As buzzwords go, it isn’t a stroke of genius: data paired with warehouse unites two of the most mundane words imaginable. Data pales before the far sexier information, and as for warehouse, well, one need only think of the final scene in Raiders of the Lost Ark, in which a treasure of incalculable worth disappears for eternity in just such a place. Add to that the facts that data warehousing is a mature technology (translation: its 15 minutes of buzz are over) and that it’s among the most expensive and complicated technology initiatives most companies tackle, and you might conclude that the only wise course is to run from it as fast as possible.
But think again, because data warehousing has several factors in its favor. For one, companies have become smarter about how to deploy it. Costs are still high and it will never qualify as a weekend project, but there are ways to ease the pain and get a faster return on investment. Also, while customer relationship management (CRM) and business analytics are red hot, they aren’t replacing data warehousing so much as highlighting what a vital part of IT infrastructure it has become. And companies are finding that data warehousing is applicable to any number of strategic efforts, from providing a unified view of the customer to sorting through terabytes of operational data in search of business insights that can boost profitability.
“Data warehousing has always plodded along quietly,” says Lynne Harvey, an analyst at Patricia Seybold Group, a Boston- based consulting firm. “It’s hard to understand–even experts have a tough time explaining it–but as companies try to cope with more and more information, all roads lead to data warehousing.”
A data warehouse differs from a typical database in that it usually extracts information from a number of sources–perhaps including external sources, such as data banks of competitor information–integrates it, and makes it accessible for analysis by nonspecialist employees, even senior management. That integration, however, is also the most difficult part of constructing a data warehouse. IDC, a market intelligence firm in Framingham, Massachusetts, surveyed 1,000 companies earlier this year and found that data integration was the biggest challenge to building a data warehouse, accounting for 70 percent or more of the effort to build one.
Such work can be expensive. Lou Agosta, director of research in data warehousing for Giga Information Group, in Cambridge, Massachusetts, says that while some projects may cost only in the low six figures, $1 million still qualifies as a small effort, and price tags above $10 million are not uncommon. But there’s good news, too, he adds. “Businesses have spent the last two decades trying to manage corporate data as an asset,” says Agosta, “and they now understand how to build data warehouses.”
Fedex: Overnight Success
A case in point is FedEx Corp. Two years ago, the Memphis-based cargo king surveyed about 200 of its business analysts and found they were spending far too much time simply trying to get their hands on important data, leaving less time to analyze it. One problem was the company’s structure: a corporate financial planning group provided support to 10 divisions, all of which shared a mainframe but relied on customized planning and reporting systems focused on a subset of data. Any efforts to take a macro view required lots of manual integration. As Joel Halvorson, FedEx’s manager of financial systems, explains, “If we were working on the business plan and needed information from a certain system, we’d have to stop and wait while someone sent us the data through a mainframe file so we could pull it into a consolidated view.”
The company’s analysts were charged with helping to boost profitability by studying operational data and looking for ways to save or capitalize on opportunities. To do that, they might need to reach into one system that tracks revenue, another that tracks pickup and delivery data, a third that tracks employee pay, and so on. That was not only time- consuming, but also confusing, because units might define operational metrics differently. Cost per package, for example, varied depending on whose system was being accessed.
So FedEx decided to build a data warehouse as a backbone for its business intelligence (BI) efforts. The project began in late 1998, although Halvorson says the real work didn’t start until February 1999, when Chicago-based ThinkFast Consulting Inc. was brought in to help guide the effort. That means the system took only about a year to begin producing results–impressive considering its scope.
The company identified 16 key sources of data to feed the warehouse hub. “The mainframe is still critical,” Halvorson says; every time a FedEx driver scans a package, that data goes straight to the mainframe. But now when business analysts use that data, they’re drawing it from the data warehouse.
And they’re tapping it through the company’s intranet, a key component of the overall design. “Our global reach made that element essential,” says Cathy Ross, vice president of corporate financial planning. But if giving employees around the globe easy access to the system was important, at the same time so was restricting the warehouse effort, at least initially. “We wanted to avoid ‘scope creep,'” says Halvorson, “so initially we stayed focused on a system that was tailored for finance.” FedEx did a fair amount of internal marketing, letting analysts know what was in the pipeline and soliciting feedback from early adopters. The company also focused heavily on training, and phased in the project, believing that demonstrating limited success early was better than springing a wildly ambitious system on users down the road.
To date, the data warehouse has been a success in several ways. Not only has the company made good on its pledge to slash the time analysts spend gathering data, but because the warehouse and related software are so easy to use, many groups beyond the finance department are clamoring for access as well. And the claim made by so many analysts–that data warehouses provide an essential element of IT infrastructure–has also proven to be true. FedEx has undergone a major reorganization this year, a circumstance that can often derail large-scale IT projects. But in this case, the warehouse effort not only stayed on track, it also helped facilitate the changes.
“A centralized, Web- accessible warehouse let us provide information to employees even as they were reassigned and regrouped,” explains Halvorson. “If we hadn’t had this in place, supporting those changes would have been much more complicated.”
One Bite At A Time
Another company reaping multiple benefits from a data warehouse is PHH Vehicle Management Services, a subsidiary of Avis Group Holdings Inc., in Hunt Valley, Maryland, which manages almost 1 million vehicles for companies around the world. Providing information on everything from fuel economy to accident prevention is a vital part of the company’s value to its clients, and as far back as 1995 PHH saw that a data warehouse would be a strategic asset.
“We began early, before there were many reports of failure to discourage us,” quips Mickey Lutz, the company’s senior vice president of IT. Unlike FedEx, which built a data warehouse purely for internal use, PHH wanted to let customers reach into the warehouse and extract the information they needed. But getting there was, in the words of data warehouse manager Jeff Wolf, “like eating an elephant one bite at a time.”
To avoid choking, PHH “time-boxed” the project over a span of three and a half years, rolling out release after release every four to nine months on a carefully controlled schedule. Each release included access to additional information sources and expanded the potential user base, so that what began as a warehouse accessible to a few hundred internal users has now grown into PHH InterActive, a Web-based BI tool that is used by more than 1,000 PHH employees and 7,000 customers.
With PHH InterActive, clients use a Web browser to query the warehouse, pulling out information on everything from vehicle maintenance to billing. To build the system, PHH had to integrate data from 15 different sources, from a mainframe to departmental Unix servers. The company began with core data pertaining to the vehicles it manages and branched out to include a number of other sources, from safety records to competitor information. “You have to stay focused,” Lutz says. “These projects get bogged down because everyone wants to put every source of data imaginable into one.” PHH, however, avoided this tendency to pile on. “When new requirements came in,” Lutz says, “others were dropped, but we kept the number of people involved and the date of each delivery rock solid.”
As for the price of the project, PHH will say only that it has cost “in the millions of dollars,” but also maintains that the price is only half what a “typical” data warehouse costs. The price was kept relatively low, Lutz says, not only through his team’s “ruthless” time- boxing approach, but also by working very closely with the businesspeople who are the warehouse’s primary users. Mary Stone, director of business analysis for PHH, says her group worked hard to spell out exactly what it wanted. “A solid partnership between technology and business staffs,” she says, “leads to good design decisions up front, which keeps the effort moving along.”
In fact, it may never stop. “I don’t see how the warehouse can ever be finished,” Stone adds, “because the business is never finished.” While most companies that have a data warehouse in place–32 percent of large companies and 23 percent of midsize companies, according to IDC–say the effort is continuous, particularly for companies such as PHH, which offer the data to customers as a form of value- added service. “Our clients want intelligence from us,” Stone says, “not simply a lowest-cost provider.” In fact, clients now act as a source of information, passing data on PHH’s competitors back to the system so they can view multiple suppliers.
PHH was acquired by Avis last year, and the parent company is already leveraging the data warehouse technology across several of its holdings. The project has made such an impact that the technology team is now routinely included in sales calls. “We provide a way for clients to manage what is, after all, a huge expense,” Lutz says, leaving little doubt that the data warehouse is providing a substantial return on investment.
Rolling Up The Sleeves
Given such success, why aren’t more companies embracing data warehousing? For one thing, return on investment (ROI) takes time. Even a project designed to show limited return in six months may require far longer to reach full flower, and that assumes all goes well. “Because a data warehouse requires persistence,” says Brenda Moncla, director of the data warehouse practice at ThinkFast, “it often fails to find the sponsorship it needs.”
Successful efforts also require involvement. “It’s never going to be a case of break-open-the- shrink-wrap,” says Giga analyst Agosta. “While you can build one without writing a single line of code, integrating all those sources of data requires rolling up your shirtsleeves and working with people who really know the business.” Even companies that use consultants to do the bulk of the work need to stay heavily involved. Halvorson says one key to FedEx’s successful implementation was that “we knew exactly where all the data that we needed was, and didn’t have to spend time on a complicated assessment phase.”
Data warehousing also faces competition from newer trends, notably CRM and BI. Many vendors in this space promise that their products will give companies a complete view of the customer without the need for additional infrastructure, but Agosta predicts that the current boom in CRM and BI will ultimately be good for data warehousing. “The dirty little secret behind CRM,” he says, “is that it presumes a unified view of the customer. The IT department is stuck having to facilitate that, and only data warehousing can do it.”
Kay Hammer, president and CEO of ETI Inc., a software firm that specializes in data integration, says E-commerce creates a need for a new type of data warehouse, one that tracks off-line inventory. “You need to be able to see inventory that doesn’t belong to you,” says Hammer. “It’s key to E-tail fulfillment, JIT, and other initiatives.”
Some vendors agree. “Ultimately, CRM has to sit on top of a customer-centric database,” says Peter Heffring, president of NCR CRM Division. “You need vast amounts of customer and prospects data, details of every transaction, two to five years’ worth of history on each customer. The only answer is a data warehouse.”
NCR has a long history in the data warehouse market, especially in the high end, where solutions based on its Teradata database are optimized to churn through data to facilitate decision-support applications. Given that, it’s not surprising that the company’s CRM division would advocate a data warehouse, but Heffring is not just a pitchman for the company’s top-of-the-line offerings. “You do get value from individual CRM applications,” he says, “and in fact that’s how almost all companies proceed–there’s virtually no true enterprisewide CRM being done yet, underpinned by a data warehouse.” And a high-end suite of CRM applications can run $3 million, so it’s no surprise that few companies are rushing to support that with a multimillion-dollar data warehouse. But as companies collect more data and move toward real-time customer analysis, says Heffring, a data warehouse becomes crucial.
Ultimately, say proponents of data warehousing, better access to information will provide unforeseen benefits. PHH, for example, didn’t originally anticipate how valuable its data warehouse would be as a sales tool. FedEx didn’t know that once its financial analysts had easier access to data its engineering staff would be lining up, eager to sort through the data for its own purposes. David Pugliese, a systems analyst at Alvion Technologies, a Web applications service provider for the mailing-list industry, in Cape Coral, Florida, maintains that “anytime a closed store of information can be made accessible to a large audience that clearly wants it, it’s worth it.”
That’s not an argument that will convince everyone, however, despite evidence that data warehousing can yield results more quickly and for less money than in years past. But for those who believe that corporate data is an underleveraged asset, breaking ground on a data warehouse may be the right move.
Keep It Clean
While data warehousing is usually positioned for its strategic value, it has other, more mundane benefits. By consolidating information about customers in a central location, for example, it’s easier to “cleanse” the data, finding duplicate or outdated entries. This is particularly true for direct marketers, who were among the first to embrace data warehousing. “Given that 16 percent of U.S. households move every year,” says Peter Heffring, president of NCR CRM Division, “companies can save as much as $50 million a year in mailing costs by weeding out incorrect or duplicate addresses from their customer records.”
The same principle holds on the B2B side. “Data decays [for example, a company changes its address or area code] at the rate of 2 to 3 percent a month,” says Sandy Stoker, vice president of information marketing for Dun & Bradstreet Corp. And according to D&B, 20 percent of U.S. companies do business under more than one name, while almost as many use a mailing address that differs from their physical location. So “data rationalization,” or the scouring of a database for conflicting, duplicate, or incomplete information, can boost efficiency. Because a data warehouse typically pulls data from multiple locations, it becomes the logical nexus for any cleansing or rationalization efforts. (Indeed, a warehouse containing less-than-pristine data is a seriously flawed tool.)
A number of software products for rationalizing data are available, and firms can also buy data that has undergone this process. D&B, for instance, sells records on 60 million businesses worldwide, which it maintains in a vast data warehouse it calls WorldBase.
Piece of Cake
There’s no doubt that data warehousing, given its expense, is an IT project for large and midsize companies. But smaller companies shouldn’t lose heart. The Eli’s Cheesecake Co., for example, has created a data mart (a more limited form of data warehouse) for very little money, and without needing much in the way of technical expertise. In fact, it was the company’s CFO, Marc Zawicki, who led the way.
Before joining the 210- employee company in November 1998, Zawicki had had some experience with OLAP, or online analytical processing, a software architecture that lets businesspeople query databases without needing to know much about their underlying structure. When he joined the Chicago-based company, Zawicki found that the lone employee who knew how to query the Microsoft Access database was swamped with requests for reports, creating a logjam. So Zawicki recommended a move to OLAP- based tools from Cognos Inc. “Now we have four or five power users of the system,” he says, “and even our president uses it to spot trends and get other key information.”
Data from the company’s ERP system is fed into the Access database, along with spreadsheet data and information from other sources. The result: it works like a data warehouse, with a price tag in the thousands of dollars, not millions.
Of course, there’s plenty it can’t do. Data from Web orders, for example, now sits in a separate system, and a CRM tool used by the sales force constitutes another source of as-yet-nonintegrated data. But even in its rudimentary form, the system shows that by bringing information from several sources together and making it available to whichever employees may benefit, companies can act more nimbly. “Now people can point-and-click and get key metrics on our performance,” Zawicki says. “That means decisions are smarter and faster, which can only be good.”