Talk about chaos. For decades, corporate executives have made strategic business decisions based on information deduced from multiple reports that IT compiled by summarizing sets of frequently conflicting data.
Business intelligence systems promise to change that by, among other things, pulling data from all internal systems plus external sources to present a single version of the truth. This truth can then be delivered to decision-makers in the form of answers to highly strategic questions such as "How many customers spend more than $1 million with our company?" or "Which clients are most likely to respond to our latest marketing campaign?"
By 2005, market research firm IDC projects the worldwide market for business intelligence software will total about $6 billion -- up from $2.5 billion this year -- signaling a voracious user appetite for the truth. But getting there isn't just about buying and deploying new software tools. Instead, it involves an arduous data-modeling effort upfront, and that can trigger widespread political battles and numerous other challenges. Here's a rundown of not-so-predictable gotchas and how some veteran users of business intelligence systems overcame them.
1. Rarely, if ever, is the truth in plain sight. Truly valuable business information must be mined from disparate and "dirty" data that resides in multiple, incompatible computer applications and databases. Fair warning: It takes many months and requires paying militant attention to detail to combine and prepare the source data that will ultimately get you to a single version of the truth.
Pfizer needed to pull data from 14 systems, each of which handles a unique slice of the business, to come up with a comprehensive financial picture of the $32 billion pharmaceutical giant. In what turned out to be a false start, Pfizer's IT group began by creating software interfaces to link the 14 individual systems to a data warehouse, which was conceived as a single source of financial information. The problem was that business and financial terms were defined differently within each of the contributing systems.
"You put all the data together and you've won the battle, but lost the war," notes Danny Siegel, senior manager of business technology at Pfizer in the US. "It took us four months just to integrate a few sources." Lesson learned: Gathering the data in a single place is only half the battle. "We saw that we had to put in place some rigorous data standards. This kicked off a six-month, totally nontechnical effort to devise a set of standards that allows users to slice and dice data in whatever context they need it," Siegel says.
Tony LoFrumento, executive director of CRM at $19 billion Morgan Stanley, which created a database that provides a holistic view of each of the firm's 5 million accounts, advises others to pay more attention to information content than to its delivery.
"People talk about business intelligence and getting to one version of the truth, and what they usually think about first is the reporting tools," says LoFrumento. "The reporting tools are great, but without the right information and data infrastructure in place, you're just spitting back what users already know, in a nicer format," he says.
2. The truth is a moving target. Defining standard business and financial terms isn't a once-and-done undertaking. Asking the right questions is also an ongoing process. "There really never is an end to the process," says Rob Cox, director of financial planning at Baltimore-based Erickson Retirement Communities, the country's largest developer of retirement housing, with 10,000 residents and 5,000 employees.
Cox's team took months to come up with standard definitions and reports as part of Erickson's implementation of analytical software from Hyperion Solutions. But it wasn't long after the software was deployed that the balking began. "People came out of the woodwork to say they needed different definitions," Cox recalls. One department wanted to include depreciation in the net income calculation, while another department didn't.
In another case, top management asked to have general and administrative costs broken out from the lines of business in which they were reported and instead have them recorded in a stand-alone manner.
Rather than create a bunch of ad hoc reports to respond to every request, Cox set up a process under which all change requests are funneled to a committee in the financial group, which reviews and then approves or denies them. Changes are then made to the central database so that the new types of reports can be produced.
Even though the numbers may be broken out differently for different departments, they still roll up to a single bottom-line number for the entire company, Cox says. "It's more work, but it's always better to take the time and change things in the database. The truth is a moving target, and you've got to be willing to stop the presses and do the work to change the database, rather than allow every user to make overriding adjustments on their reports," he adds.
At AXA Financial, a dedicated data management team within IT is responsible for what Marvin Rafe, IT director and chief architect, describes as "data stewardship." Whenever there are discrepancies across various lines of business, this IT team takes the issues to a cross-departmental data stewardship team for resolution.
3. Getting to one version of the truth is a highly politicized process.There are two schools of thought on how to get all parties on the same page. The first is to have a strong executive sponsor who has the clout to enforce data standards and the reporting of agreed-upon key performance metrics.
The second, adopted successfully at $2.1 billion automotive retailer Pep Boys, is to appeal directly to end users, such as salespeople and managers in the field "who are starving for data," according to Bob Berckman, director of database services at the 595-store chain based in Philadelphia.
"Our original approach was to go to senior management, and they liked the concept of one version of the truth, but they couldn't focus on it long enough, and after a year, it just wasn't taking hold," Berckman recalls. "Also, the senior guys were being hand-fed the information they needed, so there was no need for them to go and get it."
So Berckman switched gears to a grass-roots approach, hitting up regional managers and individual store managers for support for the now 6-year-old business intelligence system.
"If they could get accurate information in a timely manner, it made their jobs easier," so their incentive to make the project work was greater, Berckman points out.
Before business intelligence systems, Pep Boys managers needed to call the store managers in their regions for financial information. Each store reported differently, plus they reported at different times, making an accurate, comprehensive regional sales report nearly impossible to put together. Now, all stores report data the same way at the same time. "Everyone is looking at the same numbers. There is no confusion. We have one version of the truth," says Berckman.
4. The whole truth is best delivered in chunks and disclosed at prescribed times."A few years ago, we started with wanting all data as real-time as possible," recalls Scott Hicar, CIO at disk drive manufacturer Maxtor. "But we learned that you really only want to publish information at the rate it can be understood and used in decision-making. To have a new version of the truth every 15 minutes doesn't help anybody," he says. Maxtor had a major snafu with managers reading real-time updates of financial and operational reports at different times of the day.
Of course, the numbers varied, representing multiple versions of the truth, which created chaos, not clarity. The solution was to publish updates at specified times throughout the day.
Finally, IT shouldn't make any assumptions about the kinds of information business users will want next -- another misstep Maxtor made early on in its business intelligence project.
"Once we saw where the business was going and the kinds of questions they were asking, we figured they'd want to see more detailed stuff, and we started building data marts based on that," Hicar recalls. "We tried to . . . get them to higher-value information faster, but we learned that IT can't give them the truth." In other words, IT tried to rush the delivery of complex information before the users were ready. They needed time to absorb business intelligence, one step at a time.
What users want
At Landstar System, a transportation logistics company, designated power users from the business provide ongoing help desk support to other users when business intelligence problems arise. The reason is simple, says Bob Leo, director of data management and administration at Landstar. "For any BI effort to be worth it, you have to have the business community take ownership of the process."
On the IT side, technology professionals are responsible for understanding exactly what business users require from a business intelligence system. This means understanding the kinds of business questions they need to answer and the kinds of data that will generate that information.
"Some of our challenge in IT has been being able to speak to users in business terms, which is very key to the success of a business intelligence project," says Leo. "We in IT can talk about dimensions and measures and columns and building cubes, but we've also had to learn how to adapt our terminology in the requirements-gathering process to talk about what the users do every day," he says.