Drill Down

The first time I got a call from my credit card company's consumer frauddivision, I got very agitated. I checked my wallet to make sure I hadn't lost my creditcard and then called them right back. I was told that my card had been used in an unusualpattern of transactions. Okay, I thought, I had just completed a complicated day that hadstarted in Pittsburgh and ended in Cape Cod with stops in Baltimore and Boston. And I hadcharged something in each place.

But, the second time I got a call from the credit card company's fraud division, I hadnot been travelling at all. I had used the card in what I thought was a regular way tomake a fairly routine ­ if one large ­ purchase. When I asked, I was told that mypattern of purchases matched a pattern commonly used when a card has been stolen (I am notgoing to tell you what that pattern is, but I hadn't stolen anything).

Of course, the credit card company generated the information that triggered the callthrough data warehouse and data mining applications. Information floods into centralcomputers from millions of point-of-sale terminals across the world and then matched topatterns of interest to the company. Detecting fraud is one pattern. But many others areavailable.

Patterns in Profit

As data warehouse applications become more common and sophisticated in retailing, theissue of privacy is going to become more hotly contested. For example, when most peopleshop at a supermarket, they are vaguely aware that as the price information is scanned, itis also stored in a database. But, they may not know that if they are members of thesupermarket's "loyalty" program, grocers and others can develop a picture oftheir buying habits. But they can. Over time, specific shoppers can be offered specificdiscounts based on their purchasing patterns.

The leader in the field is Catalina Marketing Corp., based in St. Petersburg, Fla.,which specializes in providing point-of-sale discount coupons, as well as other servicesbased on detailed customer and purchasing information, particularly to leading supermarketchains and others.

The company has developed a $3-million data warehouse that it can use to createprofiles of the buying patterns of individuals. The data warehouse was built on a Compaq8400 SMP server running a 440 series Alpha chip. The company has accumulated close tothree terabytes of data in Red Brick Warehouse 5.1 database from Red Brick Systems (nowpart of Informix). The database itself could grow to 10 terabytes of data. The company canadd direct access storage in 200-gigabyte increments from EMC Corps and MTI Technologies.

The warehouse stores shopping patterns of nearly 150 million people each week from morethan 11,000 supermarkets. There are close to 30 billion rows of data representing 65 weeksof buying information structured using a star schema. The star schema approach acceleratesthe response to the queries. There is in excess of 84,000 queries a week.

Frankly, there are good business reasons for the system. According to industrystatistics, coupon redemption, one of Catalina's core services, varies greatly accordingto medium. The redemption rate for newspapers and magazine is about 0.6 percent; fordirect mail it is 4.3 percent. Catalina Marketing claims a redemption rate of 8.9 percent.

As for the larger picture, NCR Corp., which counts five of the world's six largestretailers, including Wal-Mart, among its data warehouse customers, recently studied 20retail NCR customers. The company found that they experienced a return on investmentranging from 300 to 1000 percent. And these are large systems. For example, Sears, Roebuckinstalled a 1.7 terabyte, NCR 3600 system, the third largest in the company's history. Theset up includes 10 application processors, 44 parser engines and 864 disk storage units.

While some people couldn't care less if anybody knows if they use Crest or Aqua Freshtoothpaste, or if they brush their teeth at all (the Sears system does not trackindividual buying patterns), others will care very much.

Some people also object to the way the key personal information about individualcustomers is obtained. Basically, supermarkets pay for it by giving discounts throughtheir loyalty programs. Many critics argue that consumers do not fully realize thetrade-off they are making. They just want the discounts.

While loyalty cards are extremely popular, the industry knows it has a potentialproblem. When Catalina Marketing launched a new program to track shoppers over time itlaunched a full-bore public relations campaign to minimize coverage of the privacy issue.

That certainly is one approach to the issue. Another is to try to engage in anabove-board public discussion of this sensitive area.

Dr. Eliot King is an Assistant Professor of Communications and Director of the New Media Center at Loyola College in Maryland. He can be reached at (410) 356-3943, or by e-mail at eking@loyolanet.campus.mci.net.