Personalization and hedonic motivation in creating customer experiences and loyalty in omnichannel retail

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Personalization and hedonic motivation in creating customer experiences and loyalty in omnichannel retail © 2020 the Authors Published version Tyrväinen, Olli; Karjaluoto, Heikki; Saarijärvi, Hannu Tyrväinen, O., Karjaluoto, H., & Saarijärvi, H. (2020). Personalization and hedonic motivation in creating customer experiences and loyalty in omnichannel retail. Journal of Retailing and Consumer Services, 57, Article 102233. https://doi.org/10.1016/j.jretconser.2020.102233 2020 Journal of Retailing and Consumer Services 57 (2020) 102233 Available online 14 August 2020 0969-6989/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Personalization and hedonic motivation in creating customer experiences and loyalty in omnichannel retail Olli Tyrv¨ ainen a , * , Heikki Karjaluoto b , Hannu Saarij¨ arvi c a Jyv¨ askyl¨ a University School of Business and Economics, P.O.Box 35, FI-40014, University of Jyv¨ askyl¨ a, Finland b Jyv¨ askyl¨ a University School of Business and Economics, University of Jyv¨ askyl¨ a, Finland c School of Management University of Tampere, Finland ARTICLE INFO Keywords: Retail Customer experience Omnichannel experience Personalization Hedonic motivation ABSTRACT This study examines the effects of personalization and hedonic motivation on customer experience and its loyalty outcomes in omnichannel retail context. The study develops eight hypotheses which are tested using two survey samples (Finland (n =2084) and Sweden (n =2334). In addition, empirical analysis includes 20 semi-structured interviews. The findings support all the hypotheses confirming the positive relationships personalization and hedonic motivation have on cognitive and emotional customer experience components. Further, the positive effects of customer experience on loyalty are confirmed. The results provide both theoretical and managerial insights for improved CX and customer loyalty. 1. Introduction Digitalization is reconfiguring retailing. This retail transformation is not only changing the channel of transaction but is redefining the nature and types of exchanges, actors, offerings, and retail settings (Hagberg et al., 2016). The integration of digital technologies into the consumer shopping experience (see Reynolds and Sundstr¨ om, 2014) is enabling new means of value creation (e.g., Pantano and Viassone, 2015; Saarij¨ arvi et al., 2014) and value capture (e.g., H¨ anninen et al., 2018; Hure et al., 2017). For example, shoppers can be offered various digital touchpoints on their path to purchase with the aim of supporting their purchase process, which will simultaneously allow the collection of information regarding their needs and preferences (Blom et al., 2017; Yurova et al., 2017). Consequently, consumers use complementary channels as an integral part of their shopping experience (see Beck and Rygl, 2015; Dholakia et al., 2010) with the aim of optimizing different channels’ benefits and costs (Lemon and Verhoef, 2016). This has exerted pressure on retailers to redesign their channels according to consumers’ processes. One outcome of this evolution is that, when considering customer experiences (CXs), retailers are striving toward omnichannel retailing, which is defined as “the synergetic management of the numerous available channels and customer touchpoints, in such a way that the CX across channels and the performance over channels is optimized” (Verhoef et al., 2015, pp. 175–176). Per the McKinsey Company (2019), omnichannel shoppers represent one-third of all shopping journeys. Toward that end, prior literature has discussed the importance of synchronizing bricks and clicks in enabling omnichannel customer journeys that are valued by consumers and that drive satisfaction, loyalty, and word of mouth (WOM) (Kumar and Reinartz, 2016; Leroi-Werelds et al., 2014). Moreover, recent studies have identified a positive relationship between successful channel integration, a firm’s performance (Blom et al., 2017; Oh et al., 2012), and consumers’ benefits (Grewal et al., 2017). Conversely, Briel (2018) has discussed the increasing demand for personalized experiences and how megatrends, such as the expanded use of mobile devices (e.g., smartphones), artificial intelligence, and real-time big data analytics, play an increasingly important role in personalization efforts toward CXs in omnichannel retail settings. Personalization has become even more important in the omnichannel context because the integration of channels holds great potential in delivering more personalized CXs (H¨ anninen et al., 2019). However, despite retailers’ increased spending on personalization systems, there is still a lack of empirical research on investment returns concerning personalization (Kalaignanam et al., 2018). More specifically, Salonen and Karjaluoto (2016) state that the focus of personalization studies should move toward integration of psychological elements, such as emotions, into personalization. Prior research has mainly examined the relationships within * Corresponding author. E-mail addresses: [email protected] (O. Tyrv¨ ainen), [email protected] (H. Karjaluoto), [email protected] (H. Saarij¨ arvi). Contents lists available at ScienceDirect Journal of Retailing and Consumer Services journal homepage: http://www.elsevier.com/locate/jretconser https://doi.org/10.1016/j.jretconser.2020.102233 Received 3 December 2019; Received in revised form 11 June 2020; Accepted 12 July 2020 Journal of Retailing and Consumer Services 57 (2020) 102233 2 personalized interactions and CXs in specific areas, such as mobile and online channels (Bilgihan et al., 2015; McLean et al., 2018; Pappas et al., 2017; Rose et al., 2012) and has relied heavily on single-channel research settings. In that respect, there have been recent calls for more empirical research to investigate omnichannel shopping in general (Hure et al., 2017) and how omnichannel strategies can enhance CXs in particular (Souiden et al., 2019). Lemon and Verhoef (2016) have suggested that future studies should focus on the relationship between the combined effects of CXs at multiple touchpoints and customer behaviors, such as loyalty intention. Hence, while hedonic motivation (HM) is as a key driver of online shopping (Arnold and Reynolds, 2003; Bridges and Florsheim, 2008), prior research has produced contradictory results about the relationship between HM and CXs (Arnold and Reynolds, 2012; Evanschitzky et al., 2014; Hubert et al., 2017; Izogo and Jayawardhena, 2018; O’Brien, 2010). Notably, Gilboa et al. (2016) stated that this relationship requires further examination in a retail context. In prior studies, researchers have investigated the role of hedonic motivation in online and offline stores. Scarpi et al. (2014) found that hedonic shopping motivation has more important role in online context. Shen et al. (2016) show that consumers prefer shopping utilitarian products from offline stores and hedonic products from online stores. In this study, we contribute to this discussion from omnichannel viewpoint. Our objective is to consider the impact of hedonic motivation on cognitive and emotional CX viewpoints in omnichannel context. In this study, we extend the abovementioned prior research by linking HM and personalization to CX and loyalty outcomes in omnichannel research setting. Understanding these interlinkages is key for addressing several important managerial questions that arise in the omnichannel environment, including deeper insight into what consumers want in such an environment. To address this purpose, we first define the key study constructs and discuss relevant studies that are related to our research framework. This is followed by the presentation of our pilot study (Study 1; n =20), which was used to further develop the study hypotheses. Study 2 presents the development and testing of the study hypotheses. We conclude this paper with a discussion of the findings’ conceptual implications, the ways in which the findings might alter retail practices, and directions for future studies. 2. Theoretical background 2.1. Customer experience A CX is defined as a consumer’s internal and subjective response to any direct or indirect contact with a company (Brakus et al., 2009; Gentile et al., 2007; Meyer and Schwager, 2007). It comprises the search, purchase, consumption, and other after-sale phases, which are elements that the retailer may or may not be able to control (Verhoef et al., 2009). CXs occur when consumers search and shop for products, receive services, and consume both (Brakus et al., 2009). A retail CX is a combination of many elements in various retail channels (Piotrowicz and Cuthbertson, 2014). Gentile et al. (2007) introduced six CX components: cognitive, emotional, sensorial, pragmatic, lifestyle, and relational. The study of cognitive and affective CX components is common in prior research (Frow and Payne, 2007; Rose et al., 2012; Tynan and McKechnie, 2009). We follow the logic of Rose et al. (2012), who stated that the online customer interprets online stores’ stimuli from both the cognitive and the emotional perspectives and uses these components to form the core construct of his/her CX. Cognitive experience refers to the flow state that occurs during shopping (Hoffman and Novak, 2009). Flow refers to states of enjoyment, involvement, and concentration (Huang, 2006) that lead to a positive, subjective experience (Hoffman and Novak, 2009). A cognitive CX results from the internal processing of incoming stimuli, such as either the review of incoming information (Frow and Payne, 2007) or online interactions, such as navigation (Novak et al., 2000). The cognitive experience results in the customer learning and developing new skills (Tynan and McKechnie, 2009). An emotional experience is a consumer’s emotional response to stimuli (Hansen, 2005). An emotional CX is entertaining, and it generates enjoyment (Tynan and McKechnie, 2009). Rose et al. (2012) viewed the CX’s affective state as the moods, feelings, and emotions that are generated by a customer’s affective system. The experience makes the customer feel good and enriched (Lemke et al., 2011). An emotional experience, whether positive or negative, influences a consumer’s information processing and builds long-term associations in a consumer’s memory, which can lead to certain behaviors, such as recommendations (Edvardsson, 2005; Lemke et al., 2011). 2.2. Hedonic motivation Jin and Kim (2003) defined shopping motivations as drivers that bring consumers to marketplaces to satisfy needs. Holbrook and Hirschman (1982) divided consumers’ shopping motivations into utilitarian and hedonic dimensions. Hedonic shopping motivation presents behavior that is related to fun, amusement, fantasy, and the sensorial stimuli aspects of consumption (Babin et al., 1994). HM is defined as a key predictor of intention to purchase in consumer behavior research (Holbrook and Hirschman, 1982) and has recently also been associated with customer satisfaction and loyalty (Vieira et al., 2018). In the technology context, HM is defined as fun or pleasure that results from technology use (Venkatesh et al., 2012). However, HM varies across different retail formats (Arnold and Reynolds, 2003). For example, in the physical store context, hedonic shoppers value enjoyment, entertainment, and exploration, whereas in the online context, they respect adventure, authority, and status (To et al., 2007). Our study focuses on HM as a component of CX. 2.3. Personalization Personalization involves providing customized content and services based on customer data (Adomavicius and Tuzhilin, 2005) and adapting offers to meet a customer’s needs (Lemke et al., 2011). In traditional offline stores, personalization refers to serving customers in a face-to-face manner to satisfy their needs (Shen and Ball, 2009). In the online context, retailers can track customers’ previous purchase habits with personalization technologies; based on this amassed data, they can then modify what to display and how to display it (Zhang et al., 2011). Personalization aims to offer the most suitable products at the optimal time and in the best place to please customers (Sunikka and Bragge, 2012). It provides benefits for both retailers and customers because individualized products, services, and interactions appeal to consumers (Ansari and Mela, 2003). Personalization also offers transaction flexibility, more targeted banner advertisements and websites, and product recommendations (Kalaignanam et al., 2018). Per Choi et al. (2017), personalizing pricing information with location-based systems results in financial gains for consumers, and personalized pricing helps consumers make more informed decisions. In addition, the quality and benefits of personalization have been shown to increase the intention to purchase (Pappas et al., 2016). In this study, we are interested in consumer perceptions of personalization in the retail environment as a driver of CXs. 2.4. Word of mouth Per Zeithaml et al. (1996), WOM is a component of customer loyalty. They defined WOM as a consumer’s likelihood to recommend and say positive things about a company to other consumers. Today, the Internet is a place for consumers to interact with one another and share their experiences (Bilgihan, 2016) on different platforms, such as social media. Compared to traditional face-to-face WOM, social media WOM differs in its non-simultaneous nature (Balaji et al., 2016), and electronic channels also provide opportunities to have discussions anonymously. O. Tyrv¨ ainen et al. Journal of Retailing and Consumer Services 57 (2020) 102233 3 Chen et al. (2017) showed that WOM in the form of online product reviews can positively affects sales. For example, negative electronic WOM can spread widely and have harmful effects on a firm’s reputation and sales (Balaji et al., 2016). In this study, we follow the same logic as Zeithaml et al. (1996); thus, we operationalize WOM as a loyalty outcome of the omnichannel CX. 2.5. Repeat Purchase Intention RPI indicates the likelihood of engaging in repurchasing behavior (Rose et al., 2012). Chiu et al. (2014) defined it as the subjective probability that a customer will purchase a product from the same retailer. RPI is widely used as a behavioral outcome in retail studies (Bilgihan, 2016; Chiu et al., 2014; Rose et al., 2012; Verhagen and van Dolen, 2009). Similar to WOM, it is conceptualized as a component of customer loyalty (Zeithaml et al., 1996). In this study, RPI is operationalized as an outcome of CX. 3. Pilot study We applied a qualitative pretest to conceptualize and identify the key themes that are related to our research model (Fig. 1). 3.1. Procedure In-depth consumer interviews (n =20) were conducted among consumers with rich prior experience in omnichannel retail in Finland (FIN) during September and October of 2018. The participants were selected via the purposeful criterion sampling method to gather knowledge about the research topic (Patton, 2002, pp. 40–46). All participants had experience as omnichannel customers (i.e., they had visited both online and offline stores of the same retailer). The definition of omnichannel retail was explained to the respondents before each interview session, and they were guided to reflect on their recent purchasing process regarding one specific retailer. The interviewees’ demographic information is presented in Table 1. The interview protocol was designed to lead the interview. Semistructured questions left space for deeper discussion and allowed the interviewees to express their own viewpoints (J¨ arvenp¨ a¨ a and Lang, 2005). All interviews were conducted in Finnish. Thus, semi-structured questions were translated from English to Finnish. The interviews were audio-recorded and transcribed into text files after each session, and the transcription was translated from Finnish to English. The average time for each interview was 18 min. In the data analysis, we followed the three-stage content analysis protocol of Miles and Huberman (1994), including data reduction, data display, and conclusion drawing. Interview responses were read carefully and then coded and organized according to the main themes, which were based on the literature review. 3.2. Results The results from the interviews are presented in Table 2. Regarding HM, our results imply its interconnection with CXs. Briefly, the interviewees described a positive CX as having hedonic components. HM was associated with unplanned spending and impulse purchases. In addition, customers appreciated a retailer’s personalized content, such as personalized advertisements, offers, and product recommendations. These were closely related to a positive CX and a consumer’s purchase behavior in the future. The interviewees were expecting retailers to utilize amassed data on their previous purchases to personalize content. Lastly, the findings suggest that a positive CX is linked to loyalty outcomes. Per the interviewees, a positive CX and positive WOM, such as recommendations from within one’s family circle, increased their purchase intentions in the same store. By contrast, repeated negative CXs were associated with ending the customer relationship with the retailer. 4. Survey To further investigate the CX’s relationships to its antecedents and outcomes, we developed eight hypotheses. Our research model (Fig. 1) Fig. 1. Research model. Table 1 Demographic profile of interviewees. Gender Occupation status Age Female Student 19 Male Student 21 Female Student 21 Female Student 21 Female Student 22 Male Student 22 Male Working 23 Male Working 23 Male Student 23 Female Student 23 Female Doctoral student 24 Male Working 25 Female Working 25 Male Self-employed 27 Male Doctoral student 27 Male Working 27 Female Working 28 Male Working 29 Female Working 31 Female Working 31 O. Tyrv¨ ainen et al. Journal of Retailing and Consumer Services 57 (2020) 102233 4 posits both HM and personalization as independent variables and demonstrates how they are linked with CX components and how CX affects both WOM and RPI. 4.1. Hypothesis development 4.1.1. The impact of hedonic motivation on customer experience Previous studies have shown the importance of HM during a customer’s shopping journey; researchers have proven this relationship in the contexts of offline search behavior (Arnold and Reynolds, 2003), time spent browsing in online stores (Kim and Eastin, 2011; Menon and Kahn, 2002), and intention to use mobile retail services (Shaw and Sergueeva, 2019). In this study, we want to examine HM’s relationship with CXs. HM has been shown to influence a CX’s emotional and cognitive components. The relationship has been studied in the retail context and in individualistic and collectivistic consumer cultures (Arnold and Reynolds, 2012; Evanschitzky et al., 2014; Izogo and Jayawardhena, 2018; O’Brien, 2010). Evanschitzky et al. (2014) showed that HM influences flow experience. Arnold and Reynolds’ (2012) study of the relationship between HM and hedonic shopping value found that consumers who are searching for hedonic experiences usually find them. Hubert et al. (2017) hypothesized that HM can be better served if retail allows for a seamlessly integrated experience. Their examination of the relationship between HM and experience response in the mobile shopping context did not find a significant effect. However, their focus was on pure mobile channels, which do not represent the entire omnichannel concept and are usually related to seamless experiences (Verhoef et al., 2015). Based on the above discussion, we propose the following hypotheses: H1a. HM has a positive direct effect on emotional CX. H1b. HM has a positive direct effect on cognitive CX. 4.1.2. The impact of personalization on customer experiences Creating personalization for customers helps retailers meet customers’ needs, which will positively influence RPI (Ha et al., 2010; Lemke et al., 2011). Personalization has been shown to decrease customers’ searches and product evaluation costs, which subsequently increase customer loyalty. Moreover, it reduces customers’ shopping efforts (Kim and Baek, 2018; Zhang et al., 2011). Personalization is positively related to CXs (Bilgihan et al., 2015; McLean et al., 2018; Pappas et al., 2017; Rose et al., 2012). Bilgihan et al. (2015) stated that recommendation systems with personalization features can engage customers in online stores. McLean et al. (2018) examined the relationship between customization and CXs within retail mobile applications and proposed that customization directly influences CXs. Rose et al. (2012) studied the relationship between web store customization and the affective components of CXs; they found that providing personalized, relevant content can result in positive customer emotions and thus influence the affective CX component. In addition, personalization increases the customer’s feeling of control and makes him/her part of creating the experience (Chang et al., 2010), which both affect the cognitive CX. Based on the above discussion, the following hypotheses are proposed for empirical testing: H2a. Personalization has a positive direct effect on emotional CX. H2b. Personalization has a positive direct effect on cognitive CX. 4.1.3. The impact of customer experience on word of mouth Brakus et al. (2009) confirmed the positive relationship between CXs and customer loyalty intentions. In this study, we focus on WOM and RPI, which have been verified as indicators of customer loyalty (Zeithaml et al., 1996). Prior studies have confirmed the relationship between CXs and WOM (Bilgihan, 2016; Keiningham et al., 2007; Klaus and Maklan, 2013; Klein et al., 2016; Maklan and Klaus, 2011). Both the emotional CX (Klaus and Maklan, 2013; Lovett et al., 2013) and the cognitive CX (Klaus and Maklan, 2013) drive positive WOM. Positive experiences influence customer loyalty intentions, and after a positive experience, consumers are more likely to recommend the brand to others (Brakus et al., 2009). Klaus and Maklan (2013) show a direct positive relationship between CX quality and WOM. Similar results were presented by Jones et al. (2006), who found that WOM is reflected in the CX with a retailer. Zhang et al. (2017) found that cognitively and affectively engaged customers are more likely to promote their brand communities. Lovett et al. (2013) studied drivers of WOM in both online and offline channels and found that consumers prefer to share their positive feelings about and experiences with a retailer. For example, brands that are considered more exciting received more WOM in their study. In addition, customers’ cognitive processes have been shown to have a positive relationship with their intentions toward recommendation (Zhang et al., 2017). Based on the preceding discussion, we advance the following hypotheses: H3a. Emotional CX has a positive direct effect on WOM. H3b. Cognitive CX has a positive direct effect on WOM. 4.1.4. The impact of customer experience on repeat purchase intention RPI has been widely used as an outcome of CXs in prior research (Bilgihan, 2016; Chiu et al., 2014; Kim and Han, 2014; Rose et al., 2012). Flow experience and purchase intention have a significant direct relationship in that shoppers remember their positive experiences, and customers who experience flow during their online shopping will be more willing to repurchase in the same store (Bilgihan, 2016; Chiu et al., 2014; Kim and Han, 2014). Purchase experience via the Internet predicts both the positive and direct intention to use the Internet for repurchasing (Shim et al., 2001). Emotional experiences influence a customer’s loyalty intentions, and the quality (positive or negative) of each experience directly affects RPI (Gountas and Gountas, 2007; Ou and Verhoef, 2017). Rose et al. (2012) estimated the total effects of both affective CX and cognitive CX on RPI and found a significant effect. Therefore, the following can be hypothesized: H4a. Emotional CX has a direct positive effect on RPI. H4b. Cognitive CX has a direct positive effect on RPI. Table 2 Study 1 findings. Theme Summary of findings Hedonic motivation Positive customer experience is described by hedonic components. Elements like “positive feelings,” “endorphins,” “happiness,” and “laugh” came forth in interviews when the interviewees were asked to describe a positive customer experience. High hedonic motivation during the shopping journey is related to change, including unplanned spending and impulse purchases. Personalization Consumers appreciate personal offers and advertisements via mobile phones and email. Product recommendations based on previous purchases and personal preferences lead to new purchases. Personalized advertisements, such as those in social media and retailers’ newsletters, are mentioned to improve a customer’s experience. Personalized advertising content is positively related to purchase behavior. Loyalty outcomes Positive customer experiences influence customer loyalty intentions. Consumers will recommend a retailer and purchase again at the same store. “I will visit again”; “it will be my first choice next time”; and “I will recommend it” were typical answers when interviewees were asked how a positive experience influences their behavior. Seriously negative customer experiences may end customer relationships. O. Tyrv¨ ainen et al. Journal of Retailing and Consumer Services 57 (2020) 102233 5 4.2. Methodology 4.2.1. Data collection Our objective was to study omnichannel CXs. Thus, our research setting was designed to guide the survey respondents to reflect on their recent purchasing process in one specific store, which was either partly or entirely conducted in an online store or an offline store. We explained the meaning of the essential terminology, the purchasing process, and both offline and online stores at the beginning of the survey form. The target population of the survey was defined as retail customers in FIN and Sweden (SWE). To achieve a suitable sample, we employed a market research firm to recruit the participants. The hypotheses were tested by using different samples from the two countries. The study was conducted in FIN and SWE. The results were based on data that were gathered during December 2018. A total of 4418 valid responses were received: FIN (n =2084) and SWE (n =2334). The demographic profile of the sample is shown in Table 3. A total of 44.0% FIN and 43.8% SWE respondents were male (49.3% of the entire population is male in FIN [Statistics Finland, 2019] and 50.3% is male in SWE [Statistics Sweden, 2019]). The sample was skewed toward young consumers (15–25 years old: 39.9% in FIN, 56.2% in SWE), whereas the percentage of this age group is 11.2% in FIN (Statistics Finland, 2019) and 11.3% in SWE (Statistics Sweden, 2019). The mean household income was between 40,000 and 50,000 € for the FIN respondents and between 30,000 and 40,000 € for the SWE respondents (the average household incomes for 2017 were 38,300 € in FIN (Statistics Finland, 2019) and 36,190 € in SWE (Statistics Sweden, 2019). Therefore, the FIN sample was slightly skewed toward a higher income, while the SWE sample represented the overall Swedish population. The mean household size was two in both samples; therefore, both samples represented the overall populations of these countries from this perspective. 4.2.2. Method biases Per Hulland et al. (2018) the nature of non-response bias should be considered by comparing the demographics of the respondent group to those of the sampling frame. The FIN and SWE populations were compared to our sample in Table 3. Thus, nonresponse bias was unlikely. Common method bias refers to contexts in which a respondent could give a construct’s true score, but because of a measurement method, he/ she does not (Kock, 2015). Common method variance (CMV) is often present in self-reported survey studies that use a single data source (Podsakoff et al., 2003). To ensure that CMV did not threaten our results, we structured the survey as follows: items in the questionnaire were placed in an alternating order; we separated the predictor and criterion variables; and we hid the respondents’ identities (Hulland et al., 2018; Podsakoff et al., 2003). In addition, the respondents were allowed to answer anonymously and were assured that there were no right or wrong answers; therefore, they were asked to answer all questions as honestly as possible. 4.2.3. Measurement scales All variables were measured on a 7-point Likert scale, ranging from “strongly disagree” to “strongly agree,” except for the variable WOM, which was measured on a 10-point scale. Previously validated instruments were used to measure the items that represented the study variables (Appendix A). For measurement, three items for measuring cognitive CX four items for measuring emotional CX were adopted from Brakus et al. (2009); three items for measuring HM were adopted from Venkatesh et al. (2012); three items for measuring personalization were adopted from Pappas et al. (2014); three items for measuring RPI were adopted from Chiu et al. (2014); and two items for measuring WOM were adopted from Lepp¨ aniemi et al. (2017). To secure equivalent data collection from both countries, we followed the protocol of Coviello and Jones (2004) for the survey’s translation from English to Finnish and Swedish. We used the same cover letter and the same items, and the items were both forward and back translated. The data collection process was similar and concurrent in both countries. The hypotheses were tested with a sample of 2084 retail customers in FIN and 2334 customers in SWE using SmartPLS 3 software (Ringle et al., 2015). Both samples were analyzed separately. Partial least squares (PLS) structural equation modeling was selected because this study focused on predictions instead of theory testing (Hair et al., 2014; Richter et al., 2016). Moreover, many variables were not normally distributed because the study was exploratory in nature, and it examined several new relationships. In these situations, PLS is a preferred method (Hair et al., 2014). To test the compositional invariance, we used the non-parametric measurement invariance of composite models procedure with 5000 permutations (Hensler et al., 2016). Because all the c values of the multiple indicator measures were not significantly different from 1 (p > 0.05), partial measurement invariance was established. In addition, we examined the structural invariance of the model and found no significant differences between the two countries (Hensler et al., 2016). Therefore, the conditions for analyzing both data sets in the same study were considered adequate. 4.3. Results 4.3.1. Measurement model The reliability and validity of the models were evaluated in both samples by composite reliability (CR), average variance extracted (AVE), Cronbach’s alpha ( α ), and factor loadings (Table 4 and Appendix A). The factor loadings were all significant and higher than 0.70 in all cases, and the CR, AVE, and α met the criteria that were set in the literature (Hair et al., 2014, p. 105). The square root of the AVE in each latent variable was compared with the other constructs to evaluate discriminant validity. The square root of the AVE was higher than the correlation between any two latent constructs (Fornell and Larcker, 1981), which verified the discriminant validity. The Heterotrait-Monotrait ratios of correlations were also all below the cut-off criterion of 0.90 (Henseler et al., 2016), which further confirmed the discriminant validity. Moreover, all the items had the highest factor loadings in the construct that they were measuring. Thus, the reliability and validity of the models were found acceptable. 4.3.2. Structural model The R 2 values for emotional and cognitive CX, WOM, and RPI were all above 0.40, indicating predictive accuracy (Fig. 2) (Henseler et al., 2009). The Stone-Geisser criterion (Q 2 ) values were above zero, indicating the models’ predictive relevance (Henseler et al., 2009). Table 3 Demographic Profile of the Respondents vs. Population. Finland Sweden N % Population N % Population Gender Male 918 44.0 49.4 1023 43.8 50.3 Female 1166 56.0 50.6 1311 56.2 49.7 Age (years) 15–24 831 39.9 11.2 1334 57.2 11.3 25–34 308 14.8 12.3 315 13.5 14.0 35–49 538 25.8 19.0 414 17.7 18.9 50–60 407 19.5 13.9 271 11.6 12.6 Household size 1 589 28.3 44.0 469 20.1 37.7 2 679 32.6 32.7 552 23.7 31.3 3 357 17.1 10.3 452 19.4 12.4 4 254 12.2 8.4 502 21.5 12.3 5 128 6.1 3.1 217 9.3 4.3 6+77 3.7 14.1 143 5.1 2.0 Data for Finnish population derived from Statistics Finland (2019) and Swedish population from Statistics Sweden (2019). O. Tyrv¨ ainen et al. Journal of Retailing and Consumer Services 57 (2020) 102233 6 The results fully support all the hypotheses in both samples (Fig. 2; Table 5). The findings show that HM had a positive direct effect on both the emotional (β FIN =0.717, p <0.01; β SWE =0.533, p <0.01) and cognitive (β FIN =0.475, p <0.01; β SWE =0.478, p <0.01) components of CX in both samples, confirming H1a and H1b. The results also support H2a and H2b, showing that personalization had a significant positive effect on CX (emotional CX: β FIN =0.136, p <0.01; β SWE =0.274, p < 0.01; cognitive CX: β FIN =0.315, p <0.01; β SWE =0.317, p <0.01). Significant positive effects in both samples were found between cognitive CX and WOM (β FIN =0.187, p <0.01; β SWE =0.189, p <0.01) and emotional CX and WOM (β IN =0.590, p <0.01; β SWE =0.500, p <0.01), confirming H3a and H3b. Finally, H4a and H4b are supported: both emotional CX (β FIN =0.546, p <0.01; β SWE =0.487, p <0.01) and cognitive CX (β FIN =0.164, p <0.01; β SWE =0.189, p <0.01) had a significant effect on RPI. Although we did not hypothesize country differences, we examined how the path coefficients differed between FIN and. SWE. With the multimethod PLS-multi-group analysis (MGA) (Parametric Test, WelchSatterthwait Test), including 5000 subsamples and 1000 permutations, we found some significant differences (p <0.05 or >0.95) between the countries (Hair et al., 2014, p. 253). These differences were related to the relationships between HM and emotional CX, personalization and emotional CX, and emotional CX and WOM. The findings indicate that these relationships were stronger in the FIN sample, except for the relationship between personalization and emotional CX, which was stronger for SWE (see Appendix B). The relationship between emotional CX and WOM was stronger in the FIN sample. When controlling for demographics (gender and age), we found a slight positive effect of gender on WOM (β FIN =0.059, p <0.01; β SWE = 0.100, p <0.01) and RPI (β FIN =0.046, p <0.01; β SWE =0.044, p <0.01) in both samples, indicating that females have a slightly higher intention to spread positive WOM and RPI. Regarding the relationship between Table 4 Discriminant validity, means and standard deviations. Finland AVE CR a α (1) (2) (3) (4) (5) (6) (7) (8) (9) HM b (1) 0.891 0.961 0.94 0,944 PERS c (2) 0.769 0.909 0.85 0,495 0877 COG d CX (3) 0.747 0.898 0.83 0,630 0550 0,864 EMO e CX (4) 0.760 0.927 0.89 0,784 0490 0,624 0872 WOM f (5) 0.946 0.972 0.94 0,700 0444 0,578 0724 0,973 RPI g (6) 0.787 0.917 0.86 0,660 0407 0,501 0648 0,701 0887 Gender (7) n/a n/a n/a 0,075 0021 0,086 0039 0,102 0073 n/a Age (8) n/a n/a n/a −0,108 −0,122 −0,180 −0,048 −0,092 −0,020 −0,174 n/a Income (9) n/a n/a n/a −0,048 −0,041 −0,038 −0,037 −0,055 −0,008 −0,061 0145 n/a Mean 5.44 4.59 4.85 5.36 7.86 5.71 n/a n/a n/a s.d. 1.26 1.44 1.30 1.21 2.02 1.19 n/a n/a n/a Sweden AVE CR a α (1) (2) (3) (4) (5) (6) (7) (8) (9) HM b (1) 0.761 0.905 0.842 0.872 PERS c (2) 0.733 0.892 0.818 0.562 0.856 COG d CX (3) 0.731 0.891 0.816 0.655 0.585 0.855 EMO e CX (4) 0.706 0.906 0.861 0.688 0.574 0.596 0.840 WOM f (5) 0.925 0.961 0.919 0.620 0.455 0.491 0.618 0.962 RPI g (6) 0.737 0.894 0.821 0.604 0.486 0.500 0.622 0.706 0.859 Gender (7) n/a n/a n/a 0.115 −0.007 0.061 0.056 0.137 0.086 n/a Age (8) n/a n/a n/a −0.087 −0.028 −0.103 −0.020 −0.014 0.026 −0.082 n/a Income (9) n/a n/a n/a 0.008 −0.004 −0.015 −0.017 −0.015 0.026 0.039 0.012 n/a Mean 5.21 4.82 4.91 5.26 7.83 5.53 n/a n/a n/a s.d. 1.53 1.60 1.60 1.49 2.30 1.50 n/a n/a n/a Notes: a CR – Composite Reliability; b C α – Cronbach’s alpha; c HM – Hedonic Motivation; d PERS – Personalization; e COG CX – Cognitive Customer Experience; f EMO CX – Emotional Customer Experience; g WOM – Word of Mouth; h RPI – Repeat Purchase Intention. n/a =Not applicable. Item measured with single item. AVE, CR and C α cannot be computed. Fig. 2. Study results. O. Tyrv¨ ainen et al. Journal of Retailing and Consumer Services 57 (2020) 102233 7 age and WOM and age and RPI, the only significant positive effect was found between age and RPI in the SWE sample (β SWE =0.078, p <0.01). This indicated that, among Swedes, the older one is, the higher the RPI. The other relationships did not exceed the threshold of significance (p < 0.01). 5. Discussion 5.1. Theoretical contributions Consumers today are presented with a large variety of opportunities through which they can search, compare, purchase, and obtain goods and services (Yrj¨ ol¨ a et al., 2018). While omnichannel retailing is becoming the new norm for many retailers, the purpose of this study was to develop and test a conceptual framework that links key retailing constructs in an omnichannel research setting. While previous studies have focused more on limited retail sections and single-channel studies, such as online retail or mobile shopping, we developed a research setting that focused on the entire omnichannel experience, as Hure et al. (2017) suggested. The results of the customer interviews and a survey of 2084 retail customers in FIN and 2334 retail customers in SWE provide both theoretical and managerial contributions concerning the effects of personalization and HM on CXs and their influence on WOM and RPI. The results provide three main theoretical contributions. As suggested by Souiden et al. (2019), we extended our empirical investigation beyond the single-channel setting by studying CX from the omnichannel perspective. In general, our findings confirm that CX is a key determinant of consumer behavior in the omnichannel context. Among both samples, the findings show the positive effects of CXs’ emotional and cognitive components on loyalty outcomes, WOM (Klaus and Maklan, 2013), and RPI (Gountas and Gountas, 2007; Ou and Verhoef, 2017; Rose et al., 2012). Therefore, the results strengthen the previous findings that are related to the relationship between CX and customer loyalty (Brakus et al., 2009). CX is a key determinant of consumer behavior in the omnichannel context. Among both samples, the findings show the positive effects of CXs’ emotional and cognitive components on loyalty outcomes, WOM (Klaus and Maklan, 2013), and RPI (Gountas and Gountas, 2007; Ou and Verhoef, 2017; Rose et al., 2012). Therefore, the results strengthen the previous findings that are related to the relationship between CX and customer loyalty (Brakus et al., 2009). Previous studies have compared online and offline stores from HM viewpoint and shown that consumers shopping hedonic products prefer online stores (Scarpi et al., 2014; Shen et al., 2016). This study contributes to this discussion by presenting the omnichannel viewpoint. The key findings of our research in both the FIN and SWE samples particularly show that emotional and cognitive CX is driven by HM—a relationship that has been identified in the online context (Evanschitzky et al., 2014; Izogo and Jayawardhena, 2018). Similarly, the results also confirm the direct relationship between personalization and CX components, which were previously studied in the online retail context (McLean et al., 2018; Pappas et al., 2014; Rose et al., 2012). The effect of personalization is slightly stronger on cognitive CX than on emotional CX in omnichannel context. Finally, even though FIN and SWE are geographically and culturally close to each other, three differences were found between these two samples through the use of the PLS-MGA. Generally, HM proved to be more related to emotional CX in FIN, while personalization was a more significant driver of emotional CX in Sweden. The relationship between emotional CX and WOM was stronger among the FIN sample. The present study is the first to demonstrate the differences that are related to CX in these countries. 5.2. Managerial implications It is widely acknowledged that CX plays an increasingly important role in retailing. Today, retailers are piloting (and both succeeding and failing) diverse omnichannel initiatives (Yrj¨ ol¨ a et al., 2018) in serving customers such as various mobile applications, click & collect concepts, home delivery or cashierless stores (Amazon Go). Clearly, there is a need for well-grounded empirical insight that can help managers design their presence in an omnichannel environment. In that respect, our findings put forth three main managerial implications for enhancing CXs in omnichannel retailing. Our results indicate that creating more personalized offerings and advertisements improves the total CX. Consequently, retailers should identify personalization as a key driver for improving CX. Moreover, although the empirical focus was placed on personalized offerings and advertisements, retailers should actively search for new ways to build the perception of personalization, which can take various forms, such as using customer data in order to provide customers with personalized information about their purchases, e.g. information regarding healthfulness or carbon footprint of their previous food purchases. Our findings demonstrate that better CXs in omnichannel retailing lead to positive effects on customer loyalty through increased WOM and RPI, which underscores that retailers have the possibility to influence customer purchasing behavior via personalization. In that respect, our findings help managers understand the interlinking nature of these constructs and consequently prioritize among different marketing activities in an omnichannel environment. This study draws the attention of retailers to HM. It is wellestablished that HM represents a key element in CX creation. Similarly, higher HM indicates a higher cognitive and emotional CX. This finding indicates that creating a fun, enjoyable, and entertaining shopping environment also leads to a more positive experience in omnichannel retailing, which increases RPI and positive WOM. Thus, while general attention is often placed in e-commerce on providing low prices and a wide product selection, retailers should always consider the potential of supporting consumers’ omnichannel behavior via more hedonic characteristics, such as emphasizing the enjoyment of using a personalized promotion. 5.3. Limitations and further research This study has certain limitations that can be addressed in future studies. Because our results were mainly drawn from a survey study, which is not a longitudinal tool, the dynamism of CXs could not be fully captured. By using a longitudinal study setting, future studies will be able to confirm the causality of the studies’ relationships. Table 5 Hypothesis testing. Hypothesized relationship Path coefficient (β) (Finland) Path coefficient (β) (Sweden) Hypothesis testing H1a: Hedonic motivation → Emotional CX 0.717*** 0.533*** Supported H1b: Hedonic motivation → Cognitive CX 0.475*** 0.478*** Supported H2a: Personalization → Emotional CX 0.136*** 0.274*** Supported H2b: Personalization → Cognitive CX 0.315*** 0.317*** Supported H3a: Emotional CX → WOM a 0.590*** 0.500*** Supported H3b: Cognitive CX → WOM a 0.187*** 0.189*** Supported H4a: Emotional CX → RPI b 0.546*** 0.487*** Supported H4b: Cognitive CX → RPI b 0.164*** 0.189*** Supported Notes: a WOM – Word of Mouth; b RPI – Repeat Purchase Intention. ***p ≤0.001. O. Tyrv¨ ainen et al. Journal of Retailing and Consumer Services 57 (2020) 102233 8 In addition, personalization was measured with items that are considered especially personalized advertisement. However, our qualitative results reveal that other issues, such as social media personalization, play a role in forming CX. Therefore, we encourage researchers to investigate personalization in a limited context. While the focus of retail research has turned into integrated omnichannel experience, customer’s preferences, demands and behavior are varying between different channels. It would be important to study if the consumers’ attitudes towards personalization differ between retail channels, such as webstores, mobile applications, and retailer’s social media channels. Finally, our convenience samples led to empirical data that might not accurately represent the opinions of retail consumers in FIN and SWE. Retailers in those two markets are in the leading position to add elements of personalization. To enhance the generalizability of our findings, additional studies should access our theory in other markets, especially those that are emerging. Appendix A. Measurement Scales Constructs and items Factor loadings c Finland Sweden Hedonic Motivation Shopping in this store is fun. 0.952 0.827 Shopping in this store is enjoyable. 0.949 0.903 Shopping in this store is very entertaining. 0.930 0.884 Personalization This retailer can provide me with personalized deals/ads that are tailored to my activity context. 0.888 0.839 This retailer can provide me with more relevant promotional information that is tailored to my preferences or personal interests. 0.904 0.873 This retailer can provide me with the kind of deals/ads that I might like. 0.837 0.856 Cognitive CX When I visit the online and B&M stores of this specific retail chain, I get inspiration and new ideas (e.g., store window displays, product displays, and tips from store personnel). 0.891 0.854 Visiting the online and B&M stores of this specific retail chain evokes curiosity and a thirst for knowledge (e.g., I want to try new products and services). 0.881 0.871 It is nice to test products in the online and B&M stores of this specific retail chain (e.g., off-line product demonstrations or using applications in the e-store). 0.818 0.839 Emotional CX I become careless while visiting the online and B&M stores of this specific retail chain (e.g., reliance on the quality of merchandise, data privacy, and security matters). 0.840 0.827 I feel relaxed while visiting the online and B&M stores of this specific retail chain (e.g., proficient customer service, functioning product delivery and returns). 0.894 0.848 I am looked after while visiting the online and B&M stores of this specific retail chain. 0.880 0.847 I feel welcome while visiting the online and B&M stores of this specific retail chain. 0.871 0.839 WOM a How likely is it that you would recommend [company X] to a colleague or friend? 0.972 0.961 How likely is it that you would say positive things about [company X] to other people? 0.974 0.963 RPI b I plan to continue using this store to purchase products. 0.905 0.879 I consider this store my first choice for transactions in the future. 0.847 0.811 It is likely that I will continue purchasing products from this store in the future. 0.909 0.884 Notes: a WOM – Word of Mouth; b RPI – Repeat Purchase Intention; c – All factor loadings significant at p <0.001. Appendix B. Multimethod multigroup analysis: Differences between countries P-value tests Relationship Path Coefficients-difference MGA Parametric test Welch-Satterthwait test Permutation HM a → COG CX c 0.003 0.536 0.923 0.924 0.926 HM → EMO CX d 0.183 0.000*** 0.000*** 0.000*** 0.000*** PERS b → COG CX 0.002 0.525 0.954*** 0.954*** 0.941 PERS → EMO CX 0.139 1.000*** 0.000*** 0.000*** 0.000*** COG CX → WOM e 0.017 0.304 0.595 0.596 0.611 COG CX → RPI f 0.045 0.918 0.172 0.174 0.207 EMO CX → WOM 0.103 0.001*** 0.001*** 0.001*** 0.004*** EMO CX → RPI 0.049 0.062 0.120 0.122 0.165 Notes: a HM – Hedonic Motivation; b PERS – Personalization; c COG CX – Cognitive Customer Experience; d EMO CX – Emotional Customer Experience; e WOM – Word of Mouth; f RPI – Repeat Purchase Intention. ***p ≤0.1 or p ≥0.9. References Adomavicius, G., Tuzhilin, A., 2005. Towards the next generation of recommender systems: a survey of the state-of-the-art and possible extensions. IEEE Trans. Knowl. Data Eng. 17 (6), 734–749. Ansari, A., Mela, C.F., 2003. E-customization. J. Market. Res. 40 (2), 131–145. Arnold, M., Reynolds, K., 2003. Hedonic shopping motivations. J. Retailing 79 (2), 77–95. Arnold, M., Reynolds, K., 2012. Approach and avoidance motivation: investigating hedonic consumption in a retail setting. J. Retailing 88 (3), 399–411. Babin, B., Darden, W., Griffin, M., 1994. Work and/or fun: measuring hedonic and utilitarian shopping value. J. Consum. Res. 20 (4), 644–656. Balaji, M.S., Khong, K.W., Chong, A.Y.L., 2016. Determinants of negative word-of-mouth communication using social networking sites. Inf. Manag. 53, 528–540. Beck, N., Rygl, D., 2015. Categorization of multiple channel retailing in multi-, cross-, and omni-channel retailing for retailers and retailing. J. Retailing Consum. Serv. 27 (C), 170–178. Bilgihan, A., 2016. Gen Y customer loyalty in online shopping: an integrated model of trust, user experience and branding. Comput. Hum. Behav. 61 (C), 103–113. Bilgihan, A., Kandampully, J., Zhang, T., 2015. Towards a unified customer experience in online shopping environmentsAntecedents and outcomes. Int. J. Qual. Serv. Sci. 8 (1), 102–119. O. Tyrv¨ ainen et al.