InformationWeek

Most analysts say the data-cleansing market is destined to grow quickly, but it's still relatively modest in size. Aberdeen Group's Creese estimates that market revenue in 1998 came to around $80 million. Besides Trillium and Vality, specialist software providers include First Logic, Group 1 Software, and Innovative Systems. Data cleansing has historically also been offered by services companies such as Acxiom Corp.

One of the biggest opportunities for using this technology, analysts say, is the booming E-commerce market. That's partly because new data warehouses and CRM applications are being built behind E-commerce sites to analyze and predict customer behavior, and they rely on both cleaning up data entered by the customer and, in some cases, identifying a customer's activity among the thousands of users of the site. "If we can't accurately create a consolidated view of the customer, we can't create predictive models for customers that come to the Web site," says Jim Davis, director of product strategy at SAS.

Another emerging trend, English says, is applying data-cleansing technology to defect prevention by operating at the data-entry point. The goal is to prevent quality problems from creeping in at the outset, rather than cleaning up data that's already been gathered and stored. When visitors to a Web site type in their name and address, for instance, the E-commerce application running at the site can use data-cleansing software to check the validity of the address against a database held in-house or at a data-cleansing vendor's site.

Sagent Technology Inc. last month rolled out a service called Centrus, joining companies such as Acxiom and HotData Inc. that provide software linking users' Web sites and other applications in real time to data-cleansing software and databases.

Along with some of these new services is a pricing model that promises to make data-cleansing technology available to a much broader range of users. HotData and Sagent are offering technology on a pay-as-you-go basis to attract smaller, up-and-coming businesses. Sagent's Centrus is priced according to which functions users buy and the volume of data cleansed: Certification that addresses are accurate and formatted to U.S. Postal Service standards costs $5 per 1,000 records for up to 25,000 records. The price falls to $2.90 per 1,000 records for more than 250,000 records. Adding consumer information such as household telephone numbers from information providers such as Experian can cost anywhere from $20 per 1,000 records to several times that amount. Users also get a set of software routines that they can build into their Web applications.

HotData is offering a similar service, called e-luminate, which is priced at $20 a month for up to 1,000 records checked. But HotData's main offering to date has been another low-priced service aimed at sales professionals. A HotData plug-in for Symantec's Act or Vivcott's Maximizer contact-management software can access HotData's data center across the Internet to check the validity of address and other contact information, also adding other business information such as key contacts at the company.

Using the HotData plug-in, salespeople can type the names and addresses of business contacts into their Act databases and click on the "HotData" button. The information is sent to HotData's site and returned, corrected, to the same fields on the salespeople's screens. At the same time, HotData can display information about the contact--such as the type and size of the business for which he or she works--gathered from third-party information providers such as Dun & Bradstreet and Experian.

English says he expects to see more low-priced products and ser-vices enter the market. Aberdeen Group's Creese says products such as HotData's are beginning to address the need to make data-cleansing pro-ducts easier to use. Suppliers "realize they must make the tools less techie" to appeal to a broader audience, he says. "Users just want to push a button."

Peter Avery, national vendor administrator at The Paint Bull, which provides products and training for auto improvements such as painting and dent repair, turned to HotData to solve an expensive problem. Paint Bull would purchase a database of prospective customers, such as auto-body repair shops, then mail them a package including a video and a catalog. Paint Bull relies on the direct-mail campaigns to generate about a third of its business, Avery says. But nearly half the packages were regularly returned as undeliverable--and with postage and contents costing around $10 per package, that rapidly added up to a big headache for the 42-person Saginaw, Mich., company.

Paint Bull decided to use HotData's service, first paying about $300 to clean up an existing database of some 14,000 records, then subscribing to the monthly $20 service. The effect was dramatic, Avery says: Undeliverable mail dropped to around 14% and stayed at that level. While checking addresses, HotData also adds useful business information about the prospect, such as annual sales, number of employees, and key executives.

Avery also uses HotData's service to check prospects that come in via Paint Bull's Web site, which now accounts for around 40% of the company's business. With Web prospects, he says, "The biggest problem is that they make mistakes when entering information."

That's one reason there's likely to be a bigger focus in E-commerce on catching errors as they're entered by including data cleansing in the initial validation of data entered at a Web site. Analysts and suppliers point out that the deeper inaccurate data gets into a company's systems, the more it costs to fix. "If direct mail bounces back or there are product shipping problems, you've got [extra] shipping costs, and lost customer satisfaction," says Johnny Anderson, HotData's president and CEO. In contrast, "if you can catch errors at the customer touch point, at the Web site, the cost to fix them is nothing."

Not everyone describes the data-quality problem in quite such dramatic terms as English, who argues that there will be a revolution in data quality akin to the quality revolution in manufacturing during the 1970s and 1980s--and that businesses that ignore the need will lose out, just as Detroit's automakers lost out to Japanese manufacturers.

But most analysts agree that data quality may have become such a pressing issue that it's time to reassess system design to ensure that inaccurate data doesn't get into systems in the first place. Says English: "Data quality is something you design in, not something you inspect out."

return to page 1