MSc. Business Administration – Marketing Track
Thesis
Student: Giliam Vos - 10868577
Thesis subject: The buffering effect of initial trust on service
recovery evaluations
Thesis supervisor: Frank Slisser
March 2016
Statement of originality
This document is written by student Giliam Vos, 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.
Abstract
The growth and expansion of online retailing is still increasing at a high rate. Within this further developing world of online retailing, the question that arises is whether conceptions and theories about the relationship between consumer and retailer still hold true in a similar vein online as they do offline. Due to the indirect and impersonal nature of online retailing, one area of particular interest is the role that trust plays in the relationship between consumer and retailer. This paper focuses on the relationship between trust and service recovery by researching the effect that initial levels of trust have on the evaluations and attitudes people have after the occurrence of service failure and the following service recovery procedure. The findings show that initial levels of trust create a buffering effect that influences the further process of the transaction between consumer and retailer and the implications this has for post service recovery trust, satisfaction and repurchase intention.
Table of Contents
Abstract………...3 1. Introduction………6 2. Literature Review………10 2.1 Risk of E-commerce………11 2.2 Definitions of trust………..…122.3 Elements of Online trust………..13
2.4 Results from online trust and engagement in e-commerce………...15
2.5 Types of service failure……….15
2.6 Service recovery online………16
2.7 Service recovery paradox………18
2.8 Trust & Service recovery……….19
2.9 Buffering effects………20
2.10 Formulating the hypotheses………...21
2.11 Hypotheses………23 3. Methodology………25 3.1 Research design………25 3.2 Measures………...28 3.3 Webshops………...30 4. Results……….31
4.1 Respondents………...31
4.2 Reliability analysis………..32
4.3 Correlations……….34
4.4 Skewness, Normality, Kurtosis………35
4.5 Hypothesis testing………...36 4.6 Mediation………..39 4.6.1 Satisfaction………...40 4.6.2 Repurchase intention……….42 5. Discussion……….43 6. Conclusion……….48 6.1 Conclusion...……….48 6.2 Limitations………...………49 6.3 Future research……….52 References………..55 Appendix……….59 Appendix A Pretest 1………..59 Appendix B Pretest 2………..60 Appendix C Survey………..62
1. Introduction
The ever-expanding nature of e-commerce and growth of online retailing is an often-discussed subject. Consumers are still increasingly opting to engage in e-commerce and engage in transactions online rather than in more traditional offline methods, such as brick and mortar stores and offline service suppliers. However, the very nature of engaging in online transactions and e-commerce means that it is impersonal, without personal interactions. As such, the barriers for engaging in e-commerce differ from traditional barriers as found in
traditional transactions, and must thus be approached and considered in a different manner. One of the most formidable barriers to people for engaging in e-commerce is often identified as a lack of trust, where people feel uneasy disclosing financial and personal information to online merchants (Wang & Emurian, 2005). Therefore, in this current time where the intertwining of online services, activities and e-commerce is steadily increasing, it is more important than ever for companies to actively seek out ways to deal with this natural tendency of distrust that people display towards e-commerce.
Research on the role of trust in e-commerce is generally focused on both finding and defining the antecedents of trust, and the effects of trust on purchasing behavior. While these are all important areas of research with managerial implications, the focus of this paper addresses a different angle by instead looking beyond that initial stage to settings where a disturbance in the trust relationship between the merchant and consumer has occurred by means of a
service failure, and consequently the service recovery effort that follows this disturbance.
Although ideally there would be no need for such a thing as service recovery, since in a perfect scenario there would be no such thing as service failure, reality is unfortunately different. Furthermore, due to the characteristics of
e-commerce, where there are few barriers stopping people from abandoning and switching companies, it is arguably even more important than before to maintain and sustain relationships with customers, especially in those moments when the incentives to switch are elevated such as when a service failure occurs. However, extending that relationship further it can be asked whether levels of trust that existed before the customer engages in a transaction also influence their levels of satisfaction after service recovery. Trust is an important and essential element of the relationship between consumer and company, and also plays an important role in customer satisfaction. By linking the role of trust with service recovery, the aim is to gain a closer look into how the interaction between a company and consumers takes place. Say for example that a customer has a high degree of trust towards a company; now if they experience a service failure, will they experience higher levels of dissatisfaction based on the fact that they had higher expectations of said company, or will that moderate their level of
disappointment? And if said customer then receives a satisfactory service recovery, how will this influence their opinion? Again, one could argue that a high level of trust combined with a successful service recovery would strengthen their initial belief and trust in the company and thus mean they would have a higher level of satisfaction, as the company delivered according to their
expectancies, despite the “hiccup” that occurred in the middle. But at the same time it could be argued that higher trust would lead to higher levels of
expectation, meaning that even though the company successfully recovered from the service failure, they would be more dissatisfied because it did not initially deliver as expected. In a way this ties into the service recovery paradox, where a lower level of satisfaction is actually raised due to the service recovery taking place, but in this case it would move from a higher level to a lower than normal level.
Following this line of reasoning, the level of trust ex ante could thus function as a buffer, which could soften the impact of the post service recovery evaluations in terms of trust, but also in terms of satisfaction and repurchase intention. Since initial trust consists of factors that can be directly or indirectly manipulated and controlled by companies, such as information quality, security and privacy protection, third-party seals, buying conditions (Kim et al., 2008), the relationship between ex ante trust and post ante trust, satisfaction and repurchase intention is worth exploring.
But at what point do people start to lose trust in a company? And is this an irreversible process, or can a company regain this loss of trust by taking certain actions or displaying certain behavior? In what way does the occurrence of service recovery influence people’s post service recovery satisfaction, trust and repurchase intention? And what buffering effect does pre-transactional trust play on these post service recovery measures? There has been little attention to the buffering effect of trust on post service recovery in an online environment,
where the online nature means that different rules apply to it, as opposed to an offline physical setting, thus leading to a research gap about the relationship between pre-transactional trust, service recovery and post-transactional evaluations.
Within this gap the focus will be on:
1. The buffering effect of trust on post service recovery evaluations.
2. The impact of the change of circumstances of service recovery taking place in an online setting.
As stated earlier the relevance of this research gap stems from different
applications. Part of it comes from the increasing rise of online retailing, which makes it relevant for research to be conducted into more various aspects of the online retail landscape and into what the consequences of this new and different landscape are for merchants. Furthermore it draws attention to an aspect of the complicated relationship between trust and service recovery processes, which as of yet is not fully understood in the online context. What has been shown is that the same set of rules and principles as a whole that exist in offline settings cannot be applied to online settings in the same manner, but what rules out of this set can be transferred and applied online is still unclear. Lastly, the concept of influencing post transaction and post service recovery measures of trust, satisfaction and repurchase intention by manipulating pre transactional aspects provides a huge opportunity for companies to increase their customer
basis as is currently often done in terms of dealing with service failure and service recovery.
Given these implications of the aforementioned reasons, this research aims to improve the understanding of and attempt to fill that gap by answering the following question:
To which extent are post service recovery levels of trust, satisfaction and repurchase intention moderated by initial levels of trust?
The next chapter reviews the existing literature on the topics of online consumer trust and service recovery. This leads to the formulation of the hypotheses and the construction of the conceptual models. After this the methodology will be introduced. The following chapter will present the results from the survey, after which these results will be discussed, the research question will be answered and the limitations of this research will be addressed. The final chapter will then discuss the conclusions that can be drawn based on this research and provide suggestions for further research in this area.
2. Literature review
This chapter looks at past research on the topics of e-commerce, trust (in online settings) and service recovery online. The first part will deal with risks and trust
in online commerce. The second part looks more specifically at service failure and the service recovery that follows it. The third part addresses buffer effects and ties that together with the research question, where the focus is on the buffering effect of trust. The goal of the literature review is to identify the elements that will be used to formulate and test the hypotheses.
2.1 Risk of e-commerce
Web based shopping removes many geographic barriers between customers and merchants, leading to a more distant relationship between the two parties (Van Slyke, Belanger, & Comunale, 2004). However, because of this distant
relationship there is room for uncertainties to emerge. The nature and characteristics of Internet cyber transactions (blind, borderless, 24/7, non instantaneous) in it self are a potential source of concern for consumers, who face uncertainty whether or not the seller will fulfill his transactional obligations (Kim, Ferrin, & Rao, 2008). Vulnerability also emerges from the fact that in general, consumers pay for a transaction ahead of time (Van Slyke et al., 2004), although the emergence of new “consumer beneficial” payment methods is steadily countering this perceived vulnerability, through companies such as PayPal and credit card companies offering increased consumer protection. Nevertheless, these vulnerabilities coupled with the existence of cybercriminals have led to an environment where trust between the business and the consumer must somehow be earned and is not a given in the way that many people
experience it in brick and mortar stores, where the face-value is considered sufficient to generate high enough levels of trust. Most online merchants are not
trusted enough for consumers to immediately engage in relationship exchanges with said online merchants (Smith, 2004). In order for e-commerce to be a worthy equivalent of traditional commerce, this trust is needed for both short- and long term effectiveness (Smith, 2004). The building of trust is therefore seen as a fundamental, yet largely unresolved issue in the development of Internet shopping (Lee & Turban, 2001). Trust can help alleviate the risk of online purchase decisions, by decreasing the risk and increasing the willingness to buy online (as opposed to buying offline) (Lee & Turban, 2001). In e-commerce, reputation and assumed beliefs about a company play an even larger role than in an offline setting because there are fewer visible signs of credibility, and at the same time greater perceived risks (Qureshi et al., 2009). As such, having a good reputation leads to trust which influences initial purchase intention and the willingness to engage in business with an online merchant.
2.2 Definitions of trust
Trust can be considered one of the foundations of e-commerce (Pavlou & Gefen, 2004), and as such in order to talk about a concept as consumer trust in e-commerce, a proper working definition of trust must first be specified. Although the importance of trust is widely recognized, there is disagreement about its definition, characteristics, antecedents, and outcomes (Lee & Turban, 2001). Kim, Ferrin, & Rao (2008) provide the following definition: "a consumer's subjective belief that the selling party or entity will fulfill its transactional
obligations as the consumer understands them." This approximates the definition that will be used throughout this paper in the sense that it details both the belief
of the consumer as well as mentioning the transactional obligations, but doesn't fully cover the entire spectrum of trust as used here. Yakov (Bart, Shankar, Sultan, & Urban, 2005) use a different, more wide definition, with a more predefined notion of the direction and content of the relationship: “Trust is a psychological state comprising the intention to accept vulnerability based on positive expectations of the intentions or behaviors of another.” This specifically mentions the positive expectations and intentions of the other involved party, and notes the
psychological state, as opposed to the previously mentioned, more subjective belief. In the context of this paper a definition that is more tailored to relational exchange will fit the bill better, such as a willingness to rely on an exchange partner in which one has confidence (Wang & Emurian, 2005). As trust implies the existence of a relationship between different parties, as does the
aforementioned willingness to rely on, this definition seems more encompassing, yet broad enough to work with. Thus in this paper the definition as provided by Wang & Emurian (2005) will be used, as it accurately captures the relationship between consumer and online merchant. After establishing this working definition, the different elements that together make up trust must be defined.
2.3 Elements of online trust
Antecedents of trust are factors that produce a sense of trustworthiness and can help determine whether consumers trust a merchant (Wang & Emurian, 2005). In order to further facilitate the studying of consumer trust, Lee & Turban (2001) have developed a research model that presents the major relationship between consumer trust in e-commerce. They distinguish between four groups of major
potential antecedents of trust: trustworthiness of the Internet merchant, trustworthiness of the Internet shopping medium, infrastructural factors, and other factors. These all potentially influence trust, whereas trust itself moderates purchase intention and perceived risk (Kim et al., 2008). Wang (2008) finds the generally observable characteristics of trust to be trustor and trustee,
vulnerability, produced actions and subjective matter. In this Lee & Turban (2001) specifically mention the shopping medium, whereas Wang (2008) puts less emphasis on it by integrating it into the trustor and trustee. This could in part be explained by the difference in publication dates, where there was a higher degree of standardization and general acceptance of certain formats of shopping mediums at the time of the latter publication by Wang than the 2001 Lee & Turban article. Standardization in terms of format as well as higher standardization of (renowned) payment methods are all potential contributors to this. Yet, separating the merchant and medium nowadays seems unnecessary as the website is commonly considered the storefront for the merchant and is thus intertwined with it in terms of identity and the level of trustworthiness it evokes.
Another aspect of online trust that is different from traditional forms of trust between consumers and merchants is the importance of post purchase trust on the internet (Ha, Janda, & Muthaly, 2010). In the absence of this post purchase trust, merchants may be unable to serve previous customers, despite their previous satisfactory experiences (Ha et al., 2010). As such, online
post-transactional trust acts as a moderator for satisfaction and repurchase intention, which in term influences further behavior. Such behavior may include engaging in interactivity with a site, revisiting a site and eventually possibly repurchasing
from a site previously interacted with (Ha et al., 2010). As such, Ha et al. (2010) argue that online trust based on previous experiences play an important role in facilitating consumers’ further behavioral intentions. Yet at the same time, it cannot be assumed that trust in itself will lead to repurchase intention in the way that it does in offline settings (Qureshi et al., 2009).
2.4 Results from online trust and engagement in e-commerce
The presence of a sufficient level of trust from a consumer towards a merchant will influence their behavior. Behavioral intent can range from clicking on websites, to using the websites and ordering products (Bart et al., 2005). Trust affects the consumer’s attitude and risk perception, which then influences the willingness to buy (Bart et al., 2005). Wang (2005) separates this into two different categories of produced action, being actual purchasing behavior and “window shopping.” Both of these are beneficial to merchants, as they result in sales or potential sales. Furthermore, even potential shoppers provide data about surfing behavior and patterns of interest that can be valuable to
merchants. However, given the obvious higher benefits from actual sales over potential sales, the focus lies on the former.
2.5 Types of service failure
When consumers decide to engage in e-commerce, or for that matter deal with any type of commerce, they naturally expect their transaction to run smoothly and without problems or complications. Unfortunately, this does not always hold
true. Issues that arise specifically in the case of e-commerce include privacy, credit card security (payment security), delivery time and ease of navigation. These issues are considered critical elements of e-service quality (Holloway & Beatty, 2003). If a company fails to deliver on these elements, customers view this as initial service delivery falling below the customer’s expectations or “zone of tolerance” (Holloway & Beatty, 2003). This can then potentially lead to the loss of customers or the emergence of negative word of mouth (Lin, Wang, & Chang, 2011). It is at this point that companies must employ service recovery strategies in order to prevent the aforementioned loss of customers to occur. Customer retention is generally cheaper than the costs associated with acquiring new customers, which gives a strong incentive to firms to put extra effort into retaining their current customers (Holloway & Beatty, 2003). Thus, if companies successfully use recovery management, they can increase customer loyalty, satisfaction, repurchase intention and a number of other desirable types of consumer behavior (Holloway & Beatty, 2003). Customer satisfaction in the context of e-commerce can be defined as “the perceived degree of contentment with regard to a customer’s prior purchase experience with a given electronic commerce firm,” and repurchase intention as “the customer’s self-reported likelihood of engaging in further repurchase behavior” (Ha et al., 2010). 2.6 Service recovery online
Service recovery can take place at three different stages: preventive (before the service failure occurs), concurrently (just as the service failure occurs) and post hoc (sometime after the service failure occurred) (Miller, Craighead, & Karwan,
2000). Miller et al. (2000) split up the recovery activities into psychological and tangible efforts. The psychological efforts revolve around showing concern for the customer’s situation, by both empathizing and apologizing. The tangible efforts involve compensation for both real and perceived damage. This is primarily to reimburse the customer, but can also be used to overcompensate the customer relative to actual costs suffered in order to atone for the bad experience. The line between what constitutes fair compensation and
overcompensation is vague, and depends on the degree of distributive justice: to what degree the customer considers the outcome fair, deserved and necessary (Lin et al., 2011). An important element of the service recovery process lies in the crucial role of frontline employees (Chang, 2008). They are often the first interaction point with customers and serve as an important bridge between the consumer and the company. This naturally poses difficulty for online service recovery, as they are online and therefore lack frontline employees, a proper physical storefront and personal face-to-face interaction. The nonexistent nature of these elements also underlines the importance of web site design, as it puts more emphasis on the role that the design of a website fulfills as the equivalent of a storefront to a business (Bart et al., 2005). The online storefront is where the perceptions of trust are built, through a decrease in the sense of vulnerability (Rose, Clark, Samouel, & Hair, 2012). Elements that heighten a consumer’s perception of integrity and trustworthiness of a merchant can be a well designed and organized user interface and accuracy of product descriptions. User-friendly search and navigation options give consumers a sense of control and safety, which in turn translates into positive conceptions about a merchant (Qureshi et al., 2009). These are all elements that add to the building of an initial feeling of
trust towards a merchant, and arguably still influence trust perceptions in potential further repurchasing contexts.
2.7 Service recovery paradox
Ultimately, customer satisfaction is one of, if not the, main goal and key outcome of the marketing process. Anderson & Srinivasan (2003) define online customer satisfaction as “the perceived degree of contentment with regard to a customer’s prior purchase experience with a given electronic commerce firm.” While service failures are unavoidable, they can be turned around and actually proven to add value and strengthen the relationship between the customer and the company (Chang, 2008). The service recovery paradox is a situation in which the post recovery levels of satisfaction exceed those prior to the service failure, if the customers receive high-level recovery service (De Matos, Henrique, & Rossi, 2007). While this has been researched and commonly believed to be true in offline settings (Rose et al., 2012), the online variant of this is an often-debated subject. There is disagreement about whether or not it is possible to generate higher levels of satisfaction after a successful service recovery when compared to before, in an online setting (Magnini, Ford, Markowski, & Honeycutt Jr, 2007). On the one hand, because customers retain a higher level of control in the
transaction, and feel more responsible for its outcome, poor service is said to have less effect on online customers (Harris, Grewal, Mohr, & Bernhardt, 2006). However, this does not necessarily imply that a successful recovery generates higher levels of satisfaction. Although trust and satisfaction have a direct effect on repurchase intention(Rose et al., 2012) the switching costs of e-commerce are
much lower, to the point of almost zero, which has been shown to have a
negative impact on the repurchase intentions of customers. Several studies have shown that service failure and the effect it has on perceived quality of service, regardless of the valuation of the service recovery, reduces repurchase intention (Cranage & Mattila, 2006). Even if customers are satisfied with the recovery effort, they will most likely not repurchase after experiencing a failure (De Matos et al., 2007). It is generally believed that the service recovery paradox doesn’t manifest itself online the way it does in offline settings, if it does at all (Harris et al., 2006). This could be considered as an indication that solely focusing on the service recovery process as a means to improve post service recovery
evaluations is not sufficient.
2.8 Trust & service recovery
Trust is a key variable in managing customer relationships. Research shows that customers who were initially satisfied with the service they received (without the occurrence of service failure) expressed higher levels of trust than those customers who experienced a failure, yet were satisfied with the service recovery process afterwards (De Matos et al., 2007). In that situation, despite reaching a level of satisfaction similar to those who did not experience failure of any kind, there is still a loss in the element of trust between the customer and the company. Given the relevance of trust in the relationship, this represents an important variable. This loss of trust will be greater or lower depending on how the customers perceive the failure. Failures can have either temporary or
subjective to how it is perceived by the customers, will lead to a decrease in trust, resulting in a decrease in repatronage intentions (De Matos et al., 2007). Trust in an online merchant captures the extent to which the merchant is
deemed competent, benevolent and to which degree he has shown high integrity in the relationship between customer and merchant (Qureshi et al., 2009)
2.9 Buffering effects
Previous research on buffering effects in the context of service recovery have looked at the satisfactory history transactions (Magnini et al., 2007) between a customer and a company and customer relationships (R. L. Hess, Ganesan, & Klein, 2003), but not at the impact of trust as a buffering effect. Hess et al. (2003) show that customer-organization relationships can help protect an organization from the negative effects of failure on customer satisfaction. A buffering effect can stem from expectations consumers hold about the continuity of their
relationship with an organization. Those customers that expect to have a longer lasting relationship tend to have lower expectations of the service recovery effort, which in turn led them to experience a higher satisfaction after the actual service recovery effort takes place. Interestingly, this is in contrast with the expectation that they would be more demanding with regard to the recovery effort. Hess et al. (2003) explain this by stating that customers with high expectancy for relationship continuity may be more tolerable and demand less immediate compensation because they consider the balance of equity across a longer horizon of exchanges. Now if this logic is extended to trust, then it could be argued that higher initial trust in a merchant could imply the expectation of a
longer lasting relationship. Trust is one of the factors that differentiate
relationships from transactions (J. Hess & Story, 2005). Thus, a high level of trust could be seen as the initiation of a relationship between a customer and a
merchant. As such, this would create a buffer effect on the post service recovery evaluations similar to the one as found by Hess in relationship buffers. Thus, this leads to question whether the level of initial trust can help buffer post service recovery outcomes. If a higher level of trust can lead to a more tolerable attitude towards service recovery processes, this would in turn lead to higher evaluations post service recovery.
2.10 Formulating the hypotheses
Based on the findings from the literature review, trust plays an essential role in the decision making process of consumers to engage in business with online merchants. Trust positively moderates repurchase intention and satisfaction. The occurrence of service recovery on the other hand, negatively influences trust, repurchase intention and satisfaction. A way to counter the negative occurrence of service recovery in traditional offline settings manifests itself through means of the service recovery paradox, where the post service recovery evaluations are higher than they would have been, had there not been a service failure and a consecutive service recovery process. However, the general consensus as found by research is that this paradox is not or hardly present in online environments, as the switching costs are so low to the point of almost zero. Now, given that post service recovery trust again moderates repurchase intention, this shows how important the factor of trust is in the whole process.
Since repurchase intention, which can be seen as a form a loyalty, is in the end a crucial - if not the ultimate - goal as this means a continuation of the relationship between the consumer and the merchant, the question becomes how to
stimulate this in such a way that the repurchase intention goes up. In the case of a negative experience, stimulating may not be possible but at the very least minimize the negative influence of the bad experience.
As the quality of the service recovery is not a decisive factor in ensuring that the post service recovery evaluations are of such a level that people feel comfortable continuing the existing relationship by engaging in repurchasing, the attention is drawn towards factors that exist before the occurrence of the service failure and recovery. This raises the question of whether an initial level of trust could
positively moderate the post service recovery evaluation levels. Moreover, service recovery processes are a personal, tailored product and experience. The initial trust that a website or vendor evokes in a consumer can to a degree be standardized, with certain elements that can be proven to generally invoke higher levels of trust. As such, this would allow a vendor to target the whole of potential consumers who visit their website, as opposed to the service recovery process that needs to be tailored to each consumer individually. Thus, arguably it would make sense to pay attention to and invest in creating a higher level of initial trust rather than focus on specific cases.
This thus led me to formulate the following hypotheses to test whether or not the initial levels of trust can have a positive impact on post service recovery
evaluations, and more general the influence of trust on customer satisfaction and repurchase processes.
2.11 Hypotheses
H1: A higher level of initial trust (pre service recovery) positively moderates the relationship between the quality of service recovery and post service recovery satisfaction levels.
H2: A higher level of initial trust (pre service recovery) positively moderates the relationship between the quality of service recovery and the level of repurchase intention.
H3: A higher level of initial trust (pre service recovery) positively moderates the relationship between the quality of service recovery and post service recovery trust levels.
H4: Post service recovery repurchase intention is mediated by post service recovery trust.
H5: Post service recovery satisfaction is mediated by post service recovery trust.
These expectations are tested by means of a 2x2 survey in which the initial levels of trust and the perceived quality of service recovery were manipulated to
respectively be high and low in different compositions. They were tested quantitatively using a research design that pitted the different combinations of high and low trust and service recovery level against one another in order to compare the changes.
By researching this, the aim is to gain a better understanding of the factors that influence customer satisfaction. Although prevention of service failure will always be preferable to having to engage in service recovery, it is important to identify other areas where there is room for improvement for the overall customer satisfaction, as that can be considered one of the main goals of marketing. Previous research that has shown that even successful recovery service still leads to a loss in trust and repurchase intention. An increase in consumer loss and switching behavior make this a relevant topic of research, as it would provide insight into ways to ideally manage this by leveraging the buffering effect of initial trust. Finally, although this research does not address the problem of avoiding the occurrence of service failure in itself, it does focus on elements that can be improved or adjusted before the occurrence of the service failure.
Figure 2. The conceptual model of hypothesis 4.
Figure 3. The conceptual model of hypothesis 5.
3. Methodology
3.1 Research design
The purpose of this study is to measure the buffering effect of service recovery quality on post service recovery valuations of trust, satisfaction and repurchase intention. This is tested in an online environment, as there are an ever-increasing number of online retailers who are both opening and expanding their businesses
and an overall move away from offline consumer retail. An experiment will be used in order to test for the hypothesis as outlined in the previous section. In the experiment, all participants are exposed to the same conditions in order for the experiment to be deemed valid, save for the planned manipulation of the
independent variable and the moderator.
In this online retailing context, the hypotheses were developed to specifically address the characteristics of the environment. In the experiment two large Dutch online retailers will be used as subjects for the testing. Although ideally the experiment would be based on real life experiences with the two subjects, due to time constraints and difficulties in finding people who have had these experiences first hand, hypothetical situations were designed and presented to participants. For the hypothetical situations vignettes will be used. A vignette is a constructed description of a situation, where a systematic combination of
characteristics is used (Atzmüller & Steiner, 2015). The vignette used in this research is a 2x2 design, varying in website used and the quality of the service recovery. Different groups of respondents get different vignette sets, but each subject within a group will receive the same vignettes. Subjects are randomly appointed to different vignette sets. The data was collected through online self-administered questionnaires.
In order for the survey to approximate real life situations as best as possible, two existing and widely known webshops were selected for the hypothetical
situations. The selection of these webshops was done by using data taken from consumers’ associations websites, where one webshop that is highly regarded as
a good, trustworthy webshop with high ratings and one webshop generally deemed less trustworthy, with low ratings were selected. The website www.klachtenkompas.nl, an initiative by the leading Dutch consumers’ association, was initially used to select the two webshops. Criteria for the selection of the webshops were that they had 1000 complaints or more. This criterion was personally set based on a review of the companies with the highest amount of complaints. Of those companies, the first two that met the search criteria of all-round online retailer, as opposed to a seller of services (post, railways) or intangible goods (e.g. internet / mobile phone provider). Within this criterion, two webshops stood out based on their positive and negative
evaluations. On the website, customers file a complaint with a company. The reason being the higher likelihood of getting a response due to the public nature and widespread usage of the website. The company either publicly responds to the complaint lodged, or the customer follows up on their earlier complaint, detailing the continuation and handling of the complaint. Complaints only count for the total amount of complaints if they are marked as resolved, either in a satisfactory or non-satisfactory manner. The comments and replies from both consumer and company are moderated so as to ensure the objectivity and prevent the unnecessary slandering of specific companies. After marking a case resolved, customers are able to rate their satisfaction with the complaint
handling and the company in general on the website.
At the time of writing, Bol.com had 1330 complaints, of which 83% were
considered satisfactory. This both met with the minimum number of complaints (1000), and was also the highest rated all-round webshop within the group of
companies with a 1000 or more complaints. Other higher ranked companies were either selling services (e.g. mobile phone contracts, internet contracts) or provided a less tangible service (postal service, railways). For a poorly
performing webshop the lowest rated webshop that met the minimum
requirement of a 1000 complaints was Bcc.nl, with a 33% satisfied rating and a 67% unsatisfied rating over a 1019 complaints. These ratings were checked on another consumer goods review site, www.tweakers.net, which confirmed their overall ratings, with a 2/5 rating or lower for bcc.nl and a 4/5 or higher rating for bol.com. While this does not necessarily mean that these shops command high or low levels of trust on an individual basis, it increases the chance of this being the general consensus, especially in the light of word of mouth and easy access to this information. This thus provides with a first step in constructing a low and high level of initial trust. Further testing is then done to confirm their status as low and high levels of initial trust when combined with a description of the respective sites. Two pretests were developed in anticipation of the survey in order to test the content of the survey questions. One pretest was run to test for the evaluation of service quality and one pretest was run to measure what level of trust people received from webshop descriptions.
3.2 Measures
The survey first asks respondents for their demographics (age, gender, nationality). They will then be asked whether they are familiar with the two webshops (respectively Bol.com and Bcc.nl) and the degree of experience they have with them (closed questions). After this they are presented with one of two
webshop descriptions, which are either high or low in initial trust. After this their initial trust is measured. The initial trust measures are adopted from Koufaris his article about the development of initial trust in online companies (Koufaris & Hampton-Sosa, 2004). Then they are presented with one of the two service recovery scenarios, which are respectively high and low in quality. After the scenario 4 more sets of questions are asked, regarding service recovery quality, trust, satisfaction and repurchase intention. A 7-point Likert scale will be used for all constructs, ranging from completely agree to completely disagree. Trust, satisfaction and repurchase intention are adapted from Zboja his paper on the impact of brand trust and satisfaction on retailer repurchase intentions (Zboja & Voorhees, 2006). The decision to use different measures for pre service recovery initial levels of trust and post service recovery measures of trust
stemmed from the fact that the initial trust measures from Koufaris are geared toward the impression people get, whereas the trust measures from Zboja are aimed at how they experienced their interaction with the retailer. The service recovery quality measures are adapted from Andreassen his paper on
antecedents to satisfaction with service recovery (Wallin Andreassen, 2000). As the survey will be administered in Dutch the constructs have been translated and checked by peers to ensure accuracy of the translation.
3.3 Webshops
Initially two pretests were run to ensure the validity of the measures used in the later survey. Since the survey would consist of two webshops, which would be representative of high and low initial trust levels, a pretest was done to make
sure that the two webshops that were selected accurately displayed high and low initial trust levels. The pretests were run with fake names (Company A, Company B). The goal of the pretest was to ensure that one company had high initial trust levels and one company had low initial trust levels based on the description that was provided along with the company. Participants in the pretest were shown both descriptions in random order to avoid specific bias by always starting with the low and going to the high or the other way around. The pretest (N=21) is shown in Appendix A. The scores for all 5 questions were between 5 and 6 on a 7 point Likert scale for the high trust company, and between 3 and 4 on a 7 point Likert scale for the low trust company.
The second pretest was done to see if the two scenarios accurately showed high and low service recovery quality. Again, participants were shown the
descriptions in random order to avoid specific bias by always starting with the low and going to the high or the other way around. The pretest (N=20) is shown in appendix B. The score for the high service recovery scenario was very high allover, ranging from 5,7 to 5,9 out of 7 for all three questions. The score for the low service recovery was very low, ranging from 1,3 to 1,8 out of 7 for all three questions. After the pretest, the scenario for the low service recovery quality was adjusted and made slightly less negative and more realistic, in order for it to not overpower the potential effect of the initial trust.
4. Results
4.1 Respondents
The population for this sample stemmed from a convenience sample, and was selected through self-selection on social media, e-mail and snowball sampling (Saunders & Lewis, 2012). The choice for convenience sampling was due to the high number of respondents needed and the limited amount of time available for data collection. The results were collected using the Qualtrics survey software. All participants were required to be of Dutch nature in order to qualify for participation in this research due to the nature of the survey, which relies on familiarity with the selected webshops. A price consisting of a voucher was provided as an incentive to boost response rate. A total of 204 responses were collected, which met the initially set required minimum of 50 respondents per group. As the questions in the questionnaire were made compulsory to answer, there was no need to discard incomplete questionnaires.
The division of male and female respondents to the survey was moderately equally divided, with 59% of the respondents being male and 41% being female. The sample as a whole was relatively young, which is an unintended effect of the convenience sample, with over 50% of the respondents falling in the 20-29 category. The potential implications of this will later be discussed in the limitations section. 97% of the respondents replied that they were Dutch and currently living in Holland. 97% also stated that they sometimes use webshops
to buy products online. An overwhelming 89% of the respondents said that they had at some point bought from Bol.com before, thus ensuring their familiarity with the website. 9% said they had not bought from Bol.com before, but they were familiar with the website. Although less people had bought from Bcc.nl before, 87% said they had either bought from there or were at least familiar with it.
The respondents were randomly assigned to one of the four conditions, of which one received 53 responses, one received 51 responses and two received 50 responses.
Groups Number of respondents
Bol.com high recovery scenario 51 Bol.com low recovery scenario 50 Bcc.nl high recovery scenario 53 Bcc.nl low recovery scenario 50 Table 1. Respondents per group.
4.2 Reliability analysis
A reliability analysis was performed in order to check whether the measures consistently reflected the construct that it is measuring, and is generally indicated by Cronbach’s Alpha. A Cronbach's Alpha coefficient is generally believed to be reliable when it measures 0.7 or higher. The results of the
Likert scales used. All scales were checked whether their reliability could be improved by deleting an item, but none could substantially (∆>0.1) improve from the deletion of an item. The inter-item correlation of the scales was also checked. All scales met the inter-item correlation of >0.3, except for one item from the trust scale, where 3 of the inter-item correlations were below 0.3 (0.245, 0.202, 0.226) and 2 were relatively low (0.348, 0.417). However, the Cronbach’s Alpha would only increase from 0,884 to 0,916, which is not a big (>0.1) improvement, and the Cronbach’s Alpha score of 0,884 is already high. Furthermore, since there was sufficient inter-item correlation with 2 of the other items within the scale, rather than no correlation with any of the items, the decision was made to keep them in the scale. Apart from the Trust scale, all other scales scored high on the Cronbach’s Alpha score, meaning there was good consistency among the items contained in the scales. None of the scales would substantially improve if an item were to be deleted.
Cronbach’s Alpha Number of items
Initial trust 0,933 5
Quality of service recovery 0,98 3
Trust 0,884 6
Satisfaction 0,938 5
Repurchase intention 0,966 3 Table 2. Reliability analysis.
4.3 Correlations: Mean SD 1 2 3 4 5 6 7 1.In. T. 4.70 1.47 1 2.Trust 3,95 1,55 ,274** ,000 1 3.SAT 3,66 1,80 ,326** ,000 ,881** ,000 1 4.Rep. Int. 3,61 2,00 ,343** ,000 ,845** ,000 ,924** ,000 1 5.QSR 3,94 2,41 ,186** ,000 ,858** ,000 ,854** ,000 ,854** ,000 1 6.Fam. Bol 1,24 0,69 -.114 ,008 -,118 0,093 -,122 ,081 -,169* 0,016 -1,29 ,066 1 7. Fam. Bcc 2,57 1,03 ,026 ,710 ,046 ,511 -,030 ,672 -,047 ,505 -,015 ,836 ,026 ,716 1 Table 3. Correlations.
1 = initial trust, 2 = trust, 3 = satisfaction, 4 = repurchase intention, 5 = quality of service recovery, 6 = familiarity with Bol.com, 7 = familiarity with Bcc.nl
The correlations table shows significant effects between the dependent variables measured. All the correlations between the dependent variables are positive, meaning that if one increases the other one also increases. All the correlations (trust, satisfaction, repurchase intention, quality of service recovery) are significant at the p <.000 level. The correlations with the control variables are
not significant except for familiarity with Bol.com and Repurchase intention, which is significant at the p <.05 level. Familiarity with Bcc.nl does not show any significant correlations with the dependent variables. This means that whether respondents were familiar or not with either Bol.com or Bcc.nl does not have a significant effect on the dependent variables. The correlations between the variables are in line with the hypotheses. Initial trust is positively related to trust, repurchase intention and satisfaction.
4.4 Skewness, Normality, Kurtosis
Normality was checked for trust, repurchase intention and satisfaction. The skewness score shows the symmetry. Trust was slightly negatively skewed with -.210 and had a kurtosis of -.930. Repurchase intention was ever so slightly skewed to the right with -.050, and had a kurtosis of -1.349. Satisfaction was very slightly skewed to the left with .063, and had a kurtosis of -1.261. All three
variables were thus approximately symmetric, but relatively flat due to their negative kurtosis scores.
Skewness Kurtosis
Trust -.210 -.930
Repurchase Intention -.050 -1.349 satisfaction ,063 -1,261 Table 4. Skewness and kurtosis results.
4.5 Hypothesis testing
First a factorial MANOVA was run to measure the different means. The choice for a MANOVA stemmed from the fact that there are multiple variables correlating with each other, as can been seen in the correlation table. The fixed factors consisted of four different groups; Bol.com with low quality service recovery, Bol.com with high quality service recovery, Bcc.nl with low quality service recovery and Bcc.nl with high quality service recovery. In this, the Bol.com groups were the groups with high initial trust and the Bcc.nl groups had low initial trust. This was corroborated by the descriptive statistics, which showed similar high scores for the groups belonging to Bol.com, and similar low scores for the groups belonging to Bcc.nl. The dependent variables included trust (post service recovery), satisfaction and repurchase intention. A second MANOVA was run that only differentiated between Bol.com and Bcc.nl.
The first MANOVA showed a significant (p = .002) Box’s M, which means that the covariance in matrices across all groups is not the same. However, Pillai’s Trace is significant at p < .000, meaning that there is significant variance between the groups. The groups showed a significant effect: F(9, 600) = 19,773 , p < .000, Pillai’s Trace = 0,686, partial η2 is .229. The test of between subjects showed significant effects for all the dependent variables on the four groups.
The table below shows the post hoc test results using a Tukey post hoc test. In order to observe the occurrence of a buffering effect, the comparisons were made between groups with the same recovery scenario. This means that
Bol.com with a high quality recovery scenario (BOLH) is coupled with Bcc.nl with a high quality recovery scenario (BCCH) and Bol.com with a low quality recovery scenario (BOLL) is coupled with Bcc.nl with a low quality recovery scenario (BCCL).
Post Hoc results
Mean dif. Std. error. Sig. Lower bound Upper bound Trust BOLH-BCCH ,752 ,179 ,000 ,287 1,217 BOLL-BCCL ,617 ,183 .005 ,143 1,091 Repurchase intention BOLH-BCCH ,954 ,259 ,002 ,283 1,625 BOLL-BCCL ,913 ,264 ,004 ,229 1,597 Satisfaction BOLH-BCCH 1,025 ,228 ,000 ,433 1,616 BOLL-BCCL ,812 ,233 ,003 ,209 1,415
Table 5. Multiple comparisons.
Trust had a mean difference of .75 in the high recovery group, with a significance of p <.000. The mean difference in the low recovery group was .62, with a
high recovery group, p = .002, and a mean difference of .91 for the low recovery group, p = .004. Satisfaction had a mean difference of 1.03 in the high recovery group, p < .000, and a mean difference of .81 in the low recovery group, p =.003. These results show a trend that the mean difference in evaluations in higher in for the high recovery scenario than for the low recovery scenario, which could indicate a stronger buffering effect of initial trust for high recovery rather than for low recovery. This is also confirmed by the significance levels, which show a trend in favor of the high recovery scenario. All 3 variables were significant at the p < .005 level, except for trust in the low recovery scenario which showed a p = .005 score. The implications of these results will be discussed later.
The results show support for the hypothesized buffering effect as the groups with high initial trust show significant difference from the groups with low initial trust in both the high and the low service recovery quality scenarios. Hypothesis 1,2 and 3 stated that a higher level of initial trust would positively influence the post service recovery evaluations of trust, repurchase intention and satisfaction for similar service recovery scenarios. If there is a consistent higher evaluation for the groups with higher initial trust then that would point to a buffering effect caused by the initial trust levels. The comparison table shows support for this, as there is a significant difference between the evaluations for Bol.com and Bcc.nl in both the high and the low service recovery situations.
After the first MANOVA, a second MANOVA was run comparing Bol.com and Bcc.nl. One group consisted of Bol.com with high recovery scenario and low recovery scenario. The other group consisted of Bcc.nl with high recovery
scenario and low recovery scenario. This test was run to see the overall effect of high initial trust (Bol.com) and low initial trust (Bcc.nl) regardless of the service recovery scenario.
Box’s test of equality of covariance matrices showed a significance of p = .518, f=,868 and Pillai’s Trace showed a significant effect f(3,200)=4.440, p = .005, Pillai’s Trace =,062. The tests of between subjects confirmed the results of the first MANOVA and gave a significance of p = .002 for trust, p < .000 for
satisfaction and p = .001 for repurchase intention. In this test there was no discrimination between the high and the low recovery situations, which further confirms the influence of the initial trust levels on the outcomes.
The MANOVA test was then run again for four different control variables: age, gender, familiarity with Bol.com and familiarity with Bcc.nl. None of the control variables had any significant influence on the results and as such were not used for further testing.
4.6 Mediation
In order to test for the mediating effect of post service recovery trust, a
regression was run using the Process macro for mediation (model4) by Hayes. The quality of the service recovery was taken as the independent variable, repurchase intention and satisfaction as the outcome variable and post service recovery trust as the mediator variable. Two separate tests were run for both hypotheses. A nonparametric resampling procedure of 2000 times was used to reduce inaccuracy in the outcomes, as the variables are not normally distributed.
The results of the mediation tests show support for hypotheses 4 and 5. Post service recovery trust as a mediator has a larger effect on satisfaction than on repurchase intention.
4.6.1 Satisfaction
M (Trust) Y (Satisfaction)
Antecedent Coeff. SE p Coeff. SE p X (QSR) A ,5533 ,0233 <.001 c’ ,2769 ,0444 <.001 M (Trust) - - - b ,6518 ,0689 <.001 Table 6. Consequent of the regression analysis.
Figure 4. Conceptual model of hypothesis 4.
The indirect effect through the mediator trust was ,3606, meaning that two respondents who differed by one unit in their reported quality of service
recovery are estimated to differ by 0.3606 units in their satisfaction as a result of the tendency for those who perceive the quality of service recovery as lower to feel less trust, which in turn translates into lower satisfaction. This indirect effect
is statistically different from zero, as can be observed by a 95% BC bootstrap confidence interval that is entirely above zero ((,2755; ,4511).
The direct effect of quality of service recovery is ,2769, meaning that someone who perceives a higher quality of service recovery but a similar level of post service recovery trust will be an estimated .2769 units higher in their
satisfaction. The direct effect is statistically different from zero, t=6,2390, p=,000, with a 95% confidence interval (,1894; ,3645).
The total effect of quality of service recovery on satisfaction is =,6376. The positive outcome means that a person who perceives the service recovery quality higher will feel greater satisfaction. The effect is statistically different from zero, t=23,3422, p=,000, and a 95% confidence interval of (,5837; 6914).
The result of the mediation test shows support for hypothesis 4, which states that trust acts as a mediator between quality of service recovery and satisfaction. The direct effect is smaller than the total effect (,2769 < ,6376), meaning that there is partial mediation.
4.6.2 Repurchase Intention
Consequent of the regression analysis
M (trust) Y
(Repurchase Intention)
Antecedent Coeff. SE p Coeff. SE p X (QSR) A ,5533 ,0233 <.001 C’ ,3995 ,0529 <.001 M (trust) - - - B ,5341 ,0820 <.001 Table 7. Consequent of the regression analysis.
Figure 5. Conceptual model of hypothesis 4.
The indirect effect through the mediator trust, was ,2955, meaning that two respondents who differed by one unit in their reported quality of service
recovery are estimated to differ by 0.2955 units in their repurchase intention as a result of the tendency for those who perceive the quality of service recovery as lower to feel less trust, which in turn translates into lower repurchase intention.
This indirect effect is statistically different from zero, as can be observed by the 95% BC bootstrap confidence interval that is entirely above zero (,1923; ,4182).
The direct effect of quality of service recovery on repurchase intention was ,3995, meaning that someone who perceives a higher quality of service recovery but a similar level of post service recovery trust will be an estimated .3995 units higher in their repurchase intention. The direct effect is statistically different from zero, t=7,5560, p=,000, and 95% confidence interval of (,2952; ,5037).
The total effect of quality of service recovery on repurchase intention was ,6949. The positive outcome means that a person who perceives the service recovery quality higher will feel higher repurchase intention. The effect is statistically different from zero t=23,3412, p=,000, with a 95% confidence interval (,6362; ,7536).
The result of the mediation test shows support for hypothesis 5, which states that trust acts as a mediator between quality of service recovery and repurchase intention. The direct effect is smaller than the total effect (,3995 < ,6949),
5. Discussion
To what degree can initial trust be used as a buffering effect in post service recovery evaluations? The main finding of this research shows the existence of a buffering effect in the relationship between initial trust, service recovery and post service recovery evaluations. As the results of this research show, initial trust can be leveraged to have a buffering effect on post service recovery evaluations. The results confirmed the hypothesized buffering effects from hypothesis 1,2 and 3. Furthermore, the importance of the role of trust in general was further underlined by the results of hypothesis 4 and 5, which showed a mediating effect for trust in relation to satisfaction and repurchase intention.
A webshop can be considered an online storefront for a retailer. Much like an offline storefront, consumers will form their initial impression based on its appearance, functioning and features. However, unlike an offline store there is no direct interaction with employees, who provide a certain level of assurance and trust. Furthermore, payments are generally done in advance, which
heightens the uncertainty of whether a seller will fulfill his transactional duties (Van Slyke et al., 2004). This uncertainty is furthered by the fact that there is no physical location to return to in case of a service failure or other type of problem. As such, the building of trust is fundamental for consumers in their decision making process of whether to engage in business with an online retailer or not (Lee & Turban, 2001). But moving beyond the initial decision to engage in business with an online merchant, there is the aspect of service failure. The
occurrence of a service failure has high implications for consumers’ post service recovery evaluations and their potential repurchase intention from the same retailer in the future. The influence of the quality of the service recovery differs between offline and online settings. In offline settings, a well-executed service recovery can lead to an increase in the post service recovery evaluations of the consumer, the service recovery paradox, and thus increase the chance of future business with the same retailer. This effect however does not hold true for online settings, where even positive evaluations of the service recovery process most likely will not lead to repurchasing after a service failure has been experienced (De Matos et al., 2007).
It was for this reason that the focus of this research was to test whether it was possible to influence post service recovery evaluations with elements other than the quality of the service recovery. Due to the importance of initial trust in the initial purchase decision, it was tested whether initial trust could also influence the post service recovery evaluations and create a buffering effect. In order to test this, a 2x2 design was created in which the initial levels of trust and the quality of the service recovery were manipulated, in order to allow comparisons between the different combinations of initial trust and quality of service
recovery.
As the results of the research showed, there are significant differences in post service recovery evaluations that stem from a difference in initial levels of trust, ceteris paribus. Respondents who were exposed to a webshop with high levels of initial trust (Bol.com) gave higher post service recovery evaluations for trust,
repurchase intention and satisfaction. This occurred regardless of whether they were shown the high quality or the low quality service recovery process.
However, there was a trend that showed higher mean differences for the high recovery scenario than for the low recovery scenario. Although the differences were still significant in the low recovery scenario, this seems to point towards a stronger buffering effect for high quality service recovery. This finding was unexpected, as the assumption was that the effect would be stronger in a low quality recovery scenario, as it leaves more room for improvement. When looking at it in this way, the buffering effect seems to be stronger in improving already positive evaluations rather than decrease or nullify negative evaluations. A possible explanation could be the relatively large difference in the high and the low recovery scenario, where the low recovery scenario was set so low that it interfered with the buffering effect. Another explanation could be that in general, the impact of a low quality service recovery is relatively greater than that of a high quality service recovery procedure. In such, the findings are in line with earlier research that found the absence of a service recovery paradox in online settings, because the switching costs are low to the point of (almost)
nonexistence.
Looking further into the differences between the high and the low quality of service recovery groups, repurchase intention had the most similar mean difference between the two groups: .95 mean difference for the high quality service recovery group, and .91 for the low quality group. Trust had a mean difference of .74 for the high quality group, and .62 for the low quality group. Satisfaction had the biggest difference, with 1.02 for the high quality group and
.81 for the low quality group. What this means is that on average, the buffering effect has the most impact on satisfaction: in the case of a high quality service recovery, it changed the average evaluation by over 1 point on a 7 point Likert scale. On the other hand, the buffering effect was the smallest for the post service recovery trust evaluations. For repurchase intention, there was a relatively small difference between the mean difference for the high and low quality group. It would thus appear that the difference between the high and low quality service recovery, in terms of how big the buffering effect is, is more or less equal in both high and low quality service recovery situations for repurchase intention, as opposed to satisfaction and trust where it appears stronger in high quality service recovery scenarios.
The second part of the analysis looked at the mediating effect of post service recovery trust on repurchase intention and satisfaction. The results supported both hypothesis 4 and 5, showing a significant mediating effect on both
repurchase intention and satisfaction. This means that respondents for a significant amount base their opinions about satisfaction and repurchase
intention on the level of trust they experience from a retailer. This is in line with the literature that stressed the importance of trust in the relationship between the consumer and the online merchant. A higher level of initial trust will lead to a higher level of post service recovery trust, regardless of the quality of the service recovery process, which in turn improves the satisfaction and the repurchase intention.
6. Conclusion
6.1 Conclusion
The research question that was posed at the beginning of this thesis was: to which extent are post service recovery levels of trust, satisfaction and repurchase intention moderated by initial levels of trust? As the results of this research show, there is a significant relationship between initial trust and trust, repurchase intention and satisfaction, in the context of online retailing. Hypothesis 1 to 3 stated that a higher level of initial trust (pre service recovery) positively moderates the relationship between the quality of service recovery and post service recovery evaluation levels for respectively satisfaction, repurchase intention and trust. The findings supported all three hypotheses, showing the buffering effect that initial trust has on post service recovery evaluations. Hypothesis 4 stated post service recovery repurchase intention is mediated by post service recovery trust, and hypothesis 5 stated that post service recovery satisfaction is mediated by post service recovery trust. Both these hypotheses were also found to be supported, and further supported the mediating effect that trust has on satisfaction and repurchase intention. The support for all of the hypotheses leads to some interesting new insights into the relationship between a consumer and a merchant in an online setting.
These findings point towards areas of interest for the enhancement of customer-retailer relationships. One of the problems that customer-retailers encountered in the
online setting that is different from the offline setting is the absence of the service recovery paradox. Now within the service recovery paradox the majority of the effort made by the merchant takes place after the service failure has occurred. In line with the findings from this research, shifting that effort to before the service failure occurrence could be a suitable solution for companies aiming to minimize the damage that they incur from service failure. The key in this lies in building good rapport with customers by increasing their
trustworthiness. This is further emphasized by the mediating role of trust in the post service recovery situation, in which it has a significant positive influence on satisfaction and repurchase intention.
Consumers place high value on their initial impression of a webshop. It is building on this initial trust that they evaluate all other aspects that occur
afterwards, whether they have a positive or negative experience with a company. Thus, having a high initial trust creates a buffering effect that can be felt
throughout the rest of the transaction up until after the initial transaction is completed in future repurchasing behavior.
6.2 Strengths and limitations
The quantitative nature of this research allows the collection of a large number of respondents. However, this also means that the results will be correlational rather than causal, and will thus more likely show the existence of relationships between different variables instead of explaining why these relationships exist. Another limitation is that due to time restraints as well as access to respondents,
this research has to rely on fictional scenarios instead of real consumer experiences. Ideally, all the participants would base their answers on real life experiences that they had with service recovery processes from the two webshops used in this research. Furthermore, the use of only two scenarios, respectively a high quality recovery scenario and a low quality recovery scenario, is a limitation. Using more scenarios would lead to a better
understanding of the strength of the buffering effect on different levels of service recovery quality, as it could show the differences between more various levels of quality of service recovery.
For the survey, it was opted to use existing webshops. This is both a strength and a limitation: by using existing webshops that the vast majority of the
respondents were familiar with, and a large number of the respondents had personal experiences with, it allowed for a less lengthy introduction of the webshops which in turn may have aided in the high the response rate.
Furthermore, the descriptions of the webshops as provided in the survey were taken directly from their own websites with no alterations done to them, in order for them to approximate reality as closely as possible. In this, it was a realistic example of two large webshops. However, a limitation is that people’s personal experiences with said webshops could distort their objectivity in answering the questions, as it could lead to a bias of sorts. Furthermore, although an attempt was made to select two webshops that were similar in product offerings, size and fame, it is naturally not possible to find an exact match. In this case, Bol.com was definitely the larger of the two with a wider selection of products ranging from electronics to books and everything in
