It's not a "computer problem", it's crappy software design

In over 20 years in IT (gee, has it been that long?), I've never claimed to be the most crackerjack programmer in the world. I know people who can crank out bug-free code at an amazing rate, but I was never one of those people. I'm pretty good at designing a normalized database, and because I came out of a liberal arts education and non-technical fields before stumbling almost by accident into IT, I understand what makes a good user interface and what makes a crappy one. This is, in my most humble opinion, a skill that is often sorely lacking in most IT projects. When you let programmers design user interfaces, you get, well, user interfaces designed by programmers. They're often convoluted, ugly, not at all intuitive, and only those who designed them understand how to use them.

Based on this article in the New York Times about the possible suspension of funding for the FBI's "computer overhaul", it sounds to me like you had a bunch of IT people using off-the-shelf software to develop applications for people who until a few years ago couldn't send e-mail:

F.B.I. officials said that design changes and “minor” technical problems prompted the suspension of parts of the third and fourth phases of the work, which is intended to allow agents to better navigate investigative files, search databases and communicate with one another.

The decision to suspend work on the $305 million program is particularly striking because the current contractor, Lockheed Martin, was announced to great fanfare in 2006 after the collapse of an earlier incarnation of the project with the Science Applications International Corporation.

“This is terribly frustrating,” Senator Charles E. Grassley, the Iowa Republican who has been a frequent critic of the F.B.I.’s computer systems, said in an interview Thursday. “We’ve been through this song and dance before. Wouldn’t you think after hundreds of millions of dollars being wasted that they’d finally get it right?”

Beyond the financial costs are concerns about the F.B.I.’s ability to handle its law enforcement and national security responsibilities with an information system still regarded as sub-par in some crucial areas.

In a paper-driven culture, the agency’s computers were so inadequate that many agents until several years ago could not send or receive e-mail messages, and had difficulty getting case histories and linking to other databases. In the aftermath of Sept. 11, 2001, agents in Florida had to send photographs of the hijackers by overnight mail to Washington because they could not send e-mail attachments.

The current project, known as Sentinel, has fixed some longstanding problems, including difficulties with e-mail and database searching, auditors have found.

But in examining recent work, officials realized that mundane problems — like slow response times, awkward display pages and screen print that was too small — were cropping up.

Posing added complications were guidelines on F.B.I. operations put in place in 2008 that gave agents more latitude to look at factors like ethnicity and religion in terrorism investigations. That type of data was not typically used by the F.B.I. when the project began, and officials said they were trying to find ways to incorporate the expanded criteria into their investigative files and computer reports.

Now, I don't know anything about the FBI's systems, nor do I know anything about what Lockheed Martin has spent $425 million developing. But based on the linked article, it appears that what Lockheed Martin has been working unsuccessfully to do for the last four years using off-the-shelf software (probably SAP or Oracle) indicates that either they have a thousand monkeys at a thousand keyboards working on the project, or (more likely), the tool they're using is so cumbersome out of the box that it's impossible to do anything with it that anyone would want to use. Given the extent to which Lockheed Martin uses SAP, my guess is that's what they're using.

Here's what ZDNet (linked above) said when the contract was awarded in 2006:

The FBI is on track for a fully computerized casefile program, dubbed Sentinel, with a $305 million grant to Lockheed Martin, FBI CIO Zalmai Azmi said. The full program, which will cost $425 million, will give agents secure Internet access to many systems within a year and the entire system will be running by late 2009, Azmi claims, according to the Washington Post.

Of course similar claims were made of Trilogy, the FBI’s previous attempt at a massive computer upgrade, which tanked on the iceberg of Virtual Case File. Azmi says Sentinel is different.

Azmi told reporters yesterday that the FBI has included an aggressive system of audits, outside management review and financial controls to guard against any problems.

"We have a number of controls in place to ensure that this program is not following in Trilogy’s footsteps," Azmi said.

Azmi said Sentinel would offer FBI agents and analysts "one-stop shopping" for access to the bureau’s dozens of incompatible databases. "The list of capabilities that this program will bring will be enormous," he said.

Still, the Justice Dept.’s IG released a report last week, casting doubt that the FBI has adequate management controls in place and that Sentinel, even if successful would deliver information-sharing capabilities across agencies. 

Two companies that are part of the Lockheed Martin project team, Computer Sciences Corp. and CACI International Inc., also played roles in the earlier Trilogy efforts, Azmi said. The main contractor on the failed software system, Science Applications International Corp., is not involved in Sentinel, officials said.

Linda Gooden, president of Lockheed Martin Information Technology, said she is "confident that the FBI has instituted the appropriate checks and balances" to guard against overruns and other problems. The company’s pay will be "performance-based," she said.

I guess that's why they've only received $305 million of the $450 million allocated for the project.

If two smart college kids can develop a search algorithm that becomes the most widely-used search tool in the world, you'd think it wouldn't be rocket science to produce something better than a system that relied in overnight snail mail until a few years ago.