Clash of the Killer Ps

Can online personalization be profitable and privacy-conscious?

For now, there's an uneasy truce between the Internet's killer Ps - personalization and privacy. Personalization - the ability of e-commerce companies to modify content to match individual customer preferences - has become one of the most critical ways of generating online sales. If the message isn't one-to-one, you're done, the thinking goes.

1pixclear.gif
1pixclear.gif
1pixclear.gif

Personalization Buster

If there's a personalization movement, there's also an effort to thwart it.

Researchers at Xerox Corp.'s Palo Alto Research Center (Parc) have created an algorithm that's designed to keep the behavior of online shoppers hidden from Web site operators.

For example, you visit an online clothing store and search for size and pricing information for shirts. But the electronic footprints you leave behind are masked in a cloud of other digital information, making it difficult for the online retailer to collect and mine specific pieces of data about you.

The concept isn't new; cryptographers have long talked about similar "oblivious-transfer" mechanisms. But the idea has even more relevance today, when commercial Internet sites harvest user clicks by the thousands, says Bernardo Huberman, a research fellow at Xerox Parc in Palo Alto, Calif.

The new algorithm could easily be added to today's Web browsers, Huberman says. Microsoft Corp., which has several joint development deals with Xerox, could, in theory, add the de-personalization algorithm to Internet Explorer, Huberman says.

A Microsoft spokeswoman says no such announcement is imminent. Meanwhile, online retailers can guard against such secretive shoppers by blocking any browser that uses the technology. "But then they would lose customers," Huberman says. "It's a trade-off of economic considerations."

A patent is pending.

1pixclear.gif

But holier than that grail, experts say, is an exquisite balance between using personal data and guarding it.

Though personalization is front and center in many e-commerce business plans, "privacy is an afterthought," says Jonathan Gaw, an analyst at International Data Corp. in Framingham, Mass.

Information technology managers are eager to use Web technology's data mining capabilities. After all, knowing who your customers are and how they behave on your site can lead to more thoughtful, targeted marketing and, presumably, a lot more sales.

However, IT staffers and marketing departments often fail to devise careful policies for using data mining results in ways that preserve customer privacy, Gaw says. That's a mistake at a time when industry and Congress are fiercely debating whether and how to legislate privacy.

"IT people have to start recognizing that you have to think about (personalization and privacy) in the development of the site," he says.

How Do I Look?

Lands' End Inc. is one company that does personalization right, experts say, even when the data being collected couldn't get more personal.

At the www.landsend.com site, women can try on clothes virtually, using online models that are created to look like them. Hair color, body size, face shape and other features are put together to form an online paper doll that users can dress and rotate.

Just 50 of the thousands of garments at Landsend.com are available for virtual tryouts. But that includes bathing suits, which many women dislike trying on under the glare of fluorescent lights at department stores.

"We have some very vocal customers" who suggested adding swimwear to the lineup, says Jeremy Hauser, research and analysis specialist at Lands' End in Dodgeville, Wis.

More than 1 million models have been built in the 18 months the feature has been up on the site, Hauser says. By no means do most customers participate: The site received more than 38 million visits in the past fiscal year, 80% of which were from returning customers.

Still, Hauser labels the feature popular. Though he declined to give a timetable, he says he plans to add male models and hundreds more garments.

Lands' End takes pains to make sure customers can control how much and what kind of information they reveal.

If women don't want to type in their exact measurements to create online models, for example, they can choose from a menu of less revealing descriptions. Bust, for example, can be "small-medium" or "larger" and waist can be "well-defined" or "undefined."

But more important, e-commerce experts say, is the fact that Lands' End doesn't reuse that information to personalize other areas of the site.

Upon logging on, for instance, a shopper with a virtual model on file isn't automatically blasted with cross-marketing come-ons about which sale items would look good on her body type.

That kind of unsolicited personalization generates bad karma on at least two fronts: First, it increases customers' privacy concerns, and, second, it decreases customer trust, says P. K. Kannan, who teaches a course in online marketing at the University of Maryland in College Park.

Do it often enough and customers "will drop you," Kannan says.

Take It Personally

Smart personalization, Kannan says, is done gradually and by giving control to the consumer.

Rather than presenting first-time users with long questionnaires to fill in about themselves, he advises asking few, if any, questions at the start.

Instead, offer optional services, such as a personal shopper or tailored e-mail newsletters. Over time, users will have to reveal more and more information about themselves as they shop and sign up for services.

But the key is keeping the customer in charge of his data rather than harvesting click patterns to make assumptions about individual Web visitors.

Indeed, a Federal Trade Commission report last month found that Internet companies largely fail to protect consumers' privacy. One particular fault: not giving customers access to their data.

In a random sampling of Web sites with more than 39,000 different visitors per month, just 20% met the FTC's standards for a fair privacy policy. Many sites failed to meet one key FTC standard: giving customers control of their personal data.

Of the 100 most popular Web sites, fewer than half - 42% - met the FTC's privacy criteria, the agency says.

The FTC declined to reveal which companies failed. But it suggested that Congress ought to write privacy legislation because industry self-regulation isn't working.

When personalization works, it becomes the glue that makes a site sticky, compelling users to return again and again.

Yahoo Inc.'s My Yahoo is a good example. Users can create personal pages that track stocks, show local weather, filter news, suggest new Web sites and offer other services.

"Over time, you get more pieces of information and can build hooks so that the consumer sticks with you," Kannan explains. "It's too time-consuming to jump to another site and set up all the same stuff."

Missteps

Other kinds of personalization techniques can backfire.

For example, Dell Computer Corp. last year let users construct a personal Web page to get tailored help in selecting computer products. Users entered details about the brands and models of their PCs and peripherals, then Dell would make recommendations.

But many users weren't diligent about updating their profiles when they changed or added products. Dell, therefore, was often working off old data and consequently, giving some bad advice.

So Dell redesigned the service last year to provide more generic online computer support instead.

Amazon.com Inc. often tries to lure wayward customers back to its Web site to shop by recommending books and CDs based on the customers' previous buys.

But that approach is often too broad.

For example, if several family members shop from the same PC, Dad might get an e-mail suggesting he come check out the latest 'N Sync CD. But it is actually 14-year-old daughter Brittany who has an obsession for the all-boy band.

Though the software exists to collect granular data on user behavior, the marketing smarts haven't caught up, says Gaw. "I don't think it's gotten so sophisticated yet".

Amazon.com in Seattle didn't respond to four interview requests. Several vendors, working through various nonprofit organizations, are working on standards for buying, selling and trading customer data.

Exchange Standards Compete

Among the leading contenders is an effort by the International Digital Enterprise Alliance. A nonprofit group in Alexandria, Va., the alliance plans to create a set of specifications based on XML, dubbed the Customer Profile Exchange (CPEX). That standard is supposed to help solve integration problems in blending customer data acquired from outside entities with a company's back-end systems for customer relationship management, order fulfillment and billing.

The first CPEX spec is due by the end of this month.

A similar effort, called the Electronic Commerce Modeling Language (ECML), is also in the works.

The ECML Alliance has created a set of uniform field names for common pieces of customer data, such as name, address and various forms of payment. ECML was designed to help online merchants collect electronic data for shipping, billing and payment.

ECML Alliance members include major credit-card companies, such as American Express Co., Discover Financial Services, MasterCard International Inc. and Visa International Inc. IBM, Microsoft Corp. and other technology vendors are also members.

Meanwhile, both IBM and Microsoft, as well as Novell Inc. and other vendors, separately offer their own digital wallets, which let users store personal data on their PCs and decide which online companies can have it.

The standards projects all compete at this point, and no single format has taken root, says Kerem Tomak, an assistant professor of MIS at the University of Texas in Austin.

"It's in flux, as is e-commerce in general," Tomak says.

But trafficking data about online shoppers - legally and in a methodical, businesslike way - may become quite profitable. Tomak tells of a University of Texas assignment where graduate business students must set up and run e-commerce businesses.

"There was a team last year that opened up an experimental storefront and sold customer demographics to other students at other universities." he says. "Interestingly enough, that was one of the most successful projects."

Related:

Copyright © 2000 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon