| 2
The moderating role of brand equity in the event of a
service failure: An empirical study into satisfaction,
intentions and behavior
Master Thesis
Author: Jeroen Aldenkamp Student number: 1383779 email: J.aldenkamp@gmail.com Phone: 06‐27086975Adress: Tuinbouwstraat 128a, 9717JP, Groningen, Netherlands
Preface
Ever since I started my master marketing I wanted to do something special with my final thesis. To some extent this has to do with my experiences on my bachelor thesis. The subject couldn’t grab my attention. I learned how difficult it was to work on a subject that’s completely uninteresting. Right then and there I decided that I would choose a more challenging and interesting subject for my master thesis. Something that could really prove I earn a masters degree, a proper challenge.
| 4
Management Summary
The purpose of this study is to investigate the way customers react to service failures and the level of recovery when customer perceived brand equity (CPBE) is taken into account. Previous literature has given ample evidence that service failures and subsequent recoveries influence customer attitudes and actual behavior. A failure decreases customer satisfaction, loyalty intentions and the share of wallet, and a good recovery can (to some extent) restore these again. Previous literature has also shown that CPBE influences this relationship. However, it remains a topic of debate whether having high CPBE is an advantage or a disadvantage in the event of a service failure?
To address this issue, this research puts forth the first large‐scale, empirical study into this topic. A three‐stage‐ least‐squares regression analysis was done where brand equity is used as a moderator in the relationship between service failures and the level of recovery and satisfaction, intentions and behavior.
The results show that brand equity can offset the effect of a service failure on customer satisfaction and loyalty intentions. However, this effect is not found to be significant on the share of wallet. So having high brand equity can be an advantage in the event of a service failure in that high CPBE customers have higher post‐ failure attitudes than low CPBE customers. This ‘buffering effect’ is fairly durable. Even as more failures occur the moderating effect of high CPBE remains present in the same way as it is present with a single service failure.
On the other hand, when looking at the level of recovery, having high brand equity is shown to be a disadvantage in customer satisfaction. The results show that the level of recovery has a stronger effect on low CPBE customers. Where high CPBE customers hardly react to any form of recovery attempt, an excellent recovery with low CPBE customers can restore their satisfaction even beyond pre‐failure levels. This is also known as the service recovery paradox. However, this effect has dissipated in loyalty intentions and share of wallet. The direct effect of the level of service recovery is not found to be significant in loyalty intentions and the share of wallet. There are however indirect effects that carry over the level of recovery from customer satisfaction to loyalty intentions and the share of wallet. These show that low CPBE customers still have a stronger reaction to the level of recovery, but this effect is not strong enough to create a service recovery paradox in customers’ loyalty intentions and the share of wallet.
| 6
Introduction and relevance of the research
Service failures are the most common reasons for customers switching service providers (Keaveney, 1995). Also, a growing number of customers is unsatisfied with recovery attempts offered by these service providers (Tax, Brown and Chandrashekaran, 1998; Andreassen, 2001). Research into service failures has traditionally been the domain of CRM literature since the impact of a service failure directly influences all kinds of customer perceptions about the firm. A service failure decreases satisfaction, creates negative word of mouth (Jones et al. 2007) and negatively influences the likelihood of retention (Sloot, Verhoef and Franses, 2005). Furthermore recovering from service failures is usually the responsibility of CRM or service managers. Also the vast majority of research describing mechanisms to offset the effects of a service failure and how to appropriately deal with them has been written with CRM in mind (for examples see: Hart, Heskett and Sasser, 1990; Mattila, 2006; Maxham and Netemyer, 2002b: Smith, Bolton and Wagner, 1999). So because the discussion of service failures is usually done from a CRM perspective, existing literature on service failures has been somewhat one‐sided. Especially since recent studies have shown that customers are remarkably forgiving if the service failure is made by a high equity brand (Brady et al. 2008). Literature has up until now shown only little interest in explaining the effects of brand building strategies on service failures. (notable exceptions: Brady et al. 2008; Roehm and Brady, 2007). Within the branding literature, this view of brand equity is, as noted by Keller and Lehmann (2006) understudied: “It could be argued that there has been somewhat of a preoccupation with brand extensions and some of the processes that lead to the development of brand equity. By contrast, there has been relatively limited effort directed toward exploring […] personal and social impacts of brands.” Research into this field would offer CRM managers new insights into why and how customers form their service evaluation and give them new tools to direct service recoveries to where they are most effective. It would give strong brands a new way to use their brand and weaker brands the incentive to build brand equity. Furthermore, it would bring branding and CRM managers closer together to help each other deal with service failures. Therefore goal of this thesis is to investigate how brand equity (e.g. brand perceptions) moderates the effect of the service failure and the level of recovery on satisfaction, intentions and behavior. This is the first large‐scale, longitudinal, non‐experimental study into this topic. This research contributes to both the brand and the CRM literature because it links the two literature streams empirically.
This thesis is organized as followed: In the first chapter I will discuss three relevant components of existing literature: CRM, using the satisfaction‐profit chain by Anderson and Mittal (2000). Service failures, where I will differentiate between single and multiple service failure(s) and service recovery. And branding, using a customer based view on brand equity. I will also show gaps in the literature and discuss the few papers that have combined CRM, service failures and branding. In the second chapter I will present a conceptual model to structure (my) further analysis. In the third chapter I will put forth my hypotheses and argumentation for these hypotheses. In the fourth chapter I will present a large‐scale longitudinal empirical study and a system of equations that I will use in the fifth chapter to test the hypotheses. The results of the analysis will be discussed along the line of the earlier mentioned satisfaction‐profit chain. The final chapters will deal with a discussion of the results, the implications, the limitations and directions for future research.
1.
Literature review
1.1. Customer Relationship Management
Previous CRM literature often focuses on the question of what drives customer retention (Bolton, 1998; Bolton and Lemon, 1999; Mittal and Kamakura, 2001; Verhoef, 2003; Gustafsson, Johnson and Roos, 2005). These and other studies into retention and churn have shown a wide range of factors that predict whether a customer stays or goes, like: affective commitment (Verhoef, 2003; Mattila, 2004; Gustafsson, Johnson and Roos, 2005), calculative commitment (Anderson and Weitz, 1992; Dwyer, Schurr and Oh, 1987; Heide and John, 1992), situational and reactional triggers (Gustafsson, Johnson and Roos, 2005), prior customer satisfaction (Mittal and Kamakura, 2001; Johnson, Herrmann and Huber, 2006), behavioral intentions (Zeithaml, Berry and Parasuraman, 1996) and attitudinal loyalty (Brady et al, 2008; Harris and Goode, 2004; Kumar and Shah, 2004, Mägi, 2003). Central to most of these studies are constructs of customer satisfaction, intentions and behavior. In their discussion of the satisfaction‐profit chain, Anderson and Mittal (2000) elaborate on the way satisfaction relates to intention. In their model satisfaction is based on performance. This satisfaction relates to retention in a positive, but often non‐linear way. And finaly, the relation between retention and profits also positive. This chain of effects is shown in figure 1.The first step in the chain is to create customer satisfaction. In a meta‐analysis of customer satisfaction,
Szymanski and Henard (2001) have identified antecedents of customer satisfaction in prior literature. They conclude that (disconfirmation of) expectations, performance, affect and equity are the main antecedents of customer satisfaction. Summarizing their results: Expectations and anticipations form satisfaction (LaTour and Peat, 1979; Oliver and DeSarbo, 1988). These anticipations are either met or not in the actual performance leading to either positive or negative disconfirmation (Oliver, 1980; Oliver, 1981; Oliver and DeSarbo, 1988). Furthermore, actual performance also has a direct effect on customer satisfaction. There is a well‐established base of authors modelling and proving a positive relationship between performance and satisfaction (Churchill and Surprenant, 1982; Hailstead, Hartman and Schmidt, 1994; Tse and Wilton, 1988). Szymanski and Henard also mention that customer satisfaction has an affective component. The argumentation for this is that “emotions elicited during consumption leave traces in the consumers’ memory and integrate into their satisfaction assessment”. The final antecedent of customer satisfaction Szymanski and Henard mention is
| 8
The second step in the chain is issue to more debate. Many different authors have investigated different views
about the way satisfaction relates to actual behavior (and profits). One way is to relate satisfaction directly to actual behavior (See arrow 1 in figure 1; e.g. Jones and Sasser, 1995; Mittal and Kamakura, 2001; Bolton, 1998). A second way is to include intentions before relating satisfaction to actual behavior (see arrows 2 and 3 in figure 1). Mittal and Kamakura (2001) investigate both views more in depth by relating satisfaction to both repurchase intent (arrow 2) and actual repurchase behavior (arrow 1). They find that the way satisfaction relates to repurchase intent is different from the way it relates to actual repurchase behavior and suggest looking at the effect of satisfaction on intentions before relating it to actual behavior.
So the second step in the chain is to relate satisfaction to intentions. One of the first papers investigating intentions (Zeithaml, Berry and Parasuraman, 1996) notes that intentions can be seen as “a promise of future behavior” and that they can be favorable or unfavorable leading to customers either defecting or not. The relationship between satisfaction and intentions has been shown to be positive but far from straightforward (Mittal and Kamakura, 2001). This relationship is susceptible to all kinds of non‐linearities caused by differences in customer characteristics. Examples of these customer characteristics could be different thresholds for repurchase depending on the age, income or education level (Cooil et al. 2007). For example, older customer might have the same satisfaction towards a brand or firm as younger customers, but a higher repurchase intention because they have had more experience with the brand or firm. Also, higher educated customers may be more likely to switch brands because it may be easier for them to find and compare different offerings available, even though they may be just as satisfied as customers with a lower education.
The third step is to relate intentions to behavior. Rust, Lemon and Zeithaml (2004) note that the most common
view of actual behavior used today is retention. The problem with retention however is that it doesn’t fully reflect “customers’ actual behavioral patterns”. From a retention perspective, defecting customers are “lost for good”. And upon repurchase, they are treated as new customers. Cooil et al. (2007) and Rust, Lemon and Zeithaml (2004) argue that it’s probably more realistic that customers form “serial monogamous or polygamous” relationships with firms they do business with. This means that customers come and go and that a firm captures only a portion of a customers’ business in a specific category. This has lead to “a growing popularity of the concept of share of wallet” (Zeithaml, 2000) and it captures actual behavior more accurately than retention (Cooil et al. 2007). The relation between intentions and actual behavior is usually characterized as fairly straightforward, linear and positive (Kumar and Shah, 2004; Zeithaml, Berry and Parasuraman, 1996).
A final addition to the satisfaction‐profit chain beyond the original paper by Anderson and Mittal (2000) is a
Using a critical incidents technique Bitner et al. (1990) were one of the first to take a closer look at service failures. Their conclusion was that the most common source for dissatisfaction was “employees inability or unwillingness to respond in service failure situations”. So it is the employee’s response to the failure (or lack thereof) that causes the dissatisfaction. In a study on customer switching behavior Keaveney (1995) shows that service failures, service recoveries and failed service recoveries explain up to 60% of ‘critical behavior’ that led to brand defects. When a service failure occurs customers suddenly find themselves in a situation where they are forced to re‐evaluate the relationship (van Doorn and Verhoef, 2008) and re‐determine whether or not they are satisfied, if they feel any loyalty towards a brand and whether or not they want to continue the relationship (Maxham III and Netemeyer, 2002a, b; Keaveney, 1995).
Service failure ‐ The direct effect of a service failure on customer satisfaction, loyalty and share of wallet is
intuitively negative. Theory suggest that the simple direct effect of a service failure is that it decreases customer evaluations (Bitner et al. 1990; Mattila, 2004; Maxham III and Netemeyer, 2002a). However, it has been shown in a study on unreported service failures that loyal customers complain less (Bolton, 1998). Or that very satisfied customers update their attitudes differently in the event of a service failure (van Doorn and Verhoef, 2008). In such situations satisfaction or loyalty could potentially increase after a service failure. Multiple service failures ‐ Research also shows that there is a difference between single and multiple service failures. “Many service relationships are ongoing and it is likely that more than one service failure will occur during the relationship” (Maxham III and Netemeyer, 2002a). And in the eyes of a customer experiencing more than one service failure, the failures might feel related even though they are caused by unrelated parts of a service system going wrong. Therefore, customers could see these sequential failures as one big failure, more or less independent of the time period (Maxham III and Netemeyer, 2002a). Also, more service failures mean more attributions of blame. So the direct effect of multiple service failures is greater than the effect of only one service failure (van Doorn and Verhoef, 2008). Service recovery ‐ The effect of service recovery on customer satisfaction, loyalty and share of wallet in general is positive. An excellent service recovery can increase satisfaction (Bitner, 1990; McCollough et al, 2000; Smith and Bolton, 1998) sometimes even beyond the pre‐failure levels. This is known as the service recovery paradox (McCollough et al, 2000; Smith and Bolton, 1998). In the context of multiple service failures Maxham and Netemeyer (2002a) suggest that if only one service failure occurs, an excellent service recovery restores customer attitudes to pre‐failure levels, and sometimes even beyond. But if multiple service failures occur, no matter how good the recovery attempt, it is impossible to restore customer attitudes to pre‐failure levels. But research shows that the majority of customers are dissatisfied with service recovery efforts (Tax, Brown and Chandrashekaran, 1998; Andreassen, 2001). This means that managers do not appear to offer the right compensation for the failures, even though there is plenty of literature available on the topic of recovery management. Examples of best practices are: acting quickly (Hart, Heskett and Sasser, 1990), offering an explanation (Mattila, 2006), fair treatment (Maxham and Netemyer, 2002b), effective complaint procedures (Tax, Brown and Chandrashekaran, 1998), fair comensation (Smith, Bolton and Wagner, 1999) and empowering front‐end employees to make amends (Tax and Brown, 1998).
1.3. Branding
| 10
The financial perspective on branding and brand equity has its roots in the merger and acquisitions boom of the
1980’s. Since then it became apparent the purchase price of a firm was largely reflected by the value of its brands (Leone et al, 2006). “The power of a brand lies in the minds of consumers and what they have experienced, learned and felt about the brand over time; brand equity can be thought of as the ‘added value’ endowed to a product in the thoughts, words and actions of consumers; there are many different ways […] the value of a brand can be manifested or exploited to benefit the firm” (Leone et al, 2006). During these transactions it became clear that brands formed one of the most important intangible assets of a firm. In the literature following these transactions, authors have tried to measure brand equity using accounting measures due to a need to explain brand equity to the financial market (Amir and Lev, 1996; Keller and Lehmann, 2006). Measures such as Tobin’s Q (Lindenberg and Ross, 1981), stock market responses (like: EVA or ROI) (Aaker and Jacobson, 1994, 2001; Lane and Jacobson, 1995; Mizik and Jacobson, 2003) and firm acquisitions (Mahajan et al. 1994). From this point of view, a brand is simply an asset that, like a plant or equipment, can easily be bought and sold.
The product perspective on branding and brand equity relates to the effect of brand extensions and price
sensitivity. “In economic terms brand equity can be seen as the degree of ‘market inefficiency’ that the firm is able to capture with its brands” (Erdem, 1998a, b; Keller and Lehmann, 2006). These market inefficiencies can be used to smoothen the introduction of brand extensions or justify price increases.
The customer perspective on branding relates to customer perceptions of a brand. It is the accumulated
experience of a customer with a brand over time (Keller, 1993; Keller and Aaker, 1998). This view has previously been linked to CRM literature (Blattberg and Deighton, 1996; Reinartz, Jacquelyn and Kumar, 2005; 2004; 2003; Rust, Zeithaml and Lemon, 2004) and will also be the view of this thesis.
Aaker (1991) has discussed the role of brand loyalty in creating brand equity. She concluded that creating brand loyalty has marketing advantages such as reduced marketing cost and better trade leverages. Furthermore, brand loyal customers are more resistant to competitive marketing strategies and create positive word of mouth (Dick and Basu, 1994). In elaborating on the work by Keller, Chaudhuri and Holbrook (2001) investigate the effects of brand trust and brand affect on brand loyalty. They find that brand trust and brand affect lead to purchase loyalty and attitudinal loyalty. Purchase loyalty increases market share and attitudinal loyalty decreases price sensitivity. So in summary, brand equity from the customers’ point of view might have a positive influence on customer satisfaction, loyalty and actual behavior.
1.4. Existing literature
Current literature streams do not provide a complete picture of how CPBE and customer relation management interact in the event of a service failure (see table 1). When discussing the effects of a service failure and the level of recovery, previous authors have identified a number of factors that influence its effect on customer evaluations of the service, such as one versus multiple service failures (Maxham III and Netemeyer, 2002a), the level of commitment (Ahluwalia, Unnava and Burnkrant, 2001) and cultural aspects (Mattila and Patterson, 2004). Only recently has (customer perceived) brand equity been added to this list. However these studies have thus far only been experimental in nature.Roehm and Brady (2007) looked at the interplay between brand equity and service failures to explain differences in customer responses. They find that in the event of a failure customers are dealing with questions about how to handle the failure and what the implications of the failure are. This coping process works as a temporary buffer because customers are too busy coping to update their attitudes towards the brand. If the firm can rectify the failure within this period high brand equity is sustained. But after the “coping effect” wears off, high equity brands are in a disadvantage. Customers expected more from the firm and the unresolved service failure decreases brand equity. So high brand equity buys a firm more time to organize their recovery attempt but if the failure is not addressed on time, the carefully build brand equity will start to deteriorate. Brady et al (2008) used a similar approach to see whether brand building can also be used as a strategy to offset service failures in a more durable way. They find that “high equity brands have more favorable satisfaction assessments and stronger behavioral intentions than low equity brand following the same failure scenario”. This means that when a failure occurs, having high brand equity creates an advantageous position that shields the firm from the full effect of the failure. Brady et al. (2008) acknowledge that a failure decreases attitudes, but this decrease in attitudes is smaller for high brand equity firms than for low brand equity firms. A second important finding is that this “buffering effect” of having high brand equity is present in all stages of the failure sequence, also after a recovery attempt. So even in a recovery, having high brand equity is an advantage.
| 12 Brady et al. (2008) acknowledge these differences in results, but argue that this is due to the fact that Mattila (2004) has used the more emotionally loaded construct of affective commitment, whereas they used the more balanced construct of brand equity that also has a rational component. This difference in findings is important to the discussion because it begs the question whether high CPBE is an advantage or a disadvantage in the event of a service failure. This research will address this question in three ways. First I will look into the effects of CPBE as a moderator in the relations between a service failure and satisfaction, intentions and behavior. This will show whether or not high CPBE offers an advantage in post‐ failure attitudes. Second, I will look into the durability of the moderating effect of CPBE as more service failures occur. So: is the moderating effect of CPBE persistent or does it deteriorate over time as more failures occur? Finally, I will investigate the moderating effect of CPBE on the relation between the level of service recovery and satisfaction, intentions and behavior. This will show if, and under what circumstances high CPBE is an advantage or disadvantage in post‐recovery attitudes.
2.
Conceptual model
| 14 FIGURE 2 Conceptual model CSt / CSt‐1 = Customer Satisfaction at time period tLOYt / LOYt‐1 = Loyalty intentions at time period t
SOWt / SOWt‐1 = Share of Wallet at time period t
I propose to use the conceptual model displayed in figure 2. As noted earlier, the satisfaction‐profit chain links satisfaction to actual behavior (e.g. Anderson and Mittal, 2000; Heskett et al. 1994; Keiningham and Zahorik, 1994). Customer satisfaction has a positive influence on loyalty intention and although previous authors have noted that is relation is often non‐linear (Anderson and Mittal, 2000; Mittal and Kamakura, 2001), this research will assume a linear relation for simplicity purposes. The nature of the relationship between customer satisfaction and loyalty intentions is not the main goal of this research. Furthermore, loyalty intentions and share of wallet also assume have a positive and linear relationship. Also, customer satisfaction, loyalty intentions and share of wallet are assumed to be influenced by carry‐over effects where previous evaluations influence current evaluations (van Doorn and Verhoef, 2008; Mittal, Kumar and Tsiros, 1999). This is displayed in grey in figure 2.
intentions and behavior (Dwyer, Schurr and Oh, 1987; Gustaffson, Johnson and Roos, 2005), increasing it too much destroys goodwill and decreases intentions and behavior in the long run (Jones et al. 2007).
Also influencing the chain are service failures and subsequent recoveries. Service failures decrease customer attitudes about satisfaction, intentions and behavior (Bitner et al. 1990; Mattila, 2004; Maxham III and Netemeyer, 2001a). An appropriate recovery attempt can, to some extent, restore these attitudes (Bitner et al. 1990; McCollough et al, 2000; Smith and Bolton, 1998).
Last are the effects of brand equity. Previous literature shows that CPBE has a positive influence on satisfaction, intentions and behavior (Aaker, 1991; Dick and Basu, 1994; Keller, 1993; Keller, 2008; Pappu and Quester, 2006). But relatively new to literature is the way CPBE influences the effect of service failures and subsequent recoveries. Brady et al. (2008) have shown that high brand equity can work as a buffer in the event of a service failure. Therefore CPBE is expected to moderate the effect of a service failure on satisfaction, intentions and behavior. A more in‐depth argumentation for this view is presented in the next chapter. CPBE also influences the way customers perceive a service recovery. Mattila (2004) has shown that emotionally bonded customers respond less positive to a recovery than customers who are not emotionally bonded. So I expect CPBE to also moderate the effect of a service recovery on satisfaction, intentions and behavior.
| 16
3.
Hypotheses
3.1. The moderating effect of CPBE on post failure attitudes in the
event of a service failure
As shown in the literature review, authors disagree on whether high CPBE should be an advantage or a disadvantage in the event of a service failure. The first view argues that high CPBE is a disadvantage based on expectancy‐disconfirmation theory and gap models of service quality (Parasuraman, Zeithaml and Berry, 1985; Oliver and Rust, 1997; Niedrich, Kiryanova and Black, 2005). These imply that a service failure is in disconfirmation with existing beliefs for customers with a positive emotional bond. These customers see a service failure as a violation of their positive attitudes towards the brand or firm, thereby amplifying the effect of a failure. For customers with no or a negative emotional bond, a service failure is a confirmation their attitudes, to some extent they might have even expected the failure to occur. This confirmation strengthens their negative perception. Furthermore Zeithaml (1993) shows that positive‐emotionally‐bonded customers have a more limited ‘zone of tolerance’ [for defects].
The other view on CPBE suggests that having high CPBE is an advantage (Brady et al. 2008). It causes customers to perceive a service failure as a deviation from typical performance (Bharadwaj, Varadarajan and Fahy, 1993; de Chernatony and Riley, 1999; Onkvisit and Shaw, 1989; Janiszewski and van Osselaer, 2000; Nowlis and Simonson, 1997; van Osselaer and Alba, 2000) and be more forgiving. In that case, CPBE can be seen as something that ‘softens the blow’. Theories supporting this view are for instance: the anchoring and adjustment model of belief updating (Smith and Bolton, 1998; Hogarth and Einhorn, 1992). This theory explains the effect of previous “favorable data points” that can work as an anchor in the evaluation that occurs after the failure. If high CPBE is seen as a favorable data point it can mitigate the effect of a service failure. The second theory that could offer an explanation is that of confirmatory bias (Darley and Gross, 1983; Hoch and Ha, 1986; Hogarth and Einhorn, 1992). This theory suggests that customer pays more attention to information that is consistent with their perceptions of existing performance leading to a situation where they simply counter argue negative information (Ahluwalia et al. 2000; Ahluwalia, Unnava and Burnkrant, 2001). In such a case, CPBE is a sort of barrier that keeps high CPBE customers from perceiving the full extent of a service failure because customer attach more value to information that confirms existing beliefs than information that disconfirms beliefs. Finally, the theory on the assimilation of positive benchmarks (Martin, 1986; Martin, Seta and Crelia, 1990; Meyers‐Levy and Sternthal, 1993; Meyers‐Levy and Tybout, 1997) suggests that existing performance benchmarks are easily mapped onto new performance. This is a mental process that saves cognitive energy by simply reapplying the previous performance judgment to the existing situation (Cohen and Basu, 1987) making the service failure feel less severe.
These two opposing views are difficult to unify because they both predict a different moderating effect of CPBE. But as Bitner et al. (1990) showed that only a limited number of failures are caused by employees’ intentional sabotage of the service, a service failure is hardly ever intentional. In this debate, attribution theory (Folkes, 1988; Weiner, 1989) can give better insights into which view explains the moderating effect better. In attribution theory a service failure could be characterized as “unstable attribution” (Folkes, 1988; Weiner, 1989). This means that customers might not attribute the full extent of the failure to the brand because they see the failure as a deviation from typical performance; a failure can happen. So I expect that the ‘soften the blow’ view offers a better explanation for the moderating effect of CPBE on the relation between service failures and satisfaction, loyalty and behavior. Therefore I pose the following hypotheses:
H1a: The negative effect of a service failure on customers’ post‐failure satisfaction is moderated by CPBE. A
H1b: The negative effect of a service failure on customers’ post‐failure loyalty intentions is moderated by
CPBE. A service failure is less severe for customers with high CPBE than for customers with low CPBE.
H1c: The negative effect of a service failure on customers’ post‐failure share of wallet is moderated by CPBE.
A service failure is less severe for customers with high CPBE than for customers with low CPBE..
3.2. The moderating effect of CPBE on postfailure attitudes in the
event of multiple service failures
There is ample evidence that multiple service failures cause a greater decrease in satisfaction, intentions or behavior; two or three failures are simply worse than only one failure. And in line with the previous paragraph, I expect that the mechanisms that explain the buffering effect of CPBE still influence customer perceptions when a customer is subjected to more failures. But with each sequential failure the moderating effect of CPBE decreases, either because a customers’ high regard for a brand can only stretch so far, or because customers might see multiple failures as one big failure (Maxham and Netemeyer, 2001a). Either way, high CPBE customers might have a tolerance for failures that low CPBE customers don’t have. But as Keller and Lehmann (2006) note: brand equity also has a performance component and if brands can’t deliver, brand equity deteriorates. So I expect that the moderating effect of CPBE on the relation between service failures and satisfaction, loyalty and behavior will deteriorate as more failures occur. Therefore I pose the following hypotheses:
H2a: The moderating effect of CPBE on customers’ post‐failure satisfaction is smaller in cases of multiple
service failures than with a single service failure.
H2b: The moderating effect of CPBE on the customers’ post‐failure loyalty intentions is smaller in cases of
multiple service failures than with a single service failure.
H2c: The moderating effect of CPBE on customers’ post‐failure share of wallet is smaller in cases of multiple
service failures than with a single service failure on share of wallet.
3.3. The moderating effect of CPBE on the relation between post
recovery attitudes and the level of service recovery
As with the discussion of service failures, the moderating effect of CPBE on service recovery can also be explained by two opposing views: high CPBE is either an advantage or a disadvantage.
The first view poses that CPBE doesn’t always have to be an advantage and it is based on the expectancy‐ disconfirmation theory (Parasuraman, Zeithaml and Berry, 1985; Oliver and Rust, 1997; Niedrich, Kiryanova and Black, 2005). Mattila (2004) has shown that a firm gains the most from a good service recovery with customers that have low affective commitment. This is because customers with low affective commitment experience positive disconfirmation: The service failure caused negative expectations, but the recovery attempt exceeds these expectations leading to a positive gap and increasing attitudes. Some authors suggest that this could lead to a service recovery paradox, where post‐recovery attitudes are higher than pre‐failure attitudes (Maxham III and Netemeyer, 2002a). Customers with high affective commitment experience a much smaller positive gap. Their negative expectations after a failure were mitigated by their higher CPBE and the effect of a recovery attempt is therefore much smaller.
| 18
a bad recovery is ignored. This effect would not be present in customers that perceive low CPBE, because a good service recovery is in disconfirmation of their existing beliefs and therefore ignored. And if they experience a poor service recovery it confirms their beliefs but does not increases their attitudes (Darley and Gross, 1983; Hoch and Ha, 1986; Hogarth and Einhorn, 1992).
Again attribution theory (Folkes, 1988; Weiner, 1989) can give better insights into which view explain the moderating effect of the level of service recovery best. In contrast to a service failure, which might not be intentional and is seen as a deviation from typical performance, a recovery attempt is always the full responsibility of a service provider. If this recovery attempt fails or doesn’t live up to expectations, the effect of this failure can be fully attributed to the service provider. This is the same for a good service recovery; all the credit is given to the service provider because it was their responsibility to solve the problem. This is also known as “stable attribution” (Folkes, 1988; Weiner, 1989). In this situation of “stable attribution” I therefore expect expectancy‐disconfirmation theory (Parasuraman, Zeithaml and Berry, 1985) to be a better predictor of the moderating effect of CPBE on the relation between the level of service recovery and customer satisfaction, intention and behavior. Therefore I pose the following hypotheses: H3a: The positive effect of a service recovery on customer satisfaction is moderated by CPBE. The effect of a recovery is weaker for customers with high CPBE then for customers with low CPBE.
H3b: The positive effect of a service recovery on loyalty intentions is moderated by CPBE. The effect of a
recovery is weaker for customers with high CPBE then for customers with low CPBE.
H3c: The positive effect of a service recovery on share of wallet is moderated by CPBE. The effect of a
recovery is weaker for customers with high CPBE then for customers with low CPBE.
4.
Methodology
4.1. Data collection
The sample used for this thesis consists of approximately 3000 customers of a large Dutch public transport provider. A mail survey was conducted under 10.000 customers in the period from 2007‐2008. Because of the longitudinal nature of this research the survey was conducted twice with a one‐year time span between surveys. In satisfaction literature a study design with two measurement points is fairly common (Mittal, Kumar and Tsiros, 1999; Verhoef, Franses and Donkers, 2002; van Doorn and Verhoef, 2008). And it has been shown in previous literature that the one‐year time span between the measurement points isn’t a problem in measuring carryover effects (Johnson, Herman and Huber, 2006). Mittal, Kumar and Tsiros (1999) even use a 21 month time period between measurements. In the first survey round (at the end of 2007) 5994 customers responded (a response rate 59,9%). In the second survey round (at the end of 2008) 3221 customers returned the questionnaire (a response rate of 32,2%). After removing customers that only took part in the first survey round and customers with too few responses in the second round a total sample of 3176 customers remained (a response rate of 31,8%). 62.2% of the sample was male and 37,8% was female. The age distribution of the sample is as followed: 3,8% were under the age of 18, 14,2% was between the age of 19 to 29 years, 11,0% was between the age of 30 to 39 years, 19,9% was between the age of 40 to 49 years, 28,6% was between the age of 50 to 59 years and 22,5% was older than 60 The respondents were asked to report on their perceived customer satisfaction, loyalty intentions, brand perceptions, their share of wallet in this specific market, switch cost, negative service encounters and recovery attempts (up to three) in the three months leading up to the questionnaire. All items were measured on a ten‐ point scale except share of wallet, which was measured as a % of spending. For a full listing of all the questions see table 2. Respondents didn’t receive a reward for filling out the questionnaire. The survey was part of a more broad study into customer behavior and perceptions at the transport provider.
4.2. Measures
In line with the method used by Mittal, Kumar and Tsiros (1999), customer satisfaction was measured by one question where customers were asked how satisfied they were on a 10‐point scale. The same was used for service recovery. It has been noted that the use of single item measures in large‐scale surveys may result in models that display estimated relationships softer than they actually are (Mittal, Kumar and Tsiros, 1999). On the other hand, these measures have been successfully used in other large‐scale surveys (Bolton and Drew, 1991; Mittal, Ross and Baldasare, 1998; Mittal, Kumar and Tsiros, 1999). The obvious advantage of using single‐ item measures in large‐scale longitudinal studies is the effect it has on the response rate. Using single‐item measures reduces the time it takes for a respondent to fill the questionnaire. This increased response rate may offset the decrease in reliability (LaBarbera and Mazursky, 1983).
Service failures were measured by asking for the total number of service failures that have occurred in the three months leading up to the survey. This data was then dummy coded in a way that each respondent had experienced either no, one or multiple service failures.
the scale. It showed two underlying factors: one related to the availability of alternatives and one relating to price. These two new constructs were also tested for internal consistency resulting in α = 0.354 for the alternatives construct and α = 0.473 for the price construct. So combining the questions into one single construct has a higher α then treating them as separate.
TABLE 2
Scale overview with source literature
Construct Source Measure
Brand equity “Brand X is a strong brand” “Brand X is a well known brand” ”Brand X is an attractive brand” “Brand X is a unique brand” “Brand X is a smart choice”
Verhoef, Langerak and Donkers, 2007. Customer satisfaction “How satisfied are you with the service offered by company X?” Mittal, Kumar and Tsiros, 1999 Loyalty intentions “I tell positive things about company X to people in my surroundings” “I recommend company X to family, friends and colleagues” “Company X is my first choice” “I am certainly going to use company X’ services again in the future”
Zeithaml, Berry and Parasuraman, 1996. Share of Wallet “% of business with company X” (“% of business with competitor Y”) (“% of business with competitor Z”) … Cooil et al. 2007
Switch Costs “I remain a customer with company X because … … I don’t have any alternative transportation” … it’s easiest for me”
… alternatives are too expensive”
... I can’t cancel my subscription on short notice” … someone else is paying for my subscription”
Verhoef, Franses and Hoekstra, 2002. (Calculative commitment) Service failure “Have you experienced anything particularly negative with company X in the past three months?” “How long ago did the negative experience happen?” “Was company X responsible for the experience?” (based on Mayor and Russel, 1982). “How serious did you perceive the experience?”
Service recovery “How good were you helped by company X to restore the failure?”
Smith, Bolton and Wagner, 1999.
(adapted from Bitner and Hubbert (1994) and Oliver and Swan (1989a, b)
4.3. Modeling
To test the conceptual model, I specified a system of equations. These models describe how current evaluations of the service are based on past evaluations, service failures and service recoveries. The variables that measure service failure are dummy coded in such a way that a respondent i either has had no service failure, only one service failure or more than one service failure at time period t. By multiplying these variables with CPBE at time period t-1, it is possible to estimate the moderating effect of CPBE (Baron and Kenny, 1986). Furthermore, the direct effects of service failures and the level of service recovery are also included along with the direct effect of CPBE in all the models. For the second model the direct effect of current customer satisfaction and switch costs are also included and for the third model the direct effect of current loyalty and current switch costs are ta n intke o account.
5.
Analysis and Results
Model fit ‐ The results of the 3SLS regression are displayed in table 5. For customer satisfaction χ2df = 2974 =
2471,34 (p < 0,0001), for loyalty intentions χ2(df = 2972) = 3697,88 (p < 0,0001) and for share of wallet χ2df = 2972 =
1584,73 (p < 0,0001). Therefore the models are all highly significant. However, the good χ2 statistics could also
be explained by the size of the sample and the resulting high number of degrees of freedom.
To assess the models performance, I also estimated a rival model without the moderation effects to compare it with the used model. The results of this analysis are displayed in table 4. These fit statistics indicate a slight improvement if the moderating effect of CPBE is included.
Model stability ‐ The model’s stability is assessed with 10 random holdout samples, each containing a random
selection of 75% of the respondents and running the model on these samples (Bolton, Lemon and Verhoef, 2008; van Doorn and Verhoef, 2008). After comparing the different models with the original model, it was found that 87% of the used variables in all three models have the same significance level as the original model. Therefore, it can be concluded that the used model is stable. TABLE 5 Comparing fit statistics of a model with and without moderation effects model
Used model Non‐moderation R2 Customer satisfaction 0,4537 0,4487 Loyalty intentions 0,5749 0,5745 Share of Wallet 0,3567 0,3560 RMSE Customer satisfaction 0,7604 0,7639 Loyalty intentions 1,0442 1,0448 Share of Wallet 26,9453 26,9578 AIC 43615,64 43644,99
5.1. Customer Satisfaction
From the models estimation results of customer satisfaction it is clear that service failures have a large impact on customer satisfaction, just as previous customer satisfaction. It is also striking that the direct effect of CPBE on customer satisfaction is fairly limited. Keller and Lehmann (2006) offer an explanation for this in that good performance is a necessity for high CPBE. Performance is therefore more important than CPBE. Furthermore, the direct effect of the level of service recovery is, as expected, positive. Interestingly, this model shows that, ceteris paribus, a good service recovery could potentially compensate for multiple service failures. Although Maxham III and Netemeyer (2002a) have shown that multiple failure make the service recovery paradox impossible, this model shows that even in the event of multiple service failures, companies can still recover and even increase customer satisfaction beyond pre‐failure levels with an excellent service recovery. Also interesting is the moderating effect of CPBE on both the relations between a service failure and satisfaction and the level of recovery and satisfaction. The moderating effect for a failure is positive and the for a recovery it is negative. So the moderating effects are opposite to the direct effects. The method for testing moderation effects is discussed by Baron and Kenny (1986). They state that in looking for moderation effects, the direct effect of both the predictor (service failure and service recovery) and the moderator variable (brand equity) have to be regressed on the outcome variable along with the product of the predictor and moderator (the interaction term). If linear moderation is assumed significance of the interaction term in regression analysis is enough to prove moderation. So there is support for both the H1a and H3a hypothesis (PSingleSF < 0,01;TABLE 6
Estimation results of the 3SLS moderated model
Customer Satisfaction Loyalty Intentions Share of Wallet Coefficients CSt 0,2164 *** CSt‐1 0,4518*** LOYt 5,3249*** LOYt‐1 0,5328 *** SoWt‐1 0,4494*** Brandt‐1 0,0912 *** 0,0818*** ‐1,4663*** Switcht 0,0805*** 2,8353*** Brandt‐1 x SingleSF 0,1067*** 0,1164*** ‐0,0219*** Brandt‐1 x MultiSF 0,1514*** 0,0777*** ‐0,7756*** Brandt‐1 x SR ‐0,0296*** ‐0,0128*** ‐0,2442*** SingleSF ‐0,9764*** ‐0,8249*** 3,774*** MultiSF ‐1,6301*** ‐0,7383*** 14,7327*** SR 0,2404*** 0,1052*** 1,3574*** constant 3,4806*** 0,7207*** ‐9,4262*** * Significant at the 10% interval ** Significant at the 5% interval *** Significant at the 1% interval
Finding support for these hypotheses means that customers with high CPBE are less susceptible to service failures and their customer satisfaction is fairly stable. On the other hand, this also means that these high CPBE customers are less susceptible to good service recoveries and that there is much less to gain with these customers. For low CPBE customers satisfaction is less stable. Service failures cause a substantial decrease in satisfaction levels and a good recovery attempt can restore satisfaction just as easily. The effects of this model are graphically illustrated in figure 3. In this figure, previous customer satisfaction (CSt‐ 1) was fixed at the sample average (6,8) and I simulated the effect of high and low CPBE in the event of a single or multiple service failures. High and low CPBE are set at 8 and 2 respectively. Figure 3 shows that high CPBE customers have a fairly stable satisfaction rating. Their satisfaction rating is above the sample average because they are simply more satisfied (illustrated by the “High BE / No SF” dotted line), just as low CPBE customers are less satisfied (illustrated by the “Low BE / No SF” dotted line). Figure 3 also shows that only one service failure marginally decreases customer satisfaction, but when multiple service failures occur the drop in satisfaction is much more pronounced. Also, no matter how good the service recovery attempt, it is impossible to restore satisfaction to pre‐failure levels for high CPBE customers.
In contrast with the fairly stable high CPBE customer, the low CPBE customers are much more susceptible to a good service recovery. If multiple failures occur a good recovery attempt is enough to restore satisfaction levels. If only one service failure occurs satisfaction levels are restored even with a mediocre recovery attempt. Even more interesting is the effect of an excellent recovery: with an excellent recovery is possible to even surpass satisfaction ratings of that of high CPBE customers. So with regard to the service recovery paradox, this research shows that a service recovery paradox is only possible with low CPBE customers.
FIGURE 3 Customer Satisfaction with high/low CPBE and Single/Multiple Service failures High CPBE Low CPBE 5,5 6,0 6,5 7,0 7,5 8,0 1 2 3 4 5 6 7 8 9 10 Cu st o m e r Sati sf ac ti on Level of service recovery Average Customer Satisfaction High BE / No SF High BE / Single SF High BE / Multiple SF 5,5 6,0 6,5 7,0 7,5 8,0 1 2 3 4 5 6 7 8 9 10 Cu st o m e r Sati sf ac ti on Level of service recovery Average Customer Satisfaction Low BE / No SF Low BE / Single SF Low BE / Multiple SF Finally, hypothesis H2a is based on the premises that the coefficient for a single service failure is larger than the coefficient for multiple service failures. For H2a it is already clear that the hypothesis will not hold because the
coefficient for a single service failure is already smaller than the one for multiple service failures. Further significance testing also shows that the difference is also non‐significant (χ2(df = 1) = 2,25; P > 0.05), This means
that there is no support for the H2a hypothesis and there is no significant difference in the buffering effect of
CPBE for single or multiple service failures.
5.2. Loyalty intentions
Just as in the formation of customer satisfaction, service failures play a large role for loyalty intentions. Customer satisfaction also plays a role and there is also a carryover effect of previous loyalty intentions. Remarkable here is the non‐significant effect of service recovery. Customers somehow do take service failures into account when (re)evaluating their intended loyalty, but disregard the recovery attempt.
An explanation could lay in the fact that customers sometimes place less value on positive information because they regard it as less helpful or less diagnostic (Ahluwalia, 2002). Asymmetric disconfirmation theory, that suggests that negative information has a greater influence than positive information (Mittal, Ross and Baldasare, 1998) and prospect theory that suggests that losses are weighed more heavily than gains (Kahneman and Tversky, 1979). This is known as the negativity effect (Ahluwalia, 2002). And compared to customer satisfaction, loyalty intentions are more long‐term and forward looking. They require customers to think about their future behavior. This calls for a certain amount of diagnostic information processing. So the, less diagnostic, positive experiences are more easily disregarded in the (re)evaluation of intentions.
There is support for the H1b hypothesis, so CPBE significantly moderates (PSingleSF < 0,05; PMultiSF < 0,1) the
FIGURE 4 Loyalty Intentions with high/low CPBE and Single/Multiple Service failures High CPBE Low CPBE 5,0 5,5 6,0 6,5 7,0 7,5 1 2 3 4 5 6 7 8 9 10 Lo yal ty In te n tio n s Level of service recovery Average Loyalty Intentions High BE / No SF High BE / Single SF High BE / Multiple SF 5,0 5,5 6,0 6,5 7,0 7,5 1 2 3 4 5 6 7 8 9 10 Lo yal ty In te n tio n s Level of service recovery Average Loyalty Intentions Low BE / No SF Low BE / Single SF Low BE / Multiple SF (χ2(df = 1) = 1,73; P > 0.05). This shows that there is also no support for H2b, meaning that there is no difference between the moderated single and moderated multiple service failure coefficients. Furthermore, along with the non‐significance of service recoveries, the moderation of a service recovery is also non‐significant and there is no support for H3b. Figure 4 graphically illustrates the estimation effects for loyalty intentions. Previous loyalty intentions (LOYt‐1) were assumed to be at the sample average (6,6) and again the
effects of high and low CPBE are set at 8 and 2 respectively. But, as can be seen in figure 4, the effect of a service recovery does indirectly affect loyalty intentions through customer satisfaction. From figure 4 it is clear that the advantage of low CPBE has dissipated. Still, customers with low CPBE react more strongly to recovery attempts, but the effect is not strong enough to restore loyalty intentions to pre‐failure levels. Figure 4 also shows that high CPBE customers that have experienced one service failure have stronger loyalty intentions than those that did not experience a service failure. This counterintuitive result implies that CPBE influences service failures in such a way that the moderation effect strengthen intentions more than the direct effect of a service failure weakens them. The effect could even be described as a ‘service failure paradox’. A somewhat similar effect has been discussed by van Doorn and Verhoef (2008). They show that there is an stronger updating process where customers that experience a service failure update their attitudes differently. Before the failure these customers are in a business as usual state. But after a failure they reevaluate their attitudes, and decide to stay. In some way they are more convinced of the relationship than before the failure occurred. This effect is however short‐lived if failures keep occurring, as can be seen from the “High BE / Multiple SF” line.
5.3. Share of Wallet
Important determinants of a customer’s share of wallet are loyalty intentions, switch costs and customer’s previous share of wallet. Even though the coefficient for SoWt‐1 is much lower than that of for example loyalty
intentions, SoWt‐1 is measured as a percentage of spending and the impact is therefore comparable.
FIGURE 5 Share of Wallet with high/low CPBE and Single/Multiple Service Failures High CPBE Low CPBE 50% 52% 54% 56% 58% 60% 62% 1 2 3 4 5 6 7 8 9 10 Sh are of W a llet Level of service recovery Average Share of Wallet High BE / No SF High BE / Single SF High BE / Multiple SF 50% 52% 54% 56% 58% 60% 62% 1 2 3 4 5 6 7 8 9 10 Sh are of W a llet Level of service recovery Average Share of Wallet Low BE / No SF Low BE / Single SF Low BE / Multiple SF Furthermore, there is no impact of service failures and the level of service recovery on the share of wallet. This holds for both the direct and the moderating effect. So H1c, H2c and H3c can therefore not be supported. The
non significance of service failures and the level of recovery can firstly be explained by the fact that the firm at which the research was done has to some extent a monopoly on the service. Customers use the service because they need to, and to a lesser extent because they want to. Another reason for the minimal effect of service failures and recoveries is that a large percentage of the respondents (56,2%) have someone else paying for their use of the service for example employers. Not having to pay for using the service might disconnect usage and satisfaction. Secondly, the big influence of switch costs on share of wallet might indicate that alternatives are expensive or unavailable, so customers are locked into a relationship with the firm. Finally the result from the model estimation show that CPBE has a significant negative impact on the share of wallet formation (P < 0,05). This counter intuitive result means that customers with high CPBE have a lower share of wallet, as is clearly shown in figure 5. Further examination shows that less frequent users are more satisfied, are more loyal, perceive higher CPBE and have a lower share of wallet than more frequent users. An explanation for this can be found in the number of service failures customer groups with certain usage frequencies experience. More frequent users simply have a higher chance of experiencing a service failure and adjust their satisfaction, loyalty intentions and CPBE accordingly.
Figure 5 is a graphical illustration of the model estimations for the share of wallet. Previous share of wallet (SoWt‐1) is assumed to be at the sample average (57,9%) and high and low CPBE are set at 8 and 2 respectively.
| 28 adjust their behavior upwards. This is again in line with the service failure paradox argumentation presented in the previous paragraph. The pre‐failure relationship is in a business as usual state. But after the failure, these customers re‐evaluate their relationship towards the service provider and decide to stay. However, if multiple failures occur, they readjust their behavior downward again.
6.
Discussion
One of the most important findings of this research is that brand building strategies can be used to mitigate the effect of service failures on satisfaction and loyalty intentions. This research has shown that customers with high CPBE respond differently to failures than customer with low CPBE. The results found in this study support the findings from earlier experimental research (Brady et al, 2008) in that CPBE mitigates the effects of service failures. Brady et al. (2008) discuss the possibility of brand managers and CRM managers helping each other to establish a strong brand in the minds of customers as a buffer for service failures. This doesn’t mean that high equity brand can do as they please and they are unaffected by service failures. As this research has shown, a service failure significantly decreases satisfaction and loyalty intentions. Brand building purely as a recovery strategy is therefore unadvisable. High equity brands are merely in an advantageous position when it comes to service failures.A second finding is that there is no significant difference in the way CPBE moderates the effects of either single or multiple service failures. This is in contrast to results found by Roehm and Brady (2007). They state that the only way to sustain the effect of a carefully build brand is to respond quickly to failure and that CPBE slowly deteriorates as customers are waiting for a appropriate service recovery. However, results from this study show that the effect of CPBE does not deteriorate when more failures occur; The buffering effect of CPBE is the same for a single failure as for multiple failures and the effect is fairly durable. All the time and money spent in carefully building a brand are not wasted in the event of a failure or poor recovery. But still, it is not advisable to keep customers waiting too long after a service failure. While the effect of CPBE doesn’t, satisfaction and loyalty quickly deteriorate while customers are waiting (Hart, Heskett and Sasser, 1990). The third conclusion from this research is that in some cases a failure is not directly negative. There seems to be a sort of service failure paradox that affects customers’ intentions. Van Doorn and Verhoef (2008) have already shown that “critical incidents can have a positive effect on customer behavior among satisfied customers who have maintained a high share with the supplier in the past.” This research builds on this in that not only “having a high share in the past” can be beneficial in the event of a service failure. Brand building can create the somewhat similar effect. This is a big incentive for managers to build brand equity.