A Simple Solution
Success can depend as much on tweaking the way people use tools as on what those tools are.
The food stamp system in Michigan was once a political embarrassment. For a time, the state-run program had one of the highest error rates in the country, with nearly one in five participants receiving either too much or too little help. An increasingly cash-strapped state government had to pay millions of dollars in federal penalties, thanks to its accuracy problems.
How the state reduced those errors is a technological success story. But as Michigan's chief information officer, Teri Takai, notes, the fix did not involve "some massive investment in technology or some massive investment in time." In terms of hardware and software, Michigan only made "fairly modest technology changes," she says.
The solution, it turned out, was usability. Simply changing the way caseworkers viewed, entered and updated data to calculate the food assistance benefits helped the state cut its error rate in half--from 14.1 percent in fiscal 2002 to 7.2 percent just two years later. That brought the program within a percentage point of the national average.
Based on the improvement, the U.S. Department of Agriculture's Food and Nutrition Service waived a $6.2 million fine for fiscal 2004 so the state could invest it in additional upgrades. And the feds are likely to allow Michigan to reinvest another $6.2 million penalty if the data for 2005 show sustained improvements.
Managers of large information technology projects often overlook usability, in part because it is hard to define and quantify. How do you measure a design or an interface? But anyone who has ever tried to place an online order on a retailer's confusing Web site or punch through a maze-like telephone menu ("dial nine for more options") knows how important usability is. Michigan's accomplishment is a reminder that technological success can depend as much on the way people use the tools they are given as it does on what those tools are.
The food stamp problems in Lansing were similar to those in other state capitals--just worse. The culprit: a classic "garbage in" problem. According to a May 2005 report by the federal Government Accountability Office, caseworkers were responsible for about two- thirds of the errors made in allocating food stamp benefits nationwide.
Michigan found that many of those errors said as much, or more, about complicated regulatory requirements as they did about carelessness. But officials also could trace many of the errors to a clunky combination of three computer systems used by caseworkers in more than 100 offices across the state. Some of those tools were still based on decades-old software applications. "It was a very old, green-screen system," says Sue Doby, an information officer from the state technology department. The three systems required caseworkers to toggle back and forth from screen to screen, use archaic commands and frequently enter the same data multiple times. That was a formula that "automatically led to errors," Doby says.
The state's long-term plans for consolidating and modernizing those systems would have taken "more time than we had before we had to fork over money to the feds," says Ann Marie Sims, director of the special projects office at the human services department. "We were looking for something we could do quickly."
Growing caseloads and increasing budget and staff constraints underscored the need for efficiency--and urgent fixes. So instead of waiting for the big upgrade, a cross-departmental team looked for ways to eliminate some of the major sources of errors, largely by making the existing tools easier to use. A Web-based application that helped workers determine eligibility could be modified to prompt workers to better document participants' income fluctuations and other financial and personal changes. The complex process of applying state and federal rules to calculate benefit levels also could be automated. The need to reenter information could be reduced by automatically transferring more data from system to system. In addition, the state could deploy a new tool to review cases and help managers spot and fix errors.
These and other tweaks were rolled out in 2003, and the results were immediate. Michigan's fiscal 2004 error rate was more than 10 points better than the program's worst performance in 1998 and 1999.
Michigan officials hope to see more improvements when outdated systems are upgraded and replaced. But the gains made already with relatively simple modifications to legacy systems show the value of focusing on the use of the technology--not just the technology itself.