Literature Review

Accepted to International Journal of Information Management

Understanding the effects of a social media service failure apology:

A comparative study of customers vs. potential customers

Authors:

Dr Danae Manika, Lecturer in Marketing

Queen Mary University of London, School of Business and Management Bancroft Building, Mile End, London, E1 4NS

Email [email protected]

Professor Savvas Papagiannidis*, Goldman Chair of Innovation

Newcastle University Business School, UK

5 Barrack Road, Newcastle upon Tyne, NE1 4SE

Email: [email protected]

Professor Michael Bourlakis, Head of Supply Chain Research Centre

Cranfield School of Management

Cranfield, Bedford MK43 0AL, UK

Email: [email protected]

* corresponding author

Understanding the effects of a social media service failure apology:

A comparative study of customers vs. potential customers

Abstract

Service failure apologies on social media are a new crisis communication outlet trend used by companies to apologise to affected customers quickly and offer solutions, ultimately to restore customers’ trust and brand loyalty. This paper contributes to the nascent literature on companies’ social media service failure apologies and fills a gap in the social commerce literature by recognising that due to the open and public nature of social media, these apologies may reach not just affected customers, but also unintended audiences such as potential customers among the general public, which could potentially damage a company’s reputation and market share. An online survey administered to 241 customers and 271 non-customers of a famous mobile phone brand, which used YouTube to apologise to its customers for a service failure incident, is used to explore potential behavioural outcomes, after exposure to the apology. Findings confirm that both customers and non-customers of the service provider may become exposed to a social media service failure apology. The hypothesised model predicts behavioural intentions to remain a customer after exposure to the social media service apology better than behavioural intentions to become a customer, even though relationships hold for both groups. Theoretical and managerial implications are discussed.

Keywords: service failure apology, social media, behaviours, customers, non-customers.

Understanding the effects of a social media service failure apology:

A comparative study of customers vs. potential customers

1. Introduction

Social interactions of Internet users, especially on social networking sites, have created a new stream in e-commerce, often referred to as social commerce (Hajli 2013). Businesses can either adopt social media by integrating social commerce functions to their existing infrastructure or even create separate self-contained operations based on their social media presence. Such activities make it possible for businesses to interact and engage more actively with customers, which in turn can result in a better understanding of their needs and the development of stronger relationships (Hajli 2014b). At the same time, social commerce empowers customers by offering a number of tools which they can use to generate relevant content and co-create value with businesses when interacting with them online (Fuller et al. 2009). The above can enhance and enrich the transacting process and result in tangible and sustainable benefits for businesses. In such a fast-paced environment, though, many things often go wrong and may result in negative consumer sentiment towards a business, which can propagate through online networks very quickly. Consequently, managers should not only consider how to create, develop and sustain online communities, but also how to respond to crisis situations when these occur.

Prior literature recognises the frequency of service failure incidents faced by the services industry (Chuang et al. 2012) and the need to apologise to affected customers in order to manage and maintain relationships (Dalziel 2014). A service failure apology is a message communicated by a representative of a company to affected customers who experience a temporary service discontinuity. With this message, the company aims to address and restore equity by stating its wrong-doing, acknowledging that the customers did not receive the arranged benefits of the service they have paid for and apologising to affected customers (Boshoff and Leong 1998). Companies that recover quickly from a service failure incident tend to have higher levels of praise and recommendations of the service from those affected (Amine 1998), which is why social media service failure apologies have recently become popular, due to their ability to reach large audiences in a short period of time (Manika et al. 2015; Dalziel 2014).

This is a phenomenon that is likely to become more significant in the future given the growing importance and popularity of social media. Consequently, studying the underlying processes and channel idiosyncrasies has practical and theoretical interest. Mattila et al. (2011) note that customers prefer a technology-oriented communication channel when dealing with technological service failures (i.e. the recovery mode should be matched with the failure type﴿. Crisis communication managers thus, have a choice between various technology-oriented communication channels, such as blogs, emails or social media among others, for disseminating a technological service failure apology. According to Kerkhof et al. (2011) the choice of medium for communicating crisis response messages has a clear effect on consumers and they found that Twitter got a more positive response than blogs. This may be explained by the fact that Twitter enables rapid feedback, which is also a cornerstone of social commerce and relevant to this study. Prior research also shows that rich media, which contain videos, enable quick information flow and feedback (Nardi, 2005), which may be vital for a crisis communication’s success and lead to different ways of information processing and rapid two-way communications. Therefore, we cannot assume that a service failure apology delivered via any technology-oriented communication channel will have the same attitudinal and behavioural effects on consumers. Thus, further investigation into the effects of communicating a social media service failure apology specifically is needed.

In addition to the above, technology-oriented communication channels such as mobile communications and emails would only reach customers affected by the service failure incident. However, audience-targeting techniques on social media are rather hard to control. A social media service failure apology may not just reach customers affected by the incident, but also customers who may not be affected by the service failure incident and even the general public, which includes non-customers / potential customers (Note: these two terms will be used interchangeably from now on in this paper), who are part of a social media network. The general public is of interest to managers and should be nurtured for the benefit of a company’s reputation (Shamma and Hassan, 2009). Companies have long recognised that maintaining existing customers is more profitable than acquiring new ones. However, the general public’s perceptions of a service provider as a result of a service failure incident (and associated apology) should also be of interest. Service providers need to recognise that these apologies on social media may also affect the general public’s perceptions of a company’s reputation, with important behavioural implications, especially when it comes to sectors with high customer churn.

This paper aims to fill a gap in the literature by critically probing into existing and related work in the area of social commerce and specifically on service failure apologies. It does so by examining the attitudinal and behavioural outcomes of an apology on social media, while also exploring differences between customers and non-customers (i.e. the general public). Given that customers have a greater familiarity with the company/brand experiencing the service failure incident than the general public, customers may have more sophisticated knowledge structures (Low and Lamb 2000; Grime et al. 2002); they thus process information from the apology differently (Manika and Gregory-Smith 2014) and attach a different level of importance to corporate associations (Bravo et al. 2009). Therefore, attitudinal and behavioural outcomes after exposure to a social media service failure apology may have different associations (stronger or weaker) between them when examining customers versus non-customers after exposure to a social media service failure apology.

In addition, this paper contributes to the general service failure apology literature by looking at the effects of subjective knowledge about the service/brand before the apology, trustworthiness of the service/brand and attitudes towards the service/brand (independent of satisfaction) after the social media service failure apology and its effects on behavioural intentions. Elliott et al. (2009) note that the use of online videos, such as those embedded on social media, leads to larger investments (the behaviour) than online text. Therefore, the link between knowledge and behaviour may be enhanced in the context of social media service failure apologies. Trust is also a cornerstone of e-commerce (Hajli 2014; Hajli et al. 2014a; Hajli et al. 2014b) and within the social media arena trust has been found to influence behavioural intentions (Han and Windsor 2011; Lu et al. 2010; Shin 2010), as interactions among the connected users increase trust (Swamynathan et al. 2008). Consequently, trust is an important predictor of behavioural intentions in the context of social media service failure apologies, aside from the general antecedent of satisfaction. Attitude is a multi-dimensional construct, which does not only include a satisfaction dimension. Other attitudinal dimensions are investigated here (see measurement of attitudes). Further relevant literature on social commerce and social media, service failure apologies’ attitudinal and behavioural outcomes is reviewed in the next section of this paper.

2. Literature review

2.1. Social Commerce and Social Media Research

Social commerce is defined in short as “commerce activities mediated by social media” (Curty and Zhang 2011, p.1). A recent review of social media research by Ngai et al. (2015) illustrates the various theories, constructs and conceptual frameworks used to understand this new phenomenon and it underscores the importance of social media as a communication vehicle for consumers, businesses and their commerce activities. Within the context of e-commerce, social media users can be seen as consumers who serve as actual or potential customers for companies and managers (Chung and Austria 2010), which is why they are considered as part of a company's marketing activities (Kim and Ko 2012). Consumers use social media to not only communicate with brands (Hajli 2014a; 2014b), but also communicate with other social media users (Hajli and Lin 2014; Chevalier and Mayzlin 2006). “While consumers are concerned about the credibility of online information, they benefit from social support and are increasingly turning to social media as a source of information and support” (Hajli et al. 2014a, p.1). As a result, social media have become a vital information and communication tool for consumers, managers, and companies.

Most research on social media has focused on (dis)satisfaction with the company/brand under normal operations. However, there is limited prior literature examining social media as a tool to disseminate a service failure apology quickly (Manika et al. 2014; Dalziel 2014), which may result in more positive post-incident and apology attitudinal and behavioural outcomes (since rapid response to crisis is imperative in service failure incidents as per Amine 1998). Domino’s Pizza, JetBlue, BlackBerry and Netflix are examples of companies that have used social media apologies, offering an indication of the rising popularity of service failure apologies on social media and the need further research (Seward 2011).

2.2. Outcomes of Service Failure Apologies

Generally, the relationship between satisfaction and behavioural intention has been confirmed by numerous recent studies in the information management arena (Kuo and Wu 2012; Udo et al. 2010). Perceived satisfaction and brand loyalty (i.e. the behavioural intention to remain a customer) after a service failure apology are two of the most commonly investigated outcomes of studies investigating companies’ apologies to customers (Amine 1998; Corritore Kracher, and Wiedenbeck 2003; Pace, et al. 2010). For instance, Amine (1998) argues that satisfaction acts as an antecedent of brand loyalty, while Wirtz and Mattila (2004) found that satisfaction was a mediator between service recovery and post-recovery behaviour (e.g. repurchase intent and negative word of mouth), and that satisfaction fully-mediated the relationship between an apology and brand loyalty.

Even though the relationship between post-incident customer satisfaction and behaviour has been confirmed in the prior service failure apology literature, Wirtz and Mattila (2004) note that responding to the service failure incident in a timely manner is essential and can affect this relationship. Social media provide an efficient way of responding to a service failure incident and communicating information to affected customers. This can work in favour of the service provider by leading to more positive post-apology customer satisfaction, given the quick company response, and more positive behavioural intentions to remain a customer. Accepting responsibility or apologising for the impact of a service failure is just not enough to ensure customers are satisfied with the company after an apology (Kellerman 2006; Pace, et al. 2010).

Combined with Kerkhof et al.'s (2011) point of view that the choice of medium for communicating crisis response messages has different effects on consumers, this paper seeks to re-examine the relationship between satisfaction and behavioural intentions, in the context of a YouTube service failure apology (see methodology for more information on this). YouTube has a medium social presence/media richness (it ranges from low to high; higher richness reduces ambiguity and uncertainty) and low self-presentation/self-disclosure (it is either low or high and indicates people’s desires to control the impressions others have of them) according to Kaplan and Haenlein (2010) (Venäläinen 2014). This is why “YouTube can then serve as a good communication channel for organizations in order to produce official statements as it is in the mid-ground by not being too media rich and leaving authority for the organization by not being high in self-disclosure” (Venäläinen 2014, p.25).

In addition, acknowledging the mistake, expressing regret and offering reassurance that the offence will not be repeated (Kellerman 2006) are key ingredients of a persuasive and successful apology. Pham and Avnet (2004) note that the persuasiveness of a message (in this case the apology) depends on the message’s ability to influence an opinion, on whether or not attitudes have changed as a result of exposure to the message, and the likelihood of influencing habits. This is consistent with the traditional definition of persuasion as a change in attitudes after exposure to a message (Petty and Cacioppo 1986). The persuasiveness of a service failure apology may also be affected by the choice of communication channel. Elliott et al. (2009), who compared the use of online videos and online text to announce restatements, found that participants watching the restatement via an online video made larger investments in the firm than did participants reading the announcement via online text. Therefore, we could infer that online video may be more persuasive than online text and social media, which allow the embedding of videos, may influence the persuasiveness of the apology. This paper thus fills a gap in research by examining these relationships after exposure to a social media service failure apology, as its effects may differ from other technology-oriented communication channels.

Based on all the above it is hypothesised that:

H1: Persuasion of the social media service failure apology will be positively related to satisfaction with the company after the social media service failure apology.

H2: Satisfaction with the company after the social media service failure apology will be positively related to behavioural intentions (to remain or become a customer).

Trust is a key antecedent to the continued use of technology-oriented products and services, as it is for their adoption (Corritore, et al. 2003; Gefen, Karahanna, and Straub 2003). Trust cannot be assumed to spill over from the pre- to the post- service failure incident stage, especially due to the relative ease with which customers can switch from one service to another. Given that a service failure apology can affect the level of trustworthiness associated with the service provider after the service failure incident, companies need to embrace such an eventuality carefully, transforming a service failure incident period to a trust-forming one, in order to maintain brand loyalty (positive behavioural intentions). Trust (benevolence and credibility dimensions as per Hajli et al. 2014b) is also a cornerstone of e-commerce (Hajli 2014; Hajli et al. 2014a; Hajli et al. 2014b) and within the social media arena, trust has also been found to influence behavioural intentions (Han and Windsor 2011; Lu et al. 2010; Shin 2010), as interactions among the connected users increase trust (Swamynathan et al. 2008).

Particularly in the context of social media, users are concerned about the credibility of information but at the same time are provided with support from other users (Hajli et al. 2014a), reducing uncertainty and ambiguity, as well as creating a feeling of being part of a community network. Thus, it is expected that for the extra support that social media provide as a channel for its users, persuasiveness of the apology will be an important predictor of trustworthiness of the service after exposure to a social media service failure apology, which in turn will greatly influence behavioural intentions. In addition, trustworthiness of the service provider after an apology has not been previously investigated in prior service failure apology literature but it is vital given the context of social media (this is another contribution of this study).

H3: Persuasion of the social media service failure apology will be positively related to the perceived trustworthiness of the service provider after exposure to the social media service failure apology.

H4: Perceived trustworthiness of the service provider after exposure to a social media service failure apology will be positively related to behavioural intentions (to remain or become a customer).

Ngai et al. (2014) also note that in the social media arena, attitude is a significant predictor of behavioural intentions. Attitudes are general predispositions towards other people, objects, and issues used to evaluate them in a favourable or unfavourable manner, which may influence behaviour (Petty, Priester and Brinol, 2002; Manika and Gregory-Smith 2014), as consumers often “tend to act favourably toward things they like … and unfavourably toward things they do not like” (Petty, Priester, and Brinol 2002, p. 2). A change in attitudes according to the Elaboration Likelihood Model of Petty and Caccioppo (1986) constitutes a persuasive message by definition. Brinol et al. (2015) note that consumers have naïve theories about persuasion, which affect message responses. Consumers develop beliefs about how persuasion works and whether a particular message tactic is acceptable or appropriate, which in turn will dictate whether the views of the persuasion message are good or bad (Freistad and Wright 1995; Brinol et al. 2015). Therefore, it is not only important in the context of service failure incidents for the apology to be persuasive but also to be perceived as appropriate and to lead to positive attitudes towards the service provider/brand.

Given the previous literature discussed earlier in this literature review (e.g., Elliott et al. 2009; Kaplan and Haenlein 2010) YouTube’s richness as a channel is more likely to translate the persuasiveness of an apology into positive attitudes towards the service provider/brand, which in turn will lead to higher behavioural intentions. This might not be the case for other technology-oriented channels where the naïve theories consumers hold about persuasion might affect the message responses negatively. For example, Brinol et al. (2015) argue that this may happen if consumers find a persuasive message insincere. YouTube as a channel may decrease this possibility due to its richness.

The socially supporting environment of social media may also lead to more trustworthy perceptions of the service brand (Hajli et al., 2014a) and satisfaction with the service/brand, which is why these constructs are investigated in this study, independent of attitudes towards the service/brand. However, it should be noted that prior consumer behaviour literature sometimes measures attitudes in a way that includes a satisfactory versus unsatisfactory dimension (e.g. Goldsmith et al. 2001) or even specific measures of attitudes based on trustworthiness (e.g. Erdem et al. 1998; Erdem et al. 2004). In this study, the measure of attitudes does not include these dimensions but treats them as separate constructs given that satisfaction is often a key construct investigated in relation to service failure apologies and trust is a key variable in e-commerce; therefore, both need to be investigated separately but without neglecting other dimensions that the attitudinal construct may include and how these additional attitudinal dimensions may affect behavioural intentions. Therefore, we expect that in the context of social media service failure apologies both the trustworthiness of and satisfaction with the service/brand will influence attitudes towards the service/brand.

H5: Persuasion of the social media service failure apology will be positively related to attitudes towards the service provider after exposure to the social media service failure apology.

H6: Satisfaction with the company after the social media service failure apology will be positively related to attitudes towards the service provider after exposure to the social media service failure apology.

H7: Perceived trustworthiness of the service provider after exposure to a social media service failure apology will be positively related to attitudes towards the service provider after this exposure.

H8: Attitudes towards the service provider after exposure to a social media service failure apology will be positively related to behavioural intentions (to remain or become a customer).

2.3. Antecedents of Service Failure Apology Outcomes

In addition, three consumer characteristics relevant to the social media service failure apology should be taken into account, given their likelihood of affecting the aforementioned attitudinal and behavioural outcomes of exposure to a social media service failure apology: prior exposure to the apology, incident familiarity, and subjective knowledge of the service provider.

More specifically, social media service failure apologies can remain available online for extended periods of time, even after the company addresses the service failure and fixes any problems associated with it, thus, social media users may also be exposed to the apology message more than once. The number of exposures to a service failure apology has important implications for the effectiveness of a message (apology). Message repetition (in this case repetition of the apology) can reduce negative outcomes, according to Berlyne’s (1970) two-factor theory. However, prior research also shows that too much repetition of a message may also result in loss of effectiveness, due to wear-out effects, as the consumer/audience gets more familiar with the message's content (Campbell and Keller 2003). Campbell and Keller’s (2003) study was conducted in a traditional advertising setting. Within the context of social media, and specifically YouTube examined here as a channel for communicating an apology, social media users are less likely to “read” and/or “replay” the apology message after a certain number of exposures to the message, as users “pull” information from social media, as opposed to in a traditional communication setting the message being “pushed” on consumers (i.e. consumers cannot control the number of exposures to the message). Therefore, given that social media allow “pulling” information and based on Berlyne’s (1970) two-factor theory we hypothesise that:

H9: Prior exposure to the social media service failure apology will be positively related to the persuasion of the apology.

Next, incident familiarity, the extent to which consumers think they are familiar with the service failure incident (this refers to a continuum of familiarity not a binary measure and could also be seen as subjective knowledge of the service failure incident – but not termed as such here to avoid confusion with subjective knowledge of the service/provider, also measured and examined in this paper), may also affect prior exposure to the social media apology. The more familiar customers are with the service failure incident (e.g. when affected by the discontinuity), the greater the likelihood will be of them being exposed to the social media service failure apology. After all, the apology is by definition targeting affected customers. Even customers not directly affected by the discontinuity who have greater perceptions of their knowledge, in this case in regards to the service failure incident, will wish to remain knowledgeable (as Golden and Stanaland (2000) note “knowledge begets knowledge”) about the incident, and therefore are more likely to be exposed to the apology message, even when the message is not directly targeted to non-affected customers.

It is also expected that incident familiarity, which, according to the above, increases exposure to the apology message, is also going to positively affect the persuasiveness of the social media service failure apology. Consumers who think they are more familiar with the incident will more likely perceive themselves as more capable of judging the apology, and understanding whether or not the apology is sincere and appropriate (Brinol et al., 2015), and thus increase their confidence in the persuasiveness of the message. As per the definition and usage of a service failure apology, the company aims to address and restore equity by stating its wrong-doing, acknowledging that the customers did not receive the arranged benefits of the service they had paid for and apologising to affected customers (Boshoff and Leong 1998). Remembering that YouTube decreases ambiguity and uncertainty, while leaving authority for the organization, it is more likely that the apology on social media will be perceived as appropriate and sincere, and thus the more customers know about the service failure incident the more likely they will be persuaded by it. In addition, for customers who have more information about the incident from a source that they trust (they bought the service after all) and consider trustworthy, an apology could have a positive impact on their perceptions of persuasion and in turn make them believe the apology. For non-customers (and customers too) the more information they can find about the case, especially when consumers see the company trying to fix the problem, the more transparent the incident may look and hence consumers may be more likely to believe the apology is a genuine one.

Incident familiarity will also be affected by the perceptions of knowledge a customer may have about the service/brand, even for non-affected customers. The subjective knowledge (Brucks, 1985) consumers have about the service/brand is similar to the construct of incident familiarity, in regards to the fact that both are perceptions of knowledge, but the difference lies in the object of the perceived knowledge. Incident familiarity is perceived knowledge of the service failure incident, while subjective knowledge in this paper refers to the perceived knowledge of the service/brand. Consumer behaviour literature does note that there is a strong link between the amount of subjective knowledge and information search (Brucks 1985; Carlson, Vincent, Hardesty, and Bearden 2009; Moorman, Brinberg, and Kidwell 2004; Raju, Lonial, and Mangold 1995.) Therefore, consumers with greater subjective knowledge about the service/brand are more likely to continue to seek information about the service/brand to remain knowledgeable (Golden and Stanaland 2000), and thus are more likely to become familiar with the service failure incident, even when not affected. Given the nature of social media, exposure to the service failure apology can be seen as an information seeking activity (Hajli 2014a).

Lastly, subjective knowledge about the service/brand will also have an impact on the behavioural intentions to remain customers, given the link between subjective knowledge and behaviour, as per Alba and Hutchinson’s review of how prior knowledge affects consumer behaviour. (2000). Subjective knowledge tends to have a greater effect on behaviour than truly objective knowledge about a product, service or brand (Brucks 1985). “Knowledge is the fundamental basis of competition” (Lopez-Nicolas and Molina-Castillo, 2008, p.102). The more someone thinks he/she knows about a service/brand the more likely he/she is to remain a customer as he/she will wish to remain knowledgeable, rather than for example go with a competitor service/brand, of which the consumers may have limited or no knowledge. The context of social media also makes a difference given that Elliott et al. (2009) found that online videos led to larger investments (the behaviour) than online text. Thus, social media may also influence the strength of the subjective knowledge behaviour link due to the fact that social media allow the embedding of social media. This relationship has not been previously investigated in a service failure apology context either and therefore it is important to examine this, the latter being another contribution of this paper.

Thus, based on all the above, we put forward the following hypotheses:

H10: Subjective knowledge of the service provider before the social media service failure apology will be positively related to behavioural intentions (to remain or become a customer).

H11: Subjective knowledge of the service provider before the social media service failure apology will be positively related to incident familiarity.

H12: Incident familiarity will be positively related to prior exposure to the social media service failure apology.

H13: Incident familiarity will be positively related to the persuasion of the social media service failure apology.

2.4. Customers versus Non-customers

This paper seeks to examine the previously aforementioned hypotheses empirically in the context of social media service failure apologies and uncover whether or not these relationships are also confirmed within the social media context, as is also the case with other technology-oriented communication channels examined in prior service failure apology literatures; the latter relates to Kerkhof’s et al. (2011) notion that different communication channels may have different effects on consumers. It should be noted that this study also examines consumer behaviour constructs not often investigated in service failure apology studies for affected (by the discontinuity) customers: trustworthiness of the service/brand after exposure to the apology and subjective knowledge of the service/brand before the apology. In addition, the study aims to confirm that, given the nature of social media, non-customers may be exposed and affected by the service failure apology as well as customers, while also exploring: a) differences if any between customers versus non-customers in terms of the constructs (antecedents, attitudinal and behavioural outcomes) identified in H1 to H13; b) differences in terms of the strength, valence, or significance of the proposed relationships; and finally c) the predicted variance in the behavioural intentions (to remain or become a customer).

Since the development of Lasswell’s (1948) verbal communication model (which identifies five elements of mass communication: Who, Says What, In Which Channel, To Whom, With What Effect) and Shannon and Weaver’s (1949) general communication system (which focuses on signal transmission and distinguishes between the information source and the transmitter, the receiver and the destination, the signal and the message, as well as the potential effects of noise that might affect the transmission) mass communication models have underlined the importance of identifying the intended target audience of a message and employing marketing tactics to reach them (Severin and Tankard 2001). Nowadays, with the development of social media, ensuring that a message reaches the pre-determined target audience may be more difficult, as content may fall outside boundaries, and reach unintended audiences, with unknown effects. Therefore, it is expected that non-affected customers and non-customers of the service/brand may also be exposed to the social media service failure apology, especially given the nature of social media. Researchers have focused on understanding social media message effects on unintended audiences especially for members of the younger generation (see Odundo 2012 for a review of the literature), but not within the context of social media service failure apologies.

In addition, as stated in the introduction, the general public’s (i.e. non-customers) perception of the brand experiencing the service failure and how the public may respond to the apology message is vital to be investigated for commerce activities and for the overall reputation of the organisation (Shamma and Hassan 2009). Service failures may also affect the perceptions of the service/brand of potential customers (non-customers or the general public) who become aware of the service failure incident and, in turn, decrease their behavioural intentions to become customers of this service/brand in the future, thus not allowing the company to increase its market share or allowing the company to become the victim of competition.

This research is the first study that has examined the aforementioned issues for customers and non-customers in the context of a social media service failure apology, by putting forth the following hypothesis, based on the rationale provided:

H14: Differences between customers and non-customers are expected for: a) the constructs included in the aforementioned hypotheses; b) the strength, valence, or significance of the relationships hypothesised (H1-H13); and c) the predicted variance explained in behavioural intentions of the hypothesised model.

Specifically, in relation to H14(a), customers and non-customers (see Bravo et al., 2009) have different levels of familiarity with a service/brand (in this study this is termed subjective knowledge, therefore it is expected that customers and non-customers will vary in terms of their level of subjective knowledge of the service/brand), which may lead to different knowledge structures and in turn perceptions and behaviours. A certain type of knowledge is acquired by experience (Raju, et al. 1995), therefore it is expected that customers will be more likely to be familiar with the service failure incident, have been exposed to the social media service failure apology, and have greater subjective knowledge of the service/brand than non-customers. Given that non-customers are not directly affected by the incident, and have lower levels of subjective knowledge of the service/brand, they may not care to criticise and judge the apology message or may be harsher critics of the apology than non-customers. As a result they may perceive the apology as less persuasive than customers, as the service failure incident may be proof that they did well to choose a different service provider and thus continue to be critics of the service provider and the associated apology, and view the brand as a second or lower rated provider than the one they have. This negative sentiment, which may result in lower persuasiveness of the apology, might lead to less favourable attitudes towards the service provider, less satisfaction with the service provider, or they might perceive the service provider as being less trustworthy, and overall have fewer behavioural intentions to become a customer compared to customers who wish to remain customers.

In relation to H14(b), we need to note that differences may arise in the mean scores of the antecedents and attitudinal and behavioural outcomes measured in this paper; the latter implies that the relationships hypothesised between the constructs may also fluctuate, not necessarily in terms of their significance and valence but most probably their strength. For example, customers are more highly involved with the company than non-customers, which may explain why persuasion of the apology may have stronger relationships with perceptions and attitudinal outcomes (i.e., satisfaction with, trustworthiness of, and attitudes towards the service/brand) for customers than for non-customers. Differences in the relationships for customers versus non-customers have not been previously investigated and, therefore, specific moderating effects of customership (customers versus non-customers) are not advanced but still explored in this paper.

Lastly, in relation to H14c, the hypothesised model proposed (H1 to H13) has been advanced, based on the service failure apology literature, to predict customers’ behavioural intentions. Therefore, it is expected that the hypothesised model will predict intentions to remain a customer better than intentions to become a customer after exposure to a social media service failure apology.

In summary, Figure 1 illustrates our conceptual model and the hypotheses advanced (i.e. H1-H13 and H14b; H14a and H14c are not depicted in the figure). The first part of the model recognises the various consumer characteristics, which may affect perceptual, attitudinal and behavioural outcomes after exposure to a social media service failure apology. The second part of the model includes constructs that have been recognised as perceptual and attitudinal outcomes of becoming exposed to a social media service failure apology. This part assumes that the consumers will have already been exposed to the apology message at least once (given that all participants in the study will be exposed to the social media service failure incident apology as part of the methodology used). It is important to note that these constructs (perceptual and attitudinal outcomes) are not essential to forming behavioural intentions to remain or become a customer (Manika and Gregory-Smith 2014). Their inclusion is based on constructs investigated in prior apology and social media studies, as discussed earlier. The following section will discuss the methodology used to test the aforementioned hypotheses and the case selected to do so.

Figure 1. Factors affecting behavioural intentions after exposure to a social media service failure apology (for customers versus non-customers).

3. Methodology

3.1. Research Context

In order to examine the outcomes of a social media apology for a service failure incident, for customers and non-customers, and explore their differences, the October 2011 service outage incident of BlackBerry and its CEO’s response to that incident was selected as the case to fulfil the objectives of this study. BlackBerry, as a service and mobile phone service provider, experienced a service outage incident, which lasted about 3 days (10-12 October 2011) and affected its customers world-wide. Other studies in the information management arena have also used the mobile technologies and services industry to explore constructs of customer satisfaction and loyalty, such as the recent study of Deng et al. (2010). BlackBerry specifically used a YouTube video, as a social media communication vehicle, to apologise to affected customers for the service failure incident and reassure them that they would do everything in their power to resolve the problem. The YouTube video apology featured BlackBerry’s CEO Mike Lazaridis, as the representative of the company (www.youtube.com/watch?v=zQ1esvGae_s). The choice of communication vehicle (YouTube) made it possible for non-customers and non-affected customers of BlackBerry also to access and view the apology. The video is still available online while the number of views keeps increasing, long after the service failure incident was resolved, illustrating the need to explore how social media service failure apologies may impact on the behavioural intentions of customers and non-customers exposed to the apology. This particular case was selected as it was one of the most recent examples of social media service failure apologies, with real-world implications and insights. Field experiments in natural settings are a better alternative to regular laboratory experiments, which have been criticised for their lack of realism, artificiality and lack of generalisability (Jimenez-Buedo and Miller 2010; Schram 2005). The case was considered an ideal one for this work as it both involved a major service failure, a world famous company that would have been also widely recognised by non-customers, and one that operates within a sector that experiences high churn.

3.2. Sampling

This study’s sample frame was specified to include both BlackBerry users (those who were BlackBerry customers when the service outage incident occurred), and non-Blackberry users, who represented the non-customers. The sample was demographically dispersed and the data collection took place in June 2012, via an Internet survey administered to an online USA consumer panel, after being pre-tested. A total of 512 participants completed the survey, out of which 241 (47.1%) were BlackBerry customers and 271 (52.9%) non-BlackBerry customers (see Table 1 for the relevant demographic characteristics for both samples). The total sample had an equal split between males and females. Of the total sample, more than half of the participants were over the age of 40 (53.1%). The majority of participants were White Americans (47.9%), with 11.1% identifying as African American, 8.2% identifying as Hispanic American, and 4.9% identifying as Asian American. 31.1% of the respondents were college graduates, followed by 21.5% who had some college but no degree. About half of the participants (44.5% or 228 out of 512) were aware of the October 2011 incident, and 12.3% (or 63 out of 512) had been exposed to BlackBerry’s apology prior to this study. A few participants from both groups (customers and non-customers) were familiar with the incident and had watched the apology on YouTube prior to this study, which confirms the relevance of an apology on social media for both customers’ and non-customers’ attitudes and behaviours.

*** PLEASE INSERT TABLE 1 ***

3.3. Survey Measurements

Established scales from prior literature were used to measure the constructs identified in Figure 1. The data collection procedure included three main stages. The first stage measured participants’ subjective knowledge of Blackberry’s products and services based on Gurhan-Canli’s (2003) scale. Participants were also asked to rate their familiarity with the service failure incident based on the Moore et al. (2005) scale and whether or not they had watched BlackBerry’s YouTube service failure apology for the October 2011 incident prior to this study. Incident familiarity was also measured as a dichotomous yes or no question to gauge consumers’ awareness of the incident (see results section for more information about how this was used). Consumers who were unsure of whether or not they knew about the October 2011 incident, and/or whether they had watched the YouTube apology, were deleted from the data set, to avoid self-reported biases (n=50).

During the second stage of the data collection process we asked participants to watch the YouTube apology, which was embedded in the survey. The video lasted 1 minute and 39 seconds. The third and final stage of the data collection process gauged participants’ perceptions of the persuasiveness of the social media apology, attitudes towards and trustworthiness of the service provider after the apology, and resulting behavioural intentions to remain or become a customer. Persuasiveness of the social media apology was measured via Pham and Avnet’s (2004) scale; attitudes towards the service provider after the apology based on Tybout et al.'s (2005) scale of attitude towards high-tech products; the perceived trustworthiness of the service provider after exposure to the apology, based on Erdem and Swait’s (2004) scale; the perceived satisfaction with the service provider after exposure to the apology based on an adapted version of Karatepe and Ekiz’s (2004) scale of complaint satisfaction; and lastly, the behavioural intentions to remain or become a BlackBerry customer, based on an adapted version of Jones, Mothersbaugh, and Beatty’s (2000) scale.

Lastly, a series of demographic questions related to age, gender, ethnicity, household income, employment status, and education followed. The aforementioned scales used in this study can be seen in Table 2. The two items of the incident familiarity scale were reverse coded (depicted in Table 2).

*** PLEASE INSERT TABLE 2 ***

3.4. Methodology and Analysis

Based on the aims of this study the methodology requires both a confirmatory and an exploratory approach. This research takes the form of an embedded experimental design within a survey and is aligned with a deductive and quantitative approach, which is often associated with confirming hypothesized relationships, such as H1 to H13. Hypotheses 1 to 13 were based on prior literature and will be empirically examined in the context of social media to uncover whether or not these relationships hold within the social media context, as with other technology-oriented communication channels examined in prior service failure apology studies, given Kerkhof et al.'s (2011) notion that different communication channels may have different effects on consumers. At the same time, new hypotheses absent from prior service failure apology literature are examined (H7; H10; H11); these have, however, been supported in general social media and consumer behaviour studies. Testing H1 to H13 requires a confirmatory approach, which is why an online survey was used for the data collection, and a structural equation modelling (SEM) approach is necessary for the analysis. As far as exploring H14 was concerned, we decided to use a statistically objective method and large scale data to compare customers versus non-customers directly in an objective, measureable manner, and we therefore continued to explore these hypotheses via a quantitative methodology. We collected data from non-customers of the service/brand via the same online survey used for customers and then: a) used ANOVAs on SPSS to explore differences, if any, between customers versus non-customers in terms of the constructs (antecedents, attitudinal and behavioural outcomes) identified in H1 to H13 (for which composite scores were created on SPSS for each construct by adding the statement scores and dividing by the number of statements); and we used a multigroup SEM technique to explore: b) differences in terms of the strength, valence, or significance of the proposed relationships; and c) the predicted variance in the behavioural intentions (to remain or become a customer). The results of H14 therefore should be seen as exploratory and may not be generalizable; however, they provide a first step for future research studies.

4. Results

4.1. Measurement Model

A confirmatory factor analysis was conducted using Mplus to test the reliability and validity of the hypothesised model constructs [see Table 2 for the results of the CFA for the total sample (N=512)]. All scales had significant factor loadings above or equal to .64 and were highly reliable and valid with Cronbach’s alphas above or equal to .91, construct reliability of above or equal to .89 and average variance extracted (AVE) scores above or equal to .73 (Fornell and Larcker 1981). The measurement model demonstrated a theoretically and statistically good overall fit (χ2=792.92, df=254, p=.00; RMSA 90% CI=.05-.06; CFI=.97; TLI=.96; SRMR=.03). While the chi-square value was significant due to sample size (N=512), the normed chi-square (χ2/df) was equal to 3.12, which is within acceptable values (Schumacker and Lomax 2004). Table 2 also shows the means and standard deviations of these constructs. As further indication of discriminant validity, none of the correlations between constructs reached .85 (Dijkstra, Buist, & Dassen, 1998), and the Fornell-Larcker criterion [AVE> (r)2] indicated that the AVEs for each construct were greater than the square of the correlation estimates. The correlations for the total sample can be seen in Table 3. The authors also confirmed there was adequate variable-to-sample ratio, normality and linearity of the data, and no signs of extreme multicollinearity indicated by the VIF (VIF<5) and tolerance (tolerance > .22) for each construct (Hair et al. 1998).

*** PLEASE INSERT TABLE 3 ***

4.2. Preliminary Analysis

A series of chi-squares and ANOVAs were conducted on SPSS to compare customers’ and non-customers’ demographic and consumer characteristics, before examining the hypotheses advanced. In terms of the participants’ demographic characteristics, both samples, customers and non-customers, had an equal split between males and females, which was a pre-requisite of the data collection process, which is why the two groups did not differ significantly in terms of gender (χ2(1)=.38, p>.05) and age (F(1,510)=.06, p>.05). However, there were significant differences between the two groups in terms of the level of education (F(1,510) =22.80, p<.01). Customers were significantly more educated than non-customers (M=4.99, SD=1.68 and M=4.28, SD=1.65, respectively), confirming past, relevant marketing studies [see for example, the paper by Allred et al. (2006) highlighting the point that online shoppers are more educated than online non-shoppers]. Education could also be a key factor in determining different perceptions of customers, as was noted in the work by Caruana (2002).

Next, customers and non-customers were compared in terms of their consumer characteristics. As expected, differences related to incident familiarity and prior exposure to the apology were significant (F(1,510)=125.37, p<.01 and χ2(1)=46.67, p<.01, respectively), indicating that customers were more likely to know about the incident, and more likely to have watched the YouTube apology than non-customers (see Table 1 for descriptive statistics for each group).

An ANOVA test was also used to examine whether or not the consumers who had watched the social media service failure apology (which could be customers and non-customers), significantly differed in terms of their level of incident familiarity (the continuous multi-item scale of incident familiarity was used for this analysis). Results indicated that consumers who knew more about the incident were more likely to have watched the apology on YouTube than consumers who were not that familiar with the incident (F(1,510)=115.88, p<.01). Out of the total sample of 512 participants, 57 consumers had both watched the apology and were familiar with the incident (incident familiarity based on the dichotomous yes/no question). From the total sample of 512 consumers 6 out of 63 who had seen the YouTube apology prior to this study were unfamiliar with the service failure incident (based on the dichotomous measure of incident familiarity). In addition, the age groups were compared in terms of their exposure to the social media apology, and findings indicated that significant differences existed (F(1,510)=4.23, p<.05), with younger adults (below the age of 41 years old) being more likely to have been exposed to the social media apology than older adults (41 years old and above).

Differences between the two groups (customers and non-customers), in incident familiarity (multi-item measure), subjective knowledge, persuasiveness of the apology, satisfaction with, trustworthiness of, attitudes towards the service provider after the apology, and behavioural intentions, were examined via a series of ANOVAs. The groups could only be compared in terms of persuasiveness of the apology, given that for all the other constructs the Levene test was significant and the homogeneity of variance assumption was not satisfied. Persuasiveness of the apology differed significantly based on customers versus non-customer groups (F(1,510) =38.09, p<.01). Customers found the apology more persuasive than non-customers (M=4.58, SD=1.23 and M=3.89, SD=1.29 respectively). It should be noted that these differences are related to testing H14a, and they serve as a preliminary analysis for the SEM analysis that follows. The findings are discussed in relation to prior literature in the next section of the paper.

4.3. Structural Equation Model Analysis

First, a structural equation model was computed with Mplus, to examine the hypothesised model depicted in Figure 1, based on the total sample (N=512), without the moderating variable (whether or not consumers were BlackBerry customers or not) relevant to H14. The model examining H1 to H13 (N=512) had statistically acceptable model fit (χ2=1152.57, df=287, p=.00; CFI=.95; TLI=.95), accounting for 66.7% of the variance in behavioural intentions to remain or become a customer. The model constructs predicted 59.4% of the variance in satisfaction, 48.0% of the variance in trustworthiness, 46.8% of the variance in attitudes, 18.5% of the variance in persuasiveness of the apology, and lastly 4.11% of the variance in incident familiarity.

Table 4 summarises the results of the hypotheses (H1-H3) tested. H5 (Persuasiveness of Apology Attitudes towards the Service Provider After Apology), H6 (Satisfaction with Service Provider After Apology Attitudes towards the Service Provider After Apology), H9 (Prior Exposure to the Apology Persuasiveness of Apology) were not supported, as relationships were not significant, while H12 (Incident Familiarity Prior Exposure to the Apology) was also not supported because the relationship was negative instead of positive. These unsupported relationships are discussed in more detail in the next section of the paper as they provide original findings. Therefore, only H1-H4, H7-H11, and H13 were supported as they were positive and significant, as predicted.

*** PLEASE INSERT TABLE 4 ***

4.4. Multigroup SEM Analysis

The last two hypotheses (H14b and H14c) were examined by breaking down the total sample into customers and non-customers, and using a multigroup SEM analysis (check section 4.2. for testing H14a). Based on the literature review, H1 to H13 were moderated by the grouping variable, namely whether or not participants are customers of BlackBerry or not, as depicted in Figure 1. The overall sample had a balance between customers (n=241) and non-customers (n=271). The multi-group SEM analysis run with Mplus indicated an acceptable model fit for both groups (Overall: χ2=1681.04, df=610, p=.00; Customer Group: χ2=778.35; Non-customer Group: χ2=902.69; CFI=.94; TLI=.93). It is evident from the smaller chi-square value of the customer group that the modelled hypotheses (H1-H13) fit the data better for the customers than the non-customer group.  In addition, the customer group explained a slightly higher variance in behavioural intentions (R2=.770) than the non-customer group (R2=.517), which supports H14c. However, based on a chi-square comparison between the first unmoderated model and the multi-group SEM model with customers versus non-customers as the grouping variable the difference was significant (Δ2(610-287)=1681.04-1152.57=528.46, p<.01), and the unmoderated model fits the data better than the moderated one. This implies that H14b, which predicts that H1 to H13 were moderated by whether or not consumers exposed to the social media service failure apology were customers of the service provider experiencing the service failure incident or not, is not supported. Thus, this result shows that the hypothesised model is relevant for both customers and non-customers, and even though it predicts customers’ behavioural intentions better than non-customers’ ones (based on the chi-square values of the multi-group SEM analysis), the hypothesised relationships H1 to H13 were not significantly different between customer and non-customer samples. Thus, H14b was not supported. These results are discussed in further detail in the next section of the paper, along with managerial implications.

5. Discussion and managerial implications

This study sheds light on the attitudinal and behavioural outcomes of exposure to a social media service failure apology and examines a hypothesised model among two samples: customers versus non-customers of the service provider in relation to the service failure incident. The following sections discuss the results and their managerial implications, first for the direct relationships examined, and then for the customers versus non-customers.

Direct Relationships Examined (H1 to H13)

This study contributes to general service failure literature by investigating subjective knowledge of the service/brand, trustworthiness and attitudes towards the service/brand after exposure to an apology and their impacts on behavioural intentions to remain a customer after a service failure incident and apology; it also provides some interesting results contradicting prior literature, which may be due to the social media context thus adding original value.

Firstly, what consumers think they know (subjective knowledge) is important for behavioural intentions, for both customers and non-customers, supporting prior literature (Alba and Hutchinson 2000; Brucks 1985). Managers should take into account what consumers think they know when designing apology messages to maintain high levels of subjective knowledge. The more the apologies increase consumers’ perceptions of what they know (subjective knowledge), the more likely they are to remain or become customers (See Table 4, b=.35, p<.01), as customers feel confident they know a lot about the service/brand and do not want to risk going with an alternative service/brand which they may not know much about; while for non-customers high perceptions of knowledge would make them feel they want to know more (knowledge begets knowledge – Golden and Stanaland, 2000) or feel that they are capable of making the right choice of service provider/brand. Attitudes towards the service/brand and trustworthiness of the service/brand had positive and significant relationships with behavioural intentions, for both customer and non-customer samples, indicating that these are important predictors of behavioural intentions to become or remain a customer. In terms of the attitudes construct, this finding is consistent with prior advertising literature and the Theory of Planned Behaviour, which states that attitudes toward a brand may impact on behaviour (Tellis and Ambler 2008). The literature about apologies also supports the claim that satisfaction following an apology is a significant predictor of intentions to remain brand loyal (Amine 1998; Pace, et al. 2010), as supported by the results of this study. However, it failed to examine other attitudinal dimensions, as our study did. This provides a new finding, which implies that satisfaction may not be the only predictor of behavioural intentions to remain a customer, but other attitudinal dimensions should be taken into account, such as the perceived reliability and quality level of the service/brand, how valuable the service/brand is perceived to be and how desirable it is after exposure to the apology message (as with the measurement of attitudes by Tybout et al., 2005). Trustworthiness is also a key cornerstone of e-commerce (Hajli et al., 2014a) and should be investigated within the context of social media. Findings show that of the three constructs of attitudes, trustworthiness and satisfaction, trustworthiness has the largest effect on behavioural intentions to remain or become a customer (β=.32 compared to β=.24 for attitudes and β=.24 for satisfaction), and it is thus an important construct determining the behavioural effect of a social media service failure apology.

This study was also the first to investigate these relationships within the context of social media service failure apologies for a non-customer sample and it found that all hypothesised relationships pertaining to these constructs hold for both samples (customers and non-customers). Managers should aim to increase positive attitudes towards, trustworthiness of, and satisfaction with the service/brand after the social media service failure apology, in order to maintain its customer base, as well as increase the likelihood of acquiring new ones. For instance, social media apologies could offer viewers incentives coupled to future service quality that act as a satisfaction guarantee, in order to encourage them to remain with or to join the company. Specifically, managers should reward customers who remained loyal to the company following the failure incident and the actual apology by providing extra loyalty points, discounts etc. In terms of the importance of trustworthiness after the social media apology, managers should aim to use highly trustworthy and expert spokespersons to communicate the apology message, such as CEOs as in BlackBerry’s case (Manika et al., 2015). Trustworthiness may be the most important predictor of behavioural intentions in a social media context, more than satisfaction, as the general service failure literature has suggested so far.

Secondly, as noted earlier, our findings produced some interesting results, which contradict prior literature and add original value to the paper (related to H5, H6, H9, and H12), some of which could be explained due to the new constructs investigated in this study, while others are due to the nature of social media. Persuasiveness of the apology did not translate to positive attitudes towards the service provider for the total sample. This may be due to the fact that the attitudinal construct measured in this study did not include the satisfaction dimension with the service/brand after the apology, but focused more on general attitudes towards the service/provider. This suggests that managers should not only focus on the satisfaction of their customers but also ensure that their service/brand remains reliable, of high quality, desirable and valuable compared to competitor brands. This may be achieved by apology messages that do not neglect to highlight the benefits of the service, aside from addressing the service failure incident, to increase the general public’s views of / attitudes towards the brand. According to consumer behaviour literature, consumers tend to bestow more attention on the beginning and ending of a message (Hoyer et al., 2013) and therefore communicators may decide to highlight how they address the service failure incident at the beginning of the apology, finishing the message in a positive tone by highlighting the benefits of being a customer. In addition, satisfaction with the service after the apology did not significantly relate to attitudes towards the service provider after the apology, even though in prior consumer behaviour literature they may be complementary attitudinal dimensions. Therefore, satisfaction with the service provider after the apology only directly influences behavioural intentions for both customers and non-customers, but does not correlate with other attitudinal dimensions. This adds further support for the recommendation to managers above regarding the necessity of highlighting the benefits of a service/brand within the apology message.

Prior exposure to the apology did not significantly affect the persuasiveness of the apology, contrary to Berlyne’s (1970) two-factor theory, which states that repetition of the apology message can reduce negative outcomes (e.g., lack of persuasiveness of the apology). This result contradicting prior literature may be due to the nature of social media. Given that social media requires users to pull information it is likely that consumers who have previously seen the apology message did not pay much attention to the message as probed by this study, as they may have perceived that they knew all there was to know about the incident, therefore the persuasiveness of the message was not affected by this study (a limitation of the study – see Conclusion section). The above, in combination with the results on trustworthiness of the service/brand after the social media service failure apology, also indicates the key importance of trust, which is developed over time between customers and service providers and it also stresses the influential role that an apology plays for preserving trust following a service failure. Managers need to develop appropriate strategies aimed at maintaining and enhancing trust with their customers as with earlier recommendations (i.e. using a trustworthy spokesperson and highlighting the benefits of the service/brand along with the apology).

A result which might also seem counter-intuitive at first is that incident familiarity (as measured by the multi-items latent construct) affected prior exposure to the apology negatively (b=-.44, p<.01), indicating that the more familiar consumers thought they were with the service failure incident the less likely they were to have been exposed to the social media apology prior to this study. This could be explained by the nature of social media. Prior consumer knowledge literature states that the more consumers think they know about an object or issue the less the likelihood of searching externally for more information (Raju, et al. 1995). Exposure to a social media message, given the nature of “pulling” information, may be seen as an information-seeking activity. However, it should be noted that based on the chi-square results on SPSS with the dichotomous measure of incident familiarity, results indicated that customers who were aware of the service failure incident were more likely to have been exposed to the social media apology. Thus, these different results may be due to the measures employed.

Customers versus Non-Customers

Next, as indicated by the results, non-customers, as well as customers, may be exposed to apologies for service failures on social media, which is why managers need to carefully consider the implications of posting service failure apologies on social media as they may influence potential customers and affect the perceptions of the general public, those who were not affected by the service failure incident. More specifically, as anticipated, customers were more likely to be exposed to the social media apology than non-customers, and consumers who were familiar with the service failure incident were more likely to have been exposed to the social media apology prior to this study. In total, out of the 228 consumers who were familiar with the apology, only 57 of them had watched the apology, indicating a greater need and effort from managers to direct consumers familiar with and probably affected by the incident to the social media apology. A few consumers reported having watched the YouTube apology without being familiar with the service failure incident (6 out of 63 consumers who had watched the apology prior to this study).

Ensuring that the social media service failure apology is first targeted at affected customers, but at the same time realising that unintended audiences may become exposed to the apology on social media, is important. This is often ignored by managers, who focus their attention on retaining their loyal customer base via the apology message, as per the definition of a service failure apology and its purpose (Boshoff and Leong 1998). Potential customers and the general public should also be of interest to managers and should be nurtured for the benefit of a company’s reputation (Shamma and Hassan, 2009). Managers need to keep track of apologies on social media, and keep records of views as well as social media user characteristics, in order to have a holistic picture of social media users exposed to the social media service failure apology. Better targeting and control of audiences exposed to the social media service failure apology is necessary, in order for managers to maximise the effect of the apology on existing customers and minimise and potential negative impact on potential customers. For instance, video apologies could be used to target customers when they log in to their protected account areas or by inviting them by email to watch the apology at private locations. At the same time highlighting benefits of the service/brand within the apology message, as recommended earlier, due to the importance of trustworthiness for behavioural intentions may also contribute to nurturing the relationship of customers, non-customers, and the general public with the company/service/brand and avoid negative sentiment, which may lead to advocacy groups and expenses on addressing a company’s reputation as per Shamma and Hassan (2009).

Consistent with social media literature, which states that younger adults tend to use social media more than older adults (Correa, Willard, and Gil de Zuniga 2010), this study found that younger adults (below the age of 41) were also more likely to have been exposed to the social media apology than older adults (41 years old and above). This finding may imply that older adults may not be so likely to be exposed to a service failure apology message on social media, and therefore, a managerial implication is that social media may be a better communication vehicle for the apology for younger adults than older adults and that the service/brand may need different communication vehicles for the apology messages for different audiences (i.e. audiences segmented by age in this case). Managers are advised to employ appropriate segmentation strategies as customers react differently depending on their demographic (and other) characteristics, and select communication vehicles that will have a greater positive effect for these audiences. This provides further evidence of the value of our study and the need to investigate the effects of different technological communication vehicles for service failure incident apologies and for different audiences, as per Kerkhof et al. (2011). Technology-oriented communication channels such as mobile communications, blogs and emails may be more appropriate for older consumers than social media.

Non-customers had less favourable perceptions of the persuasiveness of the apology than customers, as expected. Even though this is the first study to investigate perceptions of the persuasiveness of a service failure apology for non-customers, prior consumer behaviour literature has found that a certain type of knowledge is acquired by experience (Raju, et al. 1995), therefore customers who are more familiar with the service failure incident, as well as the service provider/brand itself may use different knowledge structures in judging the apology (Bravo et al., 2009) than non-customers, who may focus on e.g. comparing facts – such as one service that had a failure incident compared to one that has not had one. Consumers less involved with the apology (i.e. non-customers) may not care to criticise and judge the apology message or may be harsher critics of the apology than non-customers. As a result they may perceive the apology as less persuasive than customers, as the service failure incident may be proof that they did well to choose a different service provider than the one experiencing the failure incident and thus continue to be critics of the service provider and the associated apology, and view the brand as a second or lower rated provider than the one they have.

Service failure apology literature stresses the point that perceptions of the apology (i.e. persuasion of the apology as measured in this study) can affect satisfaction and consequently brand loyalty, as satisfaction fully mediates an apology and brand loyalty (Amine 1998; Pace, et al. 2010; Wirtz and Mattila 2004). This was supported by our results for the total sample of both customers and non-customers (Table 4), which suggests that the persuasiveness of the apology has a high and significant positive relationship with satisfaction with the service provider after the social media service failure apology (b=.77, p<.01) and in turn satisfaction has a smaller size but positive and significant effect on behavioural intentions to become or remain a customer (b=.24, p<.01). The multi-group SEM model confirmed that no differences in terms of these relationships were found between customer and non-customer samples. This is why managers need to ensure that the apology message is persuasive to ensure its success in maintaining customers’ brand loyalty, as well as increasing the behavioural intentions to become a customer for non-customers who have been exposed to the social media service failure apology. For instance, Kellerman [33] argues that a good apology will include an acknowledgment of the mistake, an acceptance of responsibility, an expression of regret, and finally an assurance that the offence will not be repeated. Using, for example, an endorser such as BlackBerry’s CEO to apologise for the incident may also prove effective in comparison with using other types of endorsers, who may be perceived as less trustworthy, credible or having less expertise in the subject matter (Manika et al. 2015).

Interestingly, even though this study did not hypothesise a difference between customers and non-customers in terms of their education level (as this was beyond the scope of the study), this was tested as a preliminary analysis for our sample. We found that consistent with statistics and other studies that have looked at Blackberry as a case company compared to other mobile phone service providers, customers were significantly more educated than non-customers, which may be related to the fact that Blackberry users are normally professionals (see for example, Costa 2011; Funtasz 2012). Thus, we can infer that there may be differences in terms of perceptions, attitudinal and behavioural outcomes to a social media service failure apology, between consumers (customers or non-customers) of different education levels and professional standing. Professionals who see their mobile phones as a key device in order to accomplish their work duties may be less willing to tolerate a service failure and could be more inclined to switch to another mobile phone brand.

Lastly, given that social media service failure apologies have not been investigated before for non-customers, these findings based on the multi-group SEM analysis highlight that the hypothesised model is not moderated by whether or not social media users are customers or not of the service provider. Both customers’ and non-customers’ behavioural intentions will be affected by the social media service failure apology and the antecedents identified in the model (Figure 1). Managers responsible for the design of service failure apologies should develop clear and consistent messages directed to current and potential customers, as both need to be nurtured for the benefit of a company’s reputation as per Shamma and Hassan (2009).

6. Conclusion and future research directions

This study makes valuable contributions to the academic literature. Specifically, this study contributes to the general service failure apology literature by examining the effects of subjective knowledge about the service/brand before the apology. It also sheds further light on the trustworthiness of the service/brand and attitudes towards the service/brand (independent of satisfaction - after the apology) although this is confirmed in prior general consumer behaviour literature (on behavioural intentions to remain a customer after a service failure incident and exposure to an apology message). Secondly, this study contributes the e-commerce and social media literature in relation to service failure apologies, given the nascent literature on social media service failure apologies. It cannot be assumed that the effects of an apology via any technological communication vehicle will be the same on consumers as per Kerkhof et al. (2011); this was confirmed by our results. Thus, this study fills an important gap in the research given the popularity of social media as a communication vehicle for service failure apologies. At the same time it has investigated key e-commerce variables such as trust. Lastly, given the nature of social media and the potential of the service failure social media apology to be exposed to non-affected customers, non-customers and the general public, our study explored social media apologies and their effects for both customers and non-customers. It also provides the first step towards building a theoretical framework of how the general public, which includes customers and non-customers of the service/brand exposed to the social media service failure apology, varies in terms of their perceptions, attitudinal and behavioural outcomes (H14a), how the constructs’ relationships may vary (H14b), and how the variance predicted in behavioural intentions by the antecedents examined may vary for customers versus non-customers (H14c).

Apart from the original contributions of this study to academia and also the respective practical implications of the findings, this study suffers from some limitations, which should be noted. Although YouTube is not like every other social medium (social media vary in richness and self-disclosure), YouTube videos can also be embedded in other social media, therefore this case presents a valid example for testing the effects of a service failure apology on social media. It should be noted, however, that the results may not be generalizable to non-technology-oriented services or to social media that do not allow the embedding of videos in them. Differences may also arise due to the unique characteristics of service providers/brands and thus, these findings cannot be generalised outside Blackberry’s context. The fact that our apology case (i.e. Blackberry) is from 2011 may also impose time effects on our results. It should also be noted that even though the testing of H14 may have required a more qualitative approach, our methodology gave us the ability to statistically compare customers and non-customers in terms of the constructs examined (H14a), the relationships among those constructs (H14b), and the variance explained by the antecedents in behavioural intentions to remain or become customers (H14c).

This work and the limitations noted above have also illustrated many avenues for future research. For example, future studies of apologies should aim to compare the use of different endorsers in terms of their effectiveness and the results of this study need to be validated with different case studies and various examples of social media apologies to ensure generalizability. Future research could analyse various customer segments based on consumer characteristics (e.g. different education levels and professional standing) and identify whether a more targeted approach is required to maximise the effectiveness of a social media service failure message. Different social media based on their level of richness should also be examined as they might have different effects. A qualitative methodology will also be necessary in understanding the reasons behind these results. Finally, future longitudinal research will be invaluable for highlighting whether there is a lasting, detrimental impact on customers and non-customers following this service failure incident (and similar ones) and whether a subsequent apology was effective or not.

References

Alba, J. W., and Hutchinson, J. W. (2000). Knowledge Calibration: What Consumers Know and What They Think They Know. Journal of Consumer Research, 27(2), 123-156.

Allred, C.R, Smith, S.M., and Swinyard, W.R. (2006) E‐shopping Lovers and Fearful Conservatives: A Market Segmentation Analysis. International Journal of Retail & Distribution Management, 34 (4/5), 308 - 333

Amine, A. (1998). Consumers' true brand loyalty: the central role of commitment. Journal of Strategic Marketing, 6(4), 305-319. doi: 10.1080/096525498346577

Berlyne, D. E. (1970). Novelty, Complexity, and Hedonic Value. Perception and Psychophysics, 8, 279–286.

Boshoff, C., and Leong, J. (1998). Empowerment, attribution and apologising as dimensions of service recovery: An experimental study. International Journal of Service Industry Management, 9(1), 24-47.

Bravo, R., Montaner, T. and Pina, J.M. (2009). The role of bank image for customers versus noncustomers, International Journal of Bank Marketing, Vol. 27 Iss 4 pp. 315 – 334.

Brinol, P., Rucker, D.D. and Petty, R.E. (2015). Naıve theories about persuasion: implications for information processing and consumer attitude change. International Journal of Advertising, Vol. 34, No. 1, 85_106, http://dx.doi.org/10.1080/02650487.2014.997080

Brucks, M. (1985). The Effects of Product Class Knowledge on Information Search Behavior. Journal of Consumer Research, 12, 1-16.

Campbell, M. C., and Keller, K. L. (2003). Brand Familiarity and Advertising Repetition Effects. Journal of Consumer Research, 30(September ), 292-304.

Caruana, A. (2002). Service Loyalty: The Effects of Service Quality and the Mediating Role of Customer Satisfaction. European Journal of Marketing, 36 (7/8), 811-828.

Carlson, J. P., Vincent, L., Hardesty, D. M., and Bearden, O. (2009). Objective and Subjective knowledge Relationships: A Quantitative Analysis of Consumer Research Findings. Journal of Consumer Research, 35(5), 864-876.

Chevalier, J., and Mayzlin, D. (2006). The Effect of Word of Mouth on Sales: Online Book Reviews. Journal Of Marketing Research, 43(3), 345-354.

Chuang, S.C., Cheng, Y.H., Chang, C.J. and Yang, S. W. (2012) The Effect of service failure types and service recovery on customer satisfaction: a mental accounting perspective” The Service Industries Journal, 32 (2): 257-271.

Chung, C., and Austria, K. (2010). Social Media Gratification and Attitude toward Social Media Marketing Messages: A Study of the Effect of Social Media Marketing Messages on Online Shopping Value. Proceedings Of The Northeast Business & Economics Association, 581-586.

Correa, T., Willard, A., and Gil de Zuniga, H. (2010). Who interacts on the Web? The intersection of users’ personality and social media use. Computers in Human Behavior, 26, 247–253.

Corritore, C. L., Kracher, B., and Wiedenbeck, S. (2003). On-line trust: Concepts, evolving themes, a model. International Journal of Human Computer Studies, 58(6), 737-758.

Costa, M. (2011). BlackBerry’s accidental youth phenomenon could become a strategic web changer. Retrieved 29th of August, 2014, from http://www.marketingweek.co.uk/blackberrys-accidental-youth-phenomenon-could-become-a-strategic-web-changer/3022716.article

Curty, G. R. and Zhang, P. (2011). Social Commerce: Looking Back and Forward. Proceedings of the American Society for Information Science and Technology, 48 (1), 1-10. DOI: 10.1002/meet.2011.14504801096

Dalziel, N. (2014) Customer Complaints and Service Recovery on Social Media: An Investigation into Barclays Bank Facebook Page (p.111-119). In ePub – European Conference on Social Media: ECSM by Sue Greener and Asher Rospigliosi.

Deng, Z., Lu, Y., Wei, K.K., and Zhang, J. (2010). Understanding customer satisfaction and loyalty: An empirical study of mobile instant messages in China. Internation journal of Information Management, 30, 289-300.

Dijkstra, A., Buist, G., and Dassen, T. (1998). A criterion-related validity study of the nursing care dependency. International Journal of Nursing Studies, 35, 163-170.

Elliott, W. B., Hodge, F. D., and Sedor, L. M. (2009). Using Online Video to Announce a Restatement: Influences on Investor Trust and Investment Decisions. doi: http://dx.doi.org/10.2139/ssrn.1466290

Erdem, T. and Swait, J. (1998). Brand Equity as a Signaling Phenomenon. Journal of Consuemr Psychology, 7, 131-157.

Erdem, T., and Swait, J. (2004). Brand Credibility, Brand Consideration, and Choice. Journal of Consumer Research, 31(1), 191-198.

Fornell, C., and Larcker, F. D. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18 (1), 39-50.

Funtasz, J. (2012). Canadian middle manager experience with mobile email technologies: A phenomenological exploration of frequent BlackBerry users. Information, Communication & Society, 15(8), 1217-1235.

Gefen, D., Karahanna, E., and Straub, D. W. (2003). Trust and TAM in online shopping: An integrated model. MIS Quarterly, 27(1), 51-90.

Golden, L.L., and Stanaland, A.J.S. (2000). Consumer Receptivity to Social Marketing Information: The Role of Self-Rated Knowledge and Knowledge Accuracy, ANZMAC 2000 http://smib.vuw.ac.nz:8081/Gardasil/ANZMAC2000/CDsite/papers/g/Golden1.PDF

Goldsmith, R.E. , Lafferty, B.A., and Newell, S. (2001). The Impact of corporate credibility and Celebrity Credibility on consumer Reaction to Advertismenets and Brands. Journal of Advertising, 29(3), 30-54.

Gregory-Smith, D, Wells, V.K, Manika, D and Graham, S (2015). An Environmental Social Marketing Intervention among Employees: Assessing Attitude and Behaviour Change, Journal of Marketing Management Vol 31 (3-4), pp 336-377.

Grime, I., Diamantopoulos, A. and Smith, G. (2002). Consumer evaluations of extensions and their effects on the core brand, European Journal of Marketing, Vol. 36 Nos 11/12, pp. 1415-38.

Hair, J. F. Jr., Anderson, R. E., Tatham, R. L., and Black W. C. (1998). Multivariate Data Analysis. 5th ed., Upper Saddle River, NJ: Prentice Hall.

Hajli, M. (2013). A research framework for social commerce adoption. Information Management and Computer Security, 21(3), 144 – 154.

Hajli, N. (2014a). A study of the impact of social media on consumers. International Journal of Market Research, 56(3), 387-404.

Hajli, N. (2014b). The role of social support on relationship quality and social commerce. Technological Forecasting & Social Change 87, 17–27.

Hajli, N. and Lin, X. (2014). Exploring the Security of Information Sharing on Social Networking Sites: The Role of Perceived Control of Information. Journal of Business Ethics, 1-13.  

Hajli, N., Sims, J., Featherman, M., Love, P. E. D. (2014a). Credibility of Information in Online Communities. Journal of Strategic Marketing, 1-16.

Hajli, M. N., Lin, X., Featherman, M. and Wang, Y. (2014b). Social Word Of Mouth: How Trust Develops in the Market. International Journal of Market Research, 56(5).

Han, B.O. and Windsor, J. (2011). Users’ willingness to pay on social network sites. Journal of Computer Information Systems, 51, 4, pp. 31-40.

Hoyer, MacInnis, and Pieters (2013), Consumer Behavior (International Student Edition, 6th Edition), published by South-Western/ Cengage learning.

Jimenez-Buedo, M., and Miller, L. M. (2010). Why a trade-off? The relationship between the external and internal validity of experiments. Theoria. Revista de Teoría, Historia y Fundamentos de la Ciencia, 25(3), 301-321.

Jones, M. A., Mothersbaugh, D. L., and Beatty, S. E. (2000). Switching barriers and repurchase intentions in services. Journal of Retailing, 76(2), 259-274.

Kaplan A. M. and Haenlein, M. (2010). Users of the world unite! The challenges and opportunities of social media. Business Horizons, 53 (1), 59-68.

Karatepe, O. M., and Ekiz, E. H. (2004). The effects of organizational responses to complaints on satisfaction and loyalty: a study of hotel guests in Northern Cyprus. Managing Service Quality, 14(6), 476-486.

Kellerman, B. (2006). When should a leader apologize and when not? Harvard Business Review 84(4), 73-81+148.

Kerkhof, P., Schultz, F., and Utz, S. (2011). How to Choose the Right Weapon. Communication Director 02/2011, 76-79.

Kim, A.J. and Ko, E. (2012). Do social media marketing activities enhance customer equity? An empirical study of luxury fashion brand. Journal of Business Research, 65, 1480–1486.

Kuo, Y-F., and Wu, C-M. (2012). Satisfaction and post-purchase intentions with service recovery of online shopping websites: Perspectives on perceived justice and emotions. International Journal of Information Management, 32, 127-138.

Lasswell, H.D. (1948). The structure and function of communication in society. In L. Bryson, ed, The Communication of ideas. New York: Harper & Bros.

Lopez-Nicolas, C. and Molina-Castillo, F.J. (2008). Customer knowledge management and e-commerce: The role of customer perceived risk. International journal of Information Management, 28, 102-113.

Low, G.S. and Lamb, C.W. (2000). The measurement and dimensionality of brand associations, Journal of Product & Brand Management, Vol. 9 No. 6, pp. 350-68.

Lu, Y, Zhao, L. S. and Wang, B. (2010). From virtual community members to c2c e-commerce buyers: trust in virtual communities and its effect on consumers’ purchase intention. Electronic Commerce Research & Applications, 9, 4, pp. 346-360.

Manika, D., and Gregory-Smith, D. (2014). Health marketing communications: An integrated conceptual framework of key determinants of health behaviour across the stages-of-change. Journal of Marketing Communications, http://www.tandfonline.com/doi/abs/10.1080/13527266.2014.946436#.Vd3vHLxVhBc

Manika, D., Papagiannidis, S. and Bourlakis, M. (2015). Can A Ceo’s Youtube Apology Following A Service Failure Win Customers’ Hearts? Technological Forecasting and Social Change, 95, p. 87–95.

Mattila, A.S., Cho, W. and Ro, H.C. (2011). The role of self-service technologies in restoring justice. Journal of Business Research, 64: 348-355.

Moore, R. S., Stammerjohan, C.A., and Coulter, R.A. (2005). Banner Advertiser-Web Site Context Congruity and Color Effects on Attention and Attitudes. Journal of Advertising, 34 (2), 71-84.

Moorman, C. D. K., Brinberg, D., and Kidwell, B. (2004). Subjective knowledge, Search Location, and Consumer Choice. Journal of Consumer Research, 31 (December), 673-680.

Nardi, B.A. (2005). Beyond Bandwidth: Dimensions of Connection in Interpersonal

Communication. Computer-Supported Cooperative Work, vol 14, 2005, pp. 91-130.

Ngai, W.T.E., Tao, S.S.C., and Moon, K.K.L. (2015). Social media research: Theories, constructs, and conceptual frameworks. International Journal of Information Management, 35, 33-44.

Odundo, E. N. (2012). The Unintended Consequences of Social Media on Media-raised Generation. Retrieved on 16/03/2015 from: http://opensiuc.lib.siu.edu/cgi/viewcontent.cgi?article=1302&context=gs_rp

Pace, K. M., Fediuk, T. A., and Botero, I. C. (2010). The acceptance of responsibility and expressions of regret in organizational apologies after a transgression. Corporate Communications: An International Journal, 15(4), 410-427.

Raju, P. S., Lonial, S. C., and Mangold, G. (1995). Differential Effects of Subjective knowledge, Objective Knowledge, and Usage Experience on Decision-making: An Exploratory Investigation. Journal of Consumer Psychology, 4, 153-180.

Petty, R., and J.T. Cacioppo (1986). The Elaboration Likelihood Model of Persuasion.” In Advances in Experimental Social Psychology (vol.19), edited by L. Berkowitz, 123-205. New York: Academic Press.

Petty, R.E., J.R. Priester, and P. Brinol (2002). Mass Media Attitude Change: Implications of the Elaboration Likelihood Model of Persuasion. In Media Effects, edited by J. Bryant and D. Zillmann, 155-198.Mahwah, New Jersey: Erlbaum.

Pham, M. T., and Avnet, T. (2004). Ideals and oughts and the reliance on affect versus substance in persuasion. Journal of Consumer Research, 30(4), 503-518.

Schram, A. (2005). Artificiality: The tension between internal and external validity in economic experiments. Journal of Economic Methodology, 12(2), 225-237.

Schumacker, R. E., and Lomax, R. G. (2004). A Beginner's Guide to Structural Equation Modeling. 2nd ed., Mahwah, NJ: Lawrence Erlbaum.

Severin, W. J. and Tankard, James, W., JR. (2001). Communication Theories: Origins, Methods, and Uses in the Mass Media (5th edition), Addison Wesley Longman, Inc.

Seward, Z. (2011). How to Say You’re Sorry: 10 CEO YouTube Apologies Retrieved 17th July, 2014, from http://blogs.wsj.com/deals/2011/10/14/how-to-say-youre-sorry-10-ceo-youtube-apologies/?mod=e2tw

Shamma, H.M., and Hassan, S.S. (2009). Customer and non-customer perspectives for examining corporate reputation, Journal of Product & Brand Management, Vol. 18 Iss 5 pp. 326 – 337.

Shannon, C., and Weaver, W. (1949). The Mathematical Theory of Communication. Urbana: University of Illinois Press.

Shin, D.H. (2010). The effects of trust, security and privacy in social networking: a security-based approach to understand the pattern of adoption. Interacting with Computers, 22, 5, pp. 428-138.

Swamynathan, G., Wilson, C., Boe, B., Almeroth, K. and Zhao, B.Y. (2008). Do social networks improve e-commerce? A study on social marketplaces. Online Social Networks, in Proceedings of the first workshop on Online social networks, 1-6, ACM.

Tellis, G. J., and Ambler, T. (2008). The Sage Handbook of Advertising. London: Sage.

Tybout, A. M., Sternthal, B., Malaviya, P., Bakamitsos, G. A., and Park, S. (2005). Information Accessibility as a Moderator of Judgements: The Role of Content versus Retrieval Ease. Journal of Consumer Research, 32, 76-85.

Udo, G.J., Bagchi, K.K., and Kirs, P.J. (2010). As assessment of customers’ e-service quality perception, satisfaction and intention. International Journal of Information Management, 30, 481-492.

Venäläinen, N. (2014). “I’m Sorry” – an analysis on CEO apologies through YouTube as part of corporate crisis communication. Accessed via: http://epub.lib.aalto.fi/en/ethesis/pdf/13631/hse_ethesis_13631.pdf

Wirtz, J., and Mattila, A. S. (2004). Consumer responses to compensation, speed of recovery and apology after a service failure. International Journal of Service Industry Management, 15(2), 150-166.

TABLES

Table 1: Sample characteristics

	Customers (n=241)	Non-customers (n=271)
Demographic Characteristics	N(%)	N(%)
Gender
Male	117(48.5%)	139(51.3%)
Female	124(51.5%)	132(48.7%)
Age
18-25	29(12.0%)	33(12.2%)
26-30	36(14.9%)	36(13.3%)
31-40	46(19.1%)	60(22.1%)
41-50	64(26.6%)	59(21.8%)
50+	66(27.4%)	83(30.6%)
Ethnicity
African American	39(16.2%)	18(6.6%)
Native American	10(4.1%)	14(5.2%)
White American	106(44.0%)	139(51.3%)
Asian American	18(7.5%)	7(2.6%)
Hispanic American	25(10.4%)	17(6.3%)
Multiracial	7(2.9%)	5(1.8%)
Non-USA Native	6(2.5%)	7(2.6%)
Other	30(12.4%)	64(23.6%)
Education
Some high school or less	3(1.2%)	1(.4%)
High school graduate or equivalent	22(9.1%)	57(21.0%)
Vocational/technical school	17(7.1%)	20(7.4%)
Some college, but no degree	42(17.4%)	68(25.1%)
College graduate (four year program)	79(32.8%)	80(29.5%)
Some graduate school	16(6.6%)	11(4.1%)
Graduate degree	51(21.2%)	25(9.2%)
Professional degree (M.D., J.D., etc.)	11(4.6%)	8(.4%)
Other	0(0%)	1(.4%)
Household Income
Less than $15,000	9(3.7%)	21(7.7%)
$15,000-$25,000	16(6.6%)	32(11.8%)
$25,001-$50,000	55(22.8%)	71(26.2%)
$50,001-$75,000	68(28.2%)	64(23.6%)
Over 75,000	89(36.9%)	76(28.0%)
Prefer not to specify	4(1.7%)	7(2.6%)
Employment Status
Full-time employed	166(68.9%)	138(50.9%)
Part-time employed	22(9.1%)	35(12.9%)
Out of work (looking for work)	7(2.9%)	20(7.4%)
A homemaker	16(6.6%)	28(10.3%)
A student	10(4.1%)	13(4.8%)
Out of work (not looking)	0 (0%)	1(.4%)
Retired	11(4.6%)	20(7.4%)
Unable to work	7(2.9%)	14(4.7%)
Other	2(.8%)	2(.7%)
Incident Familiarity (Prior to this study measured as a dichotomous variable)
Yes	150(62.2%)	78 (28.8%)
No	91(37.8%)	193(71.2%)
Prior Exposure to Apology
Yes	55(22.8%)	8(3%)
No	186(77.2%)	263(97%)

Table 2: Measurements, CFA factor loadings, AVE, CR, reliabilities, and descriptive statistics

Variables	Scale Items	CFA Loadings	N=512
Incident Familiarity (Moore, Stammerjohan and Coulter 2005)	On a scale of 1 to 7, please indicate the number that best describes the extent of your answer to the following questions.		AVE=.79 CR=.94 a=.97 M= 3.25 (SD=2.05)
	In general would you consider yourself familiar or unfamiliar with BlackBerry’s Service Failure Incident in 2011? (1=Very familiar, 7=Very unfamiliar) REVERSE CODED	.93
	Would you consider yourself informed or uninformed about BlackBerry’s Service Failure Incident in 2011? (1=Not at all informed, 7=Highly informed)	.98
	Would you consider yourself knowledgeable about BlackBerry’s Service Failure Incident in 2011? (1=Know a great deal, 7=Know nothing at all) REVERSE CODED	.96
Subjective Knowledge of Service Provider (Canli 2003)	On a scale of 1 to 7 please indicate the extent to which you agree or disagree with the following statement.		AVE=.88 CR=.96 a=.96 M=3.98 (SD=1.89)
	I know a lot about BlackBerry’s products and services. (1=Strongly disagree – 7= Strongly agree)	.93
	My knowledge of BlackBerry’s products and services is ___. (1=Inferior - 7=Superior)	.94
	My knowledge of BlackBerry’s products and services is ___. (1=Very poor – 7=Very good)	.94
Persuasiveness of the Apology (Pham and Avnet 2004)	Please indicate the extent to which you agree or disagree with the following statement. (1=Strongly disagree – 7= Strongly agree)		AVE=.77 CR=.91 a=.91 M= 4.22 (SD=1.30)
	This message influences my opinion about BlackBerry’s products and services.	.90
	This message changed my attitude toward BlackBerry’s products and services.	.88
	The message will influence my habits related to BlackBerry’s products and services.	.86
Attitudes towards the Service Provider (After Apology) (Tybout, et al. 2005)	On a scale of 1 to 7, please indicate your attitudes towards BlackBerry’s products and services.		AVE=.84 CR=.95 a=.96 M= 4.64 (SD=1.48)
	Unreliable: Reliable	.90
	Low quality: High quality	.93
	Not valuable: Valuable	.95
	Undesirable: Desirable	.89
Trustworthiness of Service Provider (After Apology) (Erdem and Swait 2004)	Please indicate the extent to which you agree or disagree with the following statements. (1=Strongly disagree – 7= Strongly agree)		AVE=.79 CR=.95 a=.96 M= 4.76 (SD=1.31)
	This brand’s product claims are believable.	.96
	Over time, my experiences with this brand have led me to expect it to keep its promises, no more and no less.	.95
	This brand has a name you can trust.	.91
	This brand doesn’t pretend to be something it isn’t.	.95
Satisfaction with Service Provider After Apology (Karatepe and Ekiz 2004)	Please indicate the extent to which you agree or disagree with the following statements. (1=Strongly disagree – 7= Strongly agree)		AVE=.73 CR=.89 a=.96 M=4.82 (SD=1.04)
	My satisfaction with this company/brand has increased.	.64
	My impression of this company/brand has improved.	.93
	I now have a more positive attitude towards this company/brand.	.96
Behavioural Intentions - to remain or become a customer (Jones, et al. 2000)	Now that you are aware of the October 2011 Blackberry network outage incident and have watched Blackberry’s CEO apology on YouTube regarding that incident, please indicate, on a scale of 1 to 7, you intentions to remain a Blackberry customer or become a Blackberry customer.		AVE=.92 CR=.98 a=.98 M= 4.33 (SD=1.77)
	Non-existent : Existent	.96
	Improbable : Probable	.97
	Impossible : Possible	.97
	Uncertain : Certain	.95
	Probably not : Probably	.93

Table 3: Correlations among constructs for the total sample (N=512)

Prior Exposure to Apology	1
Incident Familiarity	-.43**	1
Subjective Knowledge	-.33**	.61**	1
Persuasiveness of Apology	-.24**	.38**	.48**	1
Satisfaction After Apology	.03	.06	.14*	.29**	1
Trustworthiness After Apology	-.13**	.27**	.41**	.57**	.67**	1
Attitude After Apology	-.05	.12**	.25**	.37**	.62**	.68**	1
Behavioural Intentions	-.20**	.39**	.58**	.53**	.62**	.74**	.62**	1

**p≤.01, *p≤.05

Table 4: Structural Equation Model Results Examining H1 to H13 on the Total Sample

(N=512) Hypothesized Relationships	Std. Loadings	S.E.	z-scores	Hypothesis Supported?
H1: Persuasiveness of Apology Satisfaction with Service Provider After Apology	.77**	.02	35.53	Yes
H2: Satisfaction with Service Provider After ApologyBehavioural Intentions	.24**	.05	4.96	Yes
H3: Persuasiveness of Apology Trustworthiness of the Service Provider After Apology	.69**	.03	26.17	Yes
H4: Trustworthiness of the Service Provider After ApologyBehavioural Intentions	.32**	.06	5.76	Yes
H5: Persuasiveness of Apology Attitudes towards the Service Provider After Apology	-.08	.06	-1.22	No
H6: Satisfaction with Service Provider After Apology Attitudes towards the Service Provider After Apology	.04	.07	.57	No
H7: Trustworthiness of the Service Provider After Apology Attitudes towards the Service Provider After Apology	.71**	.05	14.0	Yes
H8: Attitudes towards the Service Provider After ApologyBehavioural Intentions	.24**	.04	6.26	Yes
H9: Prior Exposure to the Apology Persuasiveness of Apology	-.08	.05	-1.69	No
H10: Subjective Knowledge of the Service Provider Behavioural Intentions	.35**	.03	11.48	Yes
H11: Subjective Knowledge of the Service Provider Incident Familiarity	.64**	.03	23.34	Yes
H12: Incident Familiarity Prior Exposure to the Apology	-.44**	.04	-12.19	No
H13: Incident Familiarity Persuasiveness of Apology	.39**	.04	8.77	Yes