Log file Component Potential Marketing Application
IP address of the browser making the request; Detect at least the Internet Service Provider of
user machine name is not usually recorded the user
Country code and domain name Determine which regions of the world might
best be targeted
Hour, minute, and second of the request, in Determine some of the web habits of a user ,
addition to the date and day of the week for example, are they late night surfers, etc.
HTTP method of the request; (type of request). Know what types of requests are most
commonly made
Response status with the server (the success Improve the service of the web site
or failure of the request);
Number of bytes transferred in the transaction Determine which files are downloaded more
often
Referring URL (where the visitor was when the Determine which on-line ads are most effective
request was made to your site
User name, if authorization is required Identify a user and create a profile about them
Type of browser used by visitor Ensure compatibility of web-site with most
common browsers
Web pages on the server visited Determine potential areas of interest for the
customer
Personalization
The most popular one-to-one technology particularly in business to consumer on the Web today
is Personalization. Personalization is a technique that has been used in traditional direct
marketing in order to differentiate customers based on their values. There is so much choice on
the Web that it is becoming increasingly vital for Web marketers to differentiate their Web sites
and services by dynamically creating personalized web pages or entire web sites. Personalized
sites present visitors with real-time content tailored to their specific preferences on an ongoing
basis giving the consumer a value-added experience, providing a compelling reason to revisit the
site and helping companies build customer loyalty.
Currently there are three key technologies for personalization available, Collaborative filtering,
Rule-based reasoning and Case-based reasoning. These tools make it possible to recommend
products to customers based on purchase similarities with other customers.
Collaborative Filtering
Collaborative filtering is a recommendation system. An intelligent agent sorts previously created
profile information of users and finds other people who are like them based on their profile and
creates affinity group.
Firefly, of Cambridge, MA, markets agent technology that compares information that a consumer
has provided on surveys to data provided by large numbers of other people to recommend
products or services that the consumers might want. The user registers for Firefly Passport,
which acts as a navigational command center for the various Firefly features and sites. Firefly
uses a 1 through 7 scale rating system for recommendations. Barnes & Noble uses Firefly
Passport for book recommendations and Yahoo! uses Passport on its customized site, My
Yahoo!, which allows users to create their own web page and customized searches. Firefly claims
it now has 2.5 million unique users.
GroupLens, part of the ongoing research project on personalized recommendation systems at the
University of Minnesota and direct competitor of Firefly, offers collaborative filtering-based
NetPerceptions that matches a database of users’ preferences to other user’s input and delivers
product recommendations. Amazon uses NetPerceptions to create its book recommendations.
Over 40% of Amazon’s customers are repeat buyers.
WiseWire uses collaborative filtering to take data retrieval to the next level by learning from users
which sources are best for particular topics. The site organizes content into specific sources
called Wired. Content is delivered from a wide range of sources but filtered according to users’
tastes. WiseWire’s approach to scanning and rating content is rather unique among the products
mentioned. WiseWire combines the community-based algorithms of recommendation systems
with more traditional artificial-intelligence techniques. The company describes its system as
collaborative neural network systems.
Rule-based reasoning
Rules-based reasoning creates users profiles based on user preferences and information
requests. It allows a company to apply traditional business logic to targeting content or
advertising or products at an individual. For example, ‘if user is male and in the following age
group and in the following zip codes, show him the following content ’. It enables a fairly simple
approach toward personalization based on profile information.
BroadVision, who specializes in Web-based one-to-one software, serves Fortune 1000
companies that expect a heavy demand for their Web sites to help them create close
relationships with their customers. BroadVision’s One-to-One Web brand is built on a rule-based
reasoning technology, consisting of various software tools. It allows a Web marketer to track
individual users, dynamically change each Web page and matches an individual customer’s
tastes and preferences based on their previous on-line usage. One-to-One Web matches a new
user’s input to a set of predefined rules, and adds the templates, objects and business rules
relevant to e-commerce. Kodak and the Internet Shopping Network have built stores using this
software. One of the crucial benefits of One-to-One Web is that marketing managers can set and
reset the company’s own business rules without any technical assistance by using a feature
called Dynamic Control Centre. According to Internet Week, One-to-One has perhaps the most
far-reaching personalization system, offering an end-to-end platform for personalization, content
management, and dynamic Web publishing.
Micromass’ IntelliWeb developers tie content databases to expert system-based rules/facts
databases that are triggered when a visitor’s information is entered. It dynamically creates each
web presence in real-time, based on the current - individual - profile of each web site visitor and
personalizes anything per visitor – text, graphics, Java applets, etc.
Case-based reasoning
Case-Based Reasoning is a relatively new paradigm in the Artificial Intelligence field, in which
new problems are solved by storing, retrieving, and adapting the solutions to previously
encountered problems. It offers both a cognitive model of human problem solving and a concrete
methodology for building knowledge-based systems. CBR is based on the premise that expertise
is experiential in nature.
Cases contain and relate individual bits of knowledge about instances of things people have
experienced.
Brightware’s Brightware 1.0 enables companies to actively solicit questions from Web visitors to
engage them as sales leads so that they can turn their Net presence into a round-the-clock
selling tool. Brightware 1.0 achieves this through its sales server’s inbound marketing agent that
listens to customers, answers their questions, sends information, and refers them to sales
automatically. The inbound marketing agent automates replies to free-form Web and electronic
mail inquires based on an information extract technique combined (keywords and predefined
rules) with the power of its own Case-based reasoning engine. Inbound marketing agent handles
50 to 80 percent of incoming requests, and accurately processes 95 percent of these
instantaneously. Messages that are not understood by the system are routed to personnel. 24
hour instant response to customer’s inquiries enhances customer satisfaction.
Currently rule-based reasoning software package appears to be more widely used on the Web
than collaborative filtering and case-based reasoning. The rule-based reasoning can be
customized easily by non-technical users and allows them to change the rules on the fly without
changing application logic. while collaborative filtering takes a more automated, black box
approach, meaning users are not able to figure out what exactly happens in the system. Further,
where the knowledge in a domain is well understood, a rule-based system is likely to be more
compact and easier to use than the equivalent case-based system.
However, those technologies can be complementary, and in fact Firefly uses both collaborative
filtering and rule-based reasoning with different products. Barnes and Noble’s online marketing
uses both technologies. Brightware 1.0 also has both rule-based and case-based reasoning
functionalities.
A major drawback of personalization is that it requires a huge database to be effective, therefore
high-traffic sites such as Amazon and Barnes & Noble benefit from this technology.
Personalization is still too costly (more than $250,000 for integration) to be widely accepted by
smaller companies. WiseWire offers its software on a service-bureau basis, at $900/month for 5
subjects in order to accommodate small to medium size customers.
http://web.mit.edu/ecom/www/Project98/G2/offer.html#Personalization