2
Evaluating methods for measuring passenger satisfaction in the
airline industry: email or e-device?
Author: Erik Christiaan Berens
Address: Schaerweijdelaan 247-A
3702 GH Zeist
Phone number: +31 (0)6 51 933 529
E-mail: info@eberens.nl
Student number: 1830252
Department: University of Groningen
Faculty of Economics & Business Department of Marketing Landleven 5
9700 AV Groningen
Specialization: Marketing Management Qualification: Master thesis
Date: July 2011
Supervisors: Dr. J.T. Bouma
Dr. ir. M.C. Achterkamp
External supervisor: H. Zijlstra
Organization: KLM Royal Dutch Airlines BV
Address: Amsterdamseweg 55
1182 GP Amstelveen
3
Management Summary
As competition and pressure on revenues increases the quest to find new unique selling points for airliners has become fiercer than ever. Currently the airline industry is going through some rough transition times as more low cost carriers and competitors from the Middle East and Asia try to increase their market share. This reshaping of the airline landscape stresses the importance of robust customer satisfaction data.
Air France KLM is one of the top five players in the world wide airline industry transporting with over 1800 flights each day more 70 million passengers on an annual basis. The dynamic environment they operate in demands new customer insights to ensure a steady passenger growth. The last five years Air France KLM has used an on board pen and paper questionnaire (SCORE) to retrieve valuable passenger feedback to evaluate and/or launch new services. This methodology has proved its robustness and representativity but also knows its drawbacks as it is inflexible, has a long lead time and is considered old-fashioned by some.
The search within the customer knowledge team for a new passenger satisfaction tool has begun. Goal is to find a tool which is more flexible, less costly, up to date and above all should give instant access into the gathered data. A short feasibility study showed two possible options: an onboard eDevice or email invitations with a link to a web questionnaire after the flight. This study was conducted from a research point of view, this to find out which tool can provide the bested ata quality and representativity.
The eDevice encountered two major drawbacks; first the data quality retrieved was below average making the data less reliable than current SCORE. Secondly the volumes retrieved with the eDevice are below the 1% coverage goal of Air France KLM, the combination of both reasons crossed out the eDevice as a serious substitute for the pen & paper method. If the length of the questionnaire was to be redesigned and the user friendliness of the eDevice to be optimized, still the volumes would remain questionable on flights shorter than an hour. Combining this knowledge with high start-up and running costs, not even mentioning responsibility within the organization makes the eDevice an unsuitable substitute.
4 created some biases in the sample as the ratio FB (Flying Blue) versus non FB members gets disturbed; these and other biases can be corrected using an advanced weighting process.
A mind set change within Air France KLM comes along with this method since satisfaction scores decrease by an average of one point on each item. This is again caused by the self-selection who attracts the more motivated and on average critical passengers, but can be, if needed be compensated for using additional statistics. Having a more critical sample is of great value since this guarantees a dedicated sample. Apart from this timing should be well understood as this research has proven that after 7 days the deviation from the mean starts to significantly change.
For marketing managers it is important to understand that this decrease in satisfaction is not due to a decrease in service quality, but only due to an increase in serious, sincere respondents. This combined with the costs savings, increased flexibility for implementing ad-hoc studies and shorter lead times makes the email eSCORE, with a correct weighting process a representative method to replace the current pen and paper questionnaire.
5
Preface
After seven years of studying at the Arnhem Business School, University of Wisconsin Stout and the University of Groningen this thesis is the end of my student life and at the same time the start of my professional life.
During the last eight months at Air France KLM I have learned a great deal, not just theoretically but also personal. Therefore I would like to thank the entire Customer Knowledge/Research team both at Air France and KLM for giving me this opportunity. This dynamic, enthusiastic, supporting and friendly team really made working for Air France KLM joyful.
Furthermore I would like to say special thanks to my project team; Hans Zijlstra, Raphael van Saidof and Ronald Rood who gave me the space and opportunity to realize this project. Their knowledge and energy I encountered during this project was not just of vital importance to the project, but has also motivated me.
Special thanks also for the support, guidance and helpful contribution of my supervisor from the University of Groningen: dr. Jelle T. Bouma.
At last I would like to thank Susan van Nierop for her help, encouragement and critical reviews.
Thank you all,
6
Table of content
Chapter 1. Introduction ... 10
1. Introduction ... 10
1.1. Background ... 10
1.1.1. The Airline industry ... 10
1.1.2. Passenger satisfaction ... 11 1.2. Problem definition ... 12 1.2.1. Current disadvantages ... 13 1.2.2. Current advantages ... 13 1.2.3. Goal... 14 1.2.4. Infeasible options ... 14 1.3. Research Question ... 15 1.3.1. Conceptual model ... 16 1.3.2. Research questions ... 17 1.4. Disposition of thesis ... 17
Chapter 2. Literature Review ... 18
2.1. Introduction ... 18
2.2. Online versus traditional offline methods ... 18
2.3. Representativity ... 19
2.3.1. Volumes, coverage and response rates ... 20
7
3.3. Research design ... 33
3.3.1. Sampling size ... 33
3.3.2. Sample & population characteristics ... 34
3.3.3. Data collection ... 35
3.3.4. Measurement design and scaling ... 35
3.4. Plan of analysis ... 35
3.4.1. Statistical tests ... 35
Chapter 4. Results ... 37
4.1. Representativity ... 37
4.1.1. Volumes, response rates and coverage ... 37
4.1.2. Loyalty ... 38 4.2. Segmentation ... 41 4.2.1. Demographics ... 41 4.2.2. Behavior ... 42 4.3. Data quality ... 44 4.3.1. Variance ... 45 4.3.2. Corporate Image ... 47 4.3.3. Response time ... 47 4.3.4. Item completeness ... 48
4.4. Satisfaction and dissatisfaction ... 50
Chapter 5. Conclusion ... 53
5.1. Hypothetical conclusion ... 53
5.1.1. Volumes & response rates ... 53
5.1.2. Data Quality ... 55
5.1.1. Segmentation ... 56
5.1.4. Satisfaction ... 57
5.1.5. Overall conclusion ... 57
8
Chapter 6. Recommendations ... 60
6.1. General recommendations ... 60
6.2. Response rates & volumes ... 60
6.3. Data Quality ... 62
6.4. Segmentation ... 63
6.5. Weighting process ... 63
Chapter 7. Limitations and further research ... 64
References ... 65
Appendices ... 69
Appendix I eSCORE options and weights ... 69
Appendix II Questionnaire (KL ICA)... 70
Appendix III Flying Blue Tier Levels... 75
Appendix IV Loyalty levels explained ... 76
Appendix V Data Output ... 77
Appendix VI Screenshot eDevice - Ipad ... 92
9
Abbreviations
AFKL Air France KLM
LCC Low Cost Carrier
DL Delta Airlines
FFP Frequent Flyer Program FB Flying Blue (FFP of AFKL)
SkyTeam Global airline alliance partnering thirteen members KPI Key Performance Indicators
SCORE Current passenger satisfaction monitor of AFKL, based on pen&paper questionnaire eSCORE Electronic SCORE
IFE In-Flight Entertainment system E-device Electronic Device (e.g. Ipad)
PNR Passenger Name Record
Short Haul National flights of AF (within France)
Medium Haul AF and KL flights on the main land of Europe and Northern Africa Long Haul Intercontinental AF and KL flights
10
Chapter 1. Introduction
1.
Introduction
In this chapter an introduction of the thesis at Air France KLM (AFKL) is provided. This chapter elaborates on the background of the research area and also showing the studies importance for AFKL. Further along the chapter a problem statement, research purpose and research questions are discussed. At the end a clear overview of the thesis is presented.
1.1. Background
The thesis focuses on passenger satisfaction research in aviation. For AFKL passenger satisfaction research knows four main purposes:
Deliver segmenting and profiling information; Used for target setting and evaluation;
Evaluating, adjusting and implementing new services; Discovering trends.
Currently this data is gathered using a pen and paper questionnaire. The goal of this research is to analyze the representativity of alternative data retrieval methods as the current pen and paper process has a long lead time, equals high costs, is inflexible and is considered old-fashion. The role and importance of passenger satisfaction data for aviation, in particular AFKL is discussed further along the chapter.
1.1.1. The Airline industry
In the beginning of 2004 Air France and KLM merged into: Air France KLM. This merger enabled both airliners to reduce risks and expand their reach around the world. Currently Air France KLM jointly operates over 1800 flights each day, transporting over 70 million passengers (PAX) each year making it one of the top 5 players in the worldwide airline industry.
11 For AFKL this reshaping landscape is not passing by unnoticed. Over 60% of all the flights operated by AFKL are on the continent of Europe and face direct competition from a numerous amount of LCC’s like Easy Jet, Vueling and Ryanair. Next to LCC’s AFKL also faces major continental and intercontinental competition from the Lufthansa and British Airways group. One of the strategies applied in the past years by the big three in Europe was extending their range and product portfolio by mergers and take-over’s. AFKL currently has a number of subsidiaries flying under the umbrella of AFKL who all operate as LCC’s. One of the fully owned subsidiaries of KLM is Transavia where Air France owns BritAir and City Jet.
But there are still opportunities for growth. These are not to be found on the matured European and North American market but can be found in the Far East and the South of America. As of 2007 AFKL has teamed up with Delta Airlines, with over 161 million passengers each year is the biggest airliner in the world. This corporation means that all profits and losses made on transatlantic flights between the US and Europe are shared between the two airlines, all in order to reduce risk. Such joint ventures are the future of the airline industry, building a strong reputation and fleet on the other side of the world (Greenfield) simply takes too long and is too costly, so joint ventures are the ideal way of increasing revenues and decreasing competition.
1.1.2. Passenger satisfaction
Satisfaction in literature is referred to as: good experiences minus the bad experiences of a customer (Meyer and Schwager, 2007). These experiences are influenced by the 7 P’s of service marketing (product, price, place, promotion, people, process and physical environment), leading to satisfaction. Customer satisfaction plays a big role in a service organization as it is an antecedent of brand loyalty, customer retention, market growth and finally overall profitability (Anderson and Mittal, 2000).
12 member are transferred from the unknown passenger into a reachable customer, this enables AFKL to obtain customer insights. Still the biggest part of all passengers are not enrolled, therefore in this thesis there will be referred to passenger satisfaction, not customer satisfaction. Please bear in mind when reading this thesis that a customer can be a passenger more than once.
To goal of passenger satisfaction research is to gain insights in the variables that influence the satisfaction. By enabling yourself to retrieve this data, services can be improved for all segments. Ostrowski et al (1993), Heskett et al (1994) and others have shown a significant relation between service quality and the level of customer loyalty. They show that satisfying customer needs will increase customer loyalty, increase retention rates (lowering the chance of changing carrier), reducing the amount of competition and therefore financial risk. Building customer loyalty through good and adequate service is the only way that an airliner can obtain a permanent competitive advantage ensuring survival and growth. This relation is considered to be a key factor to success (Andreassen, 1997).
“By listening to customers, service providers can identify critical areas of service that need managerial action and those that can be promoted on differentiating features”
(Headley and Choi 1992).
As stated, behavioral data is needed to enable your company to serve passenger needs. Currently AFKL uses a pen and paper questionnaire, called SCORE. These are distributed on board by the crew according to a sampling plan based on employee number in combination with aircraft type. On each flight 1 up to 5 questionnaires are filled out anonymously generating over 550.000 completed questionnaires each year; goal for representativity equals 1% of the annual PAX but is not always met. On a monthly basis Key Performance Indicators (KPI’s) are compared to the results of SCORE. The use of SCORE is not limited to target setting as the results are also used for segmenting. To underline the importance of SCORE: SCORE is the only tool to find out who and why passengers are on the specific AFKL flights and rate services differently.
1.2. Problem definition
13 Therefore the business is searching for a new tool to measure in-flight experiences. In this thesis two options will be tested. These options are: an in-flight digital questionnaire on an eDevice (Ipad 1) and an after flight email invitation referring to a web questionnaire. This new tool should solve current disadvantages and should remain a reliable and feasible passenger satisfaction measurement tool.
1.2.1. Current disadvantages
SCORE currently has four main disadvantages: Long lead time
Before the data retrieved from the planes has been collected and scanned it takes up to 1,5 month before the data can be analyzed.
High costs
Printing, packing, distributing, collecting and scanning of questionnaires. Equaling high annual costs.
Inflexible
Changing the questionnaire equals high printing costs and a long lead time before changes can be implemented.
Considered old-fashioned by some
Paper on board, considering all possibilities these days is considered out dated by parts of the upper management.
These disadvantages are the advantages of a digital version (eSCORE). As Evans and Mathur (2005) state that the biggest strength of online surveys are their global reach, flexibility, speed, easy data analysis, low costs, ease of follow-up, and forced answering. All advantages AFKL is seeking for.
1.2.2. Current advantages
The biggest advantage of SCORE that should not be lost when implementing electronic eSCORE are: Reliable, safe method (representative and robust, also in the future)
High volumes (volumes should remain high to ensure representative analysis)
14
1.2.3. Goal
As elaborated on above the current SCORE method has its disadvantages. Therefore the main goal is to implement a new tool to measure passenger satisfaction and behavior that is contemporary, flexible, cost efficient and fast. This new tool may not harm representativity, volume and the robustness of the sample. Apart from email and the eDevice other satisfaction measurement tools were considered during my internship but where found infeasible to implement within AFKL. These are discussed in the next paragraph.
1.2.4. Infeasible options
In this thesis the focus is only on the options that can be actually implemented within AFKL. These two options were derived from the following initial list of options:
Email invitation E-device on board
In-Flight entertainment system (IFE) Card invitation
Touch points
Connectivity on board
The last four substitutes for SCORE are not feasible for implementation within AFKL. IFE was rejected due to limited availability in aircrafts, technical issues and high costs. Apart from this, IFE systems are only on intercontinental planes forcing AFKL into a mixed methodology, an option which is not preferred.
Card invitation is about distributing cards on board by the crew. These cards contain a unique link to the web questionnaire. AFKL is a member of the SkyTeam alliance, current they are using this methodology onboard for research purposes. This SkyTeam research showed that response rates are very low (±1,5%). This has various reasons; it is to be believed that by improving this methodology (better branding, clear incentives etc) it still remains impossible to increase response rates by over 400% as for a representative satisfaction research response rates of at least 6% are needed. This to avoid high workload for the crew and 100% self selection by the respondents.
15 out which segments are flying this is a no go. A solution would be an integrated ticket scanner, but this is costly and will not solve another problem which is that PAX are not able to enter greater amounts of data due to great time constraints and rush involved at most of the touch points.
The final option, connectivity on board implicates inviting and letting PAX at the same time on their own computer, smartphone or e-device complete the questionnaire onboard. This option requires WIFI outfitted planes, a technology not going to be implemented within the next 5 years. Also the self selection can cause a potential problem as some segments might become overrepresented due to the fact they are more e-savvy or travel for business purposes (usually carry a laptop).
All these options are schematically reviewed, and are weighted on several criteria according to the MoSCoW prioritization1, these can be found in appendix I.
The MoSCoW model shows that only email and eDevice appeared to be feasible. Both are characterized by a different methodology, the biggest difference is online versus offline and on board versus off board; these are big differences that might bias results. This bias maybe caused by: segmentation, device, way of inviting or differences in timing must be known on beforehand to limit the trend breach.
The following part of this thesis only focuses on the email and eDevice with their possible biases and solutions for overcoming and implementing them successful.
1.3. Research Question
For managers within AFKL it is of high importance that they receive accurate non-biased passenger satisfaction results and segmentation data as large company decision are based upon this data. Therefore the new passenger satisfaction tool should be well representative for the population and a possible trend breach should be well explainable.
Representative in this research is: at least a 1% correct representation of segments and flights and a correct review of satisfaction.
The research question:
In which way can the current pen and paper method be replaced by either email or eDevice surveys without harming representativity?
16
1.3.1. Conceptual model
The eDevice survey from a methodological perspective remains the same as current SCORE. The email invitation linking to a web questionnaire is a different approach. The biggest gap can be found when looking at the moment the questionnaire is received and completed. For the eDevice questionnaire, which is loaded on an Apple Ipad, the contact moment will be while the passenger is still experiencing the service on the plane, same as SCORE. The email version will be an after flight survey, meaning that there are possibly more variables that will influence representativity like timing and corporate image. The second biggest difference between eSCORE and the current SCORE is the shift towards a digital environment. The variables influencing representativity are schematically shown in figure 1.1 conceptual model.
17
1.3.2. Research questions
Additional questions need answering before the research question can be answered. These questions are addressed in this paragraph.
Can email increase the number volumes?
H
1 emailDo loyal customers have a higher response rate?
H
1a emailDoes the eDevice harm volumes in a negative way for short haul flights?
H
2 eDevice Can email alter the response composition negatively influencing the overallsatisfaction?
H
3 emailDoes email target different behavioral groups?
H
3a emailDoes email target passengers with a different profile?
H
3b emailWill a larger proportion of younger Dutch passengers respond?
H
4 eDeviceWill overall satisfaction decrease?
H
5 emailWill in a regression in-flight elements become of less importance?
H
5a email Will more dissatisfied passengers respond to the email invitation?H
5b email Can the eDevice cause in-flight satisfaction to increase?H
6 eDevice Will in a regression in-flight elements become of greater importance?H
6a eDevice Will overall variance for email increase and for the eDevice remain the same?H
7 email/eDeviceDoes corporate image decrease the answering variance?
H
7a emailWill an increase in time lower the answering variance?
H
7b emailCan eSCORE negatively influence data quality?
H
8 email/eDeviceCan eSCORE have a negative influence on item completeness?
H
8a email/eDeviceFurther explanation regarding the research questions and their importance as well as the literature review can be found in chapter 2.
1.4. Disposition of thesis
This report consist out of six chapters, it begins with sketching the environment AFKL is maneuvering in, apart from industry and market information the goal and set-up of this report are given. In the next chapter relevant literature is discussed using the hypothesizes above as guidance. This conceptual model is tested to get critical insights for the problem definition. In order to do so literature is combined with quantitative research to estimate the impact of each hypothesis. After this methodology and the plan of analysis are given, these will lead to the results accepting or answering the hypothesizes. Together this will lead to an overall conclusion, discussion and further research.
18
Chapter 2. Literature Review
2.1. Introduction
Searching for a new tool to measure passenger satisfaction can be seen as a new search for the truth, a truth which is seeked for during each research but which always remains research specific. What the truth is, is that airliners are not able to convert their entire passenger base into customers, over two thirds of their passengers will remain anonounimous. The size of this group explains also the importance and power of this group. Not just in terms of revenue but also in fields of (electronic) word-of-mouth as (electronic) word-of-mouth has a significant impact on customer choice. In some situations it is even seen as more effective than traditional marketing (Gruen et al, 2005). Therefore the needs, wants and evaluations of this segment and all other segments on board are of vital importance. This chapter discusses the methodological differences between online and offline, and the variables influencing representativity.
2.2. Online versus traditional offline methods
The last five years online research has become more and more attractive for industry leaders like Procter & Gamble, Unilever, and General Mills who all have increased their share of online research. Some US based companies like General Mills (one of the largest food companies) already uses online channels for 80% of their research in 2002! This is not surprising if the advantages of online research are summarized, online offers more design options, easier to retrieve responses from all over the world in a shorter time (Ilieva et al. 2002), and of course the lower cost saving aspect (Roster et al. 2004). So, it can be safely stated that online research is global, useable for B2B and B2C, flexible, fast, cheaper, ease of data entry, larger volumes, and more design and programming options.
But it also has its weakness, the biggest weakness is to be found in its representativeness, basically caused by the possibility of self-selection and availability of email addresses. The last factor might directly create a bias towards a maybe more e-savvy customer group (Duffy et al. 2005). Cole (2005) found as an addition that the web respondents were also significantly younger than pen & paper respondents.
19 changing quantitative methodologies. Therefore this study addresses two alternatives for pen & paper questionnaires in the travel industry. The goal is to gain more insights about the impact of the two different methodologies on segmentation, data quality and satisfaction and therefore representativity.
Method When How
Pen & Paper on board During experience Non Digital off line Email link to web questionnaire After the experience Digital online
eDevice on board During experience Digital off line
Table 2.1 shows the three methodologies that are compared in this research. The basic links are to be found between digital versus non digital and during experience versus after experience. Both are necessary to measure the impact of a different tool.
2.3. Representativity
The most import item in this research is representativity. The conceptual model and its hypothesis focus on four main topics who determine this representativity and usability of the new method. These topics are:
2.3.1. Volumes, coverage and response rates 2.3.2. Segmentation
2.3.3. Satisfaction 2.3.4. Data Quality
2.3.1. “Volumes, coverage and response rates” has to ensure that the volume of the sample population is large enough to do correct analysis. Statistical calculations, like indicated in Malhotra (2007) are not used by AFKL to determine the minimal sample size, but are determined by the magnitude of the analysis. The most detailed monthly analysis is done on flight and airport level, reporting requires an ‘n value’ of at least 50 for KL and 100 for AF. Although the merger, some business rules are not yet fully harmonized between the carriers, discussions regarding these inconsistencies are ongoing.
2.3.2. focuses on the respondents, answering questions like: who are they, where are they from, which flights where they on and why? Currently AFKL already segments its passengers into different homogeneous subgroups who all tend to show the same behavior, needs, wants and rate the service
20 in the same way. The satisfaction of each different segment is based on various segmentation variables defined by AFKL who tend to have major impact. The variables are:
1. Tier level / non FB: influencing loyalty and involvement
2. Experience: average nr. of flights per year: use a different framework to relate services to 3. Nationality: have different mean values for satisfaction (Bolle and Kemp, 2008)
4. Booking channel: show the relation to the airliner
5. Travel purpose: shows the difference between business and leisure passengers
The aim is to find out if satisfaction differs significantly within these homogenous subgroups caused by the different methodologies: digital versus non digital and experience versus after experience measurement.
2.3.3. satisfaction, the final and last item explains the contribution of different sub satisfaction variables to the overall satisfaction. Currently AFKL uses a regression model to explain the impact of different sub satisfaction variables to the overall satisfaction. An example is the impact of on ground satisfaction to overall satisfaction. The other two major sub satisfaction variables are punctuality and overall satisfaction on board. A regression will show the contribution of sub satisfaction questions to the overall satisfaction question for each different methodology.
2.3.4. data quality focuses on the actual quality of the data retrieved by the tool. A good representation of the population and flights is great, but when the data quality is lacking it can change the overall conclusion of the methodology.
Shortly summarized:
“What is the impact of the new methodologies on volumes, satisfaction and segmentation? Again, all leading to representativity.”
2.3.1. Volumes, coverage and response rates
21 Krosnick, 1999 and Hox and de Leeuw, 1994 explain the importance of a high response rates as follows:
“Research results can be biased if the non-response is nonrandom, and if it is in some way correlated with the variables measured in the survey. Since the process leading to non-response is usually unknown, it is often optimistically assumed that when response is high, there is no serious non response bias. Thus, a high response rate is viewed not only as desirable, but also as an important criterion by which the quality of the survey is judged.”
Past research has focused on response rates of online questionnaires and a meta analysis on response rates articles shows that there are different factors influencing response rates. Harzing, 2000 finds that the perceived distance from sender has an impact, the further away the lower the response rates. This supports the research of Cox et. al. (1974) who claim that higher involvement with the product or service increases response rate. Also numerous other studies have found this statement to be truth (van Kenhoven, 2002). This also strongly correlates with the factor “interest in survey topic” that positively stimulates response rates (Harzing 2000).
Other variables strongly influencing response rates are: personalization, although Houston and Jefferson (1975) argue that personalization of questionnaires to target groups who feel easily threatened may back-fire. More recent research supports the statement that personalization increases response rates (Roy and Berger, 2004).
Incentives are also a widely discussed topic in online research and response rates. Older research has shown that incentives have a significant positive effect as this extrinsically motivates respondents to respond (Church, 1993, Collins et al. 2000). More recent research has shown (economic) incentives might also back-fire as it suggests that a lot of effort is needed for the respondents to fill out the questionnaire (Bowles, 2009). So, incentives might work for an airliner but are left out of the scope of this research due to the inability of implementing this extra hypothesis due country specific regulations.
22 service experience but did also receive a reminder email. An option not included in this study because it conflicts with AFKL’s limited communication protocols.
Taking these response rates as a guide line and combining them with the number of email addresses available response volumes can be increased by up to 400%, receiving over 2 million questionnaires per year.
But volumes and response rates are not everything, the representativity of the group is of bigger importance (Malhotra, 2007). Therefore the next paragraphs will discuss this issue.
H1 An increase in volumes for the email eSCORE is expected H1A It is expected that loyal customers have a higher response rate
2.3.2. Segmentation
“What you give is what you get”
(Senior Customer Insights Manager, Air France KLM)
Segmentation variables are the most important and strongest variables influencing the representativity of a sample and/or method.
Within AFKL context we can distinguish two kinds of segmentation: profile information and behavior information. Both have a significant impact on all satisfaction questions. Having a bias in the segmentation will unanimously lead to different satisfaction scores biasing all other results. This is where the first and major problem for AFKL arises. Currently AFKL uses need-based segmentation as is advised by literature (Garver, 2009). Some companies still segment on demographics, this is a big risk when segmenting customers in homogenous subgroups based on profile instead of behavior.
AFKL is allowed and able to contact ±55% of its passengers by email. This is promising in terms of volumes but from a statistical point of view this might also become the biggest bottleneck of the email methodology if the 55% is not randomly selected..
23 Figure 2.1 indicates the availability of email addresses in the four booking subgroups. These subgroups are organized by: Flying Blue members and direct/indirect bookers. Indirect are the bookers who book via a travel agency, direct bookers book directly at Air France or KLM. The blue field indicates the volume of email addresses available for each group. The market shares of these groups are given in percentages in the figure.
Clear is that for the email methodology the indirect non FB group will have a less representative sample due to limited availability of email addresses. Knowing that it is impossible to increase the number of available email addresses the results will be biased towards the more direct bookers and FB members. Direct bookers and loyalty program members in general tend to be more loyal and therefore generate higher response rates and better item completeness (van Kenhove et al., 2002; Powers et al., 2009).
Important to answer is if this segment behaves significantly different from the others, and if there are variables in which they show strong correlation with other types of passengers groups making it a less homogenous group, and therefore making their underrepresentation less important.
Because it is known that the email eSCORE does not use a totally random sample on every segmentation variable a bias in segmentation is expected, especially when combining this with known literature:
Article Findings
Cole (2005) & Kaplowitz et al. (2004) web respondents are younger*
Duffy et al. (2005) web respondents are higher educated
Hoek et al. (2002) & Schillewaert (2005) differences in sociodemographics of respondents
Figure 2.1, booking channels and FB membership
24 *) Cole and Kaplowitz et al. both conducted their studies over six years ago, therefore Cole stated in his article that this effect will be minimized over time.
Booking channel, loyalty and profile characteristics can have its impact on, but the actual behavior of the passengers can also have a significant impact on the segmentation. Previous AFKL research has indicated that travel purpose is a variable that strongly influences satisfaction. The following problem occurs for this group: the Passenger Name Record (PNR) is a code uniquely created for each booking, but a booking can contain multiple passengers under the same PNR number. But the booking system used by airliners only allows passengers to enter one email address resulting in one email per PNR. It is known that business purpose travelers have a lower number of passengers per booking/PNR than leisure travelers. This creates a bias in the number of reachable leisure and business purpose travelers as relative more business travelers can be reached. Therefore it is expected that there will be an increase in business travelers. Also Harzing (2000) suggests that the further away from home the invitation comes the lower the likelihood to respond, therefore also in terms of nationality changes in response composition are expected.
The following hypotheses are expected for the segmentation of the email eSCORE method:
H3 Difference in response composition, leading to different segmentation will have a negative impact on satisfaction
Sub hypothesis that will lead to a different sample composition: H3A Differences in respondents behavior are expected H3B Differences in profile composition are expected
eDevice
The eDevice methodology correspondents with the current pen & paper methodology. There is no need to collect email addresses as pursers just have to follow the sampling plan created for them. This sampling plan indicates a few seat numbers that should receive the questionnaire.
The current pen & paper method has some on board logistical problems, the biggest one is to be found in time constraints on inner circle short haul flights (<60min flight time). On these flights cabin crew does not always have enough time to distribute these questionnaires due to other more important activities on board like cleaning up, preparing for landing etc.
25 it can be assumed that the volumes received on short haul flights will be lower than with the current pen & paper questionnaire. For the crew to cope with this and still reach high volumes they will start distributing the eDevice to passengers that are French, Dutch or English speaking as this make communication easier and requires less explanation of the device. Gibson & Gibbs (2006) have found that if nationality and/or country of birth are the same communication will take less effort. Therefore it is expected that Dutch (for KL) and French (for AF) passengers will have an increased share in the sample group.
Apart from nationality, age can also be biased by the crew. Common understanding for the crew is that younger people can better understand an Apple Ipad than the more elderly on board. Again, this makes it easier for the crew to distribute. This statement is supported by Tse et al., 2008 who found that older people (65+) have more problems with first use of new electronic equipment.
So because segmentation is in hands of the crew who sometimes are under great time pressure a drop in volumes and/or a bias towards the younger Dutch passengers is expected.
H4 Increased amount of younger Dutch respondents is expected H2 Decrease in volumes on short flights is expected
2.3.3. Satisfaction
“Different methodologies lead to different outcomes, the only thing that counts is your ability of explaining the differences.”
(Senior manager KLM)
Satisfaction results delivered by SCORE each month are of great value within AFKL, not just for comparison and benchmarking but also for (personal) target setting. Therefore changes in satisfaction should be well understood. This paragraph focuses first on the email, after this on the eDevice methodology.
26 Problem and emotion focused refer to complaining behavior of dissatisfied customers, where loyalty as an attitude is likely to cause an increase in loyal customer due to limited self selection. Dick and Basu (1994) state that there are four loyalty types:
1) Truly loyal customers: these are willing to seek a particular service from a particular provider 2) Spuriously loyal customers: motivated by impulse, convenience, and habit; all if the
conditions are right
3) Latent loyal customers: customer who are loyal because they simply have no other choice 4) Non loyal customers: not loyal to a brand or service.
Loyal customers in general are tended to give a more positive feedback (Zins, 2001) but this works differently in a service context, especially in the airline industry. Monthly AFKL research shows that also here the model of Dick and Basu (1994) is applicable. Spuriously loyal customers are frequent flyers, a group who travels above average making them more critical regarding the delivered service; this is for AFKL the biggest group of ‘loyal’ customers. The truly loyal customers are smaller and consist out of non FB and FB members. Knowing that the spuriously loyal group is the biggest and are the easiest group to reach (due to FB profile) a decreasing satisfaction is expected. Apart from attitudinal loyalty, behavior loyalty is left out because no long term behavioral data on an individual or group basis is available. Therefore defining loyalty in this research is not the combination of behavioral and attitudinal loyalty, although having both gives a better estimate (Reinartz & Kumar, 2002) only attitudinal is used.
27 Applying the theory of Heskett et al (1994) who argues that customer satisfaction leads to customer loyalty what finally leads to profitability is also applicable to new customers in the airline industry.
The step indicated with the red box in Heskett’s theory (figure 2.2) is less applicable on experienced customers who are becoming more critical as they become more experienced and probably if it was not for the loyalty program (FB) or a corporate agreement these customers will not have been loyal as they are now in the spuriously or latent loyal customers groups (Dick and Basu, 1994).
Combining these elements the self selection might lead to a different distribution, showing higher numbers of negative and average rating passengers. Therefore H5 & H5A will be tested:
H5 A decrease in overall satisfaction is expected H5A An increase of dissatisfied respondents is expected
Another antecedent moderating the overall satisfaction score is expected to be the decreasing of the importance of the in-flight elements in the overall regression model for overall satisfaction. In this model overall satisfaction is influenced by overall in-flight satisfaction, overall on ground satisfaction and punctuality. This shift is potentially caused by the difference in methodology, the shift from an in-flight questionnaire to an after flight (experience) questionnaire enables the passenger to evaluate more elements of the total journey. But also alters the passengers last top of mind experience with AFKL as this will not be the experience on board but will be the landing, exiting the plane, baggage claim and maybe complaint handling or the ride home.
28 H5B In-flight elements will become of less importance within the overall regression model
eDevice
Research of Anderson and Mittal, 2000 and Oliver and Swan, 1989 showed that an increased amount of personal attention for a single customer can lead to an increase in satisfaction and therefore profit; this is also supported by Heskett, 1994 as shown in figure 2.2. This stresses the importance of having well trained crew who are willing and able to hear the voice of the customer.
“crew needs more time to explain the function of the eDevice compared to the current pen & paper, therefore crew influence on satisfaction grades with this method is greater than with the other
methodologies”
(KLM Cityhopper – Cabin Attendant)
Because a difference is assumed between pen & paper and the eDevice on crew satisfaction this difference should even be bigger when comparing it to the email methodology due to the loss of a customer interaction point (moment of truth for crew members).
H6 Increased personal attention will lead to an increase in In-Flight satisfaction H6A In-Flight elements will gain importance in the overall regression model
2.4. Data quality
Switching methodologies always is a difficult and daunting step for companies, this shows the great value of research to a company. Because of the great value research contributes to companies the data quality retrieved from the new methodology should be at least as good as they are currently retrieving. Research has shown that the biggest concerns for companies regarding the online methods is the unevenly distribution among segments caused by accessibility of the internet/email by the different segments around the world (Craig and Douglas, 2001). This is a legit concern in 2001, but today 10 years later the greater part of the world has access to the internet and continues to rise with significant numbers. Previous research by AFKL has also shown that more than 95% of their customers has weekly or more frequently access to the internet, therefore for this study among passengers of AFKL this is not expected as a major drawback.
29 the years to come. General research done in this field shows mixed results for each purpose, industry and segment. AFKL specific this study hypothesizes that the data quality of the email method will be lower that then current pen & paper method.
2.4.1. Corporate Image
The empirical model of Andreas Zins (2001) shows a significant (P<0.05) relation of .60 airline image on satisfaction.
This empirical model is build upon the same methodology as the email eSCORE methodology, as invitations have been emailed after the flight. Literature has not designed such a model for the on board methodology. What is expected, is that image will have a far lower impact on satisfaction because more other cues of the actual service are apparent.
Keller (1993) describes image on a company level as:
“..perceptions of an organization reflected in the associations held in the consumers’ memory” (Keller, 1993)
What basically means that previous experience with the company is necessary, especially when combining this with the model of Zins (2001) who found that 60% is moderated by quality. A construct that can only be evaluated in a correct way after experiencing the actual service. This previous experience is also supported by Hu et al (2009), who found that service quality, customer satisfaction and perceived value moderate corporate image (figure 2.5).
30 Grönroos, 1988; Gummesson 1993, and Andreassen 1997, argue that corporate image is believed to create a halo effect on consumers’ satisfaction judgment. Combining this with the knowledge that emotions tend to decrease over time (Gross, 2003) it is to be expected that satisfaction measurement will be less reliable for late respondents than for early respondents. Especially for late respondents it might become more a corporate image measurement instead of a satisfaction measurement limiting the amount of variance in the answers and thereby limiting the usability and trustworthiness of the online off board method. Therefore the following is hypothesized:
H7A Corporate image will cause a decrease in the answering variance
Although variance for the email eSCORE will decrease over time an overall increase compared to SCORE is expected for the email. For the eDevice no changes in variance is expected.
H7 Overall variance is expected to increase for email and remain unchanged for the eDevice
2.4.2. Response time
Several studies from the past have indicated significant differences between early and late respondents. Findings from these studies suggest that late respondents can evaluate the service significantly different than early respondents. (Barkley and Furse 1996; Lasek et al. 1997). This difference in timing is apart from creating a satisfaction bias on flight level also able to bias the overall satisfaction for comparison with the pen&paper method. Therefore it is important to find out what the impact, both on satisfaction and item missing between early and late respondents is for eSCORE.
H7B The bigger the time gap gets the lower the answering variance gets
31 Previous research of Schaefer & Dillman, 1998 (figure 2.3) learned us that response time for email is very high. Over 50% of the respondents did answer within five days. For an evaluating research five days can be very long. Therefore H2B, should not only find out if response quality drops over days but also when it starts to significantly differ from the first days, making it unreliable. For this item we should keep into consideration its importance regarding seasonality. In the summer AFKL operates and increased amount of holiday flights and a decreased amount of business flight this lengthens the time a passenger is time abroad, for a part of the sample also enlarging the time gap.
2.4.3. Item completeness
As hypothesized before differences in segmentation and satisfaction are expected and might be explained or corrected using weighting variables based on real booking data. But all this would be useless if the data retrieved is of low quality and therefore not producing accurate, usable results. Having missing (key) results in a not randomly distributed missing items curve, the results can be seriously biased endangering the validity of the survey (Olson, 2006). Hoek et al. (2002) showed that web questionnaires have the lowest predictive validity. So, the completeness of the questionnaire should be of great concern when evaluating the quality of new tools as it may cause a bias (Mazor et al, 2002). Research of Cole (2005) found that web surveys have higher amount of missing values. Apart from item completeness, response time also has a significant impact on data quality and bias creating potential. Therefore an overall decrease in data quality for the email eSCORE method is expected:
H8 A decrease in data quality is expected for eSCORE H8A A decrease in item completeness is expected for eSCORE
32
Chapter 3. Methodology
This chapter describes the methodology applied for this research. To gain further understanding of the topic and its implications a small background of the cases used are given. The research design elaborates on the methods used to conduct, administer, collect and analyze the data. This chapter ends with the plan of analysis, explaining the different types of statistical approaches used in this research.
3.1. Empirical study
The aim of this research is to gain insights in the effect of a new methodology impacting passenger satisfaction measurements. Key items in this research are the impact of the methodologies on volumes, satisfaction and segmentation. Again, all leading to the main research question: “How can the current pen and paper method be replaced by either email or digital in-flight surveys, without harming representativity?”
3.2. Research case
AFKL has a long history in passenger satisfaction research, this continuous research for AFKL has always been conducted using a pen & paper questionnaire during flight. KLM, before the merger has also successfully used a split design, meaning that questionnaires where distributed on the ground, evaluating ground services and in the airplane, obviously evaluating in-flight elements.
The last few years a growing demand for faster, more up to date and flexible data collection became visible within the organization. A feasibility study conducted during my internship showed AFKL that there are two possible substitutes for the current pen & paper method. These are the eDevice on board (iPad) or the email invitation send to the PAX after the flight containing the link to the web questionnaire.
Both new methods require extensive testing before they can be incorporated in daily AFKL business. The data retrieved from these tests are used in this research. Both the email and the eDevice method will be compared with the current SCORE method, were results tend to deviate they will (if possible) be compared with actual booking data from the PNR.
33
3.3. Research design
According to Malhotra, 2007 the importance of a research design is that it is a blueprint for conducting a correct marketing research project.
This research focuses on three different tools of collecting customer satisfaction feedback. All three have their limitations and drawbacks as well as their unique opportunities. For AFKL research the choice selection with the highest weight would be the research perspective. This means that the final and most important input comes from this perspective. So, which tool the eDevice or the email delivers AFKL a better representation of their passengers?
In this research the independent variables are the variables that influence the representativity of the tool. For each of these variables a hypothesis is tested to analyze their impact. These variables
are: response rates and volumes, behaviour and profile segmentation, satisfaction and data quality. The dependent variable is representativeness compared to current SCORE.
3.3.1. Sampling size
Literature suggests that the sample must be a correct representation of the population (Baarda, de Goede, 2001 and Malhotra, 2007). But both sources suggest different minimum sample sizes. Since this research is problem identification research a sample size of 500 is suggested by Malhotra, 2007. Baarde and de Goede, 2001 suggest 600 respondents. In addition, they propose that a minimum sample size of 100 respondents is required to make valid inferences about the population. But Malhotra (2007) states that a sample size is influenced by the average sample size used in similar studies. Meaning that the for the sample size resource constraints need to be taken into account. An overall advice therefore by Malhotra (2007) is to have at least 200 respondents.
As described above, we found that a theoretical minimum is to be found between 100 (preferably 200) and 600 respondents. For both the pen&paper method and the email method this minimum is exceeded. But due to resource constraints the eDevice sample only contains 177 valid questionnaires, exceeding the minimum of 100 and therefore inferences about the population can be made
Methodology Sample size Collection period
Pen & Paper AFKL 49.057 April 2011 (4 weeks)
Pen & Paper KLC 6.045 Summer 2010 (4 weeks)
eDevice KLC 177 Summer 2010 (4 weeks)
Email AFKL 186.317 March & April 2011 (6 weeks)
34
3.3.2. Sample & population characteristics
The characteristics of the population and sample will differ in each dataset therefore I shortly discuss some built-in biases caused by the methodology below.
Pen & Paper has no bias in the sample. According to a sampling plan randomly passengers are invited to conduct the survey. Therefore pen & paper data is seen as a correct representation of the population. Figure 3.1 is based on this data.
For the eDevice option the research was conducted in the summer of 2010 on board of KLM Cityhopper flights. For correct comparison data from the pen & paper methodology of the same period and carrier (KLC) are used. See appendix VI for a screenshot.
For the e-mail methodology larges shifts are expected between the sample composition and the population. These are caused by the previously discussed built-in biases. For AFKL to send out an email the email address has to be known, this is only the case for direct bookers (AFKL office, website, telephone) and for flying blue members. This biases changes the FB – non FB ratio and the direct indirect bookers’ ratio as it is known. Therefore in the analysis for the email methodology these groups will be partly analysed separately. The questionnaire itself contains in all setups 95 questions. In appendix VII a screenshot can be found.
Figure 3.1 shows the overall AFKL passenger profile based on SCORE data:
35
3.3.3. Data collection
Pen & Paper data was collected in April 2011 which was the same period as the fieldwork for the email version. This was done onboard of all AFKL flights using a preformatted sampling plan linked to employee number and aircraft type. Up to seven questionnaires are equally distributed among cabins on each flight.
eDevice – data is collected on board of KLM Cityhopper flights in the summer of 2010. The same sampling plan as the pen & paper method was used. The questionnaire used is a exact copy of the paper questionnaire, only loaded onto an Ipad. After the test some interviews were conducted among cabin attendants to get a clear view on the feasibility regarding distribution and explanation on particularly short flights (<60minutes).
Email – 20 till 40 email invitations were sent per flight to a random selected number of passengers on each AFKL flight in the month of April 2011. The email send to passengers contains a link to a web questionnaire presented in multiple languages. The invitation was always send in two languages, primarily one of the carrier and secondary depending on point of sale. English was always included, this to ensure that 95%+ could read the invitation. This web questionnaire is the same as the pen & paper questionnaire, so forced answering was not incorporated (appendix VIII). This was done on purpose to exclude additional biases.
3.3.4. Measurement design and scaling
For this research it was of great importance to re-use the current pen&paper questionnaire as it can be found in Appendix II. This questionnaire is based upon a 5 point Likert-type scale with scale anchors ranging from excellent till poor, and agree completely to disagree completely. Furthermore the questionnaire is split into different sections: profile questions, overall satisfaction questions, in-flight questions, ground questions and various segmentation statements.
3.4. Plan of analysis
The previous chapter has discussed the way of collecting data and the questionnaire lay-out. This chapter will focus on the analysis of the collected data.
3.4.1. Statistical tests
Statistical tests can be grouped into two different kinds of tests. The first is the descriptive statistics; these are used to present data in a clear and compact manner. This type of statistic is manly used to describe characteristics of the sample such as demographics, profile information, and how these groups rate the different satisfaction items.
36 is limited. The groups are: FB and non-FB, explanation of the various tier levels within FB are to be found in appendix III. When no significant differences are found in the analysis between these groups, they are not split in the analysis.
The second statistical tests are the mathematical statistics; these are used to show the relation between two or more variables. The goal is to test the significance of the hypotheses. The following mathematical statistics are used:
Reliability test & Validity check
Malhotra, 2007 states that when a multi-item scale is used it should be evaluated for accuracy an applicability. Malhotra suggests using an internal consistency check to gain insights in possible measurement errors. A commonly used approach to measure this is the Cronbach Alpha. This coefficient has a range from 0 to 1, where a value of 0.6 or lower generally indicates unreliable internal consistency. When the value exceeds 0.6 the variables can be grouped into a new factor. In this research the Cronbach Alpha will play a big role in comparing answering patterns between SCORE and eSCORE.
Chi-square test
This analysis will be conducted to check if the sample is a correct and reliable sample taken from the population. It will be used for cross checking and variance checking several times. Analysis of Variance (ANOVA)
An ANOVA and its post hoc test (LSD, limited significance difference) tests are used frequently to explain and show the different opinions of the different passenger groups.
Multiple Regression Analysis
If an ANOVA has showed a significant difference between methodologies or customer groups a regression shows whether the impact of this difference is of great meaning. Also the contribution of the different sub variables to main variables is explained using a regression analysis.
(Paired) sample t-test
Serves the same cause as the ANOVA, only the t-test is used when 2 groups need comparison, where the ANOVA can take into account up to 50 groups.
37
Table 4.1, Coverage in # of questionnaire
Difference compared to SCORE
Chapter 4. Results
This following chapter presents the results of the collected data. Appendix IV and V will contain additional statistical tables, as only high level tables and figures are shown in this chapter. Data discussed in this chapter is of great value to AFKL, therefore mean scores are not shown and only the deviations between SCORE and eSCORE are given on a 10-point scale.
4.1. Representativity
All surveys conducted can suffer biases. Some research is more vulnerable to biases than others, what is known, is that AFKL satisfaction research is very vulnerable to a small shift in sample composition as it will lead to a great shift in satisfaction. Therefore the variables that influence satisfaction in the eSCORE and SCORE samples are shown. Also the impact of these segmentation variables on overall satisfaction is analyzed by using a t-test and ANOVA. As discussed before these variables are mostly split by the FB and non FB, this is done due to the fact that the sample ratio from the eSCORE methodologies are not representative for the population. A chi-square test (appendix V.1) showed on this most important profiling variable a value of X2 = 0.000 p = .001 for email. So, instead correcting this by weighting variables, these groups are for the bigger part split in the analysis.
4.1.1. Volumes, response rates and coverage
SCORE and email have run parallel making a direct comparison of volumes straight forward. eSCORE email gathered 114.264 completed questionnaires where SCORE retrieved 49.075 questionnaires in the full month of April. This means that email can double the volumes of SCORE. But volume is not all that matters, a correct representative sample is. Apart from biases in customer groups, some networks may also be overrepresented due to distance from home (Hartzinger, 2000) or local infrastructure. Below email and eDevice are discussed.
Volumes, Coverage & response rate email
In April 2011 AFKL has operated around 50.000 flights (including subsidiaries of AFKL). Table 4.1 shows the networks with their increased volumes compared to SCORE. On flight number level2 eSCORE has covered 281 more flight numbers than SCORE. As was expected a clear increase is visible for the French home market, since the average flight time of these
2
Flight number is the number given to a certain route, this is not a unique flight. Flight number + date is unique.
Network eSCORE vs SCORE
38
Figure 4.1, Response Rates
national flights is so short limiting time for the crew to distribute the questionnaire on board. The distribution among the networks is not always correct but are eliminated after an AFKL data weighting process weighted it according to the actual number of flown flights. This ensures correct distribution and eliminates biases between networks, routes, and flights. Therefore email is not only capable of increasing volumes but also increases coverage, this supports H1 (An increase in volumes for the email eSCORE is expected).
Volumes & Coverage eDevice
Average KLC flights carry around 75 people, averaging 60-70 minutes of flight time. SCORE currently delivers ±0,67 questionnaires per KLC flight (0,6 for 60minutes and 0,7 for 60+ minutes(p=0.00, t=-5,549)). This is below the target of 1% minimum, hypothesized is an even lower volume on short flights for the eDevice. A field trip/study on board has shown that including explanation an eDevice questionnaire can take up to 20 minutes. This is 5 minutes longer than the pen & paper version demanding more attention in the already busy cabin crew schedule. This is also confirmed by the cabin attendants active on these test flights, they state that: “passengers like it, but it absorbs more time than pen & paper”. This way it will be impossible for more than half of the flights to collect minimal required 1% coverage.
“Time is often too limited (especially on short flights) to distribute, explain the functionality of the iPad, and complete the questionnaire within the given time frame.”
Senior Cabin Attendant KLC
Combining this with the knowledge that during the test on 17 of the 84 flights the iPad was not distributed due to time constraints and that on other short flights it was nearly impossible to distribute supports H2 (Decrease in volumes on short flights is expected), so this hypothesis cannot be rejected.
4.1.2. Loyalty
Hypothesized is that the loyal customers have higher response rates. Figure 4.1 shows a positive correlation between response rate and loyalty
level. The highest and lowest FB levels, respectively C2000/Skipper and Jeune are loyalty levels that
39
Figure 4.2. Stated Loyalty
Sig. between all levels is (p<0.01), answering options completely and somewhat agree are group into a yes; which is shown.
Figure 4.3. Loyalty
*) loyal customers are truly loyal + spuriously loyal (Dick and Basu, 1994)
cannot be earned but are given to members according to age (Jeune) or job title (C2000/Skipper3) and therefore deviate from the others. To verify if these FB levels indeed measure stated loyalty an additional question is analyzed: “I am loyal to my preferred airlines” anchoring (5) Agree Completely to (1) Disagree Completely. In figure 4.2 the frequencies are shown. It may be clear that there are differences between loyalty and non loyalty program members. Also within the loyalty program the differences are significant (p<0.01). Additionally eSCORE tends to draw more loyal customers as on average 6% more passengers claim to be loyal than with the traditional SCORE methodology (appendix table IV.3).
As previous research has shown stated behavior from a respondent might deviate strongly with their behavior. Therefore the respondents group of AFKL is segmented into the quadrant of Dick and Basu (1994) who calculates loyalty using behavior. Figure 4.3 shows a strong increase in loyal non FB respondents compared to current SCORE. Loyalty has been based on the combination of repurchase intention, by Andreassen (1997) proven as an indicator for loyalty and the variable explaining the in,- and extrinsic motivation for choosing the specific AFKL flight. In appendix IV the detailed grading criteria and calculations
for the loyalty levels of Dick and Basu’s model can be found.
Response rates showed correlation with loyalty level. This matches the data gathered for calculating behavior loyalty. For behavioral
3 This loyalty level cannot be earned and is given only to CEO’s, European politicians, and other leaders of industry 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
