Transcript Content Personalization Overview

Web Content Personalization
Overview
Based on the Tutorials of K. Garvie Brown,
R. Wilson, M. Shamos and others
1
Content Personalization
Overview
according to K. Garvie Brown
 http://www.comnet.ca/~gbrown/personalization/
 ... with some extensions
2
Personalizing Web Resources for a User one of the basic abilities of an intelligent agent
Users
Web
Resource
3
Contents
 1. Introduction
 2. Developing a user profile
 3. Tracking Visitors
 4. Technical Requirements
 5. Conclusion
 6. Examples of Online Retailers Using
Personalization
 7. Useful Information Links
 8. Links to Personalization Systems Vendors
4
1. Introduction
 1.1. What is content personalization
 1.2. Business case for content personalization
 1.3. An important quote
 1.4. Examples of Web personalization
 1.5. Laws of Web marketing
 1.6. Learn About Your Site Visitors
5
1.1. What is Web Site Content
Personalization
Personalizing a web site means providing content
that is relevant specifically to the user.
Each person gets a customized view of the web
site. Through personalization technology, web
servers modify the pages that are viewed by each
user with the goal of providing a unique and
personal web viewing experience.
6
1.2. Business Case for Content
Personalization
Content personalization is a way to build one-to-one
relationships with customers. One-to-one relationships
enable firms to keep customers longer and sell more to them
over time. Providing an online experience that is tailored to
meet the needs, interests and personal tastes of customers,
helps to develop a sustainable one-to-one relationship.
"If we have 4.5 million customers, we shouldn't have one store, we
should have 4.5 million stores." Jeff Bezos, CEO, Amazon.com
"Customers are more likely to return to your store if you personalize
and tailor your services to their individual needs and interests." Jim
Caroll & Rick Broodhead, SELLING ONLINE, Jan. 1999
7
1.3. An Important Quote
"If personalization remains an algorithmic trick, nothing
more than a way to hoodwink people into believing your
company knows and cares about them, don't be surprised
when it backfires. Internet audiences are smarter and better
informed than the audience ever was in the "good old days"
of broadcast media. If you try to mess with people's heads
online, they will find you out - guaranteed. And then they'll
hunt you down for sport."
Christopher Locke, Editor-in-Chief, Personalization.com
8
1.4. Examples of Web
personalization (R. Kohavi)
 Greet user by name
 Remember their last shopping
basket
 Remember preferred shipping
address and
credit card
 Change home page image
 Change links
(recommended assortment)
9
1.5. Laws of Web Marketing:
The Law of Dead End Street (R. Wilson)
Setting up a website is like building a storefront
on a dead-end street. If you want any shoppers,
you must give them a reason to come.
"If you build it, they will come" - that doesn't work
on the Internet.
10
1.5. Laws of Web Marketing:
The Law of Pull and Push (R. Wilson)
Pull people to your site by your attractive
content, then push quality information to them
regularly via e-mail.
Getting an invitation to send e-mail to your visitors is key
to this strategy. Include a form that will collect their email address. To convince your visitor to give you his email address, however, you need to promise two things:
(1) that you'll e-mail him something of value, and (2) that
you won't sell or rent his address to another company,
hence the need for a clear privacy policy.
11
1.6. Learn About Your Site Visitors
(R. Wilson)
 Monitor E-Mail Inquiries and Complaints
 Provide Online Questionnaires
 Send Out E-Mail Questionnaires
 Use Cookies Strategically
 Examine Order Files
 Provide Site Personalization
 Study Your Traffic Logs
 Employ JavaScript on Your Site
12
1.6. Learn About Your Site Visitors
 IP address, e.g. 192.151.11.40.
– Anonymous, but I might know your employer
 Domain name, e.g. hp.com
– I probably know your employer
 Name, address, phone no.
– A good start
 Social security number
– I know everything
13
2. Developing a user profile
 To personalize a web site, information about each
visitor must be gathered and stored. To accomplish
this, many web sites create individual visitor or
group profiles. Below are the primary methods used:
2.1. Active Profiling
2.2. Collaborative Filtering
2.3. Passive Profiling
2.4. Preference-Based Personalization
2.5. Observing Short-Term Interests
14
2.1. Active Profiling
Web site users are asked to complete online registration forms that request
basic personal information and details about special interests. There is a
problem, however, with registration forms. According to a recent study
from Jupiter Communications, about 40 % of individuals surveyed provide
incorrect information, while more than 30 % refuse to complete the form.
Other profiling methods are therefore necessary.
2.1.1. Customer’s profile features.
Here are several examples of user profile forms:
2.1.2. My Yahoo user profile form.
2.1.3. GuestTrack user profile form.
15
2.1.4. MySmarterKids.com registration form.
2.1.1. Customer’s Profile Features
Geographic (Are the customers grouped regionally, nationally, globally?)
Cultural and Ethnic (What languages do the customers prefer to do
business in? Does ethnicity affect their tastes or buying behaviors?)
Economic conditions, income and/or purchasing power (What is the
average household income or purchasing power of your customers? What
are the economic conditions they face as individuals? As an industry?)
Power (What is the level of decision-making and title of your typical B2B
customer?)
Size of company (What company size are you best able to serve? Do you
determine this best by annual revenue or number of employees?)
Age (What is the age of the companies you do business with? Dot-com
start-ups or several decades old. What is the predominant age group of
16
your target buyers? How many children and of what age are in the family?
2.1.1. Customer’s Profile Features
(Continue)
Values, attitudes, beliefs (What are the predominant values that your customers
have in common? What is their attitude toward your kind of product or service?)
Knowledge and awareness (How much knowledge and education do your
customers have about your product or service, about your industry?)
Lifestyle (How many lifestyle characteristics can you name about your
purchasers?
CACI (http://www.caci.com/Products/MSG/Databases.html) has developed the fascinating
ACORN system of 43 closely targeted lifestyle profiles that can be tied to specific ZIP codes.
http://www.premierinsights.com/acorn.html You would be able to determine patterns for your
best customers that would guide future marketing.)
Buying patterns (How consumers of different ages and demographic groups
shop on the Web.)
Media Used (How do your targeted customers learn? What do they read? What
magazines do they subscribe to? What are their favorite websites ...?)
17
2.1.2. My Yahoo User Profile Form
18
2.1.3. GuestTrack User Profile Form
19
2.1.4. MySmarterKids.com Registration
Form
20
2.2. Collaborative Filtering
Collaborative filtering technology enables companies to deliver
personalized content based on the preferences of "like-minded"
individuals. The system "learns" more about the user's individual
preferences and adjusts content accordingly.
2.2.1. Main steps in collaborative filtering
2.2.2. “Push” Technology
2.2.3. Rating New Products
2.2.4. “Movie Critic” example
2.2.5. “Levi Strauss” experience
21
2.2.1. Main Steps in Collaborative
Filtering
Collaborative filtering technology enables companies to deliver
personalized content based on the preferences of "like-minded"
individuals. The system "learns" more about the user's individual
preferences and adjusts content accordingly.
Collaborative filtering is accomplished through the following steps:
• web site visitors are asked to complete a questionnaire
designed to identify special interests;
• the accumulated results are tabulated and analyzed
respondents are segmented into groups of "like minded"
individuals;
• the system then delivers content to individuals based on
22
the preferences of their "like minded" group.
2.2.2. “Push” Technology
For example, a visitor to an online music store
queries a server for music by a specific rock band.
The server searches for and delivers the request.
Additionally, the server suggests music that may be
of interest to the visitor, based on the preferences of
the individual's "like minded" group.
Assuming the music store maintains a record of the
visitor's earlier requests, this approach could be used
to "push" new music to the visitor.
23
2.2.3. Rating New Products
For collaborative filtering to work, users must
continuously rate new products by
completing online questionnaires. This can
become tedious, over time, and visitors may
loose interest.
24
2.2.4. “Movie Critic” example
25
2.2.4. “Movie Critic” example (continue)
26
2.2.4. “Movie Critic” example (continue)
27
2.2.4. “Movie Critic” example (continue)
28
2.2.4. “Movie Critic” example (continue)
29
2.2.4. “Movie Critic” example (continue)
30
2.2.5. “Levi Strauss” experience
One of the goals of Levi Strauss & Company (LS&Co.) is to get
closer to the consumer and thus deliver product offerings that
are more relevant to consumer needs and wants.
Online
LS&Co creates personalized style and clothing recommendations
from its online informational catalog of products. It collects and
analyzes various customer tastes, e.g. in fashion, music, sports,
and other activities, then predicts which Levi's® products are
most likely to fit the customer's individual style and needs.
Personalization Firsts
First personalization effort in the fashion industry to use
collaborative filtering.They use customer preferences in
fashions, music and activities to predict preferences for specific
clothing products without knowing what products the customer
has selected in the past.
31
2.3. Passive Profiling
In general, this approach develops a profile based on how an
anonymous user interacts with the web site.
Passive profiling can collect the following information:
 The website the user came from (good for tracking advertising
effectiveness)
 What the user clicked on while visiting the site (for determining
most popular website features)
 Purchases made (determining how demographics, psychographics,
clickstream behavior etc. relate to categories of goods purchased)
 Queries made (providing greater info about user interests)
 Content of the web pages viewed (providing greater info about user
32
interests)
2.3. Passive Profiling (continue)
The information gathered about the anonymous website user
is stored in a data base. Complex data analysis techniques are
then employed to sort the data into user profiles.
Data analysis techniques include:
2.3.1. Data mining,
2.3.2. Online Analytical Processing,
2.3.3. Pattern-matching algorithms and
concept extraction,
2.3.4. User modeling agents
33
2.3.1. Data Mining
Data Mining is a class of database
applications that look for hidden patterns in a
group of data. For example, data mining
software can help retail companies find
customers with common interests.
Data mining software discovers knowledge
in a form of previously unknown
relationships among the data.
34
2.3.1. Data Mining (Examples of
Discovered Knowledge)
 Association Rules:
80 % of customers who buy beer and sausage buy
also mustard
 Rules:
If (Age < 25 or Age > 55) and Sex = ‘Female’,
Then Probability_to_order_ fishing_ tackle < 3 %
 Functional Dependencies
 Belief Networks
 Clustering ...
35
2.3.2. Online Analytical Processing
Online Analytical Processing, a category of software tools
that provides analysis of data stored in a database.
OLAP tools enable users to analyze different dimensions of
multidimensional data (taken for e.g. from the Web page
click stream*). For example, it provides time series and
trend analysis views.
*Data Stream:
Ordered sequence of points, x1,…, xi, …, xn, that can be read only once or a
small number of times in a fixed order
36
2.3.3. Pattern-Matching Algorithms and
Concept Extraction
For example the Dynamic Reasoning Engine (DRE™) is
based on advanced pattern-matching technology that exploits
high-performance probabilistic modeling techniques. The
DRE™ performs four main functions:
 Concept matching: The DRE™ accepts a source of text as input and returns
references to documents in another source of text with the highest degree of
relevance.
 Concept extraction: The DRE™ accepts text (a training phrase, document or set
of documents) and returns an encoded representation of the most important ideas in
the source, including each concept’s specific underlying patterns of terms.
 Concept retraining: The DRE™ accepts a concept and a set of text and adapts the
concept using the text.
 Standard text search: The DRE™ accepts a Boolean term or natural language
query and returns a list of documents containing the terms ordered by relevance. 37
2.3.4. User Modeling Agents
 What is “User Modeling”?
– automated personalization
– make educated guesses about users
 Goals
– How agents model their users? => Acquisition
– How to build the model? => Learning
38
2.4. Preference-Based
Personalization
Preference-Based Personalization - monitoring the activity
of a web site user through successive visits to the site.
Based on a model of the user's interests created from the
information accumulated.
The software analyzes the content of pages viewed by the
user, and through the application, e.g. neural net technology,
and complex algorithms, develops a user profile. Based on
this profile, content that is similar to pages previously
visited are suggested to the user during future sessions. This
technology constantly monitors the users behavior, and thus
provides an continuously updated user profile.
39
2.5. Observing Short-Term Interests
The ADWIZ technique tracks and responds to short-term
interests without the need for user profiles or cookies.
The technique uses keywords inputted into search engines to
match possible user interests with ad banners.
The system tracks the click-through rates of search engine
users and determines which key words correlate with the
highest click-through.
When a user inputs a query, the system inserts an appropriate
ad banner on the page containing the search result. It is
assumed that individuals who search for the same information,
are more likely to click on the same ad banners.
[Read more in: M. Langheinrich1 at al, Unintrusive Customization Techniques for Web
Advertising, http://www.ccrl.com/adwiz/adwiz-www8.html]
40
3. Tracking Visitors
 3.1. Passwords
 3.2. Clickstream
 3.3. Online Purchase Transactions
 3.4. Cookies
 3.5. Passing Data Between Pages via URLs
 3.6. Open Profile Standard
 3.7. Location-Based Personalization
 3.8. Route-Based Personalization
 3.9. Semantic Personalization
 3.10. Privacy Issues
41
3.1. Passwords
The least controversial way to track users is to ask them
who they are.
This can be easily accomplished with user names and
passwords.
Each time a user logs onto a web site, his or her
clickstreams, search queries, banner click through etc.
are tracked and the user's profile is updated.
42
3.2. Clickstream
This is a raw log of the Web pages requested by the
user and is tracked by the web site server. Page content
can then be analyzed, and user interest profiles can be
created.
 Determine distinct visitors;
 Determine repeated visits;
 Determine popularity of different sections of the site;
 Understand when and where people leave the site.
43
3.2. Clickstream (click behavior)
CASUAL VISITOR
OFFICE
PRODUCTS
PRESENTATION
ITEMS
LASER
POINTERS
STORE
HOME PAGE
HOUSEWARES
KITCHEN
TOASTERS
SPORTING
GOODS
HUNTING
RIFLES
GOLF
CLUBS
44
3.2. Clickstream (click behavior)
PROSPECTING VISITOR
STORE
HOME PAGE
OFFICE
PRODUCTS
HOUSEWARES
PRESENTATION
ITEMS
LASER
POINTERS
LASER
1
LASER
2
KITCHEN
TOASTERS
LASER
3
SPORTING
GOODS
HUNTING
RIFLES
GOLF
CLUBS
45
3.2. Clickstream (examples of products)
 MatchLogic
 Andromedia
 E.piphany
 Broadvision
 Personify
 net.Genesis
 Accrue Software
www.matchlogic.com
www.andromedia.com
www.epiphany.com
www.broadvision.com
www.personify.com
www.netgen.com
www.accrue.com
46
3.3. Online Purchase Transactions
Purchase transaction records provide valuable data into the
preferences and interests of online customers. These records
can easily be added to a personal profile. Transactions could
include details of:
 Items purchased.
 The recipient of the purchases (i.e., whether
for personal use or for a third party).
 Items added to shopping cart (since not all
items added may be purchased).
 E-coupons offered, accepted, and used.
47
3.4. Cookies
A cookie is a small piece of information written by a web
site on to a visitor's hard drive.
Cookie files are small, comprising no more than 255
characters and can contain a variety of information,
including the name of the web site that issued them, the
pages visited, passwords, and customer account numbers.
Cookies are supposedly only retrievable by the site which
issued them, and link the information gathered to a unique
ID number assigned to the cookie.
48
3.4. Cookies (continue)
Cookies were originally designed to free a web site visitor
from the chore of typing ID's and passwords. However,
marketing consultants such as DoubleClick and MatchLogic
quickly began to utilize cookies to improve the targeting of
banner ads.
When your Internet browser first visits an appropriate web page your
browser is assigned a unique, anonymous number. This number is
recorded in a small text file that is transmitted to your computer and
stored in the cookie directory of your hard drive. Then, when you visit
each web page, DoubleClick servers can recognize your browser as a
unique, anonymous user.
49
3.4. Cookies (DoubleClick)
Merchant Cookie
Client
DoubleClick
Cookie
1. Client requests a page
2. Server sends a page with
a DoubleClick URL
Merchant
Server
e.g. Altavista
3. Text is displayed
Web Page
4. Client requests the DoubleClick page
5. DoubleClick
reads its cookie
If you choose to give u personal information
via the Internet that we or our business partners
may need -- to correspond with you, process an
order or provide you with a subscription, for
example -- it is our intent to let you know how
we will use such information. If you tell us that
you do not wish to have this information used as
a basis for further contact with you, we will
respect your wishes. We do keep track of the
domains from which people visit us. We analyze
this data for trends and statistics, and then we
discard it.
DoubleClick
Server
6. DoubleClick decides
which ads to send
50
3.5. Passing Data Between Pages via
URLs (no cookies required)
Information about a visitor's online behavior can be tracked
without the use of cookies. One method is to place data
within a Uniform Resource Locator (URL).
Uniform Resource Locator is the global address of documents and other
resources on the World Wide Web.
The first part of the address indicates what protocol to use, and the second
part specifies the IP address or the domain name where the resource is
located.
Through JavaScript, information about a visitor's actions on a
web page can be captured and placed into the URL that is
sent to a server, when the visitor requests a new page. This
data is in the form of Name/Value Pairs. A server side
51
application reads the URL, and extracts the additional data.
3.5. Passing Data Between Pages via
URLs (Name/Value Pairs example)
Browsers automatically create a name/value pair for each form element.
Names and their values are delimited by an equals (=) symbol; multiple
name/value pairs are delimited by an ampersand (&) symbol.
For example, if your page contains a form that has two text input fields
named firstName and lastName, when Fred Flintstone fills out the form
and submits it, the form's data is formatted in the search string as follows:
?firstName=Fred&lastName=Flintstone
If the ACTION attribute of the form is set to:
http://www.bedrock.com/registration.html,
then the entire submitted URL will be:
http://www.bedrock.com/registration.html?firstName=Fred&lastName=Flintstone
52
3.5. Passing Data Between Pages via
URLs (continue)
Benefit:
The key benefit of using URL's rather than cookies to track
visitor activity. Tracking visitor activity by passing data
through URL's does not leave any record on a visitor's hard
drive.
Shortcoming:
The downside, however, of using URL's is that once a visitor
leaves the web site, there is no way to identify that individual
when they return, unless they physically type in a user name
and or password.
[Read more in: D. Goodman, Passing Data Between Pages via URLs,
http://developer.netscape.com/viewsource/index_frame.html?content=goodman_url_pass/goodman_url_pass.html ]
53
3.6. Open Profile Standard
OPS is a recommendation made by Netscape to the standards
body W3C (3.6.1.). The proposal recommends a standard
format be developed that will enable web users keep
personalization records on their hard drives that can be
accessed by authorized web servers. In this proposal, users
will have access to the records, and can control the
information presented. These records will replace the use of
cookies and manual online registration.
The OPS has been examined by the W3C, and key ideas have
been incorporated in the P3 Platform for Privacy Preference
Project (P3P) (3.6.2.). This project will help to set standards
for privacy on the Web. Also HP’s client-side profile storage
54
solutions (3.6.3.) can be implemented.
3.6.1. W3C - the World Wide Web
Consortium
The World Wide Web Consortium (W3C) develops
interoperable technologies (specifications, guidelines,
software, and tools) to lead the Web to its full potential as a
forum for information, commerce, communication, and
collective understanding.
http://www.w3.org/#news
W3C also provides guidelines for designing user agents
(3.6.1.1.) that lower barriers to Web accessibility, which for
example important for people with visual, hearing,
physical, and cognitive disabilities.
55
3.6.1.1. User Agent
User agents (Web browsers, media players, plug-ins,
assistive technologies ) are any software that retrieves
and renders Web content for users.
Plug-in is a program, which runs as part of the user agent being
not part of its content. Users generally choose to include or
exclude plug-ins from their user agent.
Assistive technology is a user agent that relies on services
provided by one or more other "host" user agents. Assistive
technologies communicate data and messages with host user
agents by using and monitoring APIs (application programming
interface).
56
3.6.2. P3P
The Platform for Privacy Preference Project (P3P) is an
activity of the The World Wide Web Consortium.
The P3P enables Web sites to express their privacy practices
in a standard format that can be retrieved automatically and
interpreted easily by user agents.
P3P user agents will allow users to be informed of site
practices (in both machine- and human-readable formats) and
to automate decision-making based on these practices when
appropriate.
Thus users need not read the privacy policies at every site
they visit.
57
3.6.3. Client – Side Profile Storage
User centred architecture
( http://www.hpl.hp.com/techreports/2001/HPL-2001-
291.pdf )
User devices form a trustworthy cluster. The profile is distributed and
replicated on user devices. Each device can store part of the profile
represented by jigsaw pieces. For instance, the mobile phone and the game
console both have a copy of the same jigsaw piece. External system can
access profile information but each exchange of data is under user control.
58
3.7. Location-Based Personalization
With known preferences, a user is able to get useful
information based on where he is. He'll finds out what he
needs to know—when and where he needs to know it—and
quickly find the closest suitable restaurant, or be
automatically informed of a nearby sale. All from his own
personal mobile device.
Handset-based
solutions determine
subscriber location
using Global
Positioning System
(GPS) technology
59
3.7. Location-Based Personalization
Benefon Esc! - personal
navigation phone
60
3.8. Route-Based Personalization
Static Perspective
Dynamic Perspective
61
3.9. Semantic Personalization
 I.e. Personalization based on semantics of a
content;
 Possible e.g. in Semantic Portals [SEAL
Approach, Maedche and others];
 Use semantic bookmarks (which contain
predefined semantic query formulas, specific name,
owner ID, stylesheet);
 Use semantic logfiles (which include concepts
and relation of the ontology a user is interested in) 62
3.10. Privacy Issues
Personal profiling is a form of web site visitor surveillance.
This leads to a number of ethical considerations. For
example, should companies perform passive profiling of
users without their knowledge? Is it ethical to place cookies
on hard drives, without the owners' knowledge? In response
to similar concerns, internet privacy has now become a
major issue. Web site visitors must be convinced that any
information collected will remain confidential and secure.
The W3c's P3P project is addressing these and many other
privacy concerns.
63
3.10. Privacy Issues (Example of a
Privacy Statement by SFGate* )
* A personalized online newspaper that uses
passive profiling to personalize its content.
When you visit a Web site you can expect to be notified of:
• What personally identifiable information of yours is collected;
• What organization is collecting the information;
• How the information is used;
• With whom the information may be shared; What choices are available
to you regarding collection, use and distribution of the information;
• What kind of security procedures are in place to protect the loss, misuse
or alteration of information under the company’s control;
• How you can correct any inaccuracies in the information.
64
4. Technical Requirements
4.1. Hardware and Software Required
4.2. Support and maintenance
4.3. Cost
4.4. Personalization roadblocks
65
4.1. Hardware and Software
Required
Personalization systems require the purchase of expensive
proprietary software. Examples:
GuestTrack runs on UNIX Web servers that uses a standard CGI-BIN
configuration (e.g., Sun, SGI, Linux); requires 3-4 MB to store files and
approximately 1 MB to run; can run many copies of GuestTrack or
GT/Mail software simultaneously.
Personify Essentials - three-tier analytical application server, has
Web-based, thin-client architecture; technologies from data mining and
OLAP (Online Analytical Processing) for large-scale data analysis and
profile creation.
BroadVision provides a number of one-to-one products, designed for
large-scale internet businesses.
66
4.2. Support and Maintenance
Most of company sites claimed their personalization software
was easy to maintain, easily integrated into business systems,
and highly scaleable. They also claimed that excellent support
and training was readily available.
The guess is however that the cost of implementation and
maintenance will outweigh the initial purchase cost.
67
4.3. Cost
GuestTrack: $4,000 per Web site (IP, domain) license fee.
Annual maintenance is 20% of current license fee and
includes technical support, updates and upgrades.
BroadVision's One-To-One Enterprise system (boxes and
all), costs about $350,000 and can take up to three months to
implement.
68
4.4. Personalization Roadblocks
69
5. Conclusion
 5.1. Benefits of personalization;
 5.2. Dangers associated with online
profiling;
 5.3. How to avoid similar problems
 5.4. The Federate Trade Commission is
watching
70
5.1. Benefits of Personalization
Web site personalization has many benefits.
For the web site visitor, personalization enables a more
interesting, useful and relevant web experience.
For the web site provider, personalization allows one-toone relationship building and mass customization.
Additionally, through tracking and profiling, web site
providers can test advertising performance, and determine
the kind of information that appeals to specific market
segments and individual visitors.
71
5.2. Dangers Associated with Online
Profiling
Profiling has hidden dangers. Visitors can feel that their
privacy has been violated when their web viewing behavior
is tracked, profiled and recorded on a database.
The RealNetworks incident provides a good example of the
backlash that can occur when people discover that their
behavior has been covertly monitored.
RealNetworks provided a free CD listening software program,
RealJukebox, to 12 million users. Without informing users, the
program relayed information about personal listening habits to a
server at RealNetworks' corporate headquarters. When this secret
was discovered and made public by an internet security expert,
many consumers reacted unfavorably.
72
5.3. How to Avoid Similar Problems
Although the RealNetworks incident did not involve web
site visitor profiling, it does serve as an illustration of how
people can react to profiling when it is discovered.
To avoid this kind of problem, companies must ensure that
visitors are fully aware of any tracking and profiling that is
taking place. A privacy policy such as the revised
RealNetworks General Privacy Principles (5.3.1.)
demonstrates a framework for ensuring that users are made
aware of any profiling that is occurring. Many other web
sites that utilize profiling make similar disclosures.
73
5.3.1. RealNetworks General Privacy
Principles (revised)
• RealNetworks does not share with third parties any personal information
you provide in connection with our products, without first obtaining your
informed consent.
• RealNetworks maintains controls to provide security over personal
information, including credit card details, during and after the purchase
process.
• RealNetworks sends you email regarding products and services only if
you have indicated you wish to receive email from us. You can remove
yourself from our emailing list at any time.
• RealNetworks does not associate your personal information with usage
behavior from our products unless we have received your informed
consent to do so.
• RealNetworks responds to your privacy questions. You may contact us at
74
[email protected].
5.4. The Federal Trade Commission
is Watching
Online profiling has attracted the attention of the US Federal
Trade Commission.
FTC is interested in online profiling, and may initiate
regulatory measures. This means that firms who use online
profiling must ensure that no public relations incidents occur.
It is vital that all personal information be kept in strict
confidence. Any sale of personal data to a third party could
result in a perceived violation of trust.
However, if permission is obtained, web site visitor profiling
and content personalization can be one of the most valuable
marketing and customer service tools available today.
75
6. Examples of Online Retailers
Using Personalization
6.1. Amazon.com
6.2. MySmarterKids.com
6.3. Clinique.com
6.4. Nike.com and Moviecritic.com
76
6.1. Amazon.com
Amazon.com 's "Your Account" provides customers with a
number of personalized services. Examples include the
ability to review order histories, view or update
subscriptions email newsletters, add or edit a Special
Occasion Reminder, to name a few.
http://www.amazon.com/exec/obidos/account-accesslogin/102-4037886-9968018
77
6.2. MySmarterKids.com
MySmarterKids.com, an educational product retailer, uses
personalization to present products that are suited to a
child's individual goals and learning needs. Using
information gleaned from an online registration process,
the personalization software sets up an individual account.
When logging into the system, the personalization system
searches the product database and provides the customer
with dynamically updated product recommendations that
are specifically tailored to the child's individual needs.
http://www.smarterkids.com/help/FAQreccenter.asp
78
6.3. Clinique.com
Clinique.com, a skin product retailer, has an online
personalization system that calculates a customer's skin
type, and provides product recommendations. The system
also incorporates a registration procedure that enables the
website to "remember" the customer during future online
sessions.
http://www.clinique.com/app/nph-consult.cgi
79
6.4. Nike.com and Moviecritic.com
Nike offers a personalized service in which customers can
design their own shoe, and have it delivered to them a few
weeks later. The site uses technology that requires high
bandwidth, so loading the web page can take a while.
http://www.nike.com/nike_id/
Movie critic is an example of a web site that uses
Collaborative Filtering.
http://www.moviecritic.com/
80
7. Useful Information Links
White paper on Intelligent Agents (Software that learns about a users
preferences). http://www.firstmonday.dk/issues/issue2_3/jansen
Open Profile Standard submission by Netscape to the W3c.
http://www.w3.org/TR/NOTE-OPS-StandardPractices.html
Article: Personal data standards on the way.
http://www.infoworld.com/cgi-bin/displayArchive.pl?/99/08/i0208.57.htm
Personalization.com, an info site devoted to Website personalization.
http://www.personalization.com
Article: Data Mining Gains Favor. Discussing the application Data
Mining to web site personalization.
http://www4.zdnet.com/filters/printerfriendly/0,6061,2305865-35,00.html
81
8. Links to Personalization
Systems Vendors
ATG URL: http://www.atg.com Product: Dynamo
Andromedia URL: http://www.andromedia.com Product: Likeminds
Autonomy URL: http://www.autonomy.com Product: Autonomy Update™
Blue Martini URL: http://www.bluemartini.com Product: Blue Martini EMerchandising
Broadvision URL: http://www.broadvision.com Product: One To One
Engage URL: http://www.engage.com Product: Knowledge, ProfileServer
E.piphany URL: http://www.epiphany.com Product: E.piphany E.4
GuestTrack URL:http://www.guesttrack.com Products: GuestTrack Web
Net Perceptions URL: http://www.netperceptions.com Products: Net Perceptions
Personify URL: http://www.personify.com Product: Essentials
Vignette URL: http://www.vignette.com Product: StoryServer
82