Total knowledge management:

The day NASA's new Web portal went up was the same day the Space Shuttle Columbia went down.

The agency's Web designers and information managers had been looking forward to Feb. 1 with a degree of nervous anticipation. A months-long re-engineering of its publishing processes and Web systems was to debut that day for an expected 140,000 visitors, a typical amount of Web traffic at the popular agency.

Instead, NASA's Web site registered 75 million hits the first day.

Normally in such an emergency the agency would have pulled its home page off-line and posted an emergency alert page. Instead, it worked with its hosting contractor AT&T and dynamically redesigned the site. Flash special effects were stripped away and information about Columbia — from official announcements to background information — was posted on the fly.

"We have only a few hours of data on what our traffic was before [the accident], but it's never gone back to that level," said Jeanne Holm, chief knowledge architect of NASA's Jet Propulsion Lab in Pasadena, Calif. "People continued to stay interested in a very large way — we had half a billion hits in the first month."

Holm said the feat was possible only because NASA's internal publishing architecture had been overhauled with the end user in mind, using the principles of knowledge management. "We already had in place the people, processes and technology to handle the crisis," she said.

The old site reflected the traditional view of how information was supposed to move within the agency. Mainly, it flowed around the big scientific fiefdoms — space science, earth science and astrobiology. "That was good for us," Holm said. "But it didn't do our audience — which is the public and the media, kids and academia — much good."

The new architecture turned the process upside-down. The portal now is more audience-focused, she said, reflecting the repositioning of NASA as an integrated organization instead of a collection of quasi-independent scientific centers. Information is now designed to flow more symphonically, in the manner of collaboration, information sharing and community building.

These principles — together with tools and techniques such as information classification, communities of practice, instant messaging, expertise locators, libraries of knowledge and advanced search — have come to define the best practices of knowledge management today, experts say. What sets the approach apart from earlier theories of information management is its broad, enterprisewide view and its embedding of tools directly into the business behavior of the organization.

"I see knowledge management as more of the overarching architecture, the integration of tools and services," Holm said. "Not just information technology, but the services behind it and the processes that you use. It's the framework that enables people to make more effective decisions.

First Things First

Of course, the degree to which an agency's data might be described as a junkyard is related to the quality of its classification schemes. The worse the tagging or metadata, the more junk. Data classification — knowing and identifying explicit agency information — is one of the first principles of knowledge management. Yet it has never been more challenging, experts say, especially in today's world where the notion of teradata has become almost quaint.

But new approaches to the problem of classification are now being taken that involve trade-offs between the precision and coverage of searches and that focus more on the agency's mission and desired results. "I think a real problem for knowledge management is that there's really a spectrum of performance you are getting out of your tools and business processes," said Geoffrey Malafsky, president of Technology Intelligence International in Burke, Va. "A good enterprise architect balances those two against each other to come up with the best they can do, and I actually don't see that appreciated."

One of the problems is that current commercial tools for creating data taxonomies have a precision level of about 30 to 40 percent, Malafsky said. Spread that over a large enterprise and "would you expect the end user to be happy with the result?" Another problem is confusion that arises from competing organizational interests — from database engineers, to Web designers, to research librarians — who want different data classification schemes. Ultimately, Malafsky said, the taxonomies get "too big and unwieldy and have this odd mixture of being too granular and not covering enough information."

The first lesson that those with competing interests need to learn, Malafsky said, is that everyone's perspective is valid. "Let them be aware that they don't need to have the debate anymore," he said. "The Web designers are right; the database engineers are right; the librarians are right; the KM people are right. Everybody's right. It's just that there are different perspectives into the same information."

Essentially, the big agency players must sit down and work through their priorities. "Knowledge depends on the context of the moment," Malafsky said. "What are the top 10 issues you're really trying to use this information for on a knowledge basis?"

Having decided that, a short, knowledge-based classification system can be created based on those values, he said. A set of parallel taxonomy domains is then developed and labeled. The final piece is connecting it all back to the user interface.

"I think that is the current main stumbling block with the application of IT to KM," he said. "You have to define what those processes are" and come to the realization that "people can't just share knowledge anytime, anywhere, anyplace, because there's just too much."

Chief knowledge officers (CKOs) and chief information officers need to ask: What information do users want access to at any point in time? What data do they want clearly structured for business intelligence, decision-making or executive reports?

Moments of Learning

Many believe that knowledge management will be essential in government circles, where its emphasis on an enterprise view, teaming, expertise preservation and advanced search is well-suited to piecing together far-flung intelligence — a need that became acute after the terrorist attacks of Sept. 11, 2001.

"We've seen since [Sept. 11] a lot of activity around trying to put into place changes and practices that address the opaque character of government knowledge," said Hadley Reynolds, research director with the Delphi Group. "The very fact that you have 22 agencies [folded into the new Homeland Security Department] whose responsibilities overlap creates a pressing need for permeable borders between those organizations from an informational standpoint."

Of those agencies, one that since its start has been dedicated to homeland security is the U.S. Coast Guard. It is also one of a growing number of agencies that has a designated CKO. Like Holm and others in this new field, Coast Guard CKO Nat Heiner takes a somewhat Zen-like stance on knowledge management projects and technology.

"I'm nervous about most common characterizations of knowledge management because they pigeonhole it into areas where I don't believe it's going to be as effective as if it were deployed at the enterprise level as an area of endeavor," he said.

If pressed, he would boil knowledge management down to what occurs in an organization around "the moment of learning," which he describes as when and how an organization puts its people in touch with what they need to know to get a job done. "That can be tying knots, loading guns, it can be doing very complex skilled project management of very large and expensive IT projects; it can be managing a workforce. All of those things involve moments of learning that we don't think of as such.

"Instead, we have a knee-jerk reaction: 'If they're going to learn to shoot this weapon, they're going to have to go to school A and get this rating; if they're going to learn to put out fires shipboard, they're going to have to go to school C and get that rating.' And we have sort of a lock-step approach to this.

"In many ways that's totally appropriate," he said, "but we don't look across the enterprise to see who else is doing this, and how are they doing it, and could we do it jointly, or do it better that way. Training is like documentation and education — generally the stepchild; it's last in line."

Connecting the Dots

In its first generation, knowledge management involved identifying information that agencies had stored or stockpiled but did not really know they possessed. This has evolved, experts say, from a job of information classification toward development of "analytics," applications capable of finding the connections between telling traces of information.

"Five years ago, a search engine was an important part of knowledge management because it helped you find something that you were looking for," said Delphi's Reynolds. "We're substantially beyond that [and] now we're seeing these analytic tools that essentially [identify] trends as opposed to just finding an answer to a particular question."

Such tools, derived with software from companies such as Autonomy Inc., Convera Corp. and Verity Inc., attempt to glean meaning from content fragments, whether a formally published article, a series of e-mails or a chat transcript. In a simple example, Reynolds said a market research firm might find such a tool useful in tracking references to MP3 players on the Web.

"You would start to see around the world the rise in the discussion of that subject, and then you'll be able to identify who's talking about that subject and what are the companies involved," Reynolds said. "This kind of theme identification goes well beyond the original 'well, find me all the articles that have been written by Bill Gates.'

"This is one of the areas where technology is definitely playing a role in the advancement of KM practices," he said.

The architects of the Homeland Security Department are confronting the same challenge: deciding on ways to examine the spaces between scraps of information. The methods they choose will certainly incorporate some of the principles of knowledge management: people, processes and technology.

"In order to first connect the dots, you need to first connect the people," said the Coast Guard's Heiner. "You have to connect your people, and that's a very important part of what the knowledge officer ought to be focusing on for an enterprise."

Peer-to-Peer Collaboration

One tool being considered as a way of connecting people in the interest of homeland security is peer-to-peer (P2P) networks, which enable great numbers of people and groups to link across agency boundaries. The Defense Advanced Research Projects Agency, for one, has a deal with Groove Networks Inc. to use its Workspaces P2P platform in its Total Information Awareness project, according to Reynolds.

The project, which would use advanced analytics software to examine clues to terrorist threats that might crop up in public sources, wanted a way for project participants to collaborate. "One of the concerns that DARPA had was how do we allow the people within the 40 to 60 groups that are participating to collaborate," Reynolds said. So they are creating P2P networks between the groups using the Groove technology.

"Groupware environments are pretty much implemented on an organization-by- organization basis," he said. "But we've created these islands of collaborative information and that's contributed to the opaqueness of information. So these technologies, such as Groove, are targeting company-to- company and agency-to-agency networking."

Communities of Practice

Agencies are using another type of collaborative tool — Web-based communities of practice — to share information on education, training and careers. One of the more successful examples is the D7 University — named after the district where it was created, the Coast Guard's 7th District that serves South Florida and the Caribbean.

Built around the metaphor of a university with separate colleges, D7 is the brainchild of Joel Magnussen, director of information management for the Coast Guard's 7th District. "We purposefully designed it that way to be the first of many subliminal steps to help people overcome time and space when sharing information with each other," he said.

The site was designed to put people in touch with one another, store the interaction and provide a peer-rating system giving users recognition and an incentive to participate. That gave people some credit for participating, "which often hurts KM systems by not doing that," Magnussen said.

The site has about 250 regular members, with about a third outside the region. The five colleges are business, engineering, technology, human resources and law, together with a section called Capstone for senior executive topics — strategy, doctrine and guidance. Within the colleges, there are 510 different categories of information, which grow every day, according to Magnussen. The system also uses collaborative filtering built into the coding to locate experts when questions arise on the Web site.

"We wanted to really get to the core of how people learn from each other, and how you do work knowing that you can learn from each other," Magnussen said. That vision is embodied in the unofficial motto of D7: "What if everyone knew everything?" The question is one Magnussen asked in taking the floor at a commanding officer's conference several years ago. He asked, "What if everyone at your command knew everything that you did. Now how would you do your work?"

"It was meant to elevate the whole realm of their work and ask, if your most junior person, your most inexperienced tech person knew as much as you did, how much more powerful work could that person be doing? Also, are you ready to lead a group of people who know everything that you know?

"KM is difficult to grasp," Magnussen said. "It presumes you know how people think. And a technical tool that presumes to be able to provide somebody in California the equivalent of a water cooler exchange of little knowledge nuggets with another human being is a pretty broad stretch."

Looking Ahead

Once a government agency or private company has the right tools in place to classify, locate and share information at the moment of learning then crunch the information with the latest analytic tools, what's next? A key question to ask is, how good are the results?

According to Arthur Murray, president of TelArt Technologies Inc., a knowledge worker facing a decision at most large organizations, after sifting through all the pertinent data, will make a decision based on that information that fails about 50 percent of the time. Why? Murray said most workers are getting short-changed on the support they need to make such decisions.

The point of decision — what Heiner might call the moment of learning — needs to be broken out, Murray said. "We need to find what are all the steps that we need to go through to make a good decision," he said. "Then we want to walk the user, the analyst, the decision-maker through those things."

Murray argues that many business decisions are intuitive but that most federal workers are doing repetitive tasks, such as processing claims, analyzing budgets or approving line items. "You can go through hundreds of functions and tasks in the federal government and most have an internal logic that you can repeat. But today we call the [Internal Revenue Service] and get one answer back one time and a different one another time.There's no logic."

To establish that, CKOs and CIOs must model the decision processes of top-flight performers. "They make the right decisions consistently, and you want to capture how do they do it and make that [knowledge] available to everybody," he said.

"It's a very complex world we live in and the reason we have these mistakes, the reason that [the Washington, D.C., area] snipers were able to go through 10 different roadblocks is that we weren't asking the right questions and doing the right things," Murray said. "Those processes need to be captured and made available to the right people. We're giving them all this information but we're not helping them through it. When we do that, we will see a big increase in improvement." l