Capítulo II Los proyectos de
2.2.1 Desarrollo sostenible
A site can use many different methods to provide personalization. Many Web sites differentiate themselves based on the factors of price and effort to implement. These are the wrong factors on which to base the type of personalization. The decisive factors you should use are as follows: the reason you want to implement personalization and what you are trying to accomplish by personalizing your Web site.
Although you can use personalization types individually, I submit that a better approach is to use multiple types of personalization together, building on the strengths of each.
Nominal Personalization
Nominal personalization is the simplest of all the different personalization types. All it really does is provide the user with a friendly greeting when she returns to the Web site. It requires the user of the site to answer a simple form requesting her name.
After the user enters the information into the form, the Web site creates a cookie (see
Chapter 12 for more information), which is stored on the user's site and is retrieved by the Web site whenever the user returns. A Web site can handle the cookie in one of two ways. The first way is to store the information in the cookie itself and then, when the user returns to the site, retrieve the information from the cookie. The second way is to place a random and unique key in the cookie. When the user returns, the Web site retrieves the cookie and looks the key up in the repository of users, which provides the username to display on the Web site.
The first method is only good for storing the username. This is because the placement of any other information in the cookie is unavailable for the Web site's use because the Web site will not be able to access the cookie until the next time the user logs in. The first method also starts to infringe on the user's privacy, an issue that is covered later in this chapter.
Group Personalization
One of the easiest ways to handle multiple groups on a single Web site is by using group personalization. For example, a Web site for students might use group personalization to distinguish which grade a user is in. Once the Web site knows which grade the user is in, it can route her to the correct menus and Web pages.
Usually, group-personalized sites require some form of registration. The registration typically involves entering information on a form, which the site then stores in a user repository. The Web site administrator then verifies the information. If the user's information is satisfactory, she is e-mailed a username and password.
Two different approaches exist for handling returning users with group personalization. The first is similar to nominal personalization. A unique random key is stored in a cookie after the first time the user logs on using the supplied username and password provided in the e-mail. Each subsequent time the user logs in, the Web site retrieves the key from the cookie and logs the user in. Or, if security is an issue, it asks for a password. The second method forces the user to log in from a generic start page using the supplied username and password. No cookie is stored on the user's computer.
In both cases, privacy is protected because no user information passes over the Internet, other than in the case of the first method, in which a random key traverses the Internet.
Subscription Personalization
Subscription personalization provides a way for a site to send newsletters to users. It requires the user of the site to enter information into a simple form that requests an e- mail address to which the Web site can send the newsletter. Quite often, the Web site
will send a test e-mail to the provided e-mail address, asking the user to verify that he is in fact the person requesting the newsletter.
Basically, with subscription personalization, a Web site is sending a newsletter to a target group of users who have an expressed interest in the newsletter's content. It usually includes subtle, or not so subtle, advertisements to encourage readers to make purchases from the Web site or a partner site.
On a very simple implementation of subscription personalization, typically in conjunction with nominal personalization, the newsletter is often used just as a way for the Web site to remind the user of its existence.
Intelligent Agent Personalization
An intelligent agent is really subscription personalization on steroids. It basically is a more elaborate subscription personalization Web site requesting more information from the user. From the information provided, the intelligent agent will generate a newsletter containing articles (or hyperlinks to articles) that it believes the user will be interested in, as opposed to a generic newsletter as produced by subscription personalization.
Some intelligent agents will only examine the content of their own Web sites. Other intelligent agents can become pretty elaborate and will send out spiders all over the Internet, searching for information that the user wants.
A spider is a little program that searches a Web site for content; if it finds any, it sends the content back to the requester. What makes a spider special is that it can locate hyperlinks and jump to a hyperlinked page recursively. In theory, a spider can search the entire Internet for something. When implemented, though, most spiders are restricted to the level for which they can recursively jump because it would take a very long time to search the entire Internet. More important, however, the memory requirements would be astronomical to run a nonrestricted spider because it has to remember where it went on a stack of some type.
Push or Webcasting Personalization
Push technology is very similar to subscription personalization and intelligent agents in that a user subscribes to receive information from a Web site. The big difference is that the information does not come by way of e-mail. Instead, it goes directly, at prearranged intervals via an unobtrusive transmission, to the user's computer from the push Web site.
Unlike normal Web browsing, the content comes to the user; the user does not have to go look for it. The basic procedure is to turn on the browser to receive the push
transmissions, and then periodically the browser will display updated information without any activity on the part of the user.
The setting up of this content can get very elaborate, depending on the push Web site provider used. The process usually involves the user going through many lists of content type, selecting the topics of interest. For example, a user could select weather for her area, local and international news, information about her stock portfolio, and information about her favorite sports.
Notice that privacy is maintained at the level at which the user feels comfortable. Users have the choice to disclose as much or as little about themselves as they want. With this simple example, the push Web site knows information like in what ZIP code the user lives, in what stocks the user has an interest, and what the user's favorite sports teams are. With this information, the push Web site can place ads on the browser for
stockbrokers in the user's area or tickets to local sporting events involving the user's favorite teams.
Customization
Customization often goes under the name of My Yahoo personalization. Figure 4-3 is the real My Yahoo from which the name comes. It is a Web site that enables a user to customize the look and feel and what sections of content he wants to display. This type
of site often becomes a user's home page when he logs onto the Internet because it provides the daily information that the user wants and links to the Web sites he frequents.
Figure 4-3: My Yahoo Screen shot reproduced with permission of Yahoo! Inc. © 2000 by Yahoo! Inc. YAHOO! and the YAHOO! logo are trademarks of Yahoo! Inc.
Customization uses the same approach as group personalization when it comes to Web returns (that is, cookies or login). Not only user information is stored in the repository but also the layout of the home page, preferred content sections, hyperlinks, and any other information that the site enabled the user to customize.
Privacy is maintained because for this type of personalization, total anonymity can be retained if the site wants. The username does not need to be stored anywhere on the site to use customization. At the very least, however, customization usually is used in conjunction with nominal personalization.
Rule-Based Personalization
Rule-based personalization is a method of personalizing in which a Web site requests the help of the user to define what she is looking for from the Web site. This type of personalization requires the user to show some trust in the Web site because the user will be asked many different questions, some of which the user will not want to—or simply will not—answer.
Two schools of thought exist for setting up a user when implementing rule-based personalization:
1. The first approach is to get the user up and running as fast as possible so that he will be immediately productive. This method will instantly ask the user a series of targeted questions, each trying to help the Web site target which content to provide to the user. The user must have a true need for what this site is providing for him to willingly answer so many questions up front.
2. The second approach is to ease the user into the Web site by asking questions as needed. This method is frequently implemented by requesting information when there is an immediate need for the information. The user is aware of this need and therefore is not put off by the request. For example, when a user selects a sports link, the Web site could ask him which sports are his favorites before actually displaying the linked page. For more personal information, such as a street address, the site might only ask after the user has agreed that he wants something shipped to him. This method of setting up a user is less obtrusive and is more likely to succeed, especially if the user does not have an immediate need to use all the available functionality of the Web site.
The goal of both of these setup procedures is to create a complete profile of the user so that the site can generate content targeted to the user. When a user accesses the Web site, the personalization system will examine the profile of the user and see if any available information would be of specific interest to the user. It will then sort all the content based on weighting factors calculated from the answers provided to the questions it posted during the setup of the user's profile.
Privacy is maintained because the profile information never leaves the Web site. For a user to get access to the site, she would have to log in either with a username and password or with a cookie and, usually, a password.
Collaborative Filtering
The collaborative filtering method of personalization is by far the least obtrusive as far as the user of a site is concerned. In fact, if implemented correctly, a user may not even be aware that a profile is being made of her.
This method of personalization is based on the study of group behaviors. This means that if several users like something, the odds are good that the current user will as well. Another way of looking at it is that if a user starts a pattern that previous groups of people have already followed, it is likely that the user will be striving for the same end. Thus, a collaborative filter can look for patterns and, when it finds one, provide the next element in the pattern before the user actually requests it.
To implement collaborative filtering, the Web designer has to embed tags in each piece of content. Tags are developer-defined pieces of information that uniquely identify a piece of content. Then, as a user navigates through the Web site, the personalization engine records in its own repository all the information viewed and the pattern used to get there.
The biggest problem with collaborative filtering is that it takes time to seed the system with enough information to make it useful. The more people navigate the Web site, the more patterns the collaborative filter can determine.
Because it takes some time to get enough information into the system to make it provide any relevant personalization for the Web user, collaborative filtering provides two cheats. The first cheat is that a personalization system can load the data of another system, thus providing instant seeding. The second cheat is that many collaborative filters ask users directly about their preferences using Web form questionnaires without having to navigate there themselves. For example, a video store Web site might ask a user, after she buys a particular movie, what other videos she might be interested in that are similar to the one she just bought, instead of waiting for the user to navigate to them.
One of the key user benefits in collaborative filtering is that a user has complete privacy. Other than a randomly created unique ID key stored as a cookie, there is nothing to link the user to the data being collected. No names, addresses, or phone numbers are needed for collaborative filtering personalization.