Finding Pearls In an Ocean of Data

Data analysis can be more crucial than ever in an economic downturn.

After years of gorging on IT, CEOs and boards of directors have indigestion. They're demanding greater returns from their existing systems before buying more. One way to deliver that return is through data analysis and data mining tools, which become even more important as every dollar earned - or saved - becomes more crucial.

1pixclear.gif
1pixclear.gif

Data Analysis Tools

• IntelliVisor, an online analytic service offered by SAS Institute; offered as an ASP service hosted by Electronic Data Systems Corp.

www.sas.com

• Spotfire Decision Site from Spotfire; one of a range of data analysis and visualization tools from the company, designed to graphically display hard-to-find trends and relationships.

www.spotfire.com

• Databeacon's namesake browser-based data analysis tool; designed for ease of use.

www.databeacon.com

• Hyperion Solutions' Essbase is one of the premier online analytical processing database servers; used in conjunction with applications from Hyperion and other vendors.

www.essbase.com

• Insightful's StatServer is an analytic Web server based on the S-Plus programming language.

www.insightful.com

Analytics "makes the difference between an average implementation of e-business or [customer relationship management] and an excellent implementation," says Frank Buytendijk, a senior research analyst at Gartner Inc. in the Netherlands. "It's in analytics where you assess and evaluate the effectiveness of what you're doing."

Buytendijk predicts that the demand for skilled analysts will triple between 2000 and 2004 and that by then, companies will be able to find only half the skilled analysts they need. Given that shortage, Framingham, Mass.-based market researcher IDC predicts that the market for analytic software will grow almost 28% between 1999 and 2004 to more than $6.6 billion.

From semiconductors to pharmaceuticals, companies will be seeking and using new tools with which to better understand their businesses and even to streamline the analytic process itself.

Giving Users the Data They Need

At International Rectifier Corp., an El Segundo, Calif.-based producer of power management semiconductors, manager of financial analytics Doug Burke says Hyperion Solutions Corp.'s Essbase 5.1 has allowed the company to "get a lot more out of our AS/400" midrange system by allowing the company to extract and analyze sales data very inexpensively. Rather than being forced to e-mail huge spreadsheets, which ties up network bandwidth, users can now retrieve dynamically calculated views of just the data they need, he says.

Burke says he expects more cost savings this summer when he deploys Version 6.1, which includes attributes that will allow users to dynamically analyze data across additional dimensions (such as sales area) without having to store those calculations and thus increase the size of the database. "That's a big payoff," he says, "especially when you want to scale this thing up to hundreds of thousands or even millions of products" to analyze.

When catalog retailer Blair Corp. relaunched its Web site in December, "we were looking from Day 1 how to be more profitable," says Jeff Parnell, vice president and general manager of e-commerce. The relaunch not only added many more products to the site but was also the first iteration of the site to be promoted in the millions of catalogs the firm mails each year.

The Warren, Pa.-based company wanted a Web clickstream analysis tool to tell it which areas within the site attracted the most customers and exactly why those browsing did or didn't buy, says Darren Schott, the company's director of marketing for e-commerce.

For the answers, Blair turned to IntelliVisor, an online analytic service offered by SAS Institute Inc. in Cary, N.C., to determine such things as which product categories suffered the most cart abandonment (customers ending their shopping process without completing the order) and which generated the highest conversion rates (turning browsers into buyers.) The application service provider model was attractive, Schott says, because Blair didn't have to develop the needed know-how in-house. SAS was a natural choice of vendor because Blair was a longtime SAS customer and the company could easily integrate its clickstream data into the SAS tools it already had, says Parnell.

Parnell declined to specify how IntelliVisor has helped the bottom line, except to say, "We feel very comfortable with the ROI equation." The payoff, he says, lies in "the fact you can make better decisions and make them more quickly."

Production Push

Just because the economy has slowed doesn't mean your customers have. At Texas Instruments Inc.'s Kilby manufacturing facility in Dallas, Joe Lebowitz, director of yield and product engineering, is pushing TI's equipment and processes harder than ever to make more-advanced components and do it cost-efficiently.

For example, with each increase in a processor's clock speed, TI must control its chip fabrication equipment and processes more tightly if it is to produce enough usable chips per silicon wafer to make a profit. For each wafer, Lebowitz says, instruments may collect upward of 100,000 discrete pieces of data, from the temperature of the wafer to the purity of the air passing over it. Each piece of equipment also generates data about its own performance as often as every millisecond.

"It is absolutely essential" to have tools that can help engineers find the often subtle relationships among the data, says Lebowitz, so they can quickly find and fix production problems.

Lebowitz is evaluating Spotfire Decision Site from Somerville, Mass.-based Spotfire Inc. as one of the tools used at the Kilby facility to gather and analyze such information. For example, he says, when inspectors began finding cache memory failures in some chips, Spotfire helped them discover that one tool was etching lines of a slightly different width than other tools were. Even though the tool was operating within its specifications, it wasn't operating precisely enough to generate as many usable microprocessors per wafer as the others.

Lebowitz couldn't estimate how much Spotfire might save at the Kilby facility but says, "There are probably many cases where it can save you as much as a week's worth of work." In a semiconductor manufacturing environment, saved production time can amount to millions of dollars in revenue, he says.

Before they began using Ottawa-based Databeacon Inc.'s namesake graphical analysis tool, the IT staff at Pfizer Italiana SpA's medical manufacturing facility was bogged down with routine requests for reports from users. One day, for example, a user might ask for an inventory report sorted by the Pfizer division for which the plant made a product, the type of business (such as animal health or pharmaceutical) within Pfizer under which the product falls, and the major product group into which the product is categorized, says business analyst Luca Michella. The following day, the same user might ask for the same information sorted only by major product group and business.

Not only was it time-consuming to pull the data and load it into spreadsheets, Michella says, but day-to-day differences in how he wrote the queries could generate inaccuracies.

With Databeacon, any user with a Web browser can create those reports himself and do it with a Web-based tool that requires no cumbersome updates to the software on his PC. Similarly, the time required to do quarterly analyses of quality control has been reduced from two person-days to minutes, says Michella. Previously, users had to search the manufacturing resource planning system for rejected lot numbers and the reason the lots were rejected. Now, Michella says, users can access Databeacon reports that pinpoint problem areas and allow users to drill down for more detail.

"Professional statisticians and analysts are a fairly scarce resource," says Fred Hulting, a senior research scientist at The Pillsbury Co. in Minneapolis. To shield them from routine report requests, he's rolling out Web-based statistical applications built on the StatServer decision-support tool from Insightful Corp. in Seattle. By building commonly used reporting capabilities into StatServer, users can now generate those reports themselves rather than call in an analyst, he says.

Like others, Hulting couldn't specify the application's benefits but says they have "greatly increased the capacity of my folks to spend more time with the business and focus on the bigger issues." The new applications leverage Pillsbury's existing network infrastructure, require no client software and run on a portion of a server the company already owns.

Pillsbury spent only $40,000 to license StatServer and the underlying S-Plus language used to develop the applications. Hulting and a colleague who was experienced in S-Plus developed the applications without outside assistance, but he warns that others without such experience will need consulting help.

Looking Ahead

International Rectifier is using Essbase not only to cut the costs and time it takes to collect data but also to standardize how the data is created to improve decision-making, says Burke. Using Essbase, the company has built one multidimensional data cube for inventory analysis and will soon roll out another to analyze sales by market sector, he says.

Using these common databases, "you don't have costs defined three different ways or revenue defined four different ways" across divisions, Burke says. "Whether people like the numbers or not, everyone agrees on the numbers" and can focus more on analyzing the data than gathering it, he explains.

As more users access common data through StatServer-based applications, Pillsbury is taking the opportunity to "provide some standardization and best practices" in how data analysis is done, says Hulting.

Gartner's Buytendijk says he thinks companies should centralize analysis functions so analysts can share best practices and find ways to automate routine report-generating tasks that users should do themselves. This gives analysts in the central business-intelligence office the time they need to work on the interesting projects that will keep them from defecting to other companies, he says.

Analytics won't guarantee that your company will make it through the slowdown. But some companies are betting that it can help them not only survive but also thrive when the good times roll again.

Scheier is a freelance writer in Boylston, Mass.

Copyright © 2001 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon