Great Insights from a ZDnet Computer Shopper Article....FYI...
<<Site Building
by David S. Linthicum
Originally published in the March 1999 issue
Using Personalization to Make Web Sites More Profitable
Applying personalization to your Web site gives visitors the ability to customize what content they view and how they view it though they may not always have a choice. Personalization (sometimes called customization) on the Web is cropping up more and more, with almost all search engines, such as Yahoo and Snap, providing personalized "portals" into their space. In addition, major commerce sites, ranging from Amazon.com to Dell's Web site (www.dell.com) are finding that forms of personalization are just plain profitable.
The advantage of personalization is clear: If you're able to provide a custom view of your site to each individual user, chances are that those users will be encouraged to make return visits. Moreover, commerce sites are able to boost sales by automatically keeping track of shopping baskets as well as of credit-card and shipping information from session to session. This saves users from having to reenter information every time they visit a commerce site, and it lets the site track the preferences of the online customer. Sites like CDnow's (www.cdnow.com) have made a science out of understanding their virtual clientele. Still, there's a movement afoot to stop personalization altogether, due to the Big Brother aspect of gathering information over the Net. This column is written from the perspective that personalization makes a Web site more responsive to the individual needs of each user.
On the technical side of things, personalization is a process of keeping track of users through user IDs, passwords, IP lookup, or cookies. Site users are allowed to customize their view of a site by setting options, or the site itself may personalize the view by simply tracking user behavior.
Despite the advantages of personalization, it's not for every Web site. Before you jump feet-first into building personalization support into yours, you need to consider the benefits, the risks, the enabling technology, and its costs, and develop your approach to personalization.
Approaches to Personalization
The first fact you'll discover about personalization is that there are as many definitions as there are vendors that offer solutions. You'll run the gamut of vendor-specific buzzwords such as "one-to-one marketing," "user customization," "interactive agents," and "collaborative filtering," to name only a few. Putting the hype aside for a second, the least common denominator seems to be the technology's ability to monitor a site user's behavior. Since this is the case, pay attention to privacy issues, because in some instances, you're collecting personal data—information your users are hoping you'll never share.
There are any number of ways to enable your site for personalization, but in the end, they all take either a passive or an active approach. Active is the most primitive, requiring that your users type in the answers to a set of questions. The information obtained in this Q&A is then stored inside a database at the site or inside a cookie on the user's hard drive. (For more on cookies, see the July 1997 Site Building column, "A Cookie for Your Thoughts: Cookies Help Webmasters Harness User Habits," by John Krick.) You may also use an IP address lookup to identify users when they return. Be aware that this approach produces some erroneous results, since you may not get the true IP address of the client, depending on how the ISP is managing addresses. A more primitive approach is to require that users enter a code, a user ID, or a password to make themselves known to the site on returning. Though the active approach was popular just a few years ago, the industry is turning to stealth because many users are reluctant to answer questions online.
The passive approach is more transparent to the users. Instead of requiring that site visitors fill out forms, you can track users behind the scenes via cookies or IP address lookup mechanisms. Using this approach, you follow visitors' behavior, and based on revealed preferences, you update a cookie or a database to produce more relevant content. For instance, a user who is always browsing through the men's-wear portion of a department store's commerce site may be presumed to be male, have his cookie updated with that information, and be presented with the latest specials of interest to men, such as grills, power tools, or Dockers. The notion here is to present visitors with content most likely to captivate them, burying less relevant content in menus. Also, when a user surrenders a name, an address, and a credit-card number for a purchase, that information may be tracked as well.
A good example of a passive site laced with some active features is Amazon.com. The site is able to keep track of your movement, maintaining information about you, such as shipping and billing addresses, and one or many methods of payment. Also, Amazon.com can present you with books or music it believes you may find too compelling to pass up, based on your behavior. Other sites are learning these tricks as well, and Web-based product marketing and sales are picking up.
However, active and passive approaches provide only broad categories for personalization techniques. Several personalization methods are emerging from the confusion within this new arena, as are products that support these methods. These methods include rule-based filtering, collaborative filtering, and learning-agent technology.
Rule-based filtering is an active approach wherein the user answers a set of questions, and then the site presents content based on the responses. We've seen such questions many times, such as "What's your age?" "Are you male or female?" and "What's your gross yearly income?" On a commercial site, your answers will determine which products and ads the site displays.
Collaborative filtering provides relevant material to users by combining their own preferences with those of similar users visiting the site. For example, user A goes to a virtual-bookstore site and buys a copy of my latest book and a copy of a book by Martha Stewart. Then user B enters the site and buys a book written by Tom Clancy as well as another Martha Stewart book. As a result, the site would recommend that user B also consider my book based on the behavior of user A--a useful, if not always foolproof, system. It makes no difference that neither user A nor user B knows anything about the other visitor. Firefly (now a part of Microsoft) is the best-known company with a collaborative-filtering solution.
Learning-agent technology simply uses "click trails" to determine which content is most relevant. This is typically done at the site, not at the user level.
So which is right for your Web site? It depends on your objective and who's using your site. Feel free to mix and match. For example, collaborative filtering and rule-based filtering can coexist on the same site, and thus provide more value. It makes sense to perform active and passive information-gathering concurrently, to track both the user who will provide such information when asked and the one who won't.
Personalization Solutions
Well, you can choose to "bake" your own cookies to handle personalization, but it's often a better idea to let other people write such code for you. And there are a dozen or so products that will let you add personalization to your site. They range in price from almost free to $100K, and many are still a bit wet behind the ears.
Broadvision's One-to-One (www.broadvision.com) is an example of a high-end tool that lets sites recognize returning customers and consequently deliver more relevant content to them. This is an example of an e-commerce personalization application; solutions in this category aim to present users with products they are likely to buy. You'll find this technology in use on major sites such as US West's (www.uswest.com), The Baan Co.'s (www.baan.com), and Kodak's (kodak.photonet.com).
Net Perceptions' GroupLens (www.netperceptions.com) is one of the slickest personalization solutions out there and is already part of mega-sites such as Amazon.com and CDnow. This collaborative-filtering solution requires that users actively or passively rate content as they click, and the content they see is adjusted as they rate the site. This is, of course, another high-end tool with a price tag to match.
If you're on a budget, however, you may want go for some of the less expensive but equally valuable solutions, such as Firefly Passport (www.firefly.com). Firefly, which was developed at the MIT Media Lab, is the best collaborative-filtering solution out there in its price range, and fortunately it puts an emphasis on privacy. It's already a part of major sites such as Yahoo, ZDNet(ZDNet is the online branch of Computer Shopper's publisher, Ziff-Davis), and barnesandnoble.com.
LikeMinds' (www.likeminds.com) Preference Server, another low-end tool, uses collaborative-filtering technology that can track user behavior passively or actively, find other users with similar behaviors, and make product recommendations. Preference Server includes APIs for C++ and CGI scripts that can reach out to most relational databases through ODBC. Major clients include Cinemax-HBO's Movie Matchmaker (rw.cinemax.com/critic) and Columbia House's Total E entertainment site (www.totale.com).
The Bottom Line
So are most of you going to enable your site for personalization today? Perhaps not. However, it's inevitable that to compete with the larger sites out there, you're going to have to take some steps to understand your users better and provide content that's fitting for them. Whether you do this through an active or a passive mechanism, use a canned solution, or roll your own, depends on the goals and the objectives of your site.
Clearly, the earliest opportunities for personalization exist in the exploding world of e-commerce. There, the right content sent at the right time can make the difference between a $10 and a $100 sale, and thus affect the survival of a site. The next generation of applications is exciting and will include the ability to provide users with personalized portals into their banks, college-registration systems, and even corporate intranet sites. As the major search engines learn to support personalization, it's just a matter of time before your users will start to expect it.>>
---------------------------------------------------------------------------------------------------------
IMO, NETP's technology is cutting edge and they are in a sweet spot of this new e-commerce revolution. Once the BIG money figures this out we should be moving up towards new highs <gg>..!!
Best Regards,
Scott |
