Data miners dig a little deeper
Posted 7/11/2006 10:11 PM ET
By Michelle Kessler and Byron Acohido, USA TODAY
SAN FRANCISCO When customers sign up for a free Hotmail e-mail account from Microsoft, they're Must submit their name, age, gender and ZIP code.
But that's not all the software giant knows about them.
Microsoft takes notice of what time of day they access their inboxes. And it goes to the trouble of finding out how much money folks in their neighborhood earn.
Why? It knows a florist will pay a premium to have a coupon for roses reach males 30-40, earning good wages, who check their e-mail during lunch hour on Valentine's Day.
Microsoft is one of many companies collecting and aggregating data in new ways so sophisticated that many customers may not even realize they're being watched.
These businesses are using new software tools that can record every move a person makes online and combine that information with other data. Brick-and-mortar stores, afraid of being left behind, are ramping up data collection and processing efforts, too, says JupiterResearch analyst Patti Freeman Evans.
The result: Corporate America is creating increasingly detailed portraits of each consumer, whether they're aware of it or not.
Companies say they can be trusted to do so responsibly. Yahoo, for instance, has a strict ban on selling data from its customer registration lists. And Microsoft says it won't purchase an individual's income history just the average income from his or her ZIP code. "We're making sure there's a very bright line in the sand," says spokesman Joe Doran.
Some consumers aren't reassured. Salt Lake City lighting designer Jody Good, 54, goes to great lengths to control his personal information, including signing up for some services with false names and keeping unusually tight security settings on his PC. "I'm trying to preserve my privacy," he says.
Privacy advocates are worried, too. "Think about it: A handful of powerful entities know a tremendous amount of information about you," says Jeff Chester, executive director of the Center for Digital Democracy. "Today they manipulate you into what kind of soap to buy, tomorrow it might be who you should pray for or who you should vote for."
Internet firms are at the vanguard of the trend through a technique called behavioral targeting.
It works like this: Anyone who has registered to use any of Yahoo's free online services can be sure the tech company is paying close attention to everything they do within its network. It will notice, for instance, who uses Yahoo search to find information on SUVs. Why? It wants to sell targeted ads to SUV makers and auto loan brokers that will appear, say, on Yahoo Finance the next time that person checks his or her stocks.
Targeting isn't new. In 2000, online advertiser DoubleClick ignited a public uproar when it announced plans to cross-reference anonymous Web-surfing data with personal data collected by offline data broker Abacus Direct. Congress held hearings, and DoubleClick backed off.
But, as online advertising soars, "We're hitting the tipping point," says Bill Gossman, CEO of Revenue Science.
Merrill Lynch estimates the online ad market will grow 29% this year topping $16 billion. Researcher eMarketer estimates advertisers will spend about $1.2 billion on targeted online advertising this year and more than $2 billion by 2008.
A phalanx of online marketing specialists, including Tacoda, Revenue Science, AlmondNet, advertising.com, DrivePM and Did-it, are pushing it. They're hustling to form overlapping alliances with major media website publishers.
"Targeting has certainly become a large part of what advertisers are interested in," says Yahoo spokeswoman Nissa Anklesaria.
Targeting advocates herald the win-win-win. Publishers can charge more for relevant ads that spur sales and decrease annoyances. Microsoft says it can assist a Dinners-To-Go franchisee zero in on working moms, age 30 to 40, in a given neighborhood, with ads designed to reach them before 10 a.m., when they are likely to be planning the evening meal. "Instead of carpet bombing, it's more of a shotgun approach where you're hoping to hit the targeted customer," say Doran.
But there's one big holdout: search giant Google. Although the company scans customers' Google e-mail accounts in order to send them text ads, it hasn't yet embraced more proactive targeting.
"We're treading very carefully in this space because we put user trust foremost," says Google product manager Richard Holden.
Brick-and-mortar companies are working hard to create similarly rich data sources. Although they can't track a customer's every move, they can create basic profiles of them using loyalty cards.
Loyalty cards are typically given to customers in exchange for personal information. In return, customers get coupon-like discounts when they present their card.
Safeway assisted with kicking off the loyalty-card era in 1998, and was soon followed by rivals such as Albertsons and Kroger. Many programs are run behind the scenes by a little-known Florida firm called Catalina Marketing.
Catalina collects loyalty and bulk sales data from more than 20,000 stores, then uses it to create pictures of shoppers over time, says CEO L. Dick Buell. The picture gets clearer the more data stores collect. For example, Catalina can assist retailers determine that someone lives in an upscale area, buys diapers, and may be interested in high-end baby food.
The data Catalina receives do not contain personal information. Records are identified by an ID number only. But retailers hang on to personal information and can reattach it to records once getting them back from Catalina
That worries privacy advocates, because loyalty cards fairly rare a few years ago are spreading fast. Most large grocery chains except Wal-Mart have them. And they're moving beyond grocery stores, to outlets such as Barnes & Noble bookstores, CVS drugstores, and Exxon and Mobil gas stations.
Mining for data
The flood of new information is assisting to spawn a sister industry: data-mining software. These powerful programs sort through massive databases, looking for patterns that would take a human years to spot. Sales of data-crunching software have jumped more than 30% since 2000 and are expected to keep growing, says tech analyst Dan Vesset with researcher IDC.
"Most large companies are doing it in one area or another," says tech analyst Gareth Herschel with researcher Gartner.
In its most basic form, data mining is simple. A grocery store might put the peanut butter next to the jelly one week, and move it to a different aisle the following week. The store can then run data-mining software on the two weeks' sales receipts to learn which setup sold more peanut butter.
Technology always improving
But far more sophisticated and complex types of mining are emerging. Silicon Valley firm Sigma Dynamics has launched software that can analyze data on the fly, even if it's not stored in neat columns and rows. For example, it can read the typed notes of a customer service agent, compare them with a database of stored records, and see if any phrases match. Then it can instantly pop up a window offering a solution to the customer's problem.
Entrepreneur Jeff Jonas has created software that starts by examining the record of a known entity usually a person. It then compares that record to thousands of others, looking for patterns that might signify a relationship. Jonas designed the software for Las Vegas casinos, which wanted to better know who their customers were in part to keep out cheats. The software could identify relationships that might otherwise go unnoticed, such as the fact that a cheater and a casino employee were roommates.
The Central Intelligence Agency realized that the software could have other applications. Its venture capital arm, In-Q-Tel, funded Jonas' small company in 2001. IBM bought it in 2005 and now sells the program to businesses, including retailers and financial institutions.
Other companies are working to make data mining traditionally a high-tech discipline of statisticians and programmers accessible to average workers. One such firm, San Francisco data-mining software maker KXEN, says there's a huge demand. Data-mining technology "has been used for a long time, but only by a very small number of people," says President Ken Bendix. Although companies have long had useful data, the information was rarely used to its full potential, he says. Now that's starting to change.
But the growth in data mining creates a problem, says Stanford University professor Hector Garcia-Molina. "How can you provide that kind of useful information without violating the privacy of individuals?" he asks. Garcia-Molina is working on computer science tools that will keep databases from extracting too much personal information while slicing and dicing. The work, still in its experimental phase, "is not easy," he admits.
IBM's Jonas proposes a lower-tech solution better disclosure. Most companies do have privacy policies today, but they're generally vague, he says. "I would like to do business with companies who are using my data the way I expect them to," he says. "I want to avoid surprise."
Jeff Barnum, a 45-year-old real estate consultant from Cincinnati, agrees. He avoids filling out many forms and frequently deletes cookies from his PC, yet is willing to share his information with companies he trusts.
Barnum says his clients do the same. When visiting an open house, many people give him false names and other information. But once they get to know him, "They'll tell me anything," he says, laughing.