InformationWeek

March 15, 1999

Tapping The Pipeline

continued...page 3 of 3

Related links from our sister publications:

Planet IT Planet IT E-Commerce Technology Center

Computer Reseller News Microsoft Makes Big Pitch For E-Commerce For Small Businesses

CDnow Inc.'s capture and analysis of online buying patterns helped it achieve fourth-quarter results in which 56% of sales came from repeat customers, says VP of technology Mike Krupit. CDnow maintains data on its customers in four Oracle databases. One database logs the pages users view each time they visit, another keeps track of purchase history and shipping information, and a third catalogs preference information gathered by GroupLens from Net Perceptions. An Oracle8 database aggregates data from the other three databases in a data warehouse.

CDnow mines the data warehouse using data-analysis software from Brio Technology Inc. Among other insights, the Brio software shows how many customers respond to promotions by actually buying products. In addition, CDnow uses custom-built tools to predict which customers will buy certain products and which products will sell the most. The company is currently evaluating MineSet, data mining software from Silicon Graphics Inc. that presents data in 3-D images for predicting sales.

Not surprisingly, the growing niche of Web intermediaries sees a potentially lucrative business opportunity from Web data capture and analysis. Instill Corp. in Palo Alto, Calif., which links buyers and sellers in the food-service industry via the Web, last year turned the data it captures from purchasing transactions into a business intelligence service called Advantage.

Using proprietary data-capture tools linked to a Sybase database, Advantage helps restaurant and hotel chains identify patterns in food purchasing that they can analyze for cost savings. This year, Instill expects revenue from Advantage to equal revenue from its initial business of facilitating online transactions.

Online Information Exchange
A business plan for a data product offering is tops on the priority list for Chrome Data Corp. in Oregon City, Ore., whose Chrome.com Web site links auto dealers with institutional car buyers such as corporate vehicle fleet managers. The firm couldn't even consider such a strategy when it routed orders from a client-server, dial-up system because there was no online information exchange.

"The Web lets us gather data [stored in a Microsoft SQL Server database] on what models the users actually spec out," says Chrome Data product manager Forrest Nabors. "We've never been able to capture that before, but the Web is a much more direct link to customers. [Web intermediaries] will have the most valuable data, because what they're doing on the Web is much more sophisticated than just referring names or orders. They're capturing actual customer behavior."

One fast-growing player in this space is MatchLogic Inc. in Westminster, Colo., which was acquired last year by Excite for $89 million. MatchLogic provides integrated marketing services to help advertisers such as Charles Schwab, Dell, General Motors, and Procter & Gamble orchestrate Internet ad campaigns.

MatchLogic maintains several databases of information gathered through Web sites. One database consists of anonymous data about users' computers, collected as visitors click through sites, click on ads, ask for a brochure, or purchase goods. This research and analysis database currently collects data on 160 million Web page impressions every day--a number that Jack Garzella, director of MatchLogic's core systems group, expects to reach 500 million per day by midyear and more than 1 billion per day by year's end.

A second database contains this data in summarized or aggregated form. This is used for reporting purposes, such as judging the effectiveness of a particular ad campaign. A third database contains "self-reported" data, information voluntarily supplied by some 6 million people through Web-site registrations, sweepstakes entry forms, and responses to E-mail campaigns. This 500-Gbyte database is used to build more than 57 million demographic profiles that MatchLogic clients use for advertising campaigns that target, for example, groups with a specific combination of age, geography, and income.

MatchLogic builds custom models for its clients for targeted ad campaigns for lead generation, customer retention, and other marketing chores. "We have the largest profile database that we're aware of," Garzella says.

MatchLogic uses Oracle8.0.5 to store the data and Ardent Software's DataStage extraction, transformation, and loading tool to process and direct the data from collection points to the databases. MatchLogic uses analysis software from SAS Institute Inc., Oracle Discoverer for ad hoc queries, and an internally developed report-generation program. Sun Enterprise 4000 and 6000 servers provide the horsepower. "We're seeing great demand for these services from our larger customers," Garzella says.

That trend will only continue as the Web moves from a sales channel to a strategic and direct link to customers with unprecedented feedback potential.