Managing service recovery processes: the role of customers' age

Research and practice show that effective management of service recovery processes boosts customer satisfaction. Under this assumption, the purpose of this paper is to analyze a set of factors which may determine satisfaction with recovery processes and loyalty. We also analyze the role of age as potential moderating. Segmenting customers' samples by age may potentially contribute to more effective service recovery process management. Older customers seem to be more loyal when dealing with service providers than younger customers, while younger customers are more demanding in terms of companies' efforts. Implications for both literature and practice are included at the front-end of the paper.


Introduction
In the current highly competitive environment, consumers become more demanding and expect the excellence of any service provider (Tsai, Su 2009). The existence of multiple options becoming available to the consumer raises consumer's expectations and sometimes customers perceive a low quality service.
As a result, more and more companies are focusing on improving the service itself as well as establishing the channels to successfully handle any service failure that may happen with the objective of enhancing customer service (Salavou 2010). Additionally, market research is becoming an important tool for identifying and assessing both existing and potential consumer needs and expectations, with the objective of designing better services and enhancing personnel training methods.
However, even when very carefully designing any service, mistakes are unavoidable and even the best companies have to deal with them (Chang, Hsiao 2008;DeWitt et al. 2008;Huang 2008;Michel, Meuter 2008;Varela et al. 2008a). Such mistakes have a negative impact on consumer perception and consequently, they can affect customer's satisfaction levels (Michel, Meuter 2008). Several authors, Bitner et al. (1990), Mc-Collough and Bharadwaj (1992), Zeithaml et al. (1996) and Varela et al. (2008a), have pointed out that an effective service failure management and sometimes a timely solution can successfully restore customer satisfaction (service recovery paradox). This phenomenon has been widely studied in the service marketing area and has become of great interest for both academia and business, due to the enormous pressure most industries are facing (Maxham, Netemeyer 2002). However, literature still suggests a gap related with some consequences of service recovery processes, not about satisfaction, but about loyalty and customer-rm relationships.
As previously mentioned, several authors (e.g., Michel, Meuter 2008;Varela et al. 2008a) have studied the impact of service recovery processes on customer satisfaction levels. In particular, Tsai and Su (2009) and Varela et al. (2009) among others, nd that good service recovery processes decrease the likelihood of customer switching to the competence. Additionally, those processes are considered antecedents to customer loyalty and the relational approach (e.g., DeWitt et al. 2008;Chang, Hsiao 2008;Ok et al. 2007). In fact, they show that the effective management of service related complaints stimulates good long term relationships with customers and provides valuable feedback for those companies that wish to improve their service. On the other hand, poor service recovery processes may lead to even higher levels of dissatisfaction. Even though there has been an increasing interest on this area, several gaps in the literature still exist (Dewitt et al. 2008). Some of the main questions that still need to be addressed are i) how justice is perceived and its impact on loyalty, and ii) individual and joint consideration of attitudinal and behavioral loyalty, which would contribute to a fuller understanding of service recovery measure effectiveness. Huang (2008) explores the concept of customer satisfaction but does not present any empirical proof on the direct link between satisfaction and loyalty. He postulates that the scope of the failure, the client perception of staff efforts, and the expectations with respect to recovery are antecedents to the level of satisfaction of recovery when the service is eventually restored. He concludes the paper suggesting the interest on a wider study of the behavioral rami cations of such satisfaction. In the same line of research, Michel and Meuter (2008) focus their research on satisfaction levels after the service recovery process. They admit that working with different scales derived from one item has its limitations and the study should be repeated using multi-item scales.
Some studies (e.g., Verhoef 2003;Homburg, Giering 2001;Mittal, Kamakura 2001) point out that consumers' age tends to display divergent behavior patterns and therefore consumers perceive satisfaction differently. Yet, literature speci cally analyzing the role of age in satisfaction models is hard to come by in services marketing research. Shanin and Chan (2006) and Pai-Lin et al. (2001) are rare exceptions. In the case of service recovery research, we were not able to nd explicit evidence of previous research on this area. Therefore the work presented in this paper tries to ll in this gap in the literature. In such a context, this study aims to contribute to both, academia and business, segments by analyzing some consequences of service recovery processes, not only about satisfaction -as prior studies do but more importantly, about loyalty and customer-rm interactions and relationships. Therefore, this paper complements the existing literature and reach the following speci c objectives: i) identifying and exploring the impact of perceived satisfaction antecedents on service recovery, ii) analyzing the link between satisfaction with service recovery processes and loyalty, iii) studying the potential moderating role of the age variable in service recovery processes, and iv) re ecting on implications for both business practice and the literature. To the extent that we achieve these goals we will be contributing to lling in the gaps mentioned above.
The paper is structured as follows. Section 2 presents the literature review taking as key reference the service recovery processes literature. Section 3 presents the model of reference based on a set of hypothesis. Section 4 justi es the sector of reference for the empirical study. Sections 5 and 6 are related with the empirical study and results. The nal section of the paper discusses our contribution to literature and practice as well as presents the main conclusions of this research. Michel and Meuter (2008) and Maxham (2001) de ne service failure as a real or perceived setback or problem that occurs during customer-company interaction. Similarly, service recovery is de ned as the post-failure measures taken by the service provider aimed at resolving the service failure (Bitner et al. 1990;Grönroos 1998). Both Chang and Hsiao (2008) and Smith et al. (1999) make a distinction between complaint management and service recovery. These authors agree that in both cases, we are dealing with a reaction to a service related problem on the company's side; however, while complaint management is reactive in nature, service recovery is proactive; that is, the aim is not only to successfully deal with the complaint but also to provide a satisfactory, client oriented solution as quickly as possible.

Literature review
We now introduce the service recovery paradox (SRP) that refers to a seemingly illogical situation where -following a failure / recovery process -higher levels of customer satisfaction are achieved than in the case of customers who have not experienced any service failure (Varela et al. 2008a;Michel, Mauter 2008;McCollough, Bharadwaj 1992). Therefore, a successful service recovery process becomes a catalyst for customer satisfaction which may lead to positive word-of-mouth boosting long term relationships with clients and enhancing company-client relationships (Tsai, Su 2009). This idea has been widely studied in De Matos et al. (2007), Magnini et al. (2007), Maxham (2001), Bolton (1998), andZeithaml et al. (1996), among others.
However, we nd, after an exhaustive review of the literature, that ndings are not conclusive with respect to service recovery satisfaction levels. In recent years, authors such as Magnini et al. (2007), Hocutt et al. (2006), and Maxham and Netemeyer (2002) have aligned themselves with previous research in recognizing the positive impact of recovery procedures; on the opposite side we nd scholars such as Ok et al. (2007), Andreassen (2001), McCollough (2000, and Hocutt et al. (1997) that do not recognize such effect. Michel and Meunter (2008) point out the existence of two factors that might explain these two opposite lines of research. On one hand, the lack of consensus with respect to the recovery de nition: some studies compare clients who have experienced service failure with those who have not experienced it, while other studies compare the same client before and after a failure is experienced. In the present study we assume that clients who complain do so because their expectations have not been met. On the other hand, recovery processes are somehow scarce because only a small number of dissatised clients actually register a complaint, making it dif cult to nd a signi cant sample of clients that have received satisfactory recovery. In order to overcome this potential problem we decided to use a specialized company to obtain the empirical data for our research (see the empirical research section).
The conceptual basis for this study falls under the larger theoretical framework of SRP. However, we suppose that customers make complaints when expectations are not achieved and therefore we only analyze the perceived satisfaction with service recovery processes. The next step is to identify factors which might pave the way for satisfactory recovery.

Antecedents
In our literature research we have found key articles that consider critical variables to analyze service recovery processes. These factors that have already been analyzed in the literature are the following: intent to complain (Hocutt et al. 1997), company's image (Bontis et al. 2007;Andreassen 2001), trust (Kau, Loh 2006), service quality (McCollough 1995), and switching intentions (Varela et al. 2009;Zeithaml et al. 1996). Huang (2008) andDe Matos et al. (2007) consider service failure severity, recovery expectations, company responsibility, and perceived interest/effort when it comes to resolving service related problems, which is the approach we follow in our research (de nitions of variables and their impact on satisfaction with service recovery processes are provided in section 3).

Consequences
The literature identi es perceived satisfaction and repurchase intent as key variables for the study of the recovery process (e.g., DeWitt et al. 2008;Michel, Meuter 2008;De Matos et al. 2007). However, up to date, both variables have been studied in isolation and only a small percentage of studies, in our opinion, have grounded ndings in empirical evidence-despite implicitly recognizing the link between customer satisfaction and service loyalty / intent to repurchase.
In this line, the work carried out by DeWitt et al. (2008) revolves around the notions of justice, and loyalty-both attitudinal and behavioral (de nitions and interactions are also provided in section 3).

Moderators
Additionally, the vast majority of the studies cited recognize the potential impact of moderating factors such as type of service / industry or the demographic pro le of the sample. Key studies for understanding notions such as perceived satisfaction and customer retention strategy (e.g., Shanin, Chan 2006;Verhoef 2003;Homburg, Giering 2001;Mittal, Kamakura 2001) point out that certain variables -including gender, age, education and income level -can alter projected behavior patterns. Such research suggests that older customers are more responsible, conscientious and have more stable preferences than younger consumers and that is the reason why their loyalty and repurchase probability is higher. Meanwhile, younger customers tend to be more emotional, energetic and do not rely that strongly on their satisfaction with pure objective parameters but with more subjective evaluations than older customers. Therefore, younger customers may be more demanding in terms of effort and expectations than older customers, while the effect of justice on satisfaction and the probability of loyalty may be higher in the segment of older people than in the younger people. Moreover, reasons for different patterns of behaviour between ages may be related with psychological, cultural and other contextual factors (e.g., Shahin, Chan 2006, Bravo et al. 2008. The value of such ndings has been corroborated in marketing circles by Homburg and Giering (2001), Jones et al. (2001), Mittal and Kamakura (2001) and Bolton (1998), among others. However literature calls for renewed research efforts taking into account this variable in service recovery models.
A review of the literature, then, reveals research gaps which substantiate the present study. It follows -as authors such as DeWitt et al. (2008) -have recently suggested, that more empirical evidence of the nexus linking perceived justice, service recovery related satisfaction and customer loyalty is needed. Our study also acquires added value by looking at a speci c industry and context which has yet to be analyzed -despite its economic impact: the mobile phone sector in Spain.

Hypotheses
Based on the ideas and arguments presented in the previous sections we submit that customer satisfaction is a key factor in understanding the customer's perspective following a service recovery episode. The literature suggests, moreover, that the happier customers are, the more likely they are to maintain a long lasting relationship with the company and therefore we can integrate our approach in the relationship marketing paradigm. We would venture, then, that customer perception of company effort, expectations for recovery, the severity of the service failure, and perceived justice underlie and explain the level of customer satisfaction following service recovery measures. Client loyalty -both attitudinal and behavioral -can be understood as a function of the level of customer satisfaction. Our reference model can be found in Figure 1, followed by justi cation of our hypotheses. We also consider the moderating effect of age in the set of hypothesis.
The literature de nes perceived effort as customer perception of the energy and resources that the company has dedicated to solving the problem (Huang 2008;De Matos et al. 2007;Guenzi, Pelloni 2004). It would be logical, therefore, to assume that customer appraisal following a service recovery encounter (perceived satisfaction) is a function of perceived effort. However, we have to assume that differences between customers and employees viewpoints happened (Asghar, Rostamy 2009). A case in point is the client who, having perceived sincere concern and effort on the part of the company, -despite the lack of a satisfactory solution -demonstrates something close to what can be termed as satisfaction (Mohr, Bitner 1995). Our rst hypothesis, then, is: H1: The greater the perceived effort, the greater the perceived satisfaction following the service recovery process.
However, Mittal and Kamakura (2001), Homburg and Giering (2001) and Jones et al. (2001) seem to suggest that older clients tend to be less demanding in this regard, which suggests that perceived effort is a more relevant factor among younger customers: H1 A : Younger customers are more susceptible than older customers to perceived effort and its impact on post-recovery customer satisfaction levels.
The notion of expectations is paramount in the world of marketing, explaining the wealth of literature on the subject (e.g., Armstrong et al. 2009;Grönroos 1998 As the term applies to service recovery contexts, authors such as Ok et al. (2007), Hess et al. (2003), and Swanson and Kelley (2001) point out that recovery expectations are linked to customers' hopes of obtaining an appropriate, satisfactory solution to a given problem. Logically, higher expectations make for more demanding clients; in other words, both direct and inverse expectations-satisfaction relationships are to be expected (Huang 2008;Wirtz, Mattila 2004). This brings us to our second hypothesis:

H2:
The higher the client's expectations are with regard to service recovery, the lower the level of perceived satisfaction. In this regard, Verhoef (2003) and Mittal and Kamakura (2001) suggest that younger clients are, for the main part, more demanding than older clients and therefore tend to have lower expectations for recovery. If this is the case, we can accept that old customers have lower recovery expectations than the young ones: H2 A : Younger customers are more susceptible than older customers to the impact service recovery related expectations have on post recovery perceived satisfaction levels.
Service failure severity is de ned as the magnitude of the loss suffered by consumers due to a negative episode (Huang 2008). Loss of this nature can be triggered by tangible factors such as monetary damage, or by intangible aspects like anger or frustration. The literature suggests that the greater the magnitude of the service failure, the more dif cult it is to satisfy the client via service recovery strategies (Magnini et al. 2007;Mattila 1999;Smith, Bolton 1998;McCollough 1995). Thus, both the service recovery process and the perception of the nal result are conditioned by the magnitude of the failure to the extent that the greater the severity, the lower the level of satisfaction perceived by the client (Hoffman et al. 1995). This phenomenon is due to the fact that customers tend to perceive less justice in recovery strategies and nal outcomes as the magnitude of the failure augments (Huang 2008). In short, a severe service failure can deal a fatal blow to customer satisfaction levels.

H3:
The greater the magnitude of the service failure, the lower the level of customer satisfaction with regard to service recovery.
Additionally, the greater objectivity of the older segment respect to younger customers (e.g., Homburg, Giering 2001;Pai-Lin et al. 2001) may reinforce the effect of this link. Therefore we propose the following: H3 A : Older customers are more susceptible than their younger counterparts to the impact of service failure severity -with respect to service recovery -on perceived satisfaction levels with regard to service recovery processes.
From the perspective of justice theory, customers are asked to evaluate solutions obtained via the service recovery process in terms of fair or unfair (DeWitt et al. 2008;Chang, Hsiao 2008). In this sense, justice can be linked both to the customer-company interaction phase and to the nal outcome of the recovery process itself (Maxham, Netemeyer 2002;Tax et al. 1998). In this way when the customer receives satisfactory treatment-and, above all, recovery-high levels of perceived justice and a satisfactory end result are obtained. This idea is present in Chang and Hsiao (2008), and Varela et al. (2008a), who explicitly endorse the idea that effective service recovery boosts perceived justice and contributes towards maximizing customer satisfaction.

H4:
The higher the level of perceived justice, the greater the level of satisfaction perceived by the client throughout the service recovery process.
In this grain, authors like Mittal and Kamakura (2001) and Pai-Lin et al. (2001) suggest that older people tend to display a more highly developed sense of justice. Hence H4 A is: H4 A : Older customers are more susceptible than younger customers to perceived justice and its impact on post-recovery customer satisfaction levels. Relationship marketing postulates that satisfaction is fundamental to effective client retention (Gustafsson et al. 2005;Kim et al. 2004;Gummensson 1997;Morgan, Hunt 1994;Dwyer et al. 1987). A happy client is one whose expectations have been met and who believes, therefore, that the company in question will be able to deliver down the line (Santala, Parvinen 2007). Hence, we can assume that a satis ed client will be a loyal client (Tamosiuniene, Jasilioniene 2007). Customer loyalty can be de ned, then, as the client's commitment to repurchasing products from the same brand or company in the future, determined by both an attitudinal and a behavioral component (Oliver 1999). Attitudinal loyalty is linked to client propensity to commit to a given brand / company and must not-as Shankar et al. (2003) point out-be limited exclusively to repurchase behavior. Positive word-of-mouth, for instance, can also be considered an example of this type of customer loyalty (Bontis et al. 2007). Varela et al. (2009) suggest that the higher the level of perceived satisfaction with regard to service recovery, the lower the chances are that a client will abandon a service provider. In such a context, we can assume that the probability of repurchase/attitudinal loyalty will increase (DeWitt et al. 2008). Based on the previous arguments we propose the following set of hypothesis:

H5:
The higher the level of perceived satisfaction, the greater the degree of attitudinal loyalty displayed by the customer. H6: The higher the level of perceived satisfaction, the greater the degree of behavioral loyalty displayed by the customer. H7: Attitudinal loyalty towards a brand/company and behavioral loyalty towards a brand / company are directly proportional. However, speci c research into consumer behavior (e.g., Capraro et al. 2003;Homburg, Giering 2001;Mittal, Kamakura 2001;Pai-Lin et al. 2001) leads one to believe that older people are more prone to be loyal while younger customer -though they may also be loyal -are more likely to respond to emotional and short term results. This brings us to our last set of hypotheses: H5 A : Older customers are more susceptible than their younger counterparts to the impact of perceived satisfaction with service recovery efforts on attitudinal loyalty. H6 A : Older customers are more susceptible than younger customers to the impact of perceived satisfaction with service recovery efforts on behavioral loyalty. H7 A : The impact of attitudinal loyalty vis-à-vis behavioral loyalty affects older customers more than it does younger customers.

The mobile phone sector in Spain
We test our hypothesis in the Spanish mobile phone sector. This industry will serve as our framework for analyzing the signi cance of the proposed model. Telecommunications is currently the most aggressively competitive sector in Spain ) -and one of the most affected by globalization of services, due to the high rate of technological development. Moreover, the ever increasing range of choices enhances consumer decision making power (Maícas, Sesé 2008). Up until 2004, as Rivero and Manera (2005) point out in a recent study, mobile operators were still able to nd small pockets of potential clients who had not yet been tapped. Market saturation became the reality from that point on up to date, and capturing clients from the competition has become the only route to growth.
Yet another of the strategies of choice, aimed at maintaining activity levels within the mobile phone sector, has been the quest for new applications, models and 'limited time offers'. The result: a 4% increase in business across the industry in 2009, peaking at 52.9 million counting both company and individual clients -the equivalent of 114.6 lines per one hundred inhabitants (Juste 2010).
Such a competitive framework underscores the importance of a deep understanding of the client -of knowing exactly what customers want and expect in order to effectively position oneself in the market (Maícas et al. 2009;). Apparently, however, marketing strategies have become much more aggressive, as our pilot study revealed; a whole slew of more attractive, 'new and improved' products which, paradoxically, are relatively distant from the original market need and 'raison d'être'. Pricing strategies and 'limited time offers' appear to be aimed more at capturing new clients than fostering real customer loyalty. As a result, in the Spanish mobile phone sector a) there is very little difference between one mobile operator and another in terms of the services they offer; b) new clients are harder and harder to come by; c) the skyrocketing cost of capturing new clients means it takes longer and longer to recuperate the initial investment; d) ambitious sales objectives have driven many mobile companies to come up with strategies to draw clients away from competitors -at any cost; e) it is an industry known for high rotation and quick client turnover. In the authors' opinion, such ndings call for fostering more conservative strategies, perhaps, in order to really guarantee customer satisfaction and loyalty; to this end, strategies linked to relationship marketing and the service recovery paradox might be our best bet. In any case, we should underline that the speci c characteristics of the industry (e.g., price, commercial campaigns, accessories, etc.) may play a major role in both company's choice and service recovery perceptions.
Moreover, despite the research by Hur et al. (2010), Sesé (2009), Wieringa et al. (2007), and Lee et al. (2006) pointing out the value of the mobile phone industry as a point of reference for empirical research in the context of service marketing, up to date it has not been considered in the light of the service recovery paradox.

Introduction
Our pilot study revealed that, on average, 25% of interviewed customers had experienced some sort of a problem with their mobile operator at one time or another. Only 16%, however, had led a complaint and a mere 5% of the clients who had complained felt their problem had been resolved satisfactorily. Such ndings suggest that i) Spanish mobile operators have a long way to go when it comes to effective complaint management; and ii) collecting data of this sort is no easy task. With this in mind, the decision was made to engage a data collection service; our inclusion criteria required that survey participants be legal adults who had experienced some sort of service related problem with their mobile provider, led a formal complaint and received a response from the company in question.
The eldwork for our study was carried out in November and December, 2009; 202 surveys were compiled. All pertinent technical details can be found in Table 1.
In order to analyse the moderating role of age in service recovery processes, we split the sample in two groups (customers up to 25 years old and customers from 26 years old) following Shahin and Chan (2006), Bravo et al. (2008) and Varela et al. (2008b). Arguments to defend this split are based on lifestyles changes that happened when people are around 25 years old: different pattern of behaviour may be related with psychological, cultural and other contextual factor such as emancipation and incorporation to the labour market. We used the scales proposed by Huang (2008) to gauge perceived effort, service error severity, recovery expectations, and post-recovery satisfaction. For our assessment of perceived justice and customer loyalty (attitudinal and behavioral) we opted in favor of the scales put forth by DeWitt et al. (2008). Prior to distributing the nal survey we circulated a pretest we had eshed out in collaboration with colleagues from Marketing departments at several different universities, PhD candidates, and a small sample of potential interviewees. With the context under scrutiny in mind, pertinent reliability and validity tests were run for all proposed scales -even in cases where the scale in question had previously been tested in earlier studies. The scales that were eventually selected have been included in Appendix 1 for easy reference.
We worked with a Partial Least Squares (PLS) structural equations analysis technique to evaluate the measurement model and signi cance of the hypotheses. PLS-Graph version 03.00 build 1017 (Chin, Frye 2003) was the software of choice.

Measurement model
It should be noted here that one of the constructsperceived justice -is made operable via a molecular approach; this makes it a second-level factor which is the cause of its rst-level components or factors (Chin, Gopal 1995). Thus, it was essential to apply the approach in two phases -also referred to as hierarchical components analysis (HCA) (Lohmöller 1989;Chin, Gopal 1995). We should note here as well that perceived justice is a second-level construct which is measured using three rst-level factors: distributive justice, interactive justice and procedural justice.
With regard to our measurement model, we began by assessing the reliability of individual items. The indicators for all three samples are above the accepted 0.707 benchmark established by Carmines and Zeller (1979), as seen in Table 2. Only one item was below the accepted benchmark: If another mobile provider offered lower prices or special discounts, I would make the change (ACT L3); this item was excluded from all three samples.
In the case of construct reliability, the measurement scale of choice was composite reliability ( c ) (Werts et al. 1974). Careful scrutiny of the ndings in Appendix 1 shows all constructs in all dimensions to be reliable across the three samples: indicator values above 0.8 (Nunnally 1978).
When it came to assessing convergent validity, we turned to the average variance extracted (AVE) scale proposed by Fornell and Larcker (1981). Given that the 0.5 benchmark these authors establish is below the AVE for the different constructs/dimensions, we can af rm that convergent validity exists (see Appendix 1).
The presence of discriminant validity has been con rmed using AVE (Fornell, Larcker 1981), comparing the square root of this measurement with the correlations among constructs. Discriminant validity is present in all samples, as seen in Appendix 2.

Structural model
Following this analysis of our measurement model an assessment of the signi cance of the hypotheses proposed in the structural model is in order. It should be noted that PLS does not require that data derive from normal, or known, distributions -which explains why traditional parameter estimation techniques for testing model signi cance are considered inappropriate (Chin 1998). Yet another difference between covariance based structural equation models and PLS is that, in the latter, goodness-of-t measures are not called for (Hulland 1999). As seen in Table 2, the structural model is assessed i) using the variance value from the model (R²), and ii) considering the size of the standardized path coef cients ( ) after observing both the t values and the signi cance level obtained from the bootstrap test with 500 subsamples.
With respect to the antecedent variables for post recovery satisfaction (see Table 2 for the total sample and subsamples), we should note that neither customer expectations (H 2 ) nor service failure severity (H 3 ) have a signi cant impact on perceived satisfaction levels. On the other hand, the relationships expressed by hypotheses H 1 and H 4 -links between perceived effort and justice, and customer satisfaction -were established: in the total sample (0.330; p < 0.001 and 0.530; p < 0.001); and in both subsamples ( YC 1 = 0.349; p < 0.001 and OC = 0.328; p < 0.001) and ( YC = 0.486; p < 0.001 and OC = 0.552; p < 0.001). Table 2. Results for the structural model (total sample and younger / older customers subsamples)

Impact on endogenous variables
Total sample (N = 201)

Path coef cients ( ) T value (bootstrap)
Impact on post-service recovery satisfaction (SAT) The impact of customer satisfaction on attitudinal loyalty has been fully veri ed. On the one hand the relationship proposed in hypothesis H 5 with respect to the total sample has proven true ( = 0.538; p < 0.001); on the other hand, this relationship has been established for both subsamples ( YC = 0.525, p < 0.001 and OC = 0.545; p < 0.001).
Differences were detected between samples, however, in the case of behavioral loyalty. The relationship expressed by hypothesis H 6 -the link between customer satisfaction and behavioral loyalty -has been established as true for the total sample ( = 0.241; p < 0.001), as well as for the younger customers subsample ( YC = 0.453; p < 0.01). Satisfaction and behavioral loyalty could not be linked for older customers, however.
1 From now on: YC = Younger Customers; OC = Older Customers.
Finally, the proposed direct, positive relationship linking attitudinal and behavioral loyalty proved true for all three samples ( = 0.589; p < 0.001; YC = 0.494; p < 0.001 and OC = 0.646; p < 0.001). With regard to the explained variance of the endogenous variables (R 2 ), our research model proved to be suf ciently predictive; ndings were consistent across samples, as seen in Table 4.

Analysis of the moderating effect of the age variable
In order to contrast the moderating role of age in the model, the path coef cients between the variables (see Table 2) must be compared. Yet, questions may arise regarding whether differences among the segments obtained for each variable -re ecting the nature of the relationship -are substantial enough to warrant behavioral differences in function of age. One statistical procedure designed to verify the signi cance of these comparisons (in which a t-test is run) is the multigroup analysis 2 put forth by Chin (2000) and employed by Keil et al. (2000).
In short, for signi cant relationships, the identi ed segments suitably distinguish between different predicting variables and their dependent variables, as seen in Table 3. The intensity of the relationships proposed in hypothesis H 1A is greater for younger customers than for older customers ( YC > OC , p < 0.1); among younger customers, the impact of perceived effort on satisfaction is greater than among older customers. In the case of hypotheses H 4A , H 5A and H 7A on the contrary, our study demonstrates that justice has a greater impact on satisfaction among older customers than it does among younger customers ( OC > YC , p < 0.001). Likewise, we have established that the impact customer satisfaction has on attitudinal loyalty, and attitudinal loyalty on behavioural loyalty are more substantial among older customers. Finally, the link between satisfaction and behavioral loyalty (H 6A ) is not signi cant for the older customers subsample, while the same relationship is clearly signi cant for the younger customers subsample.
The Tippins and Sohi (2003) approach was adopted in order to test the moderating impact of attitudinal loyalty (ACT L) on post-recovery satisfaction (SAT) and behavioral loyalty (BEH L). This approach recommends an analysis of competing models in which two substantive models are gauged and evaluated for signi cant differences. In the rst model, the direct relationship linking SAT and BEH L is explored; in the second, the same relationship is examined, this time with the ACT L in a moderating role. In the case of both, the total sample and the two subsamples, the model which includes ACT L in a moderating role explains more BEH L variance than the other model. There is a positive correlation between SAT and ACT L across all samples. In the case of SAT and BEH L, on the contrary, there is deviation among samples (see Table 4).
In the case of the older customers subsample there is no direct relationship linking these variables -unlike with the total and younger customers subsamples. Finally, attitudinal loyalty (ACT L) displays a direct, positive relation with respect to behavioral loyalty (BEH L) across all three samples.
The signi cant relationship linking SAT and BEH L in the direct impact model diminishes in importance across all three samples in the mediation model.
In light of these ndings, we can af rm that ACT L plays a moderating role for SAT and BEH L. That said, this mediation is only partial for both the total and the young customers subsamples. In the case of the older customers subsample, however, mediation is unmitigated since the SAT-BEH L relationship is insigni cant in the model which includes the direct link between SAT and BEH L. Table 4 presents data corresponding to our calculation of the total impact (direct and indirect) on BEH L. We used the test proposed by Sobel (1982) to calculate the sig-   Table 4, the moderating impact of ACT L on SAT and BEH L is con rmed by the z statistic, with a value of p < 0.001 across all three samples. The magnitude of indirect impact on the total is derived from the variance accounted for (VAF) put forth by Iacobucci and Duhachek (2003). In the total sample, 56.8% of the total impact of SAT on BEH L is due to indirect impact, climbing to 73.49% in the case of older customers and dropping to approximately 36.4% for younger customers.

Discussion
Our research con rms the value of designing and executing effective service recovery strategies given their impact on perceived customer satisfaction levels. This analysis supports some pioneering research by Bitner et al. (1990), Zeithaml et al. (1996), and more recent studies by authors such as DeWitt et al. (2008) and Michel and Meuter (2008), among others. Not even the best companies are immune to making mistakes; this is something that the vast majority of consumers tend to understand. Consequently, complaint management and service recovery strategies clearly must lead to a reasonable solution if the company wishes to restate customer satisfaction and keep its image intact. In this line of thought, successful recovery from a service failure can translate into enhanced customer satisfaction, con dence and loyalty, as long as the company has effectively shown its ability to solve the problem. Our data suggests that service failure severity and expectations for recovery do not have a signi cant impact on perceived satisfaction (H 2 and H 3 ). Such results partially contradict literature. A reasonable explanation could be that the greater the range of choices, the more entitled the customer feels to receive satisfactory service from the get-go; customers expect a satisfactory solution -regardless of the magnitude of the service failure. In other words, the severity of the failure is not important and higher expectations can put a damper on otherwise positive service recovery outcomes. On the other hand, customer perceptions regarding company efforts to deal with problems (H 1 ) do have a direct and signi cant impact on customer satisfaction with service recovery. Existing studies (e.g., Huang 2008;Mohr, Bitner 1995) suggest that customers value the interest and effort companies invest in solving problems. However, we should consider that differences between customers and employees viewpoints may occur (Asghar, Rostamy 2009). It has even been pointed out that if real, sincere desire is perceived, customer satisfaction will exist even if a solution does not. The implications are clear: when service failure occurs or customers express dissatisfaction, the company should make an effort to get to the bottom of the problem and provide a solution, while making sure the client is well aware that the company is taking the corresponding steps in the right direction. Moreover, we have found that higher expectations among younger customers make perceived effort on the part of the company the essence for this segment; such ndings are in line with ideas of authors such as Mittal and Kamakura (2001), Homburg and Giering (2001) and Jones et al. (2001).
Our fourth hypothesis (H 4 ) proposes a direct relationship between perceived justice and post service recovery satisfaction. The data suggests that such a link exists, con rming some of the ideas of authors like DeWitt et al. (2008), Chang and Hsiao (2008), Maxham and Netemeyer (2002) and Tax et al. (1998). Therefore, we can assume that customers should be treated fairly and receive solutions in line with their perception of justice. In this regard, -and given that perceptions with respect to justice can vary notably between companies and clients-it would be a good idea for companies to invest in getting to know what customers expect, and what they consider fair, in order to adapt to their needs or, at the very least, help them understand that the solution provided is the most appropriate for the given problem. Additionally, our research shows that older customers are more likely to perceive justice both in the service recovery process and in the nal result than younger customers.
The rst recommendation we can make with regard to good management practices is that companies should make efforts to ensure that services are delivered well the rst time around. This, as we have indicated throughout, is much easier said than done. Whichever route is eventually taken, the company should react quickly i) to understand underlying factors and ii), to communicate with the customer. This shows the company's desire to nd a satisfactory solution to the problem. Moreover, it is essential that rms explain how the error occurred and what is being done to deal with it. For that purpose, the company must have staff capable of attending clients in a friendly, effective way. We have demonstrated that customer-company interaction and perceived effort have a clear impact on satisfaction in service recovery contexts. Hence, investing in selection, training and staff motivation activities is essential -as it is also attention to internal marketing fundamentals.
If companies wish to learn from mistakes and enhance future service they must view client complaints as both vital feedback and a management challenge that must be met.
Simplifying the ling process, or even adopting a proactive attitude -in other words, not waiting until a complaint is led before asking customers whether they are satis ed with service or not-can be interpreted positively as showing interest and investing in problem solving. In addition, such processes can yield feedback from customers regarding speci c aspects of the service(s) provided.
Lastly, this study is set out to nd a nexus between Huang's 2008 study, focused on customer satisfaction, and recent research by Michel and Meuter (2008) which takes a closer look at attitudinal and behavioral loyalty. In this regard, -taking the tenets of relationship marketing as a point of reference -our ndings for H 5 , H 6 and H 7 show a link between customer satisfaction and loyalty, and suggest that we are looking at a key relational tool with the potential to make switching costly for the client. In this line of thought, Varela et al. (2009) suggest that positive service recovery processes decrease the chances of clients jumping ship (switching). Polo and Sesé (2009), though implicitly, corroborate this thesis.
In many cases customer-company interaction could be channeled to cultivate longer lasting relationships. Knowledge gleaned from this type of feedback would equip rms to tailor services more speci cally to present needs and future expectations, while affording clients a better grasp on a service provider's actual capacity to react in the face of service failure. If the customer does eventually abandon the relationship, the company should be interested in the motive(s) and try to learn from each and every client that falls by the wayside.
Our initial multi-sample analysis -later corroborated by our impact study of moderating effects -establishes that, in service recovery contexts (as in other situations) older customers are more likely to build long term relationships and exhibit a greater degree of attitudinal and behavioral loyalty than younger customers. However, younger customers are more demanding in terms of effort than older customers. Because data suggest different patterns between age segments, we can therefore propose segmenting the customers' database and then adapting speci c recovery strategies depending on the customers' age.
All of these ndings could be harnessed by mobile companies to enhance customercompany interaction and streamline the use of available resources. Curiously, however, the customers we interviewed in both our pilot and main studies perceived a lack of interest on the part of companies when it came to problem solving and, in many cases, service recovery efforts were unsatisfactory. This situation may allow us to anticipate de cient service evaluations, bad word-of-mouth and a lackluster company image. A common behavior pattern emerges, however, among Spanish mobile providers; nearly identical policies put consumers in a position where, despite a sense of empowerment in their choice of the service provider itself, they feel abandoned and defenseless in the face of real or potential service related problems. This perception fades when consumers are bombarded with a wealth of offers for new services or to migrate to another company. Yet, most wonder why the initial interest in pleasing them disappears once they have signed up for the service. This is the reality of the mobile phone market in Spain. Given the scenario, who will be brave enough to take the rst step in this direction? Apparently, inquiring clients want to know and our research suggests that the rst company really interested on customers will get the market leadership.

Conclusions
This research corroborates the positive impact of service recovery efforts in customers' satisfaction. Our data also suggest that satisfaction with service recovery processes anticipate customers' loyalty towards the rm. This study also analyzes the moderating role of the "age" variable in service recovery scenarios. In this sense, our study contributes to narrowing the gap identi ed in the rst section of the article.
However, despite the inherent interest of the study, it is clearly not without its limitations. For one, only the Spanish mobile phone sector has been analyzed; a sector which is representative but which exhibits signi cant idiosyncrasies. The literature called for exploring sectors which had yet received little attention from researchers and our data is, for the most part, in line with results reported in previous studies. Even so, one must be cautious when extrapolating ndings across sectors: an analysis of potential structural / conjunctural similarities and differences would be in order. In this case we should recognize that speci c characteristics of the sector (e.g., price, promotional campaigns, accessories as a present, etc.) could play a major role not only in company choice but also in the service recovery outcomes perceptions.
Secondly, this is a cross-sectional study based on the opinions expressed by customers themselves. It would be interesting to carry out a longitudinal analysis of the entire complaint process, followed by an objective assessment of service provider solutions and nal outcomes. From a practical standpoint, however, getting involved in customercompany interaction can be an extremely complex endeavor; and, after all, the key to the service recovery paradox lies in customers' perceptions of how they are treated and to what extent their problems are, or are not, resolved.
Finally, with regard to potential lines for future research, it would be interesting to analyze the moderating effect of other consumer pro le variables such as profession, income and education. An international study comparing mobile company behavior patterns and customer perceptions might justify adopting a relational approach -which customers in our pilot study sample showed an interest in -vis-à-vis maintaining the more aggressive approach which, for the moment, seems to prevail in Spain.

Measurement scale items, individual reliability, composite reliability and variance extracted (total sample and subsamples)
Construct  Notes: a The data forming the diagonal line in bold corresponds to the square root of the average variance extracted (AVE) for the construct, while the rest of the numbers represent correlations between constructs, b All correlations are signi cant for p < 0.01