1
University of Amsterdam
Faculty of Economics and Business
MSc Business Administration- Marketing Track
How do different personalization implementations in
online retail shopping influence customer satisfaction
and loyalty?
Master Thesis- 29
thof June 2015
Author: Pinar Kuruuzum
Student No: 10825053
2 Statement of Originality
This document is written by Student Pinar Kuruuzum who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Contents
1. Introduction ... 6
1.1 The Effects of Personalization ... 7
1.2 Research gap & Main question ... 11
1.3 Structure of Research ... 12 2. Literature review ... 13 2.1 Loyalty ... 13 2.2 Satisfaction ... 15 2.3 Personalization ... 16 2.4 E-personalization ... 18
2.5 Personalized Product Recommendation ... 22
2.6 Personalized Promotions ... 23
2.7 Personalized Advertisements ... 24
2.8 Interaction between Satisfaction and Loyalty ... 25
2.9 Interaction between E-personalization & Satisfaction & Loyalty ... 26
3. Conceptual Framework ... 28
4. Methodology ... 29
4.1 Sample Characteristics ... 29
4.2 Design... 29
4.3 Variables & Measures ... 30
4.4 Procedure ... 31
5. Results ... 34
5.1 Correlation Check ... 37
5.2 Hierarchical Regression Analysis ... 38
5.3 Testing Hypotheses ... 41
6. Discussion ... 46
6.1 Findings ... 46
6.2 Theoretical and Managerial Implications ... 48
6.3 Limitations and Future Research... 48
References ... 50
Appendix-1 Classification Scheme for Personalization Systems ... 53
4 Appendix-3 Personalized promotion example-2 ... 55 Appendix-4 Personalized advertisement example ... 56 Appendix-5 Survey ... 57
5
Abstract
With the rise of the online shopping, online retailers have started to use various ways to attract customers and one of them is known as e-personalization. E-personalization can be implemented in different forms and it is thought to have positive effects on business results. In the present study, the influence of different personalization implementations, namely personalized product recommendations and personalized promotions, on customer satisfaction and loyalty was tested. A quasi-experiment was used as a research method and results revealed that both personalized product recommendations and personalized promotions have positive effect on e-loyalty. While satisfaction mediates the relationship between personalized product recommendations and e-loyalty, it does not mediate for personalized promotions. Also, the results showed that number of personalization implementations used has not significant effect on e-satisfaction and e-loyalty. The outcomes of this research provide further understanding about the effects of personalization implementations managerially. Having a more clear understanding of the connection between personalized product recommendations, personalized promotions, customer satisfaction and loyalty, companies can improve their customer relations by setting more effective online offerings.
6
1. Introduction
Over recent decades, the Internet and its following technology have enabled companies to market their products in more cost-effective and efficient way by building an additional sales channel (Smith, 2006). This online channel has seen a fast growth. According to eMarketer report business-to-consumer (B2C) e-commerce sales worldwide was expected to be $1.5 trillion in 2014, rising nearly 20% over 2013 (Nielsen, 2014). Besides this fast growth, the characteristics of the online market contain great opportunity for firms to more efficiently reaching customers than traditional retail stores, overcoming physical obstacles (Limayem, Khalifa, & Frini, 2000).
Online retailers use various ways to attract customers and one of the methods is known as e-personalization. It is defined as “providing relevant content or recommendations to a customer through past behavior, similarity with other users, explicitly defined preferences or individual characteristics” (Smith, 2006, p.90). For example, a retailer that sells sport equipment online can suggest football shoes to a specific customer who viewed or purchased football shorts, offer discount on football t-shirts or show football related advertisements on its website.
There are currently many online retailers providing personalized offers to customers such as Amazon, eBay, bol.com and online store of retailers (for example Philips, Body Shop, Nike). The CEO of Amazon, Bezos explains the importance of personalization by saying “The success of electronic retailers will depend on their ability to analyze each customer's tastes and create unique experiences from the moment they walk in the virtual door. If we have 4.5 million customers, we shouldn't have one store. We should have 4.5 million stores” (Walker, 1998).
Although it is expected that providing online personalized offers enable firms to retain communication or interaction with their customers by increasing their Web experience,
7 encourage the rate of website visiting, increase the chance of sales, advertisement revenue and website’s profitability (Weng & Liu, 2004), the business results of different e-personalization implementations in online retail shopping are unknown. Therefore, the interest of this study is to investigate the impact of different personalization implementations in online retail shopping on customer satisfaction and loyalty.
This section further explains the known effects of personalization using previous literature. Then, the research gap and main research question with sub-questions are presented to address this gap. The academic and managerial contributions are explained. Lastly, the structure of the thesis is shared.
1.1 The Effects of Personalization
In the academic literature online personalization has been a topic of wide discussion. Murthi and Sarkar (2003) classified the literature into two groups. The first group is personalization process and second one is personalization and firm strategy. Personalization process investigates the technical process behind it and firm strategy is concerned about business results of personalization (Murthi & Sarkar, 2003). As this research aims to identify the impact of certain personalization implementations on customer relations, it belongs to the second group of studies. The literature in the second group of studies have investigated the possible effects of personalization on different variables. The study by Thorbjørnsen, Supphellen, Nysveen, & Egil (2002) compared personalized websites and customer communities for their ability to create consumer-brand relationships measured by Brand Relationship Quality (BRQ) dimensions (love/passion, self-connection, personal commitment, intimacy, partner quality, behavioral interdependence). Initially they could not obtain any significant effect until Internet experience was included as a moderating variable. Particularly, they revealed that personalized websites
8 developed stronger consumer-brand relationships for respondents with extensive Internet experience than for respondents with limited Internet experience. On the other hand, customer communities developed stronger consumer-brand relationships for respondents with limited Internet experience than the respondents with extensive Internet experience. Because, novice users are more influenced by the third party information than experienced users (Thorbjørnsen, Supphellen, Nysveen, & Egil, 2002).
Another study by Holland and Baker (2001) also investigated the effect of personalization and community on website brand loyalty. They found that personalization contributes website loyalty via increasing switching cost of customers because once the customer has spent time and energy into the personalization process, there is a discouragement to start all over again with another site. Also, consumers’ goals in visiting a website (task or experiential) affect their propensity to be site brand loyal. While task oriented customers value the convenience of website, experiential oriented customers value enjoyable features more (Holland & Menzel Baker, 2001). The authors supported their model with the case studies of the corporate websites.
Aside from consumer-brand relationships and website brand loyalty, another study showed the effect of personalization on intention to purchase. In their research Pappas, Giannakos, & Chrissikopoulos (2012) demonstrated how the use of personalized services in online shopping effect intention to purchase as well as customers' privacy issues and enjoyment. Also, they investigated whether enjoyment and privacy issues have an effect on purchase intention. They used online survey for their empirical research and found that personalization effects enjoyment and purchase intention positively, but has no effect on privacy. The reasons for the results were explained as the way of personalized services offers the right product at the right
9 time, answering to customer’s needs and making their shopping experience easier and more enjoyable. Also, privacy concerns were not affected because customers agree and accept some privacy warnings (such as cookies) before they use the personalized service. Furthermore, results show that privacy concerns effect purchase intention negatively, while enjoyment has a positive influence on purchase intention (Pappas, Giannakos, & Chrissikopoulos, 2012).
Another study (Zhang, Agarwal, & Lucas Jr, 2011) has specified the personalized offers more, investigating the effect of personalized product recommendations (PPRs) on online store loyalty and taking into consideration the shopping efficiency as a moderator. In the research, shopping efficiency refers to product screening cost, product evaluation cost and decision making quality. They used household production function model in consumer economics literature to build a theoretical framework that explains the process through which PPRs effect customer online store loyalty. They stated that retailer learning of customer knowledge, gained to enable personalization, effects the efficiency of the online retailing activity. They tested the theoretical model with an experiment where the quality of PPRs was manipulated and concluded that retailer learning demonstrated with higher quality PPRs is linked with lower product screening cost, but higher product evaluation cost. Furthermore, higher quality PPRs create higher value for consumers in terms of higher decision making quality, which is positively related with loyalty.
Similarly, some researchers investigated the effect of customization on e-loyalty. As the term customization is related to the interest of this study, it is necessary to look into the studies in that area. Srinivasan, Anderson and Ponnavolu (2002, p.42) defined customization in their study as “the extent to which an e-retailer’s website can recognize a customer and then tailor the choice of products, services, and shopping experience for that customer”. In their research, they
10 examined the influence of the eight factors used in online retailing (customization, contact interactivity, care, community, convenience, cultivation, choice, and character) on e-loyalty. They were briefly explained as follows: contact interactivity means the dynamic nature of the engagement that happens between an e-retailer and its customers. Care represents the attention that an e-retailer gives to all the pre- and post- purchase customer interface actions in order to build short term transactions and long term customer relationships. Community refers to online social entity organized by e-retailer in order to enable customers to exchange opinions and information about the products and services. Convenience is the extent that customer thinks the website is simple, intuitive, and user friendly. Cultivation refers to the extent that an e-retailer gives relevant information and incentives to customers to increase their purchases over time. Choice is offering wide range of products and services in a category. Lastly, character refers to the image that e-retailer projects to customers through the text, style, color and graphics on its website. They found that all the factors except convenience have significant effect on e-loyalty. The result of the study also showed that loyalty positively effects word of mouth and higher willingness to pay. When elaborating the effect of customization, they stated that customization contributes the perception of increased choice for customers, signals high quality, enables to complete transaction more efficiently and therefore makes it attractive for customers to visit the website again in the future (Srinivasan, Anderson, & Ponnavolu, 2002).
Another study by Chang & Chen (2008) suggested a theoretical framework for examining the relationship between customer interface quality (customization, interactivity, convenience and character), satisfaction, switching costs and e-loyalty. Also, they stated that internet experience moderates the relationship between interface quality and other constructs and switching costs moderates the relationship between satisfaction and e-loyalty. They used a
11 questionnaire to test the framework and showed that customer interface quality, including customization, interaction, convenience and character positively affect satisfaction and e-loyalty while have no effect on switching costs. However, when internet experience was introduced, customer interface quality increases switching costs for customers with higher internet experience. Furthermore, the study revealed that convenience significantly and directly effects on e-loyalty, while customization, interactivity and character indirectly influence e-loyalty through customer satisfaction.
1.2 Research gap & Main question
The previous literature about personalization has paid large attention to its effects on customer relations constructs such as loyalty, moderators and mediators of this relationship, however none of them tested and compared the effect of specific personalized implementations empirically. All of the studies presented before investigated the concepts of personalization very broadly. Therefore, the specific effect of certain personalization implementations in online retail shopping on customer satisfaction and e-loyalty is unknown. There are some studies (Kim et al., 2006; Montgomery & Smith, 2009; Smith, 2006) introducing different personalization implementations such as personalized product recommendations and personalized promotions conceptually, however none of them tested and compared them on e- loyalty empirically. Also, none of the studies investigated the mediation effect of satisfaction in this relationship. Therefore, the objective of this study is to address this gap by answering the question and sub-questions below: • How do different personalization implementations, namely personalized product
recommendations and promotions, in online retail shopping influence customer satisfaction and loyalty?
12 • Which of the stated online personalization implementations are more strongly
connected to satisfaction and loyalty? • How are satisfaction and loyalty related?
• What is the direct/indirect effect of personalization implementations on loyalty (through satisfaction)?
Managerially, having a more clear understanding of the connection between personalized product recommendations, personalized promotions, customer satisfaction and loyalty, companies can improve their customer relations by setting more effective online offerings.
1.3 Structure of Research
The structure of the research will be as follows. First, the related literature and complementary articles will be investigated in order to form the conceptual model and hypotheses. Then, in the methodology part, sample information, research design and clarification of all variables will be presented. Next chapter will continue with results of quantitative analysis which will lead us to discussions part. The outcomes related to hypotheses will be analyzed in detail. Finally, implications, contributions and limitations will be shared.
13
2. Literature review
This chapter begins with the explanation of the constructs loyalty, satisfaction, personalization, e-personalization and different e-personalization implementations (personalized product recommendation, personalized promotions, personalized advertisements) using previous literature and practical examples. Then it continues with the interaction between these constructs to formulate the hypotheses and to build the conceptual framework.
2.1 Loyalty
Loyalty is an important construct in customer relations and it is covered widely in previous literature. The term is defined as repeated buying behavior displayed for a sustained period of time as a result of favorable attitude towards the subject (Chang & Chen, 2008) and it contains both attitudinal and behavioral elements (Casaló, Flavián, & Guinalíu, 2008). Oliver (1999, p.34) defines customer loyalty as “a deeply held commitment to re-buy or re-patronize a preferred product/service consistently in the future, thereby causing repetitive same-brand or same brand-set purchasing, despite situational influences and marketing efforts having the potential to cause switching behavior”.
According to Oliver (1999), there are four loyalty phases: (1) cognitive loyalty (2) affective loyalty (3) conative loyalty (4) action loyalty. The first phase refers to loyalty to information such as price, specifications and so on. Second phase is loyalty to a liking (e.g. “I purchase it again because I like it”). Third phase indicates intentions (e.g. “I am committed to purchasing it again”). Last phase is loyalty to action inertia, readiness to overcoming obstacles to repurchase. He states the easiest phase of loyalty to break down is cognitive while hardest is action phase.
14 The concept of loyalty has been enhanced to e-loyalty with the rise of the online consumer activities (Chang & Chen, 2008). Anderson & Srinivasan (2003) made an extensive research about loyalty in the e-commerce context and defined e-loyalty as customer’s favorable attitude toward an online retailer that drives repeat purchasing behavior. For the purpose of this study, this definition is used. Although loyalty and e-loyalty are theoretically similar, there are different aspects concerning internet based marketing and buyer behavior (Anderson & Srinivasan, 2003). For example, in e-commerce settings it is harder to build loyalty as switching costs are low and competitors are only one mouse click away (Oliver, 1999). Thus the switching cost is an important contextual factor in e-loyalty literature.
Switching cost is defined as consumer perceptions of the time, money, and effort regarding to the change from one website to another (Chang & Chen, 2008). Although the cost is associated with financial factors, psychological and emotional factors also play important role. For example, when social bonds and trust are built over a period of time between the company and customers, customers have psychological exit barriers even if they are not satisfied with the service/product (Sharma & Patterson, 2000). Therefore, it is crucial for companies to increase switching costs to retain customers.
Building customer loyalty in online markets is very important for the profitability of businesses. Because according to Reichheld & Schefter (2000) the cost of acquiring customers on the Internet is much more than retaining existing ones. Also, increasing customer retention rates by 5% leads to 25% - 95% increase in profits. They claim that without the glue of loyalty, even the best-designed e-commerce model will fail. Moreover, Srinivasan, Anderson and Ponnavolu (2002) investigated the behavioral consequences of loyalty in e-commerce and proved empirically that loyalty increases positive word-of-mouth (the degree to which a person
15 says positive things about the e-retailer to other people) and willingness to pay. They argue that loyal customers are more inelastic to price than non-loyal customers and willing to pay a premium to continue the relationship with their preferred retailers instead of bearing the burden of searching cost. Therefore, building loyalty in online markets is crucial for retailers.
2.2 Satisfaction
Satisfaction is defined as the pleasurable fulfillment of needs and expectations as a result of consuming certain products or services (Oliver Richard, 1997). In their study, Chang & Chen (2008) state that there are two types of customer satisfaction: (1) transaction-specific, (2) cumulative. Transaction-specific customer satisfaction is the post-evaluation of a specific buying situation. In contrast, cumulative customer satisfaction refers to overall assessment of the company based on experience with its products and services over time. In this study the term satisfaction refers to both types of satisfaction.
Moreover, in order to fully understand the construct, this study investigates the possible antecedents. In his research Oliver (1980) suggested a model stating expectation and expectancy disconfirmation influence satisfaction. Specifically, expectations set a reference point to make a comparative judgement. So, the results below than expectations (a negative disconfirmation) are perceived worse than reference point, while those above the expectations (positive disconfirmation) are perceived better than this base. Therefore, satisfaction can be interpreted as an additive combination of the expectations and disconfirmation (Oliver, 1980).
E-satisfaction has emerged with the rise of e-commerce and is defined as the contentment of the customer due to previous buying experience with a particular e-commerce company (Anderson & Srinivasan, 2003). As it is a new concept compared to satisfaction, there are fewer conceptual and empirical studies elaborating the term. Szymanski and Hise (2000) provided
16 initial evidence for the determinants of e-satisfaction in their study. After conducting a qualitative research they suggested a conceptual model that states online convenience, merchandising, site design, and financial security are the factors effecting e-satisfaction. In their model, online convenience refers to the time and effort advantages of the online shopping such as locating merchants, finding items and buying offerings. Merchandising represents wider product offerings and product information available online. For example, while a traditional book superstore may supply 150,000 titles, Amazon.com supplies millions of titles. The other factor, site design refers to the ambience associated with the site itself and how it functions such as uncluttered screens, simple search paths, and fast presentations. Lastly, security of financial transactions are important due to customer concerns about sharing their credit card or payment information. The authors then tested these factors with survey and concluded that site design, convenience and financial security have significant effect on e-satisfaction while merchandising (wider assortments and richer information) has very little effect. Therefore, these findings support that giving special attention to convenience, site design, and financial security may create the most positive results increasing e-satisfaction (Szymanski & Hise, 2000).
This study uses the e-satisfaction definition of Anderson and Srinivasan (2003) as the level of customer contentment resulting from prior purchasing experience with a particular website to find its role in e-personalization and e-loyalty relationship.
2.3 Personalization
The construct of personalization has been covered in the literature starting with offline context and more largely elaborated in online context. It means customizing product, services, content, communication etc. to the needs of single customers or customer groups (Riemer & Totz, 2003).
17 The initial studies about personalization are in the context of service marketing. Surprenan and Solomon (1987) investigated the ways of personalization implementations in service encounters and defined personalized service as “any behaviors occurring in the interaction intended to contribute to the individuation of the customer, that is the customer role is embellished in the encounter through specific recognition of the customer's uniqueness as an individual over and above his/her status as an anonymous service recipient” (Surprenan & Solomon, 1987, p.87). They suggested three ways of conceptualizing and implementing personalization: (1) option personalization, (2) programmed personalization and (3) customized personalization. Option personalization allows customers to choose from a set of service options. For example, Burger King's offering to "have it your way" and companies’ cafeteria-style health benefit offerings to employees, from which each person selects his or her own unique mix of benefits from the total set offered. Programmed personalization is creating the impression of personalized service by encouraging communication with customer’s names and personal information. It aims to make each person feel like an individual, not just another customer. Lastly, customized personalization aims to help customer by advising the best possible form of the service offering for his or her needs. This type of personalization is generally supplied by real estate brokers, tax consultants, physicians, therapists, and other advice service providers. The authors tested the effects of suggested constructs on satisfaction and evaluations of the service organization. The results showed that while option personalization and customized personalization positively affect satisfaction and service organization evaluation, programmed personalization has negative effect. That is, the more non-task information included by the employee the lower the level of employee's competence, the organization’s trustworthiness, and customer satisfaction (Surprenant & Solomon, 1987).
18 Another study that investigates personalization in service encounters context is by Mittal and Lassar (1996). They defined personalization as “the social content of interaction between service employees and their customers” (Mittal & Lassar, 1996, p.96) referring to employee’s personal warmth and individual attention to know customers and their needs. They assessed the role of personalization in service quality perception that is measured by SERVQUAL. They suggested that personalization improve service quality perceptions and this effect is higher for person- processing service (e.g. healthcare service) than possession- processing service (e.g. car fixing service). They tested the hypotheses with a survey and concluded that personalization positively affects service quality perceptions, especially for the person- processing services. The study also implies that improved service quality perception is important for companies because of its positive impact on customer satisfaction (Mittal & Lassar, 1996).
These studies presented here and others support the fact that personalization has positive effect on customer perceptions. Furthermore, internet technologies have introduced many new opportunities for companies to improve their personalization options.
2.4 E-personalization
Personalization has been accepted as an important method for customer relationship and web strategies in e-commerce and mobile commerce. However, it is defined, characterized and implemented in very different ways in the literature (Fan & Poole, 2006). In their article Fan and Poole (2006) suggested an extensive scheme for classifying how personalization can be implemented online. The scheme has been built on 3 dimensions of implementation choices. These 3 dimensions are (1) the aspect of the information system that is formed to provide personalization (what is personalized), (2) the target of personalization (to whom to personalize),
19 and (3) who does the personalization (whether the user or the system). (See Appendix-1 for the scheme)
The first dimension, what to personalize, consists of 4 components: the information itself (content), how the information is shown (user interface), the media through which information is delivered (channel/information access), and what users can do with the system (functionality). The second one, to whom to personalize, can be either a category of users (for example, young women interested in cooking) or individual users. The third dimension is whether personalization is made by the user (explicit) or the system (implicit). For example, all the product recommendation systems are made by system and some of the portal interface changes can be done by users (for example msn.com) (Fan & Poole, 2006). This scheme provides very broad classification of personalization implementations including all types of websites. As this study focuses on retail shopping websites, it fits in the model in terms of (1) content, (2) individual& categorical and (3) implicit. Because this study investigates the content, uses both individual & categorical users and system- made (implicit) implementations.
A more recent classification was done by Sunikka and Bragge (2008) based on previous literature including the Fan & Pole’s study. They suggested a framework with 3 dimensions: initiator of the personalization (customer or company initiated), type of products (tangible or intangible) and level of personalization (individual or group).
Although this classification is similar to Fan & Pole’s, it differs in type of products dimension and the name of the categories included (e.g. Web-customization, one-to-one personalization, micro personalization and mass customization). This thesis fits in the framework in one to one personalization and micro personalization because it investigates company initiated aspect of personalization for intangibles (web context, services) -because it examines the
20 personalization of offerings/promotions not personalization of tangible products- and both individual & group customers.
Table-1. Personalization framework
Source: (Sunikka & Bragge, 2008)
The most common technical approaches enabling e-personalization are content-based approach and collaborative filtering analysis (Lawrence, Almasi, Kotlyar, Viveros, & Duri, 2001; Weng & Liu, 2004). Content-based approach recommends products based on similarity to person’s previous purchases. In the literature it is also referred as dynamic content by Smith (2006). He argues that it could be used both in business to business (B2B) and business to consumers (B2C) transactions. For example, in B2B, buyers can see more detailed and special options about their industry specific materials and components while other external customers may not. In B2C, online retailers can speed customers buying decision by suggesting content based on information that the system has collected from customer’s previous visits and shopping cart selections. For example, the system can offer carpet cleaner to a person who bought dog food before (Lawrence et al., 2001). The other approach, collaborative filtering helps to predict a certain customer’s preferences based on the actions of similar customers. These recommendations generally communicated by “People who bought this item also bought these products” or some other similar phrases (Smith, 2006, p.94). Besides the filtering approaches
21 mentioned above, data mining is another method that enables e-personalization. It aims to profile customers by collecting and manipulating data from their online activities (shopping, demographics, social media activities, navigating etc.). This profile can be used together with other methods to create more effective interfaces, messages and product offerings and promotions (Smith, 2006).
2.4.1 E-personalization Implementations
Companies benefit from these technical approaches for implementing different e-personalization applications. After reviewing the literature, most common implementations can be listed as:
• personalized product recommendations (Montgomery & Smith, 2009; Zhang & Lucas, 2011)
• personalized promotions (Smith, 2006; Montgomery & Smith, 2009) • personalized advertisements (Kim et al., 2006)
• personalized search (Montgomery & Smith, 2009)
• personalized website interface (Wu, Im, Tremaine, Instone, & Turoff, 2003) • mass customization (Riemer & Totz, 2003)
Based on Sunikka & Bragge’s classification, first three implementations can be one-to-one personalization or micro-personalization depending on whether it is applied to individual or a group. Personalized search is one-to-one personalization as it is system initiated intangible application that is applied to an individual user. Personalized interface is web- customization as it is initiated by the user and mass customization refers to the personalization of product itself.
This study aims to investigate the effect of e-personalization implementations in online retail shopping context where the personalization is initiated by the company. Therefore
22 personalized product recommendations, personalized promotions and personalized advertisements are within this criterion. As personalized advertisements are not as common as other implementations, it is only explained conceptually and not included in this research. It can be a subject of future research when it is seen more common.
2.5 Personalized Product Recommendation
One of the important form of personalization implementations by online retailers is personalized product recommendations (Zhang and Lucas, 2011). E-commerce websites can anticipate customer’s future buying behavior through the information supplied from customer demographic data and a customer’s past viewing/ purchasing behavior. Therefore, they can recommend personalized products to customers and transform people who just browse on the Web into consumers, increasing customer loyalty and enhancing cross selling (Weng & Liu, 2004).
In their research, Zhang and Lucas (2011) developed a theoretical framework that explains the mechanism through which personalized product recommendations effect online store loyalty. They used household production function model in the consumer economics literature to develop the framework. According to the framework, retailer learning gained to enable personalization effects the efficiency of the online retailing activity. As the learning increases, the quality of PPRs increase. They suggest that higher quality PPRs are associated with greater value gained by consumers from the online retailing activity in terms of higher decision making quality, which is positively related with repurchase intention. They tested the framework with an experiment by manipulating the quality of PPRs. As a result, they proved that the quality of PPR positively effects repurchase intention and online store loyalty.
Moreover, the fact that many online book, movie and music retailers use the personalized recommender systems shows the importance of these service. The few practical examples are;
23 “Amazon, Barnes & Noble, LibraryThing, and Storycode to recommend books; Blockbuster, Eachmovie, Hollywood Video, Movielens, and Netflix to recommend movies; Audioscrobbler, CDNow, iLike, iTunes, Last.fm, MusicMatch, MSN Music, MyStrands, RealPlayerMusicStore, Rhapsody, and Napster to suggest music; TiVo for television shows; Findory for newsitems; and StumbleUpon for website recommendations” (Montgomery & Smith, 2009, p.7).
2.6 Personalized Promotions
Personalized promotions are generally in form of e-mail and banner ads (Smith, 2006). In his paper, Smith (2006) states that it is crucial to distinguish personalized e-mail promotions from mass mailing or e-mail spam. Personalized e-mail promotions should deliver valuable messages to specific customer or category of customers in order to build and maintain ongoing relationship. He explains that these special offers are sent to customers who registered the website with an e-mail address (see Appendix-2 for an example). Second type of personalized promotions, banner advertisements, are messages that appear in customer’s screen when viewing the retailer’s website. In the case of the banner ad is a correct match of audience and message, it can be an effective tool to satisfy customer needs (Smith, 2006).
Moreover, personalized promotions can be in other forms such as apps. For example Southwest airlines has a promotion system called Ding which works by providing special offers only through Ding application. Once customers download the app and give personal information (e.g. home airport, planned destinations), Ding can offer 20-25% reductions in prices over standard fares in line with customers’ interests. Also it enables the Southwest to make unique, personalized offers based on the customer‘s history and claimed preferences (see Appendix-3) (Montgomery & Smith, 2009).
24 Although previous studies explain the personalized promotions concept, they did not test the effects on customer relation constructs empirically.
2.7 Personalized Advertisements
Personalized advertisements are defined as advertisements that are selected to display to particular customers based on their demographics and past purchase behavior (Kim et al., 2006). Companies can benefit from personalized advertisements by receiving more desired consumer responses than bombarding them with irrelevant messages and having cost efficiency as mass advertising is more expensive (Pavlou & Stewart, 2000).
In the literature most of the studies about personalized advertisements (Kazienko & Adamski, 2004; Kim, Lee, Shaw, Chang, & Nelson, 2001; Kim et al., 2006) explain the different technical methods that enable the implementation.
One of the practical applications of online personalization is Philips’s website. The company is using cookies to track customers’ preferences to personalize the content, recommendations and advertisements. (Cookies are short connections distributed by the Web server and held by the client’s browser for future use. Cookies are used to identify each user and store information such as pages visited, products viewed and purchased (Pierrakos, Paliouras, Papatheodorou, & Spyropoulos, 2003)). In its website’s cookie policy, Philips explains the personalization of its website: “Our aim is to provide visitors to our website with information that is as relevant as possible to them. We therefore endeavor to adapt our site as much as possible to every visitor. We do this not only through the content of our website, but also through the advertisements shown. To make it possible for these adaptations to be carried out, we try to acquire a picture of your likely interests on the basis of the Philips websites that you visit in order to develop a segmented profile. Based on these interests, we then adapt the content and the
25 advertisements on our website for various groups of customers. For instance, based upon your surfing behavior, you may have similar interests to the 'males in the 30-to-45 age range, married with children and interested in football' category. This group will, of course, be shown different advertisements to the 'female, 20-to-30 age range, single and interested in travelling' category” (Philips cookies policy.2014).
Another practical example of personalized advertisements in social media context is Facebook. The company uses 3 ways to show personalized ads. Firstly, it uses the information users share on Facebook (ex: Pages they “like”) and other information about the users in their account (ex: age, gender, location, the devices used to access Facebook). Then it shows them relevant advertisement of products/services (See Appendix-4 for an example and Facebook’s explanation). Secondly, it shows the advertisement of companies or products that user’s friend “liked”, assuming the user also may like. Thirdly, Facebook can show the advertisement of a business in which the user visited its website outside of Facebook (About advertising on
facebook.2015). For example, the user might view coffee machines on Philip’s website (which
uses cookies to record the visit). Then Philips asks Facebook to show its coffee machine ads to this list of visitors, and visitors see these ads on their homepage.
As personalized ads are not commonly used as other personalized implementations and there are no studies testing empirically its effect, it is left for a future study to test the effects.
2.8 Interaction between Satisfaction and Loyalty
The previous literature has paid large attention to relationship between satisfaction and loyalty. They indicate that satisfaction is seen as an antecedent of loyalty, meaning increased satisfaction leading to increased loyalty (Anderson & Srinivasan, 2003; Balabanis, Reynolds, & Simintiras, 2006; Casaló et al., 2008; Gommans, Krishnan, & Scheffold, 2001; Reichheld & Schefter, 2000;
26 Srinivasan, Anderson, & Ponnavolu, 2002). Therefore, satisfying customers is important to e-stores in terms of keeping customers loyal. Satisfied customers tend to develop a positive attitude towards those e-stores (Chang & Chen, 2008). More extensive research has been done by Anderson & Srinivasan (2003) in the context of e-commerce. In their study, they found that although e-satisfaction effects e-loyalty, consumers’ individual level factors (inertia, convenience motivation, and purchase size) and firms’ business level factors (trust and perceived value offered by the firm) moderate this effect. According to the study, while convenience motivation, purchase size, trust and perceived value positively influence the effect of e-satisfaction on e-loyalty, inertia influences negatively.
Therefore, based on the studies presented, it is reasonable to suggest that;
H1: E-satisfaction has a positive direct effect on e-loyalty.
2.9 Interaction between E-personalization & Satisfaction & Loyalty
Previous studies (Holland & Menzel Baker, 2001) showed that there is a strong positive relationship between e-personalization implementations in online retail shopping and customer relationship construct satisfaction and loyalty. Following the e-personalization strategies enable customers to receive more relevant and useful information from company and will encourage them to use that company’s products and services. Loyalty will increase as well as the revenue if the company uses private, personal information appropriately (Smith, 2006).
The specific effect of personalized product recommendations on loyalty was investigated by Zhang, Agarwal, & Lucas Jr (2011)’s study. They proved that the quality of PPR positively effects repurchase intention and online store loyalty as explained in the literature review.
27 Moreover, Weng & Liu (2004) also support this claim conceptually in their paper. Therefore, it is reasonable to suggest the following hypotheses:
H2a: Personalized product recommendations have a positive direct effect on e-loyalty.
H2b: Personalized product recommendations have a positive direct effect on e-satisfaction.
H2c: Personalized product recommendations have a positive indirect effect on e-loyalty through
e-satisfaction.
The effect of personalized promotions on satisfaction and loyalty was suggested in the previous literature, but never tested empirically. Smith (2006) states that personalized promotions can be an effective tool to satisfy customer needs and Changchien, Lee, & Hsu (2004) say that these promotions have the potential to increase the success rate of promotion, and customer satisfaction and loyalty (Changchien, Lee, & Hsu, 2004). So, the following hypotheses are formed to test empirically:
H3a: Personalized promotions have a positive direct effect on e-loyalty.
H3b: Personalized promotions have a positive direct effect on e-satisfaction.
H3c: Personalized promotions have a positive indirect effect on e-loyalty through e-satisfaction.
The previously stated implementations raise the question that how satisfaction and loyalty are effected by the level of personalization (the number of personalization implementations). In his research Smith (2006) states that the greater degree of satisfaction can be achieved by leveraging more knowledge about consumer and applying to e-personalization implementations. As the technologies that support e-personalization increase, so will its results on consumers. Depending on this statement, it is reasonable to expect the following hypotheses:
28
H4a: Level of personalization (the number of personalization implementations) has a positive
direct effect on e-loyalty.
H4b: Level of personalization (the number of personalization implementations) has a positive
direct effect on e-satisfaction.
3. Conceptual Framework
Considering the previous literature presented and analyzed, following conceptual framework is formed.
Personalized Product Recommendations
29
4. Methodology
In this section the research method is described in order to answer the main research question: “How do different personalization implementations, namely personalized product recommendations and personalized promotions in online retail shopping influence customer satisfaction and loyalty?” The research method of this study is planned as quasi-experiment. This section consists of sample characteristics, design, variables & measures, procedure and strength & limitations of the research method.
4.1 Sample Characteristics
This study uses quasi-experiment targeted to online shoppers (who has done online shopping at least one time in his/her life) in Netherlands and Turkey to represent the population of online shoppers. An invitation to the experiment was sent to approximately 300 subjects via email and social media (Facebook and Linkedin) that directs them to online survey in “Qualtrics” website. Among those 170 people participated and completed the survey (%56 response rate). The proportion of female participants was 58.3%. The average age of the sample was 29.37 years; 20% of them were between 17 and 25; 50% were aged between 26 and 30. In regard to their education, 3% of the respondents had high school degree, 37.7 % had bachelor degree, and the remaining 59.3% had master and PhD degree. The occupations of the participants included students (22.6%), employed (73.8%), self-employed (1.8%) and the other (1.8%).
4.2 Design
The design of the experiment consists of 2 parts. Firstly, the respondents were asked to think about an online retailer and state it. Secondly, they were divided into 4 groups according to their familiarity with the personalization implementations in previously stated website. The 4 four groups were (1) no personalization (2) personalized product recommendations (3) personalized
30 promotions (4) both PPRs and personalized promotions. Then, they were asked questions related to their group. The group with the no personalization was used the baseline for the analyses. Stimuli was the introduction of stated personalized implementations.
4.3 Variables & Measures
The presentation of each different variable and the measures for the variables are needed to explain the procedure of the experiment. As stated in the conceptual framework, the independent variables are personalized product recommendations and personalized promotions, dependent variable is loyalty and mediator is satisfaction.
Although the literature related to each variable served as a guide in the development of the measures, well-established scales for some of the constructs were not readily available. Especially, as there is absence of questions to measure personalization implementations (personalized product recommendations and personalized promotions) the studies of Smith (2008), Ho & Bodoff (2014) and Srinivasan et al. (2002) were used to build questions. Customer satisfaction was assessed by 6-item measures adopted from Anderson and Srinivasan (2003), which developed a satisfaction scale by modifying satisfaction scale items based on Oliver (1980). Loyalty was assessed by adopted from Anderson and Srinivasan (2003), who developed a loyalty scale by modifying loyalty scale items based on Zeithaml, Berry, and Parasuraman (1996). All items were measured using a 7-point Likert-type scale ranging from“strongly disagree" (1) to "strongly agree" (7). The table below shows the development of the measures.
31 Table-2 Development of Measures
Source Scale Type Items
Personalization Implementations
Adapted from research on personalization, e-commerce (Srinivasan, Anderson and Ponnavolu, 2002; Chang and Chen, 2008; Ho and Bodof, 2014)
Seven-point Likert: Strongly disagree (1) to strongly agree (7).
Personalized Product Recommendations (a,b,c,d,e)
Personalized Promotions (a, b, c,d, f)
Written by researcher using literature on personalization (Fan and Poole, 2006; Smith, 2006).
Seven-point Likert: Strongly disagree (1) to strongly agree (7).
Personalized Product Recommendations (f,g) Personalized Promotions (e)
Part C. E-satisfaction
Adapted from Anderson and Srinivasan, 2003; Oliver, 1980
Seven-point Likert: Strongly disagree (1) to strongly agree (7).
E-satisfaction (a,b,c,d,e,f)
Part D. E-loyalty
Adapted from Anderson and Srinivasan, 2003; Zeithhaml, Berry and Parasuraman, 1996
Seven-point Likert: Strongly disagree (1) to strongly agree (7).
E-loyalty (a,b,c,d,e,f,g)
Part E. Demographic Data
Written by researcher using measures similar to Anderson and Srinivasan, 2003
Nominal and open-ended Gender, age, occupation, formal education and ethnicity
4.4 Procedure
Online survey was used as a tool to run the experiment. In the beginning of the questionnaire basic knowledge about personalization was provided to respondents with a graphic
32 representation of personalization implementations. Then they were asked to think about a website of an online retailer that they have shopped and indicate it. Afterwards they were asked to choose the online personalization implementations that are used by the online retailer they indicated and then received questions accordingly. The main idea of the experiment was to measure the relationship between different personalization implementations, satisfaction and loyalty by comparing these 4 groups. Table 3 contains a summary of the treatments. In treatment 1, the respondents indicated that none of the online personalization implementations were used by the online retailer they stated before and they only answered questions about e-satisfaction and e-loyalty. In treatment 2, respondents indicated that only personalized product recommendations were used by the online retailer they stated before and they answered questions about personalized product recommendations, e-satisfaction and e-loyalty. Similarly, in treatment 3, respondents chosen personalized promotions and answered questions about personalized promotions, e-satisfaction and e-loyalty. In the last treatment the respondents claimed both PPRs and personalized promotions and answered all the questions in the survey. Table -3 Summary of the Treatments
Treatment Procedure Number of Respondents
Treatment1 None (no personalization) 34
Treatment2 Personalized product recommendations 44
Treatment3 Personalized promotions 33
Treatment4 Both PPRs and personalized promotions 59
33 Lastly, demographic questions were asked to serve more information about the sample and not used for testing the hypotheses. So, the questionnaire consists of 5 sections: Part A. Website indication and personalization choice, Part B. Personalization implementations, Part C. Satisfaction, Part D. Loyalty, Part E. Demographic questions (see the appendix-5 for the full questionnaire).
Although the online survey enables to collect data in a cost efficient and fast way, there are several limitations. The first limitation is the data used in the study is based on self-reports, not the real behavior. Secondly, the size of the sample is relatively small to enable generalization of the findings of the study.
34
5. Results
In order to see whether there are errors in the data, frequency check was done and no errors were found. The cases with missing values were excluded. Only cases which had no missing data in any variable were analyzed. The recoding of counter- indicative items were applied. Then, construct validity was checked for personalization, e-satisfaction and e-loyalty using exploratory factor analysis (EFA). The results for personalization scales were shown in Table 4. Among 13 items, 4 of them were dropped from the analysis as these items were loaded to different factors. The results of the EFA showed that two factors (personalized product recommendations and personalized promotions) explained 58 percent of the total variance. After factor analysis, a reliability test was conducted with an internal consistency test (Cronbach’s alpha). The measurement items in this study are reliable in that the Cronbach’s alpha values for all the dimensions were more than 0.7. The alpha coefficients for personalized product recommendations and personalized promotions were 0.75, and 0.71, respectively.
Table-4 Reliability and Validity for Personalization
Scale Measured items Factor
Loading Cronbach’s Alpha Personalized product recommendations
This website makes product recommendations that match my needs.
0.859 0.75
This website enables me to order products that are tailor-made for me.
0.731
This website makes me feel that I am a unique customer.
0.636
I believe that this website is customized to my preferences.
35
Personalized promotions
The promotions that this website offers to me are tailored to my preferences.
0.565 0.71
This website sends me e-mails offering personalized promotions.
0.497
The promotions that this website offers to me match my needs.
0.588
I can buy products at more reasonable prices with the personalized promotions that this website offers.
0.882
Personalized promotions provide time efficient shopping experience.
0.608
The results of EFA analysis for e-satisfaction and e-loyalty scales were shown in Table 5. Among 7 e-loyalty items, 2 of them were dropped from the analysis as these items were loaded to different factors. The results indicated that two factors (e-satisfaction and e-loyalty) explained 64 percent of the total variance. Reliability of e-satisfaction and e-loyalty has been measured with internal consistency coefficient Cronbach’s alpha, and found to be 0.86 and 0.81 consecutively, which indicate a quite high internal consistency (Hair, Anderson, Tatham, 1998). Table-5 Reliability and Validity for E-satisfaction and E-loyalty
Scale Measured items Factor
Loading
Cronbach’s Alpha
E-satisfaction I am satisfied with my decision
to purchase from this website.
0.543 0.86
If I had to purchase again, I would feel differently about buying from this website.
0.806
36 this website was a wise one.
I feel badly regarding my decision to buy from this website.
0.865
I think I did the right thing by buying from this website.
0.604
I am unhappy that I purchased from this website.
0.908
E-loyalty I try to use this website
whenever I need to make a purchase.
0.563 0.81
When I need to make a purchase, this website is my first choice.
0.813
I like using this website. 0.694 To me this website is the best
retail website to do business with.
0.850
I believe that this is my favorite retail website
0.898
After this, skewness, kurtosis and normality tests were performed for personalized product recommendations, personalized promotions, e-satisfaction and e-loyalty. Kolmogorow-Simirnov test was used in order to check whether the variables were normally distributed. According to the results, shown in Table 6, all the variables except e-satisfaction were normally distributed or close to being a normal distribution.
37 Table-6 Skewness and Kurtosis of Scales
Scale Skewness Kurtosis Distribution
Personalized product recommendations (n=103) -0.16 -0.16 Normal distribution Personalized promotions (n=92)
-0.58 -0.19 Close to being a normal distribution
E-satisfaction (n=170)
-1.55 2.00 Non- normal distribution
E-loyalty (n=170) -0.24 -0.18 Normal distribution
In order to test the research hypotheses correlation, linear regression analysis, one-way ANOVA and Kruskal-Wallis were conducted.
5.1 Correlation Check
Pearson’s correlation coefficient was calculated to see whether the variables were correlated. The results in Table-7 showed that personalized promotions positively correlated with e-loyalty. A weak, positive and significant correlation was observed between personalized promotions and loyalty (r=0.234, p<0.05). Personalized product recommendations correlated with both e-satisfaction and e-loyalty positively and weakly (r=0.251 and r=0.197). There was a correlation of 0.321 between e-satisfaction and e-loyalty. Also, the means of e-satisfaction and e-loyalty of participants were not very high.
38 Table-7 Means, Standard Deviations and Correlations
Variables Mean SD Personalized
product recommendations Personalized promotions E-satisfaction E-loyalty 1.Personalized product recommendations 4.55 1.06 - 2.Personalized promotions 4.92 0.90 0.664** - 3. E-satisfaction 5.52 0.83 0.251* 0.089 - 4. E-loyalty 4.72 0.91 0.197* 0.234* 0.321* -
** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)
5.2 Hierarchical Regression Analysis
Hierarchical regression analysis was used to examine the relation between the independent variables (personalized product recommendations and personalized promotions) and the dependent variable (e-loyalty). Specifically, hierarchical multiple regression showed the ability of personalized product recommendations and personalized promotions to predict e-loyalty, after controlling for gender, age, occupation, education and ethnicity. As a first step of the hierarchical multiple regression, five predictors were entered: gender, age, occupation, education and ethnicity. The observed effect of personalized product recommendations and personalized promotions on e-loyalty is independent of the effect of these five demographic variables.
39 Table- 8 Hierarchical Regression Model of E-loyalty
R R2 R2 Change B SE β t Step1 0.226 0.051 0.051 Gender -0.090 0.203 -0.049 -0.445 Age -0.019 0.017 -0.139 -1.112 Occupation -0.243 0.344 -0.102 -0.709 Highest education -0.074 0.201 -0.045 -0.368 Etnicity 0.075 0.263 0.038 0.287 Step2 0.322* 0.103* 0.052* Gender 0.077 0.180 0.044 0.429 Age -0.019 0.017 -0.144 -1.163 Occupation 0.039 0.277 0.018 0.139 Highest education 0.179 0.164 0.133 1.095 Etnicity 0.121 0.207 0.067 0.583 Personalized product recommendations 0.240 0.086 0.235 2.732* Step3 0.320* 0.102* 0.051* Gender -0.061 0.199 -0.033 -0.309 Age -0.020 0.017 -0.147 -1.198 Occupation -0.229 0.336 -0.096 -0.680 Highest education -0.045 0.197 -0.027 -0.226 Etnicity 0.088 0.257 0.045 0.342 personalized promotion 0.231 0.108 0.228 2.137*
40 Step4 0.322* 0.104* 0.002 Gender -0.334 0.212 -0.201 -1.571 Age -0.231 0.133 -0.257 -1.496 Occupation -0.125 0.268 -0.072 -0.467 Highest education 0.125 0.174 0.095 0.719 Etnicity 0.408 0.259 0.235 1.576 Personalized product recommendations 0.243 0.126 0.238 2.140* Personalized promotions -0.054 0.146 -0.062 -0.371 *p<.05
In step 1 the model was not statistically significant F (5, 81) = 0.871; p >.05 which means the most variance of e-loyalty was explained by other factors. In step 2 personalized product recommendations was entered as a predictor and the total variance explained by the model as a whole was 10.3%, F(5, 80) = 2,519; p <0.05 which indicates that the introduction of personalized product recommendations explained an additional 5.2% in e-loyalty, after controlling for gender, age, occupation, education and ethnicity. In step 3 personalized promotions was entered as a predictor and the total variance explained by the model as a whole was 10.2 %, F (5, 80) = 2,413; p <0.05 which shows the introduction of personalized promotions explained an additional 5.1% in e-loyalty, after controlling for gender, age, occupation, education and ethnicity. The regression model in Model 4 was obtained by adding the variable personalized promotions to the model. The total variance explained by the model as a whole was 10.4% (F=5.193; p<0.05). As R square change was F (7, 49) =2.182; p>0.05, personalized promotions did not do a significant contribution to the model.
41
5.3 Testing Hypotheses
The first hypothesis suggests that e-satisfaction has a positive direct effect on e-loyalty. In order to test it, the linear regression analysis was conducted by taking e-satisfaction as an independent variable and e-loyalty as a dependent variable. As a result, the model was statistically significant F (1, 169) = 19.346; p <0.001 and explained 10.3 % variance in e-loyalty. It can be concluded from the analysis that e-satisfaction is a statistically significant predictor of e-loyalty (β =0.321, p<0.001). Therefore, H1 can be accepted.
The second hypothesis consists of 3 sub-hypotheses. The first one is that personalized product recommendations have a positive direct effect on e-loyalty. Second is personalized product recommendations have a positive direct effect on e-satisfaction and last one is personalized product recommendations have a positive indirect effect on loyalty through e-satisfaction. In order to test the mediation effect of e-satisfaction, a series of hierarchical regression analyses were conducted by forming the 3-step mediation analyses described by Baron and Kenny (1986). To support the mediation hypothesis, the independent variables (personalized product recommendations) would need to correlate with the mediator (e-satisfaction) as well as with the dependent variable (e-loyalty). Then, the independent variable were regressed on the dependent variable after controlling the effects of the mediator. Full mediation would be established if the relationship between the independent (personalized product recommendations) and dependent variables (e-loyalty) became non-significant. Partial mediation was established if the variance explained by the independent variable was declined in absolute size but was still different from zero after the inclusion of the mediator (Baron and Kenny 1986). As showed in Table-9 Model1 indicated that personalized product recommendations has positive direct effect on e-loyalty, β =0.197, p<0.05. Therefore H2a can be
42 accepted. Also, Model2 showed that personalized product recommendations is a statistically significant predictor of e-satisfaction (β =0.251, p<0.05). Therefore, H2b can be accepted. In model3 it was observed that, when e-satisfaction was added to personalized product recommendations, e-satisfaction became strong predictor of e-loyalty (β= 0.340, p<0.001). While the coefficient of personalized product recommendations in Model1 was β =0. 197, it decreased to β =0. 112 in Model3 and the coefficient was not statistically significant (p<0.05). These results show that there was a full mediation effect between personalized product recommendations and e-loyalty (0.001) (Baron & Kenny, 1986). Therefore H2c can be accepted. Table-9 Impact of the Personalized Product Recommendations on E-loyalty: Results of the Regression Analysis Model1 E-loyalty (β) Model2 E-satisfaction (β) Model3 E-loyalty (β) Independent variable Personalized product recommendations 0.197* 0.251* 0.112 Mediator E-satisfaction - - 0.340*** R2 R2 change F 0.039* 4.075* 0.063* 6.811* 0.383*** 0.147*** 8.601*** *p<.05; ***p<.001
The third hypothesis consists of 3 sub-hypotheses. The first one is that personalized promotions have a positive direct effect on e-loyalty. Second is personalized promotions have a positive direct effect on e-satisfaction and last one is personalized promotions have a positive indirect effect on loyalty through satisfaction. In order to test the mediation effect of
e-43 satisfaction, a series of hierarchical regression analyses were conducted by forming the 3-step mediation analyses described by Baron and Kenny (1986). As showed in Table-10 Model1 indicated that personalized promotions has positive direct effect on e-loyalty, β =0. 234, p<0.05. Therefore H3a can be accepted. Model2 showed that personalized promotion is not a statistically significant predictor of e-satisfaction (β =0.089, p>0.05). Therefore, H3b is rejected. In model3 it was observed that, when e-satisfaction was added to personalized promotions, e-satisfaction became strong predictor of e-loyalty (β= 0. 333, p<0.001). While the coefficient of personalized promotions in Model1 was β =0. 234, it decreased to β =0. 204 in Model3 and the coefficient was statistically significant (p<0.05). However, as the condition that the independent variables (personalized promotion) should correlate with the mediator (e-satisfaction) was not met, there was no mediation effect between personalized promotions and e-loyalty (Baron & Kenny, 1986). Therefore H3c can be rejected.
Table-10 Impact of the Personalized Promotions on E-loyalty: Results of the Regression Analysis Model1 E-loyalty(β) Model2 E-satisfaction (β) Model3 E-loyalty (β) Independent variable Personalized promotions 0.234* 0.089 0.204* Mediator E-satisfaction - - 0.333*** R2 R2 change F 0.055* 5.193* 0.008 0.723 0.165*** 0.110*** 8.763*** *p<.05; ***p<.001
44 In order to investigate the demographic differences between the 4 groups (None-personalized, personalized product recommendations, personalized promotions and both of them) Chi-square test was conducted. As a result, there was not a statistically significant difference found among groups.
The fourth hypothesis consists of 2 sub-hypotheses. The first one is that level of personalization (the number of personalization implementations) has a positive direct effect on e-loyalty. Second one is level of personalization (the number of personalization implementations) has a positive direct effect on e-satisfaction. In order to test the hypotheses, one-way variance analysis (ANOVA) was done. As e-loyalty had a normal distribution, ANOVA can be used. Table-11 shows the differences in e-loyalty among the experimental groups. As a result of the ANOVA there was no statistically significant difference between e-loyalty means of the 4 groups. Therefore, H4a is rejected.
Table – 11 Difference in groups for e-loyalty
Group n Mean Order F Significance
Nonpersonalized 34 4.77 2 Personalized product recommendations 44 4.57 3 2.233 0.086 Personalized promotions 33 4.49 4 Both of them 59 4.93 1
As e-satisfaction did not have normal distribution a non-parametric test Kruskal-Wallis was used instead of ANOVA. Table-12 shows the differences in e-satisfaction among the experimental groups. As a result of the analysis there was no statistically significant difference between e-satisfaction means of the 4 groups. Therefore, H4b is rejected.
45 Table – 12 Difference in groups for e-satisfaction
Group n Mean Mean rank Chi-square Significance
Nonpersonalized 34 5.58 85.00 Personalized product recommendations 44 5.31 76.23 4.328 0.228 Personalized promotions 33 5.48 80.82 Both of them 59 5.66 95.32
