As manufacturing goes global, use of BI makes for smart business

Crystal balls may be the stuff of myth, but for Tim Rey, leader of the data mining and modeling group at Midland, Mich.-based Dow Chemical, assessing the future is serious business. “We sell into practically every global market,” says Rey. “It pays for us to monitor macroeconomic forces.

By Tim Beyers, contributing editor June 1, 2007

Crystal balls may be the stuff of myth, but for Tim Rey, leader of the data mining and modeling group at Midland, Mich.-based Dow Chemical , assessing the future is serious business.

“We sell into practically every global market,” says Rey. “It pays for us to monitor macroeconomic forces.”

Dow operates 156 sites in 37 nations, resulting in $46 billion in annual revenue. With that much volume, any one of thousands of variables can adversely influence profits. Rey’s job is to use business intelligence (BI) to stack the odds in Dow’s favor.

Rey leads a seven-person flex resource team charged with making BI pervasive in an organization of 42,000 employees. Industry analysts say Dow is not alone in such an aspiration. Increasingly, globally dispersed enterprises are adopting BI tools as a way of detecting potential problems on the horizon.

This escalating thirst for analytics is driving two related phenomena: rapid growth in BI software sales, and consolidation among BI software vendors.

On the sales front, Stamford, Conn.-based Gartner predicts BI, already a $5-billion global market, will enjoy 8.6-percent average annual growth through 2011. “A 2007 survey identified BI as the top priority on the CIO wish list for the second year running,” says Gartner Analyst Dan Sommer.

These numbers seem to have caught the attention of vendors prowling for acquisitions. In February, Oracle announced a $3.3-billion deal for Hyperion, a financial data management specialist. And in May, integration software vendor TIBCO paid $195 million for BI tools supplier Spotfire.

All together now

This rush to consolidate, as analysts see it, is part of a trend toward bringing a broader range of functionality under the BI umbrella. “If you get three analysts and three vendors in the same room, you’ll get six opinions of what BI really is,” says Boris Evelson, who covers BI for Cambridge, Mass.-based Forrester Research .

Nevertheless, common elements abound when it comes to BI. All of the major vendors still supply reporting, for example. Most also offer performance management, dashboards, and data integration. But only a few offer forward-looking analytics, an area that takes on more importance as business operations become more global.

Rey, who spends more than 70 percent of Dow’s BI budget on forward-looking projects, agrees. Dow has more than 4,000 users of JMP analytical software from SAS Institute, which has proven to be about as good as any crystal ball.

Originally known within SAS as “John’s Macintosh Product,” JMP (pronounced jump) runs on Windows, Macintosh, and Linux operating systems. It was among the first to combine the statistical analysis capabilities of a spreadsheet with a graphical interface, allowing results to be displayed visually.

The latest version of JMP goes further by assessing the probability of business outcomes, as well as the ability to visually express the results from various analysis techniques, including regression and neural networks.

Standardizing success

For Rey, the process of developing a BI infrastructure began in 1990 when Dow first adopted SAP as its ERP system. An Oracle-powered data warehouse was added in 1994. Four years later, a BI system with reporting from Business Objects and Cognos would emerge.

By 1999, Dow would adopt Six Sigma to craft management best practices, but that would take the business only “part of the way there.” Better data was needed—something beyond historic reporting. Dow decided on data mining.

Extracting value from the process took two years of painstaking work. Rey credits Six Sigma for keeping the team on track. “We enforced that same discipline on data mining,” he says.

In fact, Dow moved to a standardized IT model that allowed managers to control where, and how, data is distributed. “We are very well structured in terms of processes,” says Rey. There are more than 1,000 tables in Dow’s shared data warehouse. More than 45,000 variables are tracked. Analytical reports are aimed at, among other things, earning better prices for goods and services and detecting fraud.

In late 2005, BI won numerous fans among Dow’s executive ranks. That’s when Dow’s lead manager for purchasing rail contracts became concerned that the company was overpaying to ship its products via rail car. If so, it would be an expensive problem: Railroad shipments cost Dow more than $400 million annually across North America.

Rey’s team began studying the problem but quickly found that the number of variables involved—roughly 10 to 12—would make a classic linear analysis inaccurate. Using JMP, the team created a neural network, a crude form of artificial intelligence trained to understand multiple variables and complex relationships.

The team spent months sifting through five external databases and three internal databases. Once the system was programmed to compare Dow’s payments with industry norms, all the data was merged into a model that could be accessed by JMP.

Results came quickly thereafter. Using JMP, Rey proved Dow was frequently overpaying by at least 20 percent. Today, the purchasing team still uses the software to project fair rates before entering negotiations with any of its railroad partners.

Vertical focus

Dow is just one of many companies demanding comprehensive business intelligence platforms. SAS says 11 percent of its $1.9 billion in 2006 revenue was derived from manufacturing-related BI products.

And the applications are becoming more specialized. SAS, for example, has developed a BI suite called Service Intelligence for tracking what Manufacturing Industry Marketing Manager Michael Newkirk calls the “service chain.”

“If you buy a car, from the moment you drive it off the lot, you are in the service chain,” says Newkirk. “You are going to need maintenance, repair, and upgrades over the life of that vehicle.”

Service Intelligence addresses each of these needs by drawing on six analytic components that run on top of a BI platform and in conjunction with another SAS product: Forecast Server. Some components, such as its Warranty Analysis, offer the sort of predictive intelligence that Dow craves.

“Think of the Firestone-Ford Explorer problem,” Newkirk says. “If they would have been able to detect that problem, what would it have meant in terms of brand protection and cost savings?”

But Forrester’s Evelson cautions against viewing software as a panacea. “One of the main problems with BI is that it’s very complex to build and use. Often, you’ll have to pull together more than 100 components to make a BI platform work.”

Evelson adds that getting reporting and analytics up and running may be only 20 percent of the effort. “Business intelligence is still an art, not a science,” he says.

Rey sympathizes. “Large companies like ours change their structure every two years whether we want to or not,” he says. Such shake-ups can cause practical problems if data ownership falls to individuals, but Dow won’t allow that. Dow’s data model—which prizes systematic, companywide methods for tracking and reporting on data—is practically corporate gospel.

“Sixty percent to eighty percent of the effort with business intelligence is getting data ready,” says Rey. “It’s in the data model. You’re putting up a lot of roadblocks if you don’t have that infrastructure in place.”

Data-driven management

For years, TimeWarner Retail Sales and Marketing was guessing at how many magazines to ship to its distributors. The opportunity for error was huge, as TimeWarner publishes more than 400 titles for distribution in 120,000 stores.

When a wave of consolidation hit the magazine distribution business, only five distributors remained, controlling 90 percent of TimeWarner’s network. Management decided the time was right for BI.

A Web-connected BI system powered by tools from three vendors was in place in just over eight months. The centerpiece of TimeWarner’s BI network is a package called S-PLUS from Insightful .

S-PLUS is a programming environment for building predictive analytic applications. Nick Brown, an Insightful VP, says it functions within any BI platform or data source. At TimeWarner, S-PLUS works in conjunction with Oracle’s Discover BI suite, and reporting tools from Cognos.

Dilip Patel, director of BI and information management, says the tools allow TimeWarner to gauge regional promotions. Reports are supplied regularly via the Web to marketing analysts, who now control all print orders. The result: $3.5 million in savings.

Patel says TimeWarner now runs a better business with better tools.

Dow Chemical’s Rey feels similarly, estimating that the return from implementing generic business intelligence was at least four or five to one.

“In the case of our advanced analytics projects,” concludes Rey, “it was at least ten to one.”


Related Resources