The Relationship between Service Recovery and Patronage Intentions: The Mediating Role of Relationship Quality

This study aims to investigate, using the structural equation modeling with a partial least square (PLS-SEM) approach, the relationship among service recovery strategies, relationship quality, and customer patronage intentions concerning Internet Service Providers (ISPs) in Egypt. Our findings confirmed that some service recovery strategies positively influence on both relationship quality and patronage intentions. As well as, relationship quality has a significant impact on customer patronage intentions. Furthermore, relationship quality has the role of a mediator of the relationship between SR strategies and patronage intention. Moreover, both speeds of recovery and empathy moderate the relationship between some SR strategies and relationship quality. Finally, our study proposed some theoretical and managerial implications.


Introduction
Services depend on a high level of interactions between providers and their users.Several authors have considered that this tight relationship increases the chance of a service failure (Hart, Heskett and Sasser, 1989;De-Ruyter and Wetzels, 2000;Chueh, Wang, and Liao, 2014;Komunda and Osarenkhoe, 2012).If service providers' deal with failures in the wrong way they can incur in some harmful consequences as it can drive customers away produce negative WOM, and reduced profits (Tax, Brown and Chandrashekaran, 1998;Bitner, Brown and Meuter, 2000;Tronvoll, 2010).How companies respond to service failure has been found to be a significant factor in both customer attrition (Park and Park, 2016), and customer switching behavior (McCollough, Berry and Yadav, 2000).Mittal, Ross, and Baldasare (1998) hold that a service organization negative performance in responding to service failure have a greater impact on consumer's satisfaction and on their intentions to further purchases compared to a positive performance of the service in the normal situations.
At the same time, service providers can benefit from responding to a service failure in a way customers find satisfying (Najjar, Smith and Kettinger, 2010).Accordingly, Bell and Zemke, (1987) believed that the occurrence of service failure provides a company with a real chance to rebuild consumers' trust and commitment and satisfy them.Therefore, the effective service recovery is one of the main keys to gain an advantage over the competitors, and it has become one of the most important topics in service quality literature (Michel, Bowen and Johnston, 2009).Wong, Hung, and Chow (2007) confirmed that long-term customer relationships could be considered as one of the most important assets a business organization can have.Despite this, many service companies overlook the sustainable competitive advantage that can be generated through their stable long-term relationships with customers (Morgan and Hunt, 1994;Linares, 2012: p.50), which allows service firms to reduce the negative results related to the unavoidable negative service experiences (Linares, 2012: p.39).In addition, the more positive is the relationship between service provider and its customers, the greater is the customer's likelihood to stay loyal to it and recommend its services to others (Gittell, 2002).Therefore, one of the main objective of marketing relationship management is to strengthen the link with existing customers in order to enhance customer satisfaction (Claycomb and Martin, 2002) and to build customer loyalty through the formation of mutual long-term benefits between service firms and its customers (Kanyan, Andrew, Ali and Beti, 2015).
Even if customer patronage behavior is surely not a new topic in service marketing literature (Tagashira and Minami, 2016) it is still considered a central strategy for being successful in service industries (Darian, Wiman, and Tucci, 2005;Ganesh, Reynolds and Luckett, 2007;Zolfagharian and Paswan, 2009).Moreover, it represents a major concern for managers because it enables them to identify and target those consumers most likely to purchase (Pan andZinkhan, 2006, Patney, 2010).
Information and Communications Technology (ICT) is considered one of the core components of a knowledge economy, as well as its role in supporting economic and social development.In the case of Egypt, ICT sector contributes 10.2% to the Egyptian economic growth, as well as its valuable contribution to the Egyptian GDP (economic indicators of the Egyptian economy's annual and quarterly, the Ministry of Planning, June 2016).It is a noticeable increase in the number of internet Egyptian customers, as the ratio of ADSL subscribers was increased in January-March 2016 compared to January-March 2015 by 22.5% (Information and Communications Technology Indicators Bulletin, March 2016).These elements have driven the choice of the sector we have focused on in this paper.Smith (2007) suggests that the effectiveness of various service recovery strategies changes in different settings.So that the service recovery is a contextual phenomenon, needed to be tested in different environments.Moreover, according to Lin (2006), the lack of control on all the service-delivering processes and the need to interact with different customers make services more prone to failures, errors, and complaints than manufacturing products.The interactions open the door to miscommunication, increasing the probability of a service failure.
According to the national telecommunication regulatory authority (NTRA) reports (website: www.tra.gov.eg), the telecommunications sector is more prone to service failure than many others as shown by the relative increase in the number of customers' complaints, as well as the relative delay in dealing with them by some Internet service providers.The reports show that the number of complaints filed by customers is increasing for most companies.In addition, the various providers have not shown a standard period of handling the c omplaint.This evidence encourages us to explore the perception of ADSL internet customers regarding the provider's service recovery efforts and its relation with relationship quality and patronage intentions in the eastern communities.
This study seeks to explore the role of SR strategies on the quality of the relationship between the service provider and its customers, and how both of them can help to understand customers' patronage intention.In this, we have focused on the internet sector, as these factors have not been tested before in this context.Previous studies have looked into the relationship of service recovery strategies and consumer's re-patronage intentions, and the one between SR strategies and the service recovery satisfaction.In this paper, we advance the existing knowledge complementing these traditional approaches using a broader definition of SR strategies and examining its influence on relationship quality as a whole and on patronage intentions.

Service Recovery (SR)
British Airways used service recovery in 1989 to help focus front-line staff attention on the customer (Li-hua, 2006).According to Zemke and Bell (1990), service recovery is "a thought-out process for returning aggravated customers to a state of satisfaction with the company after a service has failed to exceed the customer's expectations".SR contains all strategies, actions, and tactics a service provider undertakes to respond to a service failure, in order to satisfy customers and retain them (Andreassen, 2000;Miller, Craighead, and Karwan, 2000;Petzer, Steyn and Mostert 2009).It encompasses all the corrective actions to respond to a poor quality service (Gronroos, 1988).According to Boshoff (1999), it is effective only when a company adopts a proactive approach going beyond re-establishing the "normal state".
Service recovery (SR) has been often approached, in service marketing studies, as the results of a combination of strategies the company can adopt in order to deal with service failure.In the following lines, we report the strategies and techniques of service recovery that taken into account: The first strategy in an SR process is to make an apology (Zemke and Bell, 1990).Apologies let the customer know that the company acknowledges its responsibility for the failure (Ennew and Schoefer, 2003;Liao, 2007).Apologies help in creating an empathic link between company and its customers (Liao, 2007).Some authors (Battaglia, Borchardt, Afonso Sellitto and Medeiros Pereira, 2012;Levesque and McDougall, 2000) suggested that its effectiveness depends on the severity of the failure, and on the willingness to provide some form of compensation.
The second strategy is problem solving, often seen as the main recovery strategy.It restores the service in its normal state.It also shows the provider's skills in dealing with the failure causes, and in planning and implementing corrective actions (Battaglia et al., 2012).Problem solving should be fast (Parasuraman, Berry and Zeithaml, 1991) and effective (Levesque and McDougall, 2000) as if it fails, it will damage the relationship with the customer.
A third strategy involves explaining the causes of the failure openly.It refers to an explanation of the causes of occurrence the service problem in an obvious and specific way (Conlon and Murray, 1996;Boshoff, 1999).In other words, explain what might have caused the service failure (Liao, 2007).Therefore, it must not be an attempt or a way for the service providers to place the blame on someone else, because it may create negative outcomes (Liao, 2007).
The fourth strategy is about giving some kind of value-added compensation to restore the service value proposition (Boshoff, 1999).It often involves giving back money to the customer and it may be required when the failure could not be fixed (Levesque and McDougall, 2000), other times it comes in the form of free access to future services free and/or discounts (Ennew and Schoefer, 2003).
A fifth strategy to deal with SR is the follow-up.Following-up happens when the company contacts the consumers after recovery efforts to know if he/she has been satisfied with the recovery process itself (Ennew and Schoefer, 2003).Moreover, it refers to the situation in which the service firm provides its consumers with all information needed about his/her complaint and what is being done to resolve it (Boshoff, 1999;Boshoff, Hoogendoorn and van der Kraan, 2005).
Two other factors impact on relationship quality: speed of recovery, and empathy.The speed of recovery can be seen as a way to lessen the customer's difficulties (Battaglia, et al., 2012).Johnston and Clark (2005) hold that finding, and fixing, failures was one of the roots of SR.A delay in response increases the customer resentment and it negatively affects service quality perception (Boshoff, 1999).An emphatic answer shows that the company cares for the specific interest of each single customer (Zeithaml, Parasuraman, and Berry, 1990, p. 26) and that it tries to reduce his/hers inconvenience (Boshoff, 1999).According to Liao (2007), SR itself asks for an empathizing employee's behavior in dealing with affected customers; employees should understand the customer perspective (Krishna, Dangayach, and Jain, 2011;Battaglia et al., 2012).

Relationship Quality (RQ)
Relationship quality is the degree in which the relationship may help consumers in fulfilling some of their needs (Hennig-Thurau and Klee, 1997;p. 752).It also refers to the consumer's overall assessment of the relationship strength and the extent to which it fulfills the consumer's expectations based on their experience with the business firm (Crosby, Evans, Cowles, 1990;Smith, 1998, p. 78).Relationship quality literature confirmed that it is a multi-dimensional construct, which consists of three interrelated constructs namely customer satisfaction, trust, and commitment that may help to reduce consumer's uncertainty feelings concerning its relation with the service provider (Morgan and Hunt, 1994;Dorsch, Swanson, Kelley, 1998;Woo and Ennew, 2004;Vesel and Zabkar, 2010;Doaei, Rezaei and Khajei, 2011;Kim Jung-Eun and Lee, 2012).
Customer satisfaction is usually considered as a central element in the relationship marketing literature.Generally, satisfaction pointed to the customer's fulfillment response, it is a judgment that a product or service characteristics, or the product or service itself, may be used to reach a satisfying level of consumption-related fulfillment (Oliver, 1997, p. 13).In 1981 (p. 27), Oliver defines customer satisfaction as "Summary emotional state resulting when the emotion surrounding disconfirmed expectations is coupled with the consumer's prior feelings about the consumption experience" (Andreassen, 2000).Choi and La (2013), suggest that service recovery satisfaction (SRS), sometimes referred to as the complaint handling (Tax et al. 1998;Orsingher, Valentini and De Angelis, 2010), is the customer's assessment of how a service failure handled by the service provider.It is also described as the degree to which the consumer has promising or satisfactory feelings toward the provider recovery activities (Chang, 2004).So that, SRS is a critical issue because of its role in maintaining and strengthening the relationship between service provider and its customers (Fornell and Wernerfelt, 1987).
Customer Trust is the second pillar in this relationship quality.Anderson and Weitz, (1989) defined it as "one party's belief that its needs will be fulfilled in the future by actions undertaken by the other party" (p.312).Moorman, Deshpande, and Zaltman (1993) argue that trust is the consumer's belief or attitude toward the service provider integrity, they included in trust both attitudinal (belief, trustworthiness, confidence) and behavioral intention (reliance).Therefore, Moorman et al. (1993) define customer trust as "a willingness to rely on an exchange partner in whom one has confidence"(p.315).Morgan and Hunt (1994) define it as "confidence in an exchange partner's reliability and integrity "(p.23).Therefore, both definitions of Moorman et al. (1993) and Morgan and Hunt (1994) emphasize the importance of confidence in exchange partner's (Ok, 2004).
Last, but not least there is the Customer commitment.It has a great importance in building relationships with customers as it seen as the key predictor of customer retention (Morgan and Hunt, 1994).Moorman, Zaltman, and Deshpande (1992) defined commitment as "a continuing desire to keep a valued relationship" (p.316).Likewise, Morgan and Hunt (1994) defined it as "an exchange partner believing that an ongoing relationship with another is so important as to warrant maximum efforts at maintaining it".Building on Morgan and Hunt's (1994) definition Lacey, Suh and Morgan (2007) described it as a customer's continuous desire to stay in the relationship with a firm accompanied by his or her willingness to make efforts at maintaining that relationship.It has been defined as an inner desire to maintain a relationship with the service provider over time (Ok, 2004;Jones, Fox, Taylor and Fabrigar 2010).

Patronage Intentions (PI)
According to Rust et al. (1999) patronage intention is the customer's attitude towards the service firm as it refers to the combined effect of his/her various evaluations of the firm service dimensions (Zolfagharian and Paswan, 2009;Patney, 2010).It is an overall measure capturing likelihood and willingness to shop from the service provider, willingness to recommend to others and willingness to revisit (Grewal, Baker, Levy and Voss, 2003).It also refers to loyalty for the shopping mall, the likelihood of returning to the shop, and the likelihood of recommending the mall to others (Rahman, Wong and Yu, 2016).It contains WOM intentions, willingness to pay more, future switching intentions, intention to purchase, and repurchase/repatronage intention (Diddi and Niehm, 2016).

Service Recovery and Patronage Intentions
Starting from service failure implicitly defined as a problem of service quality (Gronroos, 1988, p. 13).Gronroos (1988) mentioned that service recovery after failure was to be considered as the provider's reaction to its poor service quality and he considered it as part of service quality dimensions.Prior studies have conf irmed a direct influence of service quality on both repurchase intentions and willingness to recommend (Zeithaml, Berry and Parasuraman, 1996).Baker, Parasuraman, Grewal, and Voss. (2002) also stated that customer perceptions of interpersonal service quality influence his/hers patronage intentions towards the company.Ogwo and Igwe (2012) found that service quality has affected positively on customer patronage.Moreover, service quality (i.e.reliability, personal interaction, and problem solving) directly influences customer's retail patronage intentions (Lee, Fairhust and Lee, 2009).In addition, Wirtz and Mattila (2004) confirmed that SR dimensions (compensation, recovery speed, and apology) effects on repatronage/repurchase intentions -as a part of patronage intentions according to Sadachar (2014), and Diddi and Niehm (2016).Therefore, we propose the following set of hypotheses: H1: (from a to e): Each service recovery strategy (Making an apology; Problem Solving; Providing Explanations; Compensation; Follow-up) has a significant positive effect on patronage intentions.

Service Recovery and Relationship Quality
Based on the literature review, the attributes of satisfaction, trust and commitment are the factors needed to measure the relationship quality construct.Previous studies found that some service recovery dimensions had a direct effect on service recovery satisfaction as a dimension of relationship quality (Tax et al., 1998;Wirtz and Mattila, 2004;Duffy, Miller and Bexley, 2006;Kim, 2007;Komunda and Osarenkhoe, 2012;Yaya, Marimon and Casadesus, 2013;Kandulapati and Bellamkonda, 2014;Stratemeyer, Geringer and Canton, 2014;Mostafa, Lages, and Sä ä ksjä rvi, 2014).In addition, SR justice further leads to customers trust (DeWitt, Nhuyen and Marshall, 2008).Therefore, we propose the following hypotheses: H2: (from a to e): Each service recovery strategy (Making an apology; Problem Solving; Providing Explanations; Compensation; Follow-up) has a significant positive effect on relationship quality.

Relationship Quality and Patronage Intentions
Park, Choi, Kim, and Rho (2015) confirmed that customer trust has a direct effect on patronage intentions.In addition, customer satisfaction affects directly patronage intentions (Sadachar, 2014;Sharma, 2015).Furthermore, customer satisfaction affects on behavioral responses, repatronage/ repurchase intentions (Choi and Chu, 2001;Wirtz and Mattila, 2004).Moreover, Wirtz and Mattila (2004) have found that SRS fully mediates the relationship between three SR strategies namely: apologizing, compensation, speed of recovery and repatronage/repurchase intentions.Therefore, we formulate the following hypotheses: H3: Relationship quality has a positive effect on patronage intentions.H4: (from a to e): Relationship quality mediates the positive effect of each service recovery strategy (Making an apology; Problem Solving; Providing Explanations; Compensation; Follow-up) on patronage intentions.

Speed of Recovery and Empathy Moderate Service Recovery Strategies Effect on Relationship Quality
The speed of Recovery is a way to reduce the customer's difficulties related to a given service failure (Battaglia, et al., 2012).A fast response can help in creating a more positive customer's service perception (Boshoff, 1999).At the same time, customers expect the company to acknowledge the difficulties service failures are imposing on them and they expect to be dealt with in an emphatic way.This behavior can be considered as a proxy for how much the company really cares about the customer's satisfaction (Zeithaml et al., 1990, p. 26).According to Liao (2007), these two elements are so important that they could even be considered as specific dimensions of Service Recovery.Other scholars have found that they are needed to really understand the customer perspective and to relate with him (Krishna et al., 2011;Battaglia et al., 2012).Therefore, we propose the following set of hypotheses:

Measures
To measure service recovery strategies, we have combined several different measures defined and validated in the literature (Boshoff, 1999;Boshoff et al., 2005, Kau andLoh, 2006;Kim, 2007;Liao, 2007; Del Rio -Lanza, Vá zquez-Casielles, and Dí az-Martí n; 2009).To measure relationship quality we depend on the measure developed by Kim et al. (2012).In order to measure patronage intentions, we have adopted the scale defined by Grewal et al. (2003) and we have combined it with items of the scale from Zolfagharian and Paswan (2009).In the end, we have used 25 items to measure service recovery, 13 items for relationship quality; and 3 items for patronage intentions rated on a 5-point Likert scale (from 1-fully disagree to 5 -fully agree).

Sample and Data Collection
We have decided to survey a sample of the Egyptian internet service user with high-speed ADSL lines in Dakahlia governorate.To sample this population and to distribute the questionnaire, which we have developed according to the literature review, we have used a social network service (Facebook) and the service provider's sites; the participants could access the Internet-based questionnaires (online survey) using Google drive as the platform.The sampling process lasted 1 month from April 28th to May 27th, 2016.The number of complete questionnaires is 430.
The questionnaire was revised after a pilot test needed to evaluate the validity and reliability of its items.The initial version was administered to people that had already suffered service failure and to several Internet professionals, to ensure that the target respondents understood the scales used.Then, we gathered data from sample units.Meanwhile, there is no a definite list of Egyptian internet subscribers we have used the convenience sampling in order to collect our data from respondents (San Martí n and Herrero, 2012).
The procedures of collecting our data from respondents were sub-divided into two stages.Firstly, two filtering questions were asked to ensure that the respondents were effectively living in the Dakahlia governorate then a second question was used to identify those who had been exposed to a service failure by their current ISP.Another step to assure the questionnaire validity was to include in it two reverse items.All the respondent features showed in the table (1).

Analysis and Results
We have analyzed our data using a PLS-SEM approach, as it is less sensitive to the multivariate normal distribution dilemma (Wang, Yeh, and Liao, 2013), moreover it is more suitable to analyze small sample sizes, and/or models with a large number of constructs, finally it is considered a good approach to avoid the multicollinearity issue in traditional SEMs (Cassel, Hackl and Westlund, 2000;Yaya et al., 2013).We have processed data using the WarpPLS software using a two steps approach, first we estimated the measurement model and then we have tested our research hypotheses (Fornell and Cha, 1994).
In order to estimate the accuracy of the PLS estimate, we have used the non-parametric techniques of resampling as bootstrapping or jackknifing (Ali, 2016: p.241).Bootstrapping method depends on the "resampling with replacement" method, which refers to that each resample holds a random arrangement of cases of the original dataset, as some cases may be repeated more than once, while some cases may not be involved at all (Sarstedt, Ringle, Smith, Reams, and Hair, 2014;p.130).Otherwise, jackknifing resampling method generates a number of resamples that equals the original sample size, and each resamples has one case deleted (Ali, 2016: p.241).
Based on the relevant advantages of jackknifing, we have chosen to adopt it (with 430 resamples) in the current study to assess the significance of both direct and indirect (mediated) parameter estimates regarding the measurement and structural models.

Measurement Model
This study has been built using a reflective measurement model.Both the construct measure's indicator reliability and internal consistency reliability are used to assess the constructs reliability, as well as, both convergent validity and discriminant validity are used to assess the validity of the constructs (Klarner, Sarstedt, Hoeck and Ringle, 2013).
The results support indicator reliability as all items loadings are over 0.50, so they are considered as statistically significant (Hair, Black, Babin, Anderson, and Tatham. 2006).Hair et al. (2006) proposed that though "generally agreed" lower limit for Cronbach's alpha value is 0.70, it still be acceptable if it reaches to 0.60 especially in social science.The results also support internal consistency of each construct as CR is greater than 0.7 (Fornell and Larcker, 1981).So, the measures are reliable and the details of this analysis are given in Table 2.
The constructs' AVEs are over 0.50, so they have convergent validity (Fornell and Larcker 1981;Hair et al. 2006), the details of this analysis are given in Table 2.The square root of the average variance extracted AVE is greater than all the other correlations in all the cases, thus confirming discriminant validity as illustrated in the following Table 3.Thus, the measurement model confirmed appropriate reliability and validity of study constructs.

Structural Model (Testing Hypothesis)
A structural model aims to test the proposed hypotheses, as it described as causal relations between latent variables.According to Klarner et al. (2013), the researchers engaged in a step-by-step analysis of the structural model to test Hypotheses from 1 to 6 comprehensibly.To begin with, in step 1 of the analysis, we only focused on the direct relationships between variables (Hypotheses 1, 2 and 3), after that we assessed the full PLS path model and, more specifically the mediator variable (Hypotheses 4).Subsequently, we tested the moderator variables (Hypotheses 5 and 6).
According to PLS-SEM, the evaluation of the model's fit depends on its ability to predict the dependent constructs.The assessment criteria are coefficient of determination (R2), predictive relevance (Q2) and the path coefficients (Sarstedt et al., 2014).The overall fit of the model will be evaluated using the following measures: Average Path Coefficient (APC); Average R-squared (ARS), Average Variance Inflation Factor (AVIF), and the Global Goodness of Fit Index (GoF).Kock (2012) recommended that APC and ARS were significant (P< 0.05) whilst the AVIF value ought to be below five.In addition, Tenenhaus GoF of ≥ 0.1, ≥ 0.25, and ≥ 0.36 indicate that the model has a small, medium, and large goodness of fit index (Tenenhaus et al., 2004).Our data has given the model a significant APC (0.116, P= 0.004), a significant ARS (0.653, P< 0.001), an appropriate value of AVIF (2.483).In addition, GoF was 0.640, which is usually sees as a "good" model fit.The overall results depicted in Table (4, 5, and 6).
In step 1, we found that the set of hypotheses H1 was partially supported.In particular we have found that the stronger effect on PI comes from the problem solving strategy (β= 0.336, p-value < 0.001), followed by providing an explanation (β= 0.164, p-value < 0.001), and then follow-up (β= 0.118, p-value = 0.007).Surprisingly the strategy of giving a compensation (β= 0.104, p-value = 0.015) had a lower, but significant, effect on patronage intentions.We have found no evidence on an effect of making an apology on PI (p-value: 0.422).
We have similar results for the H2 set of hypotheses.We found a significant positive effect from problem solving strategy (β= 0.393, p-value < 0.001), providing an explanation (β= 0.241, p-value < 0.001), and then there is the strategy of giving a compensation to the customer (β= 0.193, p-value < 0.001).Follow-up have a significant effect on RQ (β= 0.140, p-value = 0.002).Even in testing this set of hypotheses, we have found no significant effect of the making an apology strategy (p-value= 0.263).
Finally, H3 was fully supported as relationship quality has a strong, significant, positive effect on PI (β= 0.833, p-value < 0.001) The results showed that R2 of PI is 0.359, meaning that service recovery strategies as a whole can explain about 36% of the variation in internet customer's PI as a dependent variable.The R2 of SR strategies, as a whole, on RQ is 0.627.The Q2 value of PI (0.355) and RQ (0.663), which indicates the large predictive relevance of the PLS path model.Then, in step 2, we have found that relationship quality fully mediated the positive relationship between problem solving (β= 0.327, p-value< 0.001), providing explanation (β= 0.201, p-value < 0.001), compensation (β= 0.161, p-value< 0.001), follow-up (β= 0.117, p-value < 0.001) and patronage intentions.The results refer to R2 of PI is 0.685, meaning that service recovery strategies explain about 69% of the variation in internet customer's PI as a dependent variable.The Q2 value of PI is (0.593), indicating that the model has a great predictive relevance.So, the hypotheses H4 can be considered partially supported.For hypotheses H5 and H6, we found that speed of recovery moderated the positive relationship between making apology strategy (β= 0.088, p-value = 0.032), providing explanation (β= 0.094, p-value = 0.025), follow-up strategy (β= 0.103, p-value = 0.015) and relationship quality.The results refer to R2 of RQ is 0.629, meaning that service recovery strategies with the speed of recovery technique explain about 63% of the variation in RQ between internet customers and service providers as a dependent variable.We have found that our model has a great predictive relevance as the Q2 value of RQ is (0.679).
In addition, we found that empathy moderated the positive relationship between problem solving strategy (β= 0.095, p-value = 0.024), providing explanation (β= 0.090, p-value = 0.030) and relationship quality.The results refer to R2 of RQ is 0.582, meaning that service recovery strategies with the empathy technique explain about 58% of the variation in RQ between internet customers and service providers as a dependent variable.The Q2 value of RQ (0.670), which is well above zero, indicating the large predictive relevance of the PLS path model.So, the hypotheses H4a, H4b partially supported.Relationship quality = RQ, Patronage intentions= PI. -* Significant at 0.05, ** significant at 0.01

Discussion
Our study mainly aims to understand the relationship between service recovery strategies, relationship quality, and patronage intentions, as well as testing the mediating role of relationship quality in the relationship between SR strategies and consumer's patronage intentions and the moderating role of both recovery speed and empathy in the link between SR strategies and relationship quality.Moreover, providing effective practical recommendations to the internet service firm's managers.
We have found a significant positive effect for four SR strategies on PI and a similar result on RQ.The only SR strategy that was not found to have a significant effect has been making an Apology.This confirmed that SR strategies are necessary to maintain customers and create positive WOM in addition to deep the customer's intention to revisit the internet provider and continue dealing with him.Our results are also consistent with Lee et al. (2009) on that problem solving directly influences customer's patronage intentions.
SR strategies have a positive impact on RQ as well.This confirmed that SR strategies as an aspect of service quality deepen the quality of the long-term relationship between service provider and its customers.SR strategies are an effective way to increase customer satisfaction, which in turns increase trust in the provider and create customer commitment toward it.These results are on the same page with other results we have found in the existing literature (Tax et al., 1998;Wirtz and Mattila, 2004;Duffy et al., 2006;Kim, 2007;DeWitt et al., 2008;Komunda and Osarenkhoe, 2012;Yaya et al., 2013;Kandulapati and Bellamkonda, 2014;Stratemeyer et al., 2014) that SR strategies will enhance relationship quality.
We believe that the strange result for the making an apology strategy can be related to the specific Egyptian market.Egyptian Internet customers may not find this strategy effective as (1) internet services are really prone to failures, according to industry reports; (2) Egyptian customers' income is usually low so customers may prefer those strategies letting them have some sort of material compensation.These findings are consistent with Levesque and McDogugall (2000) viewpoint that an apology is better than none, but it alone is relatively ineffective when a customer endured a large loss (i.e. the case of a core service failure).
As expected from the study of the previous literature on Relationship Quality (Choi and Chu, 2001;Wirtz and Mattila, 2004;Sadachar, 2014;Park et al., 2015;Sharma, 2015) we have found a significant positive effect for it on patronage intentions.This highlights that the quality of relationship influences customer behavioral responses.This confirmed that relationship quality is very important to create or building customer loyalty and then create competitive advantage (Doaei et al., 2011).
Looking at the indirect effect of SR strategies on PI we have found that RQ fully mediated the impact of the same SR strategies (namely, problem-solving, providing explanation, compensation, and follow up).The results show that these strategies will increase the relationship quality between service providers and its customers depending on how the company dealt with the service failure and that they will help in creating a better attitude towards the company if they are carried on in a proper way.These results are partially consistent with those found by Wirtz and Mattila, 2004 i.e. service recovery satisfaction mediated the positive relationship between some SR strategies and repatronage intentions.
We have found that speed of recovery moderates the positive impact of several SR strategies (namely, apology, providing an explanation, and follow -up) on RQ.These results confirmed the role of recovery speed on improving the relationship between SR strategies and RQ.
In particular, we have found that when making an apology strategy is coupled with the speed of recovery it can have a small, but significant, positive effect on RQ.It follows that, if the provider makes an apo logy while being fast in solving the problem, it may still increase the customer satisfaction.A similar result has been found for providing explanations and following-up.These strategies will strengthen the customer's trust on the provider, which in turns affect positive word of mouth and repurchase intentions and likelihood to revisit.
Finally, our results highlight that an empathic approach enhances the effectiveness of problem solving and providing explanations on RQ.When the provider employees appear empathic and able to feel for the customer's plight or dilemma, while solving customer problems, they may be more effective in satisfying the customer, which in turn will strengthen the customers trust on the provider and maintain the relationship with pr ovider, which affect his or hers behavioral responses.These two results confirmed the important role of speed of recovery and empathy techniques to support the influence of SR strategies on relationship quality aspects.

Conclusions
This article let us improve the theoretical knowledge on the relationship between the SR strategies, relationship quality and customer patronage intentions in the internet industry context.Previous studies have found evidence that some SR strategies act as a direct cognitive antecedent to repatronage intentions as an aspect of PI (Wirtz and Mattila, 2004).In addition, problem solving acts as a direct antecedent to retail patronage intentions (Lee et al., 2009).More recent research shows that SR strategies directly affect service recovery satisfaction (Wirtz and Mattila, 2004;Kim, 2007;Duffy et al., 2006;Komunda and Osarenkhoe, 2012;Yaya et al., 2013;Kandulapati and Bellamkonda, 2014;Stratemeyer et al., 2014;Mostafa et al. (2014) andcustomer trust (DeWitt et al., 2008 ) as a sub-components of relationship quality.In addition, some aspects of relationship quality influence patronage intentions (Sadachar, 2014;Park et al., 2015;Sharma, 2015).
In this work, we have provided the contribution of providing evidence that most SR strategies (problem solving, providing an explanation, compensation, and follow-up) can really affect customer patronage intentions.In addition, our research shows that RQ mediates the effect of SR strategies (problem solving, providing an explanation, compensation, and follow-up) on PI.This indicates that the more effective the SR strategies are the better the relationship quality will be, and the more these strategies will satisfy the customer and the more they will create trust and commitment the stronger their effect will be.This highlights the need to take into account the relationship quality between service providers and their customers and its essential role in enhancing PI.Furthermore, the speed of recovery and empathy with customer's techniques can enhance the impact of some SR strategies on RQ (namely speed of recovery technique for an apology, explanations, follow-up strategies, and empathy technique for both problem-solving explanations strategies).Moreover, our finding supports the idea that apologies have no significant effect on RQ alone and, indirectly, on PI.Consequently, they cannot be considered as an effective SR strategy if they are not coupled with some other SR strategy or, at the very least, the service provider is able to deal quickly with the problem itself.
Our findings regarding making an apology strategy, open the way to a more in-depth research effort to understand which specific services from ISPs, can be really considered as core services (e.g., denial or unavailability of service), and which of them are service encounter failure / process failure (e.g., lack of civility and non-response and the lack of the existence of personal service).On the other hand, as previously noted, this can be linked with a specific cultural attribute of customers in Egypt so these results would benefit from further validation in different cultures.
This article can help ISPs management as well.We confirm that a successful relationship between service provider and its customers are largely determined by maintaining long-term valuable relations.When the customers feel that, their provider is trying to provide a high-quality service and that it is interested to satisfy them, then customers will appreciate its efforts and will have more trust in it.All this previous feelings and attitudes will help to create a stable relationship that, in turn, will reflect in customers' patronage intentions.
In order to improve that long-term relationship and to get more benefits, managers should concentrate on service recovery strategies especially the intangible ones and train employees how to apply these strategies quickly and with empathy.Furthermore, internet providers must know that essential strategies are problem-solving and providing real explanations to customer and support with appropriate compensation and do not forget to follow up customers to ensure that all is going well, all of this will enhance the service provider competitive advantage.
Finally, despite its contributions, the current study has some limitations, which provide directions for further research.First, this study mainly depends on five service recovery strategies namely: apology, problem solving, providing explanations, compensation, and follow-up to identify its effects on patronage intentions as well as it examined the moderating role of both speed of recovery and empathy in the relationship between SR strategies and PI.According to our findings, all direct paths are supported despite the ones between apology and both relationship quality and patronage intentions.Therefore, future researchers may want to examine the combined effect of apology with each of the other four strategies (e.g.apology × problem solving) to understand the consequences of a combined strategy, and this may generate some meaningful insights.Second, this study puts emphasis on exploring the effect of SR strategies on PI and it looks only at the mediating effect by relationship quality, therefore, we expect that customer gratitude might have a meaningful mediating role to investigate the relationship between SR strategies and patronage intentions.Third, our study not found a significant influence of apology as service recovery strategies on relationship quality or patronage intentions attributes, so, further research may be concentrated on retesting our proposed model in other settings, moreover, if these other contexts have a tighter interaction between providers and consumers (e.g. the hotel industry).Finally, further studies may give more attention to those issues related to a cross-cultural approach.
H5: (from a to e): Speed of recovery moderates the positive relationship between each service recovery strategy (Making an apology; Problem Solving; Providing Explanations; Compensation; Follow-up) and relationship quality H6: (from a to e): Empathy moderates the positive relationship between each service recovery strategy (Making an apology; Problem Solving; Providing Explanations; Compensation; Follow-up) and relationship quality We illustrate the theoretical model in the figure (1).

Figure 1 .
Figure 1.The theoretical model

Table 3 .
Correlations among latent variables with square roots of AVEs Making an apology =MA, Problem solving= PS, Providing explanation= PE, Speed of recovery= SPE, Empathy= EMP, Compensation= COM, Follow-up= FOL, Relationship quality = RQ, Patronage intentions= PI.

Table 5 .
Indirect effect of SR Strategies on PI

Table 6 .
Moderating effect of speed and empathy on the SR strategies effect on RQ