![Brand categorisation model [31, p. 184]](https://www.researchgate.net/profile/Stefan-Klein-4/publication/308971242/figure/fig1/AS:503848269168640@1497138196363/Brand-categorisation-model-31-p-184_Q320.jpg)
Content uploaded by Stefan Klein
Author content
All content in this area was uploaded by Stefan Klein on Jul 10, 2017
Content may be subject to copyright.
Online Search Behavior in the Air Travel Market:
Reconsidering the Consideration Set and Customer Journey Concepts
Julia Andrea Jacobs,
Stefan Klein
Christopher Patrick Holland
Maximilian Benning
University of Münster
University of Manchester
University of Münster
[firstname.lastname]@uni-muenster.de
chris.holland@mbs.ac.uk
m_benn07@uni-muenster.de
Abstract
The online air travel market is a complex and dy-
namic multi-channel environment in which consumers
use a range of decision criteria to search for their best
flight options. Online Travel Agents and Price Com-
parison Engines have transformed the search process
and enhanced market transparency. These Air Travel
Intermediaries (ATIs) are sophisticated decision sup-
port tools that enable online search and booking
across thousands of flight options for all users, regard-
less of user expertise. An experiment was conducted to
explore the detailed search behavior and processes of
29 individuals. A revised model of the customer jour-
ney as search funnel and a different operationalization
of the consideration set is described that are more re-
alistic representations of actual search behavior.
1. Introduction
The customer journey is a well-established model
in marketing [2,25,31]. It typically starts with a wide
array of purchasing options that is reduced through a
series of choices culminating in a small set of serious
options, termed the consideration set (CS). It is based
on concepts and consumer research methods that origi-
nated in the 1960s [16]. In order to study online cus-
tomer journeys, we used a search experiment for a spe-
cific product, scheduled flight tickets, as a case to re-
examine this model. We chose the airline market be-
cause it is economically important, it is highly complex
because of the number of permutations generated from
search options and the bulk of the market is now trans-
acted in the online channel.
The air travel market has distinct characteristics,
which profoundly affect the customer journey. Air
travel can be depicted as a global network of 3,275
airports (nodes) and approximately 60,000 routes (edg-
es) [28]. Flights are defined by a departure and arrival
airport and can be either direct connections between
the two or combinations of several flight segments.
These connections can be operated by one or several
airline brands, the latter operated by alliance partners
or based on IATA interlining agreements [32]. If indi-
rect connections are included, there may be several
thousands of possible combinations of routes and air-
lines for a single flight. In order to process a query for
a return flight from Frankfurt, Germany, to Honolulu,
HI, in the US, Opodo.com claims to have screened
more than 1 million flight options.
Seat capacity on every flight is constrained and the
airlines use dynamic pricing as part of a yield man-
agement strategy [21]. As flights are offered globally,
available options and prices can change continuously.
Therefore, decision support tools are needed in order to
make this complex option space not only manageable
for ordinary consumers [21], but also to give them the
assurance to have identified the best flight option to
suit their needs and preferences.
Three main types of competitors operate in this
market: the airlines, and two types of Air Travel Inter-
mediaries (ATI), which are Online Travel Agencies
(OTAs), e.g. Expedia, Travelocity and Orbitz, and
Price Comparison Sites (PCS), e.g. Kayak and Google
Flight, which also support search and comparison but
provide online referrals rather than booking functional-
ity. All three use powerful algorithms to provide con-
sumer decision support in terms of filtering flight op-
tions based on from, to, date, availability and price.
The results, depicted as a matrix or list, can be further
refined by relaxing search criteria, e.g. flexible dates
and adjacent airports or constraining the options to
specific airline brands, time of day, direct flights only
and maximum price. OTAs and PCSs offer a direct
comparison across a wide range or even a comprehen-
sive set of airlines, whereas the airlines only list flight
connections offered by them, which includes code
sharing and interlining with other airlines to provide a
connection. Global Distribution Systems (GDSs) link
the airlines’ reservation systems and provide the back-
bone for travel intermediaries to access real time in-
formation about routes, schedules, availability, and
fares and to automate reservations with the airlines.
GDSs therefore enable ATIs to provide market trans-
3940
Proceedings of the 50th Hawaii International Conference on System Sciences | 2017
URI: http://hdl.handle.net/10125/41636
ISBN: 978-0-9981331-0-2
CC-BY-NC-ND
parency [4, 32] and an easy and efficient way for con-
sumers to compare airline brands, product characteris-
tics and other search criteria related to the flight op-
tions available [8,10,15,19]. Consequently, ATIs have
established a dominant position in the online travel
market [34]. While the airlines initially supported ATIs
as distribution partners, they have more recently in-
creased the level of competitive manoeuvers, such as
enforcing the use of airline specific booking software
or introducing booking fees for intermediaries [14].
Therefore, our first research question is: Which web-
sites do consumers use for travel search, given the
online multi-channel distribution environment?
As air travel is an experience good, airlines are try-
ing to establish their brands based on the quality of the
customer experience throughout the entire service pro-
cess: from booking until the luggage is delivered. Key
features are flight schedule and connections, punctuali-
ty, safety, quality and amenities of the aircraft, in-flight
catering and entertainment [1]. However, brand prefer-
ence and for that matter brand competition only matter
when customers actually have a choice among availa-
ble flights offered by different airlines. Whether this is
the case, can only be established based on the specific
customer query and the dynamically generated list of
options available at the time of the query. Although
there may be an enormous number of theoretically pos-
sible flight options, these are filtered to a much smaller
set of options, which actually meet the customers’
baseline search criteria in terms of route and availabil-
ity, which then can be further refined by adding addi-
tional search criteria.
Air travel is a service, for which search and book-
ing occur in the “pre-consumption” phase [11,32,33].
Multiple characteristics of this service, such as origin,
destination, date and time of flight, preferred carrier,
class of travel, total journey time, seat and luggage
options all make the search and booking process a
complex and time consuming activity [19,21], in par-
ticular if they are carried out online by the consumers
themselves [22]. This leads us to the second research
question: How do consumers search online, given the
specifics of the market and the product characteristics?
The Consideration Set (CS) concept is an important
link with earlier research in a pre-Internet era and also
with studies that focused on online search behavior.
The CS has generally been measured by counting the
number of brands that the consumer actively evaluates
in their search process. In this research, the CS is re-
considered in an ATI-enabled online environment. It is
our contention that the use of online intermediaries
fundamentally changes the online search process to
such an extent that we need to model this behavior in a
more realistic and specific manner.
Previous research has investigated search breadth
and depth [20,30] including work specifically in the
airline travel market [15,18,35]. For example, Johnson
et al. [18] found relatively low search levels and noted
a distinctive power-law decay that describes the distri-
bution of the number of websites visited during the
search process. That is, a large number of users visit 2
or 3 websites, and this decays quickly so that only a
small percentage visit 4 or more websites. More recent
research into the customer journey has used online
panel data to explore the search behavior of large
groups of individual consumers [15,18,35]. Whilst this
approach is very powerful at giving a broad overview,
it is not possible to uncover the detailed search behav-
ior on the travel websites in terms of search paths and
multi-criteria search queries [8], at least not with
standard online panel data reports. But this is a crucial
part of the search process, because it is where consum-
ers make their decisions and filter the very wide range
of options down to a much smaller CS that is then
evaluated to make more detailed head to head compari-
sons [10,29]. We address this gap in the literature by
taking an experimental approach, where we are able to
investigate the search process within the airline web-
sites and also within the ATIs, so that we are able to
develop a more sophisticated and realistic measure of
search behavior and also to measure the CS by tracking
individual behavior within the ATI.
Öörni [26] conducted an experiment for airline
ticket search behavior and found that participants con-
sidered large numbers of alternatives and conducted
several search refinements on intermediaries. Our aim
is to do a modified replication of Öörni’s [26] study,
and to investigate flight ticket search in a more ad-
vanced and complex market. The purpose of this study
is to adapt and extend the traditional customer journey
model and measurement of CSs in the airline market.
An experimental approach allows us to uncover im-
portant details of the search process within websites
that have previously been represented as ‘black boxes’
within a search sequence. The detailed behavior is im-
portant because an ATI may actually constitute the
whole of the online customer journey, and even a sin-
gle airline website may contain several brands if it is
part of a group or alliance of airlines. This internal
search is therefore more representative and realistic of
the search process and gives us a more nuanced meth-
od for measuring CSs. Hence our third research ques-
tion is: How large is the number of actually considered
flight options and airline brands (consideration set size)
and how is it determined?
The structure of the paper is as follows. A review
of the literature on the online search process is fol-
lowed by a description and explanation of our method-
ology. The results from the experiment are presented,
3941
followed by a discussion of the results and conclusions,
including the research contributions, limitations and an
outline of future research opportunities.
2. Literature review
The literature review introduces brand categoriza-
tion as a relevant marketing concept and air travel spe-
cific literature on search and consumer decision sup-
port.
2.1. Stages of the online search process
Search behavior is an important stage in the cus-
tomer journey [22]. The consumer runs through the
process of evaluating available choices, which is
termed brand categorization [2,25,31]. This traditional
model composes a linear process with different search
stages that are distinct from each another. Consumers
are not able to be aware of, nor to evaluate all the
brands available on the market, therefore need to clas-
sify relevant options, so that a funneling process takes
place and certain subsets are identified [5]. Shocker et
al. [31] developed a theoretical framework for search
behavior, focusing on the funnel process, where the set
size is reduced in stages (see figure 1).
Figure 1. Brand categorisation model [31, p. 184]
In practice, it is very difficult to discern the very
nuanced difference between CS and choice set, and in
this paper we use the term CS for the sum of those
brands that are actively considered in the search pro-
cess, and from which the choice of flight is made. All
possible flight options constitute the universal set. An
individual consumer will only be aware of a sub-set of
the universal set, which is termed the awareness set
[31]. In the traditional model, the consumer’s (pre-
existing) awareness set plays a crucial role in determin-
ing whether or not a particular competitor is considered
at all. That is, if the consumer is unaware of the com-
petitor, or does not come across them in their search,
then the competitor is excluded from the search pro-
cess. But in an online environment, ATIs are a power-
ful aid to the search process and can generate a vast
range of possible options, dependent on some simple
initial search criteria and not the pre-existing
knowledge of the consumer. ATIs therefore change the
nature and scope of the initial stages of the search pro-
cess by opening up the search process to almost all
airlines and flight permutations and provide a sophisti-
cated method for filtering out and evaluating this very
large possible set of options down to a manageable
quantity. The awareness set concept is therefore argua-
bly replaced by the results of the baseline query on an
air travel site. After consumers enter the baseline
search criteria of origin, destination and time, the ATIs
and airline websites generate an initial solution space
that contains the available options, which satisfy the
baseline search criteria. The solution space is the result
of algorithmic filtering by one or more websites, and
therefore varies depending on the site(s) used. Specifi-
cally, if an airline Web site is used, the resulting solu-
tion space only represents matching flights offered by
this airline and its partners. The term space is used to
denote the idea of flexibility and ongoing refinement
during the search process. In addition, “space” illus-
trates the inclusion of further flight options and air-
ports e.g. in addition to previously included airline
brands only. The evaluation of the alternatives is an
iterative process of reviewing options, adjusting the
filter criteria, and sorting the results list, e.g. by price
or flight duration.
The next subset is the consideration set [13, p.393]
‘the theoretical construct of a consideration set is those
brands that the consumer considers seriously when
making a purchase and/or consumption decision’, also
formerly known and introduced as the evoked set [16].
Shocker et al. [31, p.183] continue with a choice set to
go deeper into set diminishment, explaining it as a ‘fi-
nal consideration set’, however, this differentiation is
not important for our purposes.
The consumer will eventually choose a flight from
a CS based on a specific set of search criteria that have
evolved over time. The CS is defined as the number of
brands contained within the set of flight options that
are actively viewed and considered by the consumer.
The flight choice is taken from the CS in the last re-
finement iteration of the search process. It has been
shown that ATIs reduce the CS size compared to par-
ticipants not using search tools [12,29]. The number of
visits to airline websites, i.e. brands, is therefore not a
reliable metric. Parra and Ruiz [29] challenged the
traditional concept of the CS and proposed a change
3942
that takes into account the nature of the online search
process.
2.2. Search behavior and decision support
across intermediaries and airline websites
Flight product characteristics, e.g. long- and short-
haul flights, and search criteria such as origin, destina-
tion, date and time of flight, preferred carrier, class of
travel, total journey time, seat and luggage options all
make the search and booking process a complex and
time consuming activity [19]. Due to the complexity of
the airline market [21], customers face an information
intensive process where they need to make extensive
decisions [27]. Customer decision support tools have
developed and improved very quickly to such an extent
that they have transformed the search experience [6,8]
and now significantly shape the airline ticket search
process [21,34]. They substitute the role of the pre-
Internet role of a travel agent [27, 32]. With easy web-
site designs [21], ATIs simplify and reduce the search
load [10,12,29]. It is now the consumer who can re-
duce the number of available flight options by defining
search criteria in an ATI. The power of the decision-
making has therefore shifted to the individual [24,27].
Based on prior research, we think that ATIs will be
increasingly used to structure the initial stages of the
search process. Clemons et al. [8] describe those online
intermediaries as major interaction points for a con-
sumer for online flight ticket search. Holland et al [15]
demonstrated that there is a clear interaction between
ATIs and direct search, and that the use of an ATI
stimulates direct search. They also found that approxi-
mately two thirds of consumers used an ATI, either on
its own, or in combination with direct search. Value
conscious customers may visit multiple ATIs because
there continues to be price variance across online in-
termediaries [6,7,8,21].
Our expectation is that consumers will use an ATI
to shape their solution space and also to refine it. Di-
rect search with airlines will only be used to gain very
detailed information about a particular flight(s), i.e. as
a further refinement of the ATI search results. For ex-
ample, this could be to gain information about seating
arrangements, stopover times, frequent-flyer programs,
to confirm brands and to add specific in-flight or air-
port services [6].
2.3. Flight ticket search criteria
The search process model shown in table 3 captures
the complexity and characteristics of the search pro-
cess, in which the consumer defines the brand catego-
rization process, or customer journey, through an itera-
tion of entering search criteria, evaluating results and
then refining the search criteria until a CS is reached
and a choice is made.
The air travel Web sites follow a dominant design,
which requires that all flight searches start with a base-
line search of origin, destination and date. The solution
space is then algorithmically generated. The search can
then be further refined using additional criteria,
[8,10,19], thereby shaping the solution space through
iterative refinements and in the detailed evaluation of
flight options and brands. We argue that an important
outcome of extensive use of ATIs is that the algorith-
mic generation of the solution space is reducing the
influence of prior knowledge such as an awareness set,
in the early part of the search process. If consumers do
not use ATIs, then the traditional brand categorization
models and customer journey models are much more
appropriate because they take into account the im-
portance of the awareness set in the search process.
3. Methodology
An experimental approach was chosen in order to
be able to explore, document and evaluate the search
process in a controlled environment [12,17,29]. Mi-
crosoft Expression was used to video the computer
screens of the participants. In this way, we were able to
compare search efforts, measure search sequences and
patterns, and ask the search subjects to explain their
search strategies. This enables the researcher to gain a
good approximation of real consumer behavior in a
realistic setting. The experiment is a modified replica-
tion of Öörni´s [26] experiment to make it comparable
and to rely on a proven method to investigate online
search behavior [29].
3.1. Experiment
A within-group laboratory experiment was con-
ducted with 29 undergraduate students enrolled in an e-
commerce course at a German university. The male:
female ration was 76:24 and the age range was 20 to 28
years. The experiment was organized as an integral
part of the course program but participation was volun-
tary and no credits were awarded. Participants were
asked to search online for a suitable one-way, long-
haul flight from Frankfurt (FRA) to Honolulu (HNL)
with a desktop computer. They were instructed to find
the best option without any further specification. They
were free to use any website they liked. Their search
behavior was documented by a video recording of their
computer screens. Before they started their search task,
participants used an ID to anonymize the recording and
the researchers certified the anonymous use of data.
After the task the students were asked to fill out an
3943
online questionnaire. Students were able to break off
the participation at any time. 29 videos were recorded
of which 24 were used for the analysis. Five of the
recordings were incomplete and unusable. Students
were given 30 minutes to complete their tasks and this
was enough time for all participants to finish the as-
signment.
We chose to design the experiment as a modified
replication of Öörni´s [26] study to focus on product
characteristics and search criteria influencing the
search pattern for online air travel behavior. Com-
pared to Öörni´s [26] study, the number of participants
was lower (29 compared to 92 students in the original
study). Although students were not awarded any course
credit for their participation, they were given a symbol-
ic token of appreciation, sweets, for their participation.
Hawaii was used again as a destination and search cri-
teria in terms of CS size were measured. Furthermore,
we have focused on online search only and the search
patterns, specifically which websites are used to con-
duct their searches and which search criteria are im-
portant regarding which search scenario. Öörni [26]
focused on the evaluation of the results based on a
comparison of traditional versus electronic markets and
did not explore the consumer search process in detail.
In our experiment we focused exclusively on the nature
of the search process, the search sequence as well as
search criteria used by the participants.
The videos were analyzed and participants’ search
activities were coded into a spreadsheet. More precise-
ly, the search patterns (start and end point), the search
sequence and their final booking page were tracked
and their importance related to search criteria. The
number of flight options and the CS were measured. If
a searcher used more than one ATI, then the results
were combined to establish the number of brands in the
CS. This approach is consistent with the technical defi-
nition of a CS [3]. If a flight option is code shared be-
tween different brands, only the first airline was ac-
counted as a brand. Finally, the participants were asked
to rank their three most important search criteria relat-
ed to the flight option chosen.
3.2. Customer journey and brand categoriza-
tion
The extensive use of ATIs makes it difficult to
evaluate the actual CS of searchers because they are
exposed to a large number of brands, presented in
some form of comparison list [34]. Whenever deci-
sions support tools (ATIs) are used, the identity and
overall number of considered brands becomes a black
box to researchers using online panel data. However,
with our experiment we aimed to open this black box
and investigate the search process in sufficient detail to
be able to count the number of brands that are consid-
ered when consumers use ATIs. The search scenario
assigned to the participants in the experiment defines
the baseline criteria, which yields the initial solution
space. This initial solution space can be refined during
several iterations by the participant into a refined set of
flight alternatives. Flight options are the sum of search
results displayed on all websites visited by a partici-
pant. Flight options are analyzed for the funnel pro-
cess, however, they are not relevant in determining the
CS size.
In the context of online air travel, the term CS be-
comes potentially ambiguous. One must first distin-
guish between flight options and then brands. One air-
line brand might offer several flight options and we
argue that to maintain consistency with earlier re-
search, then the number of brands is the key measure.
However, should the number of brands in the solution
space be counted? Or is it more realistic to analyze the
number of brands that are actively considered? The
latter approach is taken but even here one must inter-
pret the meaning of the term active in an online search
context where customers are scrolling through pages of
search results. We have measured the CS by counting
the number of airline brands visible on the Web pages
displaying the final sets, including scrolling and mouse
movements and are thus regarded as actively consid-
ered within the ATI and the airline Web sites, in case
they have been used as well. This metric is consistent
with the theoretical definition of a CS [3] but is argua-
bly a broader measure than previous measurements that
used online panel data and only counted a brand if the
user visited the actual website [15, 18, 35]. There are
two differences that have to be kept in mind: (1) code-
sharing flight options are only counted once and as-
signed to the main brand; (2) some flights, especially
long-haul flights are offered jointly by different airlines
in order to be able to cover the route. In that case, only
the main brand is counted.
4. Findings and Analysis
The presentation of findings follows the sequence of
our research questions.
4.1. Choice of Web Sites for the Search
Search for available flights can be conducted in two
ways: direct search on airline websites or by using
online intermediaries, i.e. ATIs, which can be used
separately or in combination. The results of online
search patterns are shown in table 1.
Participants preferred ATI for their search and of-
ten combined different comparison websites during
3944
their search session. Some of the participants started
their search on a general search engine rather than
looking for a travel site. The search engine listed links
to flight comparison sites and a few airlines, which
prompted the user to continue their search. They then
continued with an ATI or airline website. Some users
then used ATIs only or combined it with an airline
website during their search sequence and finished ei-
ther on an ATI or airline website. The majority of the
participants (92%) started their search on ATIs and
continued on them only without considering an airline
website (71%). 88% of the participants terminated their
search on an ATI. Airline websites were less frequent-
ed: only four concluded their search on an airline web-
site but did not book on the latter. While seven partici-
pants (29%) visited airline websites at some stage dur-
ing their search, none of the participants searched on
airline sites only. These findings are contrary to Hol-
land et al. [15], who reported that a third of German
consumers searched only on airline websites. However,
one needs to take into account the difference between
an experimental survey and an online panel used by
Holland et al. [15].
Table 1. Airline search pattern on decision support
tools and brand websites
Search pattern Website
# of par-
ticipants
%
Start search on …
ATI
22
92
Airline
2
8
Searching on …
ATI only
17
71
ATI and Airline
7
29
Start website : end
website
ATI : Airline 1 4
ATI : ATI
21
88
Airline : ATI
2
8
Total
24
100
4.2. Flight ticket search behavior
Participants showed an ongoing refinement process
of the search task by either modifying criteria (e.g.
different airport, other dates) or further constraining
criteria (e.g. earlier arrival time). Hence, their search
was a mixture of browsing or screening the results
page, refining and adding search criteria.
The search task defined baseline criteria: route
(FRA - HNL), travel period in terms of departure and
arrival time. Further search criteria were at the discre-
tion of the participants. In the post-experiment survey,
participants ranked their three most important search
criteria (table 2). The most important criteria are the
ticket price, closely followed by the flight duration and
the number of connections or stopovers. This confirms
Collins et al.’s [10] results. Flight departure and arrival
time, the attractiveness of the transit airports as well as
airline brand and the layover were less frequently men-
tioned. As a second most important factor, the top three
ranks are price, the number of connections and the
transit airport.
Table 2. Flight selection criteria
Selection criteria
Most im-
portant cri-
teria (%)
2nd most
important
(%)
3rd (%)
Price
52
26
20
Flight duration
18
16
40
Number of stops
9
26
40
Departure/ arrival
time
6 6 n/a
Transit airports
(stopover)
6 21 n/a
Airline brand
6
5
n/a
Layover
3
n/a
n/a
N
33
19
5
The combination of criteria suggests that travelers
have to make complex trade-offs, such as: how much
more am I willing to pay for a more convenient flight
in terms of overall travel time or times of departure and
arrival; or how much more am I willing to pay for a
different airline operating my flight?
The number of participants mentioning criteria de-
creases with the ranking as it was not mandatory to
select three options. As we chose students as partici-
pants, we obviously have selected a price-sensitive
customer segment. However, the high ranking of price
as selection criterion for flights is representative for a
large customer segment [1]. The baseline criteria have
been chosen to simulate a complex flight choice: direct
flights are not available for the route FRA – HNL,
travelers have the choice of taking the eastern route
(via Asia) or the Western route (via North America).
4.3. Flight options and airlines considered
The experiment design allowed us to capture the re-
finement of selection criteria and at the same time to
document the size of the initial and refined solution
spaces as well as the CSs. Table 3 shows search figures
of flight options at different stages of the funnel.
Some participants started their search on search en-
gines, with origin and destination details only and no
time restrictions (not necessarily needed), giving a high
‘origin - destination solution space’ of 8,777 flight
alternatives. This number represents the total number
of flight options for the route Frankfurt to Honolulu.
The search engine had assumed a random date (Feb.
13) and counted code-shared flights as well as flight
sold via tour operators as distinct options.
By entering all the baseline criteria with time re-
strictions, an average number of 368 flight options are
3945
available in total in the ‘origin, destination, time solu-
tion space’. All participants conducted several (at least
3) refinement iterations. The refinement stages were
either conducted on the chosen intermediary website or
other websites were selected for comparison. The flight
options during the refinement process were counted at
every iteration step. They have been decreasing from
326 to 76 flight options on average. Finally, the partic-
ipants considered on average 23 flight alternatives dur-
ing the search process and considered 11 airline
brands.
Table 3. Flight ticket search funnel
Adapted
brand cat-
egorization
Metrics and com-
ments
Experimental
results [FRA –
HNL]
Universal
set
a) All flight options
b) All airline brands
a) Appr. 60,000
routes, millions
of combinations
of routes
b) Appr. 6,000
airlines [28]
Initial
solution
space
a) Origin, destina-
tion, default time
(set by site)
b) Origin, destina-
tion, time
a) 8,777 flight
options
b) Avg. 368 flight
options
Refined
solution
space
Stepwise refinement
of search criteria
Avg. 326 – 76
flight options
Considera-
tion set
a) Flight alternatives
considered
b) Airline brands
considered
a) Avg. 23 flight
options
b) Avg. 11 brands
An example of average number of flight alterna-
tives of the participants at each search stage is given
and demonstrates the intensity of the filtering process,
reducing the available flight options from a total of
8,777 to an average of 23 flight options for the con-
sumer. In order to measure the search breadth for flight
tickets, available flight options and CS size were ob-
served for each participant. The following table 4 gives
an overview.
A total number of 24 different airline brands were
operating (segments of) the flight route from Frankfurt
to Honolulu and could have been chosen by the partic-
ipants. For each participant, ranked by ID, the total
number of flight options considered as well as his or
her CS size were calculated based on the video capture
of their scrolling and mouse behavior. The number of
flight options are calculated based on the number of
flight options displayed on various ATIs and airline
websites considered by each participant. The total
range of all the flight options considered is 8–46. On
average, participants considered 22.5 flight options
with a standard deviation of 9.75. Out of these flight
options, all airline brands considered are forming the
CS size with a total range of 5-15 brands. On average,
participants considered 11 airline brands with a stand-
ard deviation of 2.97.
Table 4. Consideration set (CS) size and number
of flight options
All available brands (N=24)
Aer Lingus, Aeroflot, Air Berlin, Air Canada, Air China,
Air France, American Airlines, Asiana Airlines, British
Airways, Cathay Pacific, China Eastern Airlines, Condor,
Delta Air Lines, Ethiopian Airlines, Iberia, Icelandair,
Japan Airlines, KLM, Korean Air, Lufthansa, SAS,
Swiss, Thomas Cook, United Airlines
ID
Number
of flight
options
CS
size ID
Number
of flight
options
CS size
1
37
12
13
35
15
2
27
12
14
18
9
3
20
6
15
11
5
4
20
8
16
16
5
5
25
12
17
21
8
6
29
7
18
38
12
7
14
10
19
17
8
8
8
5
20
24
14
9
37
12
21
15
7
10
10
5
22
16
7
11
46
9
23
15
5
12
21
8
24
20
10
5. Discussion
Our aims for this paper were to investigate charac-
teristics of the online search process for air travel
products shaped by the complexity of the air travel
market and competition in the online distribution sys-
tem (1), search patterns for flight tickets in light of
thousands of available flight options and algorithmic
search and filtering tools operating on air travel Web
sites (2) and how the theoretical concepts of customer
journey and CS changed (3). While various methods
have been used previously to uncover search behavior
online [9,15], previous findings have provided little
insight into online search paths or search intensity. We
therefore conducted a laboratory experiment with stu-
dents in order to capture the customer journey in detail.
By using an experiment, the researchers were able to
open the black box of flight search behavior and
demonstrate the need for theory concept development
and adaptation, particularly for the sales funnel and
consideration set. The airline market is different to
other markets and its specificities influence the need
for concept adaptation of the customer journey and CS.
The nature of the search process, flight search behavior
and criteria used are linked and explain the importance
3946
of being considered in combination to gather insights
into consumer behavior.
5.1. Channel competition
Online ATIs have a dominant role, as they are an
obvious choice to gather an overview of available
flight options. Across all air travel sites, the first stages
of the search process are quite similar with online
forms to enter the baseline criteria. During the search
process, the ATIs have a clear advantage in terms of
transparency. Only if the airline (brand) has already
been chosen, the airline sites are advantageous as they
generate much smaller and therefore less complex op-
tion spaces. Once the choice for a flight has been
made, airline websites provide the advantage of highly
differentiated product choices (e.g. booking class, lug-
gage requirements, leg space, ticket changes and can-
cellations fees). Lufthansa recently introduced a book-
ing fee for OTAs, which provides a disincentive for
travelers to use OTAs for the booking process [14].
The complexity of options and the variation in pric-
es and fees provide a plausible explanation for the ob-
served “site hopping”: all participants checked alterna-
tive sites in order to see whether they would encounter
different flight options or price differences.
5.2. Algorithmic search and consumer decision
support
When a consumer starts to search for a flight, they
typically have very little idea what the solution space
will look like. Even when they are familiar with routes
and schedules, the availability of seats and the airfare
are only retrieved at the time of search. Hence, search
and comparison algorithms (“engines”) determine
online search and are used by all air travel sites.
Based on the baseline criteria the algorithms dy-
namically generate the solution space, which is a tran-
sient list, as availability of seats (in a particular book-
ing class, i.e. price category) can change any second.
Some sites even signal the number of available seats at
the listed price. Depending on the characteristics of the
initial solution space, the consumer can expand the
search by relaxing the baseline criteria or reduce the
solution space, by adding constraints, such as number
of stops, time of departure, or preferred airline(s). Our
results show an alternating process of screening results
and refining parameters, typically three rounds of re-
finement. Öörni [26] found that consumers conducted
several search iterations, corroborating our results.
Once the final solution space has been established,
the remaining options are examined (“actively consid-
ered”), possibly sorted based on decision criteria and
the participants consider tradeoffs between their deci-
sion criteria. We have surveyed our participants and
asked for up to three ranked decision criteria, which
revealed different preference profiles. While the major-
ity named price as the most important criterion, facets
of convenience (flying time, number of stops and layo-
ver) scored highly in particular as second and third
criteria. The transit airport was considered by 21% as a
second criterion, which may reflect the condition of a
long haul flight and the option for an extended stopo-
ver.
The online search process shows the salience of the
decision support tools offered by the travel Web sites
in terms of filtering, refining search and sorting results.
The consumers refine the search by using a simple
interface, but in the end have to make a choice them-
selves based on their preferences and related trade-offs.
Given the transient nature of the solution space, some
sites offer the consumer to literally buy time, by tem-
porarily putting flight options on hold, in order to al-
low time to think, consult or substantiate a tentative
choice.
5.3. Refining brand categorization
Based on our analysis and findings we are suggest-
ing two amendments to the brand categorization model
or search funnel:
(1) Given the prominence of online consumer deci-
sion support tools and the product characteristics of air
travel we propose to replace the awareness set (an in-
dividual’s representation of the brand space) by the
concept of the initial solution space: algorithmically
generated, currently available options meeting the con-
sumer’s baseline criteria. Brand is only included as an
initial selection criterion in case the travelers start their
search on an airline website. The notion of solution
space includes a process view of sequential refine-
ments until a final solution space has been identified,
which more accurately reflects consumers’ prefer-
ences.
(2) The final solution space comprises the total
number of flight options that meet the customer’s de-
tailed choice criteria, which we propose to use as the
basis for counting the number of brands in the CS of
the actively considered flight options, evidenced by
scrolling, mouse movements or clicking on options.
CS size and number of flight options considered are
higher compared to prior studies [15,18,26,35] as a
result of our methodological approach. Prior studies
have operationalized the CS as the number of airline
websites visited. These results are specific to the airline
market and take a broader view of active consideration.
3947
6. Conclusion and limitations
Our study aimed at opening the black box of online
search for scheduled air travel. Based on a thorough
contextualization of the air travel market, of the com-
petition in the online multi-channel distribution system
and of the characteristics of air travel services, we have
created video protocols of realistic search processes.
The experiment revealed detailed search paths within
and across air travel sites and confirmed the expected
prevalence of ATIs in the early stages of the search
process, as search algorithms and the related consumer
decision support tools on ATI websites have consider-
ably improved over the past years. The tools have be-
come more flexible, including a higher number of pos-
sible search and filter criteria and providing higher
levels of transparency by covering all flight options
(including low cost airlines) and allowing for extend-
ing and constraining the solution space. While earlier
work [21,26] had documented the challenges for con-
sumers to search online for complex, long-haul flights,
our participants did not encounter substantial difficul-
ties and all identified suitable flight options within a
reasonably short period of time. The results also sug-
gest that airlines should focus on the booking process
and complement the ATIs, rather than compete with
them head-on during the early stages of search.
The experiment also provided insights into the dy-
namics of search, which we have divided into three
phases: (1) filtering and refining search criteria, (2)
comparing and examining the CS and (3) eventually
deciding based on trade-offs between preferences. It
revealed the key role of the decision support tools but
also their limitation when it comes to supporting the
final decision. The trade-offs might be addressed by a
new generation of decision support tools. The results of
our experimental findings suggest a revision of the
brand categorization model, specifically for online
search of airline tickets and methodological enquiries,
which reveal the flight options and brands considered
by the consumers. The amendments relate to the re-
placement of awareness set by a dynamically refined
solution space and a distinct operationalization of the
CS based on observations of the active consideration of
flight options.
Our research design has obvious limitations in
terms of the number and demographics of the partici-
pants. Yet, despite the widespread concern about stu-
dent participants in experiments, they represent a tech-
nologically savvy, but relative inexperienced customer
segment, which is quite suitable for the purpose of our
study. Laboratory experiments are not executed in a
natural environment that might represent participants’
natural search behavior as other social factors also in-
fluence the decision [26]. We are planning to extend
the sample size and the data capture by including
speak-aloud components in a future design. We will
use our findings from this study as part of a larger re-
search program, which includes a synthesis of online
panel data, weblog analysis and experiments.
7. References
[1] S. Aksoy, E. Atilgan, and S. Akinci, “Airline services
marketing by domestic and foreign firms: differences from
the customers’ viewpoint”, J. of Air Transport Management,
9, 2003, pp. 343-351.
[2] J. Brisoux, and M. Laroche, “A proposed consumer strat-
egy of simplification for categorizing brands”, In: Evolving
Marketing Thought for 1980, J. H. Summey and R. D. Taylor
(eds.), Southern Marketing Association, Carbondale, IL.
1980, pp. 112-114.
[3] J.J. Brown, and A.R. Wildt, “Consideration set measure-
ment”, J. of the Academy of Marketing Science, 20(3), 1992,
pp. 235-243.
[4] D. Buhalis, and M.C. Licata, “The future etourism inter-
mediaries”, Tourism Management, 23(3), 2002, pp. 207-220.
[5] B.M. Campbell, The Existence of Evoked Set and Deter-
minants of its Magnitude in Brand Choice Behavior. Un-
published Ph.D. dissertation, Columbia University, 1969.
[6] J. Chen, “Differences in average prices on the Internet:
Evidence from the online market for air travel”, Economic
Inquiry, 44(4), 2006, pp. 656-670.
[7] W-C. Chiou, C-C. Lin, and C. Perng, “A strategic website
evaluation of online travel agencies”, Tourism Management,
32(6), 2011, pp. 1463-1473.
[8] E.K. Clemons, I.-H. Hann, and L.M. Hitt, “Price disper-
sion and differentiation in online travel: An empirical inves-
tigation”, Management Science, 48(4), 2002, pp. 534–549.
[9] A.T. Collins, J.M. Rose, and S. Hess, “Search based in-
ternet surveys: airline stated choice”, Working paper ITLS-
WP-10-01, ITLS Sydney and ITS Leeds, 2010, pp. 1-15.
[10] A.T. Collins, J.M. Rose, and S. Hess, “Interactive stated
choice surveys: a study of air travel behavior”, Transporta-
tion, 39(1), 2012, pp. 55-79.
[11] U. Gretzel, D.R. Fesenmaier, and J.T. O’Leary, “The
Transformation of Consumer Behavior”, In: Tourism Busi-
ness Frontier, D. Buhalis and C. Costa. (eds.), Oxford, UK,
Butterworth-Heinemann, 2006, pp. 9-18.
[12] G. Häubl, and V. Trifts, “Consumer decision making in
online shopping environments: The effects of interactive
decision aids”, Marketing science,19(1), 2000, pp. 4-21.
[13] J.R. Hauser, and B. Wernerfelt, “An evaluation cost
model of consideration sets”, J. of Consumer Research,
16(4), 1990, pp. 393-408.
[14] G. Hobica, “Four Airlines to add a fee for booking with
expedia, priceline, etc.”, Huffington Post Blog, 06.05.2015,
accessed 29th August 2016,
3948
http://www.huffingtonpost.com/george-hobica/four-airlines-
to-add-a-fe_b_7518076.html.
[15] C.P. Holland, J.A. Jacobs, and S. Klein, “The role and
impact of comparison websites on the consumer search pro-
cess in the US and German airline markets”, J. of Infor-
mation Technology and Tourism, 16(1), 2016, pp. 127-148.
[16] J.A Howard, and J.N. Sheth, The Theory of Buyer Be-
havior. John Wiley & Sons, New York, 1969.
[17] J. Jacoby, R.W. Chestnut, and W.A. Fisher, “A Behav-
ioral Process Approach to Information Acquisition in Nondu-
rable Purchasing”, J. of Marketing Research, 15 (Nov), 1978,
pp. 532-544.
[18] E.J. Johnson, W.W. Moe, P.S. Fader, S. Bellman, and
G.L. Lohse, “On the depth and dynamics of online search
behavior”, Management science, 50(3), 2004, pp. 299-308.
[19] S.H. Jun, C.A. Vogt, and K.J. MacKay, “Online Infor-
mation Search Strategies: A Focus on Flights and Accommo-
dations”, J. of Travel and Tourism Marketing, 27(6), 2010,
pp. 579-595.
[20] L.R. Klein, and G.T. Ford, „Consumer search for infor-
mation in the digital age: an empirical study of prepurchase
search for automobiles”, J. of Interactive Marketing, 17(3),
2003, pp. 29-49.
[21] S. Klein, F. Köhne, and A. Öörni, “Barriers to Online
Booking of Scheduled Airline Tickets“, J. of Travel and
Tourism Marketing, 17(2/3), 2004, pp.27-39.
[22] P. Kotler, G. Armstrong, J. Saunders, and V. Wong,
Principles of marketing, 5th European edition, London, Pear-
son, 2008.
[23] I. Lewis, J. Semeijn, and A. Talalayevsky, “The impact
of information technology on travel agents”, Transportation
Journal, 37(4),1998, pp.20-26.
[24] P. Lin, C. Chen, and M. Song, “Price dispersion of
online air tickets for short distance international routes”,
Service Industries Journal, 29(11), 2009, pp. 1597-1613.
[25] C.L. Narayana, and R.J. Markin, “Consumer Behavior
and Product Performance: An Alternative Conceptualiza-
tion”, J. of Marketing, 39(4), 1975, pp. 1-6.
[26] A. Öörni, “Consumer search in electronic markets: an
experimental analysis of travel services”, European Journal
of Information Systems, 12, 2003, pp. 30-40.
[27] R. O’Keefe, and T. McEachern, “Web-based customer
decision support systems”, Communications of the ACM,
41(3), 1998, pp. 71-78.
[28] Openflights. Airport Database. Accessed 29th August
2016, http://openflights.org/data.html.
[29] J.F. Parra, and S. Ruiz, “Consideration sets in online
shopping environments: the effects of search tool and infor-
mation load”, Electronic Commerce Research and Applica-
tions, 8(5), 2009, pp. 252-262.
[30] C.M. Schaninger, and D. Sciglimpaglia, “The Influence
of Cognitive Personality Traits and Demographics on Con-
sumer Information Acquisition”, J. of Consumer Research,
8(2), 1981, pp.208-216.
[31] A.D. Shocker, M. Ben-Akiva, B. Boccara, and P. Ne-
dungadi, “Consideration set influences on consumer deci-
sion-making and choice: Issues, models, and suggestions”,
Marketing Letters, 2(3), 1991, pp. 181-197.
[32] H. Werthner and S. Klein, “ICT and the Changing Land-
scape of Global Tourism Distribution“, Electronic Markets,
9(4), 1999, pp. 256-262.
[33] H. Werthner, and S. Klein, Information, technology and
tourism: A challenging relationship, Springer Computer Sci-
ence, Springer, Vienna, Austria, 1999.
[34] Z. Xiang, V.P. Magnini, and D.R. Fesenmaier, “Infor-
mation technology and consumer behavior in travel and tour-
ism: insights from travel planning using the internet”, J. of
Retailing and Consumer Services, (22), 2015, pp. 244-249.
[35] J. Zhang, X. Fang, and O.R. Liu Sheng, “Online con-
sumer search depth: Theories and new findings”, J. of Man-
agement Information Systems, 23(3), 2006, pp. 71-95.
8. Appendix: scenario description
Please select Google as your starting homepage.
Please enter your ID code into the search bar. Now read
the following scenario: Please imagine the following situ-
ation: You just finished your Bachelor studies and you
want to do a trip with a friend to Hawaii. You have a joint
friend who´s studying in Honolulu (Hawaii) and together
you want to continue the trip from there. As you are not
sure yet at what point you will fly back, you are only in-
terested in a one-way flight Frankfurt (FRA) to Honolulu
(HNL). You have arranged to meet your friend not later
than February 14th 21.00h (local time). For personal rea-
sons you cannot leave Frankfurt before Feb. 12th. Please
search online for flights which you would choose under
the circumstances. The choice of websites is yours. Please
document your final flight choice by taking a screenshot
that documents the flight no., day, time and price. Or,
write your final choice on the sheets provided. Just look
for the price of a simple, one-way economy ticket for one
adult, i.e. without extra luggage, insurance, special seat-
ing, loyalty programs etc. You will be paying by credit
card. Your friend has asked you to find and recommend
the best options, so please identify your favorite list of
one to three options. For each option, please continue to
search up to the point where you will be asked to make a
payment. (Please don´t make any real booking!). You
may be asked to enter your personal details (feel free to
invent information e.g. Examplestreet 5, 48159 City).
Please document the top 1-3 flight choices of your flight
search results (flight no., time, day, price) in a word doc-
ument or the table provided. Briefly explain your recom-
mendation including your preferences, i.e. your ranking,
in 2-3 sentences. (Expected time for scenario: 10-15 min).
3949
