In January 1998-March 2000, a new statistical assessment and referral system to improve the job retention of welfare recipients participating in welfare-to-work programs was piloted in the welfare-to-work program serving two counties in southwestern Michigan. The assessment and referral system is based on statistical methods and uses administrative data. It is designed to be integrated into an existing intake process, require minimal (if any) additional staff, and comply with the procedures and practices of existing welfare-to-work programs. During the pilot, more than 6,000 welfare recipients participated in the program. Rather than providing all clients with essentially the same set of services, staff evaluated each client individually by collecting information about each new client's educational and employment histories and analyzing that information with a statistical assessment tool based on the statistical relationship between an individual's attributes and job retention. Clients' scores were used to determine the level and type of employment services that might help the individual find employment. An evaluation based on a random assignment design established that the statistical assessment tool yielded retention rates that were 25% higher than when participants were randomly assigned to providers. The difference in retention rates between the best and worst referral combinations was 56%. (MN)