- By Brian Robinson
- Mar 11, 2001
To help employees ex-tract meaningful information from a vast amount of raw data, organizations often turn to business intelligence (BI) applications.
Building a BI system is not the bleeding-edge undertaking it was just a few years ago. Improvements in software tools have automated much of the grunt work, such as data extraction and scrubbing.
Meanwhile, powerful new online analytical processing capabilities provide near-real-time insight into important patterns and relationships that are often buried in mountains of government data. And the growth of the Internet has made it much easier to deliver BI applications to a broader set of users.
This all makes BI a much more appealing proposition for government agencies. NASA, the Defense Department, military and intelligence agencies, the Agriculture Department, the Coast Guard and others are using BI every day to help analyze vital financial, mission-critical and readiness issues.
In addition, BI can save agencies a chunk of money when done properly. The Naval Undersea Warfare Center, for example, has a system serving just 300 to 400 regular on-site users. But for a relatively small initial investment, the center has cut an estimated $1.5 million a year from the costs of a previously paper-based system used to create and distribute time-sensitive reports about finances and operations.
"Just about every agency needs these kinds of decision-support systems, though some may not realize it yet," said Bill Smithson, group vice president for information technology at Materials, Communication and Computers Inc. (MATCOM), Alexandria, Va., a professional services firm that has worked on a number of government BI projects. "All are faced with doing more with less, plus the people they employ are less experienced. Decision-support systems can certainly help them with all of that," he said.
But only if it's done right. Perhaps more than most kinds of IT-related programs, if you build a BI system the wrong way, it's difficult to fix it later. Although each agency has its own particular needs for BI, there is a set of universal principles that organizations can follow in putting together their systems:
Make it about them. Involving users from the start might seem like a no-brainer, but the mantra of all BI-building programs is to make sure users buy into the process. And it goes further than just convincing people to use the system once it's built. They are the ones who really know how an organization runs, so having users involved from the beginning in the design of the system is the key to its eventual success. "It's the most important thing," said Tim Ries, chief information officer at the Air Force Office of Special Investigations, which has been using a BI application for the past three years to help it conduct criminal investigations more efficiently. "We went to agents in the field to ask them what they would like this system to do for them," Ries said. "Establishing those kinds of user requirements is vital to being able to design the paths we use to drill down into the data."
It's also critical that users get an early hands-on experience with the system, Smithson said, because they are the ones who will determine the eventual design of the BI system's user interface. Users may think they know what they want, he said, and the IT department will often design the interface from early discussions with users. "But [users] really only know what they want once the screen is presented to them," he said.
Throughout the BI design process, Smithson said, users should come first.
Never lose focus on the data. This is another piece of advice that sounds self-evident. After all, BI is all about the use and manipulation of data. But the attention on data is often taken only so far, which means that data issues can and do come back to bite people later. "The care we put into selecting the right data sources was the key to the success of our system," said Kelly Croft, assistant director for workforce information at the Office of Personnel Management, whose Web-based FedScope application provides agencies with a variety of data on federal workers. It was officially launched in October and is already getting rave reviews.
"We have around 70 core data elements for each employee, so when we were putting FedScope together, it was obvious we couldn't include all of them for the 1.8 million federal employees that are in our central personnel data file," Croft said. "Plus, a lot of that data is private and can't be released. So we put a lot of thought into what the data choices should be [for the BI application] while making it as useful as possible."
Organizations also don't generally understand the need to maintain the integrity of data as an ongoing process, said Tony Fisher, president and general manager of DataFlux Corp., Cary, N.C., a subsidiary of SAS Institute Inc. There are a "few enlightened organizations out there that promote data stewardship, but they are in the minority," he said. "Unfortunately, a majority can't understand the need."
Make prototypes that count. Understanding the data an organization wants to see is critical to knowing what you want out of a BI system, said Sylvia Ronayne, BI project manager at the Naval Undersea Warfare Center. "We found it was critical to our design process to first build a generic proto-type and then have people start to challenge it," she said. "Then we started to see how we could do certain things better and how to fine-tune the applications. Also, we found that once people could see the actual system, it became much more tangible to them."
But if you do build a prototype, MATCOM's Smithson warned, don't build it as a stand-alone system that could not work in the real world. Prototypes should always be built as part of an eventual production system, he said.
No change, no gain. When people start putting together a BI system, the temptation may be to build it as an online replacement for a particular paper-based system. That could be a big mistake because it would produce a closed-end system that conforms to only a narrow set of requirements. "Flexibility should be inherent in the whole BI process," said OPM's Croft, "because you know, or should know, that once you get the system up, you will gain experience through just running it, and that you will probably have to change it later on."
There's a tendency to assume that the user focus in a BI system won't change that much, but the users the application provides data to now may not be the same users the data is delivered to in the future. And the data that goes into the system will likely change as well.
Also, it's true that the software tools available for BI and decision support are much more powerful and user-friendly than in the past. But don't assume that's all that will be needed. Building a BI solution is still more art than science, and sooner or later you'll have to fall back on the expertise you have in-house to make the inevitable changes.
And one final point: Always assume it will take more time to develop a BI system than you initially anticipate because BI has traditionally been an area of underdeveloped expertise and over-blown expectations.
"Everyone thinks it will take six months or less to put these systems together, but it usually takes a lot longer," Smithson said.
Robinson is a freelance journalist based in Portland, Ore. He can be reached at firstname.lastname@example.org.
Brian Robinson is a freelance writer based in Portland, Ore.