Big Data To Improve Strategic-1

Schriftenreihe der HHL Leipzig
Graduate School of Management
Maximilian Schosser
Big Data to Improve

Strategic Network
Planning in Airlines
LEIPZIG
GRADUATE SCHOOL
OF MANAGEMENT
Schriftenreihe der HHL Leipzig
Graduate School of Management
Reihe herausgegeben von

Stephan Stubner, Leipzig, Deutschland
In dieser Schriftenreihe werden aktuelle Forschungsergebnisse aus dem B ereich
Unternehmensführung präsentiert. Die einzelnen Beiträge spiegeln die wissen
schaftliche Ausrichtung der HHL in Forschung und Lehre wider. Sie zeichnen sich
vor allem durch eine ganzheitliche, integrative Perspektive aus und sind durch den
Anspruch geprägt, Theorie und Praxis zu verbinden sowie in besonderem Maße
internationale Aspekte einzubeziehen.
Weitere Bände in der Reihe http://www.springer.com/series/12648

Maximilian Schosser
Big Data to Improve

Strategic Network
Planning in Airlines
With a foreword by Prof. Dr. Iris Hausladen
Maximilian Schosser
HHL Leipzig Graduate School of Management
Heinz-Nixdorf Chair of IT-based Logistics
Leipzig, Germany
Dissertation HHL Leipzig Graduate School of Management, 2019
Schriftenreihe der HHL Leipzig Graduate School of Management

ISBN 978-3-658-27581-5 ISBN 978-3-658-27582-2 (eBook)
https://doi.org/10.1007/9783658275822
Springer Gabler
© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2020
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with
regard to jurisdictional claims in published maps and institutional affiliations.
This Springer Gabler imprint is published by the registered company Springer Fachmedien
Wiesbaden GmbH part of Springer Nature.
The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany
For my wife Karin who filled the days of my PhD studies with pure joy.
For my mother Jutta who taught me the most important trait as a re-
searcher – curiosity.
For my father Rudolf whose perfectionist mind helped with great sugges-
tions for this thesis.
Foreword
Big data not just evolved as a popular buzzword over time but is meanwhile
seen as a high potential field for improving business processes and deci-
sions taken by responsible persons in different industry sectors, such as
airlines.
Nevertheless, available advantages of big data have not yet been ade-
quately measured from an economic perspective and have very often not
yet been exploited at all or to a reasonable level. Additionally, correspond-
ing theoretical concepts and scientific developments focusing on the area
of network planning in airlines are currently more or less missing. Those
challenging deficits make the topic considered by Mr. Schosser highly rel-
evant both from a theoretical as well as a practical perspective.
Thus, the main objective of the thesis consists in closing both the scientific
research gap and providing a solution to practitioners focusing on the as-
sessment of big data opportunities in the network planning context of air-
lines.
The author provides with the step-by-step, theoretically and empirically
based development of the framework, its elements and procedures in the
present doctoral thesis an outstanding analytical as well as conceptual per-
sonal contribution. The framework is from a content-related point of view
to be honored as a pioneering achievement and the dissertation contains
a lot of new findings that represent a starting point for further work predom-
inantly in the research and practice field of big data evaluation focused on
airline network planning that can be transferred to further use cases re-
spectively fields of applications.
The book, which is based on a dissertation at the HHL Leipzig Graduate
School of Management, is aimed equally at readers from science and prac-
tice, dealing with (big) data collection, economic evaluation of big data as
well as big data analytics.
Leipzig, May 2019 Prof. Dr. Iris Hausladen

Preface
The idea for this PhD thesis was born during a consulting project at a Eu-
ropean airline, where different departments were at completely different
maturity stages of using Big Data. While at some departments network
planning was done the old-fashioned way, with legacy IT-systems, data
and processes, other departments had already implemented a much more
agile way of integrating Big Data.
The network planning department was frequently approached by data pro-
viders with concrete offers, but there was no method to evaluate the impact
of the data for airline network planning, and most of the offers were turned
down for this reason. At the same time, there was no research available to
answer this question. The main objective of this PhD thesis is hence to
alleviate the lack of evaluation methods and provide a concise answer/ap-
proach/framework of how Big Data can create value for individual network
planning steps.
The target audience of this thesis comprises airline network planners of all
seniority levels and fellow researchers working on Big Data applications
for the transportation industry, on network optimization problems or on
commercial topics in airlines. I tried to find a middle ground between con-
tent of pure scientific interest (e.g., chapters 2 and 3), and presenting rel-
evant findings for practitioners (chapters 4, 5 and 6).
Finally, I want to thank everyone who contributed to this PhD thesis, in
particular my thesis advisor Prof. Dr. Iris Hausladen, who provided guid-
ance and challenge whenever necessary. Further, I want to thank all inter-
view partners of our airline case study group, who need to stay anonymous
due to non-disclosure agreements. Philipp Behrends, Karin Garcia and Ru-
dolf Schosser made an invaluable contribution by reviewing and comment-
ing on various drafts of this thesis. My thanks go also to all staff and stu-
dents of HHL Leipzig Graduate School of Management who made my time
there so enjoyable. Finally, I want to thank Maximilian Rothkopf for helping
to shape the idea and David Speiser for his support of my PhD project. My
PhD thesis would not have been possible without the financial support dur-
ing my educational leave granted by the Zurich Office of McKinsey & Com-
pany.
Berlin, May 2019 Maximilian Schosser

Contents
1 Introduction............................................................................................ 1
1.1 Problem and research gap definition............................................. 1
1.1.1 Practical problem ...................................................................1
1.1.2 Scientific research gap ..........................................................2
1.2 Objective of the study and research questions .............................3
2 Methodology .......................................................................................... 5
2.1 Development of research design................................................... 6
2.2 Literature review ..........................................................................11
2.2.1 Design of a structured literature review process .................11
2.2.2 Identification of keywords, databases, and journals ............13
2.2.3 Results of the structured keyword search ...........................15
2.2.4 Description of the research gap ...........................................17
2.3 Comparative case study ..............................................................18
2.3.1 Selection of the case study type ..........................................19
2.3.2 Case sampling .....................................................................20
2.3.3 Data collection and analysis techniques ..............................23
2.3.4 Research quality assurance ................................................27
3 Theoretical foundation.........................................................................29
3.1 Development of a theoretical concept .........................................29
3.2 Airlines and their business models ..............................................31
3.2.1 The airline industry ..............................................................32

XII Contents
3.2.2 Development and recent trends in the airline industry ........34
3.2.3 Airline business models .......................................................35
3.2.4 Major business processes of airlines ...................................43
3.3 Introduction to the network theory ...............................................44
3.3.1 Definition of networks ..........................................................44
3.3.2 Distinction between the network theories ............................48
3.3.3 Fundamentals of the graph theory .......................................50
3.3.4 Network flows and network optimization .............................55
3.3.5 Design of flow networks .......................................................62
3.4 Airline networks ...........................................................................64
3.4.1 General properties of airline networks .................................65
3.4.2 Types of airline networks .....................................................67
3.4.3 Airline network economics ...................................................74
3.4.4 Airline network indicators .....................................................78
3.5 Network planning in airlines ........................................................84
3.5.1 Components of network planning ........................................84
3.5.2 Long-term planning ..............................................................87
3.5.3 Spatial optimization ..............................................................96
3.5.4 Temporal optimization .........................................................99
3.5.5 Operational optimization ....................................................103
3.5.6 Network planning in cargo airlines .....................................106
3.5.7 Data needs of network planning ........................................110
3.5.8 Definition of strategic network planning .............................124

Contents XIII
3.6 Big data .....................................................................................129
3.6.1 Definition of data ................................................................129
3.6.2 Characteristics of big data .................................................130
3.6.3 The big data value chain....................................................132
3.6.4 Major types of big data ......................................................135
3.6.5 Big data types in airline and tourism research...................136
3.6.6 Issues and risks of big data ...............................................139
3.7 The RBV pertaining to data and network planning ....................140
3.7.1 The basic concept of the resource-based view .................141
3.7.2 The RBV in information systems research ........................144
3.7.3 (Big) Data in the RBV ........................................................147
3.7.4 The RBV in logistics research............................................153
3.7.5 Logistic networks in the RBV .............................................157
3.7.6 RBV-based research concept of big data in airline NP .....159
4 Status quo of strategic network planning in airlines ..........................161
4.1 The network planning process in reality ....................................161
4.1.1 Roles and responsibilities of NP departments ...................161
4.1.2 Sequence and time horizon of NP process steps .............164
4.2 Data types used for network planning .......................................170
4.2.1 Practical use of data types derived from literature ............170
4.2.2 Currently used data types in airline network planning .......174
4.2.3 Reliability and usefulness of data types ............................178
4.3 IT tools used for network planning ............................................181

XIV Contents
4.3.1 Market of IT tools for airline network planning ...................182
4.3.2 Current use of IT tools in the case study airlines...............184
4.3.3 Data capability of current IT tools ......................................186
4.4 Performance measurement of airline network planning ............188
4.4.1 Planning and optimization goals for NP departments .......189
4.4.2 KPIs currently used for airline network planning ...............191
5 Big data opportunities for airline network planning ...........................195
5.1 Information needs for airline network planning .........................195
5.1.1 Definition of information needs for airline NP ....................195
5.1.2 Satisfaction of information needs with current data use ....199
5.2 Development and evaluation of BDOs for airline NP ................204
5.2.1 Specific BDOs for airline NP information needs ................204
5.2.2 Transportation demand forecast ........................................209
5.2.3 Trend identification ............................................................223
5.2.4 Competitor monitoring .......................................................227
5.2.5 Real movements of passengers and goods ......................230
5.2.6 Real-time planning constraint monitoring ..........................235
5.2.7 Incident monitoring ............................................................238
5.3 Evaluation of BDO potential ......................................................240
5.3.1 Qualitative assessment of BDO potential ..........................240
5.3.2 Differences in BDO potential by business model ..............242
5.3.3 Implementation of BDOs in current IT systems for NP ......248
5.3.4 Feasibility assessment of BDOs ........................................250

Contents XV
6 Financial impact of big data for airline network planning ..................261
6.1 Choice of appropriate BDOs based on KPIs .............................262
6.1.1 Impact of BDOs on airline NP KPIs ...................................262
6.1.2 Suitable KPI metrics for BDOs in airline NP ......................274
6.2 Financial benefit calculation for big data opportunities .............279
6.3 Business case framework for big data in airline NP ..................288
6.3.1 Introduction of Example Air................................................289
6.3.2 Calculation of BDO impact for relevant KPI groups...........291
6.3.3 Scenario-based sensitivity analysis of KPI benefits ..........300
6.4 Cost evaluation of big data opportunities ..................................303
6.4.1 Cost composition of big data projects ................................304
6.4.2 Cost drivers for big data opportunities ...............................308
6.4.3 Cost estimation framework for BDOs ................................311
6.5 Comparison of benefit and cost potential of BDOs ...................314
7 Discussion and contrast with the scientific body of knowledge ........319
7.1 Status quo of network planning in airlines .................................320
7.1.1 Airline business models and network planning..................320
7.1.2 The network planning process in literature and practice…322
7.1.3 Current data use in literature and practice ........................324
7.2 Big data in airline network planning...........................................327
7.2.1 Big data characteristics of BDOs for airline NP .................327
7.2.2 Relevance of BDOs for other airline departments .............329
7.2.3 Integration of BDOs in corporate big data ecosystems .....331

XVI Contents
7.3 Revisiting the RBV for big data in airline network planning ......333
7.3.1 Review of the components of an adaptive NP capability ..333
7.3.2 Organizational readiness of airlines ..................................337
8 Conclusion and suggestions for further research .............................341
8.1 Conclusions on research outcome ............................................341
8.1.1 How can “big data” be defined for network planning in

airlines [RQ 1]? ............................................................................341
8.1.2 What is the status quo of the business process, IT

systems, and data use in airline network planning [RQ 2]? .........342
8.1.3 Which new “big data” opportunities are most suited to

improve network planning for airlines or replace existing data
types [RQ 3]? ..............................................................................343
8.1.4 How can the impact of big data opportunities for airline
network planning be quantified [RQ 4]? .......................................345
8.1.5 Overall conclusion .............................................................348
8.1.6 Contribution to the scientific body of knowledge ...............349
8.2 Suggestions for further research ...............................................350
8.2.1 Methodological enhancements ..........................................350
8.2.2 Content suggestions for further research ..........................351
List of Appendices ..................................................................................353
References .............................................................................................411
List of Figures
Figure 1.1 – Research objectives and research questions ....................... 4

Figure 2.1 – Overview of research methodology development ................ 5
Figure 2.2 – Four-phase research design ................................................. 9
Figure 2.3 – Tactics to ensure the quality criteria for selected research
methods............................................................................ 10
Figure 2.4 – Structured literature review process ................................... 12
Figure 2.5 – Build-up of a relevant body of literature .............................. 16
Figure 2.6 – Double tri-angulation approach ........................................... 24
Figure 2.7 – Data collection and analysis techniques of status-quo
assessment ...................................................................... 25
Figure 2.8 – Data collection and analysis techniques of BDO
evaluation ......................................................................... 26
Figure 3.1 – Evaluation of the candidate theories ................................... 30
Figure 3.2 – Structure and purpose of the theoretical foundation
chapter ............................................................................. 31
Figure 3.3 – The aviation system from a supply-demand perspective ... 32
Figure 3.4 – Overview of standard airline processes .............................. 44
Figure 3.5 – Network theory in transportation networks ......................... 50
Figure 3.6 – Connectivity matrices of an undirected, strongly
connected graph .............................................................. 52
Figure 3.7 – Overview of graph types ..................................................... 53
Figure 3.8 – Degree distribution example ............................................... 55
Figure 3.9 – Simple network with assigned link cost .............................. 57
Figure 3.10 – Maximum capacity determined by the max-flow min-cut
theorem ............................................................................ 60
Figure 3.11 – Exemplary network and solution for the minimum cost
flow problem ..................................................................... 62
Figure 3.12 – Airline network archetypes ................................................ 68
XVIII List of Figures
Figure 3.13 – Lufthansa's South America network in 1934 .................... 70

Figure 3.14 – Lorenz curves for the perfect PP and HS networks .......... 80
Figure 3.15 – Profit contribution margins in route profitability
calculation ........................................................................ 82
Figure 3.16 – Airline network planning phases ....................................... 88
Figure 3.17 – Airline demand forecasting logic ....................................... 90
Figure 3.18 – Methods for market size forecasting ................................. 92
Figure 3.19 – Number of network planning publications, using specific
data types ....................................................................... 113
Figure 3.20 – Number of analyzed contributions by problem
integration....................................................................... 117
Figure 3.21 – Frequency and combinations of network planning and
optimization problems .................................................... 118
Figure 3.22 – Comparison of the network planning process models .... 120
Figure 3.23 – Generic NP process ........................................................ 123
Figure 3.24 – Strategic network planning within the generic airline NP
process ........................................................................... 128
Figure 3.25 – Big data value chain........................................................ 133
Figure 3.26 – Structure of sub-chapter 3.7 ........................................... 141
Figure 3.27 – Research focus in theoretical context ............................. 160
Figure 4.1 – Organizational responsibility model for airline network
planning .......................................................................... 162
Figure 4.2 – Sequence and timing of NP process steps across case
study participants ........................................................... 165
Figure 4.3 – Coding results based on open questions of first interview
round .............................................................................. 171
Figure 4.4 – Clustering of data types by information object .................. 174
Figure 4.5 – Current usage of data types across case study group ..... 177
Figure 4.6 – Perceived usefulness and reliability of the 22 currently
used data types .............................................................. 179
List of Figures XIX
Figure 4.7 – Usefulness and reliability analysis of the currently used

data types ....................................................................... 180
Figure 4.8 – Market overview over commercially available IT systems
for airline network planning ............................................ 182
Figure 4.9 – IT systems currently used by the case study participants
for network planning ....................................................... 185
Figure 4.10 – Degree of automation in current NP process.................. 186
Figure 4.11 – Exemplary integration of the currently used data
sources in airline network planning IT solution .............. 188
Figure 4.12 – Performance management in NP departments .............. 190
Figure 4.13 – KPI framework for airline network planning .................... 191
Figure 4.14 – Primary and secondary NP performance targets in case
study airlines .................................................................. 193
Figure 5.1 – Information needs throughout the airline NP process ...... 196
Figure 5.2 - Allocation of currently used data sources to airline NP
information needs ........................................................... 200
Figure 5.3 – Addressed information needs of BDOs ............................ 208
Figure 5.4 – Overall qualitative potential assessment of BDOs ............ 241
Figure 5.5 – BDO qualitative potential evaluation patterns by business
model.............................................................................. 243
Figure 5.6 – Cargo vs. passenger-specific BDOs ................................. 248
Figure 5.7 – Ease of implementation assessment for current NP IT
system ............................................................................ 249
Figure 5.8 – Feasibility assessment model for big data acquisition and
analytics ......................................................................... 252
Figure 5.9 – Potential and feasibility evaluation matrix for BDOs ......... 257
Figure 6.1 – Logical process to derive cost and benefit estimate for
BDOs .............................................................................. 261
Figure 6.2 – Direct and indirect KPIs currently used by the case study
group airlines .................................................................. 263
XX List of Figures
Figure 6.3 – Estimated impact of BDOs on route and network

profitability ...................................................................... 264
Figure 6.4 – Estimated impact of BDOs on load factor ......................... 265
Figure 6.5 – Estimated impact of BDOs on forecast accuracy ............. 267
Figure 6.6 – Estimated impact of BDOs on schedule robustness ........ 268
Figure 6.7 – Estimated impact of BDOs on asset utilization ................. 269
Figure 6.8 – Estimated impact of BDOs on FTE efficiency ................... 271
Figure 6.9 – Estimated impact of BDOs on DOC and IOC ................... 272
Figure 6.10 – Relevant metrics for KPI calculation in airline NP........... 275
Figure 6.11 – Profile of Example Air ..................................................... 290
Figure 6.12 – Input data and estimates for the calculation of KPI
metrics for BDO-benefit evaluation ................................ 291
Figure 6.13 – Illustrative flight portfolio by average flight profit ............. 292
Figure 6.14 – Example benefit calculation for profit delta from flight
portfolio changes ............................................................ 293
Figure 6.15 – Example-benefit calculation for total network profit ........ 293
Figure 6.16 – Illustrative distribution of forecast deviation for Example
Air ................................................................................... 294
Figure 6.17 – Example-benefit calculation for load factor ..................... 296
Figure 6.18 – Schematized rotation plans for Example Air aircraft
types ............................................................................... 297
Figure 6.19 – Example-benefit calculation for idle hours per aircraft
type................................................................................. 297
Figure 6.20 – On-time performance of Example Air (percent and total
number of flights) ........................................................... 298
Figure 6.21 – Example-benefit calculation for on-time performance .... 299
Figure 6.22 – Cost components of BDA projects structured along the
big data value chain ....................................................... 305
Figure 7.1 – Structure of discussion in chapter 7 .................................. 319
List of Figures XXI
Figure 7.2 – Comparison of data types used by airlines and in NP

literature ......................................................................... 325
Figure 7.3 – Big data characteristics of 23 BDOs for airline NP ........... 329
Figure 7.4 – Exemplary airline big data ecosystem .............................. 333
Figure 7.5 – Revised research concept based on RBV ........................ 336
Figure 7.6 – Maturity model dimensions and maturity patterns by
airline business model.................................................... 338
List of Tables
Table 2.1 – Decision-making criteria for research approach .................... 7

Table 2.2 – Comparison of research methods .......................................... 8
Table 2.3 – Keyword extraction for the literature review ......................... 13
Table 2.4 – The case study method adopted from Eisenhardt (1989).... 18
Table 2.5 – Specification criteria for case study airlines ......................... 21
Table 2.6 – Airline case study participant overview ................................ 22
Table 3.1 – Airline business model classifications in the literature ......... 36
Table 3.2 – Key characteristics of airline business models .................... 38
Table 3.3 – Overview of transportation networks by transportation
mode ................................................................................... 46
Table 3.4 – Graph theory terminology..................................................... 51
Table 3.5 – Airline network terminology .................................................. 65
Table 3.6 – Overview of literature in airline NP processes ..................... 86
Table 3.7 – Identified data types in network planning literature ............ 111
Table 3.8 – Integrated network planning and optimization models ....... 115
Table 3.9 – Definition criteria for big data in literature .......................... 130
Table 3.10 – Social media services by Gundecha and Liu (2012, p. 3) 136
Table 3.11 – Big data types mentioned in airline and tourism literature 138
Table 3.12 – Concepts of IT resources ................................................. 145
Table 3.13 – Data resource characteristics by Levitin and Redman
(1998)................................................................................ 149
Table 3.14 – Logistic capabilities defined by Gligor and Holcomb
(2012)................................................................................ 156
Table 4.1 – Description of currently used data types derived from
literature ............................................................................ 172
Table 4.2 – Additionally coded data types not used in literature .......... 175
Table 5.1 – Classification of airline big data types ................................ 205
XXIV List of Tables
Table 5.2 – BDOs ranked by business model ....................................... 245

Table 5.3 – Feasibility assessment of BDOs ........................................ 254
Table 6.1 – Impact of BDOs on airline network planning KPIs ............. 273
Table 6.2 – Required input data and estimations for KPI metrics ......... 279
Table 6.3 – Big data project scenarios .................................................. 300
Table 6.4 – BDO benefit requirement for break-even of BD project
scenarios........................................................................... 303
Table 6.5 – Cost risks of cost drivers for value chain components ....... 309
Table 6.6 – Estimated cost risks of BDOs............................................. 312
Table 6.7 – First indicative comparison of financial benefit and cost
risks................................................................................... 316
List of Abbreviations
ACL - Airport Coordination Limited

ACMI - Aircraft, Crew, Maintenance and Insurance
ADS-B - Automatic Dependent Surveillance Broadcast
API - Application Programming Interface
ASK - Available Seat Kilometers
AWB - Air Waybill
BD - Big Data
BDA - Big Data Analytics
BDO - Big Data Opportunity
BI&A - Business Intelligence & Analytics
BT - Block Time
CAR - Cargo Airline
CASS - Cargo Accounts Settlement System
CCO - Chief Commercial Officer
CEO - Chief Executive Officer
COO - Chief Operations Officer
DOC - Direct Operating Cost
e.g. - For example (Latin: exempli gratia)
EASA - European Aviation Safety Agency
e-AWB - Electronic Air Waybill
EBIT - Earnings Before Interest and Tax
et al. - Et alii
etc. - Et cetera
EU - European Union
FDI - Foreign Direct Investment
FSC - Full-Service Carrier
FTE - Full-Time Equivalent
GB - Gigabyte
GDP - Gross Domestic Product
GDS - Global Distribution System
GPS - Global Positioning System
HS - Hub-And-Spoke
i.e. - That is (Latin: id est)
XXVI List of Abbreviations
IAG - International Airlines Group

IATA - International Air Transport Association
ICAO - International Civil Aviation Organization
IMF - International Monetary Fund
IOC - Indirect Operating Cost
IP - Internet Protocol
IS - Information Systems
IT - Information Technology
KPI - Key Performance Indicator
LCC - Low-Cost Carrier
LoRaWAN - Low Range Wide Area Network
MCT - Minimum Connection Time
MIDT - Marketing Information Data Tape
MIT - Massachusetts Institute of Technology
NLP - Natural Language Processing
NP - Network Planning
NPV - Net Present Value
OAG - Official Airline Guide
O&D - Origin & Destination
OECD - Organisation for Economic Co-operation and
Development
OR - Operations Research
p. - Page
P&RM - Pricing & Revenue Management
pp. - Pages
PP - Point-To-Point
QSI - Quality Of Service Index
RBV - Resource-Based View
RFID - Radio-Frequency Identification
RQ - Research Question
RR - Random Radial
RSK - Revenue Seat Kilometers
RTK - Revenue Ton Kilometers
SCA - Scheduled Charter Airline
SCM - Supply Chain Management
List of Abbreviations XXVII
SRS - Schedule Reference Service

TAT - Turn-Around Time
TB - Terabyte
UK - United Kingdom
US - United States
USD - United States Dollar
VHB - German Academic Association for Business
Research
VM - Virtual Machine
WTO - World Trade Organization
1 Introduction
1.1 Problem and research gap definition
1.1.1 Practical problem
Big data has become one of the most prominent buzzwords of the modern
corporate world, which is a remarkably fast feat for a new phenomenon. It
became fashionable at the onset of the 2010s, fueled by popular articles in
the Harvard Business Review (McAfee & Brynjolfsson, 2012) and the MIT
Sloan Management Review (Davenport, Barth, & Bean, 2012). Its broad
scope makes it directly or indirectly crucial for all industries, and as early
as 2011, Manyika et al. (2011, p. 1) estimated a global big data (BD) value
generation potential of more than 1 trillion US Dollar (USD).
Since airlines had been producing vast amounts of data even before the
term big data was coined, its potential value in the industry was noticed
early on (Noyes, 2014). In a recent survey, more than 60% of aviation ex-
ecutives named big data as a top priority for their company (Hodgson &
Waldmeir, 2018). The airline business is receptive to a wide range of big
data applications: the single customer view enables hyper-personalized
customer experience and marketing activities (Chen et al., 2017; Hodgson
& Waldmeir, 2018; Noyes, 2014); higher granularity of customer insights
can be utilized for customized pricing and revenue management (Boin et
al., 2017); real-time information on transfer passengers and irregular situ-
ations can help reduce delays and smoothen flight operations (Bowcott &
Dichter, 2014; Chen et al., 2017); finally, predictive maintenance can re-
duce unplanned aircraft groundings and lower maintenance cost (Brad-
bury, 2018; Chen et al., 2017).
Although airlines are at the forefront of big data adoption, not all business
processes and departments in the industry participate at the same speed.
Bertram (2017) proposes various big data applications for network plan-
ning (NP), including mobile location data and real-time flight radar data.

M. Schosser, Big Data to Improve Strategic Network Planning in Airlines,
Schriftenreihe der HHL Leipzig Graduate School of Management,
https://doi.org/10.1007/978-3-658-27582-2_1
2 Introduction
However, most NP departments in airlines have not yet adopted any big
data application in their NP process.
Investment in big data opportunities (BDOs) require the creation of a sound

business plan with clearly defined costs and benefits (Bertram, 2017). The
airline business is a low-margin industry that is very reluctant to make large
exploratory investments to prove use cases with an uncertain probability
of success (Hodgson & Waldmeir, 2018). Airline network planners cur-
rently struggle to estimate the benefit potential of BDOs in prioritizing and
formulating reliable business cases. This book addresses this practical
problem by consolidating and structuring potential BDOs for airline NP.
Subsequently, it develops a benefit evaluation methodology based on NP-
specific performance metrics.
1.1.2 Scientific research gap
Big data has experienced a rapid increase in popularity in the scientific

research community. Akoka, Comyn-Wattiau, and Laoufi (2017, p. 110)
and Govindan et al. (2018, p. 344) found a significant increase in the num-
ber of studies on big data, especially since 20141. Historically, computer
science and engineering are the major research domains investigating big
data, while business and management science has been catching up only
recently (Govindan et al., 2018, p. 346). While the scientific discussion on
big data is maturing, there is still no unanimous definition pertaining to its
characteristics (Mauro, Greco, & Grimaldi, 2015, p. 103).
Although a holistic big data value chain has been proposed by Gandomi
and Haider (2015, p. 141), little research has been conducted on the
sources and types of big data (Akoka et al., 2017, p. 111). While infor-
mation systems architecture and big data analytics techniques are under
1
Govindan et al. (2018) found an 8-fold increase of academic publications with big data in
title, abstract or full-text between 2014 and 2017.
Objective of the study and research questions 3
academic focus, big data sources are mostly treated as input factors and
not as key value contributors.
Furthermore, there is little research on big data which is specifically tar-

geted at the airline industry. Authors concentrate on individual use cases,
including predictive maintenance (Badea, Zamfiroiu, & Boncea, 2018), air
traffic optimization (Ayhan et al., 2013), and marketing and flight operations
(Chen et al., 2017). There is no scholarly contribution on the use of big
data for airline NP yet.
This study aims to bridge this gap in two ways. First, it shifts the attention
from big data architecture and analytics towards the opportunities evolving
from specific big data sources. Even the most advanced infrastructure and
analytic tools cannot create value if they use meaningless data. Second,
the study provides a perspective on big data applications for airline NP,
and thus complements research on other domains of application in airlines.
1.2 Objective of the study and research questions
The objectives of this study aim to close the scientific research gap laid out
in the previous section and address the practical problem discussed previ-
ously in order to evaluate big data opportunities for airline NP. As the first
step, the most recent research on big data and on NP in airlines needs to
be assessed. Based on the existing body of literature, a clear definition of
big data in the context of airline NP must be developed to build a solid
foundation to find solutions to the practical problem.
A deep understanding of NP processes, tools, and data types in practice

must be created before applications for BDOs in airline NP can be devel-
oped. The current state of these processes, tools and data types can then
be contrasted with the theoretical concepts of strategic airline NP.
Based on the status quo of airline NP, the information requirements of air-
line network planners can be identified. BDOs should address these infor-
mation needs to create value for airline NP. The definition of concrete
4 Introduction
BDOs and the qualitative evaluation of their applicability and usefulness

for airline NP is the most crucial objective of this study.
Finally, the development of a cost and benefit evaluation scheme for BDOs
in airline NP should address the practical problem described in the previ-
ous section. This requires the determination of relevant key performance
indicators to measure the contribution of big data opportunities in terms of
financial value.
The research objectives can be condensed into four specific research

questions, which are presented in Figure 1.1. All research questions are of
an explanatory nature since the research phenomenon has not yet been
discussed in literature.
Figure 1.1 – Research objectives and research questions2
2
Source: Own illustration.
2 Methodology
The choice of the research methodology does not depend only on the char-
acteristics of the research topic or the individual preferences of the re-
searcher. It should also be underpinned by a sound philosophical research
paradigm and must be suitable to answer the research questions and fulfil
the objectives defined in the previous chapter. Sub-chapter 2 presents the
paradigm that drives the research. Detailed descriptions of the literature
review process (sub-chapter 2.2) and of the multi-stage comparative case
study (sub-chapter 2.3) introduce the major techniques employed for the
rest of this study.
Figure 2.1 summarizes the logical development of the research methodol-

ogy, as described before.
Figure 2.1 – Overview of research methodology development3
3

https://doi.org/10.1007/978-3-658-27582-2_2
6 Methodology
2.1 Development of research design
This study adopts the pragmatic view as the research paradigm. Pragma-
tism incorporates elements of both post-positivism and constructivism by
accepting the co-existence of subjective, inter-subjective, and objective re-
ality (Antwi & Hamza, 2015, p. 223). Research problems in management
sciences, especially emerging phenomena, are rarely ever completely
quantifiable or solely based on human interactions (Wynn & Williams,
2012, p. 789). Hence, pragmatists emphasize on the consequences of ac-
tions that cause real-world problems, which are addressed in practical re-
search (Creswell, 2013, p. 6).
The four research questions defined in chapter 1 are clearly derived from
a practical problem – the vagueness of BDO benefits for airline NP. The
objective is not only to understand the causal effects of big data, mani-
fested in the expected economic impact, but also to analyze the current
processes and beliefs followed by airline network planners. Furthermore,
the recency of big data calls for a holistic investigation, as there may not
be sufficient material available to answer the research questions empiri-
cally. The combination of (1) the practical research focus, (2) an emergent
research topic, and (3) the complexity of the four research questions re-
quire the adoption of a pragmatist perspective for this research project.
Pragmatists advocate for an appropriate mix of qualitative and quantitative

methods, which best suits the research objectives (Creswell, 2013, p. 213).
Table 2.1 summarizes the relevant decision-making criteria for qualitative,
quantitative, or mixed-method approaches.
Development of research design 7
Table 2.1 – Decision-making criteria for research approach4
Criteria Quantitative approach Qualitative ap- Mixed methods ap-

proach proach
Research Confirmatory Exploratory Both
questions
Purpose Theory testing Theory building Solving practical
problems
Context  High-quality data avail-  No previous  Quantitative
able / possible to col- research on and qualitative
lect (including suffi- topic data available
ciently large N)  No or insuffi-  Emergent re-
 Comparable research cient data search topic
objects available
 Experimental set-up  Research
possible objects not
comparable
Applying these decision-making criteria, a qualitative-dominant approach

with quantitative elements appears most appropriate for this study. While
research question (RQ) 1 and RQ 2 are purely exploratory, RQ 3 and RQ
4 have a confirmatory element complementing the explanatory character
of the questions. The motivation behind conducting this study originates
from a practical problem, which should be analyzed and solved from a sci-
entific perspective. Theory building or testing is not a research objective of
this study. Furthermore, “big data” is an emerging research topic that is
increasingly garnering interest in the fields of science and practice. The
absent large-scale implementation of big data in airline NP impedes a
purely quantitative study. However, airlines are sufficiently comparable, in
that the research would benefit from a small-scale quantitative element in
order to evaluate the usefulness and potential of big data. Furthermore, a
sequential research setup is the most appropriate, since the research
questions need to be answered consecutively.
4
Own illustration based on Creswell (2013, pp. 157–211).
8 Methodology
After selecting a general research design, the most appropriate research

methods need to be determined. Yin (2017, p. 8) proposed three decision-
making criteria for this: the nature of the research question, the need for
behavioral control, and the focus on present or past events (see Table 2.2).
Since none of the research questions require behavioral control and all of
them focus on contemporary events, experiments and historic studies can
be ruled out. This leaves literature review, case study, and survey as po-
tential research methods. All these methods are suited for a qualitative-
dominant research design and can contribute towards answering the re-
search questions.
Table 2.2 – Comparison of research methods5
Research Qualita- Form of research Requires be- Focuses on

method tive/quantita- question havioral control contemporary
tive of events? events?
Experi- Quantitative How, why? Yes Yes
ment
Survey Quantitative Who, what, where, No Yes
how many, how
much?
Literature Qualitative How, why, who, No Yes/No
analysis what, where, how
many, how much?
Historic Qualitative How, why? No No
study
Case Qualitative How, why? No Yes
Study
RQ 1 is “How can ‘big data’ be defined for network planning in airlines?”

considering the existing research and theories. A structured analysis of the
literature is the most suited research method for this, as both recent and
historic literature can be explored. Although RQ 2 and RQ 3 start with
“what” and “which” respectively, both questions have an exploratory char-
acter calling for a predominantly qualitative method. For this purpose, a
5
Own illustration following Yin (2017, p. 8).
Development of research design 9
qualitative comparative case study, featuring a quantitative small-scale

survey, seems to be the best suited to answer the questions. The case
study is also extended to finally answer RQ 4 in the final research phase.
Figure 2.2 – Four-phase research design6
Figure 2.2 schematizes the four phases of the research design. The de-
tailed methodology, including the chosen techniques, are presented in the
following sub-chapters. The set-up and execution of the structured litera-
ture review is explained in section 2.2. Section 2.3 is dedicated to explain-
ing the case study set-up, including the data collection and survey design
methodology. Chapters 3–6 of this study present the theories and findings
from each respective research question (see Figure 2.2).
Ensuring high research quality is a key objective of the research design

phase. Cooper and Schindler (2000, p. 211) identify “validity” and “reliabil-
ity” as key quality measures for research in management. Reliability is a
measure to “ensure consistent results” (Cooper & Schindler, 2000, p. 215)
within and across research projects. Validity indicates the correct meas-
urement and analysis of the intended phenomena. Accordingly, Yin (2017,
p. 41) proposed reliability and three validity aspects as measures for re-
search quality, namely construct validity, internal validity, and external va-
lidity. Construct validity describes the ability of a research method or, more
6
10 Methodology
specifically, a research construct such as a survey to analyze the right re-

search subject. Internal validity minimizes systematic errors and biases in
the research process, whereas external validity ensures the generalizabil-
ity of research results.
Figure 2.3 – Tactics to ensure the quality criteria for selected research methods7
The structured literature review process proposed by Vom Brocke et al.

(2009, p. 2211–2212) aims to ensure that these quality measures are met,
arguing that literature reviews should be considered a self-contained re-
search method and thus should be subject to the very same quality
measures. Similarly, Yin (2017, p. 41) has recommended tactics for de-
signing rigorous case studies, following the aforementioned research qual-
ity criteria. For surveys (which are part of the comparative case study),
Creswell (2013, p. 155) and Cooper and Schindler (2000, p. 216) have
suggested various tactics to guarantee reliability and validity in the re-
search process. Figure 2.3 summarizes the proposed tactics to meet the
7
Own illustration following Yin (2017, p. 41), Creswell (2013, pp. 155–157), Cooper and
Schindler (2000, pp. 215–218), and Vom Brocke et al. (2009, p. 2212).
Literature review 11
discussed quality criteria. The tactics are implemented in the research de-
sign and will be further described in the sections dedicated to them.
2.2 Literature review
In order to build on established scientific theories and highlight the purpose

of a research project, the literature review is the fundamental starting point
to connect the research to the existing body of knowledge by uncovering
relevant research gaps (Vom Brocke et al., 2009, p. 2206).
2.2.1 Design of a structured literature review process
Paré et al. (2015, pp. 185–189) identified nine literature review archetypes
based on the research objective and the dominant research methods of a
project8. According to this classification, a realist review is most suited for
this study, since the research questions are already defined, and the over-
all research strategy follows a qualitative approach with quantitative ele-
ments. A realist review should follow an iterative and purposive search
strategy and include both conceptual and empirical works. The studies can
be down-selected based on pre-defined formal criteria (e.g., journal quality
or publication date) or qualitative criteria (e.g., abstract reviews).
The structured literature review process proposed by Vom Brocke et al.

(2009) includes most of the realist review criteria and is specifically in-
tended for multidisciplinary research, with strong rooting in the information
systems (IS) research domain.
The starting point of a structured literature review is the identification of

relevant search terms and knowledge containers to assure external valid-
ity. In an academic context, this means selecting the appropriate publica-
8
Please refer to Appendix 1 for a complete overview of the typology of literature review.
12 Methodology
tions and databases that pertain to the relevant literature. To define appro-
priate search terms, Vom Brocke et al. (2009, p. 2214) proposed a struc-
tured decomposition of titles or research questions as well as the inclusion
of intended theories and research paradigms. This step is crucial to guar-
antee construct validity, i.e., to ensure that the search delivers the intended
results. These activities comprise Step 1 in Figure 2.4 and are presented
in detail in the next section.
Figure 2.4 – Structured literature review process9
Step 2 entails the actual keyword search in the selected databases and
publications. The results are evaluated through a three-step approach.
First, the title is evaluated to exclude literature on completely unrelated re-
search fields (e.g., medicine) or in languages other than English, German,
or Spanish. The second iteration is a qualitative abstract screening to ex-
clude thematically irrelevant contributions (e.g., literature on air traffic con-
trol optimization). The full-text evaluation is the last iteration to confirm or
refute the actual relevance of the contribution (e.g., non-content editorials
9
Own illustration following Vom Brocke et al. (2009).
or book reviews). These steps need to be followed strictly for all individual
searches to ensure internal validity of the literature review.
Finally, Step 3 aims to identify additional literature based on a for-

ward/backward search, using the most relevant contributions identified in
Step 2. The evaluation iterations in Step 3 are similar to those of Step 2.
2.2.2 Identification of keywords, databases, and journals
The title of this study can be deconstructed as three major elements. “Big
data” describes the phenomenon that is the primary research subject. The
phenomenon is examined for a specific industry, namely “airlines,” and for
a specific business process, the “strategic network planning.” The two guid-
ing theories for this research are the resource-based view of the firm and
network theory10, which have also been included in the keywords. Syno-
nyms and adjacent search terms complement the extracted keyword in
each area (see Table 2.3).
Table 2.3 – Keyword extraction for the literature review11
Extracted keyword Source Type Synonyms and related

search terms
“big data” Title Primary research “data management”
subject “business intelligence”
“business analytics”
“airline” Title Industry “aviation”
“logistics”
“transportation”
“strategic network Title Process “strategic planning”
planning” “network planning”
“resource-based Theoretic Theory “RBV”
view” concept
“network theory” Theoretic Theory -
concept
10
Compare sub-chapter 3 for a detailed discussion on theory selection.
11
14 Methodology
The combination of the four keyword categories results in 15 different

search queries when synonyms and related search terms are included us-
ing the Boolean string “OR”12. Appendix 4 displays the aforementioned
combinations.
For the identification of the relevant journals and databases, the primary
research domains for this study must be identified first. The research do-
mains of VHB-JOURQUAL 3 ranking (VHB, 2015), curated by the German
Academic Association for Business Research (VHB), are evaluated with a
scoring model based on the relative search results for each of the extracted
keywords (see Appendix 3 for the detailed scoring results).
Logistics, operations research (OR), strategic management, and infor-

mation systems are the most relevant domains for this research. General
management science, international management, marketing, organization
science, production management, and technology, innovation & entrepre-
neurship are other domains of medium importance.
The combination of three platforms, EbscoHost, Elsevier, and Spring-

erLINK, provides full-text access to the 39 most important journals13 in the
relevant research domains (see Appendix 5 for a graphical summary).
Google Scholar complements the three access points as a meta-search
engine, as it also includes scientific reports, conference papers, and books
that are not accessible via the scientific databases.
12
Example: The complete Boolean Phrase for combination 4 would read as [“big data” OR
“data management” OR “business intelligence” OR “business analytics”] AND [“airline” OR
“aviation” OR “logistics” OR “transportation”] AND [“strategic network planning” OR “strategic
planning” OR “network planning”] AND [“resource-based view” OR “RBV” OR “network the-
ory”].
13
Based on VHB-JOURQUAL 3 ranking VHB (2015): Journals ranked A+ from all the relevant
research domains plus journals ranked A from logistics, operations research, strategic man-
agement and information systems. Appendix 4 provides a complete list of the 39 journals.
2.2.3 Results of the structured keyword search
During the initial search, the parameters were set to a full-text search and
limited to academic articles, conference proceedings, and books14. For the
initial search with one keyword only (combinations 1a to 1d in Appendix 2),
the search terms for object (big data), industry (airline), and theories (re-
source-based view OR network theory) resulted in at least 10,000 hits at
each access point, while the keyword chosen for the relevant process (stra-
tegic network planning) produced less than 1000 hits at all access points.
Therefore, the keyword selection for processes was extended with the two
decompositions “strategic planning” and “network planning.”
The initial round of the structured literature search followed the keyword
combination scheme presented in Appendix 2 and entailed 15 individual
searches. The search was commenced with the full combination of all rel-
evant keywords (4a) to identify the most relevant literature first. Subse-
quently, the search terms were relaxed by combining only three (3a–3d) or
two keywords (2a–2f). Finally, the single keyword search was repeated
(1a–1d). The titles and abstracts of the first 400 results of each search
were screened to identify the relevant works. Duplicates were only counted
in the first appearance. The detailed search results are listed in Appendix 6.
In total, 610 relevant studies were identified for the full-text evaluation, out
of which 207 passed the full-text review and were used for forward/back-
ward search. The main factor for rejection of a study was focus on different
research topics (158), such as air traffic control or communication network
planning. Other major reasons for rejection were lack of originality of re-
search content (86) and non-scientific research publications (71). The for-
ward/backward search was conducted iteratively during the full-text review
14
In Google Scholar, patents and citations were excluded.
16 Methodology
process and yielded 136 additional publications. The relevant body of liter-
ature for the theoretical aspect of this study consequently comprises 343
publications, as depicted in Figure 2.5.
Figure 2.5 – Build-up of a relevant body of literature15
Out of the 343 publications, 290 are articles from peer-reviewed academic
journals. Among these, 83 articles have been published in unranked pub-
lications, such as industry-specific journals and published conference pro-
ceedings. Appendix 7 summarizes the most frequent publications for each
VHB ranking category. Thus, most of the publications are from information
systems, operations research, and strategic management research do-
mains. The 53 remaining sources are mostly books (31), book chapters
(11), and research reports (4).
15
2.2.4 Description of the research gap
Only two useful contributions could be determined from the combination of

all keywords (combination code 4a in Appendix 6). However, these two are
general textbooks (Bazargan, 2016; Shaw, 2016) that do not discuss air-
lines, network planning, big data, and any of the theories in the same con-
text but in different chapters of the books.
The keyword combination 3a of airlines, big data, and strategic network

planning is either researched from an air traffic control perspective (Ayhan
et al., 2013) or network planning is only considered a potential application
of airline big data (Chen et al., 2017; Larsen, 2013). In fact, there is no
academic contribution that has explicitly analyzed the role of big data in
airline network planning.
Combination 3b (airlines, big data, and resource-based view or network

theory) yields research that focuses on the role of big data in the resource-
based view, where airlines serve as general examples (e.g., Erevelles, Fu-
kawa, & Swayne, 2016). In turn, network theory-centered publications
mention big data as a potential means to improve network optimization in
airlines (e.g., Akartunalı et al., 2013). None of the identified publications
elaborate the examples to an extent that can broaden the understanding
of big data opportunities in airlines.
The combination of big data, strategic network planning, and resource-

based view or network theory (combination 3c) yields useful studies on the
strategic impact of big data from a resource-based perspective (e.g.,
Grover et al., 2018)16. However, these contributions are industry-agnostic
and do not consider airline-specific network planning.
16
The resulting literature is discussed in detail in section 3.7.3.
18 Methodology
Combination 3d, yields by far the most search results from all the third-
degree keyword searches17. A rich body of literature exists on airline net-
work planning, most of which originates from operations research (see sub-
chapter 3.5 for a detailed discussion). However, most of the studies focus
on optimization methods and not on used data.
Research that stem from the lower-degree keyword searches (1a–d and
2a–f) emphasize on specific aspects of big data, airlines, strategic network
planning, and the theoretical foundations. All of the analyzed studies lack
an integrated perspective of big data in strategic network planning for air-
lines, which presents a significant gap that this study aims to bridge.
2.3 Comparative case study
The actual case study design follows the method proposed by Eisenhardt
(1989, p. 533), depicted in Table 2.4.
Table 2.4 – The case study method adopted from Eisenhardt (1989)
Step Activity Corresponding sec-

tions and chapters
Getting started Definition of research questions 1.2
Selection of research method 2
Selection of case study type 2.3.1
Selecting cases Theoretical sampling 2.3.2
Crafting Instruments & Data collection and analysis 0
Protocols techniques
Entering the field Data collection Chapter 4 and 5
Analyzing data Within case analysis Chapter 4, 5, and 6
Cross-case analysis
Shaping hypothesis Replication of logic Chapter 4, 5, and 618
Enfolding literature Comparison with supporting & Chapter 3 and 7
conflicting literature
Reaching closure Distillation of newly generated Chapter 8
knowledge
17
2’453 compared to less than 1’000 for all other 3rd-degree combinations (see Appendix 6).
18
Hypothesis on RQs developed within the respective chapter.
Comparative case study 19
The first step “Getting started” also includes the work required before de-
termining the case study design, which has been explained in the previous
sections of this chapter. This section starts with the selection of the case
study type (section 2.3.1), explains the theoretical case sampling (section
2.3.2), and presents the data collection and analysis techniques (section
2.3.3). Section 2.3.4 reviews the research quality criteria for the case study
research.
2.3.1 Selection of the case study type
The first step in the actual case study design is determining the most ap-
propriate case study type. Eisenhardt and Graebner (2007, p. 27) have
proposed two generic case study types, namely single case studies and
multiple case studies. While single case studies are the most suited to de-
scribe the existence of a specific phenomenon, multiple case studies have
more explanatory power as they enable the researcher to compare the
same phenomenon within multiple settings. Baxter and Jack (2008, p. 550)
compared the design choice with the set-up of an experimental research
design. If the purpose is to discover a phenomenon, a single experiment
or case study is sufficient. If the properties of the phenomenon need to be
analyzed, multiple experiments – or multiple cases – are required. Yin
(2017, p. 47) has also stated that although multiple case studies have more
explanatory power, resource constraints can sometimes favor a single
case study setup.
The multiple or comparative case study method is the most appropriate

design for this study for three reasons. First, the phenomenon “big data”
already exists and is widely used for other airline functions, such as pricing
and revenue management (P&RM). Thus, this research aims not to dis-
cover a completely new phenomenon but explain its application and use-
fulness in network planning. Second, the NP process in airlines differs
widely depending on the chosen business model. A single case study
would thus limit the generalizability of results to other airlines and thereby
reduce external validity. Third, multiple airlines showed keen interest in
20 Methodology
participating in this study, so the availability of cases was not a limiting

factor.
2.3.2 Case sampling
The case study guidelines developed by Eisenhardt and Graebner (2007)

and refined by Yin (2017) were followed when selecting cases among the
interested airlines and other research partners. These other research part-
ners included potential providers of big data and system developers for
airline NP software. While the airlines form the “core” comparative case
study, data providers and system developers contribute with a different and
very specific perspective on the “data source” topic.
Airline case sampling
Theoretical sampling is the most appropriate sampling approach for multi-

ple case studies, as it ensures generalizability and thus the external validity
of the research (Eisenhardt & Graebner, 2007, p. 27; Yin, 2017, p. 43). Ei-
senhardt (1989, p. 537) recommends defining relevant selection criteria
that allow the systematic comparison of a researched phenomenon across
cases.
For the airline cases, four relevant selection criteria were identified. Busi-
ness model is the most frequently used classification criterion in airline
management literature (Wittmer & Bieger, 2011, p. 31), and it will also be
used as the primary classification criterion in this study. However, other
influencing factors, such as fleet size, network type and route structure
have a considerable influence on the structure of the network process
(Goedeking, 2010, vii) and thus on the BDOs needed for its improvement.
For each selection criterion, three to four possible specifications were de-
fined (see Table 2.5).
Table 2.5 – Specification criteria for case study airlines19
Criterion Specifications
Business model  Full-service carrier (FSC)
 Low-cost carriers (LCC)
 Scheduled-charter airline
(SCA)
 Cargo airline (CAR)
Fleet size  Large (> 100 aircraft)
 Mid-sized (30–100 aircraft)
 Small (< 30 aircraft)
Network type  Multi-hub
 Single-hub
 Point-to-point
Route structure  Mostly long-haul
 Mostly short-haul
 Mixed
Four different business models are considered for the case sampling (Ster-
zenbach, Conrady, & Fichert, 2013, p. 225). Full-service or legacy carriers
(FSCs) are mostly developed from former state-owned airlines, and they
target a wide spectrum of passengers. They usually offer a premium prod-
uct and additional services, such as a frequent flyer program. Low-cost
carrier (LCCs) are the alternative draft to this model, which offers a product
with the bare minimum (“no frills”). Scheduled charter airlines (SCAs) or
“leisure carriers” focus on touristic routes and sell large seat contingents to
travel operators. Cargo airlines (CARs) focus on transporting freight only.
The business models are discussed in full details in Section 3.2.3.
Fleet size serves as a proxy for airline size, since that data is more easily
available than financial data or passenger data20. It is distinguished in three
levels, ranging from small (< 30 airplanes) to large (> 100 airplanes). Tra-
ditionally, network types are distinguished in hub-and-spoke (HS) and
19
20
Financial data for privately owned airlines is not published often; passenger data is availa-
ble consistently only for IATA member airlines.
22 Methodology
point-to-point (PP) models. However, with increased merger and acquisi-

tion activity in the airline domain, many airlines have developed multi-hub
models, leveraging the hubs of the merged airlines (Burghouwt, 2007).
Thus, point-to-point, single-hub, and multi-hub network types are distin-
guished. Route structures can be systemized by the route type. Some air-
lines serve predominantly long-haul routes (over 4,500 km), while others
serve predominantly short-haul routes (under 4,500 km), and some have
a mixed route structure.
Each criterion specification should be represented multiple times in the

case study group. Table 2.6 collates information on the nine participating
airlines, which have been anonymized for this study. All airlines are based
in Europe and participated in all the interview rounds. This ensures validity
not only across the case sample but also in time throughout the research
project (Yin, 2017, p. 43).
Table 2.6 – Airline case study participant overview21
Airline Business model Fleet size22 Network type Route type

FSC 1 Full-Service Carrier Large Multi-hub Mixed
FSC 2 Full-Service Carrier Medium Single-hub Mixed
LCC 1 Low-Cost Carrier Medium Point-to-point Short-haul
LCC 2 Low-Cost Carrier Large Point-to-point Short-haul
SCA 1 Scheduled Charter Medium Point-to-point Mixed
Airline
SCA 2 Scheduled Charter Small Point-to-point Short-haul
Airline
CAR 1 Cargo Airline Small Single-hub Long-haul
CAR 2 Cargo Airline Small Multi-hub Long-haul
CAR 3 Cargo Airline Small Single-hub Long-haul
In addition to the nine airlines, a large system development company and

ten potential data providers agreed to support the research project and
21
22
Assessed on October 4, 2016 with data from airfleets.net (2016).
complement the data collected in the comparative case study. They are
not part of the actual case study group but serve as tri-angulation partners
who help define the feasibility of potential big data opportunities.
2.3.3 Data collection and analysis techniques
The data collection process is structured as three phases, in compliance

with the research process presented in
Figure 2.1: the first phase aims to assess the status quo of NP in airlines;
the second phase intends to discover and qualitatively evaluate potential
BDOs; and the last phase aims to develop an estimation methodology for
the financial impact of investment in big data.
A double tri-angulation principle is followed throughout the study. This prin-

ciple is particularly suited for research concepts with small sample sizes
that do not allow for statistical significance testing (Berg, 2004, p. 51). The
first triangulation calls for multiple data collection techniques, whereas the
second triangulation applies to the sources.
Data from the participating airlines is collected through interviews, surveys,

and internal documents to satisfy the first tri-angulation dimension. Repre-
sentatives of system developers and data providers are also included in
the interview process to establish a tri-angulation of sources. Figure 2.6
schematizes the double tri-angulation principle.
24 Methodology
Figure 2.6 – Double tri-angulation approach23
Exploratory phase – status-quo assessment
The initial data collection in the status-quo assessment is partly exploratory

and partly explanatory. Since NP processes and potential organizational
set-ups are well-researched (refer to sub-chapter 3.5), a semi-standard-
ized questionnaire has been developed to assess these topics. In contrast,
the use of data for airline NP is not well described in literature, which re-
quires an exploratory set-up (Creswell, 2013, p. 211). In an exploratory set-
up, non-standardized questions are used to extract common themes via
coding the interview transcript, which can then be used to conduct a sec-
ond standardized interview to ensure holistic coverage and comparability
across the case study participants (Berg, 2004, p. 105). The status quo
assessment starts with non-standardized questions on the use of current
data sources in NP to extract data types by coding, which are then reiter-
ated to the interviewees in the form of a standardized online survey. Figure
23
Source: Own illustration. following Berg (2004).
2.7 summarizes the data collection and analysis process of the status quo
assessment phase, whose results are presented in chapter 4.
Figure 2.7 – Data collection and analysis techniques of status-quo assessment24
A structural coding technique is used for the analysis of the non-standard-

ized questions for the assessment of current data and information technol-
ogy (IT) system use. Structural coding is especially useful if a set of poten-
tial outcomes is analyzed across a group of multiple participants (Saldaña,
2016, p. 98). The coding object are data sources, and the initial set of
codes is derived from the data sources referenced in airline network plan-
ning literature (section 1). In addition, there is an open code – “new” – for
all data sources mentioned by airline network planners, which have not
been derived from literature.
There might be the case that an airline has not mentioned a data type in
response to the open question, even though it is being used. In order to
capture the complete picture of the data sources and IT systems used, the
first interview is complemented with a structured questionnaire that consti-
tutes all coded items, including the data types labelled as “new.” The struc-
tured questionnaire includes three questions per data source. The first
question entails a simple confirmation on whether a specific data type is
used (yes/no). The second and third questions evaluate the usefulness
and reliability of a data source. Berg (2004, pp. 105–108) recommends the
use of four-point rating scales for the measurement of usefulness. A clas-
sical Likert scale is not appropriate here, as the questionnaire does not
24
26 Methodology
capture personal attitudes towards an item but intends to rate them objec-
tively. The questionnaire for the status quo interviews is attached in Ap-
pendix 8 and the subsequent survey in Appendix 9.
Refinement phase – qualitative big data opportunity evaluation
Before the actual data collection, nine distinct information needs are de-
rived from the status quo assessment and the literature review on the NP
process. To satisfy these information needs, 23 potential BDOs are devel-
oped from three different sources, namely the initial status quo interviews,
a press review, and the analyzed body of literature.
These 23 BDOs are then presented and explained to the airline network
planners in the case study group. After the presentation, a standardized
online survey (see Appendix 10) asked for the perceived benefit potential
for the airline-specific NP. To increase granularity in the responses, a six-
point scale was chosen (Berg, 2004, pp. 105–108).
Figure 2.8 – Data collection and analysis techniques of BDO evaluation25
In addition, a feasibility scoring scheme is developed based on the existing

literature (see section 5.3.4). The individual dimension scores are aggre-
25
gated for each BDO in order to calculate a total feasibility score. The aver-
age perceived benefit potential and the total feasibility score of each BDO
are then plotted on a two-dimensional matrix. The 12 BDOs among these
with the highest benefit potential and feasibility ratio are selected for further
analysis.
Synthesis phase – Quantification methodology for BDO benefits
The last interview round was related to the key performance indicators
(KPIs), information on which was collected through the semi-standardized
interviews in the exploratory phase. The airline network planners were
asked to evaluate the impact of each of the 12 shortlisted BDOs on the
primary and secondary KPIs of their airline. Similar to the initial benefit
evaluation in the refinement phase, a six-point assessment scale was cho-
sen. The results of the structured interview were then used to inform the
development of the business case scheme to quantify the financial benefits
of the shortlisted BDOs.
2.3.4 Research quality assurance
In accordance with the data quality criteria derived from the literature in
sub-chapter 2, displayed in Figure 2.3, specific tactics have been deployed
to ensure the reliability and validity of the comparative case study.
A case study database, which includes the case study protocol, has been
created to ensure research reliability. This database contains all the inter-
view transcripts as well as the associated analyses26. The interview tran-
scripts and the report summaries were sent to the interviewees for verifi-
cation to ensure construct validity. Furthermore, specific sections of the
interviews were verified with system developers and data providers to tri-
angulate information.
26
Due to data confidentiality, the complete transcripts cannot be appended to this thesis, as
the doctoral thesis will be publicly available. The anonymized case reports can be requested
from the author.
28 Methodology
The theoretical sampling of the comparative case study has a positive in-
fluence on both internal and external validity. The comparison of different
cases enabled pattern matching, helping identify potential biases. Logical
models derived from the literature could also be tested with multiple cases,
since all sampling criteria were present in at least two cases. The research
design applied the same logic to all cases to ensure external validity. This
means that all interview protocols were standardized across cases as well
as across supporting sources.
3 Theoretical foundation
3.1 Development of a theoretical concept
When following a qualitative research design, as developed in chapter 2, it

is vital to ground the research on a relevant overarching theory (Creswell,
2013, p. 63). The research topic “big data to improve strategic network
planning in airlines” is situated on the intersection of four different research
domains – information systems, logistics, operations research, and strate-
gic management. The overarching theory needs to be applicable to all do-
mains to facilitate the creation of an overarching theoretical concept, alt-
hough additional technical theories can be used to explain certain phenom-
ena or research objects (Wynn & Williams, 2012, p. 798).
There is no literature on the prevalence of certain theories across the four

relevant research domains, but Defee et al. (2010) studied the use of over
80 theories, clustered in 13 theory groups, in the logistics research domain.
They found that competition theory, microeconomic theory, and system
theory are the most frequently used groups of theories. Each of these
groups comprises several specific theories, from which only the two most
frequently used theories are considered in this study.
Figure 3.1 compares the search results of the combinations of the consid-
ered theories herein with research domains and research topics. It identi-
fies the most appropriate specific theories for this study. It is hereby as-
sumed that a large number of search results within each domain or topic
indicates whether a theory is widely used and well established.
The most widely cited theories, across research domains and topics, are
the resource-based view, game theory, and network theory. These theo-
ries are all well-established in the three research domains (Logistics, OR,
and IS) and two research topics (big data and airlines) being studied. How-
ever, the resource-based view (RBV) is the only theory which has a strong
influence on strategic management research, and network theory is the
only well-established theory for strategic NP.

https://doi.org/10.1007/978-3-658-27582-2_3
Figure 3.1 – Evaluation of the candidate theories27
Since RBV is well-established across all relevant research domains, it is

used as an overarching theory to develop the theoretical concept. Network
theory is used to specifically underpin strategic NP. Figure 3.2 schematizes
the theoretical concept and the resulting structure of this chapter. The sec-
tions in this chapter serve two purposes. They either define the research
constructs, which are essential for the practical research presented in
chapters 4 to 6, or they provide the theoretical background on theories and
research objects.
Sub-chapters 3.2 and 3.6 define and describe the research objects “air-
lines” and “big data” respectively. After an introduction to network theory in
sub-chapter 3.3, the third research object “strategic network planning” is
27
Own illustration; the selection of candidate theories based on Defee et al. (2010); the circles
reflect a relative number of search results from EBSCOhost search combining each theory
with research area and research topics, e.g. (“resource-based view” AND “logistics”). Abso-
lute search results in title and abstract have been recorded and standardized for each re-
search domain/topic. A full circle represents >70% of the highest search result, an empty
circle <10% of the highest search result.
Airlines and their business models 31
established in sub-chapters 3.4 (airline networks) and 3.5 (network plan-

ning). Sub-chapter 3.7 finally integrates the research objects within the the-
oretical framework developed in the first sub-chapter of this chapter.
Figure 3.2 – Structure and purpose of the theoretical foundation chapter28
3.2 Airlines and their business models
This sub-chapter defines the research topic, airlines, for clarity in the fur-
ther course of this study. Section 3.2.1 delimits the airline industry within
the wider sphere of aviation systems. Developments and recent trends in
28
the airline industry are summarized in section 3.2.2. The last sections pre-
sent different business models followed by airlines (3.2.3) and the most
important business processes (3.2.4).
3.2.1 The airline industry
Airlines are one group of actors in the broader aviation system. Wittmer
and Vespermann (2011, pp. 39–40) have described the aviation system as
a classic economic demand and supply system (see Figure 3.3).
Figure 3.3 – The aviation system from a supply-demand perspective29
The supply side of the system contains the aircraft manufacturers, airlines,
airports, and aviation service providers. Aircraft manufacturers produce the
movable transportation assets, whereas airports provide the immobile in-
frastructure assets. The broader industry of aircraft manufacturing is also
referred to as the aerospace industry, comprising end-manufacturers, such
29
Source: Own illustration. adapted from Wittmer and Vespermann (2011, p. 40).
as Airbus and Boeing, as well as numerous component suppliers

(Wensveen, 2015, p. 18). Airlines use mobile and immobile assets to offer
a transportation service for passengers or goods. Wensveen (2015,
p. 164) defined airlines as “certified and scheduled air carriers” in contrast
to uncertified general aviation actors. Companies operating in the general
aviation field do not offer scheduled flights (e.g., pure charter carriers) or
are not certified for commercial transportation services (e.g., flying
schools). The supply system is facilitated by a multitude of aviation service
providers30, who enable it to function properly.
The aviation demand system mostly comprises individual passengers, who

can be categorized as individual leisure customers or business customers
(Wittmer & Vespermann, 2011, p. 40). Tour operators typically block large
capacity contingents on scheduled flights, which are then sold on their own
economic responsibility. In the air cargo industry, most demand stems from
freight forwarders, which bundle logistic services from different transporta-
tion modes, such as shipping, air cargo, and trucking, for the actual end-
customer. Both these systems are controlled by official regulators, who de-
fine the associated security standards, traffic, and passenger rights (Witt-
mer & Vespermann, 2011, p. 41).
The main task of NP is spatial and temporal asset allocation, which is a

core competence of airlines (Sterzenbach et al., 2013, p. 301), which
henceforth are the primary research focus. Since NP is a very complex
process, it relies on multiple data and system providers to cover the need
for information. These providers are part of the aviation service providers,
which are the secondary focus of this study.
30
Examples: air traffic control, aviation system providers, aviation data providers, ground han-
dling, catering, maintenance providers, aircraft leasing companies.
3.2.2 Development and recent trends in the airline industry
Historically, the development of the airline industry has been heavily influ-
enced by technological progress and changes in regulation. Most authors
such as Doganis (2010, pp. 25–63) and Gillen (2006, pp. 367–369) distin-
guish between a regulated and a deregulated era.
Until 1978, all major aviation markets were heavily regulated. All routes
needed governmental approval in specifying the permitted frequency, seat
capacity, and even the ticket price, which was calculated based on the dis-
tance and service components (Doganis, 2010, p. 36). This led to the cre-
ation of linear networks, which are economically inferior to the point-to-
point or hub-and-spoke networks (see section 3.4.2 for a detailed discus-
sion) in a free market.
The deregulation of the air transport system started with the Airline Dereg-
ulation Act in the United States in 1978 which was completely implemented
in 1985 (Gillen & Morrison, 2005, p. 162). This deregulation had a severe
impact on the airline business models and networks. The “open sky” poli-
cies31 allowed the establishment of more efficient network types, namely
the point-to-point and hub-and-spoke networks (see section 3.4.2). The
abolition of pre-defined fares led to a sharp decrease in airfare, hence in-
creasing cost pressure on the airlines (Doganis, 2010, p. 47). This sparked
both the consolidation of legacy carriers and the development of new busi-
ness models, most notably of the low-cost carriers (Gillen, 2006, pp. 370–
372).
As cross-border mergers had failed to deliver the expected synergies in

the past (Schosser & Wittmer, 2015), airlines sought alternative means of
closer cooperation in the form of strategic alliances to leverage network
effects and revenue synergies (Gillen, 2006, p. 371). As of 2016, the three
31
Airlines can connect whichever airport pair they wish to, without regulatory approval; how-
ever, limited airport slots hinder the unlimited choice of connections.
major alliances (Star Alliance, Oneworld, and Skyteam) control over 60%
of the total available passenger airline capacity (Statista, 2018).
3.2.3 Airline business models
Business models are used as the most common classification factors for
airlines, since they constitute the “fundamental differences […], i.e. the ser-
vice level offered, the regional reach, and the main function of an airline”
(Wittmer & Bieger, 2011, p. 31). Therefore, this study also adopts airline
business models as the main classification criterion.
Scholars have distinguished up to six different business models for airlines,

and a continuous space of “hybrid” business models. Prior to the deregu-
lation, airline business models could be categorized as four archetypes
(Sterzenbach et al., 2013, p. 225): first, legacy or full-service carriers
(FSCs), which served a broad, regulated network and targeted all passen-
ger types. The second archetype was leisure or scheduled charter airlines
(SCAs), which served point-to-point flights to holiday destinations, thus
mainly targeting tourists. Third, regional carriers predominantly served
business customers on low-volume routes. Finally, cargo airlines (CARs)
are easily classifiable for transporting cargo only. It is important to note that
even today, some airlines with traditional business models are not profit
maximizing companies. Pompl (2007, p. 99) noted that national prestige
(especially flag carriers), provision of public services (mostly for regional
airlines), and the stimulation of economic development are political objec-
tives that often outweigh expected economic benefits.
After the de-regulation wave in the 1980s and 1990s, various new business
models emerged. The most popular new business model are low-cost car-
riers (LCCs) that serve point-to-point networks with a very low cost struc-
ture (Shaw, 2016, pp. 100–101). At the other end of the service spectrum,
super-differentiated airlines emerged to serve high-value business routes
with a premium product as a cheaper alternative for private jets (Bieger &
Wittmer, 2011, p. 98).
In recent years, certain authors have noted the convergence of business

models into individual “hybrid” business models, which feature the charac-
teristics of more than one business model (Lohmann & Koo, 2013, p. 7).
The conceptual and empirical studies by Daft and Albers (2013, 2015)
have supported this claim. Table 3.1 provides an overview of the differen-
tiation schemes of airline business models in the literature.
Table 3.1 – Airline business model classifications in the literature32
Author Full- Sched- Re- Cargo Low- Super- “Hy-

ser- uled gional airline cost differ- brid”
vice/ charter/ carrier (CAR) carrier ent- busi-
legacy leisure (LCC) iated ness
carrier airline air- mod-
(FSC) (SCA) lines els
Bieger and x x x x x
Wittmer
(2011)
Daft and Al- x
bers (2015)
Doganis x x x x
(2010)
Gillen x x x x x
(2006)
Whyte and x x x x x
Lohmann
(2017)
Pompl x x x x x
(2007)
Shaw x x x x x
(2016)
Sterzen- x x x x x
bach et al.
(2013)
Wensveen x x x
(2015)
32
Source: Own illustration. based on literature mentioned in the table.
Full-service carriers (FSC)
Full-service carriers or legacy carriers mostly emerged from state-owned

national flag carriers or private airline ventures immediately after the de-
regulation (Whyte & Lohmann, 2017, p. 110). On the basis of their previ-
ously protected home market, FSCs usually enjoy a strong market position
due to excellent slot33 portfolios and a huge customer base. At the same
time, they suffer from a high-cost base, established during times of regula-
tion when cost minimization was not a priority for many airlines (Gillen,
2006, p. 370). After the deregulation, most FSCs built efficient hub-and-
spoke network structures with a combination of short-haul and long-haul
routes. Their fleets are often heterogeneous and comprise multiple aircraft
types from different manufacturers (Sterzenbach et al., 2013, p. 226).
FSCs target a broad customer base and thus differentiate their product
according to customer needs. This is expressed in different cabin classes,
typically including First, Business, and Economy. In-flight frills, such as
food, drinks, and entertainment, are usually free of charge. The bundling
of services in combination with extensive frequent flyer programs requires
sophisticated yield management systems, which often comprise more than
20 different booking classes (Shaw, 2016, pp. 212–214). The use of multi-
ple distribution channels, including global distribution systems (GDS), di-
rect online sales, and travel agencies, further increases the production cost
of these carriers. Table 3.2 summarizes all the individual features by busi-
ness model.
33
Airports with high traffic volumes assign landing time windows (i.e., slots) to airlines to avoid
congestion of the available infrastructure.
Table 3.2 – Key characteristics of airline business models34
Full- Sched- Re- Cargo Low- Super- “Hy-

service/ uled gional airline cost differ- brid”
legacy charter/ carrier carrier entiated carriers
carrier leisure airlines
airline
Route Short- Short- Short- Short- Previ- Short- Short-
types haul haul and haul haul ously haul and haul
and long- and only long- and
long- haul long- short- haul long-
haul haul haul, re- haul
cently
also
long-
haul
Fleet Large & Small & Small All pos- Large Very All pos-
diverse diverse sible homoge- small sible
neous
Net- Hub & Point-to- Point-to- All Point-to- Point-to- All types
work Spoke point point types point point
type
Focus Major Both Second- Cargo Past: No spe- Both
air- airports major ary air- airports Second- cific fo- major
ports and sec- ports ary air- cus and
ondary ports; second-
recently ary
all air-
ports
Target All trav- Tourists Mostly Freight Previ- Busi- All trav-
cus- elers busi- ously ness elers
tom- ness mostly travelers
ers travelers tourists;
Recently
all trav-
elers
Prod- Differ- Stand- Stand- Cargo Stand- High- A la
uct of- entiated ardized ardized ardized end full- carte
fering full-ser- full-ser- full-ser- no-frills service product
vice vice vice product product
34
Source: Own illustration. adopted from Sterzenbach et al. (2013, p. 225).
Full- Sched- Re- Cargo Low- Super- “Hy-

service/ uled gional airline cost differ- brid”
legacy charter/ carrier carrier entiated carriers
carrier leisure airlines
airline
product product product (Busi- combin-
(First, (single (single ness/ ing full-
Busi- class) class) First) service
ness, with no-
Econ- frills
omy)
Pric- Differ- Block Upmar- Spot Cheap Very up- Differ-
ing entiated capacity ket pric- and a-la- market entiated
pricing sale to ing with block carte pricing, a-la-
with tour op- stand- capac- pricing no so- carte
complex erators ard rev- ity sale with phisti- pricing
yield enue standard cated with
man- man- revenue revenue complex
age- age- manage- man- yield
ment ment ment agement man-
systems systems systems systems age-
ment
systems
Distri- Multi- Mostly Multi- Direct Direct Mostly Multi-
bution channel via tour channel sale to online busi- channel
chan- distribu- opera- distribu- freight sale to ness distribu-
nels tion tors tion for- passen- travel tion
ward- gers agen-
ers cies
Scheduled charter airlines (SCA)
Prior to the deregulation, SCAs were not permitted to market single seats;
rather, they could only sell seat allotments to tour operators (Bieger & Witt-
mer, 2011, p. 97). Post deregulation, many formerly independent airlines
were integrated into larger tourism conglomerates35 in order to reduce the
transaction cost and adjust capacity more flexibly. SCAs are characteristi-
cally smaller than FSCs, although they can serve both short-haul and long-
35
E.g., Hapag-Lloyd to Preussag (later TUI) in 1997, Condor to C&N Touristik (later Thomas
Cook) in 1997.
haul flights. They serve the connections in their point-to-point networks

with low weekly frequencies, especially for long-haul flights, reducing the
complexity and allowing for a very high level of asset utilization (Doganis,
2010, p. 166). The typical product of a SCA consists of a single Economy
Class, which offers in-flight frills similar to that of an FSC Economy Class.
The distribution process is still dominated by tour operator allotments and
also complemented with the direct sale of individual seats (Shaw, 2016,
p. 154).
Regional carriers
Regional carriers focus on connecting secondary airports with small air-

craft and mainly rely on the business travel demand between these air-
ports. In addition, many regional airlines have contracts with FSCs to feed
their hubs with traffic from smaller airports (Pompl, 2007, p. 102). They
usually incur higher operating costs than other business models and often
serve monopoly routes with limited but inelastic demand, which allows re-
gional airlines to achieve higher yields compared to other business models
(Whyte & Lohmann, 2017, p. 113). They typically offer a single class prod-
uct with full-service features in order to satisfy the demand driven by busi-
ness travelers. In some cases, they also receive government subsidies to
connect remote locations as a public service (Pompl, 2007, p. 99).
Cargo airlines (CAR)
Cargo airlines exclusively carry freight, whereas airlines following other

business models focus mainly on passengers36. Many cargo airlines are
integrated either in an aviation conglomerate (e.g., Lufthansa Cargo or
Emirates Sky Cargo) or in a global logistic forwarder (e.g., UPS, FedEx),
allowing them to leverage the hub-and-spoke network of the parent com-
panies. Independent pure players, such as Cargolux or Atlas Air, operate
36
However, the revenue share of cargo in long-haul wide-body aircraft is up to 15% (Doganis,
2010, p. 306).
loose hub-and-spoke networks, which are much more flexible to accom-

modate short-term changes. Smaller fleets of cargo airlines (less than 30
airplanes) are typically homogeneous, whereas integrated express freight
companies deploy large heterogeneous fleets. However, the business
model of cargo airlines differs substantially from passenger airlines in three
ways. First, the customers are companies and not individuals – in most
cases freight forwarders. Second, the product differentiation is based on
the cargo properties and timeliness rather than customer experience.
Third, the sale of capacity in spot markets only occurs a few days before
departure, so the network can be adjusted quickly to reflect changes in
demand, especially if there are no blocked space allotments on a specific
route (Doganis, 2010, p. 304).
Low-cost carriers (LCC)
The emergence of low-cost carriers was possibly the largest disruption in

the airline business since the deregulation. LCCs maintain a very low cost
base by combining a large, homogeneous fleet with a slim administration,
a low degree of operational complexity, and decomposition of the tradi-
tional airline product (Shaw, 2016, pp. 103–111): aircraft manufacturers
grant huge discounts for large orders, lowering unit capital cost dramati-
cally. The standardization of both fleet and product reduces training and
maintenance cost and increases flexibility, since aircrafts can be ex-
changed easily. Most low-cost airlines began as start-ups and thus avoided
building an oversized administration as well as high wage contracts, a fea-
ture of better (regulated) times. Initially, LCCs purely served point-to-point
networks between secondary (and uncongested) airports, which charged
lower fees and guaranteed faster turnaround times to maximize asset uti-
lization. The lack of transfer passengers reduced the complexity further,
eliminating the need for schedule coordination (which is especially relevant
for FSCs and regional carriers). Finally, LCCs decomposed the full-service
product into a base product, which is the flight and a piece of hand luggage,
and many ancillary services that can be assembled a la carte by the pas-
senger to best suit his preferences. Ancillary revenues can generate up to
30% of the total revenues (Doganis, 2010, p. 152). This led to a focus on
the load factor rather than average yield, since most ancillary revenues are
generated after the actual booking. At the same time, sophisticated yield
management systems, as used by FSCs, became obsolete.
Super-differentiated airlines
Super-differentiated airlines implement a focus strategy opposite to that of

LCCs. Instead of focusing on the cost leadership, these airlines focus on
providing the most luxurious travel experience to a price-insensitive cus-
tomer base (Shaw, 2016, p. 151). The majority of this market is served by
unscheduled business aviation providers that do not qualify as an airline
according to the definition developed by Wensveen (2015). There were a
small set of scheduled airlines offering scheduled Business Class-only ser-
vices on high-yield business routes in the 2000s, many of which ceased
operations during the financial crisis of 2008 (Shaw, 2016, p. 153). At the
time of this study, only one of these airlines, namely the French airline La
Compagnie still maintains scheduled Business Class-only flights as inde-
pendent airline.
Hybrid carriers
With the market entry of the new LCCs, strategic reactions of the FSCs led
to innovations in their own business model and service offering. Simulta-
neously, LCCs added premium services to their own product set, so that
the lines between these business models became blurry (Daft & Albers,
2013, p. 47). The new breed of business models has often been called the
“hybrid” business model, since they incorporate elements of both FSCs
and LCCs in varying degrees (Lohmann & Koo, 2013, p. 7). Such business
model convergence can occur in two different ways: first, airlines can alter
their product portfolio. This entails either FSCs decomposing their bundled
full-service product and offering a la carte ancillaries similar to LCCs; con-
versely, LCCs can start to bundle some of their a la carte products in some-
thing similar to full-service products (Daft & Albers, 2013, p. 50). Second-
arily, airlines can establish their own LCCs to compete directly with the new
market entrants and benefit from the low-cost structure of a newly founded
airline. This results in large airline groups operating both FSCs and LCCs
simultaneously but with clearly delimited scope and target markets (Pompl,
2007, p. 127). Airlines in North America have mostly chosen hybridization
through product portfolio adjustments, whereas European airlines have de-
ployed a mix of both hybridization approaches (Daft & Albers, 2015;
Klophaus, Conrady, & Fichert, 2012).
3.2.4 Major business processes of airlines
The business processes of airlines can be roughly clustered in three

groups: Commercial processes, operational processes, and administrative
processes (Wensveen, 2015, p. 232). Administrative processes are usu-
ally staff functions reported directly to the Chief Executive Officer (CEO),
whereas commercial processes are headed by a Chief Commercial Officer
(CCO) and operational processes by a Chief Operations Officer (COO).
Figure 3.4 provides an overview of the major processes in each category.
Operational processes are the responsibility of the classic cost centers and
thus strive to provide a satisfactory level of service at minimal cost. Hence,
efficiency is the most important innovation driver for operational processes
(Wensveen, 2015, p. 248). Commercial processes are usually revenue
centers, with the exception of NP, which can alternatively be a profit center
that coordinates commercial and operational processes (Goedeking, 2010,
p. 99). Strategic planning, including fleet planning, is usually considered an
administrative function reported directly to the CEO (Wensveen, 2015,
p. 241). Administrative processes are either cost or profit centers in the
organization, depending on the responsibilities and decision-making rights
assigned to them.
Figure 3.4 – Overview of standard airline processes37
3.3 Introduction to the network theory
Networks are omnipresent in almost all aspects of life and science. They
can take different appearances, sizes, and strengths. Networks are clearly
visible (e.g., a road network) or almost invisible (e.g., a social network).
Even in airlines, there exist numerous network structures: route networks,
airline alliance networks, computer intranet networks, and so on. In this
sub-chapter, section 3.3.1 first defines the term “network” in general and
discusses different network forms relevant for scientific research, with an
aim to clearly delimit the airline route networks from other network types.
Thereafter, section 3.3.2 presents the main scientific theories to describe
and analyze networks. The most important theories are introduced in de-
tail, namely the graph theory (section 3.3.3) and flow networks (sections
3.3.4 and 3.3.5).
3.3.1 Definition of networks
A network is a system constituting connected objects (Casti, 1995, p. 5).

These objects can take several forms, based on the type and purpose of
37
Source: Own illustration. following Wensveen (2015, pp. 233–263).
Introduction to the network theory 45
the network. In social networks, the objects are usually human actors or
groups of actors and the connections are the relationship between these
actors (Brass et al., 2004, p. 795). In technical networks, the objects can
be computers, televisions, cities, or power plants as examples, and the
connection is usually a physical connection, such as a wire, an electromag-
netic signal, or a road.
The general purpose of a network is to enable the flow of “something,”

using the connections between the objects. This “something” can be as
diverse as information, passengers, or electricity. If the flow can be physi-
cally measured and expressed in a numerical demand figure or have a
capacity constraint, the network is considered a “flow network.” Networks
without specifiable flows and unlimited capacity are called “pure networks”
(Bell & Iida, 1997, p. 19). The focus of the analysis in pure networks is
usually on network structure and connectivity. In flow networks with addi-
tional path choices, the focus is on cost functions and demand and supply
properties (Bell & Iida, 1997; Casti, 1995).
This study is concerned with transportation networks, which are classic

flow networks. Network objects, which are usually called nodes in the con-
text of transportation networks, can be points in space and time that have
interaction potential. These points can include cities, industrial centers, or
transportation facilities (Bell & Iida, 1997; Nakicenovic, 1995). In practice,
transportation facilities, such as airports, ports, or train stations, are most
frequently used as nodes in transportation networks. Network connections
or links connect these nodes and materialize in the form of roads, railways,
canals, or air and sea routes. Table 3.3 provides a summarized overview
of network properties of different transportation networks.
Table 3.3 – Overview of transportation networks by transportation mode38
Transport Network nodes Network links Individual Community

mode transporta- transportation
tion means means
Road Spatial demand Different road Cars, trucks, Buses, trucks
clusters (cities, types such as bicycles
industry centers, highways,
etc.) streets, etc.
Railway Train stations Underground or Usually N/A Trains
above-ground
rail tracks
Water Ports Canals and sea Variety of Ferries, cruise
routes boats ships, freight
ships
Air Airports Air routes General avia- Passenger air-
tion planes & craft and heli-
helicopters copters, cargo
aircraft
For the analysis and optimization of a network, network capacity is of ut-

most importance (Casti, 1995). Capacity-constrained networks bear the
risk of a flow shortage, which originates from capacity constraints in either
network objects, connections, transmission modes, or a combination of
these. Connections and transmission modes are usually only constrained
if the network flow is quantifiable39.
38
Source: Own illustration. following Nakicenovic (1995), Bell and Iida (1997), and Frybourg
and Nijkamp (1998).
39
Let us consider social networks as an example. Network flows of affect are not quantifiable
(Granovetter, 1973), and information and knowledge are at least difficult to quantify (Brass et
al., 2004). The transmission of these flows in relationships is direct, i.e., there is no need for
a transition mode. The only capacity supply constraint present in social networks are thus the
network objects themselves – the amount of information, knowledge, and affect that social
actors are willing or able to transmit in a social network.
Transportation networks possess not only capacity-constrained network

objects but also constrained connections and transmission modes. Trans-
portation facilities – the network objects of transportation networks – have
a limited capacity to serve transportation modes; e.g., airports have a lim-
ited number of slots and limited terminal buildings. Moreover, most trans-
portation networks have also capacity constraints in terms of network con-
nections and transmission modes. There is only a certain amount of traffic
that can be handled by a specific railway or road. At the same time, trans-
portation modes, such as trains, buses, airplanes, and ships, have a limited
capacity to accommodate passengers and freight.
The second property measure to distinguish network types is the emer-

gence of networks. Social networks often have a hybrid emergence. While
interpersonal networks often evolve unguided (at least in the non-profes-
sional context), interorganizational networks are generally strategically
planned (Lavie, 2006; Provan, Fish, & Sydow, 2007). However, technical
networks require dedicated infrastructure and transmission modes in order
to fulfil their function. These networks rarely ever evolve spontaneously
and need to be planned and created (Casti, 1995). Transportation net-
works is an exception here, since they may evolve without central planning,
especially in emerging countries without sophisticated infrastructure and
governmental planning. A couple of examples are need-based ride sharing
and micro-logistic concepts in rural areas (Nakicenovic, 1995).
A final decisive criteria is network reciprocity, which describes the direc-

tional flow between two network objects (Garlaschelli & Loffredo, 2004,
p. 795). In social networks, reciprocity is possible but not necessary. In
contrast, transport networks require network reciprocity. It is logically im-
possible to move passengers only in one direction, without any passengers
returning (including different return paths). After a while, there would be no
passengers left to carry, and the demand center would disappear. For
freight transportation, this phenomenon is slightly different since the car-
ried goods may be consumed at the destination, but the transportation
mode needs to return at some point to keep the network running. Counting
deadheads as bi-directional network flow, reciprocity is a necessary condi-
tion for transportation networks.
This study focuses specifically on airline route networks. Therefore, the

network objects – depending on the applied granularity level – are demand
clusters or airports. The network connections are air routes served by air-
planes as the transmission mode to enable the flow of passengers and
cargo. All of these elements underlie the capacity constraints, which are
further analyzed in section 3.3.4. Due to the asset and infrastructure spec-
ificity, airline networks are always the result of a deliberate planning pro-
cess, called the airline network planning process. The reciprocity of net-
work flows is a basic condition for all airline networks to function properly.
3.3.2 Distinction between the network theories
Network theory is an umbrella term for a multitude of individual theories

and techniques related to network analysis (Provan et al., 2007, pp. 479–
480). Network research in social sciences, including management science,
has most often referred to social networks (Borgatti et al., 2009; Brass et
al., 2004). A simple full-text keyword search on the Business Source Com-
plete database supports this statement, with “social network” yielding over
6,000 scholarly articles, whereas “transport network” or “transportation net-
work” result in less than 1,000 hits40. Despite this significant amount of
network-specific literature, there exists no unanimous network theory
across network types, not even for specific network types, such as social
network or transportation networks (Borgatti & Halgin, 2011, p. 1168).
Since airline networks are technical networks, social network analyses

cannot be applied to explain their characteristics. Thus, this study focuses
40
Retrieved via EBSCOhost on October 30, 2017 with Boolean search phrase “social net-
work” and “(“transportation network”) OR (“transport network”)” and filter for peer-reviewed
academic journals only.
on transportation networks and the associated network theories. Many

scholars have intended to define a typology of transportation network the-
ories, but there is no consensus in the broad range of research domains
studying network-related phenomena. Bell and Iida (1997) propose to dis-
tinguish static and dynamic network analysis. Static network analysis is
hereby concerned with network structure and properties, whereas dynamic
network analysis studies the flows within networks. Figure 3.5 summarizes
the concept pertaining to network theory discussed in this study. The re-
search on transportation networks has usually been fueled by quantitative
network theories (Bell & Iida, 1997, p. 17). From the static perspective,
graph theory predominates the transportation network analysis (Casti,
1995), and it is introduced alongside basic network properties in the next
section. However, transportation networks are flow networks and thus the
dynamic view has traditionally been more prevalent in academic literature
(Bell & Iida, 1997). Similar to the social network analysis, there is no single
theory on the dynamic view of transportation networks. The representation
and optimization of network flows can be summarized as the network opti-
mization theory, comprising many individual optimization issues (Magnanti
& Wong, 1984).
Figure 3.5 – Network theory in transportation networks41
Apart from theories describing the static and dynamic properties of net-
works, there also exist theories related to the impact of networks. From a
strategic management perspective, the expectation towards a network –
independent of its type – is to contribute towards building a competitive
advantage (Gulati, Nohria, & Zaheer, 2000, p. 204). When considering the
effect of networks in the management research domain, it is indispensable
to relate networks to the common theories on competitive advantage in
strategic management. For this purpose, section 3.7.5 discusses networks
in the context of the resource-based view.
3.3.3 Fundamentals of the graph theory
The graph theory has mostly been developed by mathematicians and can
be traced back to Leonhard Euler’s Königsberg bridge theorem in 1736
(Casti, 1995; Newman, Barabasi, & Watts, 2011). Every graph consists of
41
multiple nodes (Letters A to F in Figure 3.6), which are connected. Moving

through the network via different connected links creates a path. If the
origin node is also a destination node, the path is closed and thereby forms
a cycle. The matter moving through the network is a flow, which is usually
limited to the maximum capacity of the links. The terminology used in this
study is collated in Table 3.4.
Table 3.4 – Graph theory terminology42
Term used Definition Alternative Sources

in this terminol-
study ogy
Node Junction of two or more links Vertex Bell & Iida, 1997;
Casti, 1995; New-
man, 2003
Link Connection between two nodes Edge, arc Bell & Iida, 1997;
Casti, 1995
Path Sequence of nodes connected by Bell & Iida, 1997;
links in one direction to enable a Casti, 1995
flow from the first to the last node
Cycle Path where start and end point are Circuit Bell & Iida, 1997;
the same node Casti, 1995
Flow Movement between two nodes Bell & Iida, 1997
Capacity Maximum allowed flow in a time Casti, 1995
unit on a link between two nodes
Degree The number of links connected to a Connectiv- Newman, 2003
node ity
Directed graphs allow one flow direction for each link. In contrast, undi-
rected graphs allow bi-directional flows between nodes (Newman, 2003,
p. 172). All graphs can also be represented in a matrix. Figure 3.6 presents
an undirected graph with six nodes, A to F, and five links. The properties
of graph connectivity can be best elucidated using the three matrices in
42
Source: Own illustration. following the authors mentioned in the table.
Figure 3.6. The adjacency matrix summarizes whether two nodes are di-
rectly connected via a link in a binary system. The node pair AB is con-
nected by a direct link; hence, the adjacency matrix value is 1. Similarly,
node pair BC has no direct link, so the adjacency matrix value is 0.
Figure 3.6 – Connectivity matrices of an undirected, strongly connected graph43
The reachability matrix indicates whether a specific node pair is connected

by a path, which could form a direct link or a combination of multiple links.
The present graph is strongly connected; i.e., all the nodes are connected
by at least one path. The reachability matrix indicates 1 for all node com-
binations. Since it is an undirected graph, even the origin node can be
reached by moving back and forth. Finally, the distance matrix indicates by
how many links it must move between the nodes. Nodes AB are directly
connected; hence, the distance value is 1. Conversely, nodes AC are not
directly linked but reachable via the path AFC, which constitutes the direct
links AF and FC; thus, the distance is 2. If there is more than one possible
path to link two nodes, the distance of the shortest path – the geodesic
path – is shown. The longest geodesic path within a graph represents the
graph diameter. (Casti, 1995; Newman, 2003)
Besides the strongly connected graph, there are other graph types that are
depicted in Figure 3.7. A graph where every node is connected with direct
links to all other nodes is called a hyper-connected graph (Casti, 1995). If
43
graphs contain unconnected nodes or two sub-graphs that are not con-
nected, these entities are called components (Newman, 2003, p. 173). The
graph in the middle displays a two-component graph, consisting on the
components DE and ABCF. Finally, there may be different types of links
as depicted on right side of Figure 3.7. These links may differ according to
the types of flows they transmit or in the flow capacity (Newman, 2003,
p. 181). Examples of these networks are road-railway cargo networks or
high- and low-voltage electricity networks.
Figure 3.7 – Overview of graph types44
Besides graph connectivity, Albert and Barabási (2002, p. 71) have men-
tioned three key network attributes: average path length, clustering coeffi-
cient, and degree distribution. Newman (2003, p. 189) has added network
resilience and betweenness to the attribute analysis.
The average path length describes the arithmetical average of all the geo-
desic paths of a network. This measure has received a large amount of
public and scientific attention, triggered by Milgram’s (1967) small-world
problem, in which he found that the average path length of the social net-
work of all humans on earth is between 6 and 7 links. In other words, most
44
people are connected with at least half of the entire human population via
7 or fewer links. The implications of the small-world problem for graph the-
ory are discussed at the end of this section. The betweenness or centrality
measure indicates how many geodesic paths intersect at a specific node
(Newman, 2003, p. 194). This also is the theoretical base for influencer
modelling in social network analyses (Wasserman & Faust, 1994).
Transitivity or clustering in networks is a measure for the likelihood that

friends of your friends are more likely to be friends with you. In other words,
if A is connected to B, and B is connected to C, there is a high likelihood
that A is also connected to C. The importance of these “triangles” can be
quantified using the clustering coefficient (Albert & Barabási, 2002; New-
man, 2003). The coefficient measures the relation of triangle links to “nor-
mal” triples, quantified as follows:
3 ∙ number of triangles in the network
𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑖𝑛𝑔 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 =
number of connected triples of nodes
If we take the graph in Figure 3.6 as an example, we can identify one tri-
angle ABD, and ten triple relationships45. Therefore, the clustering coeffi-
∗
cient is = 0.3. As presented in Table 3.4, the degree of a node is the
number of links connecting to the node. The degree distribution indicates
the probability of a node i to have k degrees (Albert & Barabási, 2002,
p. 73). Figure 3.8 exemplifies a degree distribution for the graph introduced
earlier in this section.
45
Triple relationships in undirected networks are counted bi-directionally. For example, the
path ABE has its counterpart as EBA. The 10 triples in Figure 3.6 are ABE, AFC, BAF, CFA,
DBE, DAF, EBA, EBD, FAB, and FAD.
Figure 3.8 – Degree distribution example46
Network resilience measures the robustness of a network if individual

nodes are removed. Usually, the average path length increases with the
removal of nodes. The elasticity of the average path length in case of ran-
dom or targeted node removal is the most common indicator for network
resilience (Newman, 2003, p. 189). All of these network properties are
highly relevant for airline networks and are discussed in that context in sub-
chapter 3.4.
3.3.4 Network flows and network optimization
The network properties described in the previous sections focus primarily

on network structure and evolution. For transportation networks, however,
network flows are equally important. Hence, the optimization of network
flows through the design of network structure, flow means, and schedule
are key components of network theory. This section presents the basic
concepts of network flow and network optimization that directly impact NP
for airline route networks.
46
In comparison with static or pure networks, the analysis of flow networks

is performed with many more variables and constraints. Geospatial loca-
tion of the network nodes and the type of links are shared features of both
static and flow networks. However, flow networks always incorporate a flow
demand component that can be completely satisfied if the links are not
capacity constraint. In this case, the main concern of network optimization
is simply to find the “best” route, be it the shortest or cheapest path to get
a flow from one node to another (Bell & Iida, 1997, p. 19).
The physical properties of links in most real-world network types impose a

certain capacity constraint. This capacity constraint becomes critical as
soon as the demand exceeds the supply capacity. Now, certain flows can-
not take the shortest or cheapest path but need to be re-routed to the next
best path or cannot be transported at all (Casti, 1995). Especially in the
case of transportation networks, there are multidimensional capacity con-
straints. Consider a railroad network; each railroad has a maximum num-
ber of trains that can operate safely in a specific time window. Additionally,
for each train station, there is a maximum number of trains it can handle at
one time, which is usually constrained by the number of platforms and out-
going rail tracks. Moreover, each train also has a maximum capacity for
people or freight. In total, we find at least three network items with individ-
ual capacity constraints, namely links (railroad tracks), nodes (train sta-
tions), and transportation means (trains).
Flow networks have some more properties than static networks. The start-
ing point of a specific flow is called the origin or source. The last node of a
specific flow is the destination or sink. Nodes with a negative net flow are
called demand nodes, since more flows originate than those that arrive. In
turn, supply nodes have more arriving flows than originating flows, creating
a positive net flow. If the net flow is zero, the node is called a transshipment
node (Bazargan, 2016). If a network is separated into multiple independent
networks by deleting one or more links, the combination of eliminated links
is called the cut (Casti, 1995).
In the following paragraphs, the most important network flow issues and
their practical implications for transportation networks are introduced.
Shortest path problem
The shortest path problem is the most fundamental network flow problem
and a basis for almost all network optimization theories. First formulated
by Czech mathematician Otakar Borůvka in the 1920s (Nešetřil, Milková,
& Nešetřilová, 2001), it aims to find the optimal paths for specific source
destinations in higher-order networks. In static network theory, using a link
comes with no cost, so the optimal path is always the geodesic path47. In
a flow network, links get assigned a distance or a cost (Ahuja, Magnanti, &
Orlin, 1993, p. 94). Hence, the problem is also known as the minimum cost
problem. Consider a network such as the one depicted by the directional
graph in Figure 3.9.
Figure 3.9 – Simple network with assigned link cost48
47
The geodesic path is the path connecting two nodes with the smallest number of links.
48
Own illustration following Bazargan (2016).
Each link has an associated flow cost that can also be viewed as the travel
distance between two nodes. The geodesic paths connecting nodes A and
D are ABCD and AFED. However, if we consider the link distances (cost),
the shortest (cheapest) path from A to D is ABCED with a distance (cost)
of 10, whereas the geodesic paths have lengths of 12 (ABCD) and 14
(AFED).
Since the shortest path problem does not consider any capacity con-
straints, it is not wholly transferable to most transportation networks. How-
ever, it is very useful to understand the basic mechanism of geographic
routing that applies to all physical transportation networks (Goldman &
Nemhauser, 1967). The most common application in real transportation
networks is traffic planning on road networks, as it is applied by traditional
navigation systems without real-time traffic information for example (Zhan
& Noon, 1998, p. 66).
Consider that you want to drive from Frankfurt to Berlin at night with little
traffic, so the capacity constraint of the road system is negligible. Most nav-
igation systems provide you with two route options, the shortest and the
most economic route. When considering the shortest route, the navigation
system assigns distances to the links to calculate the shortest path. When
choosing the most economic option, the navigation system assigns travel
times (which equal cost) to the links to calculate the path with the shortest
travel time. This example demonstrates that one network – in this case the
German road network – can have varying allocations of distance and cost.
Nevertheless, the algorithm required to determine this remains the same.
Over the last 90 years, the classical shortest path problem has produced
a significant body of literature. Kruskal (1956) and Dijkstra (1959) proposed
the first solution algorithms with exact but computational heavy solution
algorithms that examined every network link. Subsequent research fo-
cused on reducing the computing time and not overly deteriorating the fol-
lowing empirical results obtained. Dreyfus (1969) and Ahuja, Magnanti,
and Orlin (1991) presented a comprehensive overview over the develop-

ment of solution algorithms. Deo and Pang (1984) provided a good sum-
mary on generalizations and specifications of the problem.
Maximum flow problem
The maximum flow problem is in many respects the counterpart of the

shortest path problem and the second nucleus of the flow optimization the-
ory. While the shortest path problem minimizes path lengths and costs
without capacity constraint, the maximum flow problem isolates capacity
constraint without assigning path length or cost (Ahuja et al., 1993, p. 167).
According to Ahuja et al. (1993, p. 184), the most important theorem of the
maximum flow problem in the entire network optimization is the max-flow
min-cut theorem introduced by Ford and Fulkerson in 1956. Based on em-
pirical observations in the Soviet railway system, they developed a simple
rule to determine the maximum flow capacity of a network. The maximum
capacity of the network is the capacity of an individual cut with the smallest
capacity of all cuts. Consider the same network as in the minimum path
problem, depicted in Figure 3.10.
Figure 3.10 – Maximum capacity determined by the max-flow min-cut theorem49
The number value now indicates the maximum capacity for each link. The
dotted lines show all possible cuts of the network, which are summarized
in the table on the right. The maximum capacity of each cut is the aggre-
gated flows that can flow from the left part of the cut to the right part of the
cut under the consideration of flow directions. For instance, cut b separates
the left part AF from the right part BCDE. The links connecting the two parts
are AB, FB, FC, and FE with an aggregated capacity of 13 flow units. Cut
c relies only on the links BC and AF to connect the left part AB and the
right part CDEF of the network. Here, the maximum capacity is only 7 flow
units, which is also the maximum capacity of the entire network.
The generic problem with “bottlenecks” limiting the capacity of large trans-
portation networks are most commonly observed in road and rail transport,
but it is also present in airline networks and shipping networks. However,
most capacity restrictions for the latter are determined by node capacity
restrictions (airports and sea ports), although waterways, such as the Suez
canal or the Panama canal, have capacity restrictions for shipping routes.
49
Source: Own illustration. following Ahuja et al. (1993, p. 184).
Minimum cost flow problem
The minimum cost flow problem is the combination of the maximum flow
and shortest path problems. In a directed network with one or more origins
and one or more destinations, each link has both capacity restriction and
associated flow cost. This is much closer to the reality of many transporta-
tion networks than the simpler problems presented earlier50 (Ahuja et al.,
1993, p. 295).
The objective of the problem is to determine how much of a single flow

commodity is routed on every possible path from the source to the desti-
nation in order to minimize the total flow cost. Consider a two-origin and
three-destination network with two transshipment nodes as depicted in Fig-
ure 3.11. The solid links have a capacity of 75 flow units and the dotted
lines of 50 flow units. The numbers without an arrow frame describe the
flow cost per link. Without capacity restriction, all commodities from B
would have routed via D, since the total transport costs are lower than or
equal to routing via C. However, the capacity restriction requires that 25
units flow via C. The resulting cost-minimizing flows are shown in the block
arrows next to the links.
50
The minimum cost flow problem has undergone multiple specifications and generalizations
that lead to a variety of special cases and sub-problems. Ahuja et al. (1993, pp. 317–337)
provided a comprehensive collection of these modifications as well as developed solution
algorithms.
Figure 3.11 – Exemplary network and solution for the minimum cost flow problem51
The minimum cost flow problem and its derivatives are of central im-
portance for transportation networks. First, it provides a sound set of solu-
tion algorithms for general routing problems. The network in Figure 3.11
could for example be understood as a simple network of a cargo airline
transporting a single commodity. The intermediate stop D has a short run-
way and thus can only be served by small aircrafts, carrying up to 50 tons
of cargo. The minimum cost flow problem here offers a simple routing al-
gorithm to optimize this network.
3.3.5 Design of flow networks
While graph theory evaluates the properties and attributes of networks,

network optimization determines optimal flows through the network; yet, a
third component is missing for a complete understanding of NP. Neither
graph theory nor network optimization recommend an optimal design of the
51
Source: Own illustration. following Bazargan (2016).
network. Graph theory assumes that it either evolves randomly or proba-

bilistically. In network optimization, the network and the corresponding
properties are usually predefined; therefore, only the flows are optimized.
However, network optimization is closely connected to network design,
since efficient networks can only be designed by understanding the opti-
mality conditions for network flows.
However, network design is in fact much more complex than network opti-
mization. Network optimization problems are characterized by solving one
well-specified problem, and more importantly, it is usually towards one sin-
gle optimization parameter. However, network design requires encom-
passing and weighing all individual optimization problems and usually has
multiple optimization parameters (Magnanti & Wong, 1984, p. 2). Consider
airline network design as an example. Besides defining the most profitable
network, you also want to design a feasible network, i.e., a network that
can be flown with reasonable schedules and that exhibits some level of
robustness towards external shocks52.
Bell and Iida (1997, p. 20) have distinguished two archetypical network de-
sign problems: the continuous and the discrete network design problems.
The continuous network design problem takes the network structure as
given and aims to change the network characteristics, such as capacities
and flow cost. Real-life examples of these design problems are the addition
of a lane to an existing road or the installation of a traffic light. In airline
networks, the determination of flight frequencies and the offered capacity
(fleet assignment) are continuous network design choices (Bell & Iida,
1997, p. 20). Discrete network design problems are concerned with the
determination of the network structure. In transportation networks, this can
include infrastructure investments, such as the construction of a new road
or railway, and the carrier route selection (Bell & Iida, 1997, p. 20). In an
52
Changes in the network structure usually entail adding and deleting destinations; shocks
such as political unrest or expired traffic rights can complement economic considerations for
network changes.
airline network, the most common discrete network design problems are
network type and route selection.
Discrete and continuous network design models can be viewed as a natu-

ral sequence 53 . Discrete network design choices usually have a longer
planning horizon and require larger investments than for continuous
choices (Guihaire & Hao, 2008, p. 1252). It is typically cheaper to repave
or widen a road than to build a completely new road. However, real-life
investment decisions require a holistic evaluation of the investment oppor-
tunities. An airline has to determine whether the newly acquired aircraft
should be used to add frequencies on an existing route (continuous design
choice) or to open a new route to passengers (discrete design choice).
Therefore, mixed models have evolved to jointly analyze discrete and con-
tinuous network design problems (Yang & Bell, 1998, p. 258).
Most continuous network design models focus on infrastructure projects,

more specifically on ground transportation, such as roads and railways
(Bell & Iida, 1997, p. 20). This might result from the greater asset intensity
of infrastructure projects than the network design of a transportation carrier
(Farahani et al., 2013, pp. 281–282). While adding a new route for airlines
or trucking companies comes with little to no asset investment, building a
new road implies substantial financing. Hence, changing the characteris-
tics of the network may often be more beneficial than changing the network
structure.
3.4 Airline networks
Airline networks represent a special form of transportation networks and

have many industry-specific properties. Section 3.4.1 presents the speci-
ficities of airline networks, following which section 3.4.2 outlines the most
53
For an extensive overview of the existing network design models, see Farahani et al.
(2013).
Airline networks 65
common types of airline networks. The underlying economics of airline net-

works are analyzed in section 3.4.3, before concluding the sub-chapter
with an overview of airline network-specific measures of efficiency, effec-
tiveness, and spatial distribution (section 3.4.4).
3.4.1 General properties of airline networks
The structure of airline networks is comparable to most other transportation

networks. As a starting point, Table 3.5 collates the airline-specific network
terminology. The nodes in an airline network are airports that are charac-
terized by providing the necessary infrastructure for airplanes, the trans-
portation means in airline networks (Burghouwt, 2007, p. 22). Essential in-
frastructural parts include runways, passenger terminals, cargo handling
facilities, connections to other ground or sea transportation systems, and
the offer of the required aviation services to enable flight operations
(Burghouwt, 2007, p. 22).
Table 3.5 – Airline network terminology54
Graph theory terminology Airline network terminology

Node Airport
Central transfer node Hub
Non-central transfer node Station
Link Route
Two unconnected nodes Unserved route
Path Itinerary
Demand source Origin
Demand sink Destination
Demand between specific sink and source O&D demand
Flow event Flight
Flow Traffic
Transportation means Airplane
54
The flows in an airline network are usually referred to as traffic. Similarly,

the demand for flow is the traffic demand. A single flow event between two
airports is called a flight, and the link between two airports is the route
(Goedeking, 2010, pp. 7–11)55. In fact, there exists almost an indefinite
number of unserved routes in every airline network, since an airline can
theoretically decide to fly to any suitable airport in the world. The routes
with regular scheduled flights are called served routes and are the links of
an airline network. In some cases, airplanes fly to more than one airport as
part of the same flow. Goedeking (2010, p. 5) calls these multi-flight routes
itineraries, which are especially relevant to cargo airlines.
If we look at the demand for airline network flows, the source of the demand
is called the origin, and the sink of the demand is called the destination.
The demand between a specific origin and destination is abbreviated as
O&D demand (Bazargan, 2016, p. 32). The demand for air travel is usually
not created at the airport itself but in the catchment area of an airport. Since
airports are connected to other ground or sea transportation systems, pas-
sengers and cargo start their individual itinerary not at the airport but at
individual homes, offices, or at manufacturing sites. Similarly, their journey
does usually not end at the destination airport but at some location in the
catchment area of those airports. These individual origins and destinations
are called the real O&D.
Depending on the centrality of the node in the network, Burghouwt (2007,

p. 24) differentiates between three types of nodes: decentral airports are
simple destinations, where no flight connections are usually offered. Cen-
tral airports can either be hubs if schedules are optimized for connection
traffic or stations if connections are possible but the airline’s schedule is
55
The terminology for links is inconsistent in literature. While some authors refer to links as
routes (Goedeking, 2010, p. 7), others use city pairs (Bazargan, 2016, p. 31) or flights
(Burghouwt, 2007, p. 22).
Airline networks 67
not optimized for it. A typical example of stations are airports that are crew
bases for low-cost airlines.
The geographic setup of airports connected via routes is the spatial dimen-
sion of an airline network (Burghouwt, 2007, p. 7). The analysis of the spa-
tial dimension is rooted in graph theory and has become increasingly pop-
ular since the 1980s (Chou, 1990; Teodorović & Guberinić, 1984). The de-
regulation of air transport in the United States in the 1970s and in Europe
in the 1990s has also attracted extensive research on the impact of dereg-
ulation on spatial structures of airline networks (Gillen, 2005; Gillen & Mor-
rison, 2005). The most important spatial trend has been the increased cen-
tralization of airline networks around the so-called hubs. Rooted in the gen-
eral network design theory (presented in the previous sub-chapter), schol-
ars have identified the hub location problem as a central design problem
for airline networks (O'Kelly & Miller, 1994).
The increasing importance of hubs and thereby of connecting flights has

significantly impacted the planning and the analysis of airline networks.
The temporal dimension of the network has become a crucial aspect: it is
not only important where an airline is flying to but also when and more
importantly with what connection times (Burghouwt, 2007, p. 8). Increasing
airport congestion and new airline alliances that emerged with code shar-
ing agreements further pronounced the importance of temporally optimized
network (Burghouwt, 2007; Goedeking, 2010). Hence, scheduling has
complemented network design as a vital activity in the NP process.
3.4.2 Types of airline networks
The structure of airline networks differs widely across business models,

history, and market conditions. From an analytic perspective, network ar-
chetypes can be defined by considering the spatial and temporal concen-
tration of the network. If a network is spatially concentrated, most traffic
flows are directed through one central airport, the hub. Conversely, if a
network is spatially dispersed, airports are connected directly and display
little difference in the number of connections (Burghouwt, 2007, p. 12). The

temporal concentration is expressed in the degree of flight coordination.
Temporally concentrated networks enable shorter connection times for the
customers and are called coordinated networks. In contrast, temporally un-
coordinated networks have random connection times (Goedeking, 2010,
p. 23). The following paragraphs highlight the four network archetypes and
conclude with a reality check pertaining to the actual network structure of
airlines.
Figure 3.12 – Airline network archetypes56
56
Source: Own illustration. following Burghouwt (2007).
Airline networks 69
Chain networks
Chain networks, also referred to as tour networks, milk-runs, or linear net-

works (Burghouwt, 2007, p. 14), are the oldest form of airline networks.
The extreme case, as depicted on the upper left corner of Figure 3.12 , is
characterized by two start/end points and several transshipment nodes.
The chain network may comprise a long itinerary performed by a single
aircraft or by a connection of multiple itineraries. In case of aircraft
changes, the schedules need to be coordinated in order to avoid long con-
nection times (Lederer & Nambimadom, 1998, p. 789).
In the early days of aviation, aircraft ranges were small and refueling stops
common in connecting distant airports. Typical examples for chain net-
works were the intercontinental networks of European and North American
airlines before the suitable long-haul aircraft was introduced in the 1970s57.
Figure 3.13 exemplifies Lufthansa’s South America network, spanning

from Berlin to Buenos Aires in 1934, which featured no less than 14 inter-
mediate stops and four aircraft changes (Airportzentrale, 2014). With the
advent of economic long-haul airplanes, such as the Boeing 747, refueling
stops became obsolete on most routes.
Since chain networks are less efficient than other network types (see sec-
tion 3.4.3), most passenger airlines have developed different network
structures. Today, chain networks are still common in cargo airlines and
with some regional airlines connecting remote airports (Gillen, 2005,
p. 61).
57
Exemplary aircraft ranges: 1930s: below 2,500 km (Junkers Ju 52 or Douglas DC-3);
1950/60s: 6,000-8,000 km (Douglas DC-6, Lockheed Constellation, Boeing 707, Douglas DC-
8); 1970s: above 10,000 km (Boeing 747, Douglas DC-10).
Figure 3.13 – Lufthansa's South America network in 193458
58
Source: Airportzentrale (2014).
Airline networks 71
Point-to-point networks
PP networks are dispersed and un-coordinated networks. The O&D con-

nections are served by direct flights rather than connecting flights passing
through a hub (Burghouwt, 2007, pp. 13–14). This has strong implications
on the number of routes necessary to connect several airports. If we con-
sider the six-airport graph in Figure 3.12, it requires 30 flights to connect
all the airports in both directions. In a centralized HS network, only ten
flights are necessary. However, as the schedules of the flights are not co-
ordinated, airlines serving PP networks do usually not offer connecting
flights at all (Goedeking, 2010, p. 32).
The standard structure in regulated markets is the PP network. Airlines

apply for route rights at the regulator and get assigned a set of routes. The
regulators usually favor direct connections, even between secondary air-
ports, and thus, most domestic airline networks in regulated markets are
PP networks to some extent. In exchange for a suboptimal network design,
regulators often limit the competition on specific routes, so that the airlines
can outweigh the cost disadvantage with higher fares (Lederer & Nambi-
madom, 1998). However, in the past 20 years, many low-cost airlines have
chosen PP networks without any regulatory intervention. The driving forces
behind this recent development are elaborated in section 3.4.3.
Hub-and-spoke networks
In the most extreme form of a HS network, all traffic flows are routed via a
single hub. The upper right graph in Figure 3.12 schematically pictures
such a “pure” HS network. The five spokes are only connected to one hub,
and any O&D demand between the spokes must change flights at the hub.
Compared to PP networks, the HS model has obvious advantages. As

mentioned in the last paragraph, much fewer flights are necessary to con-
nect all the possible O&Ds. The freed capacity can be deployed to routes
with the most demand, thus allowing for more flexibility in the schedule
design (Goedeking, 2010, p. 33). Further advantages are economies of
scale and scope, higher service levels, and an increased competitive po-
sition at the hub airports (Lederer & Nambimadom, 1998, p. 787).
Burghouwt (2007, p. 17) distinguished three types of hubs: global hubs,

specialized hubs, and regional hubs. Global hubs not only connect short-
haul with long-haul flights but also long-haul with other long-haul flights.
Frankfurt, for example, is the global Lufthansa hub, which enables travel-
ers from the United States to connect to India with reasonable connection
times. Regional hubs optimize connections from short-haul to long-haul
flights or simply from short-haul to other short-haul flights. Specialized hubs
leverage their geographic location to bundle long-haul flights to a specific
region and feed them with an associated short-haul network. Lisbon is a
good example of a specialized hub for flights to Brazil that connect to a
dense European short-haul feeder network.
Random radial networks
The spatial design of the random radial (RR) network is similar to the HS
network but lacks the temporal coordination of the hub. This network type
was the typical network for the European legacy carriers prior to the liber-
alization of traffic rights. Before the spread of the so-called sixth-freedom
rights59 or beyond rights, airlines were restricted to carry passengers only
between two countries, one of them being the home country of the airline.
Lufthansa, for example, could only carry passengers from Buenos Aires to
Frankfurt (or vice versa) but not from Buenos Aires via Frankfurt to Mum-
bai. Connecting traffic flows thus had little relevance, and the schedule had
little need for temporal coordination. With the increase of bilateral beyond-
rights traffic agreements, European home base airports, such as Frankfurt,
59
The sixth freedom right is the right to carry passengers or cargo from a second country to
a third country by stopping in one's own country. Please refer to Annex 2 of Burghouwt (2007,
p. 258) for a complete overview of the traffic right freedoms.
Airline networks 73
developed from central points of random radial networks to hubs in a HS

network (Burghouwt, 2007, pp. 12–18).
Real-life airline networks
Very few airlines actually serve a pure network archetype; rather, there are
varying shades of grey between these network types (Button & Stough,
2000, p. 26). The current networks are the result of a continuous network
development, which has been influenced by changes in regulation, com-
petition, aircraft technologies, and consumer preferences (Burghouwt,
2007, p. 2).
In the United States, for instance, networks first evolved from chain to PP
networks due to the availability of long-haul aircraft, and then after the de-
regulation of the 1970s, they developed into HS networks. All the large
United States (US) airlines have a multi-hub network, resulting from mer-
gers in a consolidating industry. In general, there has been a clear ten-
dency towards more spatial and temporal concentrations in the US airline
networks (Burghouwt, 2007; Button & Stough, 2000; Gillen, 2005).
In Europe, the situation has evolved differently. The legacy carriers moved
from chain to RR networks and then to HS networks. The spatial setup of
the legacy carrier networks has not changed substantially (Bel & Fageda,
2010), but the temporal coordination has increased significantly
(Burghouwt, 2007, pp. 3–4). The industry’s consolidation in Europe has
progressed slower than in the United States, but mergers in the past 20
years have produced more airlines with multiple hubs, such as Lufthansa,
Air France-KLM, and the International Airlines Group (IAG) 60 (Lordan,
60
Lufthansa Group currently operates with five hubs (Frankfurt, Munich, Zurich, Vienna and
Brussels), IAG with two hubs (London Heathrow and Madrid) and Air France-KLM with two
hubs (Paris Charles-de-Gaulle and Amsterdam).
2014, p. 1191). The Open Skies agreement between the US and the Eu-
ropean Union (EU) has not led to further concentration but to an increased
importance of alliances and joint ventures (Pels, 2009, p. 88).
The emergence of large low-cost airlines in Europe deserves additional

remarks. Initially, LCCs had extremely dispersed networks, although they
were far from being “pure” PP networks (Alderighi et al., 2007, p. 545). With
the opening of more bases across Europe, LCCs such as Ryanair and
Easyjet first decreased the spatial concentration of their networks; how-
ever, this trend reversed as they stationed more and more aircraft at each
base (Burghouwt, 2007, pp. 57–59). In fact, networks of large LCCs are a
combination of multiple RR networks around the operating bases.
3.4.3 Airline network economics
The size and the structure of networks have fundamental economic impli-
cations for airlines. Networks affect the cost structures, (de-)stimulate de-
mand, and influence the competitive situation of an airline (Burghouwt,
2007, pp. 2–4). The literature on airline network economics has long as-
sumed that in a deregulated market, the HS network type prevails, since
airlines can harvest economies of scale and scope, provide a broader ser-
vice offering, and increase the entry barriers to the airline industry (Goede-
king, 2010, p. 108). However, the expansion of low-cost-carriers focusing
on PP networks has led to a reconsideration of the traditional network eco-
nomics in competitive markets (Takebayashi, 2013, p. 93).
A natural efficiency driver of network development are economies of scale.

In the airline industry, economies of scale are defined as decreasing the
average cost per passenger by increasing the network size (Burghouwt,
2007, p. 26). Although economies of scale on the firm level are disputed61,
61
While Caves, Christensen, and Tretheway (1984) found a constant unit cost at the firm
level, Baltagi, Griffin, and Rich (1995) discovered economies of scale among North American
airlines.
Airline networks 75
the existence of such economies of scale at a route level is widely acknowl-

edged (Wojahn, 2001b). These cost advantages are often labelled as
economies of density, which describe the decreasing average cost per
passenger on a specific route when the number of transported passengers
increases. Economies of density may be reached by using larger aircraft
or increasing the frequency of flights on a specific route (Burghouwt, 2007,
p. 28).
While economies of density can explain the focus on few high-density

routes, it does not explain the spatial concentration of networks around a
specific hub. This is where economies of scope come to play. In the airlines
business, economies of scope are associated with “the joint production of
heterogeneous products” (Burghouwt, 2007, p. 26). The core product of an
airline is the transportation between an origin and a destination, i.e., an
O&D connection. Passengers booking transportation on different O&Ds
are considered to purchase different products. In a PP network, all passen-
gers on a specific flight have purchased the same product, since the net-
work does not permit connecting flights to a different destination. However,
in an HS network, flights usually ferry passengers who have purchased
completely different products. Imagine a Lufthansa flight from Zurich to Mu-
nich. Besides passengers flying only from Zurich to Munich, you may also
find passengers with a different origin, for example, Nairobi; you may also
find passengers whose origin is Zurich, but with a different end destination,
for example, Delhi. The economic advantage of HS networks over PP net-
works is that they allow the bundling of different products and hence reduce
the asset cost in offering these services (Wojahn, 2001b). Therefore, econ-
omies of scope can explain the formation of HS networks for airlines with
a traditional cost structure.
The bundling effects of HS networks also have a positive impact on the

demand side – the so-called network effect. Customers are more likely to
choose an airline that offers a greater variety of products (all possible
O&Ds); thus, customer retention becomes easier (Pels, Nijkamp, &
Rietveld, 2000, pp. 431–432). This positive demand network effect is also
the primary driver for alliance formation, whereby airlines can virtually grow
their network by offering code-sharing agreements with their alliance part-
ners (Pels, 2008, p. 73). The temporal coordination of flights in an HS net-
work also enables shorter connection times, thereby increases the attrac-
tiveness of the offered product (Goedeking, 2010, pp. 28–29). Several air-
lines use the quality of service index (QSI) to quantify these network ef-
fects. The QSI measures customer preferences for O&Ds, flight time, ser-
vice quality, and price as well as the disutility of delays or connections.
Based on these estimates, airlines can not only optimize the network struc-
ture but also the temporal coordination and used equipment for specific
routes (Goedeking, 2010, p. 29). Serving a HS network provides more flex-
ibility for network and schedule adjustments, since all the services are
routed through the same hub.
Other than cost advantages and demand stimulation, there is a third factor
explaining the historic development of HS networks. Concentrating traffic
on one or more hubs increases the competitive position of an airline at
these hubs. Since they provide a large share of the total traffic at these
airports, they enjoy huge market power over the airports and the associ-
ated service providers (e.g., ground operations, catering, etc.). Besides the
potential financial benefits, they often have a say on traffic right decisions
and gate allocations. This market power increases the barriers on entry in
the home markets of the HS airlines, whereby they can achieve a monop-
oly-like market position (Brueckner & Spiller, 1991; Oum, Zhang, & Zhang,
1995).
Establishing and developing a HS network also generates negative effects.

Since great traffic is routed through the hubs, these airports are prone to
congestion as they operate close to their capacity limit (Wojahn, 2001b).
Congested airports bear a higher risk of operational delays, endangering
the robustness of the network schedule. With the growing number of con-
nections, the operational complexity also increases. Passengers and their

Big Data To Improve Strategic-1

Uploaded by

Copyright:

Available Formats

Big Data To Improve Strategic-1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Big Data To Improve Strategic-1

Uploaded by

Copyright:

Available Formats

Schriftenreihe der HHL Leipzig

Graduate School of Management

Big Data to Improve

Reihe herausgegeben von

Weitere Bände in der Reihe http://www.springer.com/series/12648

Big Data to Improve

Dissertation HHL Leipzig Graduate School of Management, 2019

Schriftenreihe der HHL Leipzig Graduate School of Management

Leipzig, May 2019 Prof. Dr. Iris Hausladen

Berlin, May 2019 Maximilian Schosser

1.1 Problem and research gap definition............................................. 1

1.1.1 Practical problem ...................................................................1

1.1.2 Scientific research gap ..........................................................2

1.2 Objective of the study and research questions .............................3

2.1 Development of research design................................................... 6

2.2 Literature review ..........................................................................11

2.2.1 Design of a structured literature review process .................11

2.2.2 Identification of keywords, databases, and journals ............13

2.2.3 Results of the structured keyword search ...........................15

2.2.4 Description of the research gap ...........................................17

2.3 Comparative case study ..............................................................18

2.3.1 Selection of the case study type ..........................................19

2.3.2 Case sampling .....................................................................20

2.3.3 Data collection and analysis techniques ..............................23

2.3.4 Research quality assurance ................................................27

3.1 Development of a theoretical concept .........................................29

3.2 Airlines and their business models ..............................................31

3.2.1 The airline industry ..............................................................32

3.2.2 Development and recent trends in the airline industry ........34

3.2.3 Airline business models .......................................................35

3.2.4 Major business processes of airlines ...................................43

3.3 Introduction to the network theory ...............................................44

3.3.1 Definition of networks ..........................................................44

3.3.2 Distinction between the network theories ............................48

3.3.3 Fundamentals of the graph theory .......................................50

3.3.4 Network flows and network optimization .............................55

3.3.5 Design of flow networks .......................................................62

3.4 Airline networks ...........................................................................64

3.4.1 General properties of airline networks .................................65

3.4.2 Types of airline networks .....................................................67

3.4.3 Airline network economics ...................................................74

3.4.4 Airline network indicators .....................................................78

3.5 Network planning in airlines ........................................................84

3.5.1 Components of network planning ........................................84

3.5.2 Long-term planning ..............................................................87

3.5.3 Spatial optimization ..............................................................96

3.5.4 Temporal optimization .........................................................99

3.5.5 Operational optimization ....................................................103

3.5.6 Network planning in cargo airlines .....................................106

3.5.7 Data needs of network planning ........................................110

3.5.8 Definition of strategic network planning .............................124

3.6 Big data .....................................................................................129

3.6.1 Definition of data ................................................................129

3.6.2 Characteristics of big data .................................................130

3.6.3 The big data value chain....................................................132

3.6.4 Major types of big data ......................................................135

3.6.5 Big data types in airline and tourism research...................136

3.6.6 Issues and risks of big data ...............................................139

3.7 The RBV pertaining to data and network planning ....................140