Fail and Learn

In a complex systems and policy environment, never underestimate how those complexities and legacy systems will affect an agency's ability to achieve its information technology and management goals. That's the chief take-away that Richard Spires says he's grasped after two years running the Internal Revenue Service's modernization initiative.

The central issue has been—and to a great degree continues to be—the requirement that the agency launch new systems that can tap its existing systems so no process ever grinds to a halt, says Spires, associate CIO for Business Systems Modernization at the IRS. Spires, who spent 20 years working in industry as an IT manager before taking his IRS job, spoke at a recent breakfast in Falls Church, Va., sponsored by market researcher Input of McLean, Va.

There are three rules to live by, according to Spires:

• Make sure to require adequate documentation. An agency can't rely on the institutional knowledge of its oldest or smartest team members—ultimately everyone leaves their job. This has been an issue for the IRS because many of its systems first came online during the Kennedy administration.
• Work closely with industry and other partners. Some of the problems that arose with not focusing on the interfaces to legacy systems happened because the IRS "underestimated the complexity" of these underlying systems and failed to relay this information adequately to its former vendors and even its current Prime contractor, Computer Sciences Corp. Plus, the IRS is "never going to have the technical skill" that industry has.
• Push initiatives down into the organization. While management buy-in is important, acceptance by the front-line managers of an organization is equally crucial to a project's success. Otherwise, cultural changes never take place. Technology adoption rarely proves tougher than the culture adjustments.

At the Core

For the IRS, its Customer Account Data Engine project most typifies these issues. CADE will ultimately replace the Master File databases that contain data about the nation's taxpayers—the data that the IRS uses to handle literally every facet of its mission. Because the Master File data is the equivalent of the IRS' lifeblood, the shift to CADE must be done gradually. The agency is rolling out versions twice a year, in January and the summer.

For the 2006 filing season that begins in February, the IRS plans to use CADE files to process returns for 10 million taxpayers-and then 30 million in 2007. For this year, the IRS posted about 1.4 million 1040EZ returns to CADE. There are more than 135 million individual taxpayers, so obviously there's a long way to go, "but we're clearly on a roll here," Spires says.

An audit by the Treasury Inspector General for Tax Administration found that CADE worked as promised, posting accurate information for the returns it handled. Plus, the General Accounting Office has reported that the use of CADE let the IRS cut the average turnaround time for refunds from seven to three and a half days.

For CADE, the IRS must convert flat-file records databases—written in assembly language and dependent on batch processing—to a distributed database management system.

The issue of the legacy systems interfaces is directly tied to the ramp-up of CADE because it can currently handle only so-called "clean" accounts, those for which taxpayers require no special interactions with the tax agency. In all, that's about 60 million accounts; after that, the IRS must be able to mesh functions in its Customer Account Management (CAM) system with CADE, Spires says.

In a new strategy the IRS intends to roll out in March, the agency will detail its plans for making CAM functions available incrementally so that it can continue to phase out the Master File and expand CADE processing, says Rocky Thurston, director of the Treasury Systems Division for Management Systems Designers. The Fairfax, Va., systems integrator is helping the IRS figure out how to create just-in-time CAM functions leveraging another IRS project, the Desktop Integration System.

"We needed to refresh our strategy for modernizing the IRS, mainly because it had become stale," Spires says.

New Rules

A final pair of lessons that Spires has learned: Agencies must see results from systems projects reasonably quickly and apply investment reviews across all systems evenhandedly.

Agencies need "systems that within a year they can get some use out of—that doesn't mean they have to be fully baked," he says. Three to four years to deploy a system is unacceptable, Spires adds.

In line with that, agencies must establish investment processes that account for systems across the enterprise. At the IRS, for instance, the focus had mostly been on its biggest projects, Tier A systems, while the plans for an entire raft of Tier B systems were often ignored. Yet, the long-term goals of the IRS demand that it makes sure that both tiers of systems fit in with the agency's mission strategy and that the systems integrate with one another and any legacy applications. This uneven focus came at a cost: In its two decades working on systems modernization, the agency spent upwards of $4 billion.

The IRS really must speed up deployments and review its systems investments wisely, Spires says, because "if we don't do this, the IRS is just making its job worse and technically more complex."