Pre- and in-process: does timing of delay matter to airline passengers?

Pre- and in-process: does timing of delay

matter to airline passengers?

An analysis of timing of flight delays and customer satisfaction,

moderated by travel purpose and flying frequency.

Michelle van de Boel Master Marketing Management

Pre- and in-process: does timing of delay matter to airline passengers?

An analysis of timing of flight delays and customer satisfaction,

moderated by travel purpose and flying frequency.

MSc Marketing Management Master Thesis

January 14, 2019

First supervisor: Dr. J. van Doorn Second supervisor: Dr. J. T. Bouma

University of Groningen Faculty of Economics and Business

Table of Contents

Table of Contents ... 3

Abstract ... 5

1. Introduction ... 6

2. Theoretical framework ... 10

2.1 Literature review ... 10

2.2 Conceptual model ... 18

2.3 Hypotheses ... 19

3. Methodology ... 23

3.1 Data collection ... 23

3.2 Measurements ... 23

3.3 Plan of analysis ... 26

4. Results ... 27

4.1 Data preparation ... 27

4.2 Descriptive statistics ... 27

4.3 Testing assumptions ... 28

4.4 Analyses ... 29

4.4.1 Model 1: main effects ... 30

4.4.2 Model 2: main- and interaction effects ... 31

4.4 Robustness checks ... 33

5. Discussion ... 34

5.1 Overview of results ... 34

5.1.2 Satisfaction after pre- and in-process delays ... 35

5.1.3 The moderating role of travel purpose ... 35

5.1.4 The moderating role of number of past service encounters ... 36

5.2 Managerial implications ... 37

References ... 39

Appendices ... 45

A.

Background information the company ... 45

B.

Questionnaire ... 45

C.

Testing assumptions ... 45

Abstract

Based on the field theory of Lewin (1943) we suggest that the timing of delay matters to consumers. Depending on the stage of service delivery, pre-process delays can yield more negative responses than in-process delays. Empirical research relating individual factors and timing of service failures is scarce. In this study we investigate whether airline customers are more affected by a delay occurring at the gate (pre-process) or when seated in the aircraft (in-(pre-process). Moreover we examine the possible influence of individual factors – travel purpose and flying frequency – on the strength of the relationship between timing of the delay and customer satisfaction. We analyzed a large sample of survey data collected by the company. Results are congruent with the field theory, pre-process delays appear to have a stronger effect on customer satisfaction opposed to in-process delays. Besides, business travelers are less affected by pre-process delays than leisure travelers, moreover we found a weak effect suggesting that business passengers are more affected by process delays compared to leisure travelers. Passengers travelling frequently perceive an in-process delay less negatively than non-frequent flyers. Moreover, a weak effect seems to indicate that pre-process delays are perceived more negatively by frequent flyers. Findings can be used by designing recovery strategies as they provide a meaningful insights in understanding the customer satisfaction after pre- and in-process delays.

Keywords: Service failures, field theory, pre-process delay, in-process delay, customer satisfaction,

1. Introduction

Failing to deliver the promised service is a serious challenge for organizations in the service sector (Hess, 2008). Compared to the production process of most tangible products, intangible services are more labor-intensive which results in a greater heterogeneity (Berry, 1980). A service failure can result in many undesirable consequences for companies. For example, Dave Carroll was passenger on a flight of United Airlines in 2008. A failure occurred when his guitar was broken during the baggage handling process. After a failed recovery (no damage compensation), Dave wrote a song called: “United Breaks Guitars” which went viral and reached millions of people (Carroll, 2012). As illustrated by this example and established in literature, product and service failures can lead to dissatisfied customers (e.g. Hess, Ganesan & Klein, 2003; Tsiros, Mittal & Ross, 2004) anger (e.g. Casado Dìaz & Más Ruíz, 2002) lower repurchase intentions (Hess, 2008), complaint intentions and can even evoke a desire to hurt the company (e.g. Grégoire & Fisher, 2007). Hence, it is meaningful, for businesses and researchers, to gain a thorough understanding of service failures and their effect on customers.

There are several factors related to customer reactions to service failures. In addition the aforementioned topics, there is a lot of empirical research in the field of service failures focused on service recovery (Gonzalez, Hoffman, Ingram & LaForge, 2010; Hoffman & Kelley, 2000), perceived justice (Chang & Chang, 2010) failure attributions (Iglesias, 2009; Van Vaerenbergh, Orsingher, Vermeir, Larivière, 2014) and information supply (Butcher & Kayani, 2008; Munichor & Rafaeli, 2007; Currie & Muir, 2017). Remarkable is however, that there is little research conducted concerning the timing of the service failure.

Delays and wait are an inevitable elements in the service industry although often not planned nor desired (Butcher & Kayani, 2008). Wait can occur at different moments of the service delivery; before (pre-process), during (in-process) and after the transaction (post-process) waits. (Dubé-Rioux, Schmitt & Leclerc, 1989). Pre-process waits can be further categorized into three types: pre-scheduled waits, queue waits and post-scheduled waits (delays) as identified by both researchers Baranishyn, Cudmore and Fletcher (2010) and Taylor (1994). Pre-schedule wait occurs when the customer arrives too early for a scheduled event/appointment and has to wait for the service to start. The queue wait occurs when no specified time is indicated, usually the service is provided on a ‘first-come-first-service’ principle. Post schedule delay, indicate a delay of service. For this thesis we will focus on the latter: post schedule delays.

According to Sparks and Fredline (2007, p. 242), a “service failure, or service breakdown, can be defined as service that does not meet the customer’s expectations”. Hence, expectations do not match with the service performance resulting in a decrease of satisfaction with the service. Satisfaction is an important measurement for business as there is an overall belief that customer satisfaction stimulates an increased loyalty, (re-)purchase intentions, evoke positive worth of mouth and return on investment (Bowden, 2009). However, as illustrated by the ‘United Breaks Guitar’ example (Carroll, 2012), a service failure can have the opposite effect. We will attempt to explain the service failure – satisfaction relationship by theory. In line with the SERVQUAL model (for measuring service quality) of Parasuraman, Zeithaml and Barry (1985), Lovelock and Wirtz (2011) theorize that there is a delivery- and perception gap. The delivery gap entails that companies should make sure that the performance meets standards (if not: a service failure occurs). The perception gap depicts service delivery versus perceived service. The perception and expectations of customers are resulting in an emotional response, this response can be defined as the customer satisfaction (Mouwen, 2015). Oliver’s (1989, p. 1) definition of satisfaction is aligned with the previous as “an evaluative, effective, or emotional response”. Consumers are believed to hold different expectations and perceptions for the same service. This can possibly be explained by the individual characteristics which contribute to different customer satisfaction for services (Cheng & Tsai, 2014).

identified investigating the role of goal importance affecting different stages of delays / waits. Those studies are however mostly focused on queue waits (Meyer, 1994; Schmitt, Dubé & Leclerc, 1992) or restaurant settings (Dube-Rioux, Schmitt & Leclerc, 1989; Yang et al., 2013). Up until now, we did not find any study combining the timing of the delay with travel purpose. Another factor, different for each individual, is the travel frequency. Previous research suggests that frequent usage of a service results encourages retaliation as customers feel betrayed (suggested by Grégoire & Fisher, 2007 based on justice theory). Customers who are frequently exposed towards delays might initially feel habituation and get used to the delay without worrying. However, when delays occurring more frequently can eventually result in satiation (Karmarkar & Karmarkar, 2014). The prior frequency does not have to be limited to “the number of past encounters with the organization” but can also be defined as the “consumption of the service, [which] is the number of previous encounters with all providers within a particular service industry” (Hess at al., 2003, p. 128, 133). For this thesis we will apply the latter definition. Currently, there is a lack of empirical research examining the prior experiences in relation to timing of delay.

Overall, we state that service failures, delays in particular, negatively influence service evaluations which may affect the business. Delays in the airline industry are difficult to manage, hence, a better understanding of factors affecting the customer satisfaction after delays is desired. We will achieve this through a quantitative study conducted over a long time period. The research concerns delays in the airline industry occurring before and during the flight. Hereby we include the moderating effect of travel frequency and travel purpose. The main research question of this study can be formulated as follows: To what extent is customer satisfaction affected by the timing of the delay? This effect is further examined through the following sub-questions: To what extent is the satisfaction after a delay experience influenced by travel purpose? To what extent is the satisfaction after a delay experience influenced by prior experience?

2. Theoretical framework

This chapter provides a concise overview of established literature in which several studies are discussed concerning the topic of this thesis. We focus on a few concepts, central for this thesis: the influence of the timing of service delays, satisfaction and individual characteristics that might influence these relations. Consequently, a conceptual model is depicted – the framework of this thesis – from which hypothesis are deduced.

2.1 Literature review

Service quality and customer satisfaction are strong indicators for predicting future customer loyalty in the airline industry (Zins, 2001). Service failures diminish the customer satisfaction (see table 1) and should therefore be closely monitored by service providers. Based on literature regarding service failures (Hoffman, Kelley, Rotalsky 1995; Keaveney 1995; Mohr & Bitner 1995), researchers Smith, Bolton and Wagner (1999) identified two types of service encounter failures: outcome- and process failures. In the outcome dimension, the service provider fails to deliver the basic needs for the service (e.g. the hotel room is not available due to overbooking). In the process dimension, the manner in which the organization delivers the service is inadequate (e.g. customer has to wait for cleaning before entering the hotel room). Because process failures are not very clear in nature, it is valuable to investigate process failures in further detail.

By distinguishing delays in pre- and in-process delays we aim to discriminate between the timing of delay occurrences based on the stage of the service encounter. As proposed by Lewin’s Field theory (1943) delays occurring before the service starts (pre-process) are expected to evoke more negative affective reactions than when a delay happens in the middle of the service encounter (Dubé, Schmitt & Leclerc, 1991). Yet, little empirical research is conducted in validating this theory, moreover, there are studies arguing that in-process delays elicit more negative reactions and lower service evaluations than delays occurring further from the goal state or after goal achievement (Yang et al., 2013). Moreover, Durrande-Moreau (1999) highlights that most research focusses on situational factors (e.g. fairness, uncertainty and manipulation of environment) influencing the delay and waiting time rather than individual factors (e.g. mood, time-pressure, personal expectation).

provider. Individual factors on the other hand, concern the service value, habits, motivation, mood preceding the wait, time pressure, and are difficult to control for.

There are many studies focusing on situational factors. Some studies attempt to influence the waiting perception through interventions such as playing classical music during wait (McDonnell, 2007) or assess the availability of shopping options and reading material at train stations (Van Hagen, van Pruyn, Galetzka & Kramer, 2009). Munichor and Rafaeli (2007) conducted a study in a call center. It appears that fillers while waiting on the phone, can influence satisfaction. Most negative reactions derive from hearing apologies while waiting whereas information about the location in the queue yields positive reactions. Hence, the kind information provided is an important factor in reducing the negative effects of a delay (Butcher & Kayani, 2008) and also the timing of delivering delay information has found to significantly impact the customer satisfaction (Currie & Muir, 2017). Other situational factors, well-established in research, are: attribution of delay cause (e.g. Choi & Mattilla, 2008; Hui & Toffoli, 2002; Iglesias, 2009; Van Vaerenbergh et al., 2014) and recovery after service failure (e.g. Chang & Chang, 2010; Hess et al., 2003).

Remarkably, there are not many studies focusing on individual factors. Presumably because (while theoretically relevant) they have limited managerial implications as they are often not controllable in practice. It is difficult to identify and adjust strategy according to someone’s mood for example (Yang et al., 2013). Price is often regarded as the cost for a service. However, non-monetary elements such as time, psychical and psychic efforts are also at interplay when the consumer wants to obtain the service (Lovelock, 2001 as mentioned by Tam, 2004) The importance of those factors differ per individual, situation and depends on the nature of the service (Tam, 2004). There are two individual characteristics we aim to elaborate on: customer relationships and travel purpose.

whether pre- or in-process delays evoked more negative reactions in high- versus low need conditions. They found that, under high needs, pre-process delay resulted in a greater disaffection than in-process delays. This was not congruent with the field theory, however they suggested that those findings could be unique to the industry (restaurants) and proposed that further research amongst other industries was required. We found a couple of other researches related to the timing of delay and goal importance. The subjective goal importance is increased when the distance towards the goal is smaller. More specifically, when waiting in line, people tend to react stronger towards (illegitimate) intruders when they are closer to the goal of being served (Schmitt, Dubé & Leclerc, 1991). Besides distance towards the goal, people can also perceive the goal differently. When waiting in line for a non-attractive goal, distance from the goal affects his/her mood (Meyer, 1994). People who perceive the goal positively have a promotion focus rather than a prevention focus. Researchers found that promotion-focused consumers perceive delays further from the goal as more negative than when they are closer to the goal and vice versa (Yang et al, 2013).

Leisure travelers might have a greater promotion focus compared to business travelers. Hence, leisure trips can decrease the negative perception of a delay (Cheng & Tsai, 2014). Within the airline industry it is common sense to differentiate between economy and business class passengers. Teichert, Shehu and von Wartburg (2007) question whether this is still a valid segmentation method. Due to increased competition there is a lot of pressure on pricing with rates decreasing rapidly. This development results in a behavioral change of consumers always searching for the best deals and thus the share of consumers travelling economy class increased (Mason, 2005). The travel purpose differs per class including the customer needs. Passengers flying for business reasons attribute more value to comfort and are less price-sensitive compared to leisure trips in both business and economy class (Teichert et al., 2007). On the other hand business class indicates more desire for comfort whereas economy class indicates a greater price sensitivity.

for revenge (Grégoire, Trip & Legoux, 2009), customer vengeance (Bechwati & Morrin, 2003) or retaliation (Grégoire & Fisher, 2006, 2007).

Table 1

Overview of literature

Key concepts Author Main findings

Service delay and satisfaction

(selection)

Tam (2004) By means of an empirical study about restaurant experience researchers established the following relationships: perceived monetary costs, perceived time costs negatively affect perceived value which in turn influences customer satisfaction and post-purchase behavior. Perceived service quality has a positive influence on customer satisfaction (direct and indirect through perceived value) and consequently post-purchase behavior.

Service delay and satisfaction

(selection)

Baker (2013) A qualitative study of a four year period found that the overall satisfaction can be explained with the perceived service quality (delays, denied boarding, mishandled baggage and complaints).

Timing of delay Ben-Akiva & Lerman (1985)

Researchers observed that passengers perceive the wait duration of one minute at the bus stop more negatively than one minute spent waiting inside the bus.

Timing of delay Davis & Maggard (1988)

In the context of fast-food restaurants, it appears that queue wait duration has a stronger impact on the customer satisfaction than the actual service duration.

Timing of delay Dubé, Schmitt, Leclerc (1991)

In an experimental study students were found waiting for their teacher in the class room. Delays occurring during the pre- and post-process resulted in more negative affective reactions than delays happening during the process.

Timing of delay Hui & Thakor (1998)

opposite, in-process delays evoking more negative reactions than pre-process delays, was found for correctional delays.

Timing of delay Dabholkar and Sheng (2008)

A study assessing the impact of download delays found that: “delays near the start of the download are perceived longer than later in the process, and time pressure worsens the effect of download waiting at earlier stages of delay.” (p.1415)

Timing of delay Friman (2010) Based on positive, neutral and negative written scenario’s regarding the waiting time for the bus, Friman found that negative experiences are perceived more positively during to in-process waits (waiting in the bus) as opposed to pre-process waits (waiting at the bus station).

Timing of delay, goal importance

Dube-Rioux, Schmitt and Leclerc (1989)

In a restaurant setting, customers were more upset when a delay occurred during the pre-process (before the person had the chance to be seated and order a meal) and post-process (when the subject would like to pay) as opposed to a delay occurring in the middle of the dining experience (entrée arrives late). In contrast with the field theory they found that subjects in the high need condition (hungry) evaluated the service more negatively when a delay occurred during the pre-process compared to the in-process stage. Contrary results were found for subjects in the low need condition (not hungry).

Timing of delay, goal importance

Schmitt, Dubé and Leclerc (1992)

When waiting in line, it appears that people are more likely to raise objections when they are standing after the point of intrusion rather than when they are ahead of the intrusion point. Reactions to intrudes were stronger when the person was close to the goal of being served.

Timing of delay, goal importance

Meyer (1994) Meyer observed participants waiting in line in relation to the goal attractiveness. He found that when the goal was considered more attractive, judgements were less context related. However when the goal was considered less attractive, time spent in the queue and distance from the goal influenced the mood of the subjects.

Timing of delay, goal importance

Yang, Mattila and Hou (2013)

tested the effects for two groups with different levels of regulatory focus. Results indicate that promotion-focused consumers perceive a pre-process delay as more negative resulting in lower service quality evaluations as opposed to an in-process delay. Prevention-focused consumers, on the other hand, generate more negative emotions and lower evaluations after an in-process delay than after pre-process delay. Delays, personal characteristic travel purpose Kashyap and Bojanic (2000)

Surveys were sent to guests of an upscale hotel. The researchers found that quality perceptions of leisure and business travelers differ for the quality of… the room, public areas, staff and services, price.

Delays, personal characteristic travel purpose

Kim, Lee & Oh (2009)

Research indicates that the importance of information about crowdedness in the bus was affected by waiting time and travel time to the bus resulting in people choosing either the first or second arriving bus. Effects of bus occupancy information are different for trip purpose (commute/non-commute) and user group (white- or blue-collar workers, students, homemakers, self-employed and elderly people)

Delays, personal characteristic travel purpose

Cheng and Tsai (2014)

Tolerance for train delays appears to be differ for various passengers’ characteristics. In contrast to men, females in all age categories believe that being on a leisure trip can improve the perceived waiting time.

Customer relationship

Dagger and O’Brien (2010)

Based on a sample of service consumers from nine industries, researchers found that drivers of loyalty are different for novice and experienced consumers. Confidence benefits are more beneficial to novice customers whereas social benefits and special treatments are more beneficial to experienced customers.

Customer relationship and service failure: love is blind Hess, Ganesan and Klein (2003)

Consumers who expect the relationship to continue have lower expectations towards service recovery and attribute the service failure to a less stable cause which results in a higher level of satisfaction with the service performance after recovery.

Customer relationship and service failure: love is blind

Grégoire & Fisher (2006)

Researchers test the effect of relationship quality (RQ) on the desire for retaliation mediated by the firm’s controllability. Findings support the proposition of the “love is blind” proposition. This entails that, in case of low controllability, customers with a high RQ experience a lesser desire for retaliation after service failure than customers with a low RQ. No significant results have been found for the “love becomes hate” proposition, indicating that in cases of high controllability, high RQ customers experience a greater desire for retaliation than low RQ customers. Results do confirm that, overall, high controllability results in a greater desire for retaliation.

Customer relationship and service failure: love is blind

Vázquez-Casielles, del Río-Lanza and Díaz-Martín (2007)

Results obtained from a study in the airline industry indicate that higher customer perceptions of past service performance result in more positive attributes regarding stability (service failure’s cause is less stable) and control (service failure’s cause is less controllable) which positively influences satisfaction indirectly through emotions. Customer relationship and service failure: love becomes hate

Grégoire & Fisher (2007)

A justice-based model is designed and tested on a sample of airline customers who had complained after service failure. Results suggest that customers feel betrayal after service-failure and poor recovery which motivates them to restore fairness through retaliation. Customer relationship and service failure: love becomes hate

Grégoire, Tripp & Legoux (2009)

Customer relationship and service failure: love becomes hate

2.2 Conceptual model

Currently, little research is conducted regarding the timing of delay. Research that has been conducted is usually in the context of a queue wait or restaurant setting. With this study we want to test the effect of the timing of delays in the aviation industry. We aim to answer the proposed research questions: To what extent is customer satisfaction affected by the timing of the delay? To what extent is the satisfaction after a delay experience influenced by travel purpose? To what extent is the satisfaction after a delay experience influenced by prior experience?’ In order to answer these research questions we designed a conceptual model as depicted in figure 1.

In the model we relate the different timing of delay (independent variables) to customer satisfaction (dependent variable). In line with the field theory of Lewin (1943), we expect that pre-process delays evoke more negative emotions than in-process delays. Previous studies support this statement (a.o. Dabholkar & Sheng, 2008; Friman, 2010). Moreover, we propose that individual factors influence affective reactions after a delay experience (Yang et al., 2013). As illustrated in figure 1, the conceptual model, we examine the number of past encounters with the service and travel purpose. The first is travel purpose, defined as either business or leisure which are assumed to coincide with different needs (Teichert et al., 2007). Moreover we suggest that experienced users infer a different satisfaction from a service compared to inexperienced, or new, users (Bowden, 2009).

2.3 Hypotheses

From theory and literature we derive five hypotheses which we will present in this section. Starting with the main effect, we expect delays to have a negative effect on customer satisfaction. This assumption is derived from models distinguishing between expectancy, perception and service delivery (SERVQUAL and models of expectancy-disconfirmation). Hui, Kandampully and Juwaheer (2009) investigated the relationship between service quality and satisfaction, they developed and tested a model suggesting that high service quality increases customer satisfaction (eventually improving the company’s image, leading to customer retention). A widely used measurement of service quality is the SERVQUAL model from Parasuraman et al. (1985, revised version followed in 1991) (Lovelock & Wirtz, 2011; Yarimoglu, 2014). With the model Parasuraman, Zeithaml and Berry (1985) assess the discrepancy between service expectations and performance of the service provider, driving the perception of service quality (Ma, Harvey & Hu, 2007). Researchers propose that the performance-based measures are most reliable for measuring service-quality perception (Babakus & Boller, 1992; Lee, Lee & Yoo, 2000). One performance-based measure is timely service. In case of a delay, the service performance fails and customers are likely to be affected by it as their initial expectations are negatively disconfirmed (as clarified in the disconfirmation model of Oliver, 1980). Besides the dissatisfaction which can arise after a delay, we believe that the point of delay occurrence matters whereby we assume that pre-process and in-process delays verify in their effect on customer satisfaction.

The assumed discrepancy for the timing of the delay is based on the field theory of psychologist Kurt Lewin (1943). Even though the theory stems from 1943, scholars argue that the theory is still valuable and relevant today (for more details, see the review of Burnes & Cooke, 2013). The theory states that it is possible to understand consumer behavior through psychological forces, influencing behavior at a certain point in time (Burnes & Cooke, 2013). Those forces restrain people from obtaining their goal and are believed to be stronger when further from the goal achievement as opposed to forces close to the goal achievement. As mentioned earlier, the field theory implies that consumers perceive delays differently depending on the timing of the delay. The theory proposes that: the perceived waiting time is assumed to be longer and affective responses more negative when a delay occurs further from the goal achievement (pre-process) compared to delays happening close to the goal achievement (in-process) (Yang et al., 2013).

delay occurrence in the airline industry. Based on the field theory of Lewin (1943) the following hypothesis is formulated:

H1: Pre-process delays have a greater negative effect on customer satisfaction compared to in-process delays.

Secondly, we propose that individual factors such as the motivation for travel, or travel purpose, influence customer evaluations of the service after a pre-or in-process delay. There are several theories related to this claim. First of all, the paradox of utility after an imposed wait. A delay can lead to a positive utility; as the pleasurable anticipation of the forthcoming experience increases but on the other hand a delay can also result in a negative utility which is the aversive experience of wait (Nowlis, Mandel & Mccabe, 2004). Yang et al. (2013) aim to explain the discrepancy between positive and negative reactions towards delay by suggesting that “the positive utility dimension is more salient in predicting promotion-focused consumers’ reactions to an imposed delay, whereas the negative utility dimension is more salient in predicting prevention-focused consumers’ reactions to an imposed delay.” (Yang et al., 2013, p. 4). Promotion-focused consumers treat positive aspects as more important than negative aspects of the service in their evaluation whereas prevention-oriented consumers focus more on the negative outcomes of the service (or service element such as a delay) (Higgins, 2002). Regarding travel purpose, we expect leisure travelers to anticipate a positive experience and be more promotion focused than customers traveling for business purposes. Therefore we propose that passengers traveling for business purpose perceive a delay more negative than passengers travelling for leisure purposes.

In addition to the previous mentioned, we recognize the importance of the level of need or urgency. The field theory of Lewin (1943) proposes that different levels of need create forces of different strength and, as a consequence, more or less pressure toward the goal (Dube-Rioux et al., 1989). When translated to the aviation industry this might imply that passengers travelling for business purposes experience a greater need of arriving on time at the destination than leisure travelers (e.g. because they have meetings to attend). Hence, the desired end of arriving on time might be greater for business travelers. Consequently, we expect that a delay can be evaluated more negative by business travelers than leisure travelers. Following the field theory we expect that barriers, a delay, occurring closer to the goal (in-process) yields more negative reactions in high need conditions (Yang et al., 2013).

the theory in the context of the airline industry, we propose that a departure delay, when inside the aircraft, evokes more negative reactions for business travelers than for leisure travelers. This because business travelers have a greater urgency to arrive on time for the business meeting for example.

Based on the above mentioned theory and the assumption that, in general, delays have a greater effect on customer satisfaction for business travelers rather than leisure travelers, the following hypotheses are defined:

H2a: Pre-process delays have a weaker negative effect on customer satisfaction for passengers travelling for business purposes compared to passengers travelling for leisure purposes. H2b: In-process delays have a stronger negative effect for passengers travelling for business

purposes compared to passengers travelling for leisure purposes.

Thirdly, we will elaborate on the differential effects of pre- and in-process delays on satisfaction due to the moderating effect of the number of previous encounters with the service. In general, we assume that frequent flyers perceive delays more negatively compared to non-frequent flyers. Yet, there is some inconsistency in literature whether customer relations buffer or magnify the negative effects of service failures. The love is blind proposition implies that customers with high affective commitment towards the service provider, are more likely to ‘forgive’ the organization from a service failure as they rely more on their past experiences rather than negative post-failure attitudes (Mittila, 2004). The love turns to hate proposition is supported by theory of justice (Grégoire & Fisher, 2007). Customers with a strong relationship perceive a violation of the fairness. A lower level of fairness is assumed to be related to the outcome and process (Tax, Brown & Chandrashekaran, 1998). A violation of fairness can evoke a greater sense of betrayal which results in greater dissatisfaction or, desirability to retaliate (Gregoire & Fisher, 2007).

When considering the timing of the delay in the context of this study, customers who fly frequently are more likely to have experience with a delay or flight cancellation. They update their reference point and might experience more anxiety for the possibility of a cancelled flight compared to non-frequent flyers. Maister (2005, p. 5) presented eight statements of aspects influencing wait perceptions, two of them are: “anxiety makes waits seem longer” and “uncertain waits are longer than known, finite waits”. Both statements feel rather intuitive and are not represented in a larger theoretical framework however recognized by other researchers.

Hence, when a delay occurs during the pre-process, there is a high probability for an outcome failure (the flight will be cancelled) rather than a process failure. Logically, outcome failures have a greater effect on the customer satisfaction (Smith et al., 1999). From the company we know that flights are more often cancelled during the pre-process phase, when passengers are not boarded yet. This would imply that there is a greater level of uncertainty during the pre-process delay as passengers are not certain whether their flight will depart. When inside the airplane, there is less uncertainty for cancellation. When delays become repetitive to customers, they might have a feeling of satiation, having had enough of the experience (Karmarkar & Karmarkar, 2014). Therefore we propose that pre-process delays have a greater negative effect when passengers travel often.

On the other hand, in-process delays occur not as frequent. There is less repetitive behavior, customers are less exposed to the delay. Passengers who fly frequently however, might have more experience with delays when inside the aircraft. They are more familiar with the delay and might feel a lesser extent of anxiety and/or uncertainty whether the flight departs. A decreasing response to a repeated stimulus is also referred to as habituation (Karmarkar & Karmarkar, 2014). Moreover, passengers who fly more often might have a greater zone of tolerance regarding in-process delays as they do not have to worry about a possible outcome service failure (cancellation, rebooking etc.). The zone of tolerance is identified in the SERVQUAL measure, implying that customers have a range of acceptable service-failure (Parasuraman et al., 1985).

Hence, the following hypotheses are composed:

H3a: The higher the number of trips, the stronger negative relationship between pre-process delay and satisfaction.

3. Methodology

Based on the conceptual model and theoretical findings as described in the previous chapter, we will attribute this chapter to elaborate on the data collection, measurements and plan of analysis. Meanwhile, we will provide clarification on the variables and scales used to measure the hypothesis.

3.1 Data collection

This is a quantitative study, aiming to establish a correlational relation between ‘pre- and in-process delays’ (independent variables) and ‘customer satisfaction’ (dependent variable). This main effect is expected to be moderated by the ‘number of past encounters with the service’ and the ‘travel purpose’ of the passenger. For measuring these constructs data was extracted from operational logs and customer surveys. The operational data is objective data used to measure pre- and in-process delay. The surveys, on the other hand, are subjective to the opinions of the customers. The surveys are designed by the company and standardized in order to compare results over a longer period of time. The survey consists of an elaborate list of questions for customers to evaluate their whole customer journey. Most questions are framed as multiple choice questions with predefined categories or at times as open questions. From this questionnaire, the most relevant variables are selected to measure the constructs relevant for our research.

3.2 Measurements

The scope of this research is limited to flights of the company, departing from the Netherlands in the period of 1 December 2017 up to 30 November 2018. A dataset was developed, containing several variables from the operational- and survey data to measure the constructs – as described below. In the appendix you will find an overview questions asked in the survey (appendix B).

Pre- and in- process delays

The independent variables pre- and in-process delay are measured as the delay taking place during the boarding process, before entering the aircraft (pre-process) and when passengers are seated in the airplane, after entering the aircraft (in-process). Hereby we acknowledge that passengers can wait longer than standard procedures and still depart on time. Hence, for this research we focus on post-scheduled waits (delays) rather than pre-scheduled waits (as defined by Baranishyn et al., 2010). We will compare the delays occurring at two different points in time, before the service and during the service (pre- and in-process).

In order to explain the difference between pre- and in-process delays we will shortly explain the flight departure process. After several service encounters on the airport (check-in, passport check at the border control and luggage scan) passengers arrive at the gate for boarding. The duration of the boarding process is amongst others dependent on the type of airplane, seat occupancy rate, distance to the airplane (usually passengers board directly but sometimes they board per bus) and lastly passengers are occasionally pre-boarded – meaning that, after the boarding passes are scanned, customers wait in a waiting area before entering the plane. Consequently passengers arrive at the airplane after which the passenger doors close (PDC) when all passengers are in the aircraft. The airplane takes off, at the latest 5 minutes after the passenger doors close.

Table 2

Computation of the independent variables

Variable Calculation Condition

Pre-process delay Actual PDC – scheduled PDC If < 0 recoded to 0  Scheduled PDC Scheduled departure time – 5 min.

In-process delay Departure delay – Pre-process delay If < 0 recoded to 0  Departure delay Scheduled departure time – actual departure time If < 0 recoded to 0 Pre-process delay p* (Pre-process delay / flight duration) *100

In-process delay p* (In-process delay / flight duration) * 100

 Flight duration Scheduled arrival date, time – Scheduled departure date, time

*p = process delay as a percentage of the total flight duration

Customer satisfaction

The aim of the research is to see the effect of delays on the overall customer satisfaction which is the dependent variable. The construct is measured and tested by means of the Net Promotor Score (NPS) and overall satisfaction rate. The NPS is introduced by Reichheld (2003) to measure if consumers are likely to customers would recommend the company to others on a 1 to 10 rating scale. Answers to the question whether respondents are ‘likely to recommend the company to friends or colleagues’ result in three different groups. The ‘promotors’ advocate the company and are extremely likely to recommend (9-10 ratings), ‘passively satisfied’ customers or ‘neutrals’ (7-8 ratings) and lastly ‘detractors’ who are very unlikely to recommend (0-6 ratings). The NPS is calculated by subtracting the percentage of detractors from the percentage of promotors (Reichheld, 2003). We will use the 3 scales – promotors, neutrals and detractors – represented respectively with -100, 0 and 100. By using this scaling, the betas of the coefficients will be easier to interpret. Additionally, we take into account overall satisfaction (measured on a 1-10 scale) as a robustness check.

Number of past encounters with the service and travel purpose

Control- and additional variables

We control for several factors which might influence the results of the study. Firstly we include demographic variables such as gender and age. Secondly, we control for transfers, which customers might experience either before the measured flight (pre-transfers) or after the flight (post-transfers). Moreover we include the cabin class, being either economy or business. Lastly, we control for the customer perception regarding delays: perceived delay information. Respondents indicated whether they experienced a delay and if they did, how they perceived their delay information.

Missing values and treatment

As the questionnaire is very extensive, the company decided to randomly distribute an additional set of questions to the respondents. This means that only x% of the surveys includes the question measuring the number of trips (number of past encounters with the service provider). Hence we expect a large amount of missing values for this variable. The missing data of the variable can be categorized as missing completely at random (MCAR). This means that the reason for missing data is completely at random and is not dependent on other variables or observed characteristics of the respondent (Donders, Heijden, Stijnen & Moons, 2006).

It is important that we handle the missing data correctly otherwise the analyses can yield biased results (Hippel, 2004). There are two frequently used methods for handling missing data: listwise- and pairwise deletion. In listwise deletion “all cases with missing values are deleted” whereas for pairwise deletion “each moment is estimated separately using cases with values for the pertinent variables.” (Von Hippel, 2004, p. 161). For data MCAR, both listwise and pairwise deletion techniques are appropriate (Donders et al., 2006). In all our models we will apply listwise deletion as this yields more precise results. In this case, accurate results are preferred over preserving more of the data. Especially considering that the moderator, number of trips, is a critical predictor of our model.

3.3 Plan of analysis

4. Results

4.1 Data preparation

We started with a large dataset from the operational department, including all flights departing from the Netherlands of one year (1 December 2017 up to 30 November 2018). However, this dataset was reduced to 119,887 unique flights (one-way) after data cleaning. We removed duplicates, data of cancelled flights and wrongly administered data points (e.g. flights departing before the doors were closed). Subsequently, the dataset was merged with a subset of the survey-related data containing only variables relevant for this research. The merge was based on unique identifiers: scheduled date of departure and flight number. When analyzing the data we identified outliers and removed extreme outliers such as a departure delay of 1440 minutes, which makes 385 minutes the longest departure delay in our dataset.

4.2 Descriptive statistics

After data cleaning we have a dataset of x respondents. However, not all the respondents answered all the questions and therefore we have some missing data points, especially regarding the number of trips (N=x). However, this was expected, as mentioned previously, the corresponding question was asked to a random selection of 5% (N=x). Overall satisfaction (N=x) age (N=x) and perceived delay information (N=x) have also a somewhat lower response. All other variables have are equal or have a minimal difference from the overall sample size (N=x).

The means and standard deviations of the most important variables for this analyses are displayed in table 3. From the standard deviation we can conclude that there is variance in all predictor variables, a requirement for the regression analysis. From the table we want to briefly highlight a few aspects. “Confidential.” In figure 1 we provide more detailed information of the frequency and percentages of NPS groups, gender and age.

Table 3

Scale and descriptive statistics

Confidential

Figure 1

Confidential

4.3 Testing assumptions

In order to perform a linear regression analysis there are several assumptions that need to be verified. We already checked for outliers and variance in the variables. Now we will check whether the residuals are normally distributed, if we can detect correlation between variables and we will test for multicollinearity.

4.3.1 Normality check

In order to test whether the residuals are normally distributed we created a Q-Q plot. Hereby we examined not only NPS but also the dependent variable overall satisfaction. We plotted the observed value of the standardized residual against the expected normal distribution. The plots can be found in appendix C. From the plot we see that values deviate from the normal distribution. Therefore we conducted the Kolmogorov-Smirnov test. If the assumption of normality is violated this can imply that results are not reliable for interpretation (Razali & Wah, 2011). From the Kolmogorov-Smirnov test we can conclude that the distribution of residuals is significantly different from a normal distribution. Hence, both NPS (p<.001, statistic=.347) and overall satisfaction (p<.001, statistic=.136) are not normally distributed. There are however researchers arguing that non-normality does not necessarily affect the results of the analyses. Leliveld and Wiebenga (2014) claim that the violation of normality is not problematic in the case of a large sample size (N>200). They claim that estimates of confidence intervals and p-values are correct for interpretation based on the central limit theorem. Leeflang et al., (2015) support this by implying that large datasets are more likely to have non-normal distribution errors. When the chosen model deems appropriate, the researcher can be more lenient regarding non-normality. Since, the dataset of this research is of circumstantial size (N=x) we will not take specific measures regarding the distribution of the data if the model deems adequate.

4.3.2 (Multi-)collinearity

Multicollinearity occurs when two or more predictor variables correlate with each other and therefore have a joint effect on the dependent variable. The Variance Inflation Factor (VIF score), indicates the level of multicollinearity. Generally VIF scores should not exceed 4.0. In table 4 an overview can be found of all predictor variables and their corresponding VIF scores. From those scores we can conclude that no multicollinearity is present, as all variables have a VIF score smaller than 4.0. The highest VIF score is 2.135 for the interaction effect of in-process delays and number of trips. This is expected as those variables are not standardized before composing the interaction effect.

Table 4

Overview of VIF scores

Variable VIF Model 1 VIF Model 2a VIF Model 2b Gender 1.014 1.091 1.092 Age category 1.016 1.014 1.015 Preceding transfer 1.012 1.007 1.007 Onward transfer 1.012 1.007 1.009 Cabin class 1.015 1.036 1.037 Perceived delay information 1.027 1.197 1.201 Pre-process delay 1.024 1.138 1.703 In-process delay 1.008 1.113 1.726 Travel purpose 1.182 1.510 Number of trips 1.144 1.272 Trips x pre-process 2.115 Trips x in-process 2.153

Travel purpose x pre-process

1.369

Travel purpose x in-process

1.195

4.4 Analyses

delays. The second model is a hierarchical regression model whereby we analyze the effects of all variables including number of trips and travel purpose in the first step. Consequently, we add the interaction effects an evaluate if the amount of previous encounters with the service provider and travel purpose moderate the effect of pre- and in- process delays on satisfaction. The standardized coefficients are displayed in table 5 for all models. We evaluate the standardized beta because we want to judge the relative importance of the variable while controlling for different scales. Besides, because we have a large dataset we use a significance level of 5% in order to identify significant relations between predictor variables p ≤ .05.

4.4.1 Model 1: main effects

In order to test whether all predictor variables significantly contribute to the model, we conducted a multiple regression analysis. Hereby we included 8 control variables namely: gender, age, preceding transfer, onward transfer, cabin class, perceived delay information. Moreover, we added 2 IV’s to the model: pre-process delay and in-pre-process delay – both as a percentage (P) of the flight duration. The predictors accounted for a significant amount of variance in de dependent variable NPS R2_{= .030, p <.001. Hence, the}

variables explain 3% of the variation in the DV, NPS. This is not very large but understandable as there are many elements influencing the customer satisfaction of the flight experience. All predictors have a significant effect on the NPS. As assumed, both pre-process delays (β= -.118, p <.001) and in-process delays (β= -.055, p <.001) have a negative effect on the customer satisfaction. The results support H1, pre-process delays are found to have a larger negative effect on NPS than in-pre-process delays.

Table 5

Overview of the results

Model 1 Model 2a Model 2b

Variable β β β Gender Confidential Age Preceding transfer Onward transfer Cabin class

Perceived delay information Pre-process delay

Number of trips

Pre-process x travel purpose In-process x travel purpose Pre-process x trips In-process x trips R2 R2 adj N *** Significant at ≤ .001 ** Significant at ≤ .01 * Significant at ≤ .05

4.4.2 Model 2: main- and interaction effects

In order to test whether trip purpose and number of trips moderate the negative main effect of pre- and in-process delays, we conducted a multiple hierarchical regression analysis. In the first step we included the main effects, control variables and the predictor variables: trip purpose and number of trips (model 2a). Subsequently, we added the interaction effects (model 2b). Trip purpose is a binary variable and was therefore not standardized (Dawson, n.d.). The interaction effect is computed by multiplying the moderator with the IV’s: pre-process delay x travel purpose, in process delay x travel purpose. For the second moderation effect both the moderator and IV’s were standardized before composing the interaction effect: pre-process delay x number of trips, in-process delay x number of trips. Predictor variables pre-, in-process delay and number of trips are all standardized in this model.

Model 2a explained a significant amount of variation in the DV, R2_{= .044, p < .001. The addition of the}

interaction effects increases the explained variation of the NPS with 0.2%, R2= .046, p < .001. Even though the predictor variables travel purpose and number of trips improve the model with 1.4%, the interaction effects contribute little to the model. The predictor variables travel purpose (β= -.085, p < .001) and number of trips (β= -.070, p < .001) have a significant negative effect on NPS. These findings suggest that passengers travelling for business purposes report lower a lower NPS compared to passengers travelling for leisure purposes. Moreover, the more frequent passengers fly, the lower the NPS.

travelling for leisure purposes. There was however no significant interaction effect for the interaction between in-process delay and travel purpose (β= .027, p= .088). However, when we would apply a significance level of .10, the interaction effect would have been significant. Even though we cannot draw conclusions from these findings, results seem to indicate that travel purpose has a positive effect on the relation between in-process delay and NPS. This would suggest that passengers travelling for business purposes are more displeased with an in-process delay than passengers travelling for leisure purposes. However, further research needs to be conducted to validate these assumptions.

4.4 Robustness checks

Several robustness checks were performed. Hereby we re-estimate the model based on other variables. First we wanted to verify our results by a re-estimation of the model based on the overall satisfaction score instead of the NPS. The new model predicted 1% more of the variance in the DV as the adjusted R2

improved from R2adj= .045 in the original model, to R2adj= .055 in the newly estimated model with

satisfaction as DV. All main effects and control variables have a significant effect on the satisfaction whereby the effects are in the same direction (positive/negative). H1 is supported, preprocess delays (β= -.152, p < .001) have a stronger negative effect on customer satisfaction compared to in-process delays (β= -.081, p < .001). However, the interaction effects are different from the first analyses. We identified one significant interaction effect which was not valid in our previous analysis: in-process delays x travel purpose (β= .038, p= .021). This implies that the negative effect of in-process delays on the overall satisfaction is strengthened by the travel purpose. More precisely, passengers travelling for business purposes are less dissatisfied regarding in-process delays than passengers travelling for leisure purposes.

Moreover, we tested whether the predictor variables indicate similar effects when we do not control for the duration of the flight. We re-estimated the model were pre- and in-process delay were measured in minutes and not as a percentage of the flight duration. The model fit improved with .7%, R2adj= .052, p < .001.

Except for gender (p= .200), all main effects are significant and in the same direction whereby pre-process delays (β= -.125, p < .001) are perceived as more negative compared to in-process delays (β= -.058, p < .001). Furthermore, only the interaction between inprocess delay and number of trips is significant (β= -.041, p < .001) just as in model 2b. Hence, we can conclude that a different measurement of process delays – whereby we do not account for the flight duration – does not alter the results much. Remarkable is that travel purpose does not have a moderating effect on the relationship between both pre- and in-process delays and NPS whereas this is the case in the original model.

Lastly, we examined whether the arrival delay might influence the respondents. We conducted the same analysis as model 2b but now we included only those participants who experienced an arrival delay of 1 minute or more. The model explained .8% more of the variance of the NPS, R2

adj= .053, p < .001. In the

re-estimated model, in-process delay does not have a significant effect on NPS whereas this was the case in the original model. Possibly, when filling in the survey, customers recall their most recent experience (of arriving too late) more vividly than the experiences which are further in the past.

One last remark, we found that the R2 _{of the model is rather low, the variables explain 4% of the variance}

5. Discussion

In the previous chapter, the results of the analyses were presented. In the current chapter we will discuss those findings, as well as the managerial implications of this research. Finally, we elaborate on the limitations of this research and suggest possible future research directions.

5.1 Overview of results

The main goal of this study was to investigate whether pre- and in- process delays had a differential effect on customer satisfaction. Moreover, we expected those effects to be moderated by individual factors: travel purpose and flying frequency. The research questions were formulated as follows:

To what extent is customer satisfaction affected by the timing of the delay? To what extent is the satisfaction after a delay experience influenced by travel purpose? To what extent is the satisfaction after a delay experience influenced by prior experience?

Following from the first research question was H1, suggesting that pre-process delays have a greater impact on customer satisfaction compared to in-process delays. Results of this study support the hypothesis in the context of flight delays occurring when passengers are at the gate (pre-process) or in the airplane (process). Related to the second research question is H2, stating that business travelers perceive pre- and in-process delays differently compared to passengers traveling for leisure purposes. We found empirical support for a moderating effect of travel purpose on the relation between pre-process delays and satisfaction. However, there was only a weak moderation effect (p < .10) of travel purpose affecting the relationship between in-process delays and satisfaction. H3 is related to the third research question. The hypothesis states that the number of trips moderate the main effect. We found a weak moderation effect for the number of trips affecting the relationship between pre-process delays and satisfaction (p < .10). We did find support for the moderating effect of number of trips in situations of in-process delays affecting satisfaction. An overview of all hypothesis and test results can be found in table 6.

Table 6

Results after testing the hypothesis

Hypotheses Results

H1: Pre-process delays have a stronger negative effect on customer satisfaction compared to in-process delays.

Supported

H2a: Pre-process delays have a weaker negative effect on customer satisfaction for passengers travelling for business purposes compared to passengers travelling for leisure purposes.

Supported

H2b: In-process delays have a stronger negative effect for passengers travelling for business purposes compared to passengers travelling for leisure purposes.

Weak support

H3a: The higher the number of trips, the stronger negative relationship between pre-process delay and satisfaction.

Weak support

H3b: The higher the number of trips, the weaker negative relationship between in-process delay and satisfaction.

Supported

5.1.2 Satisfaction after pre- and in-process delays

Based on the field theory of Kurt Lewin (1943) we expected that the timing of the delay yield different emotional responses. In coherence with the theory we expected that pre-process delays evoke more negative reactions than in-process delays. Literature supported this statement (Benakiva & Lerman, 1985; Davis & Maggard, 1990; Dubé-Rioux et al., 1989, 1991; Hui & Thakor, 1998; Dabholkar & Sheng, 2008; Friman, 2010; Yang et al., 2013). However, there are only a few recent studies and there has no study been found investigating the effect of pre- and in-process delays within the airline industry.

In our study we found empirical support confirming that pre-process delays have a greater negative effect on customer satisfaction compared to in-process delays. In the context of this study this means that passengers prefer a delay when waiting at the gate, rather than a delay in the aircraft. This is in line with the field theory suggesting that the perceived waiting time is perceived as longer and more negative when a delay occurs further from the goal achievement rather than a delay happening close to the goal achievement (Yang et al., 2013). Hence, customers prefer to wait due to a service delay in the airplane, when they are closer to their goal, rather than before they enter the airplane, which is further from the goal.

5.1.3 The moderating role of travel purpose

theory we expect that barriers, a delay, occurring closer towards the goal (in-process) yields more negative reactions in high need conditions. From the means-end theory (Gutman, 1982) we expected that business travelers feel a greater urgency to depart on time (e.g. arriving on time on their business meeting). Therefore we expected that in-process delays had a greater negative effect on business travelers opposed to pre-process delays.

Findings from our study support this expectation. There was a significant interaction effect, between pre-process delay and travel purpose. This finding confirms that pre-pre-process delays have a weaker effect on satisfaction for business travelers as opposed to leisure travelers. We did not find a significant relation for the moderation effect of travel purpose on the negative relationship between in-process delays and satisfaction. Yet, results seem to indicate (p < .10) that business travelers perceive a in-process delay as more annoying than leisure passengers. However, it is very likely that business travelers experience a greater urgency to depart on time compared to leisure travelers. We tested this by re-estimating the model and included only those cases subject to an arrival delay of 1 minute or more. This resulted again in insignificant outcomes for the interaction between in-process delay and travel purpose. However, when re-estimating the model with overall satisfaction as dependent variable instead of NPS, we did find a significant moderating effect. The finding supports our hypothesis, suggesting that in-process delays have a greater negative effect on the satisfaction of business passengers compared to leisure passengers.

Hence, there is a weak support suggesting that passengers travelling for business purposes evaluate the flight experience more negatively when a delay occurred in-process rather than pre-process. We found little empirical research related to this finding. However, Schmitt, Dubé and Leclerc (1992) found that, when waiting in line, customers are more likely to raise objections when someone intrudes when almost at the cashier than at the beginning of the line. This suggests that barriers are perceived more negatively when closer to the goal.

5.1.4 The moderating role of number of past service encounters

Results partly confirm our hypotheses. There was no significant interaction effect between pre-process delay and the number of trips. However, in case of a significance level of .10, there would be a moderating effect, indicating that pre-process delays have a greater effect on the satisfaction of customers who fly more often. The weak correlation might be explained through the principle of habituation, frequent flyers might be more used to pre-process delays. We suggest to conduct further research to verify those results. Moreover, we found a moderating affect for the number of trips on the relationship between in-process delays and satisfaction, confirm our hypothesis. The effect was the significant in three of the four tested models (a model where pre- and in-process delay were measured as a percentage of the flight duration; pre- and in-process delays measured in minutes and lastly, a model including cases with arrival delay).

5.2 Managerial implications

Research in the field of flight delays has mostly been focused on optimizing operations, mainly regarding timetables (aircraft-, crew-, maintenance rosters) and revenue management (maximizing revenue while minimizing costs) (Clausen, Larsen, Larsen, Rezanova, 2010; Kohl, Larsen, Larsen, Ross, Tiourine, 2004; Løve, Sørensen, 2001). However, due to the unpredictable environment of airlines, delays are difficult to account for, think for example of crew illness or bad weather conditions (Løve & Sørensen, 2001). Therefore it might be easier for airlines to manage customer perceptions rather than controlling unforeseeable events in the operations. Hence: “if you cannot control the actual wait duration, then control the customer’s perception of it” (Taylor, 1994, p. 56).

By means of this study we hope to provide managers with new insights and improve their understanding of individual factors influencing customer satisfaction after a delay. In our research we distinguish between pre- and in-process delays, insights can therefore be applied more specifically. First of all we found that pre-process delays have a stronger negative effect on the customer satisfaction compared to in-process delays. When designing recovery strategies, we advise the management to focus on pre-process delay in particular. Moreover, we found that the delay perception differs depending on individual factors travel purpose and number of past encounters with the service. We propose that managers distinguish their approach for managing customer delay perceptions depending on customer characteristics.

Passengers who fly frequently are more susceptible to pre-process delays than in-process delays although more research should be conducted to confirm this. Managers could consider to investigate possibilities of is the provision of additional credits for the frequent flyer loyalty program for long during pre-process delays. For, most frequent flyers are member of the loyalty program. Providing certain advantages could possibly enhance the customer satisfaction of frequent flyers during or after a pre-process delay.

5.3 Limitations and further research directions

This research has some limitations. In the model we designed and tested, we included a selection of variables we deemed relevant. Nevertheless, we do recognize that there are many other variables which influence the customer satisfaction after a flight. This is confirmed by the weak R2_{, explaining only a small}

amount of the variance of the NPS. Moreover, we recognize that the interaction effects are not all significant and yield different results for the dependent variables NPS and overall satisfaction. It is recommended to conduct more research to strengthen the validity of the effects pre- and in-process delay each interacting with travel purpose and flying frequency. The measurement of number of trips is not very precise. Customers are asked how many times they flew in the last twelve months. Especially those who fly often, are not likely to remember exactly how many times they flew before. Moreover, we do not measure whether the respondents had a positive or negative experience during their previous flight.