Webcare’s effect on constructive and vindictive complainants

Abstract

Purpose

This paper aims to demonstrate that online complainants’ reactions to a company’s service recovery attempts (webcare) can significantly vary across two different types of dissatisfied customers (“vindictives” vs “constructives”), who have dramatically diverging complaint goal orientations.

Design/methodology/approach

Online multi-country survey among 812 adult consumers who recently had a dissatisfying brand experience and turned to a marketer-generated social media site to voice an online complaint for achieving their ultimate complaining goals. Scenario-based online experiment for cross-validating the survey findings.

Findings

Results suggest that “vindictive complainants” – driven dominantly by brand-adverse motives – are immune to any form of webcare, while “constructive complainants” – interested in restoring the customer-brand relationship – react more sensitively. For the latter, “no-responses” often trigger detrimental brand-related reactions (e.g. unfavorable brand image), whereas “defensive responses” are likely to stimulate post-webcare negative word-of-mouth.

Research limitations/implications

This research identifies the gains and harms of (un-)desired webcare. By doing so, it not only sheds light on the circumstances when marketers have to fear negative effects (e.g. negative word-of-mouth) but also provides insights into the conditions when such effects are unlikely. While the findings of the cross-sectional survey are validated with an online experiment, findings should be interpreted with care as other complaining contexts should be further investigated.

Practical implications

Marketers have to expect a serious “backfiring effect” from an unexpected source, namely, consumers who were initially benevolent toward the involved brand but who received an inappropriate response.

Originality/value

This research is one of the first research studies that enables marketers to identify situations when webcare is likely to backfire on the brand after a service failure.

Citation

Weitzl, W.J. (2019), "Webcare’s effect on constructive and vindictive complainants", Journal of Product & Brand Management, Vol. 28 No. 3, pp. 330-347. https://doi.org/10.1108/JPBM-04-2018-1843

1. Introduction

After unsatisfactory brand experiences, consumers can easily share their negative thoughts and feelings with many others by posting adverse comments on various online platforms including social network sites (e.g. Facebook), microblogs (e.g. Twitter) and discussion forums. Such written online statements from dissatisfied consumers that denigrate a specific brand (Laczniak et al., 2001) are basically a form of negative electronic word-of-mouth (NeWOM). The reasons why consumers engage in NeWOM are manifold. Extant literature (Grégoire et al., 2015) suggests that NeWOM motives range from constructive forms aimed at rebalancing the customer’s relationship with the brand by seeking a problem solution to more vindictive forms when complainants try to deliberately harm the brand by calling on others not to buy it.

In the search for effective means to repair the relationship with their dissatisfied customers and to mitigate negative reactions of NeWOM observers – such as unfavorable brand image evaluations, switching behavior and boycotting (Chevalier and Mayzlin, 2006) – many companies monitor public online complaints and try to interfere them with “webcare.” Van Noort and Willemsen (2012, p. 133) define webcare as “the act of engaging in online interactions with [complaining] consumers, by actively searching the web to address consumer feedback.” Even though webcare is generally regarded as an effective way to anticipate negative effects among (relatively uninvolved) observers (Lee and Song, 2010), literature implies that (involved) complainants’ brand-related reactions to recovery attempts can differ substantially, ranging from significant improvements in customer attitudes – when recovery efforts are perceived as helpful – to a magnification of the initial negative opinion or emotion (e.g. anger) – when the company’s complaint handling efforts are regarded as inappropriate or insincere (Homburg and Fuerst, 2005; Kaltcheva et al., 2013). Further, it has been shown that the same recovery strategy that proves effective for some consumers can be ineffective for others (Grégoire et al., 2009) and that complainants’ reactions to complaint handling can vary even more dramatically online (Fournier and Avery, 2011).

This research argues that these variations in recovery effectiveness can be explained by the complainant’s “webcare receptiveness” that is the extent to which a complainant desires and is favorably disposed to a corporate response that addresses the cause of the online complaint. Accordingly, two types of online complainants who differ in their receptiveness toward online service recovery attempts can be identified: Constructive complainants – whose personal or collective complaining goals can only be achieved by receiving a response from the company and who are therefore open (receptive) to webcare; and Vindictive complainants – who do not need a response from the company to achieve their goals and who regard webcare rather as an inappropriate interference in between-consumer-conversations.

While academic literature has long recognized the benefits of classifying complainants for improving recovery effectiveness (Richins, 1983), knowledge about how to effectively provide webcare to different online complainant-types is still scarce. Consequently, the contribution of this article is two-fold: first, it investigates the nature and characteristics of different types of online complainants by exploring the varying personal and collective goals they want to achieve through voicing NeWOM; second, more importantly, it identifies the role of complainant-types for determining variations in the reactions to webcare interventions. These insights are crucial to protect a brand’s image in times of critical consumers who are empowered to publicly voice their dissatisfaction on social media.

2. Conceptual background and hypotheses

2.1 Online complaining

Service failures are mistakes or problems that consumers experience while purchasing or communicating with a brand (Maxham, 2001). Such negative events lead to customer dissatisfaction and customer complaint behaviors (e.g. direct complaints), which often signal the collapse of the customer-brand relationship. The emergence of social media has not only enabled unsatisfied customers to share their experiences and to provide feedback to the involved brand but also to inform other consumers (Hennig-Thurau et al., 2010). Online complaining is an expression of dissatisfaction for the purpose of drawing attention to a perceived misconduct by a brand (or company) in order to achieve personal or collective goals (Einwiller and Steilen, 2015). These publicly visible consumer statements can lead to negative brand evaluations, deteriorated brand reputation and the dissemination of negative information among its observers (Chevalier and Mayzlin, 2006; Verhagen et al., 2013). On the other side, a growing number of studies demonstrates that online complaining is related to unfavorable brand-related reactions (e.g. future brand avoidance) by the complainants themselves. Nevertheless, some recovery actions (i.e. webcare) can help to mitigate such negative reactions (Kim et al., 2016). Scholars have long emphasized the need to categorize those who exhibit (online) complaining behaviors based on their individual characteristics (Richins, 1983; Singh, 1990a), intentions (Hagedoorn et al., 1999) or response styles (Schoefer and Diamantopoulos, 2009a, 2009b) to improve complaint handling by stimulating positive reactions (e.g. re-establishing brand-favorable attitudes and brand advocacy). In the online context, Lee and Song (2010) make a basic differentiation between “complainers” (who post complaints), “repliers” (who articulate their own opinion on the complaint problem) and “observers” (who only read the complaint or replies as interested, potential customers of the brand). Despite the growing research focusing on NeWOM observers (Weitzl and Hutzinger, 2017; Schamari and Schaefers, 2015), surprisingly little research examines complainants’ reactions and how webcare affects the attitudes and intentions of different types of online complainants.

In recent years, webcare literature (van Noort et al., 2014) has made considerable contributions to our understanding of online service recovery effectiveness. For instance, Grégoire et al. (2009) find that high-relationship quality customers feel betrayed when no recovery is offered. However, this feeling and their desire for revenge are greatly attenuated by an apology and a modest post-complaint recovery. In contrast, their research also demonstrates that low-relationship quality customers seem to be more calculative and instrumentally oriented: Here, only an expensive, high-recovery attempt has a mitigating effect on customer revenge over time. More recently, Weitzl and Einwiller (2018) identify three segments of online complainants (i.e. “constructive, unattached customers”; “constructive loyalists”; and “revengeful loyalists”) who differ dramatically in pre-failure relationship status, complaining desires and post-webcare reactions. Grégoire et al. (2018) make a distinction between “vigilante complainers” (who frame the problem as a task to be solved) and “reparation complainers” (who frame the problem as a personal and public affront to be avenged). Such findings suggest the need to consider different types of complainants and the specific circumstances of online complaining (e.g. online complainants are often the “victims” of multiple prior failures which regularly guides them to have vivid desires for revenge and brand avoidance; Bonifield and Cole, 2007) – when selecting the appropriate recovery strategy on the internet. The study at hand, extends this research by introducing a typology that differentiates complainants with respect to their underlying motives and the benefits sought when engaging in NeWOM. We argue that the motive’s goal element determines whether a person is receptive to webcare (i.e. a company’s attempt to restore the customer–brand relationship). Only if a response by the company helps to achieve the complainant’s individual goals, he/she is receptive to webcare and will react positively toward it. This assumption is implicitly supported by earlier literature (He and Harris, 2014; Weitzl and Einwiller, 2018). However, a complaint–response mismatch is likely to lead to unfavorable complainant reactions. Consequently, it is suggested that complainant-type – “constructive” vs “vindictive” complainants – should be considered as a key moderator of webcare influence on complainant reactions (Figure 1).

2.2 Types of online complainants

2.2.1 Constructive complainants

These online complainants voice NeWOM because they have personally experienced a service failure and regard online complaining as a reasonable way to rebalance their relationship with the brand. Accordingly, these complaints are accompanied by the desire to initiate (or continue) a communication process with the affected company for drawing its attention to a specific misconduct and to ultimately achieve a solution for the problem (e.g. an unsatisfactory purchase; unfriendly staff). Therefore, “constructive complainants” (CC) are receptive to webcare. They typically seek both social benefits in the form of brand communication (e.g. explanation and apology) and particularly economic benefits such as compensation offerings for their financial loss (e.g. refund and replacement) (Cambra-Fierro et al., 2015). Redress-seeking (i.e. consumers utilize online complaints to compensate their dissatisfaction by demanding a justice-restoring recovery) is a classic company-dependent complaining goal (Singh, 1990b). CC anticipate that complaining on social media is more effective than voicing a complaint in traditional channels (e.g. telephone and mail) as it empowers the consumer by increasing the (public) pressure on the brand (van Noort and Willemsen, 2012). Consequently, CC address the responsible company directly and not predominantly the general public (i.e. NeWOM observers). CC hold the strong belief that webcare is an adequate means to achieve goals that restore their customer–brand relationship and they assume that the parent company is capable of providing sought benefits required for this rebalance. Another goal of CC is to help the company with a complaint to improve its products and services (Hagedoorn et al., 1999). Ro (2015) labels this constructive criticism “friendly complaint.” Here, engaged complainants wish webcare as they want to be taken seriously and benefit from being respected and valued as a feedback-giving customer. Considering CC goals that are beneficial to the brand as they dispose individuals to view webcare as a means to derive complaining benefits. Hence:

H1.

Constructive complainants are guided by webcare-dependent goals.

2.2.2 Vindictive complainants

For vindictive complainants (VC) online complaints serve as means to reduce frustration and anxiety associated with a perceived misconduct or service failure. Here, complaints primarily enable consumers to reduce the discontent associated with negative consumption emotions by venting (Hennig-Thurau et al., 2004; Willemsen et al., 2013). They also regard NeWOM as deliberate actions to take revenge on the respective company and to punish or harm it (Grégoire et al., 2009; Ward and Ostrom, 2006). VC may also articulate and spread NeWOM to help other consumers and warn them about the brand and to call for collective retaliation against it (e.g. boycotts). Various researchers (Bronner and Hoog, 2011; Hennig-Thurau et al., 2004) suggest that, warning other consumers as an act of altruism is a key driver of NeWOM. Complaints can also be a form of amusement or entertainment (i.e. having fun and relaxation through interacting with others) at the cost of the brand (Bronner and Hoog, 2011). In sum, above complaining goals suggest that VC do not address the brand directly, but rather speak to the great number of observers to whom the complainant wants to unhesitatingly promote his/her negative sentiments. The complainants’ sought social benefits (e.g. positive recognition from fellow consumers) (Dholakia et al., 2004) motivate their engagement. They have little interest in repairing their relationship with the brand that seems destructed by the company. They have no desire to receive webcare because it would not help them to achieve their intended goals and is rather perceived as an intrusion in consumer-to-consumer conversations (Fournier and Avery, 2011). Given their webcare-independent quest for benefits, therefore:

H2.

Vindictive complainants are guided by webcare-independent goals.

2.3 Complainants’ reactions to webcare

2.3.1 Webcare response-types

Service failures and complaint handling are both critical moments of truth in the relationship between the brand and its customers (de Matos et al., 2013). They offer companies opportunities to communicate with their clients and to restore compromised brand relationships – if well-handled and addressed to the right customer (Kuo and Wu, 2012). Service recovery refers to the brand’s response and process in solving problems that result from service failures (Wuen et al., 2004). In the online context, such efforts are summarized under the term “webcare,” which includes all reactive actions aiming at influencing complainants’ (and/or NeWOM observers’) evaluations of the negative event and at restoring a brand’s image. Response strategies typically range on a continuum of the responsibility taken by the brand from defensive to accommodative responses (ACCs) (Coombs, 2007): In giving a “defensive response” (DEF) the company either insists that there is no problem, accuses the complainant of wrongdoing or shifts the blame to others, thus taking no or limited responsibility for the service failure. It typically includes “external explanations” by making factors beyond the company’s influence responsible for the dissatisfying experience (Lee and Cranage, 2014). In an ACC the company publicly acknowledges the problem and takes (at least some) responsibility for the failure. ACC can include various complaisant signals including moderately accommodative actions (e.g. a clarifying communication and apology) to highly ACCs conveying multiple cooperative signals (e.g. a detailed explanation plus a redress offer). Extant service literature differentiates between psychological and tangible recovery: the former is typically applied to rectify the problem triggered by the failure and to restore customer satisfaction (Kuo and Wu, 2012) by offering social benefits such as an apology, empathy and a sincere explanation. Tangible recovery, on the other hand, refers to a manifest compensation provided to the dissatisfied customer for reducing experienced damages by means of free services, refunds, gifts, discounts and coupons. Online, ACC regularly includes both social/psychological and economic/tangible benefits. Lee and Song (2010) propose a third strategy. the “no-response” (NOR). This is the attempt to keep the brand away from the problem by remaining silent (i.e. ignoring the complaint), making meaningless comments or engaging in futile or short-sighted, unhelpful reactions (e.g. requesting a direct contact via another complaint channel).

2.3.2 Complainant-type as a moderator of webcare effectiveness

Complainants often show varied responses to the different types of webcare (van Noort and Willemsen, 2012). To better understand these differences, the research at hand, builds on the “congruence approach” (Mahajan and Churchill, 1988). This approach suggests that when two factors match (i.e. desire to receive webcare and a company’s webcare response), outcome levels (i.e. complainants’ reactions) will be enhanced or be greater than when the two factors are mismatched. Consequently, the favorability of complainants’ reactions depends on the extent to which the nature of the benefits provided through webcare matches the benefits sought by the complainant or in other words, the complainant-type.

It can be speculated that CC perceived loss following a service failure can be offset by a gain from matching webcare. More specifically, it is here argued that webcare-responses are more likely to cause more favorable brand-related outcomes among CC (i.e. attitudes guiding complainants’ brand approach behaviors) and help to mitigate negative ones (i.e. attitudes guiding complainants’ brand avoidance behaviors) than others because of their better matching to complainants’ desires. Extant literature supports this assumption (Chang et al., 2015). An explanation can be found in “equity theory” (Adams, 1963), which implies that people try to reduce distress by restoring either their physical or psychological equity when they experience unfairness. Constructive complaining represents such an attempt to rebalance a person’s relationship with a brand. This rebalance can be achieved by obtaining specific social/psychological (e.g. apology and explanation) and economic/tangible benefits (e.g. refunds and discounts) typically conveyed in webcare. Monetary compensations – often a major component of ACC, are particularly effective as such economic gains help to improve the output-to-input ratio in the relationship (internal equity) and the balance of the output-to-input ratio of the customer and that of the brand (external equity) (Estelami, 2000).

In terms of benefits offered to CC, webcare-responses range from low (NOR) via mid (DEF) to high benefits (ACC) (del Río-Lanza et al., 2009). ACCs regularly provide both social and economic benefits that match CC complaining goals, and hence, should be the most effective strategy to stimulate pro-brand reactions. Extant literature has shown that monetary compensations, for instance, coupons, discounts and refunds, help to restore “distributive justice” (i.e. the perceived fairness of actual or tangible outcomes relative to inputs) (Kuo and Wu, 2012; McColl-Kennedy and Sparks, 2003). Psychological recovery (e.g. an apology) is generally linked to “interactional justice” (i.e. the perceived fairness of the manner in which consumers are treated during the process of recovery efforts) (Goodwin and Ross, 1992), while Tax et al. (1998) also show that an apology (i.e. a social benefit) can trigger distributive justice. Accordingly, ample research shows that ACCs offer various distress-reducing benefits (Lee and Song, 2010) and can effectively reduce feelings of aggression (Conlon and Murray, 1996), which consequently leads to more positive complainant reactions. Corporate responses accepting responsibility for a negative incidence also facilitate consumers’ trust, which fosters favorable brand evaluations and increased purchase intentions (Lee, 2005). In contrast, DEF – denying a brand’s responsibility – are more likely to increase CC perceptions of unfairness, and hence, may escalate the problem. Past research shows that a company’s rejection of responsibility for a negative event often triggers negative perceptions toward the involved brand (Lee, 2005). However, one can expect that among CC mismatch between expectations and webcare outcomes is largest when they receive a NOR, which conveys nil or negative value for the complainant who expects a clarifying response (McLaughun et al., 1983). This is consistent with earlier empirical findings (Lee and Song, 2010; van Noort and Willemsen, 2012). While DEF can certainly include an element of aggression toward the complainant, it can still demonstrate the brand’s responsiveness to the complainant’s request and can signal its willingness to clarify the circumstances of the failure. In contrast, VC perceive webcare as an invasive attempt to impede NeWOM. Hence, all forms of webcare have the same, marginal value. Thus:

H3.

For constructive complainants, accommodative webcare is more effective than no-responses in (a) enhancing webcare reactions, (b) increasing favorable post-webcare brand outcomes and (c) mitigating unfavorable post-webcare brand outcomes.

H4.

For constructive complainants, defensive webcare is more effective than no-responses in (a) enhancing webcare reactions, (b) increasing favorable post-webcare brand outcomes and (c) mitigating unfavorable post-webcare brand outcomes.

H5.

For constructive complainants, accommodative webcare is more effective than defensive webcare in (a) enhancing webcare reactions, (b) increasing favorable post-webcare brand outcomes and (c) mitigating unfavorable post-webcare brand outcomes.

H6.

For vindictive complainants, no-responses, defensive and accommodative webcare are similarly effective in (a) enhancing webcare reactions, (b) increasing favorable post-webcare brand outcomes and (c) mitigating unfavorable post-webcare brand outcomes.

It is further hypothesized that NORs – a common reality online (Einwiller and Steilen, 2015) – lead to more detrimental effects among CC awaiting a corporate reaction than among VC for whom other customers’ supportive reactions are more desirable or NeWOM is a means to an end (e.g. venting) (Weitzer et al., 2007). In contrast, for CC, NORs generate the negative impression that the company disregards complainants’ interest to restore the brand-customer relationship. According to “justice theory” and “cognitive appraisal theory,” such recovery attempts that are perceived as highly unjust result in strong emotional (e.g. frustration) and cognitive outcomes (e.g. dissatisfaction) (DeWitt et al., 2008). Hence:

H7.

For constructive complainants, no-responses lead to less (a) favorable webcare reactions, favorable post-webcare brand outcomes and (c) more unfavorable webcare brand outcomes than for vindictive complainants.

3. Research method

3.1 Study 1: Online survey

3.1.1 Research design and participants

For hypotheses testing, two studies were conducted. First, a survey approach was chosen to examine naturally occurring responses of the population of interest (i.e. customers who had recently experienced a service failure and an online recovery encounter). Surveys are a common methodological approach in the service recovery literature (Tax et al., 1998; Schoefer and Diamantopoulos, 2009a, 2009b), as they provide insights into real customer reactions following a genuine service failure and its aftermath. As online complaining is not a common phenomenon, a random sample of the general population would have been unlikely to result in a significant number of respondents with such experiences. Consequently, the common practice of applying purposive sampling was followed (Menon and Dubé, 2004; Tax et al., 1998; Schoefer and Diamantopoulos, 2009a, 2009b). Specifically, a standardized online questionnaire was administered to members of an international consumer panel. For participation, respondents had to be engaged in online complaining on the Facebook or Twitter page of a company. To reduce recall-related distortion problems, the time frame was set to a maximum of six months (Schoefer and Diamantopoulos, 2009a, 2009b). The survey was structured as follows: qualified participants were first asked to provide details about the dissatisfying incident (e.g. failure severity), followed by questions regarding their pre-failure relationship with the brand, their pre-webcare attitudes and their personal perceptions of the brand’s responses to their complaint. Questions pertaining to various post-webcare attitudes and habits followed. The study ended with socio-demographic questions. On average, it took the respondents 14 min to complete the questionnaire.

To enhance the generalizability of the results, we invited panel members from the USA and Germany between 16 and 75 years of age to participate. All participants who completed the survey obtained a small monetary compensation. After ensuring homogeneity among participants of the two countries in their key characteristics (e.g. internet usage, customer–brand relationship status, failure type and complaining history), the data were pooled for further analyses. In total, 812 valid responses (n_US = 432; n_DE = 380) were used for data analysis. The average age was 35 (SD = 11.40), 52 per cent were male. The majority of the participants were white-collar workers (69 per cent) and completed higher education (more than 50 per cent held at least a bachelor’s degree). On average, respondents spent 6.80 h per day online. This resembles the average social media user (Pew Research, 2013). The majority complained publicly online because of a service failure triggering a financial loss. Specifically, 21.6 per cent participants experienced a problem with a defective or malfunctioning product followed by 19.7 per cent participants who had problems related to a poor service quality. In addition, 59.5 per cent participants had filed a complaint directly to the company via traditional channels (e.g. telephone and mail) before complaining on social media. Webcare was desired by 75.5 per cent while a webcare-response was received by 67.7 per cent of all participants.

3.1.2 Measurements

Established scales from academic literature were used to measure the key constructs. To some, minor adaptions were necessary to fit the research context. For the German version, a forward-backward translation method with four translators was applied. Appendix 1 provides an overview of measurements for the complaint goals (measured on five-point Likert-scales ranging from 1 “I strongly disagree” to 5 “I strongly agree”), while Appendix 2 summarizes the psychometric properties of the scale used to assess complainants’ reactions. To satisfy estimate-to-sample requirements, two measurement models were evaluated by means of confirmatory factor analysis (CFA). Overall, a satisfactory model fit was obtained for both the models and also the psychometric properties pointed toward an appropriate measurement on the item-level. For instance, all factor loadings for the complaining-goal-model (Χ² = 810.20, df = 247, GFI = 0.93, AGFI = 0.91, CFI = 0.96, RMSEA = 0.05, SRMR = 0.04) ranged from 0.70 to 0.86, thus ensuring construct validity (Shimp and Sharma, 1987). Further, all composite reliability values were above the 0.70 threshold (Hair et al., 2006) indicating internal consistency. Convergent validity was supported because the average variance extracted (AVE) for all constructs exceeded the 0.50 threshold (Fornell and Larcker, 1981). The square root of the variance shared between a construct and its items were greater than the correlations between the construct and any other construct in the model (Appendix 3) indicating discriminant validity (Fornell and Larcker, 1981).

Complainant reactions were measured (reflectively) as follows: Webcare satisfaction [four items; Brock et al. (2013)]; brand satisfaction [i.e. complainant’s post-webcare brand evaluation given pre-webcare expectations; four items; Ranaweera and Jayawardhena (2014)]; brand image [i.e. complainant’s associations with the brand; four items; Van Noort and Willemsen (2012)]; brand loyalty [i.e. complainant’s tendency to purchase the brand in future; four items; Chaudhuri and Holbrook (2001), Zhou et al. (2014)]; positive WOM intention (PWOM) [i.e. the extent the complainant intends to say positive things about the brand to others in future; three items; Zeithaml et al. (1996)]; and NWOM [i.e. the extent the complainant intends to say negative things about the brand to others in future; four items; Zhou et al. (2014)]. All constructs were measured on five-point scales. For assessing the reliability and validity of these measures CFA was also applied. The final measurement model had a satisfactory fit (Χ² = 460.37, df = 194, GFI = 0.96, AGFI = 0.95, CFI = 0.98, RMSEA = 0.04, SRMR = 0.04). Again, all items loaded significantly (p < 0.001) and strongly (≥0.75) on their respective constructs ensuring construct validity (Appendix 2). Convergent validity was supported by the factors’ high AVEs (≥0.60) and CRs (≥0.86). Again, a variance-extraction test ascertained discriminant validity among the six constructs (Appendix 4). For further analysis, composite scores for each construct were calculated.

The complainant-type was assessed by asking participants to indicate whether they had desired to receive a recovery response as a reaction to their negative online comment [i.e. webcare receptiveness; two binary items; adapted from Willemsen et al. (2013)]. Webcare-type was measured (formatively) with 21 items (Cambra-Fierro et al., 2015) pertaining to the extent of defensive and accommodative signals (e.g. “The company offered me a discount.”). All covariates (e.g. failure severity) were also measured with reflective, multi-items scales taken from established literature and satisfied all psychometric requirements.

3.1.3 Non-response and common method bias

For assessing whether there was any evidence for a non-response bias among, a comparison between early and late participants was conducted (Armstrong and Overton, 1977). Specifically, a series of t-tests for independent samples was used on the key variables (i.e. complaining goals, failure perceptions and complainants’ reactions). These analyses did not reveal any significant differences (at the 5 per cent significance level) between the two groups, which provided evidence that no-response error was unlikely to be a major concern in this survey.

Because it was not possible to rule out that the results may be distorted by common method bias as the constructs were measured from the same source (Podsakoff et al., 2003), this study first conducted a single-factor CFA for all the reflective constructs. The results showed an unacceptable fit with the total variance explained by the single factor being <25 per cent in the two models. Second, a common latent factor analysis (Podsakoff et al., 2003) was applied. Here, all observed variables in the measurement models were loaded on a common method construct besides loading on their respective latent construct. The models indicated that the variance attributable to the common method factor was less than 5 per cent. Finally, the standardized loadings for each observed variable in the original models (without a common latent factor) were compared with the model including a common latent factor. This comparison showed very little differences concerning the standardized loadings suggesting that common method bias was not a problem in this study.

3.2 Study 2: Online experiment

In the second stage, a three webcare-type (NOR vs DEF vs ACC) × 2 complainant-type (CC vs VC) between-subjects online experiment was performed to cross-validate the recovery response-related hypotheses (H3-H7). This approach was chosen to increase internal validity by controlling the manipulated variables, standardization and reducing “noise” with an experimental setting. Furthermore, the use of a service failure scenario was assumed to reduce biases from memory, rationalization tendencies and consistency factors, which are possible to bias retrospective self-reports (Study 1). Subjects were chosen from a database that consisted of 600 emails belonging to university students from Austria. A total of 214 (former) undergraduate and postgraduate students agreed to participate in the study and were randomly assigned to one of the six experimental groups. All participants were exposed to a scenario in which they were told that following an online shopping experience with a fictitious company, a service problem (i.e. an underperforming product) occurred which guided them to complain publicly on the company’s Facebook brand page. The CC were told that with their online complaint they wanted to achieve specific webcare-dependent goals, while the VC had to imagine that brand-hostile complaining goals directed to fellow consumers (e.g. warning others) were the key motivators for their complaining behaviors. All participants had to write an online complaint (posting) and were then – depending on their experimental group membership – confronted with one of the three webcare-response strategies. Following this, respondents were asked to answer various questions pertaining their individual reactions (by using the same measures as in Study 1; see Appendix 2 for (satisfactory) psychometric properties of Study 2’s key measures) and their failure perceptions (i.e. controls). Subjects were dropped from further analysis if they did not experience online shopping within the past six months, were inactive on social media or found the experimental setting unrealistic. Checks on the final sample ensured that the manipulation of the two independent variables was successful: Specifically, as in Study 1, complainant-type was assessed with two binary variables (e.g. “When you made your complaint, did you want to receive a direct answer from the company online?” [yes/no]) capturing webcare receptiveness (Willemsen et al. (2013). Only respondents who consistently answered this questions correctly – according to their group membership – entered the final sample. Additional analysis revealed that webcare-unreceptive individuals (VC) had higher revengeful complaining motivation than webcare-receptive individuals (CC) [M_VC = 3.10, M_CC = 2.73; t(193) = −3.09, p < 0.01]. A similar procedure was applied to the manipulation of the second independent variable (i.e. webcare-type): Here, participants had to correctly respond to a single-choice question asking them to indicate the webcare-type received in the scenario (NOR/DEF/ACC). At the end, 195 questionnaires were usable to cross-validate the findings of Study 1. The mean age was 27 (SD = 5.34), 62 per cent were female.

4. Results

4.1 Main results (Study 1)

4.1.1 Explaining complainant-types

To test H1-H2, a hierarchical logistic regression was performed with complainant-type as the binary dependent variable (0 = webcare unreceptive; 1 = webcare receptive) (Table I).In Model 1, control variables (gender, age, education, country, complaint history prior to online complaining and pre-failure relationship strength) were introduced. A test of the full model against the constant-only model showed a significant increase, indicating that the predictors were able to distinguish those receptive to webcare from those who are not (Χ² = 40,373, df = 6, p < 0.001). Nagelkerke’s R² was 0.31 indicating a relatively high relationship between prediction and grouping. Overall prediction success was 76.2 per cent (unreceptiveness = 9.8 per cent, receptiveness = 92.1 per cent). The Wald criterion demonstrated that all variables (except gender) were (at least marginal) predictors of the complainant-type and yet, partially supported the typology of online complainants (all p ≤ 0.10). In Model 2, complaining goals were introduced. A test of the full model was again significant, indicating that the predictors distinguished between webcare receptiveness and unreceptiveness (Χ² = 55,310, df = 14, p < 0.001). Nagelkerke’s R² was 0.62, indicating a significant increase in the model’s predictive power compared to Model 1. Overall prediction success was 79 per cent (unreceptiveness = 28.6 per cent, receptiveness = 98.6 per cent). These results show that the prediction success of both complainant-types increased by introducing complaining goals. The Wald criterion implied that all webcare-dependent goals (i.e. redress-seeking, empowerment, helping the company and recognition by the company) were significant contributors to constructive complaining (all p ≤ 0.05). This supported H1. Further results show that vindictive complaining was significantly predicted by altruism toward others, entertainment and particularly revenge-taking. The odds ratio indicated that when revenge taking increases by one unit, the odds that the complainant engages in vindictive complaining are tripled (odds = 3.15). However, venting did not have any effect on webcare unreceptiveness. Therefore, H2 was only partially supported.

4.1.2 Explaining complainants’ reactions

Multivariate analysis of covariance (MANCOVA) followed by a series of univariate comparisons was used to test H3-H7. MANCOVA’s dependent variables were webcare- and post-webcare complainants’ reactions (e.g. PWOM intention). Webcare-type (NOR, DEF and ACC) and complainant-type (VC and CC) served as the two fixed factors. Further, failure severity, complaint history and pre-webcare relationship strength were included as covariates. Because of unequal group sizes, the sample was bootstrapped with 1,000 replications (Hayes, 2013). Preceding analyses revealed that the six quasi-experimental groups did not differ in various aspects that may affect complainants’ reactions toward the involved brand including socio-demographics (sex, education, age and occupation), social media usage, failure type, failure attributions (e.g. blame, controllability) relationship length, prior brand experiences and evaluations, feelings (e.g. anger) perceived and attitude toward (offline/online) complaining.

Results of initial multivariate analysis show significant main effects of webcare [Wilk’s λ = 0.88, F(121, 538) = 8.13, p < 0.001] and complainant-type [Wilk’s λ = 0.97, F(6,739) = 3.93, p < 0.001] and a significant two-way interaction between webcare and complainant-type [Wilk’s λ = 0.88, F(121, 538) = 8.25, p < 0.001] indicating a consistent pattern across the outcome variables (Table II). To test H3-H6, a series of ANCOVAs analyzing group-differences in the six post-webcare outcomes (e.g. brand image) was conducted. Again, the sample was bootstrapped with 1,000 replications to control for differences in group sizes (Table II).

In support of H3a and H3b, results revealed that among CC ACCs consistently lead to significantly more positive webcare evaluations (i.e. satisfaction) and more favorable brand reactions (i.e. brand satisfaction, attitude, loyalty and PWOM) than NOR (Table III; all p < 0.05). Similar patterns were found in support of H4a-H4b and H5a-H5b such that among CC favorable outcomes were described best by ACC > DEF > NOR [Figure 2(a) illustrates this pattern exemplified by brand image]. However, in contrast, to the expectations, webcare-types evoked different patterns when it came to the mitigation of unfavorable brand outcomes (i.e. NWOM): Here, ACC (M = 3.55, SD = 1.17) and NOR (M = 3.39, SD = 1.07) lead to a similar level of NWOM intention (p = 0.90). In contrast, DEF caused significantly higher intentions to badmouth the brand in the future (M = 4.50, SD = 1.06) dominating the other two response-types [Table III and Figure 2(b)]. Therefore, H3c-H4c were not supported.

H6 claims that VC are immune to webcare efforts and react homogeneously to marketers’ recovery efforts. In line with H6a-H6c, we found the consistent pattern of insignificant mean-differences in all outcome variables across the three webcare-types (Table II). Finally, H7a-H7c suggests that complainant-type moderates the effectiveness of NOR such that for CC this webcare-type is significantly less beneficial than for VC as it conveys the least benefits. Empirical results supported this argument to a large extent: all webcare and favorable brand-related reactions to NOR were significantly determined by a corresponding moderation effect of complainant-type. For instance, planned comparisons revealed that CC had a significantly lower brand image (M_CC = 2.51, SD_CC = 1.21) than VC [M_VC = 3.08, SD_VC = 1.47; t(308) = 3.93, p < 0.001]. Hence, H7a-H7b were supported. However, there was no empirical evidence for H7c as no significant difference between the two complainant-types (M_CC = 3.39, SD_CC = 1.06, M_VC = 3.36, SD_VC = 1.08) emerged [t(308) = 0.02, p > 0.05].

Supplemental analyses demonstrated that when comparing complainant-types, ACC significantly increased webcare and favorable brand-related reactions of CC across all outcome variables when compared to VC. In contrast, DEF (medium benefits) lead to similar positively valenced outcomes regardless of the complainant-type. For example, planned comparisons for complainant-type showed that CC receiving an ACC had a more favorable brand image (M_{ACC_CC} = 4.44, SD_{ACC_CC} = 0.97) than VC (M_{ACC_VC} = 3.71, SD_{ACC_VC} = 0.95) [t(263) = 3.70, p < 0.001]. However, brand image was not significantly different across complainants when receiving a DEF (M_{DEF_CC} = 3.50 vs M_{DEF_VC} = 3.40) [t(234) = 0.42, p > 0.05]. Besides brand image, this finding was stable across webcare satisfaction, brand satisfaction, loyalty and PWOM intentions.

4.2 Cross-validation results (Study 2)

Table IV summarizes the findings of the experimental study. Overall, by applying a similar analysis approach as for Study 1, the experiment provided sound empirical evidence for the hypothesized reactions toward webcare of VC (H3-H5). Nevertheless, also the cross-validation sample pointed toward the particularity in respect to complainants’ willingness to voice NWOM after receiving DEF: as in Study 1, DEF lead to a significantly higher intention to say negative things about the brand to known and unknown others than NOR and ACC, which resulted in similar NWOM levels. This demonstrated once again the danger of shifting the blame to others by means of defensive webcare, which ultimately triggers strong revenge desires. In respect to CC reactions, the experiment provided additional support for H6 (i.e. homogeneity of webcare-types’ impact) meaning that webcare unreceptive individuals are neither positively nor negatively affected by online complaint handling. In contrast to the survey study, however, H7 (except H7c) found empirical support (all p’s < 0.01) indicating that for CC NOR is less beneficial than for their destructive counterparts which leads, but not to higher revenge.

5. Conclusion

5.1 General discussion

5.1.1 Identifying complainant-types

Following service failures, dissatisfied consumers increasingly turn to social media to share their negative experiences, thoughts and emotions about a brand with the involved company and other consumers (i.e. NeWOM bystanders). To restore the relationship with their clients and to impede unfavorable brand-reactions among potential customers, many companies wish to intervene these public online complaints by means of “webcare” (i.e. online complaint handling). However, marketers often fail to react effectively as they are often guided by a high degree of uncertainty about the positive (negative) consequences of adequate (inadequate) responding to online criticism. This fear is justified as webcare perceived as inappropriate has been shown to backfire on a company and can undermine its intended effects (Lee and Song, 2010). Therefore, companies often prefer to remain silent and often forgo the last opportunity to win back their dissatisfied customers. In this context, this study investigates the possibilities for brands to use webcare for fostering favorable and mitigating unfavorable brand-related reactions among its dissatisfied customers who have chosen to complain online. While webcare was found to lead to heterogeneous complainant reactions, research has widely neglected the need to consider different types of complainants [see Weitzl and Einwiller (2018) for an exception]. The research at hand shows the differentiation between:

• CC – whose voice negative comments publicly on marketer-generated social media to restore their relationship with the brand after a service failure or have other brand-favoring goals (e.g. feedback); and

• VC – who strive for motivational goals that are more harmful to the brand (e.g. revenge taking) – helps to better explain the heterogeneous customer reactions to webcare.

5.1.2 Profiling complainants

CC are guided by a rich set of complaint goals enhancing their receptiveness to webcare. These complainants seek, amongst others, for redress (e.g. refunds) to restore their relationship with the brand. They use social media to increase the public pressure on the brand to find a solution, but they also regard online complaining as an act of constructive feedback that helps the brand to improve its performance (i.e. altruism toward the brand). CC use various other complaint channels before going online and have per se, a relatively strong customer–brand relationship before the service failure happened. In contrast, VC regard online complaining as a possibility to interact with other consumers and not with the brand itself. This is mirrored by various webcare-independent complaint goals, such as the desire to take revenge and to warn others. Given their receptiveness to webcare, which is grounded in dissatisfied customers’ individualistic and collectivistic complaint goals, this study demonstrates that CC are particularly sensitive to specific types of webcare responses that match their needs best.

5.1.3 Constructive complainants’ reactions

For CC different response strategies are perceived to vary extensively in their social/psychological and economic/tangible benefits offered to the complainant. As expected – given their justice-restoring complaining goal – CC respond most positively to ACCs because they are congruent with their individual desires (e.g. redress seeking). Results point to the very consistent pattern that (on average) NORs – that is, ignoring CC well-meant intention to solve the problem – lead to the most negative complainant responses (e.g. unfavorable brand image). It seems that NOR has a negative value for CC as their wish to restore and continue their relationship with the brand remains completely unappreciated by the involved company. This contrasts to previous findings (Lee and Song, 2010) indicating the importance of considering differences in the preferences of the communication targets (complainer vs observer).

For CC, DEF is – on average – not the most inappropriate response as they seem to still provide at least some clarifying information about the circumstances of the failure which satisfies problem-focused complainants. However, this research also highlights an unexpected reaction of CC to DEF: findings suggest that this specific webcare-type is particularly inappropriate to mitigate unfavorable, outward-directed brand-related reactions or more specifically, post-webcare NWOM (i.e. complainants’ willingness to badmouth the brand; NWOM). DEF is complaint handling which is incongruous to CC’ demands and which ultimately increases their strong willingness to share negative information about the brand both privately and publicly. In line with “attribution theory” and earlier research (Chang et al., 2015), DEF – responses that deny the brand’s responsibility for the failure by shifting the blame to others – seem to make the company to ultimately appear more guilty. Thus, CC view DEF as inappropriate excuses to whitewash the brand, which makes them to attribute the failure to internal and controllable problems that were preventable by the company. Research demonstrates that internal failure attribution is linked to NWOM (Hess, 2008). In addition, “cognitive appraisal theory” suggests that individuals’ evaluation of a situation triggers emotions (del Río-Lanza et al., 2009). When the response is not perceived as a remedy for the situation, it elicits negative emotions (e.g. anger) that are often vented through NWOM.

5.1.4 Vindictive complainants’ reactions

VC, in contrast, is shown to be immune against any type of webcare as these responses do not facilitate VC complaining objectives (e.g. harming the brand). Here, this research is able to demonstrate that marketers are rather powerless to mitigate detrimental effects (e.g. NWOM). However, the findings also suggest that both (desired) NOR and brand-defending responses (i.e. DEF), which are sometimes regarded as triggers to escalate the problem (Lee and Song, 2010) – don’t necessarily have this strong negative effect. Rather, VC seems to have a stable (unfavorable) attitude toward the involved brand that is unaffected by a single brand message. VC is completely unaffected by resource-intensive ACC conveying both social/psychological (e.g. sincere apology, detailed explanation and promises) and economic/tangible benefits (e.g. refunds, coupons and free gifts). Consequently, CC respond more favorably toward ACC than VC. This pattern is consistent across a variety of webcare outcomes (e.g. brand attitude and loyalty). In line with the “congruence approach” for CC, ACC conveys a high level of benefits matching their individual complaint goals (e.g. redress seeking) while for VC these responses provide no benefit at all. Interestingly, for VC ACC, i.e. responses in which the company takes responsibility for the failure – do not represent a “proof” for their grudge against the brand as for other types of complainants (Hutzinger and Weitzl, 2018), which may further stimulate negative attitudes and behaviors. Results further suggest that both complainant-types respond relatively homogeneously toward DEF. This means that this response-type provides equal value for both complainant-types. However, NOR were found to be are particularly harmful to CC who have less favorable attitudes following such a response than VC.

5.2 Theoretical implications

First, the investigation extends the evolving literature on online complainants (Ro, 2015) by identifying two complainant-types (CC vs VC) who dramatically vary in their complaining goals, benefits sought and their webcare receptiveness. Literature provides rich evidence that consumers have different motives when engaging in electronic word-of-mouth (eWOM) in general (Hennig-Thurau et al., 2004) or online public complaining (NeWOM) in particular (Sparks and Browning, 2010). However, no research has analyzed how these goals contribute to the rise of specific online complainant-types yet. This research makes the first step in this direction by segmenting online complainants (a-priori) according to their motivational background. It provides new insights for the relationships between complainants, their goals, behaviors and desires (i.e. demanding a reaction from either the brand or other consumers). Thus, the study provides new insights into the existence of heterogeneous groups of complainants that voice their discontent on brands’ social media platforms.

Second, the study contributes to the literature on webcare and appropriate response strategies directed at positive complainants’ reactions (Weitzl et al., 2018) by providing an understanding of differences in the effectiveness of alternative complaint handling strategies. Earlier research shows that consumer responses to service recoveries are variable (Kaltcheva et al., 2013). In this research, this pattern is empirically corroborated in the social media context and demonstrate that the effectiveness of (perceived) benefits provided by webcare differs significantly across complainants. Here, the study takes a step forward in providing a theoretical understanding of why these differences emerge and how webcare effectiveness can be improved. Results suggest that webcare improvements are possible when considering two important aspects:

1. Complainants differ in their reasons why they engage in NeWOM (i.e., complainant-types), which makes them to differ also in the benefits sought when complaining.

2. Different webcare responses provide different types of benefits to customers (ranging from no to various psychological/tangible benefits).

By applying a contingency framework, this research is the first to empirically show that the type of complainant (i.e. does the complainant seeks a response from the brand or fellow consumers) and the type of response jointly determine complainants’ webcare and post-webcare reactions. The research demonstrates that even in a late stage of the complaining process (when complainants go publicly online) CC reactions can be positively influenced by adequate webcare.

Third, the study contributes to the consumer engagement literature (Brodie et al., 2013) by investigating webcare as a brand communication instrument to elicit non-transactional, outward-directed behaviors. Specifically, it offers new insights on how to elicit positive engagement (PWOM) among CC and, more importantly, how to prevent negative engagement (NWOM) by selecting an appropriate response strategy. By doing so, this research answers several calls for investigating the risk of webcare backfiring effects and their circumstances (Hennig-Thurau et al., 2010; Fournier and Avery, 2011). The literature on the topic remained limited yet. However, researchers suggest that the public acceptance of responsibility for negative events (e.g. by means of ACC), may make complainants and other consumers angry and frustrated (i.e. perceived negative emotions; Lee and Song, 2010). While this research cannot support this view, it provides novel insights concerning the negative cognitive and conative consequences of “inappropriate webcare.” However, findings suggest that DEF (i.e. responses that do not solve the complainant’s problem) are likely to trigger strong negative emotions among CC (e.g. anger) which are consequently vented by means of NWOM. This causal effect is also established in the revenge-taking literature (Grégoire et al., 2009).

5.3 Practical implications

The findings have important implications for marketers looking for strategies to increase webcare effectiveness and efficiency. A major conclusion is that the complainant-type and the desired outcomes (e.g. attitudinal vs behavioral) should be the key determinants guiding the selection of the appropriate recovery strategy. Only some complainants (CC) seek benefits appeasable via webcare, while others (VC) strive for benefits beyond the company’s influence (e.g. other consumers’ reactions). Hence, the study highlights the need to identify the complainant-type as follows:

• by detecting primary complaining goals (expressed explicitly or implicitly in their negative online comments); and

• by additional complainant characteristics.

The data reveal that CC have a stronger brand-relationship than VC before the service failure happened (Table I). Their complaint history is also longer, meaning that they had used more alternative complaint channels (e.g. email and telephone) and engaged more in relationship-restoration before turning to social media than VC. This implies that marketers are able to intervene in customers’ discord early in the recovery process offline, which helps to prevent negative post-webcare reactions. Most harmful to CC inward-directed reactions is a company’s strategic silence following a public complaint. CC react with very negative attitudes, which are, however, responsible for customer–brand relationship quality (e.g. post-webcare customer satisfaction) – to NOR. Consequently, companies should closely monitor their social media outlets for customer feedback. Having this said, NOR are unlikely to stimulate “backfiring effects” in terms of negative outward-directed reactions (NWOM). Here, results reveal that DEF can cause a “spiral of negativity” – not initiated by the seemingly more hostile VC, but by CC. Therefore, marketers should be aware of the “dark side” of webcare and should be cautious when defending the attacked brand against online criticism coming from a “friendly” complainant. Providing a problem solution and including accommodative signals (e.g. apology) and customer-oriented message characteristics (e.g. personalization) can, in contrast, be the more appropriate strategy to minimize post-complaint NWOM.

Webcare directed at VC, on the other hand, is ineffective meaning that there is neither a gain (i.e. the potential to ultimately convince the dissatisfied consumer even with highly accommodative signals) nor a loss (i.e. revenge actions following a company’s attempt to defend itself against an accusation) aside from invested efforts. VC react similarly to any type of recovery strategy. However, because webcare has the potential to steer brand reactions of recovery bystanders (Weitzl and Hutzinger, 2017), NOR may not be the best corporate response. Hence, companies should consider post-complaint communication to convey accommodative signals to a brand’s future customers. In such cases, webcare can create brand-acceptance among observers even though the original complainant doesn’t value these efforts.

6. Limitations and further research

Several limitations need to be considered when interpreting the results and designing future research. Firstly, this study focuses on online complaining on marketer-created social media sites. The literature demonstrates that consumers react differently toward webcare on independent platforms such as consumer-created discussion forums (Schamari and Schaefers, 2015). Future experimental research should address this issue by investigating reactions of different complainant-types to reactive and proactive webcare across alternative social media channels and complaining contexts. Secondly, complainants were classified according to their webcare receptiveness. Further research should identify additional descriptives (e.g. customer commitment), which may help to identify and profile different types of online complainants more profoundly. Thirdly, only three classic response-types were investigated. Yet, complaint-handling actions can be more multifaceted (Cambra-Fierro et al., 2015). Therefore, further research should consider to use a more fine-grained evaluation of different response-types and message contents as reactions to varying types of failures under given situations (e.g. varying levels of “secondary dissatisfaction” after multiple attempts to achieve a company’s compliance). Future research should also address the impact of other consumers’ reactions (e.g. comments) likely to influence the reactions of VC in particular. Finally, while this research examines various attitudinal and intentional outcomes of webcare responses, subsequent studies should pay more attention to emotional reactions that may elicit extreme situations such as online firestorms and behavioral patterns (e.g. purchasing) that translate into customer profitability.

Figures

Figure 1

Determinants of complainants’ reactions

Figure 2

(a) Interaction effect of webcare and complainant-type on post-webcare brand image and (b) Interaction effect of webcare and complainant-type on post-webcare NWOM

Table I

Results of the logistic regression analysis explaining complainant-type

Variables	Model 1		Model 2
	Coefficient (SE)	Odds	Coefficient (SE)	Odds
Control variables
Gender (f)	−0.27 (0.17)	0.77	−0.30 (0.19)	0.74
Age	−0.02 (0.00)**	0.98	−0.03 (0.01)***	0.97
Education	0.19 (0.08)****	1.21	0.22 (0.08)*	1.25
Country (Germany)	−0.35 (0.20) ****	0.70	−0.35 (0.22)	0.71
Complaint history	1.40 (0.08)***	3.42	1.39 (0.10)**	3.10
Pre-failure relationship strength	1.02 (0.10)***	2.50	1.00 (0.11)***	2.39
Complaint goals
Webcare-dependent goals:
Redress seeking			2.12 (0.10)***	10.57
Empowerment			1.55 (0.15)***	9.20
Helping the company			0.40 (0.17)*	1.49
Recognition by the company			0.50 (0.12)***	1.60
Webcare-independent goals:
Altruism toward other consumers			−0.30 (0.15)*	0.69
Venting			−0.10 (0.13)	0.90
Revenge taking			−1.17 (0.12)***	3.15
Entertainment			−0.37 (0.14)**	0.69
Negelkere R²	0.31		0.62
Model X²	40,373***		55,310***

Notes:

****p < 0.10;

*p < 0.05;

**p < 0.01;

*** p<0.001

Table II

Effects of webcare and complainant-type on complainants’ reactions (Study 1)

Sources	MANCOVA			Univariate
					Webcarereactions	Post-webcare brand-related reactions
						Favorable				Unfavorable
	Wilk’slambda	F	Sig.	df	Webcaresatisf.	Brandsatisf.	Brandimage	Brandloyalty	PWOMintention	NWOMintention
Main effects
Webcare-type WC	0.88	8.13	***	1	23.88***	31.05***	23.41***	23.04***	24.22***	7.00**
Complainant-type CT	0.97	3.93	**	2	0.34	0.29	0.11	0.00	0.16	18.14***
Interaction effect
WC × CT	0.88	8.25	***	2	13.64***	16.86***	14.41***	2.50***	23.91***	19.39***
Covariates
Failure severity	0.91	12.67	***	1	6.88*	38.51***	54.37***	50.21***	53.77***	7.48**
Complaint history	0.82	28.15	***	1	49.86***	29.44***	37.39***	49.31***	80.84***	94.58***
Pre-webcare relationship strength	0.84	25.34	***	1	58.96***	105.85***	109.25***	96.08***	76.90***	10.26***

Notes:

*p < 0.05;

**p < 0.01;

***p < 0.001

Table III

Complainants’ webcare and post-webcare reactions (Study 1)

	Complainant-type
	Vindictive1	Constructive2
Dependent variable: post-webcare brand satisfaction
Webcare
Accompanieda	3.652a,2c	4.391a,1b,1c, 2b,2c
Defensiveb	3.512a,2c	3.672a,2c
Noc	3.232a,2c	2.791a,1b,1c,2a,2b
Dependent variable: Post-webcare brand image
Webcare
Accompanieda	3.712a,2c	4.441a,1b,1c, 2b,2c
Defensiveb	3.402a,2c	3.502a,2c
Noc	3.082a,2c	2.511a,1b,1c, 2a,2b
Dependent variable: Post-webcare PWOM intention
Accompanieda	3.512a,2c	4.351a,1b,1c, 2b,2c
Defensiveb	3.372a,2c	3.312a,2c
Noc	3.452a,2c	2.541a,1b,1c, 2a,2b
Dependent variable: Post-webcare brand satisfaction
Accompanieda	3.572a,2c	4.301a,1b,1c, 2b,2c
Defensiveb	3.252a,2c	3.352a,2c
Noc	3.252a,2c	2.601a,1b,1c, 2a,2b
Dependent variable: post-webcare brand loyalty
Webcare
Accompanieda	3.512a,2c	4.321a,1b,1c, 2b,2c
Defensiveb	3.322a,2c	3.342a,2c
Noc	3.412a,2c	2.591a,1b,1c, 2a,2b
Dependent variable: post-webcare NWOM intention
Webcare
Accompanieda	3.292b	3.552b
Defensiveb	3.192b	4.501a,1b,1c, 2a,2c
Noc	3.372b	3.392b

Note:

Within each cell, superscripts indicate significant differences to other cells (p < 0.05 or better)

Table IV

Complainants’ webcare and post-webcare reactions (Study 2)

	Complainant-type
	Vindictive1	Constructive2
Dependent variable: webcare satisfaction
Webcare
Accompanieda	3.192a,2c	4.831a,1b,1c, 2b,2c
Defensiveb	3.162a,2c	4.022a,2c
Noc	2.992a,2c	3.081a,1b,1c, 2a,2b
Dependent variable: Post-webcare brand image
Webcare
Accompanieda	3.342a,2c	4.881a,1b,1c, 2b,2c
Defensiveb	3.072a,2c	3.852a,2c
Noc	2.742a,2c	2.751a,1b,1c, 2a,2b
Dependent variable: post-webcare PWOM intention
Webcare
Accompanieda	3.162a,2c	4.571a,1b,1c, 2b,2c
Defensiveb	3.042a,2c	3.632a,2c
Noc	3.122a,2c	2.781a,1b,1c, 2a,2b
Dependent variable: post-webcare brand satisfaction
Webcare
Accompanieda	3.222a,2c	4.721a,1b,1c, 2b,2c
Defensiveb	2.942a,2c	3.692a,2c
Noc	2.932a,2c	2.861a,1b,1c, 2a,2b
Dependent variable: post-webcare brand loyalty
Webcare
Accompanieda	3.152a,2c	4.751a,1b,1c, 2b,2c
Defensiveb	2.992a,2c	3.672a,2c
Noc	3.072a,2c	2.841a,1b,1c, 2a,2b
Dependent variable: post-webcare NWOM intention
Webcare
Accompanieda	2.962b	3.192b
Defensiveb	2.872b	4.051a,1b,1c, 2a,2c
Noc	3.072b	3.052b

Note:

Within each cell, superscripts indicate significant differences to other cells (p < 0.05 or better)

Table AI

Confirmatory factor analysis on complaint goals

Measurement items	λ	AVE	CR	α
Redress seeking (RS)		0.66	0.89	0.89
(Yen and Tang, 2015; Yoo et al., 2013)		(0.71)	(0.91)	(0.90)
RS1: I expected to get a compensation from the company	0.83 (0.84)
RS2: I wanted a monetary refund from the company	0.80 (0.82)
RS3: I wanted to be reimbursed for my expenses	0.80 (0.84)
RS4: I desired a compensation for my troubles by the company	0.83 (0.87)
Empowerment (EM)		0.53	0.70	0.68
(Hennig-Thurau et al., 2004)		(0.57)	(0.72)	(0.71)
EM1: One has more power together with others when complaining publicly online	0.74 (0.77)
EM2: My online complaint puts more pressure on the company	0.72 (0.74)
Helping the company (HE)		0.55	0.83	0.83
(Hennig-Thurau et al., 2004; Yen and Tang, 2015)		(0.61)	(0.86)	(0.84)
HE1: I wanted to help the company to improve	0.74 (0.78)
HE2: I wanted to give the company the opportunity to resolve the problem	0.70 (0.71)
HE3: I wanted to give the company the opportunity to demonstrate their customer service quality publicly	0.76 (0.81)
HE4: I wanted to inform the company about my negative experience	0.77 (0.82)
Recognition by the company (RE)		0.55	0.71	070
(adapted from Istanbulluoglu, 2017)		(0.60)	(0.75)	(0.74)
RE1: I wanted the company to appreciate me	0.74 (0.77)
RE2: I wanted the company to respect me	0.74 (0.78)
Altruism towards other consumers (AL)		0.52	0.76	0.74
(Hennig-Thurau et al., 2004)		(0.61)	(0.82)	(0.82)
AL1: I wanted to save others from having the same negative experiences as me	0.74 (0.80)
AL2: I wanted to help others with my own experiences	0.72 (0.80)
AL3: I wanted to give others the opportunity to choose the right company	0.70 (0.75)
Venting (VE)		0.52	0.76	0.76
(Hennig-Thurau et al., 2004)		(0.59)	(0.81)	(0.79)
VE1: My complaint helped me to shake off frustration	0.71 (0.75)
VE2: I wanted to get anger off my chest	0.75 (0.81)
VE3: I wanted to express my anger about my negative experience	0.70 (0.75)
Revenge taking (RT)		0.73	0.91	0.92
(Bronner and Hoog, 2011)		(0.72)	(0.91)	(0.91)
RT1: The company earlier harmed me – now I want to harm the company	0.87 (0.87)
RT2: I want to take vengeance upon the company	0.87 (0.88)
RT3: I want to punish the company in some way	0.83 (0.80)
RT4: I thought about ways to sabotage the company	0.86 (0.84)
Entertainment (EN)		0.68	0.86	0.86
(Bronner and Hoog, 2011; Hennig-Thurau et al., 2004)		(0.69)	(0.87)	(0.85)
EN1: It is fun to communicate this way with other people	0.79 (0.80)
EN2: It was entertaining to write the negative comment	0.86 (0.85)
EN3: I enjoyed it	0.82 (0.84)

Notes:

λ – Standardized factor loadings; AVE – Average variance extracted; CR – Construct reliability and α – Cronbach’s alpha; Results for Study 2 in parentheses

Table AII

Confirmatory factor analysis on complainant reactions

	λ	AVE	CR	α
Webcare satisfaction (WS)		0.73	0.93	0.91
(Brock et al., 2013)		(0.77)	(0.93)	(0.92)
WS1: I was satisfied with the handling of my complaint	0.92 (0.91)
WS2: I was very satisfied with the complaint handling of the company	0.85 (0.87)
WS3: I felt the company provided a satisfactory solution for my problem	0.86 (0.86)
WS4: Overall, I felt the service response from the company was good	0.85 (0.87)
Brand satisfaction (BS)		0.67	0.89	0.88
(Ranaweera and Jayawardhena, 2014)		(0.72)	(0.91)	(0.89)
BS1: very displeased – very pleased	0.80 (0.83)
BS2: very unfavorable – very favorable	0.83 (0.84)
BS3: very dissatisfied – very satisfied	0.81 (0.82)
BS4: very unhappy – very happy	0.86 (0.89)
Brand image (BI)		0.70	0.91	0.91
(van Noort and Willemsen, 2012)		(0.75)	(0.92)	(0.91)
BI1: The company is good	0.87 (0.90)
BI2: The company is trustworthy	0.82 (0.85)
BI3: The company is respectable	0.80 (0.82)
BI4: The company is favorable	0.85 (0.90)
Brand loyalty (BL)		0.72	0.95	0.94
(Chaudhuri and Holbrook, 2001; Zhou et al.,2014)		(0.75)	(0.92)	(0.90)
BL1: I will spend money at this company	0.88 (0.88)
BL2: I will select the company if needed in future	0.82 (0.84)
BL3: I will buy this company’s products the next time I have the opportunity	0.87 (0.89)
BL4: I intend to keep purchasing the company’s products	0.83 (0.85)
Positive word-of-mouth intention (PW)		0.65	0.86	0.84
(Zeithaml et al., 1996)		(0.72)	(0.89)	(0.89)
PW1: I will say positive things about this company to other people	0.88 (0.90)
PW2: I will recommend the company to someone who seeks my advice	0.81 (0.85)
PW3: I will encourage friends and relatives to do business with the company	0.76 (0.80)
Negative word-of-mouth intention (NW)		0.63	0.85	0.85
(Zhou et al.,2014)		(0.68)	(0.89)	(0.88)
NW1: I will spread negative word-of-mouth about the company	0.81 (0.83)
NW2: I will tell my friends not to purchase from this company	0.87 (0.89)
NW3: I will bad-mouth this company to my friends	0.79 (0.80)
NW4: I will advise other people not to buy the company’s products	0.75 (0.77)

Notes:

λ – Standardized factor loadings; AVE – Average variance extracted; CR – Construct reliability and α – Cronbach’s alpha; Results for Study 2 in parentheses

Table AIII

Descriptive statistics, correlation and discriminant validity of complaint goals scales

	Mean	SD	1	2	3	4	5	6	7	8
1 Redress seeking	3.48	1.08	(0.66)	0.27	0.19	0.44	0.21	0.28	0.42	0.42
2 Empowerment	3.78	0.84	0.52	(0.53)	0.51	0.51	0.50	0.49	0.21	0.15
3 Helping the company	3.87	0.78	0.44	0.72	(0.55)	0.53	0.46	0.35	0.04	0.03
4 Recognition by the company	3.56	1.08	0.66	0.72	0.73	(0.55)	0.52	0.50	0.25	0.31
5 Altruism towards other cons.	3.81	0.84	0.46	0.71	0.68	0.72	(0.52)	0.44	0.14	0.14
6 Venting	3.71	0.92	0.53	0.70	0.59	0.71	0.66	(0.52)	0.25	0.28
7 Revenge taking	3.51	0.82	0.65	0.46	0.20	0.50	0.38	0.50	(0.73)	0.49
8 Entertainment	3.33	1.03	0.65	0.39	0.18	0.56	0.37	0.53	0.70	(0.68)

Notes:

The diagonal values represent the average variance extracted estimates of the construct. The values in the lower diagonal of the table present the correlations between the constructs while the values in the upper diagonal of the table present the squared correlations between the constructs. Table includes results for Study 1

Table AIV

Descriptive statistics, correlation and discriminant validity of complaint reaction scales

	Mean	SD	1	2	3	4	5	6
1 Webcare satisfaction	3.70	1.12	(0.73)	0.50	0.58	0.53	0.55	0.01
2 Brand satisfaction	3.15	1.23	0.71	(0.67)	0.61	0.56	0.55	0.01
3 Brand image	3.27	1.24	0.76	0.78	(0.70)	0.59	0.61	0.02
4 Brand loyalty	3.14	1.28	0.73	0.75	0.77	(0.72)	0.66	0.04
5 Positive word-of-mouth	3.13	1.30	0.74	0.74	0.78	0.82	(0.65)	0.09
6 Negative word-of-mouth	3.49	1.07	−0.11	−0.12	−0.15	−0.21	−0.30	(0.63)

Notes:

The diagonal values represent the average variance extracted estimates of the construct. The values in the lower diagonal of the table present the correlations between the constructs while the values in the upper diagonal of the table present the squared correlations between the constructs. Table includes results for study I