BUSINESSWEEK ONLINE : APRIL 3, 2000 ISSUE

BUSINESS WEEK E.BIZ -- COVER STORY

Weblining
Companies are using your personal data to limit your choices--and force you to pay more for products

You may think that getting graded A, B, or C ended with graduate school. Try getting Sanwa Bank to waive its $20 fee on your bounced check. Customer reps are trained to treat everyone politely. But your luck will depend on a little letter that pops up on a screen as soon as your name is punched into a computer, or when your e-mail arrives at Sanwa's server. If that letter is a ''C,'' customer reps don't exactly hustle on your behalf. That's because machines whirring at Net-speed have lumped you--often in seconds flat--with other customers whose accounts don't make much money for the bank. But if you score an ''A,'' you're right up there with the cream: Customers who generate hefty profits get bounced-check waivers, no questions asked. And B's? They're harder calls. They actually get to negotiate with the rep before their case is decided.

At First Union Bank (FTU), it's a similar story. Its Web-aided computer system, called ''Einstein,'' takes just 15 seconds to pull up the ranking on a customer. First Union won't describe the formula in any detail, other than to say that the ranking software takes stock of minimum balances, account activity, branch visits, and other variables. But the color-coding that appears in a little square on the screen offers a hint. Greens get more flexibility on credit card rates. Reds receive less--and may pay higher fees for some basic services.

Created unequal. What if the reds get mad and take their business elsewhere? Well, that's sort of the point, says Paul Rachal, who helps develop software for banks at Unisys Corp. ''This idea about 'whatever the customer wants' is gone,'' says Rachal. ''Now, it's whatever companies can afford to offer, based on each customer's worth. Not all customers are created equal.''

Today, bank computers assign ranks based on your assets and your past behavior as a customer. But soon they'll have far more information to work with. Reams of personal details are flooding into thousands of databases across the Web--some of which exist for no other purpose than to sell data to others. Want to know who's Jewish and who's Japanese? You'll find guides to that in a catalog from Acxiom Corp. (ACXM), an information broker that stockpiles names, addresses, income, race, religious affiliations, and other data on 95 million American households.

Companies have long scrutinized their customers, both to spot those who are high-value and to weed out the money losers. But with the oceans of information available on the Net, plus ever faster computers and software, companies can maintain the equivalent of profit-and-loss statements on every customer. They can sort people into more categories and, in some cases, predict how they will behave.

All this slicing and dicing will go into hyperdrive as banks, insurers, credit-card companies, retailers, media concerns, and medical businesses move the bulk of their activities onto the Net. Forrester Research says some 23% of companies are beginning to use the Net to ''micro-segment'' customers. By this time next year, the number could swell to 60%.

That's good news for companies: The more finely they can dissect your data profile, the more closely they can tally what you are likely to cost them against the profits you bring--and cut you off if you don't add up nicely. But to many people this micro-segmentation raises serious questions. Left to evolve, says Forrester analyst Bob Chatham, this technology could lead to a commercial culture in which ''high-value customers are bought and sold like derivative securities.''

Call it Weblining, an Information Age version of that nasty old practice of redlining, where lenders and other businesses mark whole neighborhoods off-limits. Cyberspace doesn't have any real geography, but that's no impediment to Weblining. At its most benign, the practice could limit your choices in products or services, or force you to pay top dollar. In a more pernicious guise, Weblining may permanently close doors to you or your business.

Old-style redlining is unacceptable because it is based on geographic stereotypes, not concrete evidence that specific individuals are poor credit risks. Webliners may claim to have more evidence against the people they snub. But their classifications could also be based on irrelevant profiling data that marketing companies and others collect on the Web. How important to your mortgage status, say, is your taste in paperbacks, political discussion groups, or clothing? Yet all these far-flung threads are getting sewn into online profiles, where they are increasingly intertwined with data on your health, your education loans, and your credit history.

This confluence of many streams of personal information, unthinkable just four years ago, is rushing into uncharted canyons. Using technology known as neural networks, many companies believe they can predict your behavior based on all these scattered facts. Cardholders will never be shown their data. And scientists who devise these programs admit that they can't vouch for their accuracy, or even say how they reach a specific conclusion. Neither can the programs' designers, some of whom admit that the technology is dicey. ''You've got to be really careful with this, so you don't start judging people on the wrong assumptions,'' says Tony Patterson, a director at HNC Software Inc. (HNCS) in San Diego, which sells this technology to banks and others. HNC has a new division, eHNC, which will market neural nets to a broader array of Internet businesses.

Are the stereotypes and segmentation that all this technology generates any less disturbing than geographic stereotypes? Lots of people are starting to ask that question. Says Ward A. Hanson, a marketing professor at Stanford University's Graduate School of Business: ''It's your data profile that will determine if you get the best service, the best price, and the best access to new products and information.''

If you do get Weblined, you may not even know it. For that, thank the wonders of ''personalization'' software, which quietly analyzes data and dishes up products or services seemingly tailored to you. On the surface, it might seem as if you are getting something really spiffy--a suggestion for a vanity checkbook, or a reminder of your daughter's birthday. In fact, somebody--or perhaps just a computer--has decided what you are fit to see, sample, or buy, based on sometimes-crude calculations of what you are worth to the firm. ''Companies can segment without being very obvious about it,'' says Forrester's Chatham. ''But just because it's personalized doesn't mean that what you're seeing and getting is first-class.''

And if you peek into the machinery of personalization, you may not like everything you see. Here's how personal it's getting on the Net: Data broker Acxiom offers a new service called InfoBase Ethnicity System, described in a 1999 marketing catalog as a ''broad and precise breakdown of ethnic, religious, and minority classifications.'' The service can, in seconds, match names against housing, income, education, and other demographic data--and identify individual or group ethnicity, designated by ''B'' for black, ''J'' for Jewish, ''W'' for white, ''N'' for ''Nipponese'' (meaning Japanese), and so on. Prices for blocks of such information start at $1,500. You can request the full names, addresses, and ages of pre-school children, or ''select parents and children by age, gender, and declared religious affiliation.'' If you have a product you would like to target to ''full-figured African American women,'' as the catalog puts it, you can get it from Acxiom--which serves a cross-section of companies from Lands' End to Conseco Insurance.

Mother lode. This spring, the company plans to move all this information into a new Net-enabled service--AbiliTec--that helps companies consolidate the information they have collected about customers and, for an extra fee, combine it with details from the data mother lode, called Acxiom Data Network. ADN, in turn, has been integrated into popular Web software programs from the likes of E.piphany, which make Acxiom's information instantly available to many more companies.

Acxiom is hardly alone. Naviant Technologies Inc. processes online product registrations for other companies such as IBM. In the process, it has gathered real-world data on more than 17 million households. Another data broker, HotData, helps its client companies link their in-house customer-tracking software with databases from credit watcher Experian and others.

The data collectors don't see any real harm in all this. John Carter, Acxiom's chief of Target Market Applications, says these indexes are just another tool to help marketers get personal with customers. Airlines, credit-card issuers, and mail-order companies have a long tradition of favoring their ''good'' customers with an arsenal of frequent-flier programs, Platinum cards, and premier rankings.

The good stuff. Indeed, segmentation can make smart business sense. The more a company knows about you, the better chance it has of persuading you to stay loyal when you are courted by a competitor. Levi Strauss & Co. says it has sold 33% more jeans--and hiked repeat-visitor traffic on its Web site by 225%--since it started using personalization and segmentation software. What's more, companies can boost efficiency by treating customers according to a ranking system--one that gives them a level of service in sync with their value to the company. Sanwa Bank estimates that its customer-service operation increased productivity by 14% last year by showering its resources on ''A'' customers, who bring the most value to the bank. At First Union, the use of such software contributed to an 18% increase in service productivity during the first two years that it was used, says Steve Boehm, general manager of First Union Direct. Such gains are especially critical at busy Web commerce sites. ''You can't offer the good stuff to everyone, because it would bring the site to its knees,'' says Richard Rovner, a senior manager at software maker SAS Institute.

Even the pruning of certain customers may be justified by efficiency gains. Before the Internet, Weyerhaueser's Wisconsin door factory had no idea which distributors were costing the company money and which were bringing in value. Now, Net-based software that handles orders is tied to ranking software, which spots distributors Weyerhaueser (WY) can afford to lose. ''The machine is smarter than we are,'' says Lee Kirchman, the plant's marketing chief. Since it implemented these programs, the company has shed roughly half its customers. But it has doubled orders to more than 800,000 doors last year.

The question is, are the results of segmentation equitable and just? It depends on which marketing basket you land in. ''For some people it's fair, and for some it's not,'' says Roger Siboni, CEO of E.piphany (EPNY), a Silicon Valley maker of e-commerce software, including segmenting programs. Marketing expert Donna Hoffman calls it good old-fashioned cherry-picking, the wooing of only the juiciest customers. ''The danger for abuse, for digital redlining, is extremely high,'' says Hoffman, an associate professor at Vanderbilt University in Nashville.

Companies that promote personalization and segmentation say that consumers get products that are appropriate to their tastes and means. But some sociologists argue that Weblining systematically limits the cultural and economic choices presented to different groups. ''There's an anti-democratic nuance to all of this,'' says New York University Sociologist Marshall Blonsky. ''If I am Weblined and judged to be of minimal value, I will never have the products and services channeled to me--or the economic opportunities--that flow to others over the Net.''

It would almost be better if the decisions about who sees what were less narrow. In fact, segmenting decisions spring from businesses' desires to forge more orderly markets. What's more, the mathematical algorithms that funnel individuals and groups of people into narrow categories aren't remotely scientific. They may incorporate the biases and intellectual limitations of the software's designers and users. Not all lower- and middle-income consumers crave a steady diet of action films. But if that's what gets served up at a local theater, that's where much of the traffic will flow.

Sure enough, Hollywood studios are starting to use the Net and sophisticated box-office data-tracking technology to speed up decisions about who gets to see which movies, which stars, and even which plot lines. Twentieth Century Fox relies on a Web-powered database, called Project Eight Ball, to place films in 30,000 theaters. The system purports to tell you which crowds prefer Wesley Snipes to Nicolas Cage.

Scientific or not, high-powered computing increases the incentive for businesses to Webline customers by making human behavior appear predictable. Visa International, for example, is using neural networks to build up elaborate behavioral profiles. Over months, these systems--which emulate the learning power of the brain--track a person's behavior online and off, then match it against models of similar personality and behavior types to predict how people will act in the future. The initial incentive was to recognize and thwart fraud. Now Visa is testing the software with 12 member banks in an effort to anticipate loan defaults. ''This gives us smarter data, and with Web-based technology, we can get that to our member banks in real time,'' says Martin Izenson, a director in Visa's risk management and security group.

Black box. The military developed much of this technology to help radar systems, for example, distinguish missiles from birds. Joel R. Reidenberg, a law professor at Fordham University in New York, thinks the technology has outpaced our ability to assess its performance or its impact. ''Neural networks are a black box,'' he says. In a period of just weeks or months, the value assumptions in the networks evolve and can no longer be analyzed with precision, even by the developers. Think what that says about the micro-profiles that are generated. ''Some of this really crosses the line into offensiveness,'' says Reidenberg.

Customers often appreciate the personalization on a Web site, or in a store. But they bristle when they learn about the rankings that go with it. Last summer, Nob Hill Foods, a supermarket chain in Gilroy, Calif., and Wild Oats Markets Inc., a health-food chain in Boulder, Colo., halted their so-called loyalty card programs, which gave cardholders discounts that others didn't get. Customers were complaining that it smacked of discrimination, and some shoppers were threatening a boycott. ''Mostly, people resented the snob factor,'' says Wild Oats spokesman Dan Hall.

In their zeal to collect more data to feed into their computer systems, some businesses that deal with customers face-to-face are forgetting their manners. Driving home from work last fall, Seattle City Planner Terry Whittman stopped at a West Seattle Pizza Hut to pick up dinner, but before she could pay, a register clerk asked for her name, address, phone number, and pizza-topping preferences. When Whittman declined, the clerk refused to fill her order. No data, no pizza. ''It was absurd,'' says Whittman, who had offered to pay cash. Pizza Hut acknowledges the incident, explaining that the ''surveys'' were aimed at building up a database of online and offline customers. Even so, Pizza Hut admits clerks overstepped company policies.

Stanford computer-science professor Eric S. Roberts says that many of the designers of the new segmenting technology didn't think of the social implications when they invented it. The moral of the story, says Tara Lemmey, president of the Electronic Frontier Foundation: ''Just because technology lets you segment more tightly doesn't mean you should.''

As Weblining evolves, will we end up constructing a more menacing marketplace? Maybe so. It could be that people who are able to pay more will be prodded to do so by segmentation software. But it is far more likely, as competition for high-value consumers increases, that people with the best profiles could also get the best prices and the best service--locking out the also-rans. Small businesses could face similar Weblining. Some of Weyerhaueser's small distributors, for example, are grumbling about the shabby way they've been treated.

It would be one thing if flexibility were built into this system. But your digital profiles and rankings never go away. Nor can you review them for errors. At the click of a mouse, your e-dossier may be available to other businesses that are keen on segmenting you even more precisely. And because the Internet is still evolving, the whole process of sorting customers will get faster, cheaper, and more subtle at every turn. The Net may be advancing at the speed of light, but it's going to start feeling more and more cramped inside the pigeonhole.

By MARCIA STEPANEK
Contributing: Heather Green, Steve Rosenbush, and Susan Zegel in New York and Robert D. Hof in San Mateo