The demand-what-you-want strategy to service recovery: achieving high customer satisfaction with low service failure compensation using anchoring and precision effects

Abstract

Purpose

Previous research suggests that the compensation offered to customers after a service failure has to be substantial to make customer satisfaction surpass that of an error-free service. However, with the right service recovery strategy, it might be possible to reduce compensation size while maintaining happy customers. The aim of the current study is to test whether an anchoring technique can be used to achieve this goal.

Design/methodology/approach

After experiencing a service failure, participants were told that there is a standard size of the compensation for service failures. The size of this standard was different depending on condition. Thereafter, participants were asked how much they would demand to be satisfied with their customer experience.

Findings

The compensation demand was relatively high on average (1,000–1,400 SEK, ≈ $120). However, telling the participants that customers typically receive 200 SEK as compensation reduced their demand to about 800 SEK (Experiment 1)—an anchoring effect. Moreover, a precise anchoring point (a typical compensation of 247 SEK) generated a lower demand than rounded anchoring points, even when the rounded anchoring point was lower (200 SEK) than the precise counterpart (Experiment 2)—a precision effect.

Implications/value

Setting a low compensation standard—yet allowing customers to actually receive compensations above the standard—can make customers more satisfied while also saving resources in demand-what-you-want service recovery situations, in particular when the compensation standard is a precise value.

Citation

Kron, N., Björkman, J., Ek, P., Pihlgren, M., Mazraeh, H., Berggren, B. and Sörqvist, P. (2023), "The demand-what-you-want strategy to service recovery: achieving high customer satisfaction with low service failure compensation using anchoring and precision effects", Journal of Service Theory and Practice, Vol. 33 No. 7, pp. 73-93. https://doi.org/10.1108/JSTP-02-2023-0029

1. Introduction

Imagine that you have booked and paid for a long-awaited overnight stay at a hotel with your partner. When you arrive at the hotel, the desk clerk tells you that the booked double room has accidently been occupied but offers you a smaller room in its place because it is the only one available. As a result, you will likely be agitated and dissatisfied with the service. Service failures such as this are quite common in the hotel industry (Lewis and McCann, 2004; Mount, 2012), as well as in many other businesses including retail (Fayos-Gardó et al., 2017), banking (Garg, 2013), tourism (Gohary et al., 2016) and the restaurant business (Ok et al., 2007). Existing literature suggests that customers, when exposed to a service failure, expect a “fair” resolution (Blodgett et al., 1997). Otherwise it can result in negative word-of-mouth (WOM), the spreading of negative reviews and customers who are unwilling to recommend the service to others (Kim et al., 2009, 2016).

Entirely avoiding service failures is impossible in most businesses, as long as human factors and circumstantial complexities play a role. Hence, an important aspect of services marketing is how to successfully recover from service failures when they occur (Van Vaerenbergh et al., 2019). In the case above, the hotel could offer complimentary drinks or food, provide sincere apologies and explanations, offer monetary compensation or use other means to amplify their service recovery to improve customer satisfaction and loyalty (Krishna et al., 2011). Recovery efforts can improve the level of customer satisfaction and under certain circumstances even surpass the level of error-free service (Edström et al., 2022; Krishna et al., 2014). This phenomenon is known as the service recovery paradox (SRP; McCollough et al., 2000), see Figure 1.

Despite efforts to implement successful service recovery processes during the last decades, the effects of recovery efforts on satisfaction levels have varied in previous research (Van Vaerenbergh et al., 2018), and the SRP does not always materialize without costly overcompensation (Edström et al., 2022; Van Vaerenbergh et al., 2019). However, recent evidence also suggests that recovery that involves (Van Vaerenbergh et al., 2018) and adapts compensation to (Nazifi et al., 2021a) the individual customer may be more effective at increasing satisfaction. In particular, involvement in determining its outcome has been proposed as a promising way forward that needs further research (Arsenovic et al., 2022).

The aim of this study is therefore to explore the potential of utilizing a psychological heuristic called anchoring and adjustment (Epley and Gilovich, 2006; Furnham and Boo, 2014) as a tool to lower the SRP threshold and involve the individual customer in the outcome of the recovery. In a first experiment, we demonstrate that presenting customers with a typical compensation value after service failure lowers their compensation demand needed to trigger high customer satisfaction and an SRP in a co-created recovery. A second experiment corroborates these findings, and shows that a precise compensation value (rather than round) further strengthens the anchoring effect, such that a higher but precise anchoring value can lead to lower compensation demand, even in comparison with a lower but round anchoring value.

Our findings and novel method make several contributions to research on managing service recovery. First, our study adds to research on co-creative and adaptive service recovery, areas which remains under-researched (Khamitov et al., 2020). Specifically, the study follows calls to investigate how involvement in the level of compensation can be used to manage the effectiveness of service recovery efforts (Arsenovic et al., 2022). By using what we term a “demand-what-you-want technique” in combination with anchoring and precision effects, we demonstrate an effective and resource-efficient way to involve the customer in the outcome of service recovery (Van Vaerenbergh et al., 2018). We thereby also contribute to the literature by answering calls for the integration of psychological perspectives on individuals in recovery situations (Khamitov et al., 2020). Finally, our study contributes to, and nuances the question of the optimal level of compensation in recovery (Gelbrich et al., 2015; Khamitov et al., 2020). In a recent study, Edström et al. (2022) tried to identify a fixed threshold for recovery efforts to trigger SRP. We build on and extend their findings by demonstrating that compensation levels are not fixed but rather constructed based on salient and available information at the moment of recovery. Our results show that heterogeneous demands and differing levels to trigger an SRP are not solely due to heterogeneity between individuals (e.g. Balaji et al., 2018; Ozgen and Kurt, 2012), but also significantly influenced by contextual and situational factors. In the following, we outline the theoretical background to the study along with hypothesis development before outlining two experiments used to test the anchoring technique in two service recovery scenarios.

2. Theoretical background

2.1 Co-creating service recovery

Previous research has shown that many factors influence the success of service recovery. The success of an apology, for example, depends on its content (Lewicki et al., 2016) but also on the timing of the apology (Min et al., 2020). Studies have also identified that personality traits (Balaji et al., 2018), the emotive state of the customer (Ozgen and Kurt, 2012), and the perceived relationship to service provider (Hazée et al., 2017; Wang et al., 2014) and brand (Wang and Zhang, 2018) impact the results of recovery. Culture has also been found to play a role in setting expectations on recovery (Kanousi, 2005) as well as service personnels' ability to improvise and adapt to the situation (Cunha et al., 2009).

To improve the effectiveness and efficiency of recovery, research and practitioners have increasingly focused on adapting (Ringberg et al., 2007), and co-creating the recovery together with individual customers (Dong et al., 2008; Xu et al., 2014). Adaptation of compensation levels to unique service episodes is often warranted as heterogeneity in context and among individual customers may require different levels of compensation to raise satisfaction levels sufficiently (Khamitov et al., 2020; Nazifi et al., 2021a). By co-creating the recovery, the customer is recognized as an active part in producing value from services (Vargo and Lusch, 2004) and involved “in taking actions to respond to a service failure” (Dong et al., 2008, p. 126).

Inviting the customer to partake in creating the recovery and compensation serves as a way to better meet the varied responses from different customers which can otherwise offset the results of compensation as a recovery tactic (Hazée et al., 2017; Ozgen and Kurt, 2012). For example, while personality traits have been found to moderate effects of service recovery on satisfaction, involving customers in recovery can negate this effect (Balaji et al., 2018).

A main mechanism behind compensation as a driver of satisfaction is the consumer's perception of whether they have been treated fairly or unfairly in the delivery of a service (Khamitov et al., 2020) – the customer's perceived justice of the service encounter (Balaji et al., 2018; Maxham and Netemeyer, 2002). The inputs and outputs of the service encounter, including failure and recovery, are weighed against each other in forming the customer's perception of justice or fairness (Adams, 1963). This perspective, rooted in Justice Theory, posits that the consumer will evaluate the episode in terms of the process (procedural justice), its outcome (distributive justice) and interaction (interactional justice) with the firm (Blau, 1964) to arrive at an overall perception of justice (Balaji et al., 2018). Recovery through compensation then serves as a way to restore the perceived injustice customers experience upon service failure (Wirtz and Mattila, 2004). A co-created service recovery further increases the customer's commitment and motivation to repair the damage caused by the service failure (Djelassi and Decoopman, 2013) and as a result, its usefulness as a recovery tactic. Co-creation of the compensation in particular, has been found to increase perceptions of fairness of the service episode (Mattila and Cranage, 2005).

2.2 Compensation in service recovery

The three dimensions of justice are highly correlated and interrelated (Hazée et al., 2017; Khamitov et al., 2020). For example, perceptions of a positive outcome often have positive spill-over effects on how fair the process and interaction are perceived to be. This is especially true in recovery involving the customer, as co-creation is by nature interactional, and where the value created cannot be separated from the service delivery (Vargo and Lusch, 2004). For example, by inviting the customer in setting the compensation level, perceptions of procedural and interactional justice are also impacted, not only perceptions of a fair or unfair outcome. Therefore, the type and magnitude of compensation have been identified as central to the success of recovery (Van Vaerenbergh et al., 2019).

Previous research indicates that different compensation types can have different effects in different situations. Monetary compensations can be more effective than apologies and explanations (Orsingher et al., 2010) and replacements or vouchers (Noone and Lee, 2010), but compensations should match the type of service failure to be fully efficient, in which case an explanation can be more efficient than monetary compensation (Nazifi et al., 2020). Quasi-monetary compensations like cryptocurrencies (Nazifi et al., 2021b) have similarly been found to have positive effects. Past research also indicates that the compensation must be quite substantial for the service recovery paradox (Figure 1) to emerge (Edström et al., 2022; Garg, 2013; Ok et al., 2007).

Additionally, the benefits from monetary compensation—in terms of customer satisfaction—are not linearly related to the monetary size of the compensation (Gelbrich et al., 2015). Instead, the relationship follows an S-shaped curve wherein the benefits of increasing the monetary size diminish as the size reaches above a certain threshold. This means that the relative effect on satisfaction is greater for a partial compensation than a full compensation like a refund or even overcompensating the customer (Gelbrich et al., 2016), because of the law of diminishing marginal utility (Noone and Lee, 2010).

2.3 Anchoring and adjustment

If very large compensations would be necessary to achieve an SRP, as past research seems to suggest (Edström et al., 2022; Garg, 2013; Gelbrich et al., 2015; Ok et al., 2007), then at first glance using such compensations for this purpose would not be a successful recovery strategy. However, if the threshold of the SRP could be lowered, this can be used to achieve high customer satisfaction while also saving resources as the recovery process becomes more efficient. A possible way to achieve this, overlooked in previous research, is to use anchoring techniques as a strategy to service recovery that has the potential to make customers more satisfied with smaller service failure compensations. In addition to making the recovery more resource efficient, this method also allows for customer participation in both the outcome and process of recovery, potentially increasing perceptions of fairness (Wirtz and Mattila, 2004) and the effectiveness of the recovery (Khamitov et al., 2020; Van Vaerenbergh et al., 2018).

An example will serve to explain the concept of anchoring and adjustment. If someone is asked whether the Eiffel Tower is higher or lower than 1,000 meters, and then asked to make an estimate of the height of the Tower, the estimation will probably end up closer to 1,000 meters (the reference value or anchoring point) than it had if the person had been asked to state the height of the Eiffel Tower without first being presented with the reference value. This difference is called an anchoring effect and seems to take place because people use the anchoring value as a starting point when making their estimate and then adjust until they reach the value they settle for (Epley and Gilovich, 2006).

Anchoring effects are very robust psychological phenomena (Furnham and Boo, 2014) and influences estimates of objective facts (such as the height of the Eiffel Tower) as well as preferential judgments (such as judgments of willingness to pay for grocery products; Andersson et al., 2021; Yoon et al., 2019). One setting in which anchoring has a tangible effect on consumer behavior is pay-what-you-want pricing (Roy et al., 2021). Willingness-to-pay estimates tend to be adjusted toward external reference prices in pay-what-you-want situations, in particular if the situation is ambiguous to the customer (e.g. if the customers have no prior knowledge of typical prices for the products they are paying for).

2.4 The contribution of the current study: anchoring and the demand-what-you-want strategy to service recovery

In an attempt to identify a compensation size threshold value at which customers become fully satisfied after a service failure, Edström et al. (2022) used the novel technique of asking the participants to state their compensation demand after experiencing a service failure. Edström et al. did not, however, use any means by which to modulate the size of this compensation demand. In the current study, we extend the original study by Edström et al. by testing whether the compensation demand can be lowered by introducing standard compensation values as anchoring points.

Asking customers what they would demand as monetary compensation after service failure, puts a customer in a situation that arguably shares psychological features with a pay-what-you-want situation. A demand-what-you-want situation is ambiguous, and subjective factors such as fairness and emotional impression are likely to influence compensation demand judgments, just as they do in pay-what-you-want situations (Chung, 2017). Demand-what-you-want situations may similarly be influenced by salient anchoring points and reference values. Specifically, if customers are told what others typically receive as monetary compensation, and then asked how much they demand, their compensation demand judgment might end up closer to the “typical compensation value” than it had if the customers had not been told about a typical compensation level. A typical compensation value is thereby a suitable device to operationalize the anchoring effect in recovery situations to lower its costs. To test this, we formulate the following hypothesis:

H1.

A lower standard offer results in a lower compensation demand.

Given the importance of WOM in most service sectors (Kim et al., 2009, 2016), it is also relevant to test whether the anchoring strategy to service recovery has any impact on customers' willingness to recommend the service to others. A co-created recovery is generally positive for WOM, and the demand-what-you-want method provides the customer flexibility to demand a compensation level that results in satisfaction to surpass that of error-free service. While we expect anchoring to lower this level, because the reference the point used in judgments of fairness and outcome are changed (Hazé e et al., 2017; Khamitov et al., 2020), the willingness to recommend the service should not be affected as long as the customer accepts the offer. To reflect this, the following hypothesis is stated:

H2.

Willingness to recommend the service to others remains the same regardless of the size of the standard offer.

Anchoring effects from numbers are not only determined by the magnitude of the number, but also by its preciseness (Janiszewski and Uy, 2008). For example, people tend to perceive precise (e.g. 297) and round (e.g. 300) numbers differently and the size of more precise prices tends to be underestimated. Precise prices (e.g. $395,425) can even be perceived as lower than round prices ($395,000), even when the precise prices are actually higher than their round counterparts (Thomas et al., 2010). Precise versus round prices can also lead to different behavioral outcomes. For example, whether a negation leads to a deal, or an impasse, can depend on whether buyers make precise or round offerings and this in turn interacts with the roundness of the sellers' list price (Petrowsky et al., 2023). Moreover, precise offers tend to reduce the size of counteroffers in comparison with round offers (Thorsteinson, 2021). This precision effect thus suggests that precise numbers can lead to stronger anchoring points.

We expect that this precision effect can be utilized to further increase the efficiency of the demand-what-you-want approach while maintaining its effectiveness. By presenting the customers with a standard compensation in precise rather than round numbers, the demand for compensation is likely to be lower, while satisfaction still surpass that of error free service. Thus, we hypothesize:

H3a.

A smaller standard compensation results in a lower compensation demand than a higher standard compensation.

H3b.

A precise standard compensation results in a lower compensation demand than a round counterpart.

3. Overview of experiments

Two experiments were designed and conducted in order to test these hypotheses. Experiment 1 tests the effect of anchoring in a demand-what-you-want situation following a service recovery in terms of lowering the SRP threshold and willingness to recommend, i.e. Hypothesis 1 and 2. The second experiment is used to corroborate the results from the first experiment and in addition to test the viability of utilizing precision effects to further decrease compensation demands.

3.1 Experiment 1

If typical compensation levels have these expected effects on demand-what-you-want judgments and willingness to recommend the services to others, then anchoring influences the SRP threshold – providing a promising strategy to manage an efficient service recovery process. Another theoretically important implication would also follow from this finding. Edström et al. (2022) tried to identify a fixed threshold value whereby a compensation is large enough for customer satisfaction to surpass that of error-free service. If anchoring lowers the threshold value, then the threshold value is not fixed or pre-defined but changes by contextual factors.

3.1.1 Methods

3.1.1.1 Design and participants

The experiment had one independent variable (anchoring size) manipulated between participants. The anchoring size variable had five conditions: 200 Swedish Krona (SEK), 450 SEK, 700 SEK, 950 SEK as standard compensation and no compensation in the control condition. The experiment also had one independent variable with two conditions (time of recommendation estimate, before vs after service recovery) manipulated within participants. Finally, the experiment had two dependent variables: estimate of monetary demand and estimate of recommendation likelihood.

A total of 225 persons were recruited as participants. All were native Swedish speakers and were recruited by reaching out to the first two authors' social networks via social media (Instagram, Messenger, WhatsApp and Snapchat). Two participants demanded compensations of more than 15 standard deviations (SD) above the sample average and were therefore removed prior to the analysis. The final sample consisted of 223 persons (148 males [66.4%] and 75 females [35.6%]) with an age range of 18–84 years (M = 34.13 years, SD = 13.65). Participants were randomly distributed across the five conditions (N = 45 in the 200 SEK condition, N = 46 in the 450 SEK condition, N = 45 in the 700 SEK condition, N = 42 in the 950 SEK condition and N = 45 in the control condition). They participated under informed consent.

3.1.1.2 Materials and measures

A questionnaire was used to collect data, based largely on the questionnaire used in Edström et al. (2022). In the first phase of the questionnaire, the participants received background information about the study, standard research ethics information (i.e. that participation was conducted on free will, that they could leave at any moment without explaining why, and that their responses were going to be treated confidentially). They were also asked about their age and gender. In the second phase, they were presented with a scenario and asked to read it carefully. The scenario was as follows:

It is time for you to go away with your partner on a long-awaited mini-holiday consisting of an overnight stay. You have booked a double room and paid 1,500 SEK for the overnight stay. When you arrive at the hotel to check in, you are informed that the booked double room is already occupied. There are no other double or larger rooms available either. You are therefore offered a smaller single room and that an extra separate single bed is carried into the room and pushed together with the other bed into a double bed.

This service failure scenario was selected because double bookings are quite frequent (https://partner.booking.com/en-gb/help/reservations/manage/all-you-need-know-about-double-bookings) and their effect on customer satisfaction one of the most severe (Lewis and McCann, 2004). After reading about the scenario, the participants were asked to rate how likely they would recommend the hotel to others. The participants made their estimates on a scale ranging from 0 (not at all likely) to 10 (very likely).

In the third phase of the questionnaire, the participants were randomly allocated to one of five experimental conditions. In four out of five conditions, the participants were told:

You have not received the room you paid for (1,500 SEK). You are therefore offered financial compensation for the event. Usually, the compensation is: XXX SEK. What amount of financial compensation in SEK would be required for you to feel MORE satisfied with your hotel stay than if you had received the booked double room as planned (enter the amount in SEK)?

The XXX value (size of usual compensation) was either set to 200 SEK (N = 45), 450 SEK (N = 46), 700 SEK (N = 45) or 950 SEK (N = 42), depending on condition. For reference, 100 Swedish Krona (SEK) is worth about $10. These sizes of the “usual compensations” were chosen against the background of the results on Edström et al. (2022). Edström et al. (2022) identified 1,200 SEK as a threshold point for the SRP, such that participants in their study was fully satisfied with the service after receiving a compensation after service failure of that size or of a large size but not lower. The 1,200 SEK value was hence taken as a starting point and the “usual compensation/anchoring points” were set below 1,200 SEK by steps of 250 SEK. In the fifth (control) condition, the participants (N = 45) were told:

You have not received the room you paid for (1,500 SEK). You are therefore offered financial compensation for the event. What amount of financial compensation in SEK would be required for you to feel MORE satisfied with your hotel stay than if you had received the booked double room as planned (enter the amount in SEK)?

In the fourth phase of the questionnaire, all participants responded to the question by reporting a figure in Swedish Krona (SEK). Finally, the participants were asked to rate how likely they would recommend the hotel to others, given that they would receive the requested compensation. The participants made their estimates on a scale ranging from 0 (not at all likely) to 10 (very likely). The entire questionnaire took about five minutes to complete.

3.1.2 Results

As can be seen in Figure 2, telling participants how much compensation customers typically receive had an effect on their subsequent demands. The demand was highest in the control condition, where the participants were not told about a typical compensation value. When told about a typical compensation value, however, the subsequent demand dropped linearly with the size of the typical compensation value, providing support for hypotheses H1. These conclusions were corroborated by a univariate analysis of variance with anchoring condition as independent variable with five levels, F(4, 218) = 5.95, p < 0.001, η_p² = 0.10. Post hoc (Bonferroni corrected) tests revealed that the control condition differed from the 200 SEK anchoring condition, p < 0.001 and the 450 SEK anchoring condition, p = 0.025, but not the other two anchoring conditions with higher anchoring values. Thus, the anchoring value (i.e. the value of the typical compensation) had to be quite low to have a lowering effect on demand estimates. However, a curve estimation analysis with the five conditions (ordered 1–5, with the control condition set to the highest value 5 and the anchoring conditions set in descending order based on the size of the anchoring values) added as an independent variable and compensation demand estimates as dependent variable showed that the relationship between anchoring size and compensation demand is best viewed as a linear relationship, R² = 0.10, F(1, 221) = 23.63, p < 0.001, although the quadratic parameter estimate was also significant, R² = 0.10, F(2, 220) = 11.76, p < 0.001. The effect of the anchor on compensation demand estimates hence appears to depend linearly on the size of the anchor, providing further support for hypothesis H1.

Participants were relatively unwilling to recommend the hotel to others after learning about the service failure. This finding serves as a manipulation check as it confirms that the service failure scenario had the expected negative effect, and is in line with previous research (Kim et al., 2009). If they were offered the money they demanded as compensation however, they were much more likely to recommend the hotel to others. The response shift did not depend on the anchoring condition (Figure 3), thereby offering support for H2. This conclusion was confirmed by a mixed within-between repeated measures analysis of variance with time of estimate as an independent variable manipulated within person and anchoring condition manipulated between persons. The analysis revealed a main effect of time of estimate, F(1, 218) = 658.34, p < 0.001, η_p² = 0.75, but no effect of anchoring condition, F(4, 218) = 0.19, p = 0.946, η_p² = 0.003, and no interaction between the variables, F(4, 218) = 0.46, p = 0.764, η_p² = 0.008.

3.1.3 Discussion

Experiment 1 found that anchoring can be used as a technique in a demand-what-you-want service recovery context to obtain a desirable customer satisfaction level while constraining the size of the service failure compensation, thus providing support for H1 and H2. Next, we turn to investigating in more detail how these standard compensation anchor points should be administered to optimize their beneficial effects.

3.2 Experiment 2

In Experiment 2, we explored whether it is possible to take advantage of the anchoring effect and the precision effect in combination to further enhance the beneficial effects of “standard compensation offerings” in a demand-what-you-want service recovery situation. Specifically, we tested whether precise standard compensations (e.g. 247 SEK), in comparison with round standard compensations (e.g. 300 SEK) lead to a smaller compensation demand from customers. To put this assumption to a critical test, we tested whether precise anchor points (e.g. 247 SEK) would lead to a smaller compensation demand even when the round counterpart (e.g. 200 SEK) is lower. If the anchoring effect is solely driven by the arithmetic value of the anchoring point, then a standard compensation of 200 SEK should lead to a smaller compensation demand than a standard compensation of 247 SEK. In turn, if a precise anchoring point makes it stronger, then perhaps a standard compensation of 247 SEK could result in a smaller compensation demand than its 200 SEK counterpart.

Experiment 2 also provided an opportunity to replicate the main findings from Experiment 1 and test whether the results pattern generalizes to a different setting. The service failure of Experiment 1 comprised of a situation in which participants were not offered the hotel room they had paid for. For Experiment 2, the participants were instead presented with a service failure in which they were given a low-quality haircut. This latter scenario is arguably regarded by many as a more severe type of service failure since it is irreversible. Thus, Experiment 2 served to test whether the anchoring technique could serve as a tool to obtain high customer satisfaction also in this context.

3.2.1 Method

3.2.1.1 Design and participants

The experiment had one independent variable with four conditions (anchoring size) manipulated between participants. The anchoring sizes were set to 300 SEK (high round anchoring point), 200 SEK (low round anchoring point) and 247 SEK (precise anchoring point). As in Experiment 1, Experiment 2 also had one independent variable with two conditions (time of recommendation estimate, before vs after service recovery) manipulated within participants and two dependent variables: estimate of monetary demand and estimate of recommendation likelihood.

A total of 249 people (58.6% female) participated in this study. Their age ranged from 18 to 81 years (M = 41.25 years, SD = 14.59). All were native Swedish speakers and were recruited by reaching out to the fourth and fifth authors' social networks via social media. They participated under informed consent.

3.2.1.2 Materials and measures

As in Experiment 1, a questionnaire was used to collect data. In the first phase of the questionnaire, the participants received background information about the study, standard research ethics information and they were also asked about their age and gender. In the second phase, they were presented with a scenario and asked to read it carefully. The scenario was as follows:

After a two months wait, it is finally time for you to go to the hairdresser. Well there you describe your hairstyle preferences to the hairdresser and show him a picture as a reference. The haircut costs 799 SEK. When you look at the results after completed haircut you get chocked. The hair is much shorter than you wished for, and the hair tips look very uneven.

Participants in all conditions (except the control condition) were also told that the hair salon provides XXX SEK as a standard compensation in these situations. The XXX value (size of standard compensation) was either set to 247 SEK (N = 57), 200 SEK (N = 55), 300 SEK (N = 59), depending on condition. The 200 SEK value was chosen because it was the lowest anchoring value used in Experiment 1 and the one that had the largest effect on participants' compensation demand. The 300 SEK value was chosen because it was the next round value above 200 SEK and should hence result in higher compensation demands than the 200 SEK anchoring point. The precise anchoring point (247 SEK) was selected because it was close to the middle of 200 SEK and 300 SEK and expressed as a precise value. In the control condition (N = 81) the participants were not told about a standard compensation.

At the second phase of the questionnaire, the participants were requested to respond to three questions. First, they were asked “how likely is it that you would recommend this hair salon to someone else?”. The participants made their estimates on a scale ranging from 1 (not at all likely) to 10 (very likely). Second, they were asked “if you would receive a compensation, how much would be needed for you to be satisfied?”. In the experimental conditions the participants were also reminded of the standard compensation prior to making their estimate. Third, they were asked “if you would receive the monetary compensation you demanded, how likely is it that you would recommend the hair salon to someone else?”. Again, the participants made their estimates on a scale ranging from 1 (not at all likely) to 10 (very likely).

3.2.2 Results

As can be seen in Figure 4, participants' compensation demand was lower when they heard about a standard compensation prior to making their demand. A lower standard compensation (200 SEK) resulted in a lower compensation demand in comparison with a higher standard compensation (300 SEK), supporting H3a. Most interestingly, a precise standard compensation (247 SEK) resulted in a lower compensation demand in comparison with another low but round standard compensation (200 SEK), even though the precise standard was in fact lower than the round counterpart. This finding supports H3b.

These conclusions were supported by an analysis of variance with standard compensation condition as independent variable and compensation demand as dependent variable, F(3, 245) = 4.48, p = 0.004, η_p² = 0.05. The difference between the precise (247 SEK) and the low round (200 SEK) standard compensation condition was significant, t(109) = 2.99, p = 0.003, as was the difference between the low round (200 SEK) and the high round (300 SEK) standard compensation condition, t(111) = 2.29, p = 0.024.

As in Experiment 1, the service failure had the expected effect on participants' willingness to recommend the service to others, which serves as a manipulation check of the experimental manipulation (Figure 5). Furthermore, if the participants had received the compensation they demanded, they had also been more willing to recommend the hairdresser to others. This conclusion was confirmed by a mixed within-between repeated measures analysis of variance with time of estimate as an independent variable manipulated within person and anchoring condition manipulated between persons. The analysis revealed a main effect of time of estimate, F(1, 245) = 476.04, p < 0.001, η_p² = 0.66 and of anchoring condition, F(3, 245) = 10.53, p < 0.001, η_p² = 0.11, and an interaction between the variables, F(3, 245) = 14.25, p < 0.001, η_p² = 0.15. The change score (before vs after compensation) was larger in all standard compensation conditions in comparison with the control condition, all ts > 4.00. Although the change score was highest in the precise standard compensation condition, there was no significant differences in change scores between the standard compensation conditions, all ts < 1.65.

3.2.3 Discussion

Experiment 2 replicates the main findings from Experiment 1, suggesting that standard compensation offerings can reduce people's compensation demands after a service failure. Experiment 2 also extends these findings by showing that a precise standard compensation makes this effect stronger. A precise (and relatively high) standard compensation can result in lower compensation demands, even in comparison with a lower standard compensation, in line with research suggesting that precise anchor values enhance the magnitude of the anchoring effect (Janiszewski and Uy, 2008; Thorsteinson, 2021). Additionally, this suggests that standard compensations have an effect on customers' perceived justice (Chebat and Slusarczyk, 2005).

4. General discussion

This study aimed to test whether typical compensation values, as an anchoring technique, influence the compensation needed to obtain SRP in demand-what-you-want situations. To this end, the first experiment revealed that anchoring has a tangible effect on demand-what-you-want judgments after a service failure, seemingly without affecting word-of-mouth. The second experiment corroborated these findings and showed how the strategy could be implemented even more efficiently by leveraging anchoring and precision effects in combination. We thereby show that anchoring can be used as a promising tool to lower the SPR threshold in a co-creative recovery. These results have both theoretical and practical implications.

4.1 Theoretical implications and interpretations

The study contributes to the literature on service recovery by introducing a novel way to manage recovery situations effectively by adapting and co-creating the compensation outcome with customers. Despite the maturity of the service recovery field, this area is still comparatively under-researched (Khamitov et al., 2020), in particular concerning how to utilize involvement in compensation levels as a way to manage recovery situations (Arsenovic et al., 2022; Van Vaerenbergh et al., 2018).

We show that a demand-what-you-want strategy using anchoring serves as an effective way to lower the costs of recovery while maintaining a high enough satisfaction level to trigger SRP. As the customer is invited to co-create the recovery, the method proposed in this paper can thereby be used to partly address issues related to different compensation levels needed to trigger SRP due to heterogeneity between customers. Previous research has identified a range of factors moderating the effect of recovery on satisfaction (Balaji et al., 2018; Hazée et al., 2017; Kanousi, 2005; Ozgen and Kurt, 2012), which increases the relevance of adaptive service recovery (Ringberg et al., 2007). The demand-what-you-want strategy has an inherent ability to adjust and adapt recovery at the individual level as compensation is determined in the interaction between service provider and customer.

In addition to a positive impact on satisfaction, a substantial number of studies have also found perceived justice to be an important factor in determining the propensity to spread negative word-of-mouth in industries ranging from Airlines (Mohd-Any et al., 2019) to online retail (Lin et al., 2011). The effects demonstrated through the experiments reported here broadly align with this view and we expect perceived justice to be a key mechanism in demand-what-you-want recovery using anchoring as well. Previous research shows that compensation, and in particular monetary compensation (Nazifi et al., 2021b; Noone and Lee, 2010; Orsingher et al., 2010), increases perceptions of fairness (Wirtz and Mattila, 2004) and by co-creating the compensation, this effect is potentially even greater (Mattila and Cranage, 2005).

By being transparent with compensation levels typically offered to other customers in similar situations, the demand-what-you-want strategy can act as a way to provide the customer with a relevant reference point useful to evaluate the perceived fairness of the outcome. To the customer, this creates a sense of being able to evaluate the offer in relation to plausible alternative compensations and self-interests (Hazé e et al., 2017; Skitka et al., 2003). If the customer believes that the response from the firm, i.e. the compensation received, is more or less as good an outcome as could be expected given the circumstances, satisfaction levels increase. This method may also positively affect judgments about procedural justice, as the standard procedure and typical compensation communicate both how similar situations involving other customers were handled in the past, and how their service episode is handled.

The overall perception of the service encounter is also a result of social and interactional processes during the service recovery (Gelbrich and Roschk, 2011). Previous research has shown that involving the customer increases motivation and commitment to recovery (Djelassi and Decoopman, 2013) and positively affects perceived service quality (Dong et al., 2008). By communicating a standard value and inviting the customer to partake in setting the compensation level, a simple form of co-creation occurs through interaction between customers and the firm as both actors integrate resources in response to the service failure. As the customer is involved in the decision-making, perceptions of procedural justice increase (Greenberg, 1990).

Our study and results also contribute to research on service recovery by addressing the question of how to arrive at an optimal level of compensation in recovery (Gelbrich et al., 2015; Khamitov et al., 2020). Previous research suggests that the compensation offered to customers after a service failure has to be substantial to make customer satisfaction surpass that of an error-free service (Edström et al., 2022; Garg, 2013; Gelbrich et al., 2015). Triggering SRP is in many contexts worthwhile, but at the same time potentially costly and resource inefficient if it requires high compensation or even overcompensation (Kim et al., 2016).

While efforts have been made to identify a fixed threshold needed for SRP to materialize (e.g. Edström et al., 2022), our study suggests that such a threshold level is not fixed, and that a demand-what-you-want strategy can utilize this fact to make recovery through compensation more efficient. We show that the compensation does not need to be particularly large if the customer is asked to make a compensation claim after a typical or standard compensation level is presented to the customer.

Based on past research, it is not clear if preferential judgments such as compensation demands are robust to the influence from anchoring, or whether they too are susceptible to anchoring effects just like judgments of objective facts (Andersson et al., 2021; Bunčić et al., 2021; Yoon et al., 2019). The susceptibility to extraneous information on these compensation demand judgments, as found in the current paper, suggests that compensation demands are not fixed. Instead, they appear to be “constructed” based on salient and available information at the moment of making the judgment, much like other judgments and decisions (Slovic, 1995). While it could be possible to identify a regular SRP threshold by simply asking customers what they demand as compensation (Edström et al., 2022), the results reported here show that this threshold is fluctuating and dependent on both the level of a standard compensation used as a mental anchor and the precision of that anchoring point.

When examining the differences between the conditions in the first experiment (Figure 2), people seemingly tend to anchor their estimates of compensation demand in the anchoring value and then adjust (upwards) until they reach a value they settle for, thus ending up with lower estimates in comparison with a condition without an anchoring point (Epley and Gilovich, 2006). As shown in Figure 4, the compensation demanded is also susceptible to precision effects (Thomas et al., 2010), where a higher but precise standard compensation value is more effective in lowering the SRP threshold than a lower round number. This aligns with previous research from other settings showing that a more precise value has a stronger anchoring effect than round or less precise values (Loschelder et al., 2016). In the context of demand-what-you-want estimates in service recovery, the precise anchoring seems to negatively affect the upward adjustment of the customer and lower it, as demonstrated by Experiment 2.

We see two mechanisms that may account for this result. First, precise numbers have been suggested to require more cognitive effort to process than round numbers, which thereby introduce feelings of uncertainty that strengthen the anchoring effects of precision (Janiszewski and Uy, 2008; Thomas et al., 2010). In the context of service recovery through compensation, the uncertainty likely relates to judging, evaluating and deciding on a level of compensation perceived as reasonable or fair. Under this perceived uncertainty, the precise value's anchoring effect on customers is more pronounced compared to a round, higher number. Second, a possible mechanism behind the results seen in the second experiment is the effect of scale granularity. Under this effect, the value offered to the customers in the precise compensation condition primes them to evaluate the offer on a mental scale of higher resolution or granularity (Janiszewski and Uy, 2008), which in turn makes them more inclined to deviate from the anchoring point (the standard compensation) in smaller increments (Frech and Loschelder, 2020). Compared to the groups offered round numbers, this group would then perceive their upward adjustments as larger deviations from the original anchor, thereby arriving at a lower final compensation value.

Considering the results from two experiments together, an additional underlying reason behind the effects could be emotion as an input affecting the evaluation process (Peters et al., 2006). The large difference between the anchoring values and the demand-what-you-want estimates indicates that customers take interpersonal “fairness” into the evaluation process when deciding upon an appropriate demand that would make them feel satisfied enough to feel that perceived justice has been served. If compensation is perceived to be large compared to what others receive, people seem to settle with that compensation size, even when the compensation the standard person receives is quite low. This aligns with the commonly held assumption in previous research (Khamitov et al., 2020) that perceived justice of outcomes is evaluated in relation to relevant reference points like other customers (Deutsch, 1985; Greenberg, 1996). The typical compensation is likely used in the customer's subjective evaluation of the compensation as more or less favorable, given plausible alternatives and outcomes.

Given the influence of descriptive norms on judgment and decision making (e.g. Andersson et al., 2022), one could expect judgments of compensation demands to end up close to the standard compensation value. However, in the condition with the lowest anchor value (200 SEK) the average compensation demand was three-four times as much as the anchor value. This further reinforces the conclusion that the anchor values are not purely processed as “cold” quantitative inputs that have full control over the demand-what-you-want judgments, but that emotion and perceived fairness also play roles in addition to the mechanisms and heuristics outlined above. Our study thereby adheres to calls for the integration of psychological perspectives at the individual level (Khamitov et al., 2020), which we utilize to propose a novel method to service recovery, and identify its likely mechanisms.

4.2 Managerial implications

Service-oriented companies must consider the decremental effect of offering larger compensations in their service recovery strategies to avoid overcompensation—that is, setting compensation sizes that waste unreasonably large resources in relation to the gains obtained in customer satisfaction (Hazarika et al., 2019). An important factor to avoid such overcompensation is the threshold of the SRP (Edström et al., 2022)—namely, the point at which compensation is large enough for customer satisfaction and the feeling of fair procedures to surpass that of error-free service. Compensations larger than this can be considered unnecessary and a waste of resources from the company's viewpoint. The findings from the current study have several managerial implications. First and foremost, they demonstrate that anchoring compensation demands has the potential to lower service recovery costs while still achieving its goal of increasing customer satisfaction through a demand-what-you-want strategy. This indicates that the overall costs for the company can be reduced if service recovery processes is managed with an anchor, and still have satisfied customers. The experiments were conducted in the hotel and hairdressing sectors, where the anchoring technique was demonstrated to be present and effective. While these are only two empirical settings, it is likely that these findings also generalize to other sectors where service failure occurs and recovery is associated with costs for the firm. In addition, the principles outlined in this paper are also likely valid in settings where the compensation is made in other than monetary form, like additional services, upgrades or future credit. Implementing this strategy should therefore be worthwhile for firms in various contexts, but we see a particular relevance in sectors where WOM plays important roles in marketing, such as e-commerce, the hospitality sectors and specialized consumer goods. While the service recovery, in line with previous research, was shown to impact WOM, the different anchoring points did not. This implies that firms can implement the strategy without trading a lower cost of service recovery against less beneficial WOM – even if the firm seeks a low compensation level.

To implement a demand-what-you-want strategy, managers should provide front-line employees a mandated range of compensation concerning typical service failures and their perceived severity. Routines are needed to specify anchoring values that correspond to these, along with a ceiling for how much employees can compensate customers in the event of service failure. Each firm should seek to test and adjust values over time to fit its specific context. While the first experiment demonstrated a linear effect of lowering anchor values on compensation demands, the absolute values serve mainly as proof of concept. It is clear from the second experiment that the precision effect incurred by using exact numbers to anchor customers' compensation demands should be utilized in this regard.

It is also important to note that implementing the anchored demand-what-you-want strategy is not based on misleading customers with typical compensation values that in reality are not typical at all. Instead, the anchoring value should preferably be framed as a standard, rather than an average or typical value of compensation.

Furthermore, the anchored demand-what-you-want strategy also presents an opportunity for managers to implement a collaborative service recovery process. A central purpose of compensating a service failure is to make the customer feel a sense of perceived fairness, as this impacts satisfaction (Albrecht et al., 2019; Ma and Zhong, 2021; Orsingher et al., 2010). Here, the collaboration process between the customer and the firm is of importance as collaboration in service recovery is positively related to a customer's emotional responses (Arsenovic et al., 2022). Conversely, should an employee act in a way that is perceived negatively, despite offering fair compensation, the experience can still affect the customer negatively (Blodgett et al., 1997). The anchored demand-what-you-want strategy could thereby increase customer's levels of perceived fairness from the compensation, because the customer takes an active part in the recovery process by adjusting its size closer to what they feel is fair. If so, the strategy could have both short- and long-term benefits for the company and be used to mitigate some of the bad-mouthing behaviors sometimes associated with monetary compensations (Arsenovic et al., 2022).

4.3 Limitations and avenues for future research

In the current study, we did not capture the usual control variable in the service failure and recovery literature but instead used the participants' intention to recommend the service to others as a check of the success of the manipulation of service failure. In particular, we used the change in intention to recommend the service before and after the compensation demand estimate as a manipulation check.

Another limitation of the current study was the use of hypothetical scenarios. While using hypothetical scenarios to elicit and study service failures is common (e.g. Khamitov et al., 2020; Lewis and McCann, 2004; Magnini et al., 2007) and recommended to obtain high experimental control of the effects under study (Edström et al., 2022; Ok et al., 2007), it is unclear whether the same pattern of results would be obtained in a real-world setting. Even if the robustness of anchoring effects suggest that the effects will manifest in real-world settings as well (Furnham and Boo, 2014), it is unclear whether the effect sizes would be similar. Testing the effects of anchoring and precision in real-world demand-what-you-want service recovery is therefore an important target for future research. Future research should also address how the anchoring and precision effects are modulated by customer individual differences and contextual factors, which were not addressed in the current study. For example, future research could address failure intentionality, failure globality and failure reversibility (Nazifi et al., 2022) in the context of the demand-what-you-want-strategy to further explore the effects of situational factors.

A final suggestion for future research is to look at the effects of other type of compensations in the demand-what-you-want context. While monetary compensation is often superior to other type of compensations such as vouchers or apologies (Noone and Lee, 2010; Orsingher et al., 2010), compensations after a service failure seem most effective when matched with the type of service failure (Nazifi et al., 2020). Future research could investigate how type of compensation interacts with type of service failure in a demand-what-you-want context.

5. Conclusions

Effective service recovery is pivotal in most service industries (Van Vaerenbergh et al., 2019), and despite much research, how to manage recovery after service failure remains a central question in service research. Increasingly, firms shift towards a co-created process involving the customer and with the ability to adapt recovery to heterogeneous needs. Following recommendations to integrate psychological perspectives at the individual level (Khamitov et al., 2020), the novel recovery method we outline offers a way to achieve satisfaction levels matching or surpassing that of error-free service, while lowering the threshold needed for this to occur. The customer partakes in creating an individualized recovery outcome that is sufficient to trigger the SRP, but also resource efficient as it lowers typical monetary compensation demands. Our study thereby aligns with calls for research on managing co-created outcomes (Arsenovic et al., 2022) and compensation levels in service recovery (Gelbrich et al., 2015; Khamitov et al., 2020). The results also show that lowering the compensation level through this method does not impact customers' willingness to recommend the service to others. The recommendations from peers are increasingly central to consumer decisions, and adequate compensation is often needed to avoid negative WOM (Piehler et al., 2019). As such, the method we propose offers firms a concrete tool to manage recovery efficiently and effectively while maintaining the positive effects of a successful recovery on customers' willingness to recommend the service to others.

Figures

Figure 1

The figure depicts the concepts involved in the service recovery paradox

Figure 2

Mean estimates of monetary compensation (in Swedish Krona, SEK) demanded after service failure in order to be more satisfied than with error-free service

Figure 3

Mean estimates of probability to recommend the hotel to others after service failure (but prior to receiving compensation) and after receiving the demanded compensation

Figure 4

Mean estimates of monetary compensation (in Swedish Krona, SEK) demanded after service failure in order to be satisfied with the service across four conditions: high round (300 SEK), low round (200 SEK), precise (247 SEK) and no (Control) standard compensation condition

Figure 5

Mean estimates of probability to recommend the hairdresser to others after service failure (but prior to receiving compensation) and after receiving the demanded compensation, across found conditions: high round (300 SEK), low round (200 SEK), precise (247 SEK) and no (Control) standard compensation condition

NK and JB contributed equally to this study. NK and JB designed Experiment 1, conducted the data collection and analysis and wrote a first draft of the manuscript. MP and HM also contributed equally to this study, designed Experiment 2, conducted the data collection and analysis. PE and BB wrote parts of and made valuable comments on the paper. PS conceived the original idea and wrote large parts of the paper. The authors declare there are no conflicts of interest.