Upload
vanxuyen
View
216
Download
2
Embed Size (px)
Citation preview
Eindhoven, December 2012
BSc Industrial Engineering and Management Science – 2010
Student identity number 0610320
in partial fulfilment of the requirements for the degree of
Master of Science
in Innovation Management
Supervisors Dr J.P.M. Wouters, TU/e, Innovation Technology Entrepreneurship and Marketing (Department of
Industrial Engineering and Innovation Sciences, School of Industrial Engineering)
Dr M.C. Willemsen, TU/e, Human Technology Interaction (Department of Industrial Engineering and
Innovation Sciences, School of Innovation Sciences)
J. Hes, BCom, Docdata Commerce, Marketing Department
Validating the e-Servicescape
An explanatory study towards web shop
conversion optimisation
by
Maarten van Haperen
TUE. School of Industrial Engineering
Series Master Theses Innovation Management
I
Executive Summary
Growth in e-commerce occurs at a tremendous rate. Over the past decade the continuing development
in electronic and mobile capabilities has led to increasingly more opportunities for e-commerce sales.
European e-commerce retail sales in 2011 were at 96 billion Euros, of which 9 billion in the
Netherlands (Thuiswinkel.org, 2012), and are expected to grow by 12.2% per year, topping 172
billion Euros in 2016 (Gill, 2012). Within the operation of web shops, maximizing turnover can be
seen as a key characteristic of success. As such increased research efforts have focussed on factors
determining the buying behaviour of online customers. Trust has been identified as the main driver of
purchase intentions online (Harris & Goode, 2010), which is in turn influenced by several factors.
These factors are summarized under the term e-servicescape: “the online environment factors that
exist during service delivery” (Harris & Goode, 2010).
What made the implementation and design of these factors challenging, is the notion that web shop
visitors are not homogeneous. Not only do different visitors have different intentions for visiting
based on utilitarian versus hedonic purposes, visitors are also heterogeneous in the sense that they are
in one of several stages part of the decision making process underlying online consumer purchase
behaviour (Butler & Peppard, 1998; Teltzrow & Berendt, 2003; Van der Heijden, Verhagen, &
Creemers, 2003). As such, different pages and sections of a web shop are oriented towards supporting
one or multiple of these stages and goals. A major deficit of the e-servicescape model as presented by
Harris and Goode (2010) is not taking into account these different phases. Furthermore, this
theoretical view opposed to the practical nature of web shop operation in which specific pages are
optimized in order to maximize revenue and purchases. The main goal of web shop operators is thus
to optimise cart value and conversion, defined as “the amount of purchasing visitors as opposed to the
total amount of visitors having expressed an interest towards a product” (Teltzrow & Berendt, 2003).
In order to further validate the e-servicescape model and in order to include the different stages of the
consumer decision making process, the following research question was developed:
Which e-servicescape factors and design rules can be used during different
stages of the consumer decision making process to optimise web shop conversion?
Based on a literature review encompassing 87 quality academic sources, 37 design rules were
extracted. These design rules were part of the e-servicescape factor “aesthetic appeal”, “layout and
functionality” or “financial security” and subsequently coupled to one or more of the consumer
decision making stages part of this research: search for information, evaluation of alternatives and
choice / purchase. Next, 13 semi-structured interviews were held with employees at a web shop
operator focussed on web shops of large retailers and brands in order to validate the model. This
resulted in the updating of 7 design rules and the adding of 10 additional design rules. Given that the
interviews were held at a single company, generalizability of the model is somewhat limited and
requires future research. The final model is depicted at the end of the main text of this thesis.
Next to the validated e-servicescape model, additional knowledge on two design rules part of the
model was created by implementing these rules at a lingerie retailer’s web shop using two field
experiments. The first experiment focussed on the inclusion of cross-selling functionality on the cart
page and the use of unique selling points (USPs) on the cart page. The results showed a negative
II
impact of cross-selling functionality on cart-to-purchase conversion rates, but not on profitability.
Furthermore the USPs did not prove to have an effect on cart-to-purchase conversion, but this might
have been a result of the sale that was active during the experiment. The second experiment focussed
on the effects of using a single-page versus a multi-page checkout on checkout-conversion. The
dataset was however too limited to draw scientific conclusion.
The limited amount of data available was not only a limitation for the checkout experiment, but also
for the cart page experiment. In made it difficult in both cases to correctly identify the effect of time
based covariates. Additionally the research setup and field nature of the experiments did not allow for
the measuring of concepts such as trust and purchase intentions, as were some technical restrictions
present.
The main scientific contributions of the research are threefold. First of all the e-servicescape model is
expanded to include consumer decision making process stages and specific and detailed design rules.
Second the field validation of the model by means of interviews allowed for the inclusion of field
knowledge, best practices and sentiment. Finally future research opportunities regarding specific
design rules and topics were identified.
The main managerial implications of the research are also threefold: the validated e-servicescape
model allows for the day-to-day use of a practical model with design rules in order to influence and
optimise visitors purchase intentions. Furthermore cross-selling was identified as having a negative
effect on cart-to-purchase conversion but no impact on revenues in the specific case of the lingerie
retailer under discussion. It implies the careful consideration of where and when to use cross-selling
to increase cart values due to the potential negative effects on conversion. The final implication is
formed by the experiment execution: it is important to carefully consider experiment setups, data
collection methods and data analysis methods as well as the way specific implications are established.
A solid academic approach and well thought-out plan are needed to derive significant and meaningful
results.
III
Preface
After six years of education, I can finally say that I do not for a second have doubts about my choice
for the education field of Industrial Engineering and specifically Innovation Management at
Eindhoven University of Technology. Balancing ‘hard’ engineering skills and knowledge and more
‘soft’ marketing and psychological skills encompassed what I was looking for and have found. The
combination of different fields of interests has both provided me with challenges and opportunities;
challenges in prioritising and in time and stress management on the one hand and opportunities for
expanding knowledge and both personal and professional enrichment on the other hand.
It feels both strange and relieving that this report marks the end of my six year education at
Eindhoven University of Technology. However it is with confidence that I now enter the career
market being the next step lying in front of me. I would like to express my gratitude to those that have
helped me in the realization of this thesis report as the ending of my Master’s Education. First of all to
my first TU/e supervisor Joost Wouters, for kindling my enthusiasm for marketing during his courses
and for his useful advice and aid in providing structure and simplicity where there first was
complexity. I also would like to express my gratitude to my second supervisor Martijn Willemsen for
supporting me and keeping me on the correct path of providing academic value in this thesis.
My thesis research would not have been possible without the support of and opportunities made
available by Docdata Commerce. I owe gratitude to my company supervisor Jurriaan Hes for
providing me with insights and his support during the realisation of my experiments and to all
Docdata Commerce employees for their support and willingness to cooperate with both my thesis
project in general and specifically the interviews I held with them.
The past six years of education and the completion of my thesis would not have been as successful or
at least triple as difficult without the continuous support of my parents Rob and Ineke, not-so-little
brother Wouter and my loving girlfriend Meike. I thank them for their criticism, support and faith.
To all those that I have not seen the chance of thanking here, please know that I am grateful for your
support, enthusiasm and spare time company. Were it during my study projects, my part time jobs, my
countless hours of practicing Irish dancing or during any other wonderful activity, I have learned that
practice makes perfect as longs as you keep an open, creative mind.
Maarten van Haperen
December 13, 2012
Scribendi recte sapere est principium et fons. - Horatius, ‘De Arte Poëtica’ 309
IV
V
Table of contents
Executive Summary ............................................................................................................................ I
Preface ................................................................................................................................................ III
1. Introduction
Research question and contribution ...................................................................................... 1 1.1.
Research approach ................................................................................................................ 2 1.2.
Research method ................................................................................................................... 2 1.3.
Thesis outline ........................................................................................................................ 2 1.4.
2. Validating and extending the e-servicescape model
Online trust............................................................................................................................ 5 2.1.
E-servicescape....................................................................................................................... 7 2.2.
Consumer decision making process ...................................................................................... 9 2.3.
Integrating the e-servicescape and the consumer decision making process ........................ 11 2.4.
Validating and extending the e-servicescape model ........................................................... 11 2.5.
3. Experiment design based on the e-servicescape model
Improving cart page conversion .......................................................................................... 17 3.1.
Improving checkout conversion .......................................................................................... 18 3.2.
4. Experiment method
Data collection .................................................................................................................... 19 4.1.
Data analysis ....................................................................................................................... 21 4.2.
Research quality .................................................................................................................. 23 4.3.
5. Experiment results and discussion
Experiment ‘Cart page’ ....................................................................................................... 25 5.1.
Experiment ‘Checkout’ ....................................................................................................... 30 5.2.
6. Conclusion ................................................................................................................................. 33
7. Reflection
Limitations .......................................................................................................................... 35 7.1.
Theoretical contributions and future research opportunities ............................................... 37 7.2.
Managerial implications ...................................................................................................... 39 7.3.
8. References ................................................................................................................................. 43
VI
Appendix A. Literature review results .................................................................................... 47
Appendix B. Company structure .............................................................................................. 49
Appendix C. Case study interview questions ....................................................................... 51
Appendix D. Case study results ............................................................................................... 57
Appendix E. Clickstream variables recorded ....................................................................... 65
Appendix F. Experiment design ............................................................................................... 69
Appendix G. Experiment results ............................................................................................... 77
1
1. Introduction
Growth in e-commerce occurs at a tremendous rate. Over the past decade the continuing development
in electronic and mobile capabilities has led to increasingly more opportunities for e-commerce sales.
European e-commerce retail sales in 2011 were at 96 billion Euros, of which 9 billion in the
Netherlands (Thuiswinkel.org, 2012), and are expected to grow by 12.2% per year, topping 172
billion Euros in 2016 (Gill, 2012). Not only the rise of fully internet-based companies is a major
development, e-commerce has also become increasingly important for retailers that operate brick-and-
mortar shops. This is due to the fact that the growth in e-commerce sales also implies the diminishing
of offline spending. A web shop is hence not merely an opportunity but has become a necessity.
Within the operation of web shops, maximizing turnover can be seen as a key characteristic of
success. As such increased research efforts have focussed on factors determining the buying
behaviour of online customers. Trust has been identified as the main driver of purchase intentions
online (Harris & Goode, 2010), which is in turn influenced by several factors. These factors are
summarized under the term e-servicescape: “the online environment factors that exist during service
delivery” (Harris & Goode, 2010). This term stemmed from research on designing the servicescape in
physical environments, with a focus on all factors during a service delivery in for example a shop or
hospital (Bitner, 1992).
What made the implementation and design of these factors challenging, is the notion that web shop
visitors are not homogeneous. Not only do different visitors have different intentions for visiting
based on utilitarian versus hedonic purposes, visitors are also heterogeneous in the sense that they are
in one of several stages part of the decision making process underlying online consumer purchase
behaviour (Butler & Peppard, 1998; Teltzrow & Berendt, 2003; Van der Heijden, Verhagen, &
Creemers, 2003). As such, different pages and sections of a web shop are oriented towards supporting
one or multiple of these stages and goals. A major deficit of the e-servicescape model as presented by
Harris and Goode (2010) is not taking into account these different phases. This opposes the practical
nature of web shop operation in which specific pages and sections are optimized in order to maximize
revenue and purchases. The main goal of web shop operators is thus to optimise cart value and
conversion, defined as “the amount of purchasing visitors as opposed to the total amount of visitors
having expressed an interest towards a product” (Teltzrow & Berendt, 2003).
Research question and contribution 1.1.
In order to further validate the e-servicescape model and in order to include the different stages of the
consumer decision making process, the following research question was developed:
Which e-servicescape factors and design rules can be used during different
stages of the consumer decision making process to optimise web shop conversion?
By providing an answer to the research question, the contributions of this research are both theoretical
and practical. The theoretical contributions are twofold. First they consist of validating the e-
servicescape model as proposed by Harris and Goode (2010) by incorporating knowledge from
different academic publication regarding e-servicescape factors and knowledge available in the field.
Second the e-servicescape model is extended to include the different stages of the consumer decision
making process and different purposes of web shop pages and sections, a criticism on the original
model that only looked at a web shop in general.
2
Research approach 1.2.
In order to answer the research question a multi-step approach was considered, as depicted in Figure
1.1. The first step in the research approach consisted of validating and detailing e-servicescape design
rules on the basis of academic publications and where applicable assigning them to specific stages in
the consumer decision making process. This resulted in a theoretical e-servicescape design rule model
adapted towards the consumer decision making process.
The second step in the research consisted of validating the theoretical framework with knowledge
available in the field of web shop operations. This resulted in a validated and extended e-servicescape
design rule model adapted towards the consumer decision making process. Two of the design rules
were investigated further in order to generate more knowledge on their implementation.
Research method 1.3.
The research approach outlined was established on the basis of three research methodologies: a
systematic literature review, a single embedded case study and a field experiment. The systematic
literature review was based on 35 quality academic sources. It established the e-servicescape model,
the consumer decision making process and a theoretical framework incorporating the design rules
based on the combination of the two. Next a series of interviews was conducted on the basis of a
single embedded case study at a web shop operator, in order to validate and extend the framework.
Two design rules were selected from the framework for further research, which was done using an
explanatory field experiment at the web shop of a large lingerie retailer.
Thesis outline 1.4.
Based on the research question and corresponding research approach, the remainder of this thesis is
outlined as depicted in Table 1.1: chapter two depicts the literature review and single embedded case
study on which the e-servicescape design rule model adapted towards the consumer decision making
process was based and validated. Chapter three focuses on establishing a set of hypotheses based on
two design rules selected for further investigation using field experiments. The experiment design and
method are described in chapter four after which the results from the experiment are presented and
discussed in chapter five. Chapter six subsequently focusses on providing conclusions drawn on the
basis of the entire research, before discussing its limitations and identifying opportunities for future
research as well as theoretical and managerial implications.
Figure 1.1 Research approach
Academic publications
Literature review
Design rule investigation
Field experiment
E-servicescape design
rule model
Knowledge from field
Case study
validate
3
Table 1.1 Thesis outline per chapter
Chapter Topic Research method
1 Introduction
2 Establishing, validating and extending the design rule model. Literature review, case study
3 Establishing hypotheses based on contradicting design rules. Experiment
4 Experiment design and method Experiment
5 Experiment results and discussion. Experiment
6 Research conclusions
7 Limitations, opportunities and implications.
4
5
2. Validating and extending the e-servicescape model
Online trust 2.1.
Due to the low search costs and effort involved in online shopping, visits without purchase intentions
will be more common online as opposed to offline in the future (Moe & Fader, 2004). Over the past
decade increasing research was performed with regards to stimulating the intention to purchase by
means of different drivers. One of the main drivers for purchase intentions as posed by several
researchers in the early 2000’s is trust (Van der Heijden et al., 2003). Other studies have shown this to
be true: “trusting beliefs regarding the web site had a significant positive effect on intention to buy
from it” (Stewart, 2003: 5) and “In particular, the emotion of trustworthiness has been emphasized as
one of the most important factors for the successful completion of commercial transactions” (Jinwoo
Kim & Moon, 1998: 2). Additionally, research has shown that poorly functional designed web shops
result in large amount of lost customers (Silverman, Bachann, & Al-Akharas, 2001; J. Song, Jones, &
Gudigantala, 2007).
2.1.1. Defining trust
In order to be able to identify the roles of trust on purchase intentions, a definition was needed:
“Online trust can be defined as an Internet user’s psychological state of risk acceptance, based upon
the positive expectations of the intentions or behaviours of an online merchant” (Y. D. Wang &
Emurian, 2005: 42). This view was supported by Lee & Turban, (2001: 79): “the willingness of a
consumer to be vulnerable to the actions of an internet merchant in an Internet shopping transaction,
based on the expectation that the Internet merchant will behave in certain agreeable ways, irrespective
of the ability of the consumer to monitor or control the Internet merchant.” The definitions show three
aspects part of the trust definition: it involves two parties (trustor and trustee) that need each other for
mutual benefits, it involves risk and it involves the trustor believing that the trustee will behave
according to the risk involving behaviour (Siau & Shen, 2003).
Shopping online brings risks that do not exist in traditional shopping, such as the absence of a
physical quality check before purchase and difficulties in safeguarding financial and privacy
information once handed over to the merchant (Lee & Turban, 2001). Both M. Lee and Turban (2001)
and Siau and Shen (2003) constituted three factors as main elements of online trustworthiness that
build on the previous definitions: ability, benevolence and integrity. Ability deals with the merchant’s
skills and competencies in performing online. Benevolence focuses on whether the consumer is
convinced that the merchant wants to do things right over merely maximizing profit. The final factor
integrity relates to the perception of the consumer that the trusted party is honest and acts in
correspondence with acceptable principles. Online trust is a combination of these three factors and is
furthermore context and situational specific.
In case of e-commerce a challenge lay in the fact that trust issues not solely involve the consumer and
the web shop merchant. There are also trust issues between the consumer and the computer system
through which that consumer makes transactions (Lee & Turban, 2001). In this research focus was put
on the type of trust mentioned first, as it was expected that interaction between consumers and
computer systems became more mainstream. This makes the web shop merchant trust issues more
relevant. Important to consider in this perspective is also the propensity to trust of individual
consumers; the way cues related to trustworthiness are magnified or reduced based on the personality
of the consumer (Lee & Turban, 2001).
6
2.1.2. Influencing trust
There are several aspects that influence a consumer’s trust in a web shop. It has to be taken into
account that trust is not merely the outcome of technical requirements but involves communication as
well (Chadwick, 2001). Hence trust can be influenced both implicit and explicit on both emotional
and cognitive levels. It can take place before visiting a web shop, during the visit and after an online
transaction (Egger, 2001). Three important aspects for influencing trust are past experience, external
factors and the e-servicescape.
The first aspect involves past experiences, influencing a consumers attitude towards and trust in web
shops in the future (Salam, Iyer, Palvia, & Singh, 2005). Bad experiences may negatively affect trust,
whilst good experiences may make it easier for a consumer to trust. The experiences do not only
constitute experiences related to the same web shop in the past, but also include experiences at other
web shops. A study by Wang and Emurian (2005) also showed that respondents with bad previous
experiences with web shops, such as being cheated on, gave comparatively lower ratings when asked
to value overall trustworthiness of a new web shop.
The second aspect involves external factors. Trust can be influenced by factors inside the scope of
control of the web shop that moderate the impact of both good and bad previous experiences as
described earlier. Such factors include a well-established offline reputation, the consumer’s perceived
size of the company, where bigger is better, and communications such as press releases,
advertisements and promotions, (L. H. Kim, Qu, & Kim, 2009; Lee & Turban, 2001; Salam et al.,
2005; Van der Heijden et al., 2003). The same goes for external factors outside the direct scope of
control of the web shop, such as news reports, evaluations and product recalls, but also guarantees,
legal rules and procedures. The latter factors influence what is called institution based trust, which in
turn will also affect purchase intentions of consumers (Salam et al., 2005). It can even lead to a
situation where a consumer does not trust a web shop, but makes a purchase nonetheless as there is
trust in the control systems and (legal) procedures in place (Van der Heijden et al., 2003).
The third and final important aspect is based on the e-servicescape model, see Figure 2.1 (Harris &
Goode, 2010). The e-servicescape was defined as “the online environment factors that exist during
service delivery” (Harris & Goode, 2010, p. 231). Hence by identifying and changing trust-
influencing factors on a web shop, both trust and correspondingly purchase intentions can be
Figure 2.1 E-Servicescape, factors and sub-factors (Harris & Goode, 2010)
7
influenced. The importance of web shop design was also stated by Wang and Emurian (2005, p. 43):
“In other words, one key consideration in fostering online trust is to build a trust-inducing e-
commerce interface.” Examining the web shop by looking whether it is from a well-known brand,
reading product information and looking for symbols of security approval are important risk-
reduction strategies by consumers (L. H. Kim et al., 2009) and are an integral part of the web shop
design. By designing the e-servicescape of a web shop in such a manner that trust and subsequently
purchase intentions are maximized, will have a direct impact on conversion. If users have higher
purchase intentions than can be converted into purchases, a web shop owner can increase revenues on
the basis of its current visitor set, without the direct necessity to attract additional visitors to gain more
visitors with a high enough purchase intention.
E-servicescape 2.2.
Once visiting a web shop consumers immediately form quick impression of the web shop, which they
seek to confirm with additional information and impressions they receive from the web shop. This
goes as far as interpreting information to suit the initial impression: on the same web shop, consumers
with a negative first impression interpret information negatively, whilst consumers with a positive
initial impression interpret the same information in a positive way (Stewart, 2003). The initial
impression can be influenced by the design of the e-servicescape. Before going into detail on the e-
servicescape, the origin and background of the concept are reviewed.
2.2.1. Concept
Bitner (1992, p. 58) has conceptualised the term servicescape as being “the built environment (i.e., the
manmade, physical surroundings as opposed to the natural or social environment)” and has posed that
it has a major impact on both consumers and employees in service organisations. Important aspects
involve ambient conditions, space and function, and signs and symbols. Although factors for brick-
and-mortar stores and web shops converge, there are some major differences as the online aspect
creates unique challenges at different transaction steps. The e-servicescape (also termed cyberscape or
virtual servicescape) was thus described as “the online environment factors that exist during service
delivery” (Harris & Goode, 2010: p. 231). This definition was widely supported by other researchers
(Jeon & Jeong, 2009; Vilnai-yavetz & Rafaeli, 2006; Williams & Dargel, 2004).
Several main factors form the e-servicescape of web shops. Aspects such as brand name cannot be
directly influenced and are therefore not part of the e-servicescape, although indicated by Fang &
Salvendy (2003) as a major factor impacting a shopper’s trust in web shop. The influential aspect that
is part of the e-servicescape is the placement and the number of displays of the brand name. The same
goes for the mention that people prefer to shop online over shopping at brick-and-mortar stores
because it is more convenient. Although this is most likely the case, the aspect by itself does not
improve purchase intentions. However, the way it is made clear in the e-servicescape to consumers
why online shopping should be preferred and is more convenient is, as well as providing a picture of a
physical building to induce trust in the online shop (Stewart, 2003).
As a starting ground for an e-servicescape literature review, the model by Harris & Goode (2010)
was used. According to their model, the e-servicescape consists of three factors as depicted in Figure
2.1. The first factor is aesthetic appeal, the second factor is layout and functionality and the third
factor is financial security. With regards to aesthetic appeal the focus lies on online ambient
conditions that create an attractive and alluring servicescape from the customer perspective. Layout
and functionality focuses on a contrasting aspect. Layout is focussed on the arrangement,
8
organisation, structure and adaptability of web shops. Functionality focuses on the extent to which
these aspects facilitate the conversion and service goals. The final factor, financial security deals with
the way in which consumers executing their purchase intention perceive the payment process as
secure and feel safe to complete it.
2.2.2. E-servicescape factor ‘Aesthetic appeal’
The first of the three factors part of the e-servicescape is aesthetic appeal. Y. D. Wang and Emurian
(2005: 51) stated that “design is more than an artistic interface.” This was supported by other
research, stating that aesthetics “deals with the sensory experience elicited by an artefact, and the
extent to which this experience tallies with individual goals and attitudes” (Vilnai-yavetz & Rafaeli,
2006: 248). The definition was supported by Cai and Xu (2011: 161) who stated that aesthetics is “a
holistic perception of design principles and individual objects (on a web shop), (…) closely connected
to attention and understanding, (…) that significantly affect human affect and emotion.” Good visual
design not only provides visual pleasure, but also comfortable reading and ease of use which (Y. D.
Wang & Emurian, 2005). Three sub-factors are part of aesthetic appeal, as seen in Figure 2.1:
originality of design, visual appeal and entertainment value (Harris & Goode, 2010).
Aesthetic appeal plays an important role in today’s consumers’ online consumption style, which
shifted from utilitarian purposes to a combination of utilitarian and hedonic purposes in which
recreation and entertainment have become more important aspects (Van der Heijden et al., 2003; J. Y.
Wang, Minor, & Wei, 2011). This shift and the different goals of users need to be taken into account
during the different purchase related stages in a web shop, moreover as the initial brief exposure of a
consumer to a web shop page immediately results in an aesthetic impression. This impression
correlates with the consumer attitude to that page and the entire web shop (Cai & Xu, 2011).
2.2.3. E-servicescape factor ‘Layout & functionality’
The second of the three factors part of the e-servicescape is layout and functionality. It encompasses
which design aspects are included on a web shop and the placement of these aspects. Four sub-factors
are part of layout and functionality, as seen in Figure 2.1: usability, relevance of information,
customisation and interactivity. The overall goal is to create a web shop with “easy-to-use navigation,
frequent updating, minimal download times, relevance to users and high quality content” (Palmer,
2002: 153).
The importance of layout and functionality was underlined further by Cai and Xu (2011: 162): “When
a web site is intuitively understandable in its design, it facilitates users’ interaction with the web site
and gives them a strong sense of control, knowledge of where to focus their attention and deep
cognitive enjoyment. As a result users may experience a state of flow whereby they have a distorted
sense of the passage of time and achieve an intrinsically enjoyable experience.”
2.2.4. E-servicescape factor ‘Financial security’
The third and final factor part of the e-servicescape is financial security. It encompasses security as
experienced while making (or planning to make) an electronic or internet payment. Financial security
is an important factor of the e-servicescape (Siau & Shen, 2003), moreover as in 2001 over 80% of
online shoppers abandoned their shopping carts before completing a transaction (Hausman & Siekpe,
2009). Although partly explainable by the fact that many web shop visitors use the shopping cart as a
wish list and comparison tool between web shops, reducing this percentage by increasing financial
security, results in a direct increase in revenue.
9
Although the penetration of internet payment is increasing, there will still be customers that have not
made (many) internet payments. Hence it is important to consider factors that determine the potential
adoption and usage of payment methods by customers. He and Mykytyn (2007) found that customer’s
willingness to adopt and use depends mostly on the overall design quality of the web shop, on
perceived risks and perceived benefits and on the payment features offered.
Liang and Lai (2002) state the importance of functional support using a good design to meet the
customer’s needs online. This is supported by Ranganathan & Ganapathy (2002) who state that the
overall quality of the web shop design is important for the performance of the web shop. Next to the
overall design quality, two factors part of the e-servicescape aspect financial security that can
influence customer’s willingness to adopt and use online payment methods: perceived security based
on perceived risk and ease of payment, as seen in Figure 2.1.
Consumer decision making process 2.3.
A major deficit of the e-servicescape by Harris and Goode (2010) is that the model focusses on web
shop at an abstract level. It does not take into consideration that consumers proceed through various
phases when making a purchase decision and that different pages and sections of the web shop
facilitate one or several of these phases.
Before focussing on the consumer decision making process it has to be noted that consumers can be
placed on a continuum of two extreme values based on behavioural characteristics (Carmel, Crawford,
& Chen, 1992; Tomes, 2000): goal-directed behaviour, focussed on making a purchase, and
experiential behaviour, focussed on browsing. The goal-directed and experiential behaviour are
respectively characterised by extrinsic versus intrinsic motivation, utilitarian benefits versus hedonic
benefits and directed versus non-directed search (Hong, Thong, & Tam, 2004).
The phases a consumers proceeds through when looking to buy a product which requires limited to
extensive problem solving behaviour, were depicted in Figure 2.2 (Butler & Peppard, 1998; Miles,
Howes, & Davies, 2000). Although the process has been set up as being linear, iterations and
feedback loops are very important as it is unlikely the consumer will follow a strictly linear approach
in his decision behaviour. The first phase constitutes a consumer realising or being attended to the fact
that a new product or service is required. Next follows the search for information in which different
alternatives are derived, followed by a phase in which these alternatives are evaluated and a phase in
which a choice on the product and purchase location are made. The final stage focuses on satisfaction
and loyalty behaviour resulting from a purchase, leading either to future purchases or disappointed
customers. As the thesis focused on conversion optimisation, focus was placed on the information
search phase, evaluation phase and purchase phase. Post-purchase behaviour was taken into thought
however, as converting existing customers to repeat buyers has proven to be over six times cheaper
than converting new customers (Silverman et al., 2001).
Figure 2.2 Online consumer purchase and decision behaviour (Butler & Peppard, 1998)
Information search
Evaluation of alternatives
Post-purchase behaviour
Choice /
purchase
Problem recognition
10
The approach towards viewing the consumer purchase and decision process as an iterative, but
generally linear process view, was coupled to the intention based view identified by Song and Zahedi
(2005), stating that a consumer visits a web shop with the intention to make a current purchase
decision, revisit the web shop in the future or to repurchase in the future. Consumers in the
information search phase will likely have the intention to revisit the web shop in the future, not
necessarily purchasing in their current visit, whilst consumers in the choice and purchase phase are
likely to have a direct purchase intention. The phase focussing on the evaluation of alternatives could
indicate both web shop visit intentions. It may be the case that customers are looking to immediately
buy after evaluating alternatives or that they may postpone the purchase action to a later visit,
indicating that there could both be a return visit or current purchase intention. The final intention by
Song and Zahedi (2005), the intention to repurchase, was coupled to the post-purchase behaviour
stage after the actual purchase. As customers evaluate their purchase on a product and on a web shop
level in this phase (Butler & Peppard, 1998), this is the moment that determines whether or not a
buyer is likely to visit the web shop again in the future with the intention to purchase again.
Focussing on the current purchase intention in the model by Song and Zahedi (2005), the consumer
buying and decision making process was linked to the business oriented customer life cycle
perspective as seen in Figure 2.4 (Teltzrow & Berendt, 2003). From a business point of view, a
company tries to get suspects (targeted customers) to visit the web shop, making them prospects.
Once on the web shop, the prospects are converted to customers, which will have different loyalty
behaviours based on purchase satisfaction. The suspects to prospects process can be coupled to the
problem recognition and information search phases by Butler and Peppard (1998), whilst the
prospects to customer process can be coupled to the evaluation and choice phase.
It was interesting to investigate the conversion process of consumers more closely, as these are
consumers who have already expressed an interest in buying and only need to complete their
purchase. The purchase process consists of four steps, as seen in Figure 2.3: seeing a product
impression, performing a product click-through, effecting a basket placement and making a product
purchase (Teltzrow & Berendt, 2003). A product impression is for example a product image on the
category overview page. Clicking on this product in order to get the product page is the following
step. If a consumer decides to buy, the product will be placed in the shopping basket. This stage poses
challenges as research by Close and Kukar-Kinney (2010) indicated that consumers also use a
shopping cart to compare products between web shops. The final stage is completing the purchase by
providing credentials and a shopping address and by paying for the product. At this time price
negotiation options such as vouchers or coupons, (multiple) shipping options and payment options
become important in the decision making by the consumer (Silverman et al., 2001).
A similar view was oriented on sequential Nominal User Tasks (NUTs). NUTs are in this case tasks a
customer has to perform in order to place an order on a web shop (Sismeiro & Bucklin, 2004). Three
tasks were identified. First a customer has to complete the product configuration, for example
selecting the product colour, desired size and quantity, and then place the product in the basket. Next
the purchase stage as identified by Teltzrow and Berendt (2003) is split in two tasks. The customer
first has to provide a complete set of personal information. This can be done by either providing all
details or by logging in if an account was made in the past. The final task is confirming the order by
providing payment data or making a payment.
11
Integrating the e-servicescape and the consumer decision making process 2.4.
Taking both the e-servicescape factors and sub-factors and the consumer decision making process, a
list of design rules to maximise purchase intentions was established on the basis of an academic
literature review including 35 quality academic sources, as depicted in the final model in Table 2.1 at
the end of this chapter and in the detailed model depicted in Appendix A. The design rules were
coupled to the consumer decision making process phases ‘information search’, ‘evaluation of
alternatives’ and ‘choice / purchase’ as these phases are most prominent on a web shop. The design
rules were selected to be applicable to one or multiple of these phases on the basis of either the article
they were extracted from (i.e. ‘provide a link back to shopping’ occurs in the cart and as such
automatically only in the ‘choice / purchase’ phase) or on the basis of common sense (i.e. ‘display
out-of-stock products and sizes’, which can only be applicable during the ‘information search’ phase).
Validating and extending the e-servicescape model 2.5.
A single embedded case study at a large web shop operator was used to validate and the model
established on the pervious pages. The company was founded in 2000, originally focussing on
providing system management hardware and services until 2005 an order fulfilment company took a
minority stake. In 2010 that company extended its share and gained sole proprietorship in the e-
commerce operations organisation that is currently employing 28 FTEs, as seen in Appendix B. The
organisation operates web shops for large brands and retailers and its portfolio includes a lingerie
retailer, a luxury leather products brand and a company selling printing supplies.
Figure 2.4 Online consumer purchase and decision behaviour (Butler & Peppard, 1998)
No sale
nC Not acquired
nP Not reached
nS
Choice /
purchase
Prospects Customers Suspects Repeat
customers
Rep.customers elsewhere
One-time customers
nM4
= nC
Figure 2.3 Online consumer purchase and decision behaviour (Butler & Peppard, 1998)
nM3 = nC
nM2
= nC
nM1
= nC
Prospects
M2 Clicked-through
M3
Placed in basket
M1 Saw impression
Customer made purchase
12
2.5.1. Case study method
The method by which the best practices were derived was a single embedded exploratory case study.
A single case study was used as only employees at the e-commerce fulfilment company were
interviewed to extract knowledge, as the case study was exploratory in nature as it was aimed at
identifying design rules and hypothesis regarding factors that influence conversion in web shops (Yin,
2009).
The unit of analyses was best practices concerning web shop conversion optimisation present within
the e-commerce company. The best practices were based on four specific cases of web shops that
made over 100,000 Euros revenue per year or processed over a million visitors per year, next to
comments that were made regarding other web shops. Using web shops with more traffic reduced the
risk of missing or overstating on optimisation techniques that are too small to measure and do not or
unintended got significant outcomes. In selecting cases a choice was made to only investigate projects
of web shops that have were recently launched on the basis of a well-known, strong and established
offline brand, or web shops that have existed for over a year, in order to be able to analyse the
servicescape factors in general. It was expected that there would be substantial moderating influences
for brands and web shops that had yet to establish themselves.
The case study data was acquired using the data collection principles of Yin (2009) in order to ensure
a basic level of reliability and validity. Reliability deals with the repeatability of a study where equal
results should follow in case of a study reproduction under the same conditions. Validity deals with
concepts, measurements and conclusions being well-founded and consists of construct validity,
whether a measurement tools measures what is intended for, internal validity, dealing with the
causality of findings and external validity, generalizability beyond the current study.
The first principle included maintaining a chain of evidence by being clear about subsequent steps in
the case study and having readers understand the structure of the research in order to make clear on
what grounds conclusions were based. The second principle was creating a case study database. It
ensured reliability by allowing other investigators to review the evidence used in the case study
report, being notes, documents, interviews and other materials, although access to this database was
limited and confidential. The third and final principle was using multiple data sources, also known as
triangulation, and ensured construct validity.
The data was collected by interviews that were held with staff that was directly involved with
operating aspects of web shops and could have impact or knowledge on how to design the e-
servicescape. The interviewees consisted of three customer support employees, four shop managers,
two marketers, the managing director, one intern and an interview with a representative of the
company that provided the payment interface, one of the largest payment service providers globally.
Two customer support employees were included to test the interview format and because they
received direct consumer feedback in their daily operations. Four shop managers were included due to
their involvement in web shop operations, with two shop managers focussing on shop management of
web shops for retailers and brands and two shop managers focussing on web shops that were fully
oriented on SAPOS, Sales At Point Of Service. Two marketing engineers were included as they
support the shop managers, the managing director was included because of his knowledge on the e-
commerce business environment and one academic intern was included due to his focus on predicting
online purchase intentions. The software engineers were not interviewed as they stated they did not
have any knowledge on web shop optimisation and were fully focussed on technology, being able to
build nearly anything required by other departments. The representative of the payment company was
included to shed light specifically on the payments section in the e-servicescape model.
13
The interviews were semi-structured, see 0, based on the e-servicescape (sub-)factors mentioned
earlier (aesthetic appeal, layout & functionality and financial security) in order to validate and confirm
knowledge from the literature based model, whilst at the same time leaving room to include other
factors not included in the model (Van Aken, Berends, & Van der Bij, 2007). As the interviews were
semi-structured based on open questions, there was ample possibilities to venture into specifics
regarding literature based and field based factors. This prevented over-focussing and ensured a clear,
broad view on the matters discussed. The results of the audio-taped interviews were reviewed by the
interviewees themselves to ensure construct validity (Yin, 2009).
2.5.2. Case study results
The results of the case study were analysed as follows: every design rule was declared either
supported or commented on per interviewee in case the interviewee had knowledge regarding the
design rule and it was discussed in the interview. After generalising these results over all
interviewees, a decision was made whether the design rule was supported or needed revision. As there
were no design rules that were contradicted by a large portion of the interviewees, no drop decision
were made. A support or comment decision was made on the basis of two characteristics: support
ratio and individual expertise.
The first characteristic is the ratio of the amount of individual supports compared to the total amount
of supports and comments. The general rule was established that at design rule should have supports
or comments in least six of the eleven interviews and that there should be a 67% majority of support
statements to come to a support decision. In case of a lower ratio, the design rule was revised to
include the comments that arose during the interviews.
The second decision characteristic is the expected knowledge of the individual interviewees,
compensating that a designer is expected to be better informed about design rules regarding aesthetics
than for example the representative of the payment service provider. This made it possible to support
or revise a design rule going against the outcome of the first decision characteristic.
Next to supporting or revising design rules, it was also possible to add new design rules. Again the
decision characteristics above were taken into account, considering that at least three interviewees
should independently mention a new decision rule in order for it to be included. In case of adding
design rules, the corresponding decision making process stages are based on common logic. The
overall model with revised and added was discussed with the marketing department and found
appropriate.
The anonymised, individual and aggregate level interview results have been included in 0.
Consecutively the e-servicescape model was updated and validated, as can be seen in 0, resulting in
the overall model as depicted on the following pages in Table 2.1.
14
Table 2.1 Validated e-servicescape model
D
ec
isio
n m
akin
g
pro
ce
ss
sta
ge
Choice / purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
Inclu
de o
rig
inal desig
ns a
nd s
igns s
uch a
s lo
gos.
Anim
ate
lo
gos f
or
incre
ased e
ffectiveness a
nd im
pact, b
ut sparin
gly
to a
void
dis
tractio
n.
Whils
t adherin
g t
o s
tandard
and c
om
mon d
esig
n r
ule
s, m
ake s
ure
the d
esig
n fits t
he w
eb s
hop a
nd b
rand p
ropositio
n.
If p
ossib
le, im
ple
me
nt re
fere
nces to a
n o
fflin
e b
rand a
nd r
eta
iler.
Desig
n s
hould
be c
olo
urf
ul by a
good s
ele
ctio
n,
pla
cem
ent
and c
om
bin
atio
n o
f colo
urs
.
Desig
n s
hould
be d
ivers
e,
by v
isual richness,
dynam
ics, novelty a
nd c
reativity.
Desig
n s
hould
be s
imp
le b
y s
how
ing u
nity,
hom
ogeneity, cla
rity
, ord
erlin
ess a
nd b
ala
nce.
Desig
n s
hould
show
cra
ftsm
anship
by m
odern
ity a
nd in
tegra
tin
g s
implic
ity,
div
ers
ity a
nd c
olo
urf
uln
ess.
Pro
vid
e la
rge s
ize h
igh q
ualit
y p
roduct im
ages s
upport
ed b
y s
chem
atic p
roduct chara
cte
ristics.
Pro
vid
e lo
gos, cert
ific
ate
s a
nd o
ther
vis
ual cues e
arly o
n to e
nhance f
eelin
gs o
f tr
ust.
Do n
ot
dis
tract users
with a
esth
etic d
esig
ns d
urin
g c
heckout.
Th
e c
are
ful use o
f people
on p
ictu
res c
an p
rovid
e c
onte
xt
and tra
nsfe
r em
otio
n a
nd f
eelin
g.
Cre
ate
a d
esig
n t
hat flo
ws f
luently fro
m h
om
e p
age t
o c
heckout
with focus o
n s
upport
for
decis
ion a
nd t
ransactio
n p
rocesses
Use a
consis
tent to
ne o
f voic
e t
hat suits t
he t
arg
et audie
nce.
Be s
carc
e w
ith v
ivid
ente
rtain
me
nt as it decre
ases s
hoppin
g c
art
use.
Cre
ate
ente
rtain
me
nt
by p
rovid
ing thoughtf
ul use o
f co
lour
and typogra
phy b
ased o
n f
unctio
nalit
y.
Cre
ate
ente
rtain
me
nt
by s
ocia
l aspects
, in
tera
ctive e
lem
ents
and in
spiratio
nal desig
n.
Pro
vid
e e
nte
rtain
me
nt
by r
egula
rly u
pdatin
g the w
eb s
hop s
o c
onsum
ers
get th
e f
eelin
g it
evolv
es.
Ori
gin
ality
of
de
sig
n
Vis
ual
ap
pe
al
En
tert
ain
men
t
valu
e
Aesthetic appeal
15
Table 2.1 Validated e-servicescape model (continued)
D
ec
isio
n m
akin
g
pro
ce
ss
sta
ge
Choice / purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
Build
mu
ltip
le w
ays o
f navig
atio
n b
ased o
n e
ase
-of-
use b
y d
iffe
rent ty
pes o
f consum
ers
and the a
ctio
ns it fa
cili
tate
s t
hat contin
uously
show
s
the b
readth
and d
epth
of th
e w
eb s
hop.
Consid
er
that th
e s
ize a
nd locatio
n o
f te
xt
and g
raphic
s d
ete
rmin
e u
sers
’ attentio
n b
ased o
n F
-shaped s
cannin
g p
att
ern
s.
Cre
ate
a c
lean a
nd u
nclu
ttere
d d
esig
n, w
ithout unnecessary
text
and g
raphic
s a
nd m
inim
um
lo
adin
g t
ime
s a
nd s
yste
m c
rashes, th
at
behaves a
s u
ser
expect.
Pro
vid
e c
lear
org
anis
atio
n a
nd layout
without
dis
tractio
ns.
Pro
vid
e a
lin
k b
ack t
o s
hoppin
g.
Cre
ate
a c
onsis
tent
and lo
gic
al user
flo
w fro
m h
om
e p
age to c
heckout.
Pro
vid
e c
onta
ct in
form
atio
n,
pre
fera
bly
inclu
din
g a
(fr
ee)
num
ber,
to r
each t
he c
onsum
er
support
depart
me
nt.
Sta
te c
om
petitive a
dvanta
ges r
ega
rdin
g t
he q
ualit
y o
f pro
duct offerin
gs a
nd s
erv
ices c
learly t
hro
ughout
the w
eb s
hop.
Sta
te in
form
atio
n r
egard
ing p
rice, fe
atu
res, in
vento
ry in
form
atio
n a
nd o
rder
rela
ted c
harg
es a
s e
arly o
n a
s p
ossib
le.
Pro
vid
e in
form
atio
n that is
accura
te,
consis
tent
and s
pecific
, support
ed b
y full
siz
e p
ictu
res.
Pro
vid
e in
form
atio
n that is
accura
te,
consis
tent
and s
pecific
.
Dis
pla
y o
ut-
of-
sto
ck s
izes,
but re
move p
erm
anent out-
of-
sto
ck p
roducts
and c
olo
urs
.
Pro
vid
e in
form
atio
n fro
m a
consum
er
poin
t of vie
w w
hils
t keepin
g t
hem
in
a c
ontin
uous f
low
.
Th
e lo
catio
n, ty
pe a
nd im
ple
me
nta
tio
n o
f cro
ss-s
elli
ng,
especia
lly in c
ase o
f lim
ited d
ata
and b
usin
ess r
ule
, should
be c
onsid
ere
d d
ue t
o
conflic
tin
g r
esults.
Specify c
usto
mis
atio
n t
ow
ard
s d
ecis
ion a
nd t
ransactio
n p
rocesses.
Add f
eatu
res s
upport
ing d
irect
inte
ractivity b
etw
een v
isitors
and s
ale
s o
r support
em
plo
yees.
Add in
tera
ctive functio
nalit
y t
hat
is p
ote
ntia
lly u
sefu
l or
influ
ences s
ite u
sage a
nd n
avig
atio
n.
Change t
ext and c
olo
urs
when h
overin
g o
ver
actio
nable
text
and im
age
s.
Usab
ilit
y
Rele
van
ce o
f
info
rmati
on
Cu
sto
mis
ati
on
Inte
racti
vit
y
Layout & functionality
16
Table 2.1 Validated e-servicescape model (continued)
D
ec
isio
n m
akin
g
pro
ce
ss
sta
ge
Choice / purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
Dis
pla
y t
ruste
d a
nd in
dependent seals
and c
ert
ific
ate
s o
f appro
val th
roughout
the w
eb s
hop.
Ask o
nly
str
ictly n
ecessary
in
form
atio
n a
nd e
xclu
de m
ark
etin
g q
uestions.
Explic
itly
sta
te w
hat in
form
atio
n is s
tore
d a
nd n
ot sto
red.
Dis
pla
y t
ruste
d a
nd in
dependent seals
and c
ert
ific
ate
s o
f appro
val.
Cre
ate
a c
onsis
tent
and lo
gic
al user
flo
w fro
m h
om
e p
age to c
heckout.
Allo
w for
checkout com
ple
tio
n w
ithout
regis
tratio
n o
r usin
g a
n a
ccount.
Pro
vid
e a
ctio
nable
feedback a
nd e
rror
me
ssages a
nd o
nly
if str
ictly n
ecessary
.
Pro
vid
e in
form
atio
n r
egard
ing t
he d
iffe
rent checkout ste
ps a
s w
ell
as t
he c
urr
ent
locatio
n.
Pro
vid
e the o
ptio
n o
f cre
dit c
ard
paym
ents
, re
gula
r paym
ent ty
pes a
nd p
aym
ent ty
pes that fu
nctio
n a
s the e
xte
nsio
n o
f exis
tin
g m
eth
ods.
Ta
ke th
e pro
ducts
sold
and diffe
rent
targ
et
audie
nces in
to account
when desig
nin
g s
ingle
or
multi-page checkouts
both
for
speed and
confirm
atio
n.
Perc
eiv
ed
secu
rity
Ease o
f u
se
Financial security
17
3. Experiment design based on the e-servicescape model
On the basis of the validated e-servicescape model, two design rules were selected for further
investigation by means of field experiments. The selection was based both on academic value to the
field and on practical opportunities available to execute experiments in order to generate knowledge.
Because of the nature of the field experiments, it was not possible to measure purchase intentions and
other psychological concepts at a visitor level. Instead conversion, “the amount of purchasing visitors
as opposed to the total amount of visitors having expressed an interest towards a product” (Teltzrow
& Berendt, 2003), was used to identify the effects of experiment variations. This was based on the
logic that by increasing trust, consumers will have higher purchase intentions, subsequently resulting
in overall higher web shop purchase rates.
The first design rule selected focussed on cross-sell functionality: “The location, type and
implementation of cross-selling, especially in case of limited data and business rule, should be
considered due to conflicting results.” Considering that the topic of recommendation engines is a
highly active research field, a decision was made to focus on generating on the effects of cross-selling
at one particular page of the web shop: the shopping cart. The cart page was selected as cross-selling
is an important variable for the cart page influencing the balance of getting visitors to enter checkout
and complete their order on the one hand and stimulating higher cart values on the other end.
Hypotheses regarding the design rule are established in paragraph 3.1.
The second design rule selected focussed on the type of checkout used: “Take the products sold and
different target audiences into account when designing single or multi-page checkouts both for speed
and confirmation.” Little academic research has been done on the topic of single-page and multi-page
checkouts. As such a field experiment is used in order to create a first indication towards the effect
size and direction of different checkout variations on checkout conversion rates. The hypothesis
regarding the design rule is established in paragraph 3.2.
Improving cart page conversion 3.1.
When comparing to a clean and minimised cart page design, the main benefit of adding cross-selling
would be to increase revenue and as such web shop profitability. Several interviewees however also
pointed out critical remarks. These remarks focussed on situations with little or insufficient data and
resources to successfully implement cross-selling on the cart page, in which cases cross-selling could
have a negative impact on cart to purchase conversion rate due to the offering of non-matching
products, subsequently leading to declining web shop revenue. As such a hypothesis was stated to
investigate the effect of cross-selling in the cart compared to a transaction oriented cart design on web
shop revenue and conversion rate. It was hypothesised that the extended revenue from cross-selling
counterweighs the decrease in conversion rate:
Hypothesis 1A A clean cart page design oriented on completing a transaction performs equal
to a cart page design oriented on enhancing cart value when compared on
cart-to-purchase conversion rate and cart value.
Next to cross-selling the effect of an additional design rule was investigated. Several design rules
focused on providing important information regarding ordering as early on in processes as possible so
that customers are informed beforehand and are not to be brought in doubt regarding the order
conditions late in the process. As such it was thought to be beneficial to again explicitly state unique
18
selling points (USPs) of the web shop in the cart, with both the goals to inform and persuade potential
buyers. Given the nature of the situation described, a hypothesis was stated to compare the cart page
design enhanced with USPs with a clean transaction oriented cart page design. It was hypothesised
that the USPs would have a positive effect on the cart-to-purchase conversion compared to the
transaction oriented cart page design:
Hypothesis 1B A transaction oriented cart page design supported by USPs has a higher cart-
to-purchase conversion rate than a transaction oriented cart page design
without USPs.
Improving checkout conversion 3.2.
Four factors were addressed in the ease of use design rule focussed on the type of checkout to be used:
speed, confirmation, product and target audience. In the setting in which this experiment was able to
run, product and target audience were already set. The products were lingerie articles in the low-to-
medium price range and the target audience consisted of females between the ages of sixteen and
fifty. The need for the remaining two factors, speed and confirmation, were investigated using two
checkout designs: a single-page and a multi-page checkout. It is expected that the target audience in
this specific case has sufficient knowledge with purchasing and paying online and with the internet in
general. Furthermore it is expected that once they have selected the products of their choice and are
ready to proceed to checkout, they want to pay swiftly with less confirmation rather than in a more
time and click consuming manner with more confirmation. The latter is supported by Bucklin and
Sismeiro (2003) stating that operators should consider pages with more information on each page to
reduce the number of page views needed to complete a transaction. As such it was hypothesised that a
single-page checkout outperforms a multi-page checkout when it comes to checkout-to-purchase
conversion rates:
Hypothesis 2 A single-page checkout has a higher checkout-to-purchase conversion rate
than a multi-page checkout.
19
4. Experiment method
The hypotheses established in the previous chapter were tested in two separate on-site field
experiments. This chapter details the method used for collecting data in paragraph 4.1 and the method
of data analysis in paragraph 4.2. The quality safeguarding of the experiment data is also discussed, in
paragraph 4.3.
Data collection 4.1.
The on-site field experiments were run at a large lingerie retailer’s web shop. The lingerie shop
formula, operated in the Netherlands by an e-commerce fulfilment partner, focuses on being personal,
service oriented and stocking high quality lingerie products with a decent price/quality balance. The
web shop was launched after three months of developing in February 2012. At the time of the
experiment nearly 500 products with over 4000 SKU’s (different sizes and colours) of three brands
were sold online, whilst the web shop processed over 130.000 unique visitors each month.
The web shop pages and page variations that were part of the field experiment were identified and
(re)designed according to the design rules under investigation. After approval by the retailer and the
fulfilment partner following design iterations, the experiments were executed. In the case of the cart
page and checkout experiments, A/B software tools were used to respectively equally assign visitors
to different cart page variations and to randomly assign visitors to different checkout variations in a
one (single-page) to four (multi-page) ratio. An open source web analytics software package was used
to record and anonymously store individual click stream data on a page view level, as depicted in 0.
4.1.1. Experiment ‘Cart page’
In order to test the hypotheses that providing USPs on the cart page positively influences conversion
and that orienting the cart page towards increasing shopping cart value influences conversion and cart
value, two variations were designed. These variations were based on a control condition, which is
depicted version next to the other variations in the conceptual design in Figure 4.1 and the actual
design in 0.
The first variation was the ‘Control’ variation. It encompassed no signs or functionality of either
Figure 4.1 Cart page designs; Left: variation 1 (Clean, control), middle: variation 2 (USPs), right: variation 3 (Cross sell)
20
cross-selling or additional USPs and was considered the most functional oriented cart page design.
The variation was based on the design rules not to distract users with aesthetics and to provide clear
organisation and layout without distractions.
The second variation ‘USPs’ focuses on clearly depicting unique selling points and other important
information on the web shop early on in the checkout process. This way consumers should be less
distracted and perceive less risk later on in the checkout process, as based on the design rules to state
advantages regarding the services of the web shop and providing checkout and order related
information as early on in the process as possible. The USPs used were delivery time, return policy,
the ability to pay in a secure way and kind service offered by the web shop. Furthermore the logos of
several banks were depicted. These specific USPs were used as they are promoted throughout the web
shop, hence providing consistency and not providing too much new information to the consumer, as
the remainder of the cart page already requires the processing of new information.
The third and final variation ‘Cross sell’ is focussed on providing the opportunity of recommending
articles to consumers in order to stimulate cart value, as such investigating the effect of cross-selling
in the cart based on relative low amounts of relational data between products available . It showed the
label ‘matching articles’ on top with two matching products below, based on a product-based
predefined set of matching products that related on the topic of whether or not a product is from the
same designer line. In case of several products, the recommendations were selected randomly
(although persistent in the case of a page refresh) from the set of recommendations available. The
recommendations were displayed with a product image, brand name, product name and price. In case
the product depicted was part of a promotion, all prices including mark-offs were shown. When
clicking on one of the products the product popped out and showed again the product image, brand
and product name, but this time supported by detailed product information, the article number and the
option to choose a colour and size as well as a button to directly add the product to the shopping cart.
4.1.2. Experiment ‘Checkout’
In order to test the influence of minimising the steps and actions a consumer must complete to place
an order whilst balancing confirmation of information, two checkout designs were tested: a multi-
page and a single-page checkout. These checkouts included the checkout steps as depicted in Figure
4.2. The first step focussed on acquiring the personal information of the consumer, albeit that it was
proceeded in the multi-page checkout by acquiring the consumers’ e-mailadress to check whether an
account already existed or not. The second step was focussed on determining the invoice address and
the shipping address. The third and final step aimed at completing the purchase by selecting a
payment method and subsequently either entering payment information or temporarily leaving the
retailer’s web shop to do so. After successful submission of the personal and shipment information
and completing the payment procedure, a success page with order information was shown.
The first design was a multi-page based checkout, focussed on confirming at every step the
Success
Order confirmed
Cart Step 1
Login / register
Step 2
Select adress
Step 3
Choose payment
Step 1*
Enter e-mailadress
Step 3*
Cancel payment
Figure 4.2 Checkout flow
21
information users entered in preceding steps. As depicted in Table 4.1 and shown in 0, the different
checkout steps were spread out over several pages. Next to the required information, the right hand
side of the page showed an overview of the order to be placed, as well as several unique selling points
for the web shop and a clickable DigiCert security seal. Furthermore all information not strictly
necessary for completing the checkout was removed from both the header and the footer.
The second design was a single-page based checkout, focussed on letting users complete the checkout
process as quickly as possible. As depicted in Table 4.1 and shown in 0, the different checkout steps
were shown on a single-page below one another. Due to this design no e-mailadress was required to
be entered by consumers to enter the checkout and as such consumers were only able to register for an
account after completion of their order. As with the multi-page checkout, the right hand side of the
page showed an overview of the order and additional information, whilst the header and footer were
stripped of non-vital information.
Table 4.1 Implementation of checkout steps
Checkout step Multi-page Single-page
Enter e-mailadress Page 1 Not included
Login / register Page 1 Login: 1st page section
Register: 2nd
page section
Select shipment and invoice address Page 2 3rd
page section
Choose payment method Page 3 4th
page section
Success Success-page Success-page
Data analysis 4.2.
The experiment design by Montgomery and Runger (Montgomery & Runger, 2007), shown in Figure
4.3, requires the identification of a dependent variable, determined on the basis of a hypotheses,
controllable (independent) variables, being the different design variations based on the e-servicescape
factors playing a role in conversion optimisation, and finally uncontrollable factors. One of these
uncontrollable factors was considered to be the day of the week on a working day (Monday to
Thursday) versus weekend level (Friday until Sunday), as it not was not possible to gather enough
longitudinal data to conduct a viable analysis on the factor. An additional uncontrollable factor was
Input
Visitors
Web shop
conversion process
Output
Transactions (y)
z1 z2 zn
Uncontrollable (noise) factors
…
Controllable factors
x1 x2 xn
…
Figure 4.3 Experiment design (Montgomery & Runger, 2007)
22
considered to be the time of day, which was split into morning (from 6.00 AM until 12.00 PM),
afternoon (from 12.00 PM until 6.00 PM) and evening / night (6.00 PM until 6.00 AM), as one would
expect that visitors behave differently during morning and evening hours. A final uncontrollable
factor considered was cart value.
Based on the click stream data several significant variables were calculated focussed both on the
general data and the specific individual experiments as explained in the subsequent sub-paragraphs.
After collection, corrupt data was removed from the dataset, specifically mobile visitors as after the
experiments it showed that a cookie bug was preventing half the mobile users from paying. Next the
data was analysed on the basis of Logistic Regression and Analysis of Co-Variance (ANCOVA)
techniques, suitable for analysing A/B/n- and multivariate experiments, using the statistical software
package SPSS. The choice for the specific method is dependent on whether the dependent variable is
dichotomous or continuous. On the basis of the analyses the hypotheses were tested and conclusions
were drawn.
4.2.1. Analysis experiment ‘Cart page’
As implied in the hypotheses regarding the cart design, the aim of the cart experiment was to optimise
conversion, defined as “a consumer that visits the cart page and the success page” (based on (Butler &
Peppard, 1998)). As such all the consumers completing their purchase received value 1, whilst others
received value 0.
{ ( )
( )
The combination of the dependent variable, the independent variable being the cart variation and the
uncontrollable factors, formed up the model as depicted in Table 4.2.
4.2.2. Analysis experiment ‘Checkout ‘
As implied in the hypotheses regarding the checkout design, the aim of the checkout experiment was
to optimise conversion, defined as “a consumer that enters the checkout and reaches the success page”
(based on (Butler & Peppard, 1998)). As such all the consumers completing their purchase received
value 1, whilst others received value 0. A limitation in the measurement possibilities of the single-
page checkout was neglecting the situation in which consumers started entering personal data on the
single-page checkout page without completing all information.
{ ( )
( )
The combination of the dependent variable, the independent variable being the checkout variation,
and the uncontrollable factors, formed up the experiment model as depicted in Table 4.2.
Table 4.2 Experiment models
Variable Experiment
Cart page Checkout
Dependent Complete purchase Complete purchase
Independent Cart variation Checkout variation
Uncontrollable Weekend Time of day Cart value
Weekend Time of day Cart value
23
Research quality 4.3.
Using an experiment as a means to identify factors and their magnitude requires taking several quality
dimensions into account, being reliability and construct, internal and external validity. Reliability
deals with the repeatability of a study. If the study were to be repeated under the same conditions, the
same results should follow. Securing reliability was done by using a structured, documented way of
working both whilst setting up the experiment and whilst performing the data analysis.
Construct validity deals with whether the measurement tool actually measures the concept being
studied. This is one of the major issues in the field experiments, as the field setup did not allow for the
measuring of characteristics and psychological concepts such as trust and purchase intention at a
visitor level. Instead key performance indicators focussing on conversion were identified based on
academic literature. Future research should focus on measuring the underlying concepts that likely
resulted in the outcome of the current experiment.
Next to construct validity, internal validity deals with the causality of results. This was safeguarded
and made open to discussion by identifying the conversion process in the literature review and using
indicators to measure several steps in this process during the experiments, even though the experiment
did not allow for the measuring of characteristics and psychological concepts as discussed before.
Additionally, external validity deals with the generalizability of the study, which was controlled for
by keeping the uncontrollable factors at a minimum. In order to do so, these factors and their potential
effects on the output variables were identified as much as possible and included as covariates. Clearly
describing the context of the experiment furthermore makes clear to what level the results are
extendable to different contexts.
24
25
5. Experiment results and discussion
On the basis of experiments designed in the previous chapters, data was gathered and analysed. This
chapter details the results of the individual independently run analyses, starting with the cart page
experiment in paragraph 5.1, followed by the checkout experiment in paragraph 5.2.
For all experiments it goes that data generated by the web shop operators both directly and via means
of additional software tools was excluded. Furthermore visitors using a mobile device were excluded
due to technical issues on mobile platforms influencing the behaviour of visitors. Additionally,
analysis assumptions regarding independence of observations were considered tenable, as visitors
shown more than one experiment variation were excluded from the dataset.
Experiment ‘Cart page’ 5.1.
Data for the cart page experiment was gathered between June 20, 2012 and July 6, 2012. In this 17
day time period 58,937 visitors visited the web shop of which 2,736 visitors (4.7%) was displayed one
of the cart page variations. This is taking into account the exclusion of visitors that were shown
multiple variations and visitors that showed anomalies in their cart and checkout pattern. Additionally,
inspection of the data and potential outliers resulted in the identification of one outlier in the ‘cross
sell’ variation where a cart value of over 300 Euros occurred, whereas throughout all variations the
next maximum values were all around 200 Euros. Even though the purchase has been validated, the
case was excluded from the dataset in the ‘value’-base model as there was a clear impact on the
results threatening generalizability. The main characteristics of the main variables in the cart
experiment dataset can be found in Table 5.1 and in 0 for the covariates day of the week, time of day
and cart value.
Table 5.1 Descriptive data cart experiment
Variation Visitors
Final step completed Conversion
Cart
Lo
gin
/
reg
iste
r
Sh
ipm
en
t
Pa
ym
en
t
me
tho
d
Can
cel
Su
cce
ss
Can
cel &
su
ccess
Mean Std. dev.
Total 2,763 1129 304 117 159 25 1012 17 37.2% .484
Clean 923 352
38,1% 99
10,7% 43
4,7% 63
6,8% 10
1,1% 348
37,7% 8
0,9% 38.6% .487
Unique selling points
899 380
42,3% 88
9,8% 38
4,2% 46
5,1% 9
1,0% 332
36,9% 6
0,7% 37.6% .485
Cross selling
941 397
42,2% 117
12,4% 36
3,8% 50
5,3% 6
0,6% 332
35,3% 3
0,3% 35.6% .479
The descriptives depict that a large portion of users do not get past the cart page. The ‘clean’ page
appears to have the highest cart-to-checkout conversion rate, but shows the largest portions of visitors
leaving the checkout on the shipment- and payment method-page when compared to the other
variations. This is opposed to the cross-selling variation that sees most checkout visitors drop out at
the login/register-page.
26
5.1.1. Baseline conversion model - Logistic regression
The analysis performed is a logistic regression. Three models were analysed. The first model only
included the purchase variable as dependent variable and the cart variation as the independent
variable. Next cart value was entered as independent value and finally day of the week and time of the
day. Before conducting the regression the accompanying assumptions are discussed. Due to the
logistic nature of the dataset, we assume a linear relationship between the logit of the outcome
variable and the combined independent variables. Additionally we assume that no important variables
are omitted and no extraneous variables are included. Based on the dataset and variables available
these assumptions are tenable. Two additional assumptions are that the observations are independent
and that the independent variables are measured without error, which are both tenable due to the
experiment setup. A final assumption is the absence of multicollinearity in the independent variables.
Given the VIF values, included in 0, for the time of day dummy variables (morning and afternoon)
and a condition index for a dimension that is slightly larger than others, there appears to be
multicollinearity between the variables. Given the nature of the variable this was to be expected,
however the implication of erratic changes in the coefficient estimates in case of small changes in the
model and the data should be taken into mind.
The results of the logistic regression can be found in Table 5.2. Starting by comparing the
performance of the different models, the constant predicted chance of purchase is with values of 60%
to 70% drastically too high in all models when compared to the descriptive conversion rate of 37.2%.
Furthermore the adding of the covariates did improve the performance of the model, even though the
value of the chi2-test did not approximate a significant result.
Table 5.2 Results logistic regression experiment ‘cart page’
Model 1 Model 2 Model 3
Variable B Wald
Odds-
ratio B Wald
Odds-
ratio B Wald
Odds-
ratio
Experiment
variation
USP 0.030 0.088 1.030 0.039 0.149 1.039 0.037 0.134 1.037
Cross-selling -0.103 1.093 0.902 -0.110 1.246 0.895 -0.108 1.183 0.898
Cart value Not included 0.04 10.185 1.004* 0.004 10.221 1.004*
Day of week: Weekend Not included Not included -0.101 1.526 0.904
Time of day Morning
Not included Not included -0.051 0.206 1.053
Afternoon -0.010 0.013 0.990
Constant -0.0336 23.391 0.715* -0.511 33.167 0.600* -0.573 25.705 0.628*
Model performance Model 1 Model 2 Model 3
Hosmer and Lemeshow
chi2-test
0.000 (p = 1.000) 6.722 (p = 0.567) 10.804 (p = 0.213)
R2 Nagelkerke 0.001 0.007 0.008
Cox & Snell 0.001 0.005 0.006
* = p < 0.001
Interpreting the results from the logistic regression is done by looking at the odds ratio rather than at
the coefficient, as it provides an intuitive interpretation: for example a constant odds ratio of 0.3
implies a 30% predicted chance and a variable odds-ratio of 1.1 implies an increase of 10% per 1
increase in the variable, meaning that a variable value of 3 results in a predicted chance of 39%. When
looking at the second model (including cart value as a significant covariate), it shows that the
predicted chance of completing an order increases with 0.4% per Euro of cart value on the constant,
which is relatively large given the average order value of 38.41 Euros (implying 9.2% base point
27
average increase in predicted chance of purchase). Time of day and weekday versus weekend does not
significantly influence the predicted chance of completing a purchase.
Taking into account all models with all covariates, there were no significant influences of the cart
page designs on conversion. The approximate 3% to 4% increase over the 60% to 70% baseline
predicted chance of completing a purchase is highly insignificant with p-values around 0.72. The
stronger negative influence of the cross-selling enabled cart page design of approximately 10% over
the 60% to 70% baseline predicted chance of completing purchase is insignificant at the p-value of
0.27. More data is needed to investigate the effects of the cart page designs and more longitudinal
data is needed to get more insights into the effects of the covariates.
5.1.2. Linear conversion model - ANCOVA
Given the fact that the cart page design involving cross-selling variation did not significantly differ
from the clean cart page design at a small enough α-level, and given that the descriptive analysis of
the dataset showed the cart variation likely having an influence on the completion behaviour at
different checkout steps, an additional model was created in which all checkout steps were weighted
equally. As such a value of 0.00 means that a consumer did not get past the cart page, whilst a value
of 0.25 implied the visitor reaching the login / register page, a value of 0.50 implied reaching the
select address page, a value of 0.75 implied reaching the payment method selection page and a value
of 1.00 implied a consumer completing the purchase. Additionally the ‘payment cancel’ step was
considered to be half way the third step and the success page, as such having a value of 0.875.
An Analysis of Co-Variance was run on the dataset with the covariates weekday versus weekend,
time of visit and cart value as covariates and the conversion variable as dependent variable. The
covariates were included in the model as they improved the quality of the model during the logistic
regression analysis. The ANCOVA implied the testing of four assumptions. Next to the assumption of
independence of the observations, which is considered tenable as discussed before, the assumption of
a normally distributed population was violated as expected due to the nature of the conversion
variable. This does however not necessarily result in issues with the analysis, as ANCOVA is fairly
robust to violation of the normality assumption and as such the non-normal distribution only had a
small effect on the Type I error rates. Additionally the assumption of homogeneity of variance was
tenable, pcart_conv_linear = 0.334 for Levene’s Test of Equality of Error Variances (α = 0.05). The
assumption of homogeneity of regression slopes was not tenable for the cart value variable (p < 0.05),
but no suitable dummy-coding scheme was identified that both resolved the violation and kept the
model easily interpretable. As such it has to be taken into account during the discussion that the cart
value covariate may display different effect types at different variable levels.
The result of the ANCOVA can be found in 0. Neither the covariate weekday versus weekend nor
time of day was significantly related to the linear conversion rate at any level or showed F-values that
were close to or larger than the critical F-value (F(1, 2,496) = 5.02). The covariate cart value did
prove to be significantly related to the linear conversion rate at F(1, 2,496) = 12.89, p = 0.000,
displaying F-values far above critical levels. The effect size proved to be relatively high with B =
0.001 (t(2,496) = 3.590; p = 0.000), which given the average order value of 38.41 Euros implies an
average increase in conversion of nearly 4%.
After controlling for the covariates, the effect of the cart page design on the linear conversion rate was
still not significant at the α = 0.05 level, but did show significance near the α = 0.1 level with F(2,
2,496) = 2.090, p = 0.124. Still the critical F-value (F(2, 2,490) = 3.69) was not reached. The result
did however give an indication towards the direction of the cart page design effects. Contrasts
28
revealed that the results again point towards small or no differences between the clean cart page
design and the cart page design incorporating USPs (p = 0.139) Looking at the differences between
the clean cart page design and the cross-selling enabled cart page design, a nearly significant
difference was found with p = 0.051, also depicted in Figure 5.1.
5.1.3. Including cart value – ANCOVA
As stated in the hypothesis, the focus should not only be on conversion as well as on overall revenue.
The experiment influenced cross-selling options available and as such potentially influences cart
values. In order to identify those effects, an additional model was created in which the values for the
linear conversion model, showing more significant results than the baseline model, were multiplied by
the cart value of the visitor.
An ANCOVA was performed with the valued conversion as dependent variable, the cart page
variation as independent variable and time of the day and weekday versus weekend as covariates.
Given the dependent variable, cart value was no longer included as covariate. Running an ANCOVA
implied the testing of four assumptions. Next to the assumption of independence of the observations,
which is considered tenable as discussed earlier, the assumption of a normally distributed population
was violated as expected due to the nature of the conversion variable. This does however not
necessarily result in issues with the analysis, as ANCOVA is fairly robust to violation of the normality
assumption and as such the non-normal distribution only had a small effect on the Type I error rates.
Additionally the assumption of homogeneity of variance was tenable, pcart_conv_linear_value = 0.213 for
Levene’s Test of Equality of Error Variances (α = 0.05), as was the assumption of homogeneity of
regression slopes, which was tested during the running of the models (p < 0.05).
The result of the ANCOVA can be found in 0. Of the covariates, only time of day, specifically
morning, was found significantly related to the valued linear conversion at F(1, 2,497) = 4.066, p =
0.044, which is close but not above the critical F-value (F(1, 2,497) = 5.02). The effect size proved to
be high with B = 3.317 (t(2,497) = 2.016, p = 0.044). The other time of day covariates and weekday
Figure 5.1 Mean values and confidence intervals (α = 0.05) of conversion variables
0.532
0.523
0.489
0.502
0.491
0.459
0.563
0.554
0.520
0,45
0,49
0,53
0,57
Clean USP Cross-selling
Cart experiment 'Linear conversion'
23.40
21.67
22.84
21.40
19.61
20.84
25.39
23.74
24.83
18,00
22,00
26,00
Clean USP Cross-selling
Cart experiment 'Linear valued conversion'
0.57
0.53
0.49
0.45
26
22
18
29
versus weekend variables were not significantly related to linear conversion rate at any day of the
week, as were not F-values depicted that were larger or close to the critical F-value.
After controlling for the covariates, the effect of the cart page design on the valued linear conversion
rate was found not significant with F(2, 2,491) = 0.178, p = 0.488, lower than the critical F-value
(F(2, 2,497) = 3.69). Contrasts also revealed only non-significant differences, as seen in Figure 5.1.
5.1.4. Discussion
The first hypothesis underlying the cart page experiment was stated as follows: “A cart page design
oriented on completing a transaction performs equal in revenue based on cart-to-purchase
conversion rate and cart value as a cart page design oriented on enhancing cart value.”
On the basis of the experiment results the first hypothesis was partly rejected (p = 0.051). Looking
merely at cart-to-purchase conversion, no statistically significant effect of the cart page design
variation was found using a logistic regression. However, when taking into account the different steps
part of the checkout using a linear conversion value model, a statistically significant effect was found
using an ANCOVA: the clean cart page design outperformed the cart page design providing cross-
selling functionality. This rejects the hypothesis on the cart-to-purchase conversion rate aspect.
It was noted that in both the general and detailed model the cart value proved to be a significant
covariate. However, when using a model in which the linear conversion value was multiplied by the
cart value, no significant effects of the cart page designs were found and only the time of day
covariate morning was found significant. The cart pages designs as such performed equal on the
revenue aspect (p = 0.488).
From the results one can deducted that it is important to find a balance between the two designs which
can be dependent on moderating and environmental business factors. On the one hand conversion is
important, but web shop operators should also take into account the revenue made from orders.
Higher value orders result in higher profits given product margins and shipping costs, whilst on the
other hand a high conversion rates provide the opportunity to achieve economies of scale or to clear
out old stock.
The second hypothesis underlying the cart page experiment was stated as follows: “A transaction
oriented cart page design supported by USPs has a higher cart-to-purchase conversion rate than a
transaction oriented cart page design without USPs.”
On the basis of the experiments result the second hypothesis is rejected (p = 0.139). There was no
statistically significant effect of providing USPs in the cart page design on cart-to-purchase
conversion rates.
Three important aspects influenced the results. First of all the presence of only approximately 900
data points per cart variation made measuring clear differences between the two variations difficult,
given there was a difference in cart-to-purchase conversion rate of only 1%. More data is needed to
clearly measure the effects of providing USPs.
The second aspect influencing results was the fact that the experiment was run during a sale period. It
was expected that the sales had a larger impact on purchase intentions than the USPs. As such the
USPs did no longer have an additional effect on purchase intentions and subsequently conversion
rates. This also correlates to the third and final aspect influencing the results, which was the content of
the USPs. The USPs were considered relatively weak, being more selling points in general than being
30
unique to the web shop. Depicting stronger USPs, which was not approved for this experiment, or
moving the USPs to a location on the cart page with more central focus would likely resolve in
different results.
Experiment ‘Checkout’ 5.2.
The experiment time range was set from August 24, 2012 to September 2, 2012. During the ten day
period 21,699 visitors visited the web shop, of which 548 visitors (2.5%) was displayed one of the
cart page variations (NCheckout ‘Multi-page’ = 450, NCheckout ‘Single-page’ = 98). In total 364 orders NCheckout ‘Multi-
page’ = 296, NCheckout ‘Single-page’ = 68) were placed, resulting in a checkout conversion of 66.4 % (MCheckout
‘Multi-page’ = 0.66, MCheckout ‘Single-page’ = 0.69). Inspection of the data and did not result in the
identification of outliers.
5.2.1. Results
The conversion rates of both checkout variations were close to one another with a difference of only
3% and a small amount of data points. As such first a chi-square test was performed, which equalled p
= 0.493. Therefore it was expected that further analysis would not result in a model with significant
variables. However, in order to get an estimate towards the size of the effect and to analyse the role of
cart value and time of day, a logistic regression was run. The model included the dependent variable
as the purchase variable, the checkout variation as the independent variable, as well as the time of the
day and cart value. Due to the short experiment run time, the covariate day of the week was excluded
from the model.
Before conducting the regression the accompanying assumptions are discussed. Due to the logistic
nature of the dataset, we assume a linear relationship between the logit of the outcome variable and
the combined independent variables. Additionally we assume that no important variables are omitted
and no extraneous variables are included. Based on the dataset and variables available these
assumptions are tenable. Two additional assumptions are that the observations are independent and
that the independent variables are measured without error, which are both tenable due to the
experiment setup. A final assumption is the absence of multicollinearity in the independent variables.
Given the relatively high VIF values, included in 0, for the time of day dummy variables (morning,
and afternoon) and a condition index for a dimension that is substantially larger than others, there
appears to be multicollinearity between the variables. Given the nature of the variable this was to be
expected, but not considered a major property given the spurious nature of the dataset.
The results of the three logistic regression runs, using no covariates, cart value as a covariate and with
time of the day as well as cart value as covariates, can be found in 0. Interpreting the results from the
logistic regression is done by looking at the odds ratio rather than at the coefficient, as it provides an
intuitive interpretation. Furthermore, given the low value for N, focus was on identifying both
significant results and relatively large odds ratios that provide an indication of the effect to be
investigated in future research. All three models did not perform extremely well. The experiment
variation was not significant in all models. The model including all variables provided significant
results that were in line with the cart experiments on the topic of cart value and morning hours being
significant. However for all models and interpretations, there is a high risk of over fitting the data
given the relatively low amount of data points. This also suits the predicted baseline chance of
completing purchase which is higher than 100%.
31
5.2.2. Discussion
The hypothesis underlying the checkout experiment was stated as follows: “A single-page checkout
has a higher checkout-to-purchase conversion rate than a multi-page checkout.”
Given the spurious nature of the dataset, one can only state that more data needs to be gathered
including data on covariates, as they appear to have a clear impact on conversion. No significant
results can be extracted regarding the single-page versus multi-page discussion other than carefully
stating they point, as expected from the descriptive data, towards a positive effect of the single-page
checkout design on checkout-to-purchase conversion compared to a multi-page checkout design. It
has to be noted that these preliminary unsupported results were specific to this case and the specific
checkout designs of this experiment.
Despite the inability to draw conclusion from the experiment, it is worthwhile to mention that
potential positive results perceived by companies redesigning their multi-page checkout into a single-
page checkout are caused by the mere fact that they are working on building new and optimised
checkouts. It could very well be that the improved results would also have been achieved by radically
optimising their multi-page checkout. In order to identify the relevance of academic research into
checkout design and checkout types, additional practical and laboratory research over multiple web
shops is needed.
32
33
6. Conclusion
Effective design of web shops is a key web shop success factor. The e-servicescape model by Harris
and Goode (2010) provides a good starting point for building web shops that increase consumer
purchase intentions and as such revenues and profitability. A major deficit of the model is however
that it does not take into account the different goals and behaviours of individual web shop visitors.
Past research has shown consumer to proceed through different phases in a consumer decision making
process before actually making an online purchase. As such the following research question was
posed:
Which e-servicescape factors and design rules can be used during different
stages of the consumer decision making process to optimise web shop conversion?
Three factors were established on a literature review and a validation oriented single embedded case
study: aesthetic appeal, layout and functionality, and perceived security. In total 44 design rules were
placed under these factors that were coupled to the applicable consumer decision making process
stages of search for information, evaluation of alternatives and choice / purchase. The final model is
depicted in Table 7.1 at the end of the next chapter.
With regards to the e-servicescape factor visual appeal, the main conclusion was drawn that
originality is not a necessity and that it is more important to provide a design that adheres to
consumers’ expectations based on (existing) brand values and that flows fluently from homepage to
checkout. Furthermore the importance of product images was stated several times as it can provide
context to images and can even transfer emotions and feelings regarding a web shop and specific
products. Although not researched often in the past, the role of product photography appears to be one
of vital importance to the success of a web shop. More specifically even, discussion focussed towards
the effect on conversion rates of using product photography displayed on models.
With regards to the e-servicescape factor layout and functionality, two main conclusions were drawn.
First of all it showed important to continuously take the end user into account when designing these e-
servicescape aspects in a web shop and adhere to expectations of consumers in order to create a
logical continuous flow from homepage to checkout. Secondly the role of cross-selling in a web shop
was discussed. As in academic literature, contradicting findings were found on the type, location and
implementation of cross-selling in a web shop. The discussion revolved around the type of cross-
selling and the way it should be presented throughout the web shop on the one hand and specifically
on the usage of cross-selling on cart page. Both potential benefits, such as an increase in cart value,
and potential disadvantages, a decrease in cart-to-purchase conversion due to consumers brought into
doubt, were mentioned.
With regards to the final e-servicescape factor financial security one of the main conclusion was that
it consumers should feel as safe as possible by invoking feelings of trust and security using logos,
certificates and statements. At the same time the effect of these cues is limited in case of existing
brands and retailers and they can even undermine feelings of trust if they are to prominent and distract
the user from entering his personal information and focussing on checking the security of the web
shop instead. The second conclusion was that a checkout should be made as easy to complete as
possible and that it should provide every payment method that a consumer could potentially want to
use, as long as it is well known and don’t make other consumers doubt the security of the web shop.
Regarding ease of use, the interviewees also focussed on the usage of a single-page or a multi-page
34
checkout. In literature little research has been done in this are as of yet, even though getting
consumers to complete the checkout process can be seen as a core activity of web shop owners.
The validated e-servicescape model shows that different design rules and approaches should be
considered for different stages in the consumer decision making process and as such on different
pages. In order to generate more academic knowledge on two topics covered by the design rules, two
field experiments were performed. Even though conclusions were drawn from these results, notion
should be made that the experiments featured a relatively low amount of data points and are as such
tentative and specific to the case of the lingerie retailer under discussion.
The first experiment focussed on the role of cross-selling on the cart page and its effect on revenue
and cart-to-purchase conversion. On the basis of the results it was concluded that elements on the cart
page that potentially distract users from proceeding to checkout including, though not exclusively,
cross-selling functionality, have a negatively influence on the cart-to-purchase conversion rate. Focus
on conversion should however be balanced out against higher cart values and as such revenues which
may be increased by means of cross-selling functionality.
The second experiment focussed on a recent development in the field: the testing and usage of single-
page and multi-page checkouts. The dataset regarding this experiment was highly spurious, making it
impossible to draw definitive conclusions. More research is needed into the topic of checkout design
and one should strongly take into account the presence of covariates, of which at least time of day and
cart value were identified as important.
35
7. Reflection
The aim of theoretical research is to contribute to science. However, as with every research there are
limitations to the current research that need to be taken into account, which is done in paragraph 7.1.
Taken these limitations into account, the academic implications of the research are discussed in
paragraph 0, including the identification of future research opportunities. The chapter ends with the
identification of managerial implications in paragraph 7.3.
Limitations 7.1.
This thesis research had several limitations to it that are important to consider when establishing
generalisations and implications. The limitations can be divided into three categories: limitations due
to the research design, limitations due to the research execution and technical limitations.
7.1.1. Research design
The research design, using a combination of academic literature, field based interviews and two field
experiments mainly provided limitations on the aspect of generalizability. Although an external case
study was part of the original research design, contact with twenty e-commerce companies did not
result in the opportunity of interviews. Reasons varied from a lack of interest in cooperating in the
research, to insufficient resources partly due to the summer period in which the research was
executed, to declining cooperation due to the competitive position of the respondents to the company
where the thesis internship was performed. The remaining in-company interviews limit the
generalizability of the e-servicescape model as it only focusses on the knowledge of employees in one
company, albeit that the interviewees come from different departments of a company that has
operated different types of web shops both in the past and at this point in time. As a result the model
and design principles established in this case study should be tested, confirmed and deepened out
further both at other companies and in different industries than the online apparel and fashion retail
industry. Although web shops with fast moving consumer goods, being printer supplies, were
covered, the interviewees showed that different design rules may apply based on the web shop owners
goal of a web shop: purchase and retention or solely purchase.
Next to the theoretical and field work in order to establish the e-servicescape model, two experiments
were executed. Three main limitations were present in the experiment design, of which the first was
focussed on the limited time span of the experiment. This resulted in the difficult interpretation of the
data due to the inability to correctly measure time-based covariates. More data was needed to gain
more insights into these covariates, which should also provide the opportunity to investigate
interaction effects between variables and the opportunity to design models that perform better on an
overall scale.
Next to the limitations due to a limited time span, generalizability is threatened by only including a
single web shop. In order to be able to better generalize the results, the experiments will need to be
repeated on different web shops both in similar and different industries, in order to establish the effect
of different design rule interpretation on consumer purchase intentions on a web shop. Moreover,
specific attributes, characteristics and propositions of web shops may result in very different results
amongst web shops that at first sight appear to be equal. The experiment results do however provide
guidelines on the direction of effects and aspects to consider when designing and optimising different
web shop pages and clusters of pages.
36
Perhaps even the most important limitation regarding the experiment design is however that due to the
field nature of the experiments, it was not possible to measure trust and purchase intentions and that
instead conversion was chosen as a measurement. This threatened construct validity and internal
validity and provides a strong recommendation for future research
7.1.2. Research execution
The academic literature review and validating interviews provided two main limitations. The first
limitation is that, given the scope of the master thesis research, the academic literature needed to
balance between being high-level and generic and being detailed. A choice was made to assume the
model of Harris and Goode (2010) in the interpretation that three specific e-servicescape factors and
sub-factors influence trust and subsequently purchase intentions, and to focus this research on
identifying design rules that influence these (sub-)factors. As such the validated e-servicescape model
should be tested in a laboratory setting to identify the causal relationships of the design rules and
measure their influence on both trust and purchase intentions directly.
A second limitation is in the interview execution. The interviews focussed on identifying aspects
influencing e-servicescape sub-factors and the testing of design-rules without literally depicting the
design rules but by incorporating them into the style of questioning. This was done in order not to
direct the interviewees and to gain as much data as was possible. However, this also implies a small
limitation to the validity of the e-servicescape model validation. Given the fact that the design rules
were however incorporated into the questioning, this was not considered an issue
The execution of the experiments resulted in three additional limitations. The first limitation focusses
on the translation of the hypotheses into experiments. Given the fact that the experiments were
executed at an actual web shop, being operated by a third party company, not all proposed and desired
experiment designs could be tested. The necessary approval of both the web shop owner and operator
resulted in more conservative experiment designs. This limited the potential effects of experiment
variations as they showed more resemblance to one another and as such made it more difficult to
derive statistically significant results and conclusions. This was enhanced by the second limitation,
also discussed during the discussion of the experiment results, which was the restricted period of time
the experiments were allowed to run. This lead to small amounts of data points available that
hampered the results analyses and drawing of conclusions: models used were of low performance and
interpretation was challenging given the expected role of covariates for which too little data was
available on the one hand and given the small differences between the different experiment variations
on the other hand. Although both limitations results in limited construct validity and external validity
of the research, the results and conclusion do provide directions towards the effects that can be
expected when making e-servicescape design decisions as well as the direction of the results of
comparable experiments.
The time period in which the experiments were conducted created the third limitation; in case of the
cart page experiment a sale period occurred and a new collection of swim wear was made available.
During the checkout experiment a sale occurred as well. In order to enhance validity, similar
experiments should be executed once more during periods with new collection, during sale periods
and during periods where there is no strong marketing campaign active. It is expected that the
different marketing campaigns attract different types of consumers, for example oriented towards
bargains during sale periods, which might result in a preference for displaying a specific type of cart
page or checkout during that period. The current results as such provide a direction for expected
effects and future research.
37
7.1.3. Technical limitations
The tools available at the time of the experiment as well as the implementation of the tools provided
several limitations to the experiment results It is tenable that the observed results are influenced
largely by the (absence of a) marketing campaign active during the experiment period. Therefore it is
important to replicate the results of this experiment at other web shops during comparable time frames
in order to establish where the balance lies between the design rules focussing on providing
navigational functionality and inspirational design on the one hand and guiding users to products of
their interest as direct and with as few clicks as possible on the other hand.
With respect to the cart page and checkout experiment there were some technical limitations in
measuring the amount of information entered by consumers in the case of the checkout experiment
and with measuring the cart value in case of both experiments. The inability to measure the amount of
data entered in the case of the single-page experiment variation limited the options of building a more
refined model next to the high level checkout-to-purchase conversion model, in order to analyse the
effect of the checkout variation on entering checkout information and completing a purchase. The cart
value was measured at the final page a web shop visitor visited, instead of when the consumers
entered the checkout. This implied that inaccurate data might exist where consumers entered the
checkout with a cart filled with products, exited the checkout, changed the cart contents and
subsequently left the web shop. It was however expected these cases were at a minimum as consumers
that exited checkout and cleared out their entire cart were excluded from the dataset based on their
cart value of zero. Furthermore the likelihood of this limitation having a major impact on the
experiments results was found relatively small and as such does not provide further implications for
generalisation and applicability of the conclusion.
Theoretical contributions and future research opportunities 7.2.
The contributions of this research are threefold. First of all the theoretical e-servicescape model
provided an overview of knowledge available in the academic research field on optimising conversion
using an e-servicescape perspective. The model provided can both be used to identify research fields
that yet require more theoretical investigation and as a starting point for quantifying the effects of
certain e-servicescape characteristics, for which this study was too limited.
The second contribution of this research lies in the subsequent step of including field data. The field
of e-commerce is changing rapidly, leading to past research results that are no longer fully accurate or
at worst even obsolete. Although there were limitations to the results of the case study, it did provide
insights into the current sentiment and knowledge available in the field regarding the implementation
of e-servicescape design rules. This leads towards the identification of future research opportunities
and directions both in confirming the results and performing additional explanatory research to further
identify the effects of the design rules.
The third theoretical contribution of this research was formed by the experiments. They identified the
potential strong impact of marketing campaigns on conversion rates at different web shop stages,
which requires further explanatory research in case of both experiments. Furthermore, although the
scale was of the experiment was too small to provide results, a first step was made in research
regarding single-page and multi-page checkouts which provides a first step towards further
investigation in the research field. Future research should focus on extending the experiment at
different web shops in different industries during different types of marketing campaigns over longer
periods of time, in order to provide more insights into factors and moderators playing a role in
checkout conversion rates. At the same time future research should focus on further identifying the
38
usage of cross-selling in the checkout. Additional dependent variables, types of cross-selling and
design choices are needed to gain more insights on the influence of both cross-selling and moderators
on conversion rates.
As stated, the most important future research opportunities lay in the identification of factors
influencing the success of cross-selling in the cart and the use of multi-page and single-page
checkouts based on the experiments. In general this research should focus on acquiring additional data
to ensure reliability and on quantifying effects and confirming effects identified in this study, and the
identification of moderating factors such as time of day, day of week and different marketing
campaigns. On the basis of the validated e-servicescape model future research opportunities were also
identified as being the identification of the role of using product on model photography over sole
product photography and the identification of the role of the category page (whether it should be
functional and oriented on navigation, or whether it should be focussed on inspiring visitors).
Additionally purchasing online by consumers via mobile communication devices such as smartphones
and tablets is becoming more mainstream, which also leads to a need of more research towards the
experiment results and on the implementation of design rules in a so-called ‘mobile environment’.
39
Managerial implications 7.3.
Next to theoretical contributions, the research also provided several managerial implications. First of
all the e-servicescape model combining theory and practice may be used by managers as a tool and
guideline on the tactical level when designing or optimising the e-servicescape of a web shop.
Although large web shop operators may find the model beneficial, web shops with limited resources
or relatively small amounts of visitors that are limited testing abilities could find the model to be a
starting point to optimise their web shop on the basis of theoretically and practical grounded
knowledge regarding web shop aspects. The final model is depicted in Table 7.1 at the next pages.
Additional managerial implications were created by the experiment results regarding the cart page
design. Even though intuition and the analogy to offline checkout bargains might lead to the inclusion
of cross-selling on the cart page, the functionality may prove detrimental to conversion rate. Careful
considerations regarding the implementation of the functionality should be taken into account, as well
as multiple dependent variables such as conversion rate and average cart value in order to establish
which version works best for a specific web shop.
Finally next to the specific implications from the experiments, the importance of both extensive
testing and deliberate experiment designs was shown. On the one hand, results might very well not be
as expected, but more important the experiment implications should be established meticulously. The
experiments part of this research are a clear example of the latter. Small amounts of data lead to
difficult analyses and interpretations of data and results which might, in case of wrong types of
analyses, result in spurious conclusions based on insufficient or inadequate data or might at least lead
to limited generalizability of conclusions even within a single web shop.
40
Table 7.1 Validated e-servicescape model
De
cis
ion
ma
kin
g
pro
ce
ss
sta
ge
Choice / purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
Inclu
de o
rig
inal desig
ns a
nd s
igns s
uch a
s lo
gos.
Anim
ate
lo
gos f
or
incre
ased e
ffectiveness a
nd im
pact, b
ut sparin
gly
to a
void
dis
tractio
n.
Whils
t adherin
g t
o s
tandard
and c
om
mon d
esig
n r
ule
s, m
ake s
ure
the d
esig
n fits t
he w
eb s
hop a
nd b
rand p
ropositio
n.
If p
ossib
le, im
ple
me
nt re
fere
nces to a
n o
fflin
e b
rand a
nd r
eta
iler.
Desig
n s
hould
be c
olo
urf
ul by a
good s
ele
ctio
n,
pla
cem
ent
and c
om
bin
atio
n o
f colo
urs
.
Desig
n s
hould
be d
ivers
e,
by v
isual richness,
dynam
ics, novelty a
nd c
reativity.
Desig
n s
hould
be s
imp
le b
y s
how
ing u
nity,
hom
ogeneity, cla
rity
, ord
erlin
ess a
nd b
ala
nce.
Desig
n s
hould
show
cra
ftsm
anship
by m
odern
ity a
nd in
tegra
tin
g s
implic
ity,
div
ers
ity a
nd c
olo
urf
uln
ess.
Pro
vid
e la
rge s
ize h
igh q
ualit
y p
roduct im
ages s
upport
ed b
y s
chem
atic p
roduct chara
cte
ristics.
Pro
vid
e lo
gos, cert
ific
ate
s a
nd o
ther
vis
ual cues e
arly o
n to e
nhance f
eelin
gs o
f tr
ust.
Do n
ot
dis
tract users
with a
esth
etic d
esig
ns d
urin
g c
heckout.
Th
e c
are
ful use o
f people
on p
ictu
res c
an p
rovid
e c
onte
xt
and tra
nsfe
r em
otio
n a
nd f
eelin
g.
Cre
ate
a d
esig
n t
hat flo
ws f
luently fro
m h
om
e p
age t
o c
heckout
with focus o
n s
upport
for
decis
ion a
nd t
ransactio
n p
rocesses
Use a
consis
tent to
ne o
f voic
e t
hat suits t
he t
arg
et audie
nce.
Be s
carc
e w
ith v
ivid
ente
rtain
me
nt as it decre
ases s
hoppin
g c
art
use.
Cre
ate
ente
rtain
me
nt
by p
rovid
ing thoughtf
ul use o
f colo
ur
and typogra
phy b
ased o
n f
unctio
nalit
y.
Cre
ate
ente
rtain
me
nt
by s
ocia
l aspects
, in
tera
ctive e
lem
ents
and in
spiratio
nal desig
n.
Pro
vid
e e
nte
rtain
me
nt
by r
egula
rly u
pdatin
g the w
eb s
hop s
o c
onsum
ers
get th
e f
eelin
g it
evolv
es.
Ori
gin
ality
of
de
sig
n
Vis
ual
ap
pe
al
En
tert
ain
men
t
valu
e
Aesthetic appeal
41
Table 7.1 Validated e-servicescape model (continued)
De
cis
ion
ma
kin
g
pro
ce
ss
sta
ge
Choice / purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
Build
mu
ltip
le w
ays o
f navig
atio
n b
ased o
n e
ase
-of-
use b
y d
iffe
rent ty
pes o
f consum
ers
and the a
ctio
ns it fa
cili
tate
s t
hat contin
uously
show
s
the b
readth
and d
epth
of th
e w
eb s
hop.
Consid
er
that th
e s
ize a
nd locatio
n o
f te
xt
and g
raphic
s d
ete
rmin
e u
sers
’ attentio
n b
ased o
n F
-shaped s
cannin
g p
att
ern
s.
Cre
ate
a c
lean a
nd u
nclu
ttere
d d
esig
n, w
ithout unnecessary
text
and g
raphic
s a
nd m
inim
um
lo
adin
g t
ime
s a
nd s
yste
m c
rashes, th
at
behaves a
s u
ser
expect.
Pro
vid
e c
lear
org
anis
atio
n a
nd layout
without
dis
tractio
ns.
Pro
vid
e a
lin
k b
ack t
o s
hoppin
g.
Cre
ate
a c
onsis
tent
and lo
gic
al user
flo
w fro
m h
om
e p
age to c
heckout.
Pro
vid
e c
onta
ct in
form
atio
n,
pre
fera
bly
inclu
din
g a
(fr
ee)
num
ber,
to r
each t
he c
onsum
er
support
depart
me
nt.
Sta
te c
om
petitive a
dvanta
ges r
ega
rdin
g t
he q
ualit
y o
f pro
duct offerin
gs a
nd s
erv
ices c
learly t
hro
ughout
the w
eb s
hop.
Sta
te in
form
atio
n r
egard
ing p
rice, fe
atu
res, in
vento
ry in
form
atio
n a
nd o
rder
rela
ted c
harg
es a
s e
arly o
n a
s p
ossib
le.
Pro
vid
e in
form
atio
n that is
accura
te,
consis
tent
and s
pecific
, support
ed b
y full
siz
e p
ictu
res.
Pro
vid
e in
form
atio
n that is
accura
te,
consis
tent
and s
pecific
.
Dis
pla
y o
ut-
of-
sto
ck s
izes,
but re
move p
erm
anent out-
of-
sto
ck p
roducts
and c
olo
urs
.
Pro
vid
e in
form
atio
n fro
m a
consum
er
poin
t of vie
w w
hils
t keepin
g t
hem
in
a c
ontin
uous f
low
.
Th
e lo
catio
n, ty
pe a
nd im
ple
me
nta
tio
n o
f cro
ss-s
elli
ng,
especia
lly in c
ase o
f lim
ited d
ata
and b
usin
ess r
ule
, should
be c
onsid
ere
d d
ue t
o
conflic
tin
g r
esults.
Specify c
usto
mis
atio
n t
ow
ard
s d
ecis
ion a
nd t
ransactio
n p
rocesses.
Add f
eatu
res s
upport
ing d
irect
inte
ractivity b
etw
een v
isitors
and s
ale
s o
r support
em
plo
yees.
Add in
tera
ctive functio
nalit
y t
hat
is p
ote
ntia
lly u
sefu
l or
influ
ences s
ite u
sage a
nd n
avig
atio
n.
Change t
ext and c
olo
urs
when h
overin
g o
ver
actio
nable
text
and im
age
s.
Usab
ilit
y
Rele
van
ce o
f
info
rmati
on
Cu
sto
mis
ati
on
Inte
racti
vit
y
Layout & functionality
42
Table 7.1 Validated e-servicescape model (continued)
De
cis
ion
ma
kin
g
pro
ce
ss
sta
ge
Choice / purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
Dis
pla
y t
ruste
d a
nd in
dependent seals
and c
ert
ific
ate
s o
f appro
val th
roughout
the w
eb s
hop.
Ask o
nly
str
ictly n
ecessary
in
form
atio
n a
nd e
xclu
de m
ark
etin
g q
uestions.
Explic
itly
sta
te w
hat in
form
atio
n is s
tore
d a
nd n
ot sto
red.
Dis
pla
y t
ruste
d a
nd in
dependent seals
and c
ert
ific
ate
s o
f appro
val.
Cre
ate
a c
onsis
tent
and lo
gic
al user
flo
w fro
m h
om
e p
age to c
heckout.
Allo
w for
checkout com
ple
tio
n w
ithout
regis
tratio
n o
r usin
g a
n a
ccount.
.Pro
vid
e a
ctio
nable
feedback a
nd e
rror
me
ssages a
nd o
nly
if str
ictly n
ecessary
Pro
vid
e in
form
atio
n r
egard
ing t
he d
iffe
rent checkout ste
ps a
s w
ell
as t
he c
urr
ent
locatio
n.
Pro
vid
e the o
ptio
n o
f cre
dit c
ard
paym
ents
, re
gula
r paym
ent ty
pes a
nd p
aym
ent ty
pes that fu
nctio
n a
s the e
xte
nsio
n o
f exis
tin
g m
eth
ods.
Ta
ke th
e pro
ducts
sold
and diffe
rent
targ
et
audie
nces in
to account
when desig
nin
g s
ingle
or
multi-
page checkouts
both
for
speed and
confirm
atio
n.
Perc
eiv
ed
secu
rity
Ease o
f u
se
Financial security
43
8. References
Agarwal, A., & Hedge, A. (2008). The Impact of Web Page Usability Guideline Implementation on
Aesthetics and Perceptions of the E-Retailer. Proceedings of the Human Factors and Ergonomics
Society Annual Meeting, 52(6), 528–532.
Bitner, M. (1992). Servicescapes: The Impact of Physical Surroundings on Customers and
Employees. Journal of Marketing, 56, 57–71.
Blanco, C. F., Sarasa, R. G., & Sanclemente, C. O. (2010). Effects of visual and textual information in
online product presentations: looking for the best combination in website design. European Journal of
Information Systems, 19(6), 668–686.
Bucklin, R., & Sismeiro, C. (2003). A Model of Web on Site Browsing Behavior Estimated
Clickstream Data. Journal of Marketing Research, 40(3), 249–267. Retrieved from
http://www.jstor.org/stable/10.2307/30038857
Butler, P., & Peppard, J. (1998). Consumer Purchasing on the Internet: Processes and Prospects.
European Management Journal, 16(5), 600–610.
Cai, S., & Xu, Y. (2011). Designing Not Just for Pleasure: Effects of Web Site Aesthetics on
Consumer Shopping Value. International Journal of Electronic Commerce, 15(4), 159–188.
Carmel, E., Crawford, S., & Chen, H. (1992). Browsing in hypertext: A cognitive study. IEEE
Transactions on Systems, Man and Cybernetics, 22(5), 865–884.
Chadwick, S. A. (2001). Communicating trust in e-commerce interactions. Management
Communication Quarterly, 14(4), 653–658.
Childers, T., Carr, C., & Peck, J. (2002). Hedonic and utilitarian motivations for online retail
shopping behavior. Journal of Retailing, 77(4), 511–535.
Close, A. G., & Kukar-Kinney, M. (2010). Beyond buying: Motivations behind consumers’ online
shopping cart use. Journal of Business Research, 63(9-10), 986–992.
Craven, J., Johnson, F., & Butters, G. (2010). The usability and functionality of an online catalogue.
Aslib Proceedings, 62(1), 70–84.
Docdata N.V. (2012). Docdata N.V. Jaarverslag 2011 (p. 121). Waalwijk.
Egger, F. N. (2001). Affective Design of E-Commerce User Interfaces : How to Maximise Perceived
Trustworthiness. In Helander, Khalid, & Tham (Eds.), Proceedings of The International Conference
on Affective Human Factors Design. London: Asean Academic Press.
Fang, X., & Salvendy, G. (2003). Customer-Centered Rules for Design of E-Commerce Web Sites
Xiaowen Fang and Gavriel Salvendy. Communications of the ACM, 46(12), 332–336.
Fink, D., & Laupase, R. (2000). Perceptions of web site design characteristics: a Malaysian/Australian
comparison. Internet Research, 10(1), 44–55.
Garbarino, E., & Strahilevitz, M. (2004). Gender differences in the perceived risk of buying online
and the effects of receiving a site recommendation. Journal of Business Research, 57(7), 768–775.
44
Gefen, D., Karahanna, E., & Straub, D. W. (2003). Trust and TAM in online shopping: An integrated
model. MIS quarterly, 27(1), 51–90.
Gill, M. (2012). European Online Retail Forecast: 2011 to 2016. Cambridge.
Harris, L. C., & Goode, M. M. H. (2010). Online servicescapes, trust, and purchase intentions.
Journal of Services Marketing, 24(3), 230–243.
Hausman, A. V., & Siekpe, J. S. (2009). The effect of web interface features on consumer online
purchase intentions. Journal of Business Research, 62(1), 5–13.
He, F., & Mykytyn, P. P. (2007). Decision factors for the adoption of an online payment system by
customers. International Journal of E-Business Research, 3(4), 1–33.
Hong, W., Thong, J. Y. L., & Tam, K. Y. (2004). The Effects of Information Format and Shopping
Task on Consumers ’ Online Shopping Behavior: A Cognitive Fit Perspective. Journal of
Management Information Systems, 21(3), 149–184.
Jang, E., & Burns, L. D. (2004). Components of apparel retailing Web sites. Journal of Fashion
Marketing and Management, 8(4), 375–388.
Jeon, M. M., & Jeong, M. (2009). A Conceptual Framework to Measure E- Servicescape on a B & B
Website. International CHRIE Conference-Refereed Track (pp. 1–8).
Kim, J., & Moon, J. Y. (1998). Designing towards emotional usability in customer interfaces—
trustworthiness of cyber-banking system interfaces. Interacting with Computers, 10(1), 1–29.
Kim, L. H., Qu, H., & Kim, D. J. (2009). A Study of Perceived Risk and Risk Reduction of
Purchasing Air Tickets Online. Journal of Travel & Tourism Marketing, 26(3), 203–224.
Kwon, K.-N., & Lee, J. (2003). Concerns About Payment Security of Internet Purchases: A
Perspective on Current On-Line Shoppers. Clothing and Textiles Research Journal, 21(4), 174–184.
Laroche, M., McDougall, G. H. G., Bergeron, J., & Yang, Z. (2004). Exploring How Intangibility
Affects Perceived Risk. Journal of Service Research, 6(4), 373–389.
Lavie, T., & Tractinsky, N. (2004). Assessing dimension of perceived visual aesthetics of web sites.
International Journal of Human-Computer Studies, 60, 269–298.
Lee, M., & Turban, E. (2001). A trust model for consumer internet shopping. International Journal of
Electronic Commerce, 6(1), 75–91.
Liang, T., & Lai, H. (2002). Effect of store design on consumer purchases: van empirical study of on-
line bookstores. Information and Management, 39, 431–444.
Ling, C., Salvendy, G., & Purdue University. (2006). Importance of Usability Considerations to
Purchase Intention on E-Commerce Website with Different User Groups. Proceedings of the Human
Factors and Ergonomics Society Annual Meeting, 50(14), 1457–1460.
Manganari, E. E., Siomkos, G. J., Rigopoulou, I. D., & Vrechopoulos, A. P. (2011). Virtual store
layout effects on consumer behaviour: Applying an environmental psychology approach in the online
travel industry. Internet Research, 21(3), 326–346. doi:10.1108/10662241111139336
45
Miles, G. E., Howes, A., & Davies, A. (2000). A framework for understanding human factors in web
based electronic commerce. International Journal of Human-Computer Studies, 52, 131–163.
Moe, W. W., & Fader, P. S. (2004). Dynamic Conversion Behavior at E-Commerce Sites.
Management Science, 50(3), 326–335.
Montgomery, D., & Runger, G. (2007). Applied Statistics and Probability for Engineers (4th ed.).
Hoboken: John Wiley & Sons, Inc.
Moshagen, M., & Thielsch, M. T. (2010). Facets of visual aesthetics. International Journal of
Human-Computer Studies, 68(10), 689–709.
Palmer, J. W. (2002). Web Site Usability, Design, and Performance Metrics. Information Systems
Research, 13(2), 151–167.
Salam, A. F., Iyer, L., Palvia, P., & Singh, R. (2005). Trust in E-commerce. Communications of the
ACM, 48(2), 72–77.
Siau, B. K., & Shen, Z. (2003). Building Customer Trust in Mobile Commerce. Communications of
the ACM, 46(4), 91–94.
Silverman, B. B. G., Bachann, M., & Al-Akharas, K. (2001). Implications of buyer decision theory
for design of e-commerce Web sites. International Journal of Human-Computer Studies, 55(5), 815–
844.
Sismeiro, C., & Bucklin, R. E. (2004). Modeling Purchase Behavior at an E-Commerce Web Site: A
Task Completion Approach. Journal of Marketing Research, 41(3), 306–323.
Song, J. H., & Zinkhan, G. M. (2008). Determinants of Perceived Web Site Interactivity. Journal of
Marketing, 72, 99–113.
Song, J., Jones, D., & Gudigantala, N. (2007). The effects of incorporating compensatory choice
strategies in Web-based consumer decision support systems. Decision Support Systems, 43(2), 359–
374.
Song, J., & Zahedi, F. (2005). A Theoretical Approach to Web Design in E-Commerce: A Belief
Reinforcement Model. Management Science, 51(8), 1219–1235.
Stewart, K. J. (2003). Trust Transfer on the World Wide Web. Organization Science, 14(1), 5–17.
Teltzrow, M., & Berendt, B. (2003). Web-Usage-Based Success Metrics for Multi-Channel
Businesses. Proceedings of the Fifth WEBKDD workshop: Webmining as a Premise to Effective and
Intelligent Web Applications (pp. 17–27). Washington DC.
Thirumalai, S., & Sinha, K. K. (2011). Customization of the online purchase process in electronic
retailing and customer satisfaction : An online field study. Journal of Operations Management, 29(5),
477–487.
Thuiswinkel.org. (2012). Thuiswinkel Markt Monitor 2011-2 (p. 74).
Tomes, E. G. (2000). Understanding and facilitating the browsing of electronic text. International
Journal of Human-Computer Studies, 52(3), 423–452.
46
Tucker, S.-P. (2008). E-commerce standard user interface: an E-menu system. Industrial Management
& Data Systems, 108(8), 1009–1028.
Van Aken, J., Berends, H., & Van der Bij, H. (2007). Problem Solving in Organizations. Cambridge:
Cambridge University Press.
Van der Heijden, H., Verhagen, T., & Creemers, M. (2003). Understanding online purchase
intentions: contributions from technology and trust perspectives. European Journal of Information
Systems, 12(1), 41–48.
Vilnai-yavetz, I., & Rafaeli, A. (2006). Aesthetics and Professionalism of Virtual Servicescapes.
Journal of Service Research, 8(3), 245–259.
Vrechopoulos, A. P. (2005). Designing Alternative Store Layouts for Internet Retailing. In G. J.
Doukidis & A. P. Vrechopoulos (Eds.), Consumer Driven Electronic Transformation (pp. 91–109).
Berlin, Heidelberg: Springer Berlin Heidelberg.
Wang, J. Y., Minor, M. S., & Wei, J. (2011). Aesthetics and the online shopping environment:
Understanding consumer responses. Journal of Retailing, 87(1), 46–58.
Wang, Y. D., & Emurian, H. H. (2005). Trust in E-Commerce: Consideration of Interface Design
Factors. Journal of Electronic Commerce in Organizations, 3(4), 42.
Wang, Y. J., Hernandez, M. D., & Minor, M. S. (2010). Web aesthetics effects on perceived online
service quality and satisfaction in an e-tail environment: The moderating role of purchase task.
Journal of Business Research, 63(9-10), 935–942.
Williams, R., & Dargel, M. (2004). From servicescape to “cyberscape.” Marketing Intelligence &
Planning, 22(3), 310–320.
Wu, G., Hu, X., & Wu, Y. (2010). Effects of Perceived Interactivity, Perceived Web Assurance and
Disposition to Trust on Initial Online Trust. Journal of Computer-Mediated Communication, 16(1), 1–
26.
Yan, R. (2009). Product categories, returns policy and pricing strategy for e-marketers. Journal of
Product & Brand Management, 18(6), 452–460.
Yin, R. (2009). Case Study Research: Design and Methods (4th ed.). Thousand Oaks: Sage Inc.
Zeithaml, V. A., Parasuraman, A., & Malhotra, A. (2002). Service quality delivery through web sites:
a critical review of extant knowledge. Journal of the Academy of Marketing Science, 30(4), 362–375.
Zhang, P., Von Dran, G., Blake, P., & Pipithsuksunt, V. (2001). A Comparison of the Most Important
Website Features in Different Domains: An Empirical Study of User Perceptions. e-Service Journal,
1(1), 77–91.
Appendices - 47
Appendix A. Literature review results
Table A.1 Literature based e-servicescape model
Lit
era
ture
So
urc
e
(Fin
k &
La
up
ase
, 2
00
0;
Harr
is &
Go
od
e,
201
0)
(Fa
ng
& S
alv
end
y,
20
03
; F
ink &
La
up
ase
, 2
00
0)
(Mo
sh
age
n &
Th
iels
ch
, 2
01
0)
(Fa
ng
& S
alv
end
y,
20
03
;
Mo
sh
ag
en &
Th
iels
ch
, 2
01
0)
(Child
ers
, C
arr
, &
Pe
ck,
200
2;
La
vie
& T
ractin
sky,
20
04;
Mo
sh
ag
en &
Th
iels
ch
, 2
01
0)
(Mo
sh
age
n &
Th
iels
ch
, 2
01
0)
(Ag
arw
al &
He
dg
e, 2
00
8;
Bla
nco
,
Sa
rasa
, &
Sa
ncle
me
nte
, 2
010
;
Fa
ng
& S
alv
en
dy, 2
00
3;
Y.
D.
Wa
ng
& E
mu
rian
, 20
05
)
(Hau
sm
an
& S
iekp
e, 2
00
9; L
. H
.
Kim
et
al.,
20
09
; Y
. D
. W
an
g &
Em
uri
an
, 2
00
5)
(Cai &
Xu
, 2
01
1;
Clo
se
& K
ukar-
Kin
ne
y,
20
10
; Y
. J.
Wan
g,
Hern
an
de
z, &
Min
or,
20
10)
(Clo
se
& K
uka
r-K
inn
ey, 2
01
0)
(Harr
is &
Go
od
e,
201
0)
(Cai &
Xu
, 2
01
1;
Y.
J.
Wa
ng
et
al.,
20
10)
(Eg
ge
r, 2
00
1; F
an
g &
Sa
lven
dy,
20
03
; P
alm
er,
20
02
; S
ilve
rma
n
et
al.,
20
01
; T
ucke
r, 2
00
8;
Zh
an
g,
Vo
n D
ran
, B
lake
, &
Pip
ith
su
ksu
nt,
20
01
)
(Ag
arw
al &
He
dg
e, 2
00
8; F
an
g &
Sa
lven
dy,
20
03)
(Cra
ve
n,
Jo
hn
so
n, &
Bu
tte
rs,
20
10
; F
an
g &
Sa
lven
dy, 2
00
3;
Lin
g,
Sa
lven
dy,
& P
urd
ue
Univ
ers
ity,
20
06;
Pa
lme
r, 2
00
2;
Silv
erm
an
et
al.,
20
01)
(J.
Kim
& M
oo
n, 1
99
8)
(Fa
ng
& S
alv
end
y,
20
03
)
(Fa
ng
& S
alv
end
y,
20
03
)
(Ja
ng
& B
urn
s, 2
00
4;
Silv
erm
an
et
al.,
20
01
; Y
an
, 2
00
9)
(Clo
se
& K
uka
r-K
inn
ey, 2
01
0;
Eg
ge
r, 2
001
)
(Fa
ng
& S
alv
end
y,
20
03
;
La
roch
e, M
cD
ou
ga
ll, B
erg
ero
n,
&
Ya
ng
, 2
00
4; Z
ha
ng
et
al., 2
00
1)
(Eg
ge
r, 2
00
1)
(Fa
ng
& S
alv
end
y,
20
03
)
De
cis
ion
ma
kin
g
pro
ce
ss
sta
ge 3*
2*
1*
Des
ign
ru
le
Inclu
de o
rig
inal desig
ns a
nd s
igns s
uch a
s lo
gos
Anim
ate
lo
gos f
or
incre
ased e
ffectiveness a
nd im
pact, b
ut sparin
gly
to a
void
dis
tractio
n.
Desig
n s
hould
be c
olo
urf
ul by a
good s
ele
ctio
n,
pla
cem
ent
and c
om
bin
atio
n o
f colo
urs
.
Desig
n s
hould
be d
ivers
e,
by v
isual richness,
dynam
ics, novelty a
nd c
reativity.
Desig
n s
hould
be s
imp
le b
y s
how
ing u
nity,
hom
ogeneity, cla
rity
, ord
erlin
ess a
nd b
ala
nce.
Desig
n
should
show
cra
ftsm
anship
by
mo
dern
ity
and
inte
gra
ting
sim
plic
ity,
div
ers
ity
and
colo
urf
uln
ess.
Pro
vid
e la
rge s
ize h
igh q
ualit
y p
roduct im
ages s
upport
ed b
y s
chem
atic p
roduct chara
cte
ristics.
Pro
vid
e lo
gos, cert
ific
ate
s a
nd o
ther
vis
ual cues e
arly o
n to e
nhance f
eelin
gs o
f tr
ust.
Do n
ot
dis
tract users
with a
esth
etic d
esig
ns d
urin
g c
heckout.
Be s
carc
e w
ith v
ivid
ente
rtain
me
nt as it decre
ases s
hoppin
g c
art
use.
Cre
ate
ente
rtain
me
nt
by p
rovid
ing thoughtf
ul use o
f colo
ur
and typogra
phy b
ased o
n f
unctio
nalit
y.
Cre
ate
ente
rtain
me
nt
by s
ocia
l aspects
, in
tera
ctive e
lem
ents
and in
spiratio
nal desig
n.
Build
mu
ltip
le w
ays o
f navig
atio
n b
ased o
n e
ase
-of-
use b
y d
iffe
rent
types o
f consum
ers
and t
he
actio
ns it fa
cili
tate
s t
hat contin
uously
show
s the b
readth
and d
epth
of
the w
eb s
hop.
Consid
er
that
the s
ize a
nd l
ocation o
f te
xt
and g
raphic
s d
ete
rmin
e u
sers
’ att
entio
n b
ased o
n F
-shaped s
cannin
g p
att
ern
s.
Cre
ate
a cle
an and unclu
ttere
d desig
n,
without
unnecessary
te
xt
and gra
phic
s and m
inim
um
loadin
g t
ime
s a
nd s
yste
m c
rashes, th
at behaves a
s u
ser
expect.
Pro
vid
e c
lear
org
anis
atio
n a
nd layout
without
dis
tractio
ns.
Pro
vid
e a
lin
k b
ack t
o s
hoppin
g.
Pro
vid
e c
onta
ct in
form
atio
n in
clu
din
g a
(fr
ee)
num
ber
to r
each t
he c
onsum
er
support
depart
me
nt.
Sta
te com
petitive advanta
ges re
gard
ing th
e qualit
y of
pro
duct
offerin
gs and serv
ices cle
arly
thro
ughout
the w
eb s
hop.
Sta
te i
nfo
rma
tio
n r
egard
ing p
rice,
featu
res,
invento
ry i
nfo
rmatio
n a
nd o
rder
rela
ted c
harg
es a
s
early o
n a
s p
ossib
le.
Pro
vid
e in
form
atio
n that is
accura
te, consis
tent
and s
pecific
, support
ed b
y full
siz
e p
ictu
res.
Pro
vid
e in
form
atio
n that is
accura
te, consis
tent
and s
pecific
.
Dis
pla
y o
ut-
of-
sto
ck p
roducts
and s
izes.
Ori
gin
ality
of
de
sig
n
Vis
ual
ap
pe
al
En
tert
ain
men
t
valu
e
Usab
ilit
y
Rele
van
ce o
f
info
rmati
on
Aesthetic appeal Layout &
functionality *
1 =
In
form
atio
n s
ea
rch
, 2
= E
va
lua
tio
n o
f a
lte
rna
tive
s, 3
= C
ho
ice
/ P
urc
ha
se
Appendices - 48
Table A.1 Literature based e-servicescape model (continued)
Lit
era
ture
So
urc
e
(Th
irum
ala
i &
Sin
ha
, 2
01
1;
Ze
ith
am
l, P
ara
su
ram
an
, &
Ma
lho
tra,
20
02)
(Th
irum
ala
i &
Sin
ha
, 2
01
1)
(Viln
ai-
ya
ve
tz &
Ra
fae
li, 2
00
6)
(J.
H.
So
ng
& Z
inkha
n,
200
8;
Wu
, H
u,
& W
u,
20
10)
(Wu
et
al.,
20
10
)
(L.
H.
Kim
et
al.,
20
09;
Y.
D.
Wa
ng
& E
mu
rian
, 20
05
) (F
an
g &
Sa
lvend
y,
20
03
;
Ga
rba
rino
& S
tra
hile
vitz,
20
04
)
(Kw
on
& L
ee
, 2
003
)
(L.
H.
Kim
et
al.,
20
09;
Y.
D.
Wa
ng
& E
mu
rian
, 20
05
)
(Eg
ge
r, 2
00
1; F
an
g &
Sa
lven
dy,
20
03
)
(Eg
ge
r, 2
00
1;
Laro
ch
e e
t a
l.,
20
04
)
(Eg
ge
r, 2
00
1;
Laro
ch
e e
t a
l.,
20
04
) (G
efe
n,
Kara
ha
nna
, &
Str
au
b,
20
03
; H
e &
Mykyty
n,
20
07
;
Lia
ng
& L
ai, 2
00
2)
De
cis
ion
ma
kin
g
pro
ce
ss
sta
ge
3
2
1
Des
ign
ru
le
Consid
er
where
to p
lace c
usto
mis
atio
n a
s t
here
are
conflic
tin
g r
esults.
Specify c
usto
mis
atio
n t
ow
ard
s d
ecis
ion a
nd t
ransactio
n p
rocesses.
Add f
eatu
res s
upport
ing d
irect
inte
ractivity b
etw
een v
isitors
and s
ale
s o
r support
em
plo
yees.
Add in
tera
ctive functio
nalit
y t
hat
is p
ote
ntia
lly u
sefu
l or
influ
ences s
ite u
sage a
nd n
avig
atio
n.
Change t
ext and c
olo
urs
when h
overin
g o
ver
actio
nable
text
and im
ages.
Dis
pla
y t
ruste
d a
nd in
de
pendent seals
and c
ert
ific
ate
s o
f appro
val th
roughout
the w
eb s
hop.
Ask o
nly
str
ictly n
ecessary
in
form
atio
n a
nd e
xclu
de m
ark
etin
g q
uestions.
Explic
itly
sta
te w
hat in
form
atio
n is s
tore
d a
nd n
ot sto
red.
Dis
pla
y t
ruste
d a
nd in
dependent seals
and c
ert
ific
ate
s o
f appro
val.
Allo
w for
checkout com
ple
tio
n w
ithout
regis
tratio
n o
r usin
g a
n a
ccount.
Pro
vid
e a
ctio
nable
feedback a
nd e
rror
me
ssages a
nd o
nly
if str
ictly n
ecessary
.
Pro
vid
e in
form
atio
n r
egard
ing t
he d
iffe
rent checkout ste
ps a
s w
ell
as t
he c
urr
ent
locatio
n.
Pro
vid
e th
e optio
n of
cre
dit card
paym
ents
, re
gula
r paym
ent
types and paym
ent
types th
at
functio
n a
s t
he e
xte
nsio
n o
f exis
ting m
eth
ods.
Cu
sto
mis
ati
on
Inte
racti
vit
y
Perc
eiv
ed
secu
rity
Ease o
f u
se
Layout & functionality
Financial security
*
1 =
In
form
atio
n s
ea
rch
, 2
= E
va
lua
tio
n o
f a
lte
rna
tive
s, 3
= C
ho
ice
/ P
urc
ha
se
*
1 =
In
form
atio
n s
ea
rch
, 2
= E
va
lua
tio
n o
f a
lte
rna
tive
s, 3
= C
ho
ice
/ P
urc
ha
se
Appendices - 49
Appendix B. Company structure
Figure B.1 Company structure (Docdata N.V., 2012)
Appendices - 50
Appendices - 51
Appendix C. Case study interview questions
The interview questions were held in Dutch. An English translation is printed in italic.
C.1 Welkom / Welcome
o Het interview is vertrouwelijk, dus er kan vrij (zowel lovend als kritisch) gesproken worden.
The interview is confidential, so you can speak freely in both positive and negative
statements.
o Het interview wordt opgenomen (audio) om later uit te kunnen werken en in geval van twijfel
over de uitwerking terug te kunnen luisteren.
The interview will be recorded (audio) in order to write a transcript and listen to the
interview again in case of the need for clarification.
o De opnames worden na afloop van het onderzoek vernietigd.
The recordings will be destroyed at the end of the thesis research.
o De uitwerking / inzichten worden ter goedkeuring voorgelegd aan de geïnterviewde.
The findings taken from the interview will be shown to the interviewee with a request for
approval.
o Heb je vooraf vragen of opmerkingen?
Are there any questions before we begin?
o Ben je akkoord met bovenstaande punten?
Do you acknowledge the points mentioned above?
C.2 Inleiding / Introduction
o De achtergrond van het onderzoek wordt uitegelegd: Afstudeeronderzoek aan de TU
Eindhoven naar conversie optimalisatie.
The background of the research is discussed: TU Eindhoven thesis research on conversion
optimisation.
o Doel van het interview is het verzamelen van best practices en validatie van het model.
Aim of the interview is to gather best practices and validate the e-servicescape model used.
o Voorbeelden tijdens het interview graag gericht op grote cases.
The examples in the interview should preferably be focused on large cases.
C.3 Functie / Function
o Wat is je achtergrond (opleiding, werkervaring)?
What is your background (education, experience)?
o Wat is je rol binnen Docdata Commerce?
What is your position at Docdata Commerce?
o Wat zijn je verantwoordelijkheden?
What are your responsibilities?
C.4 Optimalisatieproces / Optimisation process
o Hoe wordt op dit moment conversie geoptimaliseerd?
What are the current practices of conversion optimisation?
Appendices - 52
o Idee
Idea generation
o Uitwerking
Working out the idea in detail
o Voorstel
Proposing a web shop change
o Processtappen
Which steps are followed from proposal to execution?
o Implementatie
Implementing the idea
o Testen
Testing
o Feedback
Feedback
o Hoe wordt succes / falen bepaald?
How are success and failure identified?
o Wat vindt je van dit proces?
What is your view on this process?
o Welke sterke punten en verbeterpunten zijn er?
What are the strong and weak points of the process?
C.5 Aankoopproces / Consumer decision making process
o Wat zijn de verschillende fases die een consument doorloopt bij het maken van een keuze?
What are the different customer decision making process phases?
o Welke concrete acties op een website horen hierbij?
Which visitor actions on a web shop coincide with these phases?
C.6 Optimalisatiefactoren algemeen / Optimisation factors in general
o Gebruik hierbij eventueel het aankoopproces als richtlijn, maar gebruik het e-servicescape
model niet.
Use the customer decision making process as a guideline, but don’t use the e-servicescape
model.
o Welke web shop factoren spelen volgens jou een rol bij het optimaliseren van conversie?
Which web shop factors play a role when optimising conversion?
o Op welk psychologisch proces heeft deze factor invloed?
Which psychological process is influenced by this factor?
o Wat is de grootte van het effect van deze factor op de conversie?
What is the effect size of the factor on conversion?
o Kun je een voorbeeld geven?
Can you give an example?
C.7 E-Servicescape / E-servicescape
o Uitleggen wat de e-servicescape inhoudt.
Explain the e-servicescape concept.
o Gebruik het algemene model (zonder subfactoren)
Appendices - 53
Use the abstract model (without sub-factors)
o Kun je vertellen wat je onder ieder van deze factoren vindt vallen in het algemeen?
Can you explain what the factors encompass in general?
o Ontbreken er nog factoren?
Are there factors missing?
o Focus op factor 1, laat de subfactoren niet zien.
Focus on factor 1, but don’t show sub-factors.
o Welke sub-factoren vallen hier onder?
Which sub-factors are part of this factor?
o Laat subfactoren zien.
Show the sub-factors.
o Kun je per subfactor vertellen wat je daar onder verstaat?
Please explain for every sub-factor what it encompasses.
o Kun je een voorbeeld noemen van hoe je die subfactor in de praktijk gebruikt (hebt)?
Can you name an example of how you use(d) that sub-factor in practice?
o Welke theorie zit hier achter?
What theory is behind you explanation?
o Wat is het effect van deze subfactor?
What is the effect of the sub-factor?
o Mis je nog factoren?
Do you miss any sub-factors?
o Herhaal het voorgaande proces voor factor 2.
Repeat the above process for factor 2.
o Herhaal het voorgaande proces voor factor 3.
Repeat the above process for factor 3.
o Zijn er nog factoren die je mist overall?
Do you miss any factors or sub-factors?
C.8 Afsluiting / Final remarks
o Zijn er nog andere zaken die van belang zijn bij conversieoptimalisatie?
Are there any other (non-e-servicescape) factors important when optimising for conversion?
o Zijn er nog andere dingen die je aan wil dragen?
Is there anything else you would like to add?
o Heb je nog vragen of opmerkingen over het algemeen?
Do you have any question or remarks in general?
o Wil je een terugkoppeling van de algemene resultaten?
Would you like to receive feedback on the generic research results?
Appendices - 54
C.9 Figures
Figure C.1 E-servicescape, trust and purchase intention
Figure C.2 E-servicescape factors
Figure C.3 E-servicescape factor ‘Aesthetic appeal’
Appendices - 55
Figure C.4 E-servicescape factor ‘Layout & functionality’
Figure C.6 Full E-servicescape model
Figure C.5 E-servicescape factor ‘Financial security’
Appendices - 56
Appendices - 57
Appendix D. Case study results
D.1 Individual results Table D.1 Individual case study results
Appendices - 58
Table D.1 Individual case study results (continued)
Appendices - 59
Table D.1 Individual case study results (continued)
Appendices - 60
Table D.1 Individual case study results (continued)
Appendices - 61
D.2 Aggregated results Table D.2 Validated e-servicescape model
D
es
ign
ru
le
Inclu
de o
rig
inal desig
ns a
nd s
igns s
uch a
s lo
gos.
Anim
ate
lo
gos f
or
incre
ased e
ffectiveness a
nd im
pact, b
ut sparin
gly
to a
void
dis
tractio
n.
Whils
t adherin
g t
o s
tandard
and c
om
mon d
esig
n r
ule
s, m
ake s
ure
the d
esig
n fits t
he w
eb
shop a
nd b
rand p
ropositio
n.
If p
ossib
le, im
ple
me
nt re
fere
nces to a
n o
fflin
e b
rand a
nd r
eta
iler.
Desig
n s
hou
ld b
e c
olo
urf
ul by a
good s
ele
ctio
n,
pla
cem
ent
and c
om
bin
atio
n o
f colo
urs
.
Desig
n s
hould
be d
ivers
e,
by v
isual richness,
dynam
ics, novelty a
nd c
reativity.
Desig
n s
hould
be s
imp
le b
y s
how
ing u
nity,
hom
ogeneity, cla
rity
, ord
erlin
ess a
nd b
ala
nce.
Desig
n s
hould
show
cra
ftsm
anship
by m
odern
ity a
nd i
nte
gra
tin
g s
implic
ity,
div
ers
ity a
nd
colo
urf
uln
ess.
Pro
vid
e
larg
e
siz
e
hig
h
qualit
y
pro
duct
ima
ges
support
ed
by
schem
atic
pro
duct
chara
cte
ristics.
Pro
vid
e lo
gos, cert
ific
ate
s a
nd o
ther
vis
ual cues e
arly o
n to e
nhance f
eelin
gs o
f tr
ust.
Do n
ot
dis
tract users
with a
esth
etic d
esig
ns d
urin
g c
heckout.
Th
e c
are
ful use o
f people
on p
ictu
res c
an p
rovid
e c
onte
xt
and tra
nsfe
r em
otio
n a
nd f
eelin
g.
Cre
ate
a d
esig
n t
hat flo
ws f
luently fro
m h
om
e p
age t
o c
heckout
with focus o
n s
upport
for
decis
ion a
nd tra
nsactio
n p
rocesses
Use a
consis
tent to
ne o
f voic
e t
hat suits t
he t
arg
et audie
nce.
Be s
carc
e w
ith v
ivid
ente
rtain
me
nt as it decre
ases s
hoppin
g c
art
use.
Cre
ate
ente
rtain
me
nt
by
pro
vid
ing
thoughtf
ul
use
of
colo
ur
and
typogra
phy
ba
sed
on
functio
nalit
y.
Cre
ate
ente
rtain
me
nt
by s
ocia
l aspects
, in
tera
ctive e
lem
ents
and in
spiratio
nal desig
n.
Pro
vid
e e
nte
rtain
me
nt
by r
egula
rly u
pdatin
g the w
eb s
hop s
o c
onsum
ers
get th
e f
eelin
g it
evolv
es.
Cas
e s
tud
y D
ec
isio
n
Support
Support
Add
Add
Support
Support
Support
Support
Support
Support
Support
Add
Add
Add
Support
Support
Support
Add
Ex
pe
rtis
e
Desig
n
Ma
rke
tin
g
Ma
rke
tin
g
Desig
n
Ma
rke
tin
g
Desig
n
Ma
rke
tin
g
Pa
ym
en
ts
Desig
n
Pa
ym
en
ts
Desig
n
Ma
rke
tin
g
Desig
n
Rati
o
7 /
7
3 /
3
7
2
6 /
6
6 /
6
6 /
6
5 /
5
8 /
8
5 /
5
8 /
8
3
2
3
5 /
5
6 /
6
7 /
7
1
Dec
isio
n m
akin
g
pro
ce
ss
sta
ge
Choice / Purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
1.1
.1
1.1
.2
New
New
1.2
.1
1.2
.2
1.2
.3
1.2
.4
1.2
.5
1.2
.6
1.2
.7
New
New
New
1.3
.1
1.3
.2
1.3
.3
New
Ori
gin
ality
of
de
sig
n
Vis
ual
ap
pe
al
En
tert
ain
men
t valu
e
Aesthetic appeal
Appendices - 62
Table D.2 Validated e-servicescape model (continued)
De
sig
n r
ule
Build
mu
ltip
le w
ays o
f navig
atio
n b
ased o
n e
ase
-of-
use b
y d
iffe
rent ty
pes o
f consum
ers
and
the a
ctio
ns it fa
cili
tate
s t
hat continuously
show
s the b
readth
and d
epth
of
the w
eb s
hop.
Consid
er
that th
e s
ize a
nd locatio
n o
f te
xt
and g
raphic
s d
ete
rmin
e u
sers
’ attentio
n b
ased o
n
F-s
haped s
cannin
g p
att
ern
s.
Cre
ate
a c
lean a
nd u
nclu
ttere
d d
esig
n, w
ithout unnecessary
text
and g
raphic
s a
nd
min
imum
lo
adin
g t
ime
s a
nd s
yste
m c
rashes, th
at behaves a
s u
ser
expect.
Pro
vid
e c
lear
org
anis
atio
n a
nd layout
without
dis
tractio
ns.
Pro
vid
e a
lin
k b
ack t
o s
hoppin
g.
Cre
ate
a c
onsis
tent
and lo
gic
al user
flo
w fro
m h
om
e p
age to c
heckout.
Pro
vid
e c
onta
ct in
form
atio
n,
pre
fera
bly
inclu
din
g a
(fr
ee)
num
ber,
to r
each t
he c
onsum
er
support
depart
me
nt.
Sta
te c
om
petitive a
dvanta
ges r
egard
ing t
he q
ualit
y o
f pro
duct offerin
gs a
nd s
erv
ices c
learly
thro
ughout
the w
eb s
hop.
Sta
te in
form
atio
n r
egard
ing p
rice, fe
atu
res, in
vento
ry in
form
atio
n a
nd o
rder
rela
ted c
ha
rges
as e
arly o
n a
s p
ossib
le.
Pro
vid
e in
form
atio
n that is
accura
te,
consis
tent
and s
pecific
, support
ed b
y full
siz
e p
ictu
res.
Pro
vid
e in
form
atio
n that is
accura
te,
consis
tent
and s
pecific
.
Dis
pla
y o
ut-
of-
sto
ck p
roducts
and s
izes,
but re
move p
erm
anent out-
of-
sto
ck p
roducts
and
colo
urs
.
Pro
vid
e in
form
atio
n fro
m a
consum
er
poin
t of vie
w w
hils
t keepin
g them
in
a c
ontin
uous
flo
w.
Consid
er
where
to p
lace c
usto
mis
atio
n a
s t
here
are
conflic
tin
g r
esults.
Th
e lo
catio
n, ty
pe a
nd im
ple
me
nta
tio
n o
f cro
ss-s
elli
ng,
especia
lly in c
ase o
f lim
ited d
ata
and b
usin
ess r
ule
, should
be c
onsid
ere
d d
ue t
o c
onflic
tin
g r
esults.
Specify c
usto
mis
atio
n t
ow
ard
s d
ecis
ion a
nd t
ransactio
n p
rocesses.
Add f
eatu
res s
upport
ing d
irect
inte
ractivity b
etw
een v
isitors
and s
ale
s o
r support
em
plo
yees.
Add in
tera
ctive functio
nalit
y t
hat
is p
ote
ntia
lly u
sefu
l or
influ
ences s
ite u
sage a
nd
navig
atio
n.
Change t
ext and c
olo
urs
when h
overin
g o
ver
actio
nable
text
and im
ages.
Cas
e s
tud
y D
ec
isio
n
Support
Support
Support
Support
Support
Add
Revis
e
Support
Support
Support
Support
Revis
e
Add
Revis
e
Support
Support
Support
Support
Ex
pe
rtis
e
Desig
n
Ma
rke
tin
g
Ma
rke
tin
g
Su
pp
ort
Desig
n
Ma
rke
tin
g
Ma
rke
tin
g
Mg
t. d
ire
cto
r
Com
me
nts
all
inte
rvie
we
es
Desig
n
Ma
rke
tin
g
Desig
n
Ma
rke
tin
g
Rati
o
6 /
6
8 /
8
6 /
7
8 /
8
4 /
4
3
7 /
9
8 /
9
9 /
10
7 /
8
7 /
8
2 /
3
2
8 /
8
8 /
8
1 /
3
4 /
6
4 /
4
Dec
isio
n m
akin
g
pro
ce
ss
sta
ge
Choice / Purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
2.1
.1
2.1
.2
2.2
.3
2.1
.4
2.1
.5
New
2.2
.1
2.2
.2
2.2
.3
2.2
.4
2.1
.5
2.1
.6
New
2.3
.1
2.3
.2
2.4
.1
2.4
.2
2.4
.3
Usab
ilit
y
Rele
van
ce o
f
info
rmati
on
Cu
sto
mis
ati
on
Inte
racti
vit
y
Layout & functionality
Appendices - 63
Table D.2 Validated e-servicescape model (continued)
De
sig
n r
ule
Dis
pla
y tr
uste
d and in
dependent
seals
and cert
ific
ate
s of
appro
val
thro
ughout
the w
eb
shop.
Ask o
nly
str
ictly n
ecessary
in
form
atio
n a
nd e
xclu
de m
ark
etin
g q
uestions.
Explic
itly
sta
te w
hat in
form
atio
n is s
tore
d a
nd n
ot sto
red.
Dis
pla
y t
ruste
d a
nd in
dependent seals
and c
ert
ific
ate
s o
f appro
val.
Cre
ate
a c
onsis
tent
and lo
gic
al user
flo
w fro
m h
om
e p
age to c
heckout.
Allo
w for
checkout com
ple
tio
n w
ithout
regis
tratio
n o
r usin
g a
n a
ccount.
Pro
vid
e a
ctio
nable
feedback a
nd e
rror
me
ssages a
nd o
nly
if str
ictly n
ecessary
.
Pro
vid
e in
form
atio
n r
egard
ing t
he d
iffe
rent checkout ste
ps a
s w
ell
as t
he c
urr
ent
locatio
n.
Pro
vid
e t
he o
ptio
n o
f cre
dit c
ard
paym
ents
, re
gula
r paym
ent
types a
nd p
aym
ent
types t
hat
functio
n a
s t
he e
xte
nsio
n o
f exis
ting m
eth
ods.
Ta
ke t
he p
roducts
sold
and d
iffe
rent
targ
et
audie
nces i
nto
account
when d
esig
nin
g s
ingle
or
mu
lti-page c
heckouts
both
for
speed a
nd c
onfirm
atio
n.
Cas
e s
tud
y D
ec
isio
n
Support
Support
Support
Support
Add
No s
upp
ort
,
kee
p b
ase
d
on litera
ture
Support
Support
Revis
e
Add
Ex
pe
rtis
e
Desig
n
Ma
rke
tin
g
Ma
rke
tin
g
Pa
ym
en
ts
Su
pp
ort
Rati
o
9 /
11
1 /
2
2 /
2
8 /
10
5
1 /
1
3 /
3
9 /
9
8 /
12
7
Dec
isio
n m
akin
g
pro
ce
ss
sta
ge
Choice / Purchase
Evaluation of alternatives
Information search
Des
ign
ru
le
3.1
.2
3.2
.3
3.1
.4
New
2.2
.1
2.2
.2
2.2
.3
2.2
.4
New
Perc
eiv
ed
secu
rity
Ease o
f u
se
Financial security
Appendices - 64
Appendices - 65
Appendix E. Clickstream variables recorded
Table E.1 Clickstream variables recorded
Re
ma
rks
No
t d
isp
layed
at
“Bra
nds”
and
“Tre
nd
” m
enu
ite
ms
On
ly a
va
ilable
at p
rod
uct
pa
ges a
nd
pro
du
ct-
pop
ups
Ava
ilab
le w
he
n t
he c
art
has a
min
imum
of 1
pro
duct
Ava
ilab
le o
n t
he
ho
me
pa
ge
an
d L
ing
eri
e-c
ate
go
ry p
age
s
an
d -
pro
duct
pa
ges
Pa
ge
relo
ad a
fte
r clic
k
Pa
ge
relo
ad a
fte
r clic
k
On
ly m
easu
red w
he
n v
isito
r is
log
ged
in
Pro
du
ct
de
tails
dis
pla
ye
d in
po
pu
p
Me
asu
red
at
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
Eve
ry p
age
(action
)
Eve
ry p
age
(action
)
Eve
ry p
age
(action
)
Eve
ry p
age
(action
)
Pro
du
ct
pag
es
Pro
du
ct
pag
es
Pro
du
ct
pag
es
Pro
du
ct
pag
es
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Pro
du
ct
pag
es (
actio
n)
Me
asu
rin
g p
oin
t
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
On
clic
k
On
clic
k
1st f
ly o
ut o
r up
da
te o
f th
e fly
ou
t
Mo
use
ho
ve
red
on
min
i-b
asket
fly o
ut
for
at le
ast
2 s
eco
nd
s
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
En
larg
em
en
t vis
ible
fo
r a
t le
ast
2 s
eco
nd
s
On
th
um
bn
ail
clic
k
On
clic
k
On
clic
k
On
clic
k
On
clic
k
On
clic
k
On
clic
k
On
clic
k r
ela
ted
pro
duct
Va
ria
ble
Da
te /
tim
e
UR
L:
bre
adcru
mb
UR
L:
“ori
gin
al”
Pa
ge
title
Re
ferr
er
Tim
e s
pe
nt
on
pre
vio
us p
age
Co
okie
ID
(u
niq
ue
)
Bro
wse
r ch
ara
cte
ristics
Clic
k o
n m
enu
fly
ou
t ba
nn
er
“Ad
d t
o c
art
” b
utto
n
Exp
and
min
i-b
aske
t
Exp
and
“W
ha
t’s m
y s
ize
”- t
ab
Pro
du
ct
SK
U
Pro
du
ct
nam
e
Pro
du
ct
cate
go
ry
Pro
du
ct
price
Exp
and
pro
du
ct
ima
ge
Se
lect
oth
er
pro
duct
ima
ge
Se
lect
diffe
ren
t siz
e
Se
lect
diffe
ren
t co
lou
r
Ad
d t
o w
ishlis
t
Dis
pla
y “
ad
vic
e”-
tab
Dis
pla
y “
de
live
ry”-
tab
Dis
pla
y “
un
sa
tisfied
”-ta
b
Exp
and
re
late
d p
rodu
ct
Appendices - 66
Table E.1 Clickstream variables recorded (continued)
Re
ma
rks
On
eve
ry A
JA
X-r
eq
uest
(a-
synch
ron
ous p
rod
uct
loa
din
g)
Ne
xt
UR
L u
se
d t
o d
ete
rmin
es
wh
ich
ba
nn
er
was c
licked
An
sw
er
is e
xp
an
ded
afte
r clic
k
Co
nta
ct
form
is e
xp
and
ed
aft
er
clic
k
No
ch
eck if co
nta
ct fo
rm w
as
actu
ally
se
nd
In m
ini-
ba
ske
t via
AJA
X
req
uest
Pro
du
ct
de
tails
are
dis
pla
ye
d
in p
op
up
Pro
du
ct
de
tails
are
dis
pla
ye
d
in p
op
up
Po
pu
p w
ith
sh
ipp
ing
co
sts
is
dis
pla
ye
d
Eve
nt a
fte
r vou
che
r va
lidity
ch
eck
Am
ou
nt
inclu
din
g d
isco
un
ts
an
d p
rom
otio
ns
Me
asu
red
at
Pro
du
ct
ove
rvie
w a
nd
se
arc
h r
esu
lts
Hom
e p
ag
e
FA
Q p
ag
es
FA
Q p
ag
es
FA
Q p
ag
es
Cart
pa
ge
Cart
pa
ge
Cart
pa
ge
Cart
pa
ge
Cart
pa
ge
Cart
pa
ge
(action
)
Cart
pa
ge
(action
)
Cart
pa
ge
(action
Cart
pa
ge
(action
)
Cart
pa
ge
(action
)
Che
cko
ut
(actio
n)
Che
cko
ut
(actio
n)
Che
cko
ut
(actio
n)
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Me
asu
rin
g p
oin
t
Pa
ge
lo
ad
, filte
ring
, filte
r
exp
an
din
g ,
ne
xt
pag
e
On
clic
k
On
clic
k q
uestio
n
On
clic
k N
o-b
utto
n
On
clic
k S
en
d-b
utt
on
On
pro
duct
loa
din
g o
r u
pd
ating
On
pro
duct
loa
din
g o
r u
pd
ating
On
pro
duct
loa
din
g o
r u
pd
ating
On
pro
duct
loa
din
g o
r u
pd
ating
On
pro
duct
loa
din
g o
r u
pd
ating
On
clic
k
On
clic
k
On
2 s
eco
nd
hove
r o
n in
fo-
bu
tto
n
On
clic
k
On
clic
k
On
clic
k
On
clic
k
On
clic
k
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Va
ria
ble
AJA
X r
eq
ue
st fo
r p
rod
ucts
Clic
k o
n h
om
e-p
ag
e “
slid
e b
ann
er”
Exp
and
FA
Q q
ue
stio
n
FA
Q q
ue
stio
n a
nsw
ere
d?
: N
o
FA
Q q
ue
stio
n a
nsw
ere
d?
: N
o, se
nd
fo
rm
Pro
du
ct
SK
U
Pro
du
ct
nam
e
Pro
du
ct
price
Pro
du
ct
am
ou
nt
Cart
valu
e
Dis
pla
y r
ela
ted
pro
duct
Dis
pla
y r
ela
ted
pro
duct
Exp
and
sh
ippin
g c
osts
ta
ble
“Con
tin
ue
sh
opp
ing”
bu
tto
n
Vo
uche
r co
de
activate
d
Exp
and
FA
Q-p
opu
p
Exp
and
co
nta
ct
po
p-u
p
Clic
k o
n D
igiC
ert
log
o
Ord
er
ID
Date
/ tim
e o
f pu
rcha
se
Ord
er
am
ou
nt
VA
T
Sh
ipp
ing
costs
Appendices - 67
Table E.1 Clickstream variables recorded (continued)
Re
ma
rks
Ne
w d
ata
ba
se r
ow
pe
r
pro
duct
Me
asu
red
at
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Ord
er
con
firm
ed
Me
asu
rin
g p
oin
t
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Pa
ge
lo
ad
Va
ria
ble
Am
ou
nt
of p
rod
ucts
ord
ere
d
Pro
du
ct
SK
U
Pro
du
ct
nam
e
Pro
du
ct
price
Pro
du
ct
am
ou
nt
Pro
du
ct
cate
go
ry
Am
ou
nt
of p
rod
ucts
ord
ere
d
Am
ou
nt
of p
rod
ucts
ord
ere
d
Am
ou
nt
of p
rod
ucts
ord
ere
d
Appendices - 68
Appendices - 69
Appendix F. Experiment design
F.1 Cart page
Figure F.1 Web shop cart variation ‘Clean’
Appendices - 70
Figure F.2 Web shop cart variation ‘USP’
Appendices - 71
Figure F.3 Web shop cart variation ‘Cross sell’
on click
Appendices - 72
F.2 Checkout
on click
Figure F.4 Web shop checkout variation ‘Multipage – Step 1’
Appendices - 73
Figure F.5 Top: Web shop checkout variation ‘Multipage – Step 2’ Bottom: Web shop checkout variation ‘Multipage – Step 3’
Appendices - 74
Figure F.6 Web shop checkout variation ‘Single-page’
Appendices - 75
Figure F.7 Checkout success page
Appendices - 76
Appendices - 77
Appendix G. Experiment results
G.1 Descriptive statistics covariates Table G.1 Descriptive statistics of covariates
Variable
Experiment ‘Cart page’ Model 3 Experiment ‘Checkout’
Clean USP Cross- selling Multi-page Single-page
Time of the day
Morning 189
22.1% 153
19.1% 155
18.2% 112
24.9% 21
21.4%
Afternoon 350
40.9% 328
41.1% 375
44.0% 182
40.4% 49
0.50%
Evening / Night 315
36.9% 316
39.6% 322
37.8% 156
34.7% 28
28.6%
Day of the week
Weekday 396
46.4% 397
49.8% 391
45.9% n/a n/a
Weekend 458
53.6% 400
50.2% 461
54.1% n/a n/a
Mean cart value (Euros) 42.20 40.09 43.82 28.88 30.83
G.2 Logistic regression: Collinearity statistics
Table G.2 Collinearity statistics experiments ‘cart page’ and ‘checkout’
Variable Experiment ‘Cart page’ Experiment ‘Checkout’
Tolerance VIF Tolerance VIF
Experiment variaton 0.998 1.002 0.993 1.007
Cart value 0.998 1.002 0.996 1.004
Time of the day: Morning 0.818 1.223 0.764 1.309
Time of the day: Afternoon 0.820 1.220 0.762 1.312
Day of the week: Weekend 0.999 1.001 n/a n/a
Appendices - 78
G.3 Cart page: Linear model
Table G.3 ANCOVA ‘Cart page’, Dependent: Conversion linear, Test of Between-subjects effects
Source Type III
sum of squares Df
Mean square
F Sig. Observed
powerb
Corrected model 4.022 a 6 0.670 3.284 0.003 0.936
Intercept 106.136 1 106.136 519.894 0.000 1.000
Cart value 2.631 1 2.631 12.887 0.000 0.431
Day of the week: Weekend
0.267 1 0.267 1.307 0.253 0.208
Time of day: Morning
0.149 1 0.149 0.731 0.393 0.137
Time of day: Afternoon
0.033 1 0.033 0.160 0.689 0.069
Cart variation 0.853 2 0.427 2.090 0.124 0.948
Error 509.557 2,496 0.204
Total 1,176.474 2,503
Corrected total 513.579 2,502
a
R Squared = 0.008 (Adjusted R Squared = 0.005)
b Computed using alpha = 0.05
Table G.4 ANCOVA ‘Cart page’, Dependent: Conversion linear, Parameter estimates
Parameter B Std. Error t Sig.
95% Confidence interval
Observed power
b Lower
bound Upper bound
Intercept 0.456 0.025 18.270 0.000 0.407 0.505 1.000
Cart value 0.001 0.000 3.590 0.000 0.000 0,002 0.948
Day of the week: Weekend
-0.021 0.018 -1.143 0.253 -0.056 0,015 0.208
Time of day: Morning
0.021 0.025 0.855 0.393 -0.028 0,071 0.137
Time of day: Afternoon
-0.08 0.020 -0.400 0.689 -0.048 0,032 0.069
Cart variation 1 -0.043 0.022 -1.956 0.051 -0.086 -0.000 0,498
Cart variation 2 -0.010 0.022 -0.441 0.659 -0.054 0,034 0.316
Cart variation 3 0.00a . . . . . .
a
This parameter is set to zero because it is redundant
b Computed using alpha = 0.05
Appendices - 79
Table G.5 ANCOVA ‘Cart page’, Dependent: Conversion linear, Pairwise comparison
Cart page variation Mean difference (I - J)
Std. error Sig.b
95% Confidence interval For difference
b
(I) (J) Lower bound Upper bound
1 2 3
-0.033 -0.043
0.022 0.022
0.417 0.152
-0.086 -0.095
0.020 0.010
2 1 3
0.033 -0.010
0.022 0.022
0.417 1.000
-0.020 -0.063
0.086 0.044
3 1 2
0.043 0.010
0.022 0.022
0.152 1.000
-0.010 -0.044
0.095 0.063
Based on estimated marginal means
b Adjustment for multiple comparisons: Bonferroni.
G.4 Cart page: Valued model
Table G.6 ANCOVA ‘Cart page’, Dependent: Conversion linear valued, Test of Between-subjects effects
Source Type III
sum of squares Df
Mean square
F Sig. Observed
powerb
Corrected model 5,254.202a 5 1050.840 1.191 0.311 0.429
Intercept 317,318.064 1 317,318.064 359.631 0.000 1.000
Day of the week: Weekend
62.399 1 62.399 0.071 0.790 0.058
Time of day: Morning
3,587.725 1 3,587.725 4.066 0.044 0.522
Time of day: Afternoon
138.452 1 138.452 0.157 0.692 0.068
Cart variation 1,266.847 2 633.423 0.718 0.488 0.172
Error 2,203,211.685 2,497 882.343
Total 3,493,262.227 2,503
Corrected total 2,208,465.887 2,502
a
R Squared = 0.012 (Adjusted R Squared = 0.007)
b Computed using alpha = 0.05
Table G.7 ANCOVA ‘Cart page’, Dependent: Conversion linear valued, Parameter estimates
Parameter B Std. Error t Sig.
95% Confidence interval
Observed
powerb Lower
bound Upper bound
Intercept 22.683 1.434 15.821 0.000 19.871 25.494 1.000
Day of the week: Weekend
-0.316 1.190 -0.266 0.790 -2.650 2.017 0.058
Hour – Morning 3.317 1.645 2.016 0.044 0.091 6.542 0.522
Hour – Afternoon 0.526 1.329 0.396 0.692 -2.079 3.132 0.068
Cart variation 1 -0.559 1.440 -0.389 0.698 -3.382 2.263 0.067
Cart variation 2 -1.725 1.464 -1.178 0.239 -.4.597 1.146 0.218
Cart variation 3 0.00a . . . . . .
a
This parameter is set to zero because it is redundant
b Computed using alpha = 0.05
Appendices - 80
Table G.8 ANCOVA ‘Cart page’, Dependent: Conversion linear valued, Pairwise comparison
Cart page variation Mean difference (I - J)
Std. error Sig.b
95% Confidence interval For difference
b
(I) (J) Lower bound Upper bound
1 2 3
1.166 -0.559
1.465 1.440
1.000 1.000
-2.344 -4.008
4.675 2.889
2 1 2
-1.166 -1.725
1.465 1.464
1.000 0.716
-4.675 -5.233
2.344 1.783
3 1 2
0.559 1.725
1.440 1.464
1.000 0.716
-2.889 -1.783
4.008 5.233
Based on estimated marginal means
b Adjustment for multiple comparisons: Bonferroni.
G.5 Checkout Table G.9 Results logistic regression experiment ‘checkout’
Model 1 Model 2 Model 3
Variable B Wald
Odds-
ratio B Wald
Odds-
ratio B Wald
Odds-
ratio
Experiment variation:
Single-page 0.165 0.470 1.179 0.147 0.371 1.159 0.155 0.404 1.167
Cart value Not included 0.010 4.806 1.010* 0.010 0.004 5.256*
Hour of visit Morning
Not included Not included 0.494 0.250 1.638*
Afternoon 0.112 0.288 1.118
Constant 0.653 43.248 1.922* 0.388 6.352 1.474* 0.211 1.094 1.235
Model performance Model 1 Model 2 Model 3
Hosmer and Lemeshow
chi2-test
0.000 (p = .) 12.734 (p = 0.121) 15.520 (p = 0.050)
R2 Nagelkerke 0.001 0.014 0.025
Cox & Snell 0.001 0.010 0.018
* p < 0.05