The $230 Million DebacleBy David F. Carr | Posted 2003-03-19 Email Print
A 6-year-old Florida girl in foster care disappeared. In New Jersey, a 7-year-old died. How computer systems fail to protect the most vulnerable.The $230 Million Debacle">
Just a few months before the Rilya Wilson scandal broke, the South Florida Sun-Sentinel newspaper ran a story about HomeSafenet headlined "Computerized Tracking Project Years Behind And $198 Million Over Original Budget," explaining that the project wouldn't be finished until 2005. As the year went on, newspapers across the state began routinely referring to HomeSafenet as a "$230 million debacle."
CIO Niewenhous was forced to resign, and the State Technology Office took on more responsibility for the project.
Meanwhile, several usability-focused upgrades have been released. But except for proceeding with the statewide rollout of protective investigations modules, Florida is slowing work on upgrades while considering more radical measuressuch as the possibility of replacing some HomeSafenet software with off-the-shelf alternatives. Ben Harris, chief information officer for human services, says at a minimum he wants to discover some best practices to imitate from commercial products and other states' systems.
Niewenhous complains that the scandal-focused news reports that led to his ouster exaggerated the growth in the project's budget and slighted the developers. "This was something a lot of us really put our heart and soul into," he says.
While Florida's SACWIS planning began in 1994, it took until 1997 to get a federally approved project plan and issue a request for proposals. In 1998, the state convinced one of two bidders, Unisys, to lower its bid from $60 million to $54 milliona figure that still wound up being rejected as too high.
The HomeSafenet project began in 1999, with Niewenhous's $210.5 million plan for DCF to act as its own integrator. In 2001, after reality-checking their progress, the DCF developers revised their estimate to $230 million (actually a $42 million increase because it didn't include post-production costs factored into the earlier figure). They also changed the projected completion date from 2003 to 2005.
But Niewenhous says the "$198 million over budget" headline is an exaggeration; it compared his cost projection to a $32 million figure discussed in the early 1990s, before any serious planning or analysis had been done. Even the Unisys bid isn't directly comparable, he says, because it doesn't include the salaries of state workers assigned to the project.
The Unisys bid also was based on transferring software originally developed for other states. Florida decided its system should be Web-based for greater reach, and Niewenhous thought it made more sense to create a system specifically architected for the Web rather than trying to graft a Web interface onto an imported system.
DCF selected the WebLogic application server from BEA Systems and began the task of migrating data out of six existing computer systems that HomeSafenet would ultimately replace. Beta testing started in November 2000, and Release 1.0 was deployed statewide in July 2001. By spring 2002, HomeSafenet had gone through an architectural update that included migrating to a newer version of WebLogic and making more use of Enterprise JavaBeans. Taking advantage of more advanced Java capabilities eliminated the need for a separate team of developers that had been writing code for BEA's Tuxedo middleware. The application server also moved from SCO Unix to Red Hat Linux.
"Without being stupid bleeding-edge, we were having success with technology in a way that's atypical of government," Niewenhous says. He suggests his biggest mistake was authorizing early releases of the system without adequate support for generating reports.While developers struggled with the Unix version of Cognos, and ultimately switched to Crystal Reports, users were left without a good way of getting data out of the system.
Whether Niewenhous made the right call by developing from scratch is debatable. AMS, a consulting firm, has taken the same basic SACWIS design originally created for Connecticut to several states and generations of technology, including a Web implementation in Wisconsin. That experience has produced a more user-friendly system, says Patricia Mellon, a senior principal in the AMS public sector practice.
The charge that the user interface needs improvement is "a fair observation," Niewenhous says. But he also had to deal with requirements outside his control, like a dictate that users not be allowed to edit data once it was recorded in the system. Though intended to block fraud, this also made it impossible to fix unintentional mistakes. Niewenhous says "it took forever" to alter that requirement to allow edits, with the appropriate audit trail.
Ultimately, the goal should be making the system "child-friendly" more than "user-friendly," Niewenhous says. "The purpose isn't to make the worker's life easierit's to protect children."