Brand advocacy in the frontline: how does it affect customer satisfaction?

Purpose – Many organizations expect their service engineers, or frontline employees (FLEs), to behave as brand advocates by engaging in favorable communication about the brand and its offerings toward customers. However, this approach is not without risk as customers may be disappointed or even frustrated with brand advocacy behavior in many service encounters. The purpose of this paper is to study the impact of FLEs’ brand advocacy on customer satisfaction with the service encounter, and identify the conditions under which the effects are detrimental. This paper specifically considers service issue severity and product newness as contingency conditions. Design/methodology/approach – Building on social identification theory, the paper builds a conceptual model, which is empirically tested using a data set that matches data from service engineers, customers, and archival records from the after-sales service department of a globally operating business-to-business print and document management solutions provider. Findings – This paper finds that brand advocacy behavior harms customer satisfaction especially in service encounters that involve simple service issues (e.g. maintenance) for products that are new to the market. Fortunately, brand identification can compensate this negative effect under many service conditions. While the joint effect of brand identification and advocacy is most beneficial for severe service issues of new products, no effect on customer satisfaction was found for established products. Practical implications – This paper identifies those service situations in which brand advocacy is advisable and guides managers toward achieving more favorable customer evaluations. Originality/value – Past research has considered several FLE branding activities in the frontline but the effects of brand advocacy have not been isolated. In addition, most studies have assumed the effects of employee brand-related behaviors on customer satisfaction to be universally positive rather than negative and focused on antecedents and not on moderators and consequences.


Introduction
More and more companies expect their frontline employees (FLEs) to build and strengthen the brand by engaging in brand advocacy behavior, which reflects the display of favorable communication of their brand's values and offerings toward customers during service encounters. Firms like Zappos and Starbucks have experienced that "brand advocates" increase brand awareness in the market (Walter, 2013). Lowe's, Southwest Airlines, Ford, Domino's, Bank of America, General Electric, and Verizon have featured FLEs in their advertising campaigns because these workers are perceived by the audience as credible and, thus, help "humanize" the brand (O'Leary, 2010). Many organizations also rely on FLE brand advocacy in brand repositioning and new product launches (Miller et al., 2014).
However, using FLEs as brand advocates may not be beneficial in all customer encounters. Specifically, when a purchased product needs a maintenance or repair service, an FLE's favorable communication about the brand may backfire because it does not directly contribute to addressing the customer's issue. As a famous example, overly enthusiastic virtual agents at Sprint PCS and Dove infuriated clients who called to voice complaints (Spencer, 2003). Customers may perceive brand advocacy in after-sales service encounters as a business policy that gets in the way of reaching their goal. In a 2013 Accenture study, 79 percent of the respondents indicated to find such policies highly or even extremely frustrating (Accenture, 2013). On the other hand, reinforcing brand values and product benefits during after-sales services may reassure customers that they have made the right purchase decision and signal the accountability of the FLE and the firm as a whole to solve the customer issue (Aaker et al., 2004;Hess, 2008).
Despite the potential risks of brand advocacy in the frontline, marketing literature has remained silent on its outcomes for at least two reasons. First, scholars have focused on the general alignment of employee behavior with the brand promise (Sirianni et al., 2013) and frontliners "living the brand" (De Chernatony et al., 2006, p. 825). Consequently, they developed and used comprehensive and multidimensional concepts such as brand citizenship behaviors (e.g. Chang et al., 2012;King and Grace, 2012). In contrast, brand advocacy is unidimensional, and focuses on favorable communication about a brand's values and offerings. The effects of this behavior have not been isolated. Second, most studies have assumed the effects of employee brand-related behaviors on customer evaluations (i.e. satisfaction) to be universally positive and focused on antecedents, not consequences (e.g. Morhart et al., 2009). However, as the negative anecdotal evidence and survey results illustrate, managers clearly need to know when brand advocacy is appropriate and when not to use this instrument.
The aim of this paper is to study the impact of FLEs' brand advocacy on customer satisfaction with the service encounter and identify the conditions under which the effects are most and least negative. Considering both brand advocacy behavior and brand identification, this paper builds on social identity theory (Mael and Ashforth, 1992) and the branded service encounters framework (Sirianni et al., 2013) to develop a contingency model. The contingencies are two important service portfolio characteristics that may determine the strength of brand advocacy efforts. First, service issue severity reflects the degree of inconvenience borne by customers experiencing an issue. Because the severity of many service issues can be reasonably judged before allocating FLEs to the job (e.g. through remote monitoring of products, information of connected Internet-of-Things products, or through call-center problem categorization), it is a key variable for service managers in allocating service people (Batt, 2007). Second, product newness refers to the extent to which the products in the FLE's portfolio were recently launched on the market. New product introductions are critical to the growth and continued survival of a firm (Pauwels et al., 2004), but in their strive to be first to market companies often release products that have not all kinks ironed out and, thus, need service. Consistent with this, service managers use product newness to build service portfolios and distribute service jobs over the workforce. FLEs' service portfolios, thus, involve a key service-and product-related characteristic, respectively, and are argued to affect the relationship between brand advocacy and customer evaluations. This paper makes the following important contributions to literature. First, it extends the work on FLE brand attitudes and behaviors by considering brand attitude (i.e. identification), brand behavior (i.e. advocacy), and customer evaluations (i.e. satisfaction) in a single conceptual framework. Previous studies have concentrated on exploring the nomological network of brand attitudes and behaviors without considering the relationship to customer evaluations (e.g. Helm et al., 2016;Piehler et al., 2016). Other studies related either brand attitudes or behaviors to customer evaluations, but not both (e.g. Baker et al., 2014;). The value of considering brand identification and advocacy in one framework is underscored by our results that show differential effects of both concepts on customer satisfaction with the service.

231
Brand advocacy in the frontline Second, our focus on brand advocacy adds detail to studies on brand citizenship behavior that typically aggregate different brand-related behaviors into a higher-order concept and consider both on and off the job behavior (e.g. Baker et al., 2014;King and Grace, 2012). Past research has used the term brand advocacy to identify customers who are "intending to try new products of the brand, spreading favorable word-of-mouth, and being resilient to negative information" (Pai et al., 2015, p. 686; also see Stokburger-Sauer et al., 2012). Brand advocacy has also been investigated among retail salespeople as their tendency to recommend one specific brand in a multi-brand store (Badrinarayanan and Laverie, 2011). However, brand advocacy behavior of FLEs has not yet been considered.
Third, this paper adds to brand and service management literature in business-to-business (B2B) settings. While brand research has flourished in consumer studies, brands are equally important in B2B markets (Homburg et al., 2010). Especially in a time where customers are empowered toward self-service, equipment becomes increasingly smart and self-diagnostic, and remote monitoring reduces the number of face-to-face service encounters, managers and scholars are unsure whether FLEs can meaningfully profile the brand in customer encounters (Wünderlich et al., 2012). The model and hypotheses are empirically tested using a data set that matches data from service engineers, customers, and archival records from a globally operating B2B print and document management solutions provider.
Finally, we focus on service portfolios that service managers generally use to segment the market and organize and instruct FLEs (Batt, 2007) and show that careful service portfolio design can neutralize the negative effect of brand advocacy on customer satisfaction without eliminating the positive effect of brand identification. Service issue severity and product newness do not present encounter-specific and independent characteristics of a particular service job, but rather make up an FLE's structural work environment. Because there is very little academic empirical evidence on how to optimize FLEs' performance based on their service portfolios (see Van der Heijden et al. (2013) for an exception), this is an important contribution to the FLE management.

Theoretical background
The role of FLEs in brand management Brand management has become an area of interest in the marketing discipline over the last decades because brands reflect consumers' perceptions and are strong financial indicators of organizations (Vomberg et al., 2015). A brand is a cluster of functional and emotional values that provides a unique and desirable experience for stakeholders (De Chernatony et al., 2006). Although brand values can be communicated to the market in a variety of ways (e.g. advertising, public relations), the credibility and persuasiveness of FLEs' advocacy behavior have made these workers a key element in firms' integrated marketing communications (Sirianni et al., 2013). Work on brand advocacy by FLEs can be traced to two research streams: internal branding and boundary spanning. These two streams are discussed next.

Internal branding
Internal branding research focuses on how firms can foster employee behavior that is consistent with their organizations' (aspired) brand values. The top half of Table I provides an overview of recent empirical work in this domain. Active communication of meaningful brand information (Baker et al., 2014;) stimulates employees to internalize brand values in their self-concept so that they are better equipped to fulfill the promises inherent in the brand. Internalization results in high brand identification, which refers to a sense of "oneness" with the brand values such that the success or failures of the brand are perceived as one's own (Burmann and Zeplin, 2005 Brand identification promotes two types of brand-related behaviors. First, brand-building behaviors reflect "employees' contribution (both on and off the job) to an organization's customer-oriented branding efforts" (Morhart et al., 2009, p. 123). It includes positive word-of-mouth, but in contrast to brand advocacy, this concept has an off-the-job focus (i.e. toward friends and family) and hence is unlikely to affect service encounter evaluations. Second, brand citizenship behaviors represent individual voluntary brand-focused actions outside of role expectations that enhance the performance of the organization. The original concept consists of seven dimensions that include brand endorsement, which resembles brand advocacy (cf., Burmann and Zeplin, 2005). Remarkably, studies typically merge these dimensions into one overarching factor (e.g. Baker et al., 2014;King and Grace, 2012), or select a subset of activities that excludes brand endorsement (e.g. Chang et al., 2012). Consequently, also studies on brand citizenship behavior provide limited insight into the nature and role of brand advocacy.

Boundary spanning
Boundary spanning represents the second stream in FLE brand advocacy research and evolved around the concept of organizational citizenship behavior (OCB; Bettencourt et al., 2001). OCB refers to "individual behavior that is discretionary, not directly or explicitly recognized by the formal reward system, and that in the aggregate promotes the effective functioning of the organization" (Organ et al., 2005, p. 8). Service scholars extended and refined general OCBs; service delivery, internal influence, and external representation were consequently identified as service-oriented OCBs, or boundary-spanning behaviors. External representation reflects employees as vocal advocates to outsiders of the organization's image, goods, and services (Bettencourt et al., 2005) and, thus, closely relates to brand advocacy. Remarkably, the bottom half of Table I indicates that boundary spanning research has left the outcomes of external representation unexplored. Although brand advocacy behavior is related to employee positive word-of-mouth, brand endorsement, and external representation, it is a different concept because it represents the extent to which an employee engages in favorable communication of the brand's values and offerings toward customers during service encounters. Thus, it has a clear on-the-job focus whereas the other concepts also consider friends and family as recipients of brand communication. Consequently, where previous concepts hold value for human resource managers looking to hire suitable job candidates (Andreassen and Lanseng, 2010), brand advocacy behavior is targeted toward customers and, thus, valuable for service and marketing managers.

Conceptual framework
Based on the rich internal branding and boundary spanning literature, brand identification is considered central to the internalization of brand values and, thus, the motivation of employees to act as brand advocates. This paper, therefore, builds on social identity theory (Mael and Ashforth, 1992) and the branded service encounters framework (Sirianni et al., 2013) to construct a conceptual model and argue for the expected effects. Whereas the former theory may explain the role of brand identification, the latter helps to understand the conditional effects of brand advocacy behavior because it argues that familiarity of customers with the brand and its "experiences" (e.g. after-sales service encounters) determines how FLE brand-related behavior is evaluated. Brand familiarity reflects "a customer's background knowledge acquired as a result of direct and indirect experiences with a brand" (Sirianni et al., 2013, p. 110). The two service portfolio characteristics, service issue severity and product newness, directly relate to customer feelings of familiarity. Customers facing a simple maintenance service job experience a familiar and routinized service encounter. In contrast, severe issues such as a complex failure will feel less routinized and familiar to customers. Similarly, customers are less 235 Brand advocacy in the frontline familiar with new than established products from a brand. After-sales service for a new product will thus feel less routinized. The next section discusses how such associations may influence the effects of brand advocacy. Figure 1 summarizes the conceptual model.

Hypotheses
In line with social identity theory, marketing scholars have shown that FLE's experienced sense of "oneness" with the brand's values makes him or her more intrinsically motivated to behave in a manner that benefits the brand's interests because brand values are now accepted as own values (Morhart et al., 2009). In other words, there is congruence between brand and self-interest. Talking positively about the brand one identifies with provides a way to express one's own identity (Bhattacharya and Sen, 2003). Brand identification is, thus, likely to drive brand advocacy behavior.
FLEs' brand identification is also posited to directly and positively influence customer satisfaction with the after-sales service encounter. Employees who identify with their brand perceive the brand's success or failures as one's own and expend more energy to move the organization forward (Anaza and Rutherford, 2012;Burmann and Zeplin, 2005). They are better equipped to fulfill the explicit and implicit promises inherent in the brand in customer interactions (King and Grace, 2012). Employees then consider whether the potential outcomes of a service action are customer-oriented or violate the brand values (Morhart et al., 2009). Customers appreciate such a mindful effort that lives up to the brand's promises which, thus, will be reflected in their satisfaction level. Therefore:

H1. FLE brand identification positively influences (a) FLE brand advocacy behavior and
(b) customer satisfaction with the service encounter.
Regardless of whether it is maintenance or a response to failure, a service issue implies the frustration of the customers' product enjoyment or even entire business processes[1].
Customers then form expectations about acceptable service actions, or a service norm. They generally expect the company to address service issues in an efficient and timely manner (Hess et al., 2003) and in an authentic way (Gruber, 2011  differs greatly from the customer's anchor triggers a contrast effect and will be interpreted as inappropriate and more negative than it actually is. Because temporary unavailability of a product establishes a "loss frame" and a negative anchor in the customers mind (Heidenreich et al., 2015), any positive brand communication such as brand advocacy is likely to be highlighted as inappropriate and then cause dissatisfaction. Also, when employee behavior does not directly contribute to addressing the service issue at hand, customers experience difficulty to process and understand the service (Lee and Labroo, 2004). They consequently elicit a more negative response than upon alignment (Sirianni et al., 2013). Therefore: H2. FLE brand advocacy behavior negatively influences customer satisfaction with the service encounter.
It is important to know whether the negative effect of brand advocacy behavior occurs universally, or whether conditions can be identified that alleviate or even nullify this effect. First the role of service issue severity is considered. Low severity refers to a routine, maintenance-like task, while high severity indicates a non-routine failure task that requires a substantial time investment of the FLE to fix the issue. The former and latter, thus, involve a limited and major degree of customer inconvenience respectively. A routinized encounter between firm and customer triggers a top-down customer processing style (Schwarz, 2002). This processing style "is accompanied by less focused attention and relies on customers' general knowledge of the category such that they may judge the employee more on his or her category membership (i.e. working for that brand) than his or her brand-aligned or misaligned behavior" (Sirianni et al., 2013, p. 110). Brand advocacy clearly indicates the employee-brand relationship. Although these relationships are typically made explicit by tangibles such as company clothing or branded tools (Shao et al., 2004), verbal actions may add clarity. For instance, a person wearing a shirt with a brand logo may work for that brand, but there may be many other reasons for wearing this shirt. When he/she also talks about the brand releasing great products and enjoying working there, customers will infer that the person is truly a member of the brand community, or organization. This advocacy behavior is valued more by customers in routinized after-sales service encounters than by those facing unusual service situations because of the top-down processing style of customers in the former category.
FLEs advocating the brand toward customers who experience a severe service issue provide an inconsistent picture, since the positive brand message is incongruent with the negative experience. Customers have just experienced an incident that is indicative of low product quality. This does not align with any positive messages on the brand. Customers are then likely to perceive brand advocacy behavior to be inconsistent with their view of the brand and to violate their service norm of prompt response. They will be less satisfied with the service provided. Therefore: H3. Service issue severity strengthens the negative influence of FLE brand advocacy behavior on customer satisfaction with the service encounter.
Customer expectations also differ for new vs established products (Ahearne et al., 2010).
A product that has recently been released to the market is less familiar to both customer and FLE. Lacking product knowledge, customers systematically process service encounters on new products and focus their attention on the specific details of the service employee and how his/her behavior solves the issue (Hilton and Darley, 1991;Schwarz, 2002). Poorly aligned with tackling the issue at hand, brand advocacy behavior may then be considered inappropriate and more negatively relate to satisfaction. In addition, in the absence of past experiences with a (new) product, customer expectations on the product's quality are, at least partially, a function of a firm's advertising activities (Kopalle and Lehmann, 1995).

237
Brand advocacy in the frontline Brand advocacy entails a positive promotional effort and thus raises the customer's level of expectation of product quality. Unfortunately, these expectations are negatively disconfirmed by the fact that the product needs repair or other service assistance. The enhanced (negative) gap between expectations and experience translates to lower customer satisfaction. Therefore: H4. Product newness strengthens the negative influence of FLE brand advocacy behavior on customer satisfaction with the service encounter.

Methodology
Research setting and data collection The conceptual framework is tested using a sample of field service engineers working for a major international manufacturer of print and document management solutions for professional environments. These FLEs specialize in delivering onsite maintenance and repair services and have unique, individual portfolios of products and customers serviced.
Maintenance is pre-scheduled and customers report a product failure by contacting customer support by phone, e-mail, or interface on their machine. The company's after-sales procedures and the important role of FLEs in these processes are representative for many capital equipment manufacturers that are transitioning from products to services providers (cf., Oliva and Kallenberg, 2003).
Multi-source data were collected using employee surveys, customer satisfaction surveys, and the firm's archival records. The latter included objective measures of employee performance and product performance that could be traced to the individual FLEs' service visits. The data set, thus, involved matched subjective and objective data sources, which is typically hard to achieve in B2B settings. FLE data were collected using a paper-and-pencil survey which was distributed and collected during monthly meetings of FLEs with their managers at the company's headquarters. To facilitate truthful responses, surveys were handed out after the manager had left the room, promised confidentiality, and offered the respondents the opportunity to receive a summary of the results. From a total of 184 distributed surveys, 134 usable responses were received, resulting in a response rate of 72.8 percent. Five respondents were discarded in the analysis because of missing data. With one exception, all remaining 129 respondents were men. Such male dominance corresponds with labor force statistics for technical service jobs (e.g. US Bureau of Labor Statistics, 2010). Their mean age was 46.5 years (SD ¼ 11.8), and their experience averaged 15.8 years (SD ¼ 12.2).

Measurement
In the FLE survey, existing literature was used to operationalize all latent constructs with multi-item scales. Items were measured using seven-point Likert scales (1 ¼ "strongly disagree" and 7 ¼ "strongly agree"), unless indicated otherwise. The measures were pretested with eight service employees and fine-tuned the items according to their feedback. Table II provides an overview of the measures of the study's latent constructs.
FLE brand identification was measured with four items adapted from Mael and Ashforth (1992). Although FLE external representation behavior is a well-defined construct in the literature (e.g. Bettencourt and Brown, 2003), brand advocacy is rarely measured. As indicated, previous occurrences of brand advocacy in literature have a customer focus (e.g. Stokburger-Sauer et al., 2012) or pertain to salesperson recommendation behavior (e.g. Badrinarayanan and Laverie, 2011); the associated scales are unsuitable for the research purposes. Consequently, the items to measure FLE brand advocacy behavior were selected using a multi-step approach. First, for three days and with three different service engineers, FLEcustomer interactions were unobtrusively observed and their conversations analyzed. Brand advocacy generally took 238 JOSM 29,2 the form of brand endorsement and positive word-of-mouth. It could happen during all phases of the service process but often began with small talk at the start of the encounter. Although engineers adapted their brand advocacy behavior somewhat based on the service situation at hand, the average levels of brand advocacy between individual engineers differed significantly. Second, internal branding and boundary spanning literature were studied (e.g. Bettencourt and Brown, 2003;Morhart et al., 2009) to identify related items to draw on. Third, combining the insights from the observational study and the literature review, three items were adapted from extant literature that captured the breadth of brand advocacy behavior and applied to a wide range of B2B service interactions. In a small test with eight FLEs, these were tested and refined. Based on this feedback, some clear examples of brand advocacy were provided in the introduction of the question in the survey, and instructed employees to specifically reflect on their behavior during the past six months.
The product newness scale used two items, was specifically developed for this study, and tapped the extent to which the products that the FLE worked on were new and recently introduced to the market. Industry-specific investigations were conducted to define the average product life cycle. As a result, "new products" were defined as those introduced in the 18 months preceding the survey.
Archival data were used to assess service issue severity. Specifically, the number of "escalations" relative to the employee's total number of service visits was employed. Company quality guidelines dictate that the service job should be passed back to the organization ("escalated") if an issue cannot be satisfactorily addressed in one visit. Escalation does not reflect a lack of competence because each employee is certified to service the products in his/her portfolio. Because some products are more likely to have more severe issues than others, escalated service visits are not included in FLEs' speed and quality performance scores.
The dependent measure of customer satisfaction with the service encounter was assessed using an online customer survey. The firm continuously evaluated service encounters by randomly surveying customers after a service visit. Not every single encounter was evaluated to prevent overburdening of customers. For the purpose of this study customers were first

239
Brand advocacy in the frontline provided with a screening question to verify they had contact with the FLE. This ensured that they reported their evaluation of the entire service process, not just of post-encounter machine performance. Thereafter, customers indicated their level of satisfaction with the service encounter on a 1 to 10 scale rating with a higher score indicative of higher satisfaction. The data of customers were matched with the data of the FLEs involved and then aggregated to generate a single score per employee. Previous research shows that such scores can reliably be used in analyses on FLE behavior and performance (e.g. Baker et al., 2014). The average number of customers replying per employee was 6.13 (SD ¼ 3.98).
To allow for correct model estimation, several controls were added. First, an FLE's overall service speed may be an important driver of customer satisfaction with an after-sales service encounter. It was operationalized using an objective performance measure from the firm's archival records. The company monitors the duration of each FLE's service visit, which is standardized in accordance with norms that prescribe the targeted duration of a single visit for a specific product type and issue combination. The standardized scores are aggregated to a monthly average per employee to yield a mean time to repair score that indicates whether each employee has conducted service visits faster or slower than the norm. Together with FLEs' (standardized) average number of service visits per month, this score is indicative of an FLE's overall service speed. Second, an employee's overall service quality was included. The company monitors every product's uptime (i.e. operating hours without errors) following an FLE's service visit. These uptime scores are standardized to product-specific norms. A monthly average score then indicates whether an FLE has made the company's products fall below or exceed the uptime norms [2].
Third, in addition to the objective archival data, several (employee) survey-based control variables were used. Questions on FLE age and job experience (i.e. years of experience with current tasks) were included. Finally, the survey included manager brand communication, which reflects the extent to which a service employee's manager actively helps to understand how the employee's tasks connect to brand values. The measure used three items from Zhang and Bartol (2010) and was adapted to the research context. Each FLE was asked to score their manager, because individual interpretations of managerial actions most substantially shape employee attitudes and behavior (Di Mascio, 2010).

Analyses
The data were analyzed in three stages. First, a confirmatory factor analysis was conducted in which the control variables were included as covariates. Overall fit measures display a good fit of the measurement model: χ 2 (97) ¼ 134.25, CFI ¼ 0.96, TLI ¼ 0.94, RMSEA ¼ 0.06. All individual items loaded significantly on their respective construct ( factor scores W0.50). The composite reliability of all constructs was well above the suggested minimum value of 0.70. The average variance extracted (AVE) for each construct exceeded the commonly accepted threshold value of 0.50. were met for all constructs, as for any construct, the square root of its AVE exceeded the correlations with any other study construct.
Second, the possibility of a nested data structure was tested. Although FLEs worked autonomously, each reported to one of 14 managers. By calculating the intraclass correlation coefficient (ICC) for manager brand communication, the percentage of total variance that could be attributed to differences across groups can be determined. The ICC was 0.03, a very low value, so concerns about the nested nature of the data are unwarranted. In three subsequent interviews, company managers confirmed to customize their brand communication activities per FLE.
Third, the items of each construct were averaged to create the latent variables and standardized these variables to obtain standardized coefficients and allow for reliable moderation analyses. The hypotheses were then tested using Hayes' PROCESS tool in SPSS (Hayes, 2012). This software is specifically designed to handle integrated conditional process models and allows multiple moderators to operate in the same model calculation. It also enables a detailed floodlight analysis of the effects of X on Y rather than a spotlight analysis. Floodlight analyses assess interaction effects at multiple values of the moderator. The resulting detailed understanding of the effect may allow to identify the particular regions along the continuum of one or more moderators where the effect of X on Y is significant and where it is not. This practice was advocated long ago by Johnson and Neyman (1936), but since then applied only sparsely in marketing research (Spiller et al., 2013). PROCESS provides asymmetric bias-corrected bootstrap confidence intervals (CIs) for inferences about the conditional indirect effects at the 10th, 25th, 50th, 75th, and 90th percentiles of the moderator. Two models were estimated: a core model without moderation effects and the hypothesized model. In total, 5,000 bootstrap samples were used to estimate the 95 percent CIs for the indirect effects.
The core model was calculated using PROCESS template model 4 (see Hayes, 2013 for all templates). The control variables were entered as covariates for the mediator and the dependent variable. The core model takes the form of the following set of linear equations: Here, SIS and PN denote service issue severity and product newness, respectively. Table IV reports the standardized estimates for the core model (left-hand side) and hypothesized model (right). The core model explains 39.7 percent in FLE brand advocacy behavior and 22.7 percent in customer satisfaction with the service encounter. Adding the interaction effects in the hypothesized model further increases explained variance in customer satisfaction to 29.0 percent, indicating a meaningful model extension. In contrast, a model that consisted of only the control variables explained a mere 21.2 percent of variance in brand advocacy behavior and 12.8 percent in customer satisfaction with the service encounter.

Results
Consistent with H1a and H1b, brand identification positively related to brand advocacy behavior ( β ¼ 0.434, t ¼ 6.080) and customer satisfaction ( β ¼ 0.310, t ¼ 3.760). In support of H2, brand advocacy behavior related negatively to customer satisfaction ( β ¼ −0.255, t ¼ −2.782). So, brand advocacy behavior partially mediated between brand identification and encounter-specific customer satisfaction. This is also evidenced by the 95% CI of the indirect effect of this mediated relationship, which does not contain the value 0: (−0.247; −0.017).
With regard to the moderators, the interaction term of brand advocacy behavior with service issue severity related significantly and positively to customer satisfaction ( β ¼ 0.128, t ¼ 2.001). This is remarkable, as H3 predicted the opposite effect. Consequently H3 was rejected. The Discussion section further details this effect. Product newness strengthened the negative effect of brand advocacy behavior on customer satisfaction ( β ¼ −0.195, t ¼ −2.598), which lent support to H4.
Finally, the result of the control variables show that age ( β ¼ −0.285, t ¼ −2.782) and manager brand communication ( β ¼ 0.212, t ¼ 2.818) were significantly related to brand advocacy behavior. This suggests that manager attention to brand communication breeds FLE brand advocacy, and that younger employees are more engaged in brand advocacy than their older colleagues. Moreover, overall service quality related significantly to customer satisfaction ( β ¼ 0.207, t ¼ 2.865). All these results are intuitive and, thus, offer face validity to the data and increase the confidence in the outcomes.
Hence, the key to mitigate the negative effect of brand advocacy behavior on customer satisfaction lies in the ability to adapt the configuration of frontliners' service portfolios. For a more detailed investigation, the conditional effect of brand advocacy behavior on customer satisfaction were calculated at the 10th, 25th, 50th, 75th, and 90th percentiles of service issue severity and product newness based on 5,000 bootstrap samples for bias corrected CIs. Table V displays the results in the "Customer Satisfaction with Service Encounter" column. Significant conditional effect conditions are printed in italic. To aid interpretation, Figure 2 Panel A plots the conditional relationshipthe effect was calculated by taking the point estimate for conditions where the CI did not contain the value 0 and assuming no effect otherwise.
The plot indicates that for FLEs servicing established (rather than new) products, brand advocacy behavior does not affect customer satisfaction with the service encounter. Even for service on slightly newer products brand advocacy behavior is unrelated to satisfaction scores, but only if the service issue is severe. When the service issue becomes less severe (i.e. more maintenance) and product newness increases, brand advocacy behavior has an increasingly negative effect on customer satisfaction.

Additional analyses: full moderated mediation model
Based on the previous analyses, managers would need to reallocate service tasks over frontliners according to their propensity to promote the brand. Because brand identification also has a positive effect on customer satisfaction the question is at which point the joint effect of brand advocacy behavior and brand identification on customer satisfaction peaks. Although this could be extrapolated from Figure 2 Panel A by adding a constant to the estimations (i.e. in the hypothesized model BI's effects are not moderated), this would provide an incomplete picture because the level of service issue severity and product newness should be controlled for. Therefore, a full moderated mediation model was estimated using PROCESS model template 76, described by the equations: BAB ¼ i BAB þa 1 BI þa 2 SIS þa 3 PN þa 4 BI Â SIS þa 5 BI Â PN þa 6 MBC þa 7 AGE þa 8 JOBEXPþa 9 OSQ þa 10 OSSþe BAB CSAT ¼ i CSAT þc 0 1 BIþc 0 2 SIS þc 0 3 PN þb 1 BABþb 2 BAB Â SIS þb 3 BAB Â PNþc 0 4 BI Â SIS þc 0 5 BI Â PN þb 4 MBC þb 5 AGE þb 6 JOBEXPþb 7 OSQ þb 8 OSS þe CSAT Table V (right hand side) displays the results for brand identification's direct effect on customer satisfaction and its indirect effect through brand advocacy behavior for every possible combination of the moderators' values. Figure 2 Panel B plots the total effect of brand identification, reflecting the sum of the values for the direct and indirect effects for every condition. Results indicate that FLEs' brand identification is an important instrument to counterbalance the negative effects of their brand advocacy behavior. In 68 percent (17/25) of all combinations of product newness and service issue severity, brand identification neutralizes the negative effect of brand advocacy or even relates positively to customer satisfaction with service. Looking in more detail, three things stand out. First, as the plot in Figure 2 Panel B shows, there are no effects of brand attitudes and behaviors in the frontline when product newness is low. In other words, brand identification appears to not influence a customer's satisfaction when the service involves established products. Second, the total effect of brand identification peaks when both product newness and  Brand advocacy in the frontline product newness and service issue severity have the same score, i.e. both low, both medium, or both high. So, brand attitudes and behaviors tend to benefit customer satisfaction more when product newness and service issue severity are aligned then when they are misaligned (e.g. newness is high and severity is low). This provides important managerial implications, which will be discussed later.

Alternative moderators and explanations
Finally, alternative contingencies and explanations were explored. For instance, it could be that customers are less concerned with the inappropriateness of brand advocacy if the FLE's performance was very good or even excellent. Although a credible line of reasoning, no significant moderating effects were found of overall service quality ( β ¼ −0.038, t ¼ −0.536) and overall service speed ( β ¼ −0.013, t ¼ −0.127) on the relationship between brand advocacy and customer satisfaction [3]. This provides further evidence that customers' dissatisfaction with brand advocacy behavior is specific to the encounter rather than associated with the general skills and performance of an FLE across customers and settings.

Discussion
The aim of this study was to uncover the effects of brand advocacy behavior in after-sales frontline service. Academic literature studied brand identification of FLEs, but did not test its effects empirically. In addition, past studies did not isolate the effects of brand advocacy behavior from brand identification. While the potential power of brand-related behaviors by FLEs has been heralded, anecdotal accounts about negative outcomes of brand advocacy rendered the current research appropriate and necessary. The implications of the work are discussed next.
Theoretical implications Effects of brand attitudes and behaviors in the frontline. This paper extends work on internal branding and boundary spanning by substantiating the relationships of brand attitudes and behaviors with customer evaluations. Specifically, the concept of brand advocacy behavior is introduced, thereby adding considerable detail to studies that employ broad conceptualizations such as brand citizenship behavior and generally do not investigate the outcomes of these behaviors (e.g. King and Grace, 2012;Morhart et al., 2009). For instance, while Baker et al. (2014) reported a positive effect of brand citizenship behavior on customer satisfaction the results offer a more nuanced picture. Easier to observe and judge than attitudes, brand advocacy behavior plays an important role in customers' evaluations of the service and brand. However, the effect is generally negative, not positive. The results also confirm a partially mediated positive effect of brand identification of FLEs on customer satisfaction with the service encounter. These contrasting effects demonstrate the value of distinguishing between brand-related attitudes and behaviors in the frontline.
Branding literature generally holds that FLE branding efforts yield desirable effects when employee behavior is aligned with brand promises and customer expectations (Sirianni et al., 2013). Behavior that is not aligned with customer expectations and interests may negatively affect customer responses. For instance, in their recent study, Relling et al. (2016) show that positive word-of-mouth about a brand does not relate to, or may even hinder, active customer participation in online communities because discussion content is not aligned with customer expectations to find objective information about a brand. Using similar logic, the results show that customers do not appreciate brand advocacy behavior in many after-sales service situations. However, it seems that especially the alignment of brand identification with brand advocacy behavior is important. Compared to the more observable brand advocacy behavior, brand 246 JOSM 29,2 identification is the deeper attitude of the FLE and can balance the detrimental effects of brand advocacy behavior.
Conditional effects of the service portfolio. Service portfolio characteristics of FLEs are of pivotal importance for understanding the positive or negative effects of brand advocacy behavior in the frontline. When products are new to the market, the unfamiliarity with the service situation leads customers involved to have a processing style that stresses the specific details of how the FLE addresses the service issue at hand (Hilton and Darley, 1991). Brand advocacy is consequently judged as inappropriate behavior which creates customer dissatisfaction with the service encounter.
Interestingly, this logic did not hold for the moderating effect of service issue severity. Customers facing simple routinized repairs were expected to engage in top-down processing of information. Customers would then appreciate FLEs' signals of brand membership through his/her brand advocacy behavior. However, the findings indicate the opposite: the more inconvenience a service issue brings for the customer, the less brand advocacy behavior harms customer satisfaction. Possibly, severe service issues represent out-of-the-ordinary events in customers' relationship with a brand which trigger customers to update evaluations of the relationship (Van Doorn and Verhoef, 2008). Despite the increasing empowerment of B2B customers as co-producers of after-sales service, a severe service issue limits customers' possibilities to self-or even co-produce a solution. Then customers are very reliant on the FLE's expertise. In such situations, evaluations of capabilities, efforts, trustworthiness, supportiveness, and accountability of the FLE gauge the significance of and formulate responses to transgressions (Aaker et al., 2004). Complementing the resolution of a severe service issue, FLEs' brand advocacy behavior stresses the fact that the firm has taken responsibility and is willing to do so in future too. It results in a more positive evaluation. In contrast, low severity issues could possibly have been remedied by customers themselves. Under these conditions brand advocacy does not highlight unique brand values and expertise, but instead may make customers feel naive for not self-producing a solution. This generates a negative rather than the originally anticipated positive response.

Managerial implications
Guaranteeing the continuity and productivity of B2B customers' business processes, FLEs such as service engineers have become highly instrumental to meet operational performance goals and please customers in after-sales service encounters. Although the associated service encounters traditionally have a relationship-building function, the effort to reinforce the brand's offerings and values through FLEs' brand advocacy behavior has a potential dark side. Only by building FLEs' brand identification and carefully considering service portfolio characteristics, the positive effects of branding efforts in the frontline can be enjoyed.
First, managers should try to build employees' brand identification and at the same time train employees when and how to advocate for the brand based on the context of the service encounter. Customers are generally satisfied with the service offered by employees who experience a sense of "oneness" with the brand's values and diligently deliver the brand promise, but they are often dissatisfied with promotional FLEs. Although brand advocacy does not necessarily lengthen the service encounterin the sample of this research brand advocacy was not significantly correlated with mean time to repairit does decrease customers' satisfaction with the service encounter. Managers should help FLEs understand the subtle difference between "living" and "promoting" the brand. In-house role playing games where managers ask FLEs to convince them about the brand may clarify these differences. This may be done for extant personnel or for new hires during recruitment processes. Managers can then enhance brand identification through consistent brand 247 Brand advocacy in the frontline communications to their staff. For instance, videotaped service scenarios and idealized responses may be used to train FLEs to adopt the right brand attitude and behavior.
Second, a manager should make FLEs aware of when to engage in and when to refrain from such behavior. Brand advocacy's dark side does not occur universally across after-sales service encounters but depends on the product and service characteristics in the encounter. Managers are advised to consider their FLE's service portfolio in terms of product newness and service issue severity. For instance, the results show that in service jobs involving new (rather than established) products, brand advocacy harms customer satisfaction. This is an important implication because many managers feel that especially new products should go along with intense marketing actions to convince customers on the product's benefits. After-sales service on such products should break from the pre-sales promotional activities to prevent negative customer evaluations. On the other hand, for established products, branding attitudes and behaviors of FLEs do not seem to have any effect on customer satisfaction. Hence, managers may consider allocating newly hired employees, who may still have to build their emotional attachment to the brand, to service jobs involving well-established products only. Finally, FLEs' brand identification seems to work especially well in service encounters where product newness and service issue severity are aligned (e.g. both medium or both high). In these instances, the negative effects of brand advocacy behavior are limited, while customers' appreciation for employees "living the brand" is high. Passionate FLEs may be allocated to these types of jobs. In sum, managers can use Figure 2 to instruct FLEs' brand advocacy behavior based on their specific service portfolio characteristics and, thus, optimize the firm's total relationship-building effort in the after-sales frontline.

Limitations and further research
This study has several limitations that offer opportunities for further research. First, data were used from service engineers of a single B2B firm that is typical for modern-day manufacturers offering after-sales service. Although the results are considered generalizable to other B2B firms, additional studies could help to confirm this claim. Second, although this research controlled for several personal characteristics of the FLE, other traits may be investigated. For instance, employees who are high in personal skills such as emotional intelligence (Gabbott et al., 2011) or who have the ability to establish rapport (DeWitt and Brady, 2003) may be better able to understand when brand advocacy is (in)appropriate and consequently regulate their behavior. Finally, the concept of brand advocacy behavior may be developed further. Recent studies have investigated whether and how FLEs can combine service and sales tasks (cf., Jasmand et al., 2012). Although brand attitudes and behaviors could be placed on a continuum that describes frontline behavior as an increasingly profit-oriented activity from brand identification to brand advocacy to selling behavior, little is known about the relationship between these constructs, particularly in B2B service settings. This is an intriguing field for future research.

Notes
1. In this study, we focus on B2B field service engineers who work on print and document management solutions that are essential in customers' everyday business processes. If such solutions receive maintenance, the machines generally cannot be used for a certain period. For example, when a copier at a university department is serviced, employees cannot print, scan, or copy documents. This frustrates the workflow, and may cause dissatisfaction of users, particularly if not preannounced or when maintenance takes longer than expected.
2. To ensure product uptime truly reflects the FLE's service quality, rather than inherent product quality, it was checked whether one FLE's uptime score differed from another FLE's score for the same product. FLEs' average performance was compared for three individual products over a oneyear period and found large differences in the uptime scores of FLEs for each of the three products. The serious variation suggests that the measure is a good indicator for FLE quality performance. For greater robustness, the potential effects of "lemons," or products that constantly break down or are hard to fix, were ruled out. No single product produced uptime scores that consistently violated product-specific norms.