consumers’ preferred criteria for hotel online booking
TRANSCRIPT
Consumers’ Preferred Criteria for Hotel Online Booking
Astrid Dickingera,
Josef Mazanecb
a Department of New Media Technology
MODUL University Vienna, Austria
b Institute for Tourism and Leisure Studies
Vienna University of Economics and Business Administration
Abstract
Online bookings of hotels have increased drastically throughout recent years. Studies in tourism and hospitality have investigated the relevance of hotel attributes influencing choice but did not yet explore them in an online booking setting. This paper presents findings about consumers’ stated preferences for decision criteria from an adaptive conjoint study among 346 respondents. The results show that recommendations of friends and online reviews are the most important factors that influence online hotel booking. Partitioning the importance values of the decision criteria reveals group-specific differences indicating the presence of market segments.
Keywords: hotel attributes, adaptive conjoint analysis, partitioning, vector quantization.
1 Introduction
The importance of the Internet as information medium particularly in tourism has
been identified and discussed by numerous researchers (Gursoy & McLeary, 2004).
However, it has long been stated that the Internet, yet, is not used to its full extent as a
platform for bookings and recommendations (Klein, Köhne & Öörni, 2004).
Nevertheless, current figures show that the purchases made online are multiplying. An
increase in online travel sales of 22% to EUR 46.8 billion is predicted during 2007
(Marcussen, 2007). The major part of online tourism sales, 56%, is generated by air
travel, hotels account for 16%, package tours 16%, rail 8% and rental cars 4%
(Marcussen, 2007). Online hotel booking, the second most important product in
online tourism revenue is the focus of this research. Hotels make use of the internet to
not only offer services to customers but to simultaneously provide a platform for
customers who give feedback on their stay in a specific hotel. As a result in addition
to the well known star ratings hotels get their individual quality ratings by former
guests. Previous research shows that word-of-mouth is a major driver for hotel
purchase decisions (Dubé & Renaghan, 2000a). Online reviews and online user
generated content is considered as a tool for word-of- mouth on the Internet. Others
found that for every second customer the star category is a basis for taking purchase
decisions (Callan, 1998). These findings must be interpreted cautiously as various
authors found that there are significant differences regarding those drivers in different
market segments (Lewis, 1984; Wind, Green, Shifflet & Scarborough, 1989; Mehta &
Vera, 1990). Research on online booking shows that the complexity of the travel
product hinders end user adoption of online booking systems (Klein, Köhne & Öörni,
2004). Most studies have focused on the attributes of the booking system such as
usability, system quality, service quality, usefulness, customer loyalty, and
information quality to assess factors influencing online booking behavior (Law &
Leung, 2002; Law & Leung, 2000; Law & Chan, 2004). Usability problems of hotel
websites have a negative effect on online bookings and revisit intentions while
increased depth of information and customer interaction have positive effects
(Essawy, 2006). In this research hotel not system attributes are investigated.
The aim of this study is to identify the critical factors that influence online booking
through an adaptive conjoint study. The decision criteria included are pictures of hotel
and hotel room, recommendations of a friend, review of the hotel, star rating,
location, and price.
The paper contributes to both, research and practice in several ways: (i) the relevant
attributes for online bookings are identified; (ii) the results indicate that there are
group-specific differences regarding the importance of these attributes, hence
strategies for communicating with these customer segments have to be developed
accordingly; (iii) from a research perspective the study shows how customer
heterogeneity might be handled; (iv) industry players get recommendation on how to
present themselves in an online booking portal. They learn whether or not to include
pictures and reviews and which elements to emphasize.
2 Literature Review
A number of researchers have examined attributes of hotels and their importance in
customers’ selection processes. A famous example is provided by Wind, Green,
Shifflet & Scarborough (1989). These authors demonstrate how to employ hybrid
conjoint analysis to develop “Courtyard by Marriott”. They included product attribute
evaluations and combined them with actual needs of customer segments leading to a
successful introduction to the market and, above that, elaborated a method that
allowed Marriott to design additional successful lodging concepts. Their study
included almost 200 facets in the seven attribute groups external factors, rooms, food-
related services, lounge facilities, services, facilities for leisure-time activities, and
security factors (Wind, Green, Shifflet & Scarborough, 1989). Hu & Hiemstra (1996)
also used hybrid conjoint analysis but focused on meeting planners and their choice
process. Among other popular methods to identify relevant attributes are
multidimensional scaling, multiple discriminant analysis (McCleary, Weaver &
Hutchinson, 1993; Lieux, Weaver & McCleary, 1994), regression analysis (Lewis,
1985; Richard & Sundaram, 1994), and comparison of means for gaining segment-
specific differences (Mehta & Vera, 1990; McCleary, Weaver & Lan, 1994; Callan,
1998; Dubé & Renaghan; 2000a). A detailed overview of important attribute
categories and the respective studies are itemized in the following table.
Table 1. Review of relevant hotel attributes
Authors and focus
of the study
Methods Number of
levels
Categories of attributes
Lewis, R. (1984)
Business and leisure
travelers.
Factor analysis,
analysis of variance
66 Service quality, overall feeling, security,
upscale services, food & beverage, image,
price/quality, aesthetics/décor/ambient,
amenities, beverage quality, room/bath
condition, health facilities, reputation, quiet,
room attributes, price and value, reservations
and front desk, location.
Lewis, R. (1985).
Business and leisure
travelers.
Regression analysis 17 As above.
Wind, Green,
Shifflet &
Scarborough (1989)
Potential travelers.
Hybrid conjoint,
Multidimensional
scaling, multiple
discriminant analysis,
ELASTICON (for
pricing).
Close to
200
External factors, rooms, food-related
services, lounge facilities, services, facilities
for leisure-time activities, security factors
Mehta, S. & Vera, A.
(1990). Individual
travelers.
Comparison of means 26 Security, room/bathroom/overall furniture &
decor, service in restaurants/bars/telephone/
overall/check in, reputation/image, business
center facilities, class appeal of hotel, low
price food and beverages, cleanliness,
variety of restaurants/bars, location, music
and entertainment.
McCleary, K.,
Weaver, P. &
Hutchinson, J.
(1993). Business
travelers.
Multiple discriminant
analysis
56 Basic product, business services, frequent
traveler programs, banquet/meeting
facilities, advertising/public relations,
location, non-smoking rooms.
McCleary, K.,
Weaver, P. & Lan,
L. (1994). Business
women.
MANOVA 53 Business services and facilities, security
facilities, basic facilities, personal services,
free extras, convenient eating facilities,
airline or hotel reward program, special
room features, airport or meeting hotel,
Authors and focus
of the study
Methods Number of
levels
Categories of attributes
price, advertising/parking, fitness facilities.
Richard, M.D. &
Sundaram, D.S.
(1994). Leisure
travelers.
Regression, factor
analysis
29 Reception, accommodation, departure, food
building, bathroom.
Lieux, E. M.,
Weaver, P.A. &
McCleary, K.
(1994). Senior
tourists.
Multiple discriminant
analysis
8 Budget, economy, luxury budget, midprice,
upscale, luxury, bed and breakfast or
country inn, family owned independent.
Hu, C. & Hiemstra,
S. J. (1996). Meeting
planners.
Hybrid conjoint
analysis
22 Price range, functional properties of meeting
rooms, conference planning procedure, guest
room comfort, food and beverage function,
location.
Bell, R.A. & Morey,
R.C. (1996).
Corporate Managers.
Logit model 9 Guaranteed last-room availability, flexible
cancellation, electronic interface, free local
calls, free breakfast, convenient distance
from workplace, free airport shuttle,
exceptional business amenities, special rate.
Callan, R. (1998).
Leisure and business
travelers.
Grading scheme 166 Location and image, price, competence,
access, security, additional services,
tangibles, service provider’s understanding
of the customer.
Dubé & Renaghan
(2000a). Leisure and
business travelers.
Frequencies 1275
drivers
12; Location, Value for money, brand name
and reputation physical property, guest-room
design and amenities, meeting-room design
and amenities, bathroom furniture and
amenities, service functional, service
interpersonal, food and beverage related
services, quality standards, marketing, other.
Dubé & Renaghan
(2000b).
Intermediaries.
Frequencies No
indication
Convenient location, value for money,
communication with intermediary, brand
name and reputation deals and incentives,
quality of services interpersonal &
functional, guest rooms, facility design.
The most frequently mentioned attributes that influence choice in leisure and business
traveler settings are: location, service, star rating, security, food and beverage, image,
price, room and hotel attributes, facilities for leisure time activities.
The majority of conjoint studies about hotel attributes tried to convey a full taxonomy
of all attributes of a hotel, which is not the purpose of this study. Only attributes
typically presented in an online booking setting will be included since the focus of the
paper is on factors driving online booking. A review of online booking platforms
reveals that the most commonly featured attributes are star rating, price, location/map,
photographs of the rooms and the hotel, evaluations of previous visitors, information
about the rooms, information about the hotel amenities and facilities, information on
the surrounding area (www.expedia.de, www.booking.com, www.hotels.com,
www.travelocity.com, www.venere.com, www.tiscover.com). Since online booking is
investigated in this study it is important to include exactly the attributes relevant in
that scenario. Therefore, many attributes that were included in hotel studies are of no
particular relevance. The attributes included with levels in parentheses were price
(50€, 70€, 90€), online evaluation (very good reviews, average reviews, bad reviews,
no reviews), location (in the center, close to the center, off center), pictures of rooms
(average pictures, very nice pictures, no pictures), pictures of the hotel (average
pictures, very nice pictures, no pictures), star rating (three star, four star, five star),
and recommendation of a friend (a friend recommended it, a friend said it is average,
a friend discouraged, no opinion of a friend).
3 Methodology
3.1 Measurement and Study Design
Respondents were recruited by means of an online survey. The first webpage outlined
the aim of the study and invited persons interested to participate. First, respondents
were asked about their Internet usage for information search and online booking as
well as the category of hotel they usually book. Given these data respondents that
never used the Internet for information search or booking could be excluded from the
analyses. The filtering was important as the scenario the respondents were confronted
with included the booking of a four-star hotel in Barcelona. Only respondents that
normally stay in such a hotel and know about prices and services should be included
in the study. Then, an adaptive conjoint design with a total of 22 paired comparisons
followed. The software used is the adaptive conjoint tool version 6.0.6 from Sawtooth
Software Inc (Sawtooth, 2007). The conjoint procedures represent an indirect method
of measuring the tourists’ perceived utility of a combination of hotel attributes. They
do not force the respondents into an artificial task of evaluating hotel attributes
separately and sequentially. Instead, one lets them assess complete hotel descriptions
in a way they encounter in a real-world choice situation.
The conjoint design involved the seven attributes mentioned and a total of 23 levels
which were incorporated in 22 paired comparisons. The instrument used for
measuring preferential differences between pairs was a 9-point scale with anchors
“strongly favor left” to “strongly favor right” and neutral in the middle.
3.2 Conjoint and Partitioning Analysis
Sawthoth’s adaptive conjoint analyzer (ACA) extracts the part-worth utilities for the
attribute levels. However, it does not yield importance values of the attributes. These
importances are considered to be a ‘natural’ basis for building market segments
defined in terms of the combination of weights they attach to the attributes of a choice
alternative. So one has to reconstruct them. They are easily derived from the part-
worth utilities uijk of level j of attribute i for respondent k. The importance of a hotel
attribute i for respondent k is given by
ji
kinkikinki
kinkikinki
ik
ii
ii
uuuu
uuuua
,
11
11
)],...,min(),...,[max(
100*)],...,min(),...,max([
where ni denotes the number of levels of attribute i. The computation assures that the
importances of the hotel attributes are defined on a percentage scale and therefore
fully comparable over respondents.
The ACA employs an up-to-date modeling and powerful parameter estimation
method viz. a Hierarchical Bayes (HB) model with Markov-Chain-Monte-Carlo
(MCMC) estimation. As demonstrated by Tüchler, Frühwirth-Schnatter & Otter
(2004) the HB-MCMC methodology assists in effectively detecting respondent
heterogeneity in a principled manner. By delivering individual sets of attribute
importance values for each individual respondent the HB-MCMC model captures
continuous heterogeneity. (A latent class model would be an example of how to
control discrete heterogeneity within a sample of respondents.) The ACA assumes
normally distributed part-worth utilities, but, owing to its underlying Bayesian
estimation of individual parameter values, offers the analyst an ideal opportunity for
determining importance-based customer segments.
Fig. 1. (W)SSI for 2-10 classes1
Vector quantization as outlined in Mazanec & Strasser (2000) was employed for
partitioning the attribute importance data. Two decisions had to be made during the
partitioning analysis: (i) the dot product of the (normalized) vectors of hotel attribute
importances serves as similarity measure which is particularly robust against outliers.2
(ii) the number of classes (‘prototypes’ in neural networks parlance; ‘segments’ or
‘types’ in psychometrics and marketing jargon) was determined by means of the
Weighted Simple Structure Index introduced in Mazanec (2001) and evaluated in
Dimitriadou, Dolnicar & Weingessel (2002). The WSSI assists in identifying
interesting partitions i.e. solutions with strongly contrasting variable patterns.
According to Figure 1 either 5 or 9 classes appear to be promising. As two classes
dropped in size to about 2% in the 9-classes solutions the 5-classes result was chosen.
1 The SSI reflects the ease of interpretability owing to the amount of contrast among the
classes’ mean weight vectors. Interpreting the partitioning solution is the easier the more hotel
attributes exhibit marked differences between the customer segments. Managerial usability of
the results, however, will be limited once the size of the classes becomes too small. Therefore,
the weighted SSI penalizes contrasting values achieved for small segments. 2 This means that the importance vectors are rescaled to unit length to become points lying on
the surface of a 7-dimensional hypersphere. The similarity for a pair of hotel attribute
importances then corresponds to the directional cosine between the two sets of importance
values.
Finally, an attempt of profiling the importance segments in terms of external (passive)
variables was undertaken.
4 Results
4.1 Sample Composition
A total of 346 fully completed questionnaires were included in the analyses. Gender is
distributed nearly evenly with 52% male and 48% female respondents. 15% of the
interviewees graduated from University or College. Regarding the usage of the
Internet as an information source for hotels 86% indicate to consult it always or often.
Only the remaining 9% use this information source sometimes, 5% seldom. 43%
always/often use it for booking, a third sometimes, 20% seldom and 8% never. 77%
use online reviews of former guests in their information search which emphasizes
including online reviews on a booking website. 65% indicate that they posted a hotel
review themselves. Comparing this figure with the percentage for booking online it
appears that after booking online many customers are invited to evaluate the hotel
after their stay, a feature actually implemented by most booking platforms.
4.2 The Optimal Online Booking Scenario
Consider the aggregate results first. The average importance values in the entire
sample range between 8% and 21%. The most important attribute of a hotel
description is ‘recommendation by a friend’ followed by ‘hotel review’. This is in line
with previous research claiming that the hotel image with word-of-mouth as one of its
propagators is most influential when it comes to hotel choice (Dubé & Renaghan
2000a). Pictures of the room, the only visible cue allowing some sort of prior
inspection of the travel product before actually going there is the next important
attribute. The room seems to be more important than the hotel as such. This is
followed by price, location, and, finally, star category. Interestingly, star category,
supposedly an attribute of highest relevance (Callan, 1998) turns out to be negligible
in an online booking scenario. At least, this finding holds in the simultaneous
presence of reviews and/or recommendation by friends.
Compiling the maximum part-worth-utilities the ideal online offer should include
recommendations of a friend (generating an average part-worth of 2.29), favorable
reviews (2.02), pleasant pictures of a hotel room (1.34), a price of €50 (1.33), a
central location (.67), and a four-star rating (.28).
Hotel Attributes Average Importances Std. Deviation
Price 13.69 5.30
Hotel Review 19.41 4.46
Location 11.13 4.93
Pictures of room 13.86 3.42
Pictures of hotel 12.53 3.37
Star category 08.39 3.80
Recommendation of a friend 20.95 4.28
Table 1. Average importances
Now take account of the tourists’ heterogeneity in terms of the importance of hotel
attributes for online booking. The partitioning results recommend distinguishing five
different types of consumer groups as depicted in Figure 2. Groups one to five
comprise 18%, 26%, 17%, 23%, and 16% of the sample. The numeric labels in the
bar charts of Figure 2 read as follows: 1=price, 2=online review, 3=location,
4=pictures of rooms, 5=pictures of the hotel, 6=star category, 7=recommendation of a
friend.
Fig. 2. Hotel attribute importances in per cent by tourist segment
Type 1, Price indifferent. This segment shows lowest importance regarding the price
of the hotel. The visual cues shown on the booking platforms such as pictures,
evaluations and finally the star rating are of more importance. The recommendation of
a friend is also relevant for this group. Type 2, Recommendation seeker. Type 2
exhibits the highest importance levels for online reviews and the recommendation of a
friend. Pictures are somewhat important while location and star rating appear to be
nearly irrelevant. Type 3, Star rating indifferent traveler. For these online bookers the
star rating is not an important feature regarding the choice of a hotel. All other
attributes show average levels of importance. Type 4, Friends trustee. In this group
the recommendation of a friend is most important, followed by online review, price
and location. Unimportant are the star rating and the pictures presented online. Type
5, Price and recommendation. For the fifth type price and recommendations, both
from friends and online, are of highest importance. The other attributes are similar to
type two with low importance of location and star rating and medium importance of
the pictures shown online.
Membership in the importance-based customer segments turns out to be independent
from the respondents’ gender, education, job position, and travel experience.
However, the familiarity with city trips is associated with the importance
classification. ANOVA results are significant (p<.05) and indicate that more
experienced travelers (Type 3) consider recommendations to be less important than
inexperienced travelers (Types 2 and 4). Furthermore, there is a significant (p<.05)
difference between the customer segments regarding the usage of online reviews.
Types 2 and 4 use reviews as an information source before booking significantly more
frequently. This conclusively corresponds with the high importance of
recommendations for these customer groups. As to the amount of trust in the online
reviewers Type 4 and 5 significantly (p<.02) more often think that reviewers are
honest. There is also a difference regarding the perceived well-meaning and
benevolence of reviewers. The results show that in the view of Type 1 and 3 tourists
online reviewers appear to be less well-meaning (p<.03) and benevolent (p<.04) than
for members of Types 5, 4, and 2.
5 Conclusion and Future Research
With a share of 70% direct sales already account for the bulk of online sales
(Marcussen, 2007). Therefore, it is imperative that hotels optimally present
themselves and exactly provide the information required by their customers. A simple
measure such as including pictorial elements significantly raises the average part-
worth utility from -1.41 (no picture) to 1.34 (pleasant photo of a hotel room).
A change from a hotel description with no customer credentials to one with very good
reviews provokes an average increase in the part-worth utilities from -.32 to 2.02.
Within a particularly credential-sensitive segment such as Type 2 the increase is still
more drastic pushing the part-worth from -.16 to 2.50. A change in the star rating, by
comparison, only achieves an average increase from -.51 (three-star) to .28 (four-star)
with even smaller effects for the customer Types 3, 4, and 5.
Under an online setting the traditional star rating apparently does no longer fulfill its
purpose. When tourists are offered pictorial cues in conjunction with credentials
(online and personal) they tend to disregard the star ratings. For two customer types
(Type 2 and 4) representing approximately 50% of the market this effect is
particularly pronounced. It also holds for market segments with an extremely weak
(Type 1) or strong (Type 5) focus on price. Overall, an ideal online booking platform
is expected to justify their quality claims pictorially and by referring to previous
customers’ experience rather than by pointing to their star rating.
Note that the results were gained by analyzing respondents’ reported behavior. The
major limitations to overcome in future studies about featuring hotel attributes online
originate from this ‘stated preferences’ set-up. Observing the customers’ unbiased
reactions by unobtrusive measurement techniques may change the picture. Expressing
a decent amount of distrust vis-à-vis the star rating systems may be indicative of the
ideal self-image of a rational person. Also, the role of price is likely to become more
significant if the online booking exercise entails an actual purchase.
References
Bell, R.A. & Morey, R.C. (1996). Purchase Situation Modeling: The Case of Hotel Selection Criteria for Corporate Travel Departments. Journal of Travel Research (Summer): 57-63.
Callan, R. (1998). Attributional Analysis of Customers’ Hotel Selection Criteria by U.K. Grading Scheme Categories. Journal of Travel Research, 36 (Winter): 20-34.
Dimitriadou, E., Dolnicar, S. & Weingessel, A. (2002). An Examination of Indexes for Determining the Number of Clusters in Binary Data Sets. Psychometrika 67 (1): 137-160.
Dubé, L. & Renahan L.M. (2000a). Creating Visible Customer Value: How Customers View Best-Practice Champions. Cornell Hotel and Restaurant Administration Quarterly 41 (February): 62-72.
Dubé, L. & Renahan L.M. (2000b). Marketing Your Hotel to and Through Intermediaries. Cornell Hotel and Restaurant Administration Quarterly 41 (February): 73-83.
Essawy, M. (2006). Testing the Usability of Hotel Websites: The Springboard for Customer Relationship Building. Information Technology & Tourism 8: 47-70.
Gursoy, D. & McLeary, K.W. (2003). An Integrative Model of Tourists' Information Search Behavior. Annals of Tourism Research 31 (2): 353-373.
Hu, C. & Hiemstra, S. J. (1996). Hybrid Conjoint Analysis as a Research Technique to Measure Meeting Planners’ Preferences in Hotel Selection. Journal of Travel Research (Fall): 62-69.
Law, R. & Leung, R. (2000). A Study of Airline’s Online Reservation Services on the Internet. Journal of Travel Research 39 (2): 202-211.
Law, R. & Leung, K. (2002). Online Airfare Reservation Services: A Study of Asian-Based and North American-Based Travel Web Sites. Information Technology & Tourism 5: 25-33.
Law, R. & Chan, S. (2004). Internet and Tourism – Part XIV: hotels.com. Journal of Travel and Tourism Marketing 17 (4): 79-81.
Lewis, R. (1985). Predicting Hotel Choice: The Factors Underlying Perception. Cornell Hotel and Restaurant Administration Quarterly 26 (February): 82-96.
Lewis, R. (1984). Isolating Differences in Hotel Attributes. Cornell Hotel and Restaurant Administration Quarterly 25 (November): 64-77.
Lieux, E. M., Weaver, P.A. & McCleary, K. (1994). Lodging Preferences of the Senior Tourism Market. Annals of Tourism Research 21 (4): 712-728.
Klein, S., Köhne, F., Öörni, A. (2004). Barriers to Online Booking of Scheduled Airline Tickets. Journal of Travel & Tourism Marketing 17 (2/3): 27-39.
Marcussen, C. (2007). Trends in European Internet Distribution - of Travel and Tourism Services. http://www.crt.dk/uk/staff/chm/trends.htm
Mazanec, J. & Strasser H. (2000). A Nonparametric Approach to Perceptions-Based Market Segmentation: Foundations. Vienna-New York: Springer.
Mazanec, J. A. (2001). Neural Market Structure Analysis: Novel Topology-Sensitive Methodology. European Journal of Marketing 35 (7-8): 894-916.
McCleary, K., Weaver, P. & Hutchinson, J. (1993). Hotel Selection Factors as They Relate to Business Travel Situations. Journal of Travel Research 32 (2): 42-48.
McCleary, K., Weaver, P. & Lan, L. (1994). Gender-based Differences in Business Traveler’s Lodging Preferences. Cornell Hotel and Restaurant Administration Quarterly 35 (April): 51-58.
Mehta, S. & Vera, A. (1990). Segmentation in Singapore. Cornell Hotel and Restaurant Administration Quarterly 31 (May): 80-87.
Richard, M.D. & Sundaram, D.S. (1994). A Model of Lodging Repeat Choice Intentions. Annals of Tourism Research 21 (4): 745-755.
Sawtooth, Inc. (2007). SSI Web v6 Software for Web Interviewing and Conjoint Analysis. Bryan Orme, Editor (Updated 23 March 2007) Sawtooth Software, Inc. Sequim, WA http://www.sawtoothsoftware.com
Tüchler, R., Frühwirth-Schnatter, S. & Otter, T. (2004). Bayesian Analysis of the Heterogeneity Model. Journal of Business and Economic Statistics 22 (1): 2-15.
Wind, J., Green P., Shifflet, D. & Scarborough M. (1989). Courtyard by Marriott: Designing a Hotel Facility with Consumer-Based Marketing Models. INTERFACES 19 (1): 25-47.
Acknowledgements
Information Diffusion across Interactive Online Media (www.idiom.at) is funded by
the Austrian Ministry of Transport, Innovation & Technology and the Austrian
Research Promotion Agency within the strategic objective FIT-IT (www.fit-it.at).