Dear Customer, Thank You for Your Review: The Service Failure-recovery Dyadic Interactions in the Restaurant Industry

The advent of Web 2.0 has encouraged restaurant customers to post online reviews, and oftentimes, not in favor of the company. When a service failure occurs, the customer may voice their complaints publicly online. The company, on the other hand, has the opportunity to respond to these complaints and use it as a part of their service recovery strategy. While some companies are responding to negative reviews, only a few have the knowledge on how to do it effectively. Built on perceived justice framework: distributive, procedural, interactional; and service failure severity type: outcome-process, major-minor, present study intends to understand different resolution styles adopted by the company to varying types of customer complaint. The findings outline: (1) the vast majority of the company exhibits only a low level of responsiveness to complaints; (2) there seems to be a correlation between physical and psychological loss with time loss, severe emotions and switching intentions; (3) however, different strategies depending upon service failure severity are yet to be implemented by the company; (4) while components in interactional justice are mostly performed, rude responses are also frequently applied. Further elaboration of the findings and insights for marketing practice are discussed in the text.


INTRODUCTION
"Dear Manager, We were just at your restaurant and we ordered large goulash soup, paprika chicken and wine. My family and I felt very humiliated and insulted by the waiter at that time around 7 PM. We had asked for an extra bowl for the soup so our young daughter can try and he flat out refused, told us to order an extra bowl of soup and at the end just said "just eat it" and walked away. We felt very offended and felt he was treating us extremely poorly especially since this is not an outrageous request. Is this the way you treat your customers locally or to tourists like us? This experience has given us a very negative impression of your restaurant and I hope you will consider this matter seriously. I am also sharing our experience with other potential customers to ensure they consider this before visiting your restaurant. Hope you will improve the level of service and avoid waiters with such poor ability." Response from the Owner "I feel very offended that you write such lies and steal my time with the answer. I know the problem is: no free water. Why don't you write this? Hope your kid is smarter." The above customer complaint-company response interaction appeared on Google reviews, one of the most prominent third-party online customer review (OCR) platforms that accommodates both the company and customer in the hospitality and service industry. Despite the benefits of multiple reviews on OCR platforms, company's online exposure is also increasing as the company is exposed to negative reviews that are virtually apparent to local and global audience. It can be seen from the above interaction that a customer encountered a service failure at a restaurant. The disappointment of the service quality prompted the customer to voice an online complaint in real-time. Moreover, the complaint was addressed to the restaurant management, but at the same time, it also encouraged potential customers to re-evaluate their considerations before making a final decision. The restaurant management, on the other hand, did not show an intention to retain its existing customer or to gain trust of the potential customers.
In a business context, the base-value of the marketing concept is customer satisfaction. To increase business growth, it is fundamental for marketers to be able to identify customer needs and to strive to satisfy these needs. The extended effects of customer satisfaction have also been assumed to have a close linkage with customer loyalty. However, as with most things with two polar opposites, customer satisfaction comes with a downside; it is prone to fall to the opposite side of the sphere: dissatisfaction. Customer dissatisfaction, oftentimes, leads to more complicated layers than that of customer satisfaction. Dissatisfied customers who have encountered with service failures are likely to seek redress (i.e., repair, exchange, refund or other compensations); engage in negative word-of-mouth behavior that has the potential to discourage other customers to purchase; decrease the volume or frequency of purchase; pledge to never re-patronize and switch to competitors; or completely exit the market. With that being said, it is wise to say that the thorough implementation of marketing concept includes the ability of marketers to resolve service failures and turn customers to a state of satisfaction. Fornell and Wernerfelt (1988) introduced the term "defensive marketing" that deals with the protection of one company's market share from its competitors. While service failure is inevitable, specifically in the service industry, the company may have the opportunity to respond to service failures. Such response is acknowledged as service recovery. Grönroos (1988) suggested that "service recovery refers to the actions a service provider takes in response to service failure". The term "the recovery paradox" even proposes that a highly effective service recovery may lead to higher satisfaction and loyalty rates than if the service failure had never occurred in the first place (McCollough and Bharadwaj, 1992).
Nowadays, the market dynamic is changing its shape as a result of globalization and technology disruption. It expands a wider range of interaction between marketers and customers or customers and their peers, at and through multiple touchpoints. The integration of Web 2.0 into our day-to-day lives has had a profound effect on the way marketers and customers excessively rely on social media platforms to communicate than what they do in the faceto-face settings (Kujath, 2011). Thus, the ubiquity of social media and its significant importance as a communication channel must be utilized by marketers to further develop their marketing communication strategy (Abney et al., 2017). Extant research posits that social media platforms have the ability to disseminate information that is larger in audience size and proximity (Hennig-Thurau et al., 2014;Purnawirawan et al., 2012;Jeong and Jang, 2011;Hennig-Thurau et al., 2004) and to impact purchase intention (Tata et al., 2019;Zhu and Zhang, 2010;Jansen et al., 2009;Liu, 2006). Further, it has also been shown that today's customers use social media as a complaint tool (Schaefers and Schamari, 2016).
The company is aware of the widespread utilization of social media to voice complaints as well as the nature of its virtual presence that they "reluctance of publicly handling complaints" (Einwiller and Steilen, 2015;p.7).
Prior research suggests that company's service recovery effort by responding to negative online customer reviews (OCRs) signals a commitment to improve service quality and affects potential customer's trust toward company credibility; subsequently it influences their purchase intention (Olson and Ro, 2020;Sparks and Bradley, 2014;Pantelidis, 2010;Ye et al., 2008). While the company in the hospitality and service industry is becoming aware of the importance to manage negative OCRs, these companies are still unsure how to best respond to customer complaints (Sparks et al., 2016;Schaefers and Schamari, 2016;Sparks and Bradley, 2014;Xie et al., 2014). In the recent years, there has been a growing interest in these critical areas of service management. A number of scholars have called for more research to examine service recovery effectiveness to re-gain customer confidence and influence purchase intention (Olson and Ro, 2019;Sparks et al., 2016;Sparks and Bradley, 2014;Min et al., 2014) as there are still output discrepancies in the existing literature.
Furthermore, there has yet ample evidence in the methodological variations in measuring service recovery antecedents, processes and outcome (Michel, 2001). The author further stressed that the critical incident technique has been widely used in hospitality and service research, however this technique may cause recall bias as a result of the time lag between a service failure occurrence and the interview. Experimental design using scenarios is also a popular method and it may increase the internal validity, but as it lacks customer emotion's as opposed to what they feel in a real setting, it often decreases the external validity. The thirdmethod that is also used regularly employs customer complaint database addressed directly to the company (e.g., Point-of-Sales complaints, guests satisfaction survey, comment cards). An advantage of this method is that the data is based on real customer complaints expressing actual service failures. However, only a small number of customers who actually write complaints at the Point-of-Sales.
This present study intends to address the gap and extends prior research by building it upon an empirical research using a content analysis on a third-party website that is also based on actual customer complaints and company responses. Using online complaints as a database will not only deal with the interactions between the dissatisfied customer and the company, but also its potential effects on future customers. Further, this study serves as a preliminary study in the production of service recovery guidelines in formulating an effective compensation strategy as well as the appropriate response to negative OCRs, specifically in the hospitality and service industry. The contribution of this study is twofold. First, from an academic standpoint, the study advances theoretical understanding of service recovery. Second, this study presents insights for marketing practice to optimize OCRs including the negative ones as a part of their company multichannel footprint and communication strategy.

How do Customers Utilize OCR Platforms to Push-and-pull Information?
With the rise of virtual communities, a new type of online wordof-mouth has gained its popularity, which is referred to as online customer reviews (OCRs). Formally, OCRs can be defined as an aggregation of user-generated evaluations of the company, independently from those companies. In other words, customers articulate their opinions on third-party websites (Beuscart et al., 2016;Mudambi and Schuff, 2010). Previous studies have found that OCRs are more credible and trustworthy than professional reviews or even firm-generated information (Senecal and Nantel, 2004;Bickart and Schindler, 2001). The source credibility of OCRs is also noted in several number of studies (e.g., Ruiz-Mafe et al., 2018;Kim et al., 2017;Beuscart et al., 2016;Elwalda and Lu, 2016). According to Park et al. (2007), potential customers trust OCRs as references in their decision-making process as the customer usually provide honest information based on their experience with the product or service. In addition, as the barriers to voice negative experiences on OCR platforms are low, OCRs have become more popular as a complaint tool. OCRs does not only enable the dissatisfied customer to avoid direct confrontations with the company when a complaint is raised (Hong and Lee, 2005), but also enables the customer to receive the conformity from their peers who have experienced similar problems.
Derived from these arguments, a pre-test study is thus conducted to cross-examine the attitudes of restaurant customers in Hungary toward OCRs. A self-completion survey with non-probability sampling was carried out through social media from November 2020 to January 2021. A total number of 185 Hungarian residents (locals and foreigners) as well as non-residents which comprised of male and female; aged 18 and above; across different occupations; and financial well-being status were participating in the survey. Quota sampling was set on a natural fall-out basis to be reflective of market share. Table 1 demonstrates restaurant customer attitudes toward OCRs in Hungary.
Based on the empirical evidence that has been gathered, only around 30 per cent of dissatisfied customers were likely to engage in negative OCRs. These negative OCRs, however, could influence the number of potential customers who trusted and relied on OCRs when selecting a restaurant which almost double in size (T2B=56.8%). It is also important to note that while good reviews or ratings were important for potential customers (T2B=53.5%), effective service recovery in response to negative reviews was more fundamental to increase customer confidence toward a restaurant (T2B=64.9%, P < 0.05).

Developing Service Failure Severity versus Perceived Justice Frameworks
A stream of research has demonstrated that a successful service recovery management is the cornerstone in building foundation for a long-term relationship with the customer who has encountered service failures (Maxham, 2001;Smith et al., 1999;Tax et al., 1998). However, service failures can range from the ordinary cases to more major cases (Goodwin and Ross, 1992;Berry and Parasuraman, 1991;Gilly and Gelb, 1982) and perceived severity of cases has been identified as the mediating factor of an effective service recovery (McCollough et al., 2000;Smith et al., 1999;Limbrick, 1993;Zeithaml et al., 1993;Bell and Ridge, 1992). It is also found that service failure severity has a significant strong effect on service recovery satisfaction (Olson and Ro, 2020).
Service failure severity refers to a customer's perceived intensity of a service problem. The more intense or severe the service failure, the greater the customer's perceived loss (Weun et al., 2004;p.135).
Further, Grönroos (1988); Parasuraman et al. (1985) suggested that service quality depends on both the outcome and the process of the service itself. Thus, customer loss may also occur along these two factors. Smith et al. (1999); Gilly and Gelb (1982) explained that outcome failure causes a direct monetary loss, while process failure does not have a direct impact on monetary loss (e.g., time loss). Against this backdrop, this study divides service failure severity type into outcome-process if the complaint is constructive, and major vs. minor if it does not contain a description of the problem. Outcome loss will be measured through two different constructs: financial loss and physical loss. Process loss will cover psychological and time loss constructs. Whereas, major failure refers to the major incidents that lead to frustration or anger, minor failure refers to the non-catastrophic incidents leading to unpleasantness. Accordingly, Objective 1 is thus posited: • Objective 1 -To identify varying levels of service failure severity leading to negative OCRs In service recovery research, justice theory has been widely used as a theoretical framework (e.g., Ha and Jang, 2009;Smith et al., 1999;Sparks and McColl-Kennedy, 1998;Tax et al., 1998). Prior research suggested that in returning the dissatisfied customer to a state of satisfaction and repurchase intention, the outcome is mainly affected by customer-perceived justice in the service recovery process (  service recovery justice is defined as the fairness of the way service failures are managed from three dimensional approaches: distributive justice; procedural justice; and interactional justice (Gelbrich and Roschk, 2011;Orshinger et al., 2010;Ha and Jang, 2009;McColl-Kennedy and Sparks, 2003;Blodgett et al., 1997). Distributive justice reflects the tangible outcome of the service recovery (i.e., redress), procedural justice concerns the procedures used to reach the outcomes (i.e., assurance), while interactional justice deals with the way the customer is treated during the service recovery (i.e., acknowledgement) (Wang et al., 2011;Tax et al., 1998;Blodgett et al., 1997;Lind and Tyler, 1988;Thibaut and Walker, 1975). Using distributive, procedural and interactional justice constructs, Objective 2 is then formulated: • Objective 2 -To identify different service failure resolution styles adopted by restaurant establishments during service recovery process.

Study Design
In accordance with the two objectives proposed in this study, relevant empirical study with the use of a content analysis was undertaken in the restaurant industry, specifically restaurants operating in larger cities located in Hungary. The content analysis comprises solely OCRs posted on Google reviews. Google reviews was selected considering that based on the findings of a survey conducted by BrightLocal (2018), "it consistently gains the highest number of new reviews" in comparison to other online review platforms (e.g., TripAdvisor, Facebook, Yelp, Foursquare).
During the sampling and data collection phase, restaurants were methodically selected. First, they were categorized according to price range and star rating. A total number of 150 full-service restaurants across Hungary, ranging from price range $ to $$$$ and had a star rating with an upper bound five-star and a lower bound one-star were identified. The selected restaurants then stratified based on the price range and star rating groups. In the final data, there were 50 restaurants which had a star rating equal to or higher than 4.5 (coded as Top-rated); 50 restaurants with a star rating in a range of 4.0 to 4.4 (coded as Moderate); and 50 restaurants were rated lower than 3.9 (coded as Bottom-rated). In the price range breakdown, there were 75 restaurants in a price range of $ to $$ (coded as Low-priced); and 75 restaurants ranging from $$$ to $$$$ (coded as High-priced).
For the analysis, only ten randomly selected reviews per restaurant were included in order to maintain a more robust data. These reviews were sorted by the lowest rating to ensure they were contained of one or more complaints. For coding purposes, a codebook that captured pre-developed constructs and other elements in the customer complaint-company response interactions was created. All (n=1500) selected reviews were coded based on the codebook, and computed into the data set using Excel and then further analyzed using SPSS Statistics 24. Significant test among each group was performed with a two-tailed Z-test to determine if the differences that were present between variables were significant at 99%, 95% or 90% confidence level. A Z-test is used if: (1) The test statistic of each sample comes from a normal distribution; (2) the sample size is large and the population variance is known (Sprinthall, 2011).

RESULTS
The results of service failure severity analysis revealed that based on the total number of complaint case, the highest frequency of service failure found to occur at the bottom-rated restaurants (580 cases) in the low-priced range (858 cases), being financial loss; psychological loss; and the major incidents as the most frequent failures. However, these high scores do not necessarily correspond to the severity level. Table 2 presents the full details of service failure severity and its classification.
Customers who reported that they would not recommend or would not go back appeared to be significantly higher for the low-priced restaurants (37%, P < 0.01) in moderate (39%, P < 0.05) and bottom-rated (39%, P < 0.05) categories. The significant score in low-priced category was pulled-down by psychological loss score in which its significance was also found in moderate category.
Restaurants along lower price levels in moderate category were significantly superior in a number of complaints related to psychological loss among peers (32%, P < 0.05), while those in bottom-rated category were superior in receiving complaints linked to physical loss (56 of 500 or 10%, P < 0.05). Psychological loss measures the frontline staff and management attitude (i.e., rude, impolite, unfriendly behavior) when delivering the service and handling issues which reflects customer mental well-being during service encounter. Physical loss, on the other hand, represents poor quality of food that causes a decrease in customer's physical health; in some cases, the issues can be major. It should also be noted that the number of customers who experienced frustration/anger was significantly high at the same time where frequency of complaints about physical loss was also high. Furthermore, although there was no significance different between the low-priced and high-priced categories, time loss emerged as significant within restaurants in moderate (62 of 500 or 11%, P < 0.05) and bottom-rated (54 of 500 or 9%, P < 0.05) categories; subsequently, it added to the disappointment toward the overall service quality. This finding suggests that if physical loss and psychological loss were supported by time loss and the incidents that triggered frustration/anger (157 of 500 or 27%, P < 0.05) it could lead to discouragement for potential customers to try and switching intentions. The causal relationship between service failure and behavioral outcomes as well as perceived loss and emotion, however, were not statistically tested in this study.
With regard to the severity level of service failure, low-priced restaurants under moderate and bottom-rated categories clearly have produced the most severe service failures, while other restaurants have only caused common failures that did not influence behavioral outcomes. Restaurants in high-priced category received most complaints about financial loss (22%, P < 0.01) which were on parity among star rating group (a=19%; b=20%; c=18%), whereas top-rated restaurants (28%, P < 0.05) in low-priced category (23%, P < 0.01) have only caused unpleasantness due to very minute ordinary service failures. Financial loss represents customer disappointment toward the actual price they need to pay as compared to how much they are willing to pay for the overall experience, it includes additional service tax and gratuity. The low perceived price justice could be explained by the large gap between customer expectations and company performance. Intuitively, this makes sense as the larger the gap, the higher the disappointment and the perceived loss.
In regard to company response to customer complaint, results indicated that on average, only around 21 per cent of complaints received a reply from the company. Overall, the frequency of reply scores across categories and groups were consistent, with the exception of bottom-rated restaurants which performed below average (8%, P < 0.05). The similar pattern also emerged in the total response within each category, the bottom-rated category's score appeared to be the lowest in comparison with the other categories (31%, P < 0.05). Based on the total frequency with the value equal to or above 5 per cent, timeliness (11%) showed the highest in procedural justice construct. Of the total complaint the customer sends and gets a reply, all companies responded in a timely manner (Timeliness/Total replies=100%). Assurance that represents a statement in expressing a promise to not repeat the failure made up 5 per cent of the total responses.
In interactional justice approach, courteous manner (12%); expression of regret (7%); and explanation of process (5%) were enacted more frequent. These numbers showed that the baseline in common apologies have been applied by the company. While the proportion of corrective action in distributive justice was relatively small (2%), the inclusion of complimentary was expected to be under careful consideration as it should be dependent upon the problem and whether comping is possible. However, as significant differences between groups in most dimensions were not observed signify inconsistency of responses given by the company. In other words, different strategies depending upon the severity of the problem were omitted. Table 3 illustrates different resolution styles adopted by the company within perceived justice framework.
Interestingly, a high-frequency response did not necessarily associate with an effective service recovery as there were companies responded in a confrontational manner despite the severity of the complaint, specifically top-rated restaurants (9%, P < 0.05) in low-priced category (6%, P < 0.10), which accounted for a total of 5 percent. These findings highlight that some companies had a lack of awareness on the importance of service recovery for the longevity of their business. Sensibly, top-rated restaurants may be defensive toward negative OCRs as it is more difficult to sympathize with the minority of which their voices challenge the conformity from majority voices that is already manifested as valid. While these companies may not be aware of the importance of service recovery, they may be aware of the effects of negative OCRs, hence the effort to rectify the situation from escalating to viral scope. In contrast, other companies seemed to be unwilling to get into the game and disregarded those negative OCRs that may partly due to a lack of knowledge on how to deal with it effectively.
Furthermore, as moderate restaurants in low-priced category received more severe complaints among peers, these restaurants appeared to be the most effective in responding to customer complaints, even though there was still a low number of rude or defensive responses (b=3%, P < 0.05; d=6%, P < 0.10).
In this study, social presence is defined as the response of the transgressor with a name and job title. According to Hess et al. (2009);Gunawardena (1995), the objective of social presence is to provide a sense of human contact, sensitivity and being real in communication.

DISCUSSION AND CONCLUSION
The main goal of this research is to gain insight into service failure-service recovery dyadic interactions on OCR platforms.
One complaint/sample can contain one or more cases. *Tallied per each complaint/sample -adds up to 100%. A-E: Denotes significant difference at 99% confidence level among group. a-e: Denotes significant difference at 95% confidence level among group Specifically, this study examines different resolution styles adopted by the restaurant in responding to varying levels of customer complaint on Google reviews. Some key takeaways from this study underline that in general, the vast majority of the company exhibits only a low level of responsiveness to negative OCRs regardless of the issue; the most severe complaint cases affect behavioral outcomes are linked to psychological and physical loss with time loss as moderating factor. On top of that, major incidents lead to frustration/anger also play an important role; however, different service recovery strategies upon service failure severity are yet to be implemented; and while the baseline in common apologies (i.e., courteous manner, expression of regret, explanation of process) is mostly performed, confrontational responses are also vigorously applied by the company. Summarizing, it is noteworthy that the majority of restaurants in Hungary are yet to fully embrace the benefits of social media-specifically OCR platforms-as an important touchpoint to establish their interactions with the customer.
The unwillingness of the company in responding to complaints was in line with previous research. Result of a study by Einwiller and Steilen (2015) showed that nearly half of the complaints voiced on Facebook and Twitter did not receive any corporate response. Beuscart et al. (2016) examined the reception of negative OCRs in the restaurant industry from the supply side of the business. Results from this study explained that most restaurants consider negative OCRs which came from the lay judgments as harsh and difficult to cope with as it was a major challenge to figure out how to reply in an appropriate manner so that they were reluctant to deal with such reviews. As a matter of fact, it is a conventional wisdom that in returning customer confidence, providing responses in the favor of dissatisfied customers are better than no responses. Tomlinson et al. (2004) proposed that in expressing an apology, the transgressor acknowledges the unfortunate events which more effective than no apologies. In the reference to the pretest study, the number of participants who reported the importance of service recovery when selecting a restaurant was significantly higher compared to positive reviews with no occurrences of service failure. In other words, proper responses to complaints should yield higher satisfaction toward service quality. The company must synergize their efforts and online airtime on social media with those of the customer's (Song et al., 2016).
However, the high number of rude responses observed in this study contradicts prior research which indicated that defensive responses were rarely or even zero-applied due to the awareness of the dire impacts. This seems counterintuitive and is a valuable finding in the service recovery context as, in fact, more companies are becoming aware of negative OCRs in leading public sentiment that it creates an entirely new industry involving firms that specifically deal with social media crisis management (Gellman, 2014). Some restaurants evaluated in this study may not be aware of the impacts of their rude responses or they may accept complaints as a personal attack that it is hard to compose their emotions when reading the negative reviews. It could also be explained by different individualism/collectivism between cultures as noted by Liu and McClure (2001). Nevertheless, this study brings about an insight for marketing practice. Existing literature review demonstrated that positive results will be produced more when the responsibility of a transgression is acknowledged by the transgressor than if the transgressor deflects or denies the responsibility (e.g., Kramer and Lewicki, 2010;Schlenker and Darby, 1981;Scott and Lyman, 1968). Research also showed that a poor recovery process (no responses and rude responses) can decrease sales revenue (Temkin Group, 2017). A poor recovery process may have an impact on a lost customer and subsequently, it leads to a loss of profits. As an illustration, lost customers from the Responses between dimensions can be overlapping. *Tallied per each reply/sample -adds up to 100%. a-e: Denotes significant difference at 95% confidence level among group. (X) A -(X) E : Enotes significant difference at 90% confidence level among group poor handling of negative OCRs can be estimated by multiplying total complaints, number of dissatisfied customers who are likely to voice complaints and number of complainants who vow to stop doing business with the company.
Furthermore, concerning the implementation of different strategies depending upon the severity of the problem, the result challenges findings from prior research (e.g., Abney et al., 2017;Weun et al., 2004) which found that customer satisfaction with the service recovery was significantly affected by service failure severity. Considering the importance to adopt different recovery strategies to be effective, the result, thus, presented an opportunity for the restaurant to convert. At its essence, Lewicki et al. (2016) has demonstrated that neither undercompensating nor overcompensating is effective, the restaurant may take service failure severity into account. In light of complimentary or redress, if not implemented considerably it may have an impact on sales in the long run. Referred to the findings, the restaurant could focus on major incidents related to physical loss, for instance food poisoning as well as if the wait time to receive orders reaches a certain time period. However, Haesevoets et al. (2013) also proposed that compensation has to be provided along with an apology as it can better preserve the ongoing relationship with the customer than compensation without an apology.
Issues that deal with the frontline staff attitude could be redeemed by sending a proper apology consisted of relevant components in perceived justice framework as results elicited that some components are yet to be optimized in the company responses. According to van Laer and de Ruyter (2010), the more severe the problem is, the more components have to be included in the apology. Further, the high number of unconstructive criticisms emphasized that issue probing should be incorporated in the structure. Darby and Schlenker (1989, p. 354) stated that what component is to be included in the structure recognizes the transgressor's intensity of remorse, the more complete the response is, the more they "may seem to suffer remorse." As in any research, this study has to be acknowledged with certain limitations. First, this study did not measure the effect of different resolution styles on the actual sales. Second, the impact of ways the company responds on the complainant and potential customer's trust and purchase intention was also not examined. These bear an opportunity for future research to carry out an experimental study using scenarios and control variables to be able to capture its effects on those factors. In addition, future research could also explore the depth of service recovery issues in the restaurant industry particularly in Hungary and/or throughout Europe to answer the 'whys' from the company point of view.