Customer relationship satisfaction and revenge behaviors: Examining the effects of power

A customer’s revenge can significantly impact a company’s reputation. After a service failure, a desire for revenge is created and customers can decide to either commit direct or indirect revenge. With direct revenge, customers have a direct altercation with the company’s staff and albeit negative, it allows companies the opportunity for service recovery. Indirect revenge, on the other hand, is much more damaging where customers partake in negative Word Of Mouth (WOM) both online and offline. Therefore, better understanding customer revenge behaviour is important for marketing managers. Two key issues have not been resolved in the customer revenge debate: first, what the influence of relationship quality is. Does relationship quality negatively impact desire for revenge or the so-called “love is blind” approach; or does relationship quality increase the likelihood of desire for revenge and eventual revenge behaviour; or the “love is hate” approach. The second issue is what the role of consumer power and empowerment has on whether consumers partake in direct or indirect revenge. Through a self-completion survey, this study finds support for the “love is blind” approach, finding that relationship satisfaction reduces the desire for revenge. In addition, it shows that power was positively related to customers’ desire for revenge and empowerment positively related to their direct revenge behaviours, the better of the two evils. The study outlines how firms can use customer relationship management to manage customer satisfaction and empowerment in order to protect their equity and reputation in the event of a service failure.


INTRODUCTION
Revenge and retaliation are terms used in the literature to describe a multitude of negative actions a customer may take towards a firm in the event of a service failure (Bechwati and Morrin, 2003;Fitness, 2001;Gregoire and Fisher, 2006). Firms have had to acknowledge that customers are wielding increasing power over them and have the potential to harm their equity and reputation thereby deterring future customers (Hansen et al., 1996;Ward and Ostrom, 2006). It is in the best interest of firms to be aware of the path a consumer may take from frustration and dissatisfaction to radical action, so that they may protect themselves against aggressive brand *Corresponding author. E-mail: elsamari.botha@uct.ac.za. Tel: +27 (0) 21 650 4053. Fax: +27 (0) 86 719 7658. attacks (Ling, 2008). Increasingly, it has become imperative for firms to realise the possible implications of consumer retaliatory behaviours, especially now that consumers have easy access to complaint channels like HelloPeter.com, which have the potential to reach millions (Ward and Ostrom, 2006). Current theory on revenge behaviour states this: In light of a service failure, customers take a retrospective view on their relationship with the firm and acquires an idea of how satisfied they are with the relationship. Relationship satisfaction can then be defined as how content a customer is within their relationship with a firm (Crosby et al., 1990;Gregoire and Fisher, 2006;Naude and Buttle, 2000). Relationship satisfaction then in turn, is seen to have an influence on the customer's desire to take revenge upon the defaulting firm. How relationship satisfaction impact desire for revenge, however, is still uncertain.
Customer relationship marketing (CRM) states that a firm's relationships with its customers are important to create customer loyalty and profitability and can, in the event of a service failure, protect the firm against customer revenge (Berry, 1995;De Wulf et al., 2001;Hess et al., 2003). However, in recent years researchers have found the opposite; that a firm's most loyal and satisfied customer may possess the greatest desire for revenge (Gregoire et al., 2010;Ling, 2008). However, there is still a debate over whether strong customer relationships result in increased desire for revenge (that is, "I loved you. How could you do this to me?") or a decreased desire for revenge (that is, "I love you and therefore I forgive you.").
The desire for revenge is a human reaction that occurs in situations when people feel betrayed or wronged by an entity or person (Fitness, 2001;Bechwati and Morrin, 2003). This reaction is an impulse to remedy the insult or injury by retaliation, and may or may not manifest itself as actual revenge behaviours.
Revenge behaviours are the actions that a consumer takes against a firm after service failure, with the intention to cause harm or get even with the firm (Cheng et al., 2006;Gregoire et al., 2010;Hibbard et al., 2001;Ward and Ostrom, 2006;Zourigg et al., 2009). Revenge behaviours are classified as either direct or indirect. Direct revenge refers to in-person interactions between the firm and representatives of the firm; such as vindictive complaining, aggression or non-verbal communication (Eisenberger et al., 2004;Gregoire et al., 2010;McColl-Kennedy et al., 2009). On the other hand, indirect revenge refers to actions that do not involve contact with the firm in their deployment; such as negative word of mouth and patronage reduction (Gregoire and Fisher, 2006;Gregoire et al., 2010).
This study seeks to investigate the influence of relationship satisfaction on the customer's reaction following a service failure. This study also considers the amount of power the customer feels they hold over the offending firm and how empowered customers feel in their relationship with the firm. Customers who feel empowered are more likely to be direct about their dissatisfaction, as they feel their complaint will be adequately resolved if they approach the firm themselves (Hansen, 1997). When customers approach the firm directly, the firm faces substantially less damage than when a customer employs indirect tactics.
Both the desire for revenge, as well as, direct revenge behaviours are influenced by the power customer feels they possess over the firm. An important distinction must be made concerning this power. Power and empowerment are considered to be two separate constructs. Power is conceptualised according to bases, which are the sources on which the actual effect and extent of power is dependent (Belaya and Hanf, 2009). The power bases that have been found to be present in consumers in their relationships with firms are reward, expert and coercive power (Bove and Robertson, 2005). Empowerment, on the other hand, refers to the customer's perceived ability to influence the firm in a manner that he or she will find advantageous and is not comprised of specific bases of power (Gregoire et al., 2010;Goodwin and Verhage, 1989;Menon and Bansal, 2007). Desire for revenge may be exacerbated by the bases of power a customer possesses over the firm, whereas direct revenge behaviours may be influenced by how empowered a customer feels.
In essence, the purpose of this study can be summarised in two points: First, it aims to add support for CRM and its importance with regard to customer revenge. That is to say that a customer who is satisfied with their relationship with a firm will, in the event of a service failure, have a dampened desire for retaliation; therefore buffering the firm from consumer revenge. Second, this study aims to highlight the importance of power in the context of customer revenge. Specifically that customer power can positively influence their desire for revenge and empowerment, in turn, can positively influence direct revenge behaviours. This paper will proceed by first providing an understanding of marketing literature pertaining to service failure and customer relationships, customer revenge and power. These constructs will be used to motivate a conceptual model depicting their interactions. This will be followed by a description of the research methodology employed and the results obtained from the data collected. Lastly, this study will finish with a discussion and conclusion with regard to the results and will also address future research and the limitations of the study.

SERVICE FAILURES AND CUSTOMER RELATIONSHIPS
A service failure represents a critical incident which occurs between a firm and a customer. This incident is a malfunction in the service the firm offers, which inconveniences the customer to some extent (Smith and Bolton, 1998). The failure is a violation of the promises or performances a customer expects from the firm and, as such, undermines the customer's relationship with the firm (Colgate ad Norris, 2001;Hess et al., 2003;Palmer et al., 2000). Services are variable and cannot be standardised to a particular level in the manner a commodity can (Buttle and Burton, 2001;Hess et al., 2003). This implies that all firms, which offer any element of a service in their operations, will inevitably present service failures (Buttle and Burton, 2001;Colgate and Norris, 2001).
A service failure, regardless of the cause, can provoke a customer to exhibit costly behaviours for the firm, such as defection (Buttle and Burton, 2001;Colgate and Norris, 2001). A firm may attempt to mitigate such costs by undertaking service recovery (Hess et al., 2003;Weun et al., 2004;Wirtz and Matilla, 2004). However, if the recovery efforts fail it can create greater frustration and dissatisfaction than experienced during the original failure (Smith and Bolton, 1998;Wirtz and Matilla, 2004).
The customer experiences such frustration and dissatisfaction because the service failure represents a breakdown of the relationship and its norms. There is agreement in the literature that a relationship is negatively influenced by service failures (Bejou and Palmer, 1998;Weun et al., 2004). However, there is division regarding whether a relationship of high quality magnifies or buffers the effect of a service failure (Bejou and Palmer, 1998;Hess et al., 2003).
Relationship measures have been consistently challenging to define due to the interchange-ability of terms used and the numerous definitions available (Bove and Johnson, 2001;Naude and Buttle, 2000;Storbacka et al., 1994). In this study, the term relationship satisfaction will be used to describe the relationship between the customer and firm. Satisfaction represents a vitally important factor in relationships and in their maintenance and is highly correlated to the quality of a relationship (Naude and Buttle, 2000;Hendrick, 1988;Storbacka et al., 1994). Relationship satisfaction was defined by Hallowell (1996) as the difference between the perception a person has of the value received from the relationship and the value expected. It describes how content a customer is, over time, with regards to the goods and services received from the firm (Crosby et al., 1990;De Wulf et al., 2001;Gregoire and Fisher, 2006;Naude and Buttle, 2000). It is important to note that in this study, relationship satisfaction is considered and measured from the customer's perspective, not that of the firm, as it is the customer's feelings and actions being investigated (Aaker et al., 2004;Gregoire and Fisher, 2006;Hess et al., 2003).
Relationship satisfaction is important to firms because a high level of satisfaction has been found to lead to more frequent interaction and spending, which contributes to the firm's profitability (De Wulf et al., 2001;Hennig-Thurau and Klee, 1997). Additionally, a high quality relationship with a high level of satisfaction has been found to protect a firm from customer attacks in the event of a service failure, by dampening the customer's desire to seek justice (Hess et al., 2003;Tax and Brown, 1998). However, recent literature has suggested that the opposite may be true and some studies have found a high quality relationship could magnify a customer's desire for revenge. These studies suggest that CRM could in fact pose a threat to a firm's profitability and reputation in the event of a service failure (Aaker et al., 2004;Fisher, 2006, 2008a;Gregoire et al., 2009a). Gregoire and Fisher (2008b) dub these two opposing views as the traditional "love is blind" versus the more recent "love turns to hate" understanding.
"Love is blind" describes the situation whereby a Mdakane et al. 10447 customer ignores negative information that is not in line with the relationship norms which they are accustomed to with a firm. Customers disregard the failure due to the personal elements in their relationship with the firm (Gregoire and Fisher, 2006;Herr et al., 1983). The customer uses the firm's past performance and relationship norms to mitigate the service failure and maintain the relationship as if the event had not occurred. This is done in order to avoid the discomfort of the discrepancy felt by the customer and therefore, the customer's desire for revenge is dampened (Gregoire and Fisher, 2006;Hess et al., 2003;Herr et al., 1983). As a result, the customer attributes the service failure to an unstable, uncontrollable cause (Bhattacharya et al., 1995;Gregoire and Fisher, 2006;Hess et al., 2003).
On the other hand, the idea of "love turns to hate" states that when a customer is involved in a high quality relationship with a firm and a service failure occurs, it violates the relationships norms and the customer feels betrayed (Aaker et al., 2004;Colgate and Norris, 2001;Fisher, 2006, 2008b;Gregoire et al., 2009a). This betrayal magnifies the customer's desire for justice and can lead the customer to take more extreme action than would be taken if the relationship were of low quality (Aaker et al., 2004;Hess et al., 2003). This can be significant, especially if the customer has tried to bring the service failure to the firm's attention on more than one occasion (Gregoire et al., 2009a).

FROM SERVICE FAILURE TO REVENGE
The desire for revenge is a deep-seated human reaction that occurs when people feel betrayed or wronged by an entity or person (Fitness, 2001;Bechwati and Morrin, 2003). The desire for revenge is an impulse to counteract the pain and powerlessness an individual feels after being insulted or undermined (Fitness, 2001;McColl-Kennedy et al., 2009). In consumer behaviour, the desire for revenge stems from a service failure that violates the norms of the relationship (Fitness, 2001;Ward and Ostrom, 2006). This violation prompts the customer to seek justice and creates the need to equalise the damage caused by the firm using punishment (McColl-Kennedy et al., 2009;Ward and Ostrom, 2006). However this desire may be diluted in the presence of relationship satisfaction (Hess et al., 2003;Tax and Brown, 1998). This leads to the first hypothesis of this study: H 1 : In the event of service failure/s, a customer's satisfaction within her relationship with a firm will have a negative influence on the customer's desire for revenge.
It is important to make a distinction between desire for revenge and revenge behaviours. Desire for revenge refers to the need felt to cause damage to the property or reputation of a firm, whereas revenge behaviours refer to the act of carrying out this desire (Gregoire et al., 2010;McColl-Kennedy et al., 2009). Desire for revenge may not always translate into revenge behaviours, as the customer may feel powerless to act, feel their actions would be immoral or they could fear retaliation (Fitness, 2001;Gregoire and Fisher, 2008a;Gregoire et al., 2009b). Although the two constructs are separate, they are closely linked as revenge behaviours cannot occur without there being an initial desire for revenge. Moreover, desire for revenge is the energy that motivates a customer to perform revenge behaviours (Bechwati and Morrin, 2003;Gregoire et al., 2009aGregoire et al., , 2010. As the desire for revenge is a need, it can either be repressed or acted upon (Gregoire et al., 2010;Gregoire and Fisher, 2006;Zourigg et al., 2009). These revenge behaviours are defined as a categorisation of actions performed by consumers in order to warn other consumers against the firm responsible, to cause harm to the firm or to gain closure surrounding the incident (Bougie et al., 2003;Cheng et al., 2006;Gregoire et al., 2010;Gregoire and Fisher, 2006). Two types of revenge behaviour, each with different antecedents, can be observed: Direct and indirect revenge (Gregoire et al., 2010).

INDIRECT REVENGE BEHAVIOURS
Indirect revenge behaviours are actions taken by a customer 'behind the scenes', meaning there is no direct contact between the firm and the customer (Gregoire et al., 2010). The harm is generally intended for the firm rather than its employees or representatives. Indirect revenge behaviours can take the form of negative word of mouth, public complaining, negative contributions and patronage reduction (Gregoire and Fisher, 2006;Gregoire et al., 2010).
Negative word of mouth is utilised in the hope of ruining the status of the firm, as well as, to warn others against the firm (Cheng et al., 2006;Gregoire and Fisher, 2006;Gregoire et al., 2010;Schoefer and Diamantopoulos, 2008). Patronage reduction, in turn, refers to the action taken by a customer to switch to a competing firm (Gregoire and Fisher, 2006;Roehm and Brady, 2007;Schoefer and Diamantopoulos, 2008). Public complaining and negative contributions occur through social networks and third party websites, and are engaged in to alert the public to the misbehaviours of a firm, as well as, to avenge the failure (Gregoire et al., 2010;Ward and Ostrom, 2006). The example below is an extract from a complaint web-site. The customer made use of the site to voice their frustration: 'SAA voyager is absolutely useless! I flew to Munich and back in 2010. A day after getting back I had to claim voyager credits. They said it would take 3 days! Three days later I am still waiting. They say they will request it today and then I have to wait 45 DAYS AGAIN! That is very ridiculous!' (hellopeter.com, 2010).
Consumers can carry out revenge as earlier illustrated, by posting negative remarks and contributions on the internet. Consumers can easily make use of social media such as Facebook, Twitter, YouTube and public complaining web-sites to vent their frustration and post complaints after experiencing a service failure, and they have the potential to reach large audiences (Gregoire et al., 2009a;Kerr et al., 2006;Wei and Miao, 2011). Although firms attempt to monitor such websites, it is challenging to control due to the customer's anonymity (Gregoire et al., 2010). As these platforms provide anonymity to customers, it is a safe place for them to vent their frustration, making it far easier for customers to act on their desire for revenge (Gregoire et al., 2010). This leads to the second hypothesis of this study: H 2 : In the event of service failure/s, a customer's desire for revenge will have a positive influence on her acting out indirect revenge behaviours.

DIRECT REVENGE BEHAVIOURS
Direct revenge behaviours take place face to face between a customer and a firm after a service failure (Gregoire et al., 2010). Direct revenge can take the form of non-verbal communication, vindictive complaining or aggression (Eisenberger et al., 2004;Gregoire et al., 2010;McColl-Kennedy et al., 2009). Non-verbal communication occurs when customers show their dissatisfaction by using facial expressions and body language such as sneering, glaring and rolling of the eyes (Eisenberger et al., 2004;McColl-Kennedy et al., 2009). Vindictive complaining occurs when a customer uses verbal attacks on employees to voice frustration and to distract firm operations (Gregoire et al., 2010;Gregoire and Fisher, 2006). Aggression can take the form of slamming doors, vandalising or destroying property and physical confrontations with employees (Gregoire et al., 2010;McColl-Kennedy et al., 2009). An example of a customer, who vandalised the property of Comcast, an American telephone company, after the event of a service failure, is illustrated as follows: 'It had never occurred to me to take a hammer to a phone company before, but I was just so upset. After I hit the keyboard, I turned to this blonde who had been there the previous Friday, the one who told me to wait for the manager, and I said, "Nowdo I have your attention?" (Tucker, 2007)' Customers have been known to carry out their desire for revenge through direct forms of revenge in reaction to a service failure. This leads to the third hypothesis this study aims to test: H 3 : In the event of service failure/s, a customer's desire for revenge will have a positive influence on her acting out direct revenge behaviours.
Direct acts of revenge, although harmful, are likely to have short lived effects for a firm, due to the firm being able to take immediate actions to manage the situation and prevent further damage (Eisenberger et al., 2004;McColl-Kennedy et al., 2009). On the other hand, Ward and Ostrom (2006) have identified that indirect methods of revenge can have lengthy, damaging effects on the company's reputation. A key determinant of whether or not a customer will elect to utilise revenge behaviours and harm the firm is the amount of perceived power possessed by that customer (Gregoire et al., 2010;Ward and Ostrom, 2006).

REVENGE AND POWER
A determinant of whether or not a customer will elect to utilise revenge is the amount of power the customers perceives themselves to possess (Gregoire et al., 2010). Customers have power over an organisation because they can control whether or not an organisation is able to achieve its sales goals and other organisational objectives (Fine et al., 1999). Both power and empowerment can separately and independently influence a consumer's desire for revenge and revenge behaviour.

Power
Power is defined as the potential ability of one party to influence the actions of another (Brown et al., 1995;El-Ansary and Stern, 1972;Hunt and Nevin, 1974). The possession of power must be accompanied by the means or opportunity to apply it in order for the influence objectives to be achieved (Frazier, 1983). The actual effect and extent of power is dependent on its source, referred to as power bases (Belaya and Hanf, 2009). French and Raven (1959) described power according to five bases, namely coercive, reward, legitimate, expert and referent power.
In a relationship between a customer and a firm, the literature states that the effective power bases are reward, expert and coercive power; whereas legitimate and referent power have not proven to be effective (Bove and Robertson, 2005;Fine et al., 1999;Hansen, 1997). In light of this, the study will use only in reward, expert and coercive power. Reward power refers to the ability of the influencer to provide desirable positive benefits (Bagraim et al., 2010;French and Raven, 1959). Expert Mdakane et al. 10449 power arises from influence exercised as a result of special skill, expertise, or knowledge (Bagraim et al., 2010;Belaya and Hanf, 2009;French and Raven, 1959). Coercive power is the ability to influence another party based on instilling a fear of negative consequences (Bagraim et al., 2010;Belaya and Hanf, 2009;French and Raven, 1959). Hansen (1997) elaborated on the concept of customer power bases by stating that different power bases manifest themselves as different types of retaliation behaviours. There is a scarcity of research regarding the effect that power has on the desire for revenge and revenge behaviours. As the desire for revenge precedes all revenge behaviour undertaken by a customer it may be affected by the power a customer holds over a firm (Augusto de Matos et al., 2011). Customers who possess power feel they are able to cause damage to the organisation, be it the organisation's profitability or reputation (Ward and Ostrom, 2006). Therefore, these customers feel a greater desire for revenge should the firm fail to perform to expected standards: H 4 : Power (as measured by its bases) positively influences desire for revenge.

Empowerment
Some authors have not elected to view power according to the power bases, but rather to divide power according to whether or not customers feel empowered in their relationship with a firmor not (Gregoire et al., 2010;Goodwin and Verhage, 1989;Menon and Bansal, 2007). Empowerment is regarded as the customer's perception of possessing an increased ability to affect others intentionally (the defecting firm) and prevent undesired results (Wathieu et al., 2002). Empowerment differs from power in that it is focused on the consumers benefiting from the increased ability to influence the firm (Hunter and Garnefeld, 2008). Fang et al. (2008) proposed that consumers with higher perceptions of empowerment were more inclined to perform destructive actions against firms. This may be attributed to the fact that customers who feel empowered have reduced fears of the retaliatory actions that a firm may take against them (Aquino et al., 2001;Gregoire et al., 2010). These customers also feel they are able to cause more damage to the firm, be it the firm's profitability or reputation (Gregoire et al., 2010). However, customers who feel disempowered perceive themselves as being dependent on the firm (Gregoire et al., 2010;Menon and Bansal, 2007). These customers want to ensure a positive interaction with the firm (Lee, 2009;Menon and Bansal, 2007). This implies that in the event of a service failure, they are less likely to engage in revenge behaviours (Gregoire et al., 2010). Alternatively, disempowered customers can opt to utilise indirect revenge, due to its anonymity (Aquino et al., 2001;Gregoire et al., 2010;Ward and Ostrom, 2006).
An example from workplace literature may be given to better explain this concept. Aquino et al. (2006) were able to show that employees who felt more empowered approached their line managers following a personal offense at work. On the other hand, employees who felt disempowered were found to undermine covertly the leadership with their line manager. Covert tactics, such as negative word of mouth, pose a greater threat as these render more damage to the reputation of the firm and do not allow for a transparent resolution process to take place (Aquino et al., 2006). Applying these findings to the relationship between a customer and a firm, this study posits that a customer who feels more empowered will take a direct approach when enacting revenge on a firm following a service failure. This study aims to add to the literature, which is currently lacking, regarding power and revenge behaviours. The following hypothesis is proposed: H 5 : A sense of empowerment (possessing power) held by a customer a positively influences direct revenge. Figure 1 is a diagrammatic representation of the constructs and hypotheses proposed by this study. It is an extension and combination of two models: the first having been proposed by Gregoire and Fisher (2006) and the second by Gregoire et al. (2010). Gregoire and Fisher's (2006) model looked at the interaction effect of relationship quality and controllability on a customer's desire for revenge. Gregoire et al. (2010) extended on the 2006 model by examining revenge behaviours broken into two categories: direct and indirect revenge. They also considered the moderating effect of power on direct revenge behaviours. Figure 1 adds to the literature by firstly considering power, as measured by its bases, as a direct influence on desire for revenge, and separately considers empowerment as a direct influence on direct revenge behaviours.

METHODOLOGY
A conclusive, descriptive research design was used through the use of an online survey. A Qualtrics link was emailed to the target population. The following areas first address the target population and sampling that was used. Thereafter the data collection procedures and measures are discussed.

Sampling
The target population for this study comprised of university students. Using a student sample is in line with many of the studies which address consumer reactions to service failures and consumer revenge (Bechwati and Morrin, 2003;Gregoire and Fisher, 2006;Hess et al., 2003;Liu, 2009;Smith and Bolton, 1998). In a questionnaire, students were required to recall a recent event of service failure (Aquino et al., 2001;Gregoire and Fisher, 2006). This study opted to use convenience sampling as no list of students could be obtained for random sampling. This sampling method was in line with what previous studies have used (Gregoire et al., 2010;Schoefer and Diamantopoulos, 2008).
This study aimed to realise a sample size of 150 students. This size was based on previous research (Liu, 2009) and this sample size was identified as an adequate size for online surveys (Zhang, 1998). We also did not want to make the sample too large because we used structural equation modelling and did not want to introduce any biases in the data. A link to the survey was posted on the faculty website and 271 respondent's participated in the survey. However, only 153 students completed 80% or more of the survey. Of those 153 responses, one was removed due to it being unsatisfactory because the respondent only answered the extreme positive values (7) for the whole   questionnaire. This resulted in a sample size of 152 usable questionnaires. The gender divide of respondents was biased towards females, with 65.1% of respondents in the sample being female and only 33.6% of the sample being male. Of the respondents, 95.4% of the students were between the age of 18 and 25, which is to be expected given the sample was of students.

Data collection procedures
The questionnaire was pretested on a convenience sample of ten students. No changes were made to the measurement instrument and data collection subsequently ran for one month during August 2011 and was collected using the software Qualtrics via a link on the faculty website. In order to increase the sample size necessary, the survey had an attached incentive whereby respondents who completed the questionnaire were entered into a lottery with a small cash prize. This was consistent with previous studies that offered incentives to their student respondents in order to gain an adequate sample size (Gregoire et al., 2010;Roehm and Brady, 2007).

Measures
All the constructs in this study were measured using a seven-point Likert scale, based on relevant previous literature. The Likert scale ranged from 1, indicating Strongly Disagree to 7, indicating Strongly Agree. Table 1 summarizes the studies that were used to compile the measures. These scales performed well and rendered high reliability scores in all these studies.

RESULTS
The questionnaire measured the relevant constructs in order to run statistical analysis to test the proposed hypotheses. Table 2 shows a summary of the descriptive statistics for the stated relationship satisfaction, desire for revenge, revenge behaviours, power and empowerment. It shows each construct's mean rating and standard deviation. From Table 2 it can be seen that respondents' mean score for relationship satisfaction was 5.07, indicating they felt somewhat satisfied with their relationship with the firm prior to the service failure. A mean score of 3.67 was recorded for desire for revenge, indicating a neutral feeling with a slight inclination towards not feeling the desire. Direct revenge received a mean score of 2.87 indicating that respondents did not act out direct revenge behaviours. The mean score of 4.39 for indirect revenge behaviours reflected that respondents in general were neutral towards indirect methods, but with a slight inclination. It can be seen that respondents felt they possessed a little power over the firm due to expert, reward, and coercive power with a mean score of 4.37, but felt somewhat disempowered, with a mean rating of 3.2. The standard deviations for these constructs ranged from 0.91 to 1.65, which indicates a general consistency of answers, with the empowerment and desire for revenge having the largest standard deviations. This implies respondents have larger discrepancies with regard to these constructs, and perhaps stronger reactions. This would be logical for a construct such a desire for revenge, which is a powerful emotion (Fitness, 2001;Bechwati and Morrin, 2003).

PLS measurement model
The proposed model shown in Figure 1 was tested using partial least squares (PLS) analysis. PLS was favoured as it allows the incorporation of a large number of constructs for a smaller sample size, which is ideal for testing the proposed model. Before the model could be interpreted, the reliability and validity of the model was examined. The reliability of the tested model was regarded through internal consistency reliability and indicator reliability. Table 3 provides information that was needed to evaluate the reliability of this study. It can be seen in Table 3 that the Cronbach's alpha was greater than 0.7 for all constructs with the exception of power measured according to its bases. Power received a Cronbach's alpha of 0.642, which is considered as satisfactory by many studies (Malhotra, 2010). The composite reliability showed that all constructs with the exception of power received a value which is greater than 0.8. Power received a value of 0.657, which again is considered as satisfactory. Cronbach's alpha and composite reliability values indicate that the model exhibited internal consistency reliability. Indicator reliability was established by examining how each item loaded on the relevant construct. Most items received a loading greater than 0.7, with a few exceptions. Some individual items on direct revenge, indirect revenge and power received lower loadings, but collectively all loaded significantly. Together, this provided reliability for the model. Both convergent and discriminant validity were also tested. Convergent validity was established through examining the significance of each item loading. All constructs had their items load significantly with the exception of power. The second question measuring reward power was insignificant in its loading with a tvalue of 0.295. However, this item was kept in the model as its removal only improved the value very slightly. Additionally, literature supports its inclusion (Brown et al., 1995;El-Ansary and Stern, 1972;Hansen, 1997).
To further examine convergent validity, AVEs or average variance explained by the latent variables, which can be seen in Table 3, all proved acceptable with the exception of power (AVE = 0.23), direct revenge (AVE = 0.41) and indirect revenge (AVE = 0.42). The AVE of each construct had to be above 0.5 to meet the convergent validity criteria. Discriminant validity was established through examining the Fornall-Lacker Criterion, which can be assessed by looking at the latent variable correlation table (Table 4) as well as through an examination of cross loadings.
As can be seen from Table 4, discriminant validity was confirmed through two points: first, the square roots of AVEs along the diagonal were greater than all the other correlations that fell below or beside the value. Second, loading of an indicator on its assigned latent variable should be higher than its cross loadings on all other latent variables. An examination of the cross loadings revealed that discriminant validity was confirmed for each variable with the exception of power. Power had some items from empowerment which were greater than the items used in measuring reward power. Again, these items were retained in the model for validity purposes, as well as, the fact that their exclusion caused a decrease in the reliability of the measure.

PLS structural model
Given that the reliability and validity of the measurement model have been addressed, the structural model can be examined. Firstly, the fit of the structural model must be assessed, by examining the R square values of the latent endogenous variables in the model. The R-squares can be seen in Table 3 and indicate the percentage of variance explained by constructs in the model. Direct revenge received an R square of 34%, indirect revenge an R square of 22.9% and desire for revenge an R square of 14.7%. From the recommendations stated by Chin et al. (2003) direct revenge and indirect revenge showed a moderate fit while desire for revenge reported a weak fit. However, it is important to consider not only the R square, but also the level of significance and the path coefficients. The R square value holds less power if the path coefficients are one, significant, and two, above 0.2 or below -0.2 (Chin, 1998). In this model, these two additional considerations prove to be true, and the lower R square value therefore did not cause too much concern when interpreted together with the path coefficients.
The second consideration of the structural model is to examine the effect size of introducing power and empowerment into the model. The effect sizes of these additions were 0.0135 on desire for revenge, 0.0151 on direct revenge and 0.0013 on indirect revenge. These results showed that there was little difference between the model without power and empowerment and the model that includes these. This result indicates the model is stable and that the inclusion of variables does not produce drastic changes (Cohen, 1988). Overall, both the measurement and structural model provided a valid and reliable base for the findings of the model. Figure 2 shows the tested model. The standardized path coefficients and t-values for the models were examined to show the significance of the relationship between the variables. The t-values were computed through the bootstrapping procdure and a 5% significance level (that is, α = 0.05) was used to test the hypotheses. All hypothesized relationships were significant at the 5% level. The first hypothesis (H 1 ) indicated that the association between relationship satisfaction and desire for revenge was negative with a path coefficient of -0.204, and an associated t-value of 2.418. The second hypothesis (H 2 ) indicated a positive relationship between desire for revenge and direct retaliatory behaviour with a path coefficient of 0.538, and a t-value of 8.643. The third hypothesis (H 3 ) tested, indicated that desire for revenge has a strong positive relationship with indirect revenge behaviours with a path coefficient of 0.479 and an associated t-value of 7.861. The forth hypothesis (H 4 ) tested indicated a significant positive relationship between power and desire for revenge with a path coefficient of 0.342, and a t-value of 2.147. The last hypothesis (H 5 ) tested, indicated that empowerment has a positive relationship with direct revenge behaviours with a path coefficient of 0.154 and an associated t-value of 2.128.
These results supported the validity of the model, as all path coefficients were above 0.2 (or below -0.2), with the exception of empowerment and all relationships were found to be significant. This helped mitigate the concern surrounding the low R square on desire for revenge.

DISCUSSION AND CONCLUSIONS
The purpose of this study was twofold. It evaluated, within the context of a service failure, the effect a relationship with high satisfaction between a customer and firm would have on the desire for and potential acting out of retaliation against the firm. The study also aimed to establish the role of power in the context of customer revenge; specifically the effects of power and empowerment on desire for revenge and direct revenge behaviours respectively. This study contributes to the literature by providing support for the "love is blind" approach as it found that relationship satisfaction decreases customers' desire for revenge. Additionally, this study adds to the limited body of literature pertaining to power and customer revenge. This study found that power, as measured by its bases, has a positive influence on desire for revenge, as well as that empowerment has a positive effect on direct revenge behaviours. These findings have some interesting implications.
Firstly, as desire for revenge was found to have a strong influence on revenge behaviours (both direct and indirect), high relationship satisfaction between a firm and its customers could be greatly advantageous. This result supports relationship marketing literature and suggests that the desire for revenge and therefore revenge behaviours, post service failure, are buffered by a relationship of high satisfaction, supporting the notion that customer relationships are still very important to a firm. Building relationships and satisfying customers could shield the firm from customer attacks, bringing relationship marketing under the spotlight.
Although recent research by Gregoire and Fisher (2006) and Gregoire et al. (2010) have questioned the benefits of relationship marketing in the context of service failures and revenge theory, many studies still find relationship quality and satisfaction are a significant shield for firms. Berry (2002) stated that relationship marketing is vital in the context of any firm which has an aspect of service in its offering. Firms can influence relationship quality by building satisfaction in the relationship. This can be achieved through firms providing value and quality in all aspects of a firm's offering (Berry, 1995;Morgan and Hunt, 1994;Hallowell, 1996).
As expected, this study found that desire for revenge has a positive influence on both direct and indirect revenge behaviours, but has a greater influence on direct. The implications of these findings is, first, once a desire for revenge has occurred, it is very likely that the desire will be acted upon. From this study and from what is suggested earlier, relationship quality and relationship satisfaction are some ways to influence a decrease in desire for revenge. However, if desire for revenge does manifest into revenge actions, firms need to be able to deal with such situations. Due to the overt nature of some direct revenge behaviours, firms can easily identify such customers and take immediate measures. One way firms can address direct revenge behaviours is by training staff to handle such situations, by banning customers from the firm's premises or demanding payment for damaged property. These actions could deter potential revenge attacks. On the other hand, indirect revenge behaviours are difficult for firms to manage as the actions occur outside the firm's control and the implications are often far worse than that of direct revenge behaviours.
It is preferable, in the instance of a service failure, that a customer elect to utilise direct revenge. This inflicts less damage to the reputation of the firm as they are afforded more control when customers act out direct revenge behaviours. If consumers elect to use online complaint forums or word of mouth to family and friends, the firm has no way of mitigating the damage. This implies that a situation which could have been solved through effective complaint management now has the potential to cause great harm to the firm's reputation. When a customer complains directly to a firm, that effectively gives the firm a second chance. If there is good complaint resolution management ingrained in the firm, it should be able to recover effectively and enhance the relationship (Gruber et al., 2011). Consumers that elect to complain directly to the firm show they still have an interest in maintaining the relationship, by giving the firm an opportunity to resolve the problem. Hansen (1997) found that power determined the type of reaction a customer has to a service failure. He found that high powered stakeholders were prone to making certain types of complaints consistent with those described as direct revenge in the study. This supports the finding that customers, who felt empowered, were prone to direct revenge. Aquino et al. (2006) similarly found that in an employer-employee relationship, employees with a higher sense of empowerment refrained from revenge by pursuing forgiveness and resolution in firms where there were clear procedures for remedy in place.
Therefore, consumers may be less inclined to use indirect methods of revenge if they feel that their complaints would be effectively resolved. In order to ensure that direct revenge is utilised, if consumer revenge behaviour is inevitable, firms must ensure that customers feel empowered in their relationship with the firm. A customer that feels empowered feels valued and believes that their complaints will be taken seriously and will be resolved immediately.
Customer empowerment can be built by firms clearly indicating how customers' complaints are a priority for the firm. Customers should feel that their complaint will be handled professionally and effectively, and that the situation will end favourably. Firms should also have proactive consumer complaint management systems in place so that customers feel their complaints are being resolved. If there are large time lags between the time a consumer places a complaint and the response from a representative of the firm, then the consumer is more likely to vent to family, friends, social media and other online platforms. In terms of consumer complaint management, firms should be consistent with measures such as guarantees on refunds and replacements. The process for receiving refunds and replacements should not be too laborious or lengthy.
Last, this study found that power has a positive influence on the desire for revenge. This implies that in a relationship between a customer and a firm, if the customer feels they possess power over the firm, they will have a greater desire to act upon revenge. For firms to be able to reduce desire for revenge they must decrease this power, by influencing reward, expert and coercive power.
Firms can decrease a customer's perceived expert power by establishing they are experts in the field and are completely knowledgeable in the industry. Training staff to understand fully the service or product would help the firm be considered an expert. Coercive power can be decreased by the firm being impenetrable to fear and threats. Although firms should work at developing strong positive relationships with their customers, the firm should not allow itself to be bullied by powerful customers.

LIMITATIONS AND FUTURE RESEARCH
Given that the study was retrospectively based, which required respondents to recall an event, problems Mdakane et al. 10455 associated with memory bias and loss may have affected the accuracy of customers' recall, which could have potentially impacted the findings of this study (Gregoire and Fisher, 2006;Gregoire et al., 2010;Schoefer and Diamantopoulos, 2008). Future research should screen consumers on the time that has elapsed between the occurrence of the service failure and the survey completion. It is suggested that a time period of three months or less since the event be employed, in order to reduce the likelihood and effect of memory bias. They should also possibly use a scenario-based questionnaire to get all respondents to answer according to the same service failure. Doing this, however, might remove the emotional element (or at least some of the emotional intensity of the service failure) from the revenge model. Another issue faced in this study was that respondents were asked to complete the survey after the occurrence of service failure. This is regarded as an issue because service failures usually change customers' attitudes and mindsets toward the service provider. Subsequently, retrospective measures of relationship quality regarding customers' previous attitudes toward the service provider may be biased (Gregoire et al., 2009b). For future research measuring relationship quality before the occurrence of a service failure would provide a more accurate gauge of the relationship's quality.
This study made use of a sample consisting of students, which could lead to some limitations. First, the use of students can cause age-related bias in the findings. This is due to different age groups presenting differences in their willingness to adjust perceptions, which can lead to differences in the manner in which they react to service failures. In addition, the range of service experiences of students is limited and lastly, studies have found that student samples can have significant nonresponse bias. It has been found that students who complete surveys are more likely to be high achieving females (Porter and Whitcomb, 2005). Future research is needed to validate the findings of this study by employing a wider sample base consisting of non-students.
Lastly, a great deal of research concerning consumer empowerment has been conducted in an online environment (Fang et al., 2008;Hunter and Garnefeld, 2008;Pires et al., 2006). Online consumers are considered to be more empowered due to greater access to information and product alternatives (Pires et al., 2006). Therefore, for future research, the conceptual model could be further explored using online consumers.