=Paper=
{{Paper
|id=Vol-1582/8Adaji
|storemode=property
|title=Evaluating Personalization and Persuasion in E-Commerce
|pdfUrl=https://ceur-ws.org/Vol-1582/8Adaji.pdf
|volume=Vol-1582
|authors=Ifeoma Adaji,Julita Vassileva
|dblpUrl=https://dblp.org/rec/conf/persuasive/AdajiV16a
}}
==Evaluating Personalization and Persuasion in E-Commerce==
https://ceur-ws.org/Vol-1582/8Adaji.pdf
Evaluating Personalization and Persuasion in E-
Commerce
Ifeoma Adaji, Julita Vassileva
University of Saskatchewan, Saskatoon, Canada
ita811@mail.usask.ca,yiv905@mail.usask.ca
Abstract. The use of personalization and persuasion has been shown to opti-
mize customers’ shopping experience in e-commerce. This study aims to iden-
tify the personalization methods and persuasive principles that make an e-
commerce company successful. Using Amazon as a case study, we evaluated
the personalization methods implemented using an existing process framework.
We also applied the PSD model to Amazon to evaluate the persuasive principles
it uses. Our results show that all the principles of the PSD model were imple-
mented in Amazon. This study can serve as a guide to e-commerce businesses
and software developers for building or improving existing e-commerce plat-
forms.
Keywords: Personalization, persuasive technology, e-commerce
1 Introduction
In order to succeed, e-sellers have to give their clients reasons to choose them over
their competitors. E-businesses have to offer their clients a shopping experience that
is pertinent to who the customers are and guided by their needs [1]. One way to opti-
mize a customer’s shopping experience is by providing personalized contents with the
use of persuasive techniques [1], [2], [3]. Using Amazon1 as a case study, this paper
aims at identifying the personalization methods and persuasive principles that make
an e-business successful. This study can serve as a guide for e-business developers to
build successful e-commerce platforms or to improve on existing ones.
2 Related Work and Methods
2.1 Amazon
Amazon is an e-commerce company that started out as an online bookstore, but now
sells other items including clothes, electronics, furniture, food, jewelry and toys 1.
Amazon encourages users to review and rate products they purchase. These reviews
and ratings (along with other metrics) are used by Amazon to build product recom-
mendations for users2. Reviews can be marked as helpful by other users and they can
also comment on reviews and ask questions about products. Answers to these ques-
tions can be up voted or down voted based on how useful users find them. Amazon
uses a ranking system where users are ranked based on how helpful they are to the
community. Ranking in Amazon is based on several factors including how helpful
1
http://www.amazon.ca/
2
About recommendations http://www.amazon.com/gp/help/customer/display.html?nodeId=16465251
Copyright © by the paper’s authors. Copying permitted for private and academic
purposes.
In: R. Orji, M. Reisinger, M. Busch, A. Dijkstra, A. Stibe, M. Tscheligi (eds.): Pro-
ceedings of the Personalization in Persuasive Technology Workshop, Persuasive
Technology 2016, Salzburg, Austria, 05-04-2016, published at http://ceur-ws.org
108 Evaluating Personalization and Persuasion in E-Commerce
other users find a review, how often a user writes a review and how many reviews a
user has written3. Amazon rewards top reviewers with a Hall of Fame badge. These
badges are rewarded to reviewers who rank 1000 or better3.
2.2 Process Framework for E-commerce Personalization
To evaluate personalization in Amazon, we used the framework for e-commerce per-
sonalization developed by Kaptein and Parvinen [4]. We used this model because it
was the only framework we found that evaluated personalization in e-commerce. The
model postulates that there are several requirements responsible for personalization to
succeed and these are grouped into two main categories; requirements regarding cus-
tomer behavior and requirements regarding technology. The three requirements re-
garding customer behavior are; 1) the personalized content presented to a user must
have an effect on the outcome of the business. 2) The effect should be different for
each customer – it should be heterogeneous. 3) The effect should be stable to a large
extent.
On the other hand, the requirements regarding technology consists of the tech-
nology implemented by an e-business in order to tailor contents so specific users.
These requirements are: 1) Ability to measure the effect of personalization, 2) Ability
to manipulate content, 3) Ability to scale the algorithm used for personalization.
In this study, we only evaluated the requirements regarding customer behavior as
these can be inferred from a system. In the future, we intend to evaluate the require-
ments regarding technology.
2.3 Persuasive Systems Design Model (PSD)
PSD is a framework for designing and evaluating persuasive systems. It categorizes
and maps the elements of persuasion in a system and also describes the software func-
tionality expected in the end product [5]. The PSD model suggests three phases of
development and evaluation; understanding the key issues behind persuasive systems,
analyzing the persuasion context and designing of system qualities. We however only
evaluated the third phase, designing of system qualities, because we are concerned
with identifying the persuasive principles adopted in the design of a system. We used
this model for evaluating Amazon because the model was developed specifically for
designing and evaluating persuasive systems and it also describes the content and
software functionality that a typical persuasive system should have.
3 Research Method and Results
3.1 Evaluating Personalization in Amazon
Using the process framework for e-commerce personalization developed by Kaptein
and Parvinen [4], we evaluated personalization in Amazon based on the requirements
of the customer’s behavior. The implementation of these requirements in Amazon are
described in this section.
3
How ranking works http://www.amazon.com/review/guidelines/top-reviewers.html/
Evaluating Personalization and Persuasion in E-Commerce 109
Evaluating Amazon based on the customer’s behavior as described in section 2.2,
it is evident that Amazon personalizes content using several means. Amazon changes
the content displayed to users on the home page based on the last item that user
looked up. If for instance a user searches for a digital camera and views the product
description of one of the cameras displayed, the next time that customer returns to the
home page, Amazon will display several camera suggestions to the user. In addition,
Amazon personalizes content presented to users by allowing them personalize the
adverts they receive from Amazon. Personalized ads displayed to a user are based on
information about the user, like previously viewed products and purchases made on
Amazon4.
The effect of the various content on each customer can only be evaluated by car-
rying out a user study which we plan to undertake in the future. This user study will
ask individual Amazon users about the implementation of personalization by Amazon
and to what degree it is persuasive.
3.2 Evaluating Persuasion in Amazon
In this study, we evaluated the persuasiveness of Amazon as an e-commerce business
using the PSD model [5]. In this section, we identified the persuasive principles of the
PSD model and how they were implemented in Amazon. We focused on the third
phase of the model; design of system qualities. This stage is important as it focuses on
the principles of persuasion that should be adopted in making a system more engag-
ing. The principles in this phase are classified into four categories: providing primary
task support, dialogue support, system credibility support and social support [5].
This study is still work in progress; in the next phase, we plan to validate all the
principles that were identified using a user study, to verify that these principles work
as we assume they do.
PRIMARY TASK SUPPORT.
The persuasion principles in this category support users of a system in achieving their
primary objective or goal. For each principle, we identified at least one implementa-
tion in Amazon. We plan to validate all the identified principles by carrying out a user
study on Amazon’s users. This study will verify if the implementation of these princi-
ples persuade users to use Amazon. The principles in this category and how they were
implemented in Amazon include the following:
Reduction: The reduction principle assets that in order to be more persuasive, a sys-
tem should reduce complex tasks into simpler ones. A typical example in Amazon is
the use of 1-Click. From the preview page of a product, users can use the “buy now
with 1-Click” button to purchase a product without having to add the item to cart,
proceed to check out, preview the payment and shipping address details and then
place the order. Here, the task of buying an item has been reduced to a single click
event.
4
Amazon Ad Preferences https://www.amazon.ca/gp/dra/info?ie=UTF8&ref_=ya_advprf
110 Evaluating Personalization and Persuasion in E-Commerce
Personalization and Tailoring: The personalization principle of the PSD model
states that the more personalized content is available in a system, the more persuasive
the system will be. Similarly, the tailoring principle states that a system that provides
tailored content is likely to be more persuasive than one that doesn’t. The personaliza-
tion and tailoring principles are similar, hence, both are merged in this review. Users
can personalize recommendations on Amazon by rating items previously purchased or
by selecting previously purchased items for Amazon to include in future recommen-
dations. In addition, users can change the language for browsing, shopping and com-
munication on Amazon, as well as manage their payment method and options. They
can also view and manage their browsing history as well as review their wish list
settings. Amazon allows users subscribe to personalized ads based on their activities
on other sites where Amazon provides ads or content. All these, according to Ama-
zon, provide a more personalized experience for the user 5.
Self-Monitoring: According to the PSD model, a persuasive system should allow
users monitor their performance or status. Once logged in, Amazon users can check
their purchase history, previous reviews they have written, helpful votes they have
received and their ranking.
Simulation: In order to be persuasive, a system should enable users see the relation-
ship between cause and effect. In Amazon, before purchasing a book, users can view
some of its contents using the “Look Inside” link.
Rehearsal: A system could change people’s behavior if there exists a means where
users can rehearse a target behavior. This is very similar to simulation. An example of
rehearsal in Amazon is that users can browse for items, view product description and
read reviews without having to sign in or register.
DIALOGUE SUPPORT.
The design principles in this category bring about human-computer communication
with the aim of steering users towards their goal. In Amazon, human-computer com-
munication is implemented through reviews, ratings and communication between
buyers and sellers. For each principle, we identified at least one implementation in
Amazon. We plan to validate all the identified principles by carrying out a user study
on Amazon’s users. This study will verify if the implementation of these principles
are persuasive to users. For example, if a review reminder sent by Amazon actually
persuade users to write a review. The principles in this category and how they were
implemented in Amazon include the following:
Praise. The principle of praise according to the PSD model states that a system’s use
of praise can make the system more persuasive. Amazon implements praise in the
form of helpful votes. For each review a user writes, other reviewers can vote that
review as being helpful or not. A user could be persuaded to review more products if
their reviews are usually voted as being helpful.
5
Improve recommendations in Amazon https://www.amazon.ca/gp/yourstore/iyr?ie=UTF8&ref_=ya_improve_recommendations
Evaluating Personalization and Persuasion in E-Commerce 111
Rewards. The PSD model postulates that systems that reward their users for perform-
ing target behaviors could have more persuasive abilities. Amazon rewards users who
have written several helpful reviews with hall of fame reviewer badges. These badges
are earned by users who earn a ranking of 1000 or better on Amazon. They receive a
hall of fame badge and are listed on the hall of fame page for life. Rankings are earned
by writing helpful reviews very often. The usefulness of a review is decided by other
users in the community6.
Reminders. According to the PSD model, systems that remind users to carry out a
target behavior is more likely to be persuasive. Amazon relies greatly on reviews and
ratings given by users after a purchase to ensure that future recommendations to that
user are accurate7. Hence, once a user makes a purchase, after the expected delivery
date, Amazon sends an email to the user urging him or her to rate and review the item
they purchased. Amazon continuously sends a reminder until such review and rating
is carried out by the user.
Suggestion. The suggestion principle asserts that users are expected to achieve their
target behavior if the system offers suggestions while in use. In Amazon, users are
offered suggestions while typing in the name of a product in the search bar. The sug-
gested product is further classified based on the various departments the product oc-
curs in, giving the user several options to choose from. Amazon also offers sugges-
tions to users in the form of the frequently bought together feature which shows what
items are commonly purchased together by other users based on the current content of
the user’s shopping cart.
Similarity. This principle according to the PSD model states that users are more per-
suaded if a system behaves in a way that is similar to their behavior. In other words, a
system should mimic its users in specific ways. Amazon implements similarity using
the “customers who bought this items also bought” feature. With this feature, custom-
ers can see what other items similar users have bought.
Liking. Going by the PSD model, a system that is liked by its users is likely to be
more persuasive. In order words, a system should look appealing to its users. As this
feature is subjective and can only be determined by users, we did not review this per-
suasive principle. However, the proposed user survey which is a continuation of this
study, will include questions to confirm this principle.
Social role. This principle of the PSD model states that systems should adopt a social
role to make them more persuasive. Amazon has an active social community where
users can ask and answer specific questions about products. Questions and answers
can be upvoted or downvoted based on how helpful they are. Users who participate
improve their current ranking and can earn rewards. Amazon also has an active re-
view system where users can review products and earn rewards while doing so. This
social role amazon plays could be persuasive to some users.
6
http://www.amazon.com/review/guidelines/top-reviewers.html/
7
https://www.amazon.ca/gp/yourstore/iyr?ie=UTF8&ref_=ya_improve_recommendations
112 Evaluating Personalization and Persuasion in E-Commerce
SOCIAL SUPPORT.
This category of design principles describes how to design a persuasive system by
leveraging on social influence. We identified at least one implementation of each
principle. The principles in this category and their implementation on Amazon in-
clude the following:
Social learning. According to this principle, a system can be more persuasive if users
can learn from other users in the system. In Amazon, product ratings and reviews and
their helpfulness are public, hence users can learn from other customers who have
bought a product they are interested in. Users can also learn about sellers based on the
reviews of sellers done by other users that have had dealings with such sellers in the
past. Customers can also learn more about products from other community members
by asking specific questions from those that bought similar items in the past. Users
can also review previously asked and answered questions to learn more about specific
products.
Social comparison. This principle states that a system is more persuasive if it allows
users compare their performance with others in the system. Amazon uses a ranking
system which gives users a rank based on their participation and helpfulness in the
community. This rank is visible on each user’s profile (though it can be de-activated
to prevent public viewing), hence users can compare their performance to that of oth-
er customers in the system. This feature could also help users decide how seriously to
consider another customer’s review. The list of hall of fame reviewers is also public,
hence users can compare their performance to those of others.
Normative influence. The normative influence of the PSD model asserts that a sys-
tem should leverage peer pressure in order to persuade users to carry out their target
behavior. Amazon’s ranking system could be a source of normative influence among
users, whereby customers hoping to rank high in the system buy more products in
order to write reviews that other users could consider helpful, thereby improving their
ranking among their peers.
Social Facilitation. This principle asserts that a user would be more persuaded to use
a system is he/she is able to recognize others also using that system. In Amazon, this
could mean that users are more persuaded to rate a product if other users have rated it
or that users are more compelled to give a high rating if other users did same. In our
proposed user study, we will verify the implementation of this principle by asking
users what persuades them to rate a product.
Cooperation. According to the PSD model, a system that brings about cooperation
among its users could be more persuasive than one that does not. Cooperation in Am-
azon is evident in product reviews where a user describes his/her experience with a
product and other users concur with that user citing their own experience. This could
persuade skeptical customers to take a decision about a product. The customer ques-
tions and answers product feature could also promote cooperation, where users can
comment on existing answers to improve on it.
Evaluating Personalization and Persuasion in E-Commerce 113
Competition. This principle stems from humans’ natural urge to compete and states
that for systems to be persuasive, they should provide means of competition among
users. The ranking system of Amazon could lead to competition among users, where
by customers who want to improve their ranking buy more products in order to write
more reviews with the hope of earning helpful votes.
Recognition. This principle according to the PSD model asserts that a persuasive
system should offer public recognition to users who perform their target behavior.
Amazon recognizes helpful reviews by stating how many people found a review help-
ful. By clicking on a user’s profile, one can see the percentage of helpful reviews that
user has written. One can also view the ranking of that user. In addition to recognizing
helpful reviews, Amazon recognizes helpful reviewers by awarding the hall of fame
reviewer badge to the top ranked reviewers. This badge is visible on all reviews writ-
ten by recipients of the award, hence such awardees are publicly recognized.
4 Conclusion and Future Work
In this study, we identified the personalization and persuasion methods adopted by a
successful e-commerce business, Amazon. Using an existing process framework, we
identified the methods of personalization implemented by Amazon. With the PSD
model, we evaluated the persuasion principles adopted by Amazon. We were able to
identify the implementation of all of the principles of the PSD model. This study can
guide developers and other stake holders in building successful e-commerce business-
es.
This study is work in progress. In the future, we plan to validate the findings in
this paper by carrying out a user study to confirm the effectiveness of the principles
identified. We also intend on evaluating the system credibility support principles of
the PSD model, as this category of principles cannot be inferred from the system but
through a user study.
References
1. B. Schafer, J. Konstan and J. Riedl, “Recommender systems in e-commerce,” in
Proceedings of the 1st ACM conference on Electronic commerce, 1999.
2. R. Rao, S. Khan, Z. Begum and C. Divakar, “Design of E-Commerce Personalized Service
Model Based on Web Mining Classification Technique,” in IEEE International Conference
on Computational Intelligence and Computing Research, 2013.
3. S. Subramaniam, “Frontiers in E-commerce Personalization,” in Proceedings of the 20th
ACM SIGKDD international conference on Knowledge discovery and data mining, 2014.
4. M. Kaptein and P. Parvinen, “Advancing E-Commerce Personalization: Process Framework
and Case Study,” International Journal of Electronic Commerce, pp. 7-33, 2015.
5. H. Oinas-Kukkonen and M. Harjumaa, “A Systematic Framework for Designing and
Evaluating Persuasive Systems,” in Persuasive Technology, 2008