171Average value for the store calculated from Global Brand Survey dataFamily friendly store image (Mean of 5-Point Scale) 3.994 .174Average value for store calculated from store level survey dataProducts in stock store image (Mean of 5-Point Scale) 3.711 .175Average value for store calculated from store level survey dataStore size (10,000 square meters, source: internal records)c 2.891 .657Online purchase option4 (Indicator value of 1 if customers can purchase online from this retailer in the given country, zero

otherwise).375

Consumer factors: shopper goalsd

Measure from store level survey data: What was your main reason for visiting Store X today? [Select only one.]” Categories available were classifiedas: buying, browsing, and searching.

Control variables (socio-economic) factors (n = 32)e

Hofstede (2003) individualism index 56.938 22.653Hofstede (2003) uncertainty avoidance index 62.875 25.798Growth rate of industry in country (2011–2012, Euromonitor) −1.722 5.330Disposable income in country 2012 (,000 euros, Euromonitor) 36.947 2.207Urban (% of population, Euromonitor) 77.772 12.511

a Erdem, Swait, and Valenzuela (2006), Erdem, Zhao, and Valenzuela (2004).b Grewal, Levy, and Kumar (2009).c Van Birgelen, De Jong, and De Ruyter (2006).

C

emfserrs

C

m

tcgotfscaaais

d Grewal, Levy, and Kumar (2009), Puccinelli et al. (2009).e References for control variables are reported in Web Appendix.

lue Effect Size = g(Brand Factors,Store Factors,

Consumer Factors,Covariates). (5)

In meta-analyses that synthesize highly diverse studies, theffect sizes are also influenced by the purpose, study design, andethodology of the studies. In contrast, we are using estimates

rom identical equations across stores and markets. Since ourtudy design and methodology are the same in all models, theffect sizes are directly comparable. Most meta-analyses use cor-ected correlation coefficients because these are the most compa-able statistics across a variety of study designs. In our case, thetudy design is identical, so there are no method factors in Eq. (5).

ontrol variablesWe included a standard set of eleven covariates in all our

eta-analyses to assess the moderating effect of market fac-

un

10

ors (economic and cultural) on the importance of experiencelues (as measured by effect sizes). Market factors (industryrowth as well as disposable income in a given country) werebtained from Euromonitor and publicly available databases;hey were available for 32 countries. Hofstede’s cultural indicesor each country were obtained from Hofstede (2003). We con-idered household income (Seiders et al. 2005), proportion ofustomers who are loyalty club members (Seiders et al. 2005),nd national levels of trust in the retailer (Hunneman, Verhoef,nd Sloot 2015). Factors such as disposable income sometimesppear in our meta-analyses. When these variables are omitted,t is because they were not statistically significant (p > .1). Mea-ures of these variables were not available for all markets; we
ltimately obtained 842 observations for each meta-analysis. Weow describe the measures.

AR

TIC

LE

IN P

RE

SS

+Model

RE

TAIL

-804;

No.

of Pages

21

R.N

. B

olton et

al.

Journal of

Retailing

xxx (xxx,

xxxx) xxx–xxx

Table 2bPearson correlations for stage two, meta-analysis model variables.

Frustrating Ideas and inspiration Ease-of-use Expectancydisconfirmation

Frontlineemployees

Waitingtime

Service brandpromise

Brandquality

Share ofwallet

Pleasant andrelaxing

Ideas and inspiration .08Ease-of-use −.02 −.09Expectancy disconfirmation .04 −.05 −.06Frontline employees −.02 .12 −.13 .04Waiting time .09 −.11 −.09 −.04 −.16Service brand promise −.01 .00 .01 .01 .02 .08Brand quality −.14 −.14 −.01 .11 .01 .02 .69Share of wallet .09 .09 .06 −.11 −.01 .06 .30 .09Pleasant and relaxing −.02 .01 −.04 .03 −.02 .09 .83 .70 .25Easy-to-find .05 −.19 −.06 −.03 −.08 .03 .25 .34 −.03 .23Family friendly .08 −.13 −.12 .04 −.08 .04 .14 .19 −.31 .21Products I like .06 −.16 −.12 .06 −.10 −.01 .10 .23 −.28 .10Products in stock .05 −.11 −.11 −.03 −.12 .04 .03 .07 −.25 .09Store size .05 .05 −.02 −.01 −.04 −.06 −.14 −.19 .02 −.23Online purchase .13 .13 −.01 −.05 −.11 −.02 −.41 −.45 .06 −.32Individualism −.01 −.05 −.03 .01 −.11 −.11 −.43 −.24 −.26 −.45Uncertainty avoidance .03 .16 −.07 .00 −.03 .00 .32 .20 .07 .46Industry growth rate −.03 .03 −.06 .09 .02 .02 .00 .17 −.29 .09Disposable income .04 .02 .05 .00 −.04 .01 −.56 −.48 −.10 −.54Urban −.05 −.09 .06 −.05 −.02 .03 −.48 −.16 −.02 −.35

Easy-to-find Family friendly Products I like Products in stock Store size Online purchase Individualism Uncertainty avoidance Industry growth rate Disposable income

Family friendly .67Products I like .73 .83Products in stock .61 .78 .75Store size −.14 −.10 −.04 −.04Online purchase .10 .16 .07 .21 .09Individualism .27 .22 .19 .23 .11 .54Uncertainty avoidance −.24 −.14 −.10 −.16 .03 −.34 −.49Industry growth rate −.11 .14 .20 .15 −.03 −.01 −.01 .03Disposable income −.11 −.09 −.13 −.02 .00 .53 .50 −.46 .08Urban −.07 −.30 −.24 −.17 −.02 .10 .33 −.39 .14 .49

11


R

EirEwwa(vaasl

Catsnos(atuonirw

S

p2dimvWtaTftmmwctTt

fdcatlt(seTeontaftasfatt(c

Tntmc

BB

ePp(ePaShftti

.N. Bolton et al.

conomic factors. We controlled for market differences byntroducing indicator variables representing four of the fiveegions where the retailer operates—Asia-Pacific, Easternurope, Western Europe, and North America. In this way, weere able to control for multiple differences between marketsith a parsimonious set of variables. We included economic vari-

bles to capture differences across countries within each regionTalukdar, Sudhir, and Ainslie 2002). A high growth rate pro-ides opportunities for retailers because customers will learnbout brands and potentially change their preferences (Swaitnd Adamowicz 2001). Last, we controlled for store size (inquare meters) as a surrogate for attractiveness (based on retailocation literature).

ulture. Culture operates by influencing customers’ motivesnd emotions, cognitive processes (e.g., abstract vs. concretehinking), categorization and information processing, and deci-ion making (De Mooij and Hofstede 2011). Cultural andational characteristics can be expected to moderate the effectsf brand image appeal, advertising, and perceptions of brandervice quality (e.g., Hsieh, Pan, and Setiono 2004). Steenkamp2001) argued that Hofstede’s four dimensions of cultural vari-tion are useful because they reflect four fundamental issues ofhe relationship between the individual and the group (individ-alism), social inequality (power distance), social implicationsf gender (masculinity), and the handling of uncertainty in eco-omic and social processes (uncertainty avoidance). We expectndividualism and uncertainty avoidance to play an importantole in explaining customers’ responses to experience clueshen making satisfaction judgments.

tep 7: Estimation of the Meta-Analyses

To perform the regression meta-analyses, we used Com-rehensive Meta Analysis (CMA) Version 3 (Borenstein et al.015). In the regression meta-analyses, the elasticity effect is theependent variable, and the variance in the dependent variables explained using the moderators enumerated above, such aseasures of brand, store, and consumer goals, as well as control

ariables, such as customer descriptors and economic factors.e used HLM and employed a random-effects model because

he studies were drawn from different populations (markets),nd the true effect size varied from one population to the next.he results are reported in Tables 3 and 4. We report the results

or the random effects model and not for the fixed effects forhree reasons. First, there is heterogeneity across countries, as

easured by the Q-statistic (p < .001 for all models). Second, theeta-analysis is performed across countries, and random effectsill take into account the country effects (Antweiler 2001). We

ould not take into account country differences at Step 3 since
he individual models are based on data from within a country.hird, fixed effects models are more likely to have Type I errors
han random effects models (Hunter and Schmidt 2000).

rate

12


Results

Table 3 shows the results for emotional experience clues,rustration, and ideas and inspiration as well as expectancy-isconfirmation. Since expectancy-disconfirmation could not belassified a priori as an emotional or functional clue, its meta-nalysis is discussed as a post hoc analysis. Table 4 showshe results for functional experience clues, including front-ine employees, waiting time, and ease-of-use. While we triedo keep these models consistent for ease of comparison, weconsistently) omitted non-significant variables to avoid over-pecification. The tables show the coefficients and standardrrors for the context variables included in each meta-analysis.he constant can be interpreted as the average effect size for thexperience clue; it is always substantial, and the null hypothesisf no effect is always rejected (p < .01). All equations have sig-ificant (p < .01) Q-statistics (a measure of homogeneity amonghe studies; if the null hypothesis fails to be rejected, the studiesre homogenous), and explanatory power averages .25 (rangingrom .12 to .42, as measured by analog R-squared). To interprethe results, we need to take into account that the dependent vari-bles are actually effect sizes, as measured in the initial customeratisfaction models (steps 3 and 4). The negative coefficientsrom the meta-regression are interpreted to reduce the effect sizend the positive coefficients to increase effect sizes. For example,he negative coefficients for holistic brand quality in the emo-ional experience models show that emotional experience cluesfrustration, ideas and inspiration) have less of an impact onustomer satisfaction when the holistic brand quality is strong.

The results of the hypotheses testing are summarized inable 6: 42% of the tests (14 of 33) of the null hypothesis ofo effect are rejected (p < .05). As expected, the effect sizes ofhe moderator variables are smaller than the effect sizes of the

ain effects. The tests of moderating effects are not affected byommon method bias (Siemsen, Roth, and Oliveira 2010).

rand Factors: Holistic Brand Quality and the Servicerand Promise

H1a stated that Holistic Brand Quality would negatively mod-rate experience clues, and H1b stated that the Service Brandromise would positively moderate experience clues. H1a is sup-orted in three of the five meta-analyses, as follows: frustrationp < .01), ideas and inspiration (p < .01), and frontline employ-es (p < .05). Regarding H1b, the effects of the Service Brandromise were significant (with the expected sign) in two meta-nalyses, frustration (p < .01) and ideas and inspiration (p < .01).ee “Brand Factors” in Tables 3 and 4. Taken together, theseypotheses find support in five of the ten tests, providing supportor H1a and H1b. With the exception of frustration, the magni-ude of the moderating effects for holistic brand quality andhe service brand promise are small due to the lack of variabil-ty across stores and countries. First, this result implies that the
etailer is effective in communicating a consistent brand promiseround the world (thereby limiting variation on these brand fac-ors). Second, since we analyze effect sizes for a single servicencounter, a tiny effect on a given visit may add up to a huge



Table 3Meta-analysis results for contextual factors influencing importance of experience clues.

Dependent → emotional experience clues

Moderator (hypothesis, predicted sign)↓ Frustration Ideas and inspiration Expectancy-disconfirmation

Coefficient† Standard error Coefficient† Standard error Coefficient† Standard error

Constant −.5095*** .0828 .0323*** .0113 .0624*** .0152Brand factors

Holistic brand quality (H1a −) −.0028*** .0004 −.0003*** .0001 .0002*** .0001Service brand promise (H1b +) .0019*** .0005 .0003*** .0001 −.0003*** .0001Brand share of wallet .0005 .0003 .0001 .0001 −.0001 .0001

Store factorsPleasant and relaxing (H2a +) −.0002 .0005 −.0001 .0001 .0001 .0001Products-I-like (H2a +) −.0038 .0399 −.0031 .0056 .0123** .0049Easy-to-find (H2b −) .0336 .0263 −.0068 .0037 −.0104*** .0040Products in stock −.0667*** .0232 −.0021 .0034 −.0136*** .0033Family friendly .1323*** .0328 .0016 .0049 NA NAOnline purchase option (H3 +) −.0002 .0085 .0065*** .0010 −.0018 .0013

Consumer factors: shopper goalsBrowsing (hedonic) goal (H4b +) .0365*** .0066 .0042*** .0014 −.0015 .0013Searching goal .0027 .0066 .0027*** .0014 −.0002 .0012

Control variables: market factorsAsia Pacific −.0509*** .0155 NA NA .0007 .0027Europe – East −.1016*** .0191 NA NA .0087*** .0029Europe – West −.0296*** .0106 NA NA .0029 .0016North America −.0329*** .0128 NA NA .0049** .0021Individualism −.0004 .0002 NA NA NA NAUncertainty avoidance .0006*** .0002 .0001*** .0000 −.0001*** .0000Industry growth .0025*** .0007 .0003*** .0001 −.0002 .0001Disposable income NA NA .0000 .0000 .0001** .0000Urban NA NA NA NA −.0002*** .0001Store size .0019 .0036 .0001 .0006 −.0001 .0006

Meta-analysis fit statisticsQ-statistic 16,824 1602 1528Degrees of freedom 841 841 841Analog R-squared .17 .42 .34

Shaded areas indicate the test results for the hypotheses presented in the paper.tion p

*

awvi

vtvet

SU

ewVtht

wasvhiv

hpe(Hasmmm

† Z values, confidence limits and p-values can be calculated from the informa** p < .01.** p < .05.

dvantage over many shopping trips. Third, this moderator effectould be much larger in a multi-brand study, where there is moreariation in holistic brand quality. We demonstrate this featuren a replication study.

Recall that we included brand share of wallet as a controlariable. It does not increase the explained variation in any ofhe models (p > .05). We speculate that share of wallet (plus Ser-ice Brand Promise, p < .01) creates familiarity so that customersxpectations are more likely to be met, diminishing the impor-ance of expectancy-disconfirmation as an experience clue.

tore Factors: Store Image Providing Hedonic andtilitarian Value

H2a and H2b predicted that store image would magnify theffect of experience clues (i.e., a positive moderating effect)hen they are congruent with the store’s promises of Hedonic
alue or Utilitarian Value. To test these hypotheses, we inves-igated whether a moderating variable, when significant, wouldave the expected sign. We did not expect all store image factorso moderate all experience clues—unless the retailer’s strategies
ia

13

rovided. Two sided p-values are indicated by asterisks.

ere extremely effective. We classified the retailer’s store images follows. Pleasant and relaxing and products-I-like were con-idered to promise hedonic value and easy-to-find a utilitarianalue. It also promised being family friendly, which could beedonic (as a place my family enjoys) and functional (provid-ng the amenities I need while shopping with family members)alue, so we do not use it to test our hypotheses.

H2a predicted that when a retailer’s store image promisesedonic value (e.g., pleasant and relaxing environment in thatarticular market), customers will assign greater weight tomotional clues (Table 3) and less weight to functional cluesTable 4). The predictions for emotional clues are not supported.owever, support for H2a is provided by the fact that both pleas-

nt and relaxing and products-I-like are significant, with negativeigns in the models for functional clues (Table 4) negativelyoderating incongruent clues. Pleasant and relaxing negativelyoderates ease-of-use (p < .01), and products-I-like negativelyoderates frontline employees’ performance (p < .01).
H2b predicted that when a brand store image promises util-
tarian value (e.g., high ratings of the store on easy-to-find in particular market), a customer would assign greater weight



Table 4Meta-regression analysis results for contextual factors influencing importance of functional experience clues.

Dependent → functional experience clues

Moderator↓ Ease of use Frontline employees Waiting time

Coefficient† Standard error Coefficient Standard error Coefficient Standard error†

Constant .0312*** .0116 .0893*** .0141 −.0486*** .0114Brand factors

Holistic brand quality (H1a −) .0001 .0001 −.0001** .0001 −.0001 .0000Service brand promise (H1b +) −.0000 .0001 −.0001 .0001 .0001 .0000Brand share of wallet .0000 .0000 .0001 .0001 .0000 .0000

Store FactorsPleasant and relaxing (H2a −) −.0002*** .0001 .0001 .0001 .0000 .0001Products-I-like (H2a −) −.0061 .0047 −.0209*** .0056 NA NAEasy-to-find (H2b +) .0064 .0034 .0094** .0038 .0085*** .0028Family friendly −.0047 .0040 .0021 .0046 .0016 .0040Online purchase option (H3 +) −.0009 .0010 −.0024 .0012 .0024*** .0009

Consumer factors: shopper goalsBrowsing (vs. buying) Goal (H4a −) .0003 .0012 −.0029** .0013 −.0010 .0010Searching (vs. buying) Goal .0011 .0012 −.0017 .0012 −.0020** .0010

Control variables: market factorsAsia Pacific NA NA .0011 .0022 .0053** .0020Europe – East NA NA −.0039 .0026 .0017 .0019Europe – West NA NA −.0050*** .0015 .0014 .0010North America NA NA −.0002 .0018 .0001 .0015Individualism .0000 .0000 −.0001 .0000 −.0001*** .0000Industry growth −.0000 .0001 .0003** .0001 .0000 .0001Urban −.0000 .0001 −.0003*** .0001 .0001*** .0001Store size (×10,000 m2) −.0002 .0005 −.0009 .0005 .0002 .0004

Meta-analysis fit statisticsQ-statistic 1361 1536 1194Degrees of freedom 841 841 841Analog R-squared .12 .25 .24

Shaded areas indicate the test results for the hypotheses presented in the paper.† Z values, confidence limits and exact p values can be calculated from the information provided. Two sided p-values are indicated by asterisks.

*

tnttaov

S

cicnOt

C

w(

trcst(tcltcbatmtcft

** p < .01.** p < .05.

o functional clues and less weight to emotional clues. H2b isot supported for emotional clues (Table 3) but is supported forwo of the functional clues (Table 4). Easy-to-find is a posi-ive and significant moderator for frontline employees (p < .05)nd waiting time (p < .01). In sum, this retailer’s brand promisef hedonic value is less effective than its promise of utilitarianalue.

tore Factors: Online Purchasing Channel

H3 posited that customers in markets with an Online Pur-hasing Channel will weigh shared experience clues (ideas andnspiration, waiting time) more heavily and unique experiencelues (frontline employees) less heavily. This moderator is sig-ificant in two relevant meta-analyses, partly supporting H3.nline shopping enhances the importance of ideas and inspira-

ion (p < .01, Table 3) and waiting time (p < .01, Table 4).

onsumer Factors: Browsing as a Hedonic Goal

H4a and H4b posited that customers would assign greatereight to experience clues with higher cognitive or affective fit

i.e., a positive moderating effect). The meta-analysis parame-

teg

14

ers for customers who are browsing and searching are estimatedelative to those who are buying (which is subsumed in theonstant). Hence, we test these hypotheses by examining theignificance of the Browsing coefficients only because they cap-ures the difference between hedonic (browsing) and utilitarianbuying) segments. H4a predicted that customers who have utili-arian goals would assign greater weight to functional clues. Theoefficient of Browsing is negative (p < .05, Table 4) for front-ine employees but not significant for ease-of-use and waitingime (p > .10), providing partial support for H4a. This result indi-ates that customers pay more attention to store personnel whenuying than when browsing. H4b predicted that customers whore Browsing would assign greater weight to emotional clueshan customers who are buying. This hypothesis of a positive

oderating effect is supported for two emotional clues (frus-ration and ideas and inspiration, p < .01, Table 3). Browsingustomers seek ideas and inspiration and do not want to berustrated, so they pay more attention to both clues (but noto expectancy-disconfirmation).

We did not formulate hypotheses comparing search goals
o buying goals because both are considered utilitarian. How-ver, they are significantly different in two instances. Searchoals (vs. buying) positively moderates ideas and inspiration


R

atstt(

P

eieseta(cHIV((tb

C

caauirDmaaD(etac

R

rbiwrlt

aarr

idapWmsrbaAwstttedeceacr

mPsrWBhedtsb

S

sbvaiQs

.N. Bolton et al.

nd negatively moderates waiting time. One possible explana-ion is that although search is usually associated with problemolving, searching in hedonic environments and solving aes-hetic/functional problems makes searching customers sensitiveo ideas and inspiration (p < .01) and more willing to waitp < .01).

ost Hoc Analysis of Expectancy-Disconfirmation

The meta-analysis for expectancy-disconfirmation is inter-sting because this variable is considered highly importantn the satisfaction literature (Oliver 2014). Recall that themotional/functional distinction is irrelevant for this con-truct. The last column of Table 3 reveals many moderatorffects—although we did not make any predictions. First, Holis-ic Brand Quality has a positive moderating effect (p < .01)nd Service Brand Promise has a negative moderating effectp < .01) on disconfirmation, consistent with other experiencelues (H1a and H1b). Second, store image factors that promiseedonic Value positively moderate disconfirmation—products-

-like (p < .05). A store image factor that promises Utilitarianalue, easy-to-find, has a negative effect on disconfirmationp < .01). These results suggest that a concrete construal levelservice brand promise, store image of easy-to-find) diminisheshe importance of expectancy-disconfirmation, whereas holisticeliefs heighten its importance.

ontrol Variables: Market Factors

Market factors moderate the effect of experience clues onustomer satisfaction. Industry growth and uncertainty avoid-nce enhance the importance of both frustration and ideasnd inspiration in determining the customer’s holistic eval-ation of the experience (p < .01). Individualism reduces themportance of waiting time (p < .01), and uncertainty avoidanceeduces the importance of expectancy-disconfirmation (p < .01).isposable income positively moderated the effect of disconfir-ation on satisfaction. In urban settings, customers pay more

ttention to waiting time when judging satisfaction (p < .01)nd less attention to frontline employee availability (p < .01).isconfirmation was negatively moderated by urban setting

p < .01). Regional covariates were significant for frustration,xpectancy-disconfirmation, frontline employees, and waitingime, indicating differences in customer preferences or the man-gement of the retail brand. Regional variables that did notontribute significantly to the model (p > .1) were removed.

eplication Study: Testing the Robustness of the Key Results

We assessed the robustness and external validity of the keyesults by replicating the effects of holistic brand quality, servicerand promise, and store image (Products-I-Like) as moderatorsn the relationship between experience clues and satisfaction
ith the service encounter. Our goal was to ascertain that the
esults are valid for retailers in general. Since the meta-analysesooked at effect sizes across markets, we aimed to show thathe moderating effects of the brand and store factors are larger

eoao

15


cross brands within a single market than for a single brandcross markets. This difference arises because our meta-analyseselied on (some) measures based on aggregated data, whereas theeplication study utilizes individual-level data (Ostroff 1993).

We replicated the satisfaction survey for retailers in the samendustry in the USA. The study sample consisted of 600 ran-omly selected U.S. consumers, aged eighteen and over, fromn online panel provided by Qualtrics (Prolific). The partici-ants were given a small monetary incentive for participating.e measured the variables as in our main study. Instead of usingeta-analysis, we estimated a general linear model for customer

atisfaction with the service encounter as a function of two expe-ience clues (one functional and one emotional) and with holisticrand quality, service brand promise, and store image as moder-tors. We also included disconfirmation and additional controls.ll variables were standardized, and the model was estimatedith OLS. Since the variables are standardized, the coefficients

hown in Table 5 can be interpreted as effect sizes. Note thathe non-significant control variables have been excluded fromhis table. The results are consistent with H1a, which predictshat Holistic Brand Quality negatively moderates functionalxperience clues (ease-of-use, p < .05), and with H1b, which pre-icts that Service Brand Promise positively moderates functionalxperience clues (ease-of-use, p < .05). We did not find signifi-ant effects for emotional experience clues. We also tested theffect of a store promising Hedonic Value (e.g., products-I-like)nd found a negative moderating effect on functional experiencelues (ease-of-use, p < .05), supporting H2a. See Table 5 for theesults.

This replication study provides additional support for theoderating effects of Holistic Brand Quality, Service Brandromise, and a store image that promises Hedonic Value. Ithows that the results are not only valid for the cooperatingetailer but also for competing retailers in the same industry.

e were also able to replicate the opposing effects of Holisticrand Quality and Service Brand Promise, showing that theyave differing effects on the same experience clue. Finally, theffect sizes in this study (estimated across brands at the respon-ent level) are much larger than the corresponding effect sizes inhe meta-analyses (for a single brand across stores). Hence, theyhow that the magnitude of the moderating effects is relevant forusiness managers.

ummary

Based on a global study covering 32 countries, this researchought to untangle the intricate relationships among many retailranding variables, as summarized in Table 6. It examinedariables with wide applicability that are managerially action-ble and critical in shaping the customer experience. Theres strong support for customer beliefs about Holistic Branduality and Service Brand Promise as moderators of the relation-

hip between experience clues and satisfaction with the service
ncounter. Holistic Brand Quality negatively moderated threef the five clues, and Service Brand Promise positively moder-ted two clues. We also found strong support for the importancef congruence between functional experience clues and store



Table 5Results of the replication study.

a. Descriptive statistics and correlations

Mean Std. dev. 1. 2. 3. 4.

1. Satisfaction with the retail service encounter 4.01 .926 .296** .543** .548**2. Expectancy-Disconfirmation 3.47 1.246 .276** .248**3. Ideas and Inspiration (Emotional Clue) 3.59 1.091 .454**4. Ease-of-use (functional clue) 3.75 1.044

b. Regression model (OLS)

Dependent variable: satisfaction with the retail service encounter

Standardized coefficient Standard error Hypothesis

Constant .105 .036Main effectsExpectancy-disconfirmation .099** .033Ideas and inspiration (emotional clue) .281** .062Ease-of-use (functional clue) .298** .063ModeratorsBrand quality × ideas and inspiration (−) −.118ns .077 H1a

Brand quality × ease-of-use (−) −.167** .084 H1a

Brand promise × ideas and inspiration (+) .025ns .085 H1b

Brand promise × ease-of-use (+) .196** .083 H1b

Hedonic Value (products I like) × ease-of-use k(−) −.072** .031 H2a

A

T s p > .

iTCetoth

etbmcRaeee

T

R

ftTw

cgsetscsicfieOew

U

Tpmmaft2g

djusted R-squared .43

wo sided p-values are indicated by asterisks. ***p < .01. **p < .05. ns indicate

mage factors that signal Utilitarian Value (two of five tests).he retailer seems to be less effective in signaling Hedonic Value.ognitive and affective fit between customer goals and experi-nce clues also mattered; customers pay more attention to clueshat align with their goals (one of three tests). The consistencyf our results across markets with different economic condi-ions, cultures, and customers provides strong support for ourypotheses.

Discussion

Brand and store factors help customers interpret their retailxperiences. Our study explored a comprehensive set of func-ional and emotional experience clues relevant to global retailrands. Consistent with research on context-dependent judg-ents, contextual factors moderated the effect of experience

lues on customer satisfaction with retail service encounters.etailers can leverage brand and store factors to magnify favor-ble clues and diminish unfavorable clues, enhancing customers’valuations of service encounters. However, there are differ-nces in how brand and store factors operate on functional versusmotional experience clues.

heoretical Implications

etailer brand management of customer experiencesVerhoef et al. (2009) emphasized the role of the brand in

orming the customer experience; if a customer is primed withhe brand, it is likely to influence the entire customer experience.hey recommended that research should examine the extent tohich brand beliefs moderate the effects of other determinants of

cabf

16

05. Standardized coefficients are effect sizes.

ustomer experience. Our research answers this call by investi-ating how customer beliefs about holistic brand quality and theervice brand promise can strengthen and weaken the effects ofxperience clues on customer satisfaction. Our findings showhat retail brand strategy can shape customers’ responses toervice encounters in the following two ways: through explicitommunications that create beliefs about brand quality and theervice brand promise and through implicit brand promises thatnform customers’ in-store experiences. For example, when austomer learns that a store promises that it will be easy tond products, this belief will magnify the importance of experi-nce clues, such as friendly employees and short waiting times.ur findings provide detailed knowledge on how retailers can

nhance the customer experience by familiarizing consumersith their brand.

sefulness of a systems perspectiveOur research addresses a key research priority identified by

ax, McCutcheon, and Wilkinson (2013)—the need for a com-lex adaptive system perspective. In retailing, several conceptualodels have been introduced to capture the diverse factors thatay influence the customer experience (see, e.g., Grewal, Levy,

nd Kumar 2009; Verhoef et al. 2009). They include macroactors, firm-controlled factors, consumer factors, and situa-ional factors (Grewal, Levy, and Kumar 2009; Verhoef et al.009). However, empirical research has actually never investi-ated and tested the many factors identified in these complex
onceptual models simultaneously. The present research oper-tionalizes retailing as a complex adaptive system includingrand, store, and consumer factors while controlling for marketactors. It quantified how these factors systematically moderate



Table 6Summary of findings.†

Hypothesis Finding Conclusion

H1a: holistic brandquality

When the retailer has created favorable beliefs aboutholistic brand quality in a market, customers weighfunctional and emotional experience clues less heavilycompared with markets with less favorable beliefs(negative moderating effect).

Negative moderating effect for twoemotional clues: frustration and ideasand inspiration.

Supported (3 of 5tests).a

Negative moderating effect for onefunctional clue: frontline employees.

H1b: service brandpromise

When the retailer has created high levels of belief in theservice brand promise in a market, customers weighfunctional and emotional experience clues more heavilycompared with markets with low levels of belief(positive moderating effect).

Positive moderating effect on twoemotional clues (ideas andinspiration and frustration). Nopositive moderating effects onfunctional clues.

Partially supported (2of 5 tests).a

H2a: hedonic valuecongruency

When the brand store image promises hedonic value (e.g.,pleasant and relaxing environment, products-I-like), customerswill weigh emotional clues (e.g., frustration, ideas andinspiration) more heavily and functional clues (e.g.,ease-of-use, frontline employees, waiting time) less heavily.

No positive moderating effects. Partially supportednegative moderatingeffect only (2 of 5tests).a

Negative moderating effects ofpleasant and relaxing on ease-of-useand “products-I-like” on frontlineemployees.

H2b: utilitarian valuecongruency

When the brand store image promises Utilitarian Value(e.g., easy-to-find), customers will weigh functionalclues (ease-of-use, frontline employees, and waitingtime) more heavily, and emotional clues (e.g.,frustration, ideas and inspiration) less heavily.

Positive moderating effect ofeasy-to-find on all two functionalclues (employees, waiting time).

Partially supportedpositive moderatingeffect only (2 of 5tests).No negative moderating effect of

easy-to-find.H3: online

purchaseavailable

Customers in markets where the retailer makes onlinepurchase available will weigh shared experience clues(e.g., ideas and inspiration, waiting time) more heavilyand unique experience clues (e.g., frontline employees)less heavily than customers in markets that do not offeronline purchase.

Increases the weight assigned to ideasand inspiration and waiting time

Partially supported (2of 3 tests).

H4a: utilitariangoal

Customers who are Buying (a utilitarian goal) weighfunctional clues (e.g., ease-of-use, frontline employees,waiting time) more heavily than customers pursuing ahedonic goal.

Negative moderating effect ofbrowsing (vs. buying) on frontlineemployee availability; the other twocoefficients are also negative but notsignificant.

Partially supported (1of 3)

H4b: hedonic goal Customers who are Browsing (a hedonic goal) weighemotional clues (e.g., frustration, ideas and inspiration)more heavily than customers pursuing a utilitarian goal.

Moderating effect of browsing(versus buying) for both emotionalclues: ideas and inspiration andfrustration.

Supported (2 of 2tests).

sts ar

ttfidolWaeo

S

ccptuo

esVeaitatsbs

U

s

† For the supported and partially supported hypotheses, 14 of 23 or 61% of tea Replicated in robustness study.

he influence of experience clues on customer satisfaction withhe service encounter. Our findings demonstrate that contextualactors within the retailer’s control (and some beyond its control)nfluence the retailer’s ability to deliver on its brand promiseuring the service encounter. They deepen our understandingf retail conditions and management opportunities and chal-enges from a systems perspective (e.g., Tax, McCutcheon, and

ilkinson 2013). They also provide insights about the multiple,ctionable ways retailers can manage and enhance the customerxperience as well as adapt to shopper goals and market factorsutside their control.

tore image as a retail brand differentiation mechanismA retailer’s physical store image contributes to the holistic

ustomer experience, but the introduction of digital channelshanges how consumers evaluate the experience clues in the
hysical store. It magnifies the importance of experience clueshat are similar across channels and diminishes the importance ofnique clues in the store. This finding highlights the importancef store image in a given market and demonstrates how experi-
ecbm

17

e statistically significant at p < .05.

nce clues can be designed and managed to influence customers’atisfaction judgments (Brakus, Schmitt, and Zarantonello 2009;erhoef et al. 2009). When retailers consistently deliver in-storexperience clues that are congruent with the brand’s positioningnd customer beliefs, they create a powerful store image. Storemage factors then moderate the importance of other clues inhe service encounter, suggesting new ways to design the brandnd manage retail service encounters to fit (rather than shape)he socio-economic factors, enabling customers to achieve theirhopping goals. Thus, all firms can capitalize on brand equityy mirroring customers’ brand beliefs and expectations in thetore image.

sefulness of the meta-analytic approach for a single brandThe present research makes a methodological contribution by

howing how a meta-analytic approach can identify diverse mod-
rating effects for a global brand. Meta-analysis is increasinglyommon in marketing research, especially for literature reviews,ut it is also used in estimating effects across multiple experi-ents in consumer research (McShane and Böckenholt 2017).


R

Osltdonitorac

M

aamcecwBtsama

M

ogsetitptNecbtstt2fsne

Rm

oraecblqsotbishp

H

eeoetpctaslphtap

A

sfitrboaitcct

.N. Bolton et al.

ur research can be used to better understand retail brands andervice by drawing on multiple, diverse data sources. We ana-yzed a single retailer’s 400 stores operating in 32 countries;his scope ensures that the findings on the moderating effects ofiverse retail brand contexts are robust. Meta-analyses that focusn a single retail brand allow researchers to control for a largeumber of potential confounding factors, isolating and measur-ng the effects of contextual factors. Our approach could be usedo identify (possibly differing) nuances in success factors forther retail brands. Large datasets are increasingly available toesearchers, and our approach provides a straightforward way ofssessing moderators without over-parameterizing models andonfounding effects.

anagerial Implications

To profitably manage a global brand, retailers must man-ge the contextual factors that are within their control whiledapting to those outside their control. Brand, store factors, andarket factors influence the importance of experience clues in

ustomers’ holistic evaluations of their service encounters andxplain considerable variance in the effect sizes of experientiallues. These findings are especially crucial for global brands,hich typically receive scarce research attention (Steenkamp,atra, and Alden 2003). For this global retailer’s managers,

hese findings confirm the importance of key success factorsuch as brand quality (which makes customers more forgiving)nd enhancing beliefs about its service brand promise (whichakes customers pay attention to the specific experience clues

ligned with the brand).

anaging branded service encounters by monitoring storesThis research can help to guide the design and management

f branded service encounters to suit the characteristics of aiven market. Service encounter management must focus on thealient aspects of retail experiences that managers can influ-nce. Customers learn the retailer’s brand promises about howheir stores offer hedonic (pleasant and attractive) and utilitar-an (easy-to-find) value, and they expect their service encounterso be congruent with these promises. The participating retailerrimarily promises utilitarian value and provides multiple func-ional experience clues that are congruent with this promise.evertheless, customers also consider emotional clues when

valuating their (holistic) service encounters, and these are espe-ially important for customers who are browsing rather thanuying. For this reason, the retailer must ensure that its emo-ional clues are also congruent with its brand promise. Thistudy shows how branded service encounters can heighten cus-omer attention in important ways, highlighting the value ofouchpoint monitoring for retailers (Homburg, Jozic, and Kuehnl017). By collecting store performance metrics and utilizing
eedback mechanisms, retailers can manage activities at theirtores to align with shoppers’ goals and attentional mecha-isms, thereby improving customer experience during servicencounters.
bprfi

18


etail branding strategies for new versus establishedarketsThe weight customers assign to experience clues will depend

n how the retailer has developed its brand over time. When aetailer enters a new market, customer beliefs and expectationsbout holistic brand quality and service brand promise are not yetstablished. Thus, it is crucial to perform well on all experiencelues. In this situation, expectancy-disconfirmation is likely toe important in customer satisfaction judgments. As customersearn about the retail brand, favorable beliefs about holistic branduality and the service brand promise can shift their attention topecific experience clues on subsequent store visits. Perceptionsf brand quality act as a buffer so customers become less atten-ive to incongruent clues. For example, the cooperating retailrand’s quality shields it (somewhat) from failure in providingdeas and inspiration, helpful frontline employees, and short in-tore waits. Concrete beliefs about the service brand promiseeighten customer attention to the experience clues used in itsromise to predict future satisfaction with the service encounter.

ow retailers can support shopper goalsBeyond shaping service encounters through brand and store

lements, retailers must align their strategies to support differ-nt shopper goals. To encourage browsing in stores and buyingnline, customers’ in-store experiences should be enriched bymotional clues (e.g., providing ideas and inspiration). In con-rast, if retailers seek to encourage online browsing and in-storeurchase or pick up, the retailer should emphasize functionallues (e.g., short in-store waiting time). Retailers cannot controlhe goals that customers choose to pursue, so they must providelternative paths that enable each segment of customers to pur-ue their goals successfully. One option is to design the storeayout and atmospherics so that specific areas of the store sup-ort each goal. For example, consumer electronic stores oftenave attractive in-store demonstration areas where browsing cus-omers can interact with products and employees, while otherreas are designed for customers searching for alternatives orurchasing.

dapting the store to the online channelWhen the retailer makes online purchasing available, con-

umers respond to experience clues in the store differently. Thisnding suggests that the cooperating retailer needs to strengthen

he design of shared experience clues, such as ideas and inspi-ation and waiting time. This implication is counter intuitiveecause the retailer might expect customers to seek inspirationnline so that it should focus on providing more help and guid-nce through frontline employees in the store. As this examplellustrates, this pattern of findings places further emphasis onhe importance of moving from multi-channel retailing to omni-hannel retailing (Verhoef, Kannan, and Inman 2015). If theonsumer places extra emphasis on the shared experience clues,hen the same experience clues are important, but they need not
e designed in the exact same way across channels. For exam-le, over the course of the present COVID-19 pandemic, severaletailers have increased their focus on the online channel. Ourndings suggest that when consumers start returning to the store,


R

tdmsf

A

agisumccmatiti

fgcrfiaut

sitwsperss

mivfircidh

te

scpfapemci(tp

fj

A

A

B

B

B

B

B

B

B

B

B

B

.N. Bolton et al.

heir response to key experience clues will be different. When theistinction between physical and online starts to vanish, retailersust perform well in all channels for key experience clues. We

peculate that consumers might be more forgiving if the retailerails on the unique experience clues.

dapting to marketsRetailers must align their brand strategy with market char-

cteristics. Our meta-analyses show that, in markets with highrowth rates, emotional clues, such as frustration and ideas andnspiration, are more important to customers, while ease-of-usetores and waiting time are less important (perhaps due to prod-ct scarcity). This finding highlights how customers in growingarkets seek hedonic value from retail service encounters. It is

onsistent with evidence that growth strategies require firms toonnect emotionally with customers (Reinartz et al. 2011). Inarkets with a high level of individualism (such as in the USA

nd Australia), frustration and waiting time are less importanto customers (perhaps due to the prevalence of self-service), butf uncertainty avoidance is high (as in Eastern European coun-ries), less frustration and offering ideas and inspiration are moremportant.

Limitations and Future Research

Our research has some limitations that open up avenuesor further research. First, although our meta-analyses inte-rated data from more than a million customers across 32ountries, these related to a single cooperating retailer. Ourobustness study indicates that effect sizes will vary under dif-erent conditions. Hence, replications and extensions that varyn method factors—by studying different brands, categories,nd stores—will make it possible to quantify the effect sizesnder different conditions and identify additional relevant fac-ors (McShane et al. 2019).

Second, although we relied on well-established satisfactioncales and measured dimensions of the customer experiencedentified in prior research, Lemon and Verhoef (2016) observehat scales measuring the entire customer experience are notell developed. Future research should develop comprehensive

cales for measuring the customer experience. For exam-le, expectancy-disconfirmation (i.e., better/same/worse thanxpected) is a key antecedent of customer satisfaction with theetail service experience. This study’s results suggest that ithould be measured at the attribute level rather than for the entireervice encounter.

Third, our meta-analyses integrated multiple data sources toeasure market, brand, store, and consumer factors and thereby

dentify ways in which retailers can design and manage ser-ice encounters. Since brand and store factors are somewhatrm- and industry-specific, more work is needed to expand andefine these factors. Interestingly, our post hoc analyses indi-
ated that a concrete construal level (service brand promise, storemage of easy-to-find) diminished the importance of expectancy-isconfirmation, whereas holistic beliefs about the brand or storeeighten its importance. Expectancy-disconfirmation is central
B

19


o understanding customer satisfaction with the retail servicencounter, so this finding warrants additional investigation.

Fourth, we worked with a large retailer that operates 400tores around the world. The retailer’s standardized store con-ept helped us control for sources of variation that would beresent in a study involving multiple retailers. The present studyocuses on variation across countries and stores rather than vari-tions in retail brand strategy (e.g., everyday low price versusremium brand strategy). Future research could examine differ-nt retail brand strategies. Fifth, this study focused on whetheroderating effects are consistently significant across different

ountries. We could not investigate variations due to changesn retail brand strategy. A longitudinal study of market factorssuch as internet penetration and urbanization) would be usefulo reveal how these and other global factors influence customerreference structures.

Appendix A. Supplementary data

Supplementary material related to this article can beound, in the online version, at doi:https://doi.org/10.1016/.jretai.2021.03.004.

References

ilawadi, Kusum L. and Kevin L. Keller (2004), “Understanding Retail Brand-ing: Conceptual Insights and Research Priorities,” Journal of Retailing, 80(4), 331–42.

ntweiler, Werner (2001), “Nested Random Effects Estimation in UnbalancedPanel Data,” Journal of Econometrics, 101 (2), 295–313.

erry, Leonard L. (2000), “Cultivating Service Brand Equity,” Journal of theAcademy of Marketing Science, 28 (1), 128–37.

erry, Leonard L., Lewis P. Carbone and Stephan H. Haeckel (2002), “Managingthe Total Customer Experience,” MIT Sloan Management Review, 43 (3), 85.

haradwaj, Sundar G., P. Rajan Varadarajan and John Fahy (1993), “Sustain-able Competitive Advantage in Service Industries: A Conceptual Model andResearch Propositions,” Journal of Marketing, 57 (4), 83–99.

hatnagar, Namita, Nicholas Lurie and Valarie A. Zeithaml (2003), ReasoningAbout Online and Offline Service Experiences: The Role of Domain-Specificity in the Formation of Service Expectations, ACR North AmericanAdvances.

itner, Mary Jo (1995), “Building Service Relationships: It’s All AboutPromises,” Journal of the Academy of Marketing Science, 23 (40), 246–51.

itner, Mary Jo and Helen S. Wang (2014), “Service Encounters in ServiceMarketing Research,” in Handbook of Service Marketing Research, RustR. T. and Huang M. H., eds. Cheltenham, UK: Edward Elgar Publishing,221–43.

loch, Peter and Marsha L. Richins (1983), “Shopping Without Purchase: AnInvestigation of Consumer Browsing Behavior,” in Advances in ConsumerResearch, 10, Bagozzi R. P. and Tybout A. M., eds. Ann Abor, MI: Associ-ation for Consumer Research, 389–93.

orenstein, Michael (2009), “Effect Sizes for Continuous Data,” in The Hand-book of Research Synthesis and Meta-Analysis, Cooper Harris, Hedges LarryV. and Valentine Jeffrey C., eds (2nd ed.). , 221–35.

orenstein, Michael, Larry V. Hedges, Julian P.T. Higgins and Han-nah Rothstein (2015), CMA (Comprehensive Meta Analysis) Version 3,https://www.meta-analysis.com/downloads/MRManual.pdf

oulding, William, Ajay Kalra and Richard Staelin (1999), “The Quality Double
Whammy,” Marketing Science, 18 (4), 463–84.
rakus, J. Josko, Bernd H. Schmitt and Lia Zarantonello (2009), “Brand Expe-rience: What Is It? How Is It Measured? Does It Affect Loyalty?,” Journalof Marketing, 73 (3), 52–68.



http://refhub.elsevier.com/S0022-4359(21)00025-7/sbref0005




























































































































































































































































































https://www.meta-analysis.com/downloads/MRManual.pdf














































R

B

C

D

D

D

E

E

E

G

G

G

G

H

H

H

H

H

H

H

H

I

J

K

K

L

L

L

M

M

M

N

N

O

O

O

P

P

P

R

R

S

S

S

.N. Bolton et al.

rodie, Roderick J., James R.M. Whittome and Gregory J. Brush (2009), “Inves-tigating the Service Brand: A Customer Value Perspective,” Journal ofBusiness Research, 62 (3), 345–55.

hitturi, Ravindra, Rajagopal Raghunathan and Vijay Mahajan (2007), “Formversus Function: How the Intensities of Specific Emotions Evoked in Func-tional versus Hedonic Trade-offs Mediate Product Preferences,” Journal ofMarketing Research, 44 (4), 702–14.

e Mooij, Mareike and Geert Hofstede (2011), “Cross-Cultural ConsumerBehavior: A Review of Research Findings,” Journal of International Con-sumer Marketing, 23 (3-4), 181–92.

har, Ravi and Eunice Y. Kim (2007), “Seeing the Forest or the Trees:” Implica-tions of Construal Level Theory for Consumer Choice,” Journal of ConsumerPsychology, 17 (2), 96–100.

har, Ravi and Itamar Simonson (2003), “The Effect of Forced Choice onChoice,” Journal of Marketing Research, 40 (2), 146–60.

rdem, Tülin, Joffre Swait, Susan Broniarczyk, Dipankar Chakravarti, Jean-Noel Kapferer, Michael Keane, John Roberts, Jan-Benedict E. Steenkampand Florian Zettelmeyer (1999), “Brand Equity, Consumer Learning andChoice,” Marketing Letters, 10 (3), 301–18.

rdem, Tülin, Joffre Swait and Ana Valenzuela (2006), “Brands as Signals: ACross-Country Validation Study,” Journal of Marketing, 70 (1), 34–49.

rdem, Tülin, Ying Zhao and Ana Valenzuela (2004), “Performance of StoreBrands: A Cross-Country Analysis of Consumer Store-brand Preferences,Perceptions, and Risk,” Journal of Marketing Research, 41 (1), 86–100.

illespie, Brian, Darrel D. Muehling and Ioannis Kareklas (2018), “Fitting Prod-uct Placements: Affective Fit and Cognitive Fit as Determinants of ConsumerEvaluations of Placed Brands,” Journal of Business Research, 82 (January),90–102.

rewal, Dhruv, Michael Levy and Donald R. Lehmann (2004), “Retail Brandingand Customer Loyalty: An Overview,” Journal of Retailing, 4 (80), ix–xii.

rewal, Dhruv, Michael Levy and Vijay Kumar (2009), “Customer ExperienceManagement in Retailing: An Organizing Framework,” Journal of Retailing,85 (1), 1–14.

rewal, Dhruv, Anne L. Roggeveen and Jens Nordfält (2017), “The Future ofRetailing,” Journal of Retailing, 93 (1), 1–6.

amilton, Rebecca W. and Debora V. Thompson (2007), “Is There a Substi-tute for Direct Experience? Comparing Consumers’ Preferences after Directand Indirect Product Experiences,” Journal of Consumer Research, 34 (4),546–55.

artman, Katherine B. and Rosann L. Spiro (2005), “Recapturing Store Imagein Customer-Based Store Equity: A Construct Conceptualization,” Journalof Business Research, 58 (8), 1112–20.

ofstede, Geert (2003), Culture’s Consequences: Comparing Values, Behaviors,Institutions and Organizations Across Nations, 2nd edition Sage Publica-tions.

omburg, Christian, Danijel Jozic and Christina Kuehnl (2017), “CustomerExperience Management: Toward Implementing an Evolving MarketingConcept,” Journal of the Academy of Marketing Science, 45 (3), 377–401.

ong, Weiyin, James Y.L. Thong and K.Y. Tam (2004), “The Effects of Informa-tion Format and Shopping Task on Consumers’ Online Shopping Behavior:A Cognitive Fit Perspective,” Journal of Management Information Systems,21 (3), 149–84.

sieh, Ming-Huei, Shan-Ling Pan and Rudy Setiono (2004), “Product-,Corporate-, and Country-Image Dimensions and Purchase Behavior: AMulti-Country Analysis,” Journal of the Academy of Marketing Science,32 (3), 251–70.

unneman, Auke, Peter C. Verhoef and Laurens M. Sloot (2015), “The Impact ofConsumer Confidence on Store Satisfaction and Share of Wallet Formation,”Journal of Retailing, 91 (3), 516–32.

unter, John E. and Frank L. Schmidt (2000), “Fixed Effects vs. Random EffectsMeta-Analysis Models: Implications for Cumulative Research Knowledge,”International Journal of Selection and Assessment, 8 (4), 275–92.

zard, Carroll E. (1991), The Psychology of Emotions, Springer Science &Business Media.

ohnson, Blair T. and Tania B. Huedo-Medina (2013), Meta-Analytic Statisti-cal Inferences for Continuous Measure Outcomes as a Function of EffectSize Metric and Other Assumptions, Rockville, MD: Agency for HealthcareResearch and Quality. Report No.: 13-EHC075-EF

S

20


ardes, Frank R., Steven S. Posavac and Maria L. Cronley (2004), “ConsumerInference: A Review of Processes, Bases, and Judgment Contexts,” Journalof Consumer Psychology, 14 (3), 230–56.

eller, Kevin L. (2003), “Brand Synthesis: The Multidimensionality of BrandKnowledge,” Journal of Consumer Research, 29 (4), 595–600.

emon, Katherine N. and Peter C. Verhoef (2016), “Understanding CustomerExperience Throughout the Customer Journey,” Journal of Marketing, 80(6), 69–96.

ichtenstein, Donald R., Richard G. Netemeyer and Scot Burton (1995),“Assessing the Domain Specificity of Deal Proneness: A Field Study,” Jour-nal of Consumer Research, 22 (3), 314–26.

ynch, John G. Jr and Thomas K. Srull (1982), “Memory and Attentional Fac-tors in Consumer Choice: Concepts and Research Methods,” Journal ofConsumer Research, 9 (1), 18–37.

ason, Charlotte H. and William D. Perreault Jr (1991), “Collinearity, Power,and Interpretation of Multiple Regression Analysis,” Journal of MarketingResearch, 28 (3), 268–80.

cShane, Blakeley B. and Ulf Böckenholt (2017), “Single Paper Meta-Analysis:Benefits for a Study Summary, Theory Testing and Replicability,” Journalof Consumer Research, 43 (6), 1048–63.

cShane, Blakeley B., Jennifer L. Tackett, Ulf Böckenholt and Andrew Gelman(2019), “Large-Scale Replication Projects in Contemporary PsychologicalResearch,” The American Statistician, 73 (sup1), 99–105.

yer, Prashanth U. (1997), “A Study of the Relationships Between Cogni-tive Appraisals and Consumption Emotions,” Journal of the Academy ofMarketing Science, 25 (4), 296–304.

yffenegger, Bettina, Harley Krohmer, Wayne D. Hoyer and Lucia Malaer(2015), “Service Brand Relationship Quality: Hot or Cold?,” Journal ofService Research, 18 (1), 90–106.

liver, Richard L. (2014), Satisfaction: A Behavioral Perspective on the Con-sumer: A Behavioral Perspective on the Consumer, United Kingdom:Routledge.

stroff, Cheri (1993), “Comparing Correlations Based on Individual-Level andAggregated Data,” Journal of Applied Psychology, 78 (4), 569–82.

strom, Amy L., A. Parasuraman, David E. Bowen, Lia Patrício and ChristopherA. Voss (2015), “Service Research Priorities in a Rapidly Changing Context,”Journal of Service Research, 18 (2), 127–59.

astor, Dena A. and Rory A. Lazowski (2018), “On the Multilevel Nature ofMeta-Analysis: A Tutorial, Comparison of Software Programs, and Dis-cussion of Analytic Choices,” Multivariate Behavioral Research, 53 (1),74–89.

ayne, John W., James R. Bettman and Eric J. Johnson (1992), “BehavioralDecision Research: A Constructive Processing Perspective,” Annual Reviewof Psychology, 43 (1), 87–131.

uccinelli, Nancy M., Ronald C. Goodstein, Dhruv Grewal, Robert Price, PriyaRaghubir and David Stewart (2009), “Customer Experience Management inRetailing: Understanding the Buying Process,” Journal of Retailing, 85 (1),15–30.

einartz, Werner, Benedict Dellaert, Manfred Krafft, V. Kumar and RajanVaradarajan (2011), “Retailing Innovations in a Globalizing Retail MarketEnvironment,” Journal of Retailing, 87 (S 1), S53–66.

ichins, Marsha L. (1997), “Measuring Emotions in the Consumption Experi-ence,” Journal of Consumer Research, 24 (2), 127–46.

chmitt, Bernd, J. Josko Brakus and Lia Zarantonello (2015), “From ExperientialPsychology to Consumer Experience,” Journal of Consumer Psychology, 25(1), 166–71.

eiders, Kathleen, Glenn B. Voss, Dhruv Grewal and Andrea L. Godfrey (2005),“Do Satisfied Customers Buy More? Examining Moderating Influences in aRetailing Context,” Journal of Marketing, 69 (4), 26–43.

, , Andrea L. Godfrey and DhruvGrewal (2007), “SERVCON: Development and Validation of a Multidimen-sional Service Convenience Scale,” Journal of the academy of MarketingScience, 35 (1), 144–56.

iemsen, Enno, Aleda Roth and Pedro Oliveira (2010), “Common Method
Bias in Regression Models with Linear, Quadratic, and Interaction Effects,”Organizational Research Methods, 13 (3), 456–76.
irianni, Nancy J., Mary Jo Bitner, Stephen W. Brown and Naomi Mandel(2013), “Branded Service Encounters: Strategically Aligning Employee




























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































R

S

S

S

S

S

S

T

T

T

V

V

V

V

V

V

W

Ypetitive Advantage, Englewood Cliffs, NJ: Prentice-Hall.

Zomerdijk, Leonieke G. and Christopher A. Voss (2010), “Service Design

.N. Bolton et al.

Behavior with the Brand Positioning,” Journal of Marketing, 77 (6),108–23.

lovic, Paul and Sarah Lichtenstein (1983), “Preference Reversals: A BroaderPerspective,” The American Economic Review, 73 (4), 596–605.

nelders, Dirk and Jan P.L. Schoormans (2004), “An Exploratory Study of theRelation Between Concrete and Abstract Product Attributes,” Journal ofEconomic Psychology, 25 (6), 803–20.

teenkamp, Jan-Benedict E.M. (2001), “The Role of National Culture in Interna-tional Marketing Research,” International Marketing Review, 18 (1), 30–44.

teenkamp, Jan-Benedict B.E.M., Rajeev Batra and Dana Alden (2003), “HowPerceived Brand Globalness Creates Brand Value,” Journal of InternationalBusiness Studies, 34 (1), 53–65.

wait, Joffre and Wiktor Adamowicz (2001), “Choice Environment, Mar-ket Complexity, and Consumer Behavior: A Theoretical and EmpiricalApproach for Incorporating Decision Complexity into Models of ConsumerChoice,” Organizational Behavior and Human Decision Processes, 86 (2),141–67.

zymanski, David M. and David H. Henard (2001), “Customer Satisfaction:A Meta-Analysis of the Empirical Evidence,” Journal of the Academy ofMarketing Science, 29 (1), 16–35.

alukdar, Debabrata, K. Sudhir and Andrew Ainslie (2002), “Investigating NewProduct Diffusion Across Products and Countries,” Marketing Science, 21(1), 97–114.

ax, Stephen S., David McCutcheon and Ian F. Wilkinson (2013), “The ServiceDelivery Network (SDN): A Customer-Centric Perspective of the Customer
Journey,” Journal of Service Research, 16 (4), 454–70.
aylor, Steven A. (1997), “Assessing Regression-Based Importance Weights forQuality Perceptions and Satisfaction Judgements in the Presence of HigherOrder and/or Interaction Effects,” Journal of Retailing, 73 (1), 135–59.

21


an Birgelen, Marcel, Ad De Jong and Ko De Ruyter (2006), “Multi-ChannelService Retailing: The Effects of Channel Performance Satisfaction onBehavioral Intentions,” Journal of Retailing, 82 (4), 367–77.

erhagen, Tibert and Willemijn Van Dolen (2009), “Online Purchase Intentions:A Multi-Channel Store Image Perspective,” Information & Management, 46(2), 77–82.

erhoef, Peter C., Katherine N. Lemon, A. Parasuraman, Anne Roggeveen,Michael Tsiros and Leonard A. Schlesinger (2009), “Customer ExperienceCreation: Determinants, Dynamics and Management Strategies,” Journal ofRetailing, 85 (1), 31–41.

erhoef, Peter C., Pallassana K. Kannan and J. Jeffrey Inman (2015), “FromMulti-Channel Retailing to Omni-Channel Retailing: Introduction to theSpecial Issue on Multi-Channel Retailing,” Journal of Retailing, 91 (2),174–81.

oss, Glenn B., Andrea Godfrey and Kathleen Seiders (2010), “How Com-plementarity and Substitution Alter the Customer Satisfaction–RepurchaseLink,” Journal of Marketing, 74 (6), 111–27.

oss, Kevin E., Eric R. Spangenberg and Bianca Grohmann (2003), “Measuringthe Hedonic and Utilitarian Dimensions of Consumer Attitude,” Journal ofMarketing Research, 40 (3), 310–20.

eber, Elke U. and Eric J. Johnson (2009), “Mindful Judgment and DecisionMaking,” Annual review of psychology, 60, 53–85.

ip, George S. (1995), Total Global Strategy: Managing for World-wide Com-

for Experience-Centric Services,” Journal of Service Research, 13 (1),67–82.






























































































































































































































































































































































































































































































No.of Pages21 ARTICLE IN PRESS

Documents