InformationWeek.com

Welcome Guest. | Log In| Register | Membership Benefits

March 27, 2000

Business Intelligence Pays Dividends
Companies and governments use software to reduce losses due to fraud

By Talia Baron

Illustration by Hogan/Braun

Related links:

Telcos Turn To Analytical Tools To Stay In Touch (3/13/00)

The Drive To Analyze (2/14/00)

TechEncyclopedia

Need a definition of a technology term? Look it up here:

Send Us Your Feedback

ach year in California, as many as 500,000 residents fail to file their state income taxes. As a result, the California Franchise Tax Board is forced to tackle a tough problem: collecting the money. But recently implemented business-intelligence software from IBM is helping California tax authorities collect an estimated $36 million annually from residents who haven't filed their returns.

The use of business-intelligence software--which includes user query and reporting tools, online analytical processing, executive information systems, data mining, and packaged data marts--is growing among companies and government agencies that need to understand the stores of information contained in their data warehouses, then ferret out fraud and other criminal activities that can cost millions of dollars a year. According to Dataquest, the business-intelligence software market is expected to grow from $2.8 billion in 1999 to $3.7 billion this year.

By using OLAP and data mining software, the California tax board can analyze tax returns stored in its data warehouse and identify residents who haven't filed their taxes. The system lets the tax board check historical and third-party data, such as federal W-2 forms, to identify residents who haven't filed the necessary returns, as well as residents who may have filed in previous years, but not in 1999. Checking against historical data already in the system, the tax board can pick up on discrepancies and investigate for possible fraud. In 1998, for example, the system identified 37,000 new nonfilers by examining W-2, 1099, and 1098 tax forms; by cross-checking against these forms, the tax board tracked the residents' actual income against what they had reported.

"The new system is critical to our operational efficiency and our ability to serve our customers, the taxpayers," says Patrick Hill, public affairs spokesman with the California Franchise Tax Board in Sacramento. He notes that by catching those who don't pay, the board is chipping away at the "free-rider" problem in California--and as a result, people who don't pay their taxes won't coast on the coattails of those who do. "The system lets us compare data from other state agencies, financial institutions and banks, and employers against what income the taxpayer is actually reporting," Hill says. "If there is a discrepancy, the system will send out an automated notice requesting additional information from the taxpayer." It's then up to the taxpayer to contact the tax board, clear up the matter, and pay any overdue taxes. If that doesn't happen, the tax board will escalate the matter and may garnish the person's wages to collect the outstanding debt.

The tax board's system, which will be implemented by December 2001, consists of an IBM RS/6000 SP Server, DB2 Universal Database, and IBM's DB2 OLAP Server. The system cost about $29 million. Implementation began in 1998--and the system has already paid for itself in reclaimed revenue.

The keys to any business-intelligence effort are data mining tools, which can help companies reduce fraud, anticipate resource demand, increase customer acquisition, and curb customer attrition. Data mining is the process of selecting, exploring, and modeling large amounts of data to uncover previously unknown patterns, associations, and trends within and between rows and columns of information. So while OLAP tools may tell a user what has already happened and why, data mining tools can alert a user to what might happen in the future.

In practice, data mining has been around since the 1980s. But it's only recently that a broad range of companies have started investing heavily in the technology, according to Mark Smith, program director for the application delivery strategy practice at Meta Group. "More companies now have basic information supply chains and data warehouse architectures in place, but these tools don't provide a lot of information about the data," he says. That means companies are looking to advanced technologies to turn their data into real assets--and data mining tools are crucial to that effort, Smith says.

Many companies are realizing that they need to understand their core data better in order to improve their business operations. For that reason, vendors such as Cognos, Hyperion Solutions, IBM, Information Builders, and Oracle are thriving in the business-intelligence software market.

Telecommunications companies, stock exchanges, credit-card and insurance companies, banks, and securities firms all rely on internally developed or prepackaged applications to detect fraudulent use of their services. Visa International Inc., for example, uses a fraud-detection system based on an array of MicroStrategy Inc.'s business-intelligence software: MicroStrategy Web, MicroStrategy Administrator, and MicroStrategy Intelligence Server. In 1999, the system helped Visa contain its losses due to fraud to approximately $1.4 billion. That same year, Visa processed 35 billion transactions worldwide, for a total of $1.4 trillion in global sales.

The MicroStrategy system lets about 150 risk-management analysts identify, analyze, and monitor fraud and its concentrations in various countries, says Abani Heller, program manager for international risk management and security at Visa's Foster City, Calif., office.

"We've had a very robust fraud-detection system from the beginning," he says, explaining that Visa used to rely on homegrown applications running on a legacy mainframe system. "The problem was that a user running one query would have to wait an hour for only one answer. And the old system didn't provide flexibility, an intuitive interface, or the ability to do multiple analyses."

continued...page 2