Getting Personal on the Web

Personalization promises to teach corporations all they want to learn about customers. But the fervor for such tools may diminish unless standards and ethical concerns are faced.

Personalization is hot. Some observers say it is the model that will make the Internet the top marketing tool for all types of business. While the personalization concept has been used for years by direct-mail vendors and telemarketers, the World Wide Web promises to bring the concept to new levels, providing organizations with data that was virtually impossible to gather previously.

Companies from all industries, and of all shapes and sizes, are scrambling to develop Web-based systems that can personalize interactions with individual customers. The efforts are spreading through traditional industries, as well as the new so-called "" businesses. And software vendors are moving quickly to supply technologies that can "mine" data for unique characteristics. Suppliers range from start-up Internet firms to software giant Computer Associates International Inc., Islandia, N.Y., whose neural network-based

Neugent technology is said by Chairman and Chief Executive Charles Wang to offer "dynamic personalization. This can give each customer a different view" of a Web page that is based on their past activities on a site.

Still, the road to personalization will not be completely smooth, observers note. The rapid spread of Web personalization has resulted in a lack of a clear definition of the model, similar to other concepts, like CASE and EAI, that spread like wildfire through IT shops. If the trend toward disparate technologies continues, they say, personalization will cause confusion among information technology operations rather than solve problems. Some businesses are now looking to standards and industry cooperation to resolve the issues.

A marketing sea change

Industry observers have alternately described Web personalization as "the Internet's next big thing," "the Holy Grail of e-commerce" and "an ill-defined placeholder for a grab bag of technologies." To some, personalization is customizing the content of a Web site; to others, it represents a marketing sea change with far-reaching implications for anyone selling anything, anywhere.

"Many marketers look at it as a way to sell 2.1 percent more hula hoops this quarter," said Christopher Locke, editor-in-chief of and co-author of The Cluetrain Manifesto: The End of Business As Usual. Locke's Web site, sponsored by personalization vendor Net Perceptions, provides listings of vendors, analysts, industry associations, books and news, as well as pointers to other Web-based resources on personalization and related technologies.

"But what's going on is a fundamental shift in the nature of markets and marketing," continued Locke. "Instead of just figuring out what products you'd be interested in, [personalization software] can figure out what people you'd be interested in and build a market conversation that throws power back to customers."

Added Locke, "This is much more profound than just moving the merchandise. If someone asks me point-blank what kind of books I like, I usually can't give them a particularly useful answer off the top of my head. Even if I sit down to think about it, I can't come up with much more than a few old favorites and the titles I'm currently reading. But if I were to buy a lot of books online, that site could, potentially, provide a much more complete picture of my reading preferences."

Despite the lack of a clear definition and some emerging ethical concerns, a basic advantage has become very clear: Internet personalization can tailor marketing messages to individuals in a far more interactive manner than snail mail, E-mail or telephone models. Marketing mavens have started to call the Internet piece of the personalization puzzle the ultimate one-to-one marketing model.

"The Internet can give consumers a personalized, customized experience," said Larry Ponemon, a partner and global leader of the compliance risk management program at New York City-based consulting firm PricewaterhouseCoopers. "That is important — it's what makes the Internet more than a television set."

The Web clearly allows the personalization model to broadly expand beyond telemarketing and even hyperlinks embedded in E-mail messages. While tra- ditional, offline markets are stuck with top-down, market-research approaches, personalization tools can explore market behavior from the bottom up (in other words, from the data). Web-based personalization systems can use raw information to discover meaningful correlations in the data and employ increasingly sophisticated software to analyze and exploit the shopping behavior of individual customers. For example, an untold number of consumers are now greeted by name at retail Web sites like and CDNow. But both sites also do a lot more than just remember the names of customers; they can also offer product recommendations based on the purchase histories of other customers with tastes similar to a shopper's own.

Observers note that to date, organizations rely on two technologies to build personalization systems: collaborative filtering and profiling. Collaborative filtering calls for using the experiences of an individual consumer to shape electronic responses to that customer. Profiling aggregates the data from multiple Web sites to create a site configured to the needs of an individual consumer. Application server and portal suppliers are quickly starting to unveil tools that promise to make it easier for corporations to add personalization capabilities to Web sites.

Choosing your personalization model

Despite some disagreement on nomenclature, most analysts agree that there are four basic types of personalization models: name recognition, check-box, rules-based and behavioral preference-based. "Keep in mind that the edges blur. These are useful categories for thinking from 40,000 feet. But when you get down to the nitty-gritty, a lot of [personalization products] use all four [models]," said Locke at

Name recognition is probably the most basic form of personalization, as well as the easiest to incorporate into most systems. Included in this model is the letter from Ed McMahon or the "Hi Bob" on someone's MyYahoo! Web page. It is a kind of mass personalization that is in many ways decidedly impersonal. A few years ago, greeting people by name marked a turning point in mass-marketing techniques; it continues to be used by businesses because it still has value. "Most people like to be acknowledged by name," noted Locke.

Check-box personalization involves information customers provide on questionnaires, surveys and other online forms. (Think of the registration page for commercial software.) Users provide this information about themselves by checking answer boxes. A Web site using check-box personalization presents content based on a customer's answers.

Segmentation and rules-based personalization uses a demographic, geographic or psychographic profile, or other information to divide or segment large populations into smaller groups. Data such as income level, geographic location and buying history is aggregated to identify groups of people. Web sites using these types of personalization systems deliver content based on "if this, then that" rules processing.

Preference-based personalization seeks to understand the behavioral preferences of a specific, individual user and then delivers Web site content specifically targeted to that user. This type of personalization, also known as collaborative filtering, is database-driven and operates on more complex algorithms. It can also factor more details into each interaction. Preference-based personalization works by tracking user activity, comparing that activity with the behavior of other users, and then predicting what the user would like to see next. Some preference-based personalization systems can even "learn" and become more precise over time.

Preference-based tools are better at providing real-time personalization functionality. Real-time personalization refers to a site's ability to modify a response to a user based on a changing perception of the user as the interaction occurs. Rules-based and check-box model approaches are not as good at this. They can, however, assimilate new data and alter responses while the user is interacting with the site, provided all potential responses and alternatives have been accounted for.

Collaborative filtering: Personalization from the bottom up

Traditional market analysis techniques require access to data on large groups of people, called customer segments, to define mathematically meaningful relationships. Generally speaking, you need a lot of people to make the math work.

Collaborative filtering is used to look at a single customer, one that is unknown to a company. The emergence of sophisticated collaborative filtering tools is a key reason for the strong emergence of corporate interest in personalization techniques. Such tools make it possible to generate statistically significant predictions and recommendations about an individual, despite the availability of very little information about that person.

If you have ever browsed and clicked on a book, you have probably seen collaborative filtering tools at work. That message at the bottom of the page —"Customers who bought this book also bought:" followed by a list of titles — is a recommendation generated by software that collates buying patterns across a large number of customers. The patterns that emerge from this kind of data are much more than the sum of their parts. What surfaces is something Locke refers to as "the largely unconscious process by which social communities assess and measure value." This emergent behavior of complex systems, he noted, is the real guts of the latest and most sophisticated personalization techniques.

"It's organization and coordination that takes place without explicit planning or premeditation," explained Locke. "Patterns emerge among otherwise unconnected readers, or among any group of people with a shared interest. In many cases, these are communities that aren't even aware of themselves. Collaborative filtering tools are new and very powerful tools that fit the Internet largely be- cause they are capable of defining these micromarkets."

Here is how the concept works: The unknown user provides answers to a small number of questions (the "collaborative" part), which the personalization tool compares with the known preferences of customers in its database (the "filtering" part). The tool then generates a temporary "community" of customers with these shared preferences. It can now extrapolate recommendations based on other preferences the community shares.

The Album Advisor page on the CDNow Web site is another example of collaborative filtering. On the page, a buyer of a gift is asked to provide the titles of three recording artists the gift recipient likes. The personalization tool — in this case, the Net Perceptions for E-Commerce product — uses these three data points as the basis for its recommendations.

"When you input the three artists, the Net Perceptions tool creates a kind of neighborhood of people who like those three artists," said Steve VanTassel, Net Perceptions' vice president of commerce solutions. "Then it looks for things those neighbors have in common beyond those three artists. Where it finds the greatest degree of commonality, it says, 'Hey, this is what I think this person is also going to like.'"

The collaborative filtering process has its roots in research conducted at the University of Minnesota in the early 1990s. Net Perceptions, an Eden Prairie, Minn.-based personalization tools provider, was founded in 1996 to expand that research and later commercialized the concept. "We create the customer segments —this 'neighborhood' — from the bottom up or from the individual out," said VanTassel. "Once another action is taken, that neighborhood is disbanded and a new one is created. That's why we call it 'dynamic.'"

Unlike rules-based personalization tools, which can determine in advance a set of conditions that guide the site's response to its customers, collaborative filtering technologies impose no business context. Nonetheless, the two models are often used in collaboration on the same Web site. A movie retailer, for example, might want a rule stating that anyone known to be under 17 years old will not be shown movies rated R or above. Now that the playing field is established, the site still might want to make recommendations, which is where the collaborative filtering would kick in.

B2B personalization

Beyond business-to-consumer e-commerce applications, the personalization model is finding its way into business-to-business (B2B) applications. "Knowing the interests and profiles of your partners and suppliers helps you recommend meaningful products and services," said Todd Boes, director of product marketing at Allaire Corp., Cambridge, Mass. In March, Allaire acquired the Open Sesame profiling and personalization technologies from Browne Internet Solutions.

"Personalization, in a very large sense, is about taking general-purpose software and turning it into something that is specific to a customer or the customer's customer," said Joshua Greenbaum, principal at Enterprise Applications Consulting, Berkeley, Calif. "On the business side, personalization has a lot to do with being able to capture the rules and business processes that define the way you do business.

"[For example,] I have a different way I want to give incentives to my customers and partners — maybe I want to give them a discount when they buy in bulk, or I have a different approach to managing my supply chain," he added. "I want to build those rules into my software so that I'm automating as many processes as possible."

According to Boes at Allaire Corp., personalization technologies and strategies do not appear exclusively in retail and business environments, nor are those necessarily the best applications of the process.

"It's definitely being used to customize content delivery at information portals and pure content sites, such as CNN or ESPN," he said. "Because CNN knows that I happen to be interested in financial news and I follow the technology industry, that's the content I'm going to see when I go there. In fact, a lot of personalization is going to be used on the service and support side, improving daily interactions with customers — purchases, support, refunds, that sort of thing. I think it's better utilized in non-retail environments, where the personal interactions are more meaningful."

Boes said Allaire is beginning to shy away from the personalization moniker because the term is limiting. Instead, Allaire uses the term "customer intelligence" to describe the process. "A lot of the information gathered for personalization purposes is useful across organizations," said Boes. "Analyzing that information, whether you're in finance, marketing or support, is critical."

'Pushing the ethical envelope'

The concept of personalization has posed ethical concerns to a variety of organizations, however, that fear some companies might misuse personal information gathered with or without the permission of its customers. "I am part of the group that believes in the Web, that it marks a revolutionary business change," said Price- waterhouseCoopers' Ponemon, who also founded the firm's privacy practice. "But we are starting to see areas that are pushing the ethical envelope."

PricewaterhouseCoopers is a founding member of the fledgling Personalization Consortium, formed in April by 26 companies of all sizes "to promote the responsible and beneficial use of technology for personalizing consumer and business relationships." Among the founding members of the Wakefield, Mass.-based consortium are American Airlines, BroadVision, E.piphany, KPMG Consulting and Servicesoft. Said Ponemon before the consortium was unveiled during the Spring Internet World conference in Los Angeles: "If we don't do this now, we will face more and more government regulation. We have to have some accountability to government, but we want industry to set up a system of self-regulation" that can include realistic rules.

Precisely where personalization and privacy collide is anyone's guess. Privacy advocates fear any trend that puts personal information in the hands of those who might abuse it. Proponents of personalization swear they take every precaution, such as anonymous identifiers and using aggregated data to protect the privacy of their customers. Still, many users simply do not like to surrender personal data.

"People have been reluctant to use cookies, and there are people who just don't want to give up any information about themselves," said Allaire's Boes. "But people are warming up to the idea because it gives them a better experience. They're beginning to see it less as an invasion of privacy and more as a benefit."

The Personalization Consortium recently asked 20,000 Web users about personalization and online privacy. The survey found 51% of respondents willing to exchange personal information for better service. In contrast, only 15% of respondents said they would be unwilling to share any information with Web marketers, even if it meant improved service.

When respondents were asked what personal information they would provide to a Web site in order to customize their online experience, 76% said they would divulge their hobbies and interests, 81% would provide an address, 95% would provide an E-mail address, and 96% said they would supply their name.

Eventually, the advantages of personalization — improved service and faster access to preferred information — can outweigh the disadvantages of sharing personal information with unknown merchants, observers say. And the ethical concerns can be addressed, noted Ponemon at PricewaterhouseCoopers. Until then, e-commerce organizations will continue to have to provide consumers with privacy controls that allow them to choose their level of personalization, and to even opt out altogether, he said. Allaire's Boes breaks personalization strategies down into two approaches: explicit and implicit. Explicit personalization gets a user to volunteer information (rules-based). The forms users have to fill out to create a MyYahoo! page are part of an explicit personalization strategy. The implicit approach leaves it up to the software to analyze user behavior (collaborative filtering).

But for all the heat this concept is generating, surprisingly few corporations are utilizing sophisticated personalization technologies or strategies, said Boes. "There are a lot of folks who are doing very basic things. We see lots of simple, straightforward explicit personalization. But we're not seeing many using the implicit approach," he said.

"In the online world, you can find a better price with a mouse click," added Boes. "How do you keep people at your site and get them to purchase from you? It's going to come down to customer service, to that vendor knowing you better than anyone else, and [being able to] service your needs and provide support when you need it to keep you from going anywhere else."

John K. Waters is a freelance writer based in Palo Alto, Calif. Michael W. Bucken also contributed to this article.

The language of personalization

Personalization is an old idea driven by new technologies and new jargon. Here is a short list of definitions to help you wade through some of the emerging nomenclature.

Customer relationship management (CRM) — a technique that brings information about customers into a central repository. Serves to provide a better "read" on individual customers and market trends. CRM tools integrate all company operations that involve customers, including marketing, sales and service.

Customization — a kind of personalization technique that allows the customer to control the content of a Web page in order to "customize" it to reflect their particular interests. Usually, there is a form to fill out. MyYahoo! and MyExcite are examples of customization.

One-to-one marketing — what personalization is all about. In this context, it refers to techniques and technologies that gather information about individual customers, as well as to the automation of marketing efforts designed to communicate with individual customers.

Permission marketing — a term probably coined by Seth Godin and Don Peppers, whose book, Permission Marketing: Turning Strangers Into Friends, and Friends into Customers explores the concept of enlisting the customer as a partner in your marketing efforts. The authors argue that traditional forms of "interruption advertising" — which bombard consumers from magazines, direct mail and broadcast media — are less effective than those designed to get a customer’s permission with some kind of "bait" (such as a free sample, a big discount, a contest, an 800-number or an opinion survey).

Relationship marketing — a broad category of marketing activities that involves the management of ongoing relationships with customers, partners and/or vendors.

Touch points — "places" where the enterprise connects with the customer, including the call center, direct mail and Web sites. CRM tools allow organizations to connect company touch points with a common database that allows for the sharing of information through a central repository.

— John K. Waters

Privacy poll
Strogly Disagree Disagree Neither Agree nor Disagree Agree Strongly Agree
A privacy statement is necessary for me to share personal information. 5% 11% 24% 31% 27%
I find it helpful and convenient when a Web site remembers basic information about me (e.g., my name and address). 4% 5% 17% 42% 31%
I find it helpful and convenient when a Web site remembers more personal information about me (e.g., my preferred colors, music or delivery options). 7% 13% 30% 31% 19%
Banner ads and “pop ups” are an invasion of my privacy. 8% 21% 35% 16% 19%
I am willing to give information about myself in order to receive an online experience truly personalized for me. 4% 11% 33% 41% 10%
It bothers me when a Web site requests personal information I have already provided (e.g., my mailing address). 5% 9% 23% 38% 24%
Source: Personalization Consortium, April 2000

A closer look at the personalization space
As fast as things are moving in the personalization space, it would be difficult to offer a definitive list of product vendors. The following listing contains the names of some of the major players as of press time.
Company Location URL Address Product/Technology
Allaire Corp. Cambridge, Mass. ColdFusion, Open Sesame Profiling Server
Annuncio Software Inc. Mountain View, Calif. Annuncio Bright
Art Technology Group Inc. Cambridge, Mass. Dynamo
Autonomy Inc. San Francisco, Calif. Content Server
beFree Marlborough, Mass. Adaptive relationship modeling technology (gained through acquisition of TriVida Inc., Culver City, Calif.)
Blaze Software San Jose, Calif. Blaze Advisor, Blaze Expert
Broadbase Software Inc. Menlo Park, Calif. Broadbase EPM
BroadVision Inc. Redwood City, Calif.
OneToOne/ SessionMgr/home_page.jsp
Computer Associates International Inc. Islandia, N.Y. Neugents
Cyber Dialogue New York, N.Y. Telescope
Edify Corp. Santa Clara, Calif. Electronic Workforce
eHNC San Diego, Calif. SelectCast, SelectResponse
Elity Systems Inc. Somerset, N.J. Elity Insight Engine
Engage Andover, Mass. Engage ProfileServer
E.piphany Inc. San Mateo, Calif. E.piphany E.4 System, E.piphany Real-Time Recommendation Engine
GuestTrack Inc. Los Angeles, Calif. gt/index.html GuestTrack
Gustos Software LLC Laguna Hills, Calif. . Gustos Guide
ILOG Inc. Mountain View, Calif.
Components for e-Business
iLux Corp. Newark, Calif. Suite 2000
Inference Corp. San Francisco, Calif. k-Commerce Sales
Macromedia Inc. San Francisco, Calif. Aria, Likeminds
Manna Inc. Wellesley, Mass. FrontMind
MatchLogic Inc. Westminster, Colo. MatchLogic
Metabyte Inc. Fremont, Calif. MbTV
MicroMass Communications Inc. Cary, N.C IntelliWeb
net.Genesis Corp. Cambridge, Mass. net.Analysis
Net Perceptions Inc. Eden Prairie, Minn. Recommendation Engine Suite
NovuWeb Inc. Manhattan Beach, Calif. Genius Server
Open Market Inc. Burlington, Mass. IPS, Xcelerate, Transact
Personify Inc. San Francisco, Calif. Personify Essentials, Personify Snapshot Palo Alto, Calif. Responsys Interact
Tian Software Co. Denver, Colo. e-volving/ASP
Vignette Corp. Austin, Texas StoryServer, netCustomer
Younology New York, N.Y. OrbjectWare
Source: John K. Waters