Living its first blissful decade of existence with no competition in its business of providing outsourced payroll services, Paychex Inc. paid scant attention to preserving existing clients–that is, until 2002, when a client attrition study documented that more than 20% of the client base had evaporated. Not only did each loss represent the elimination of prospective revenue, but Paychex actually absorbed a loss with each departure, thanks to front-loaded costs from commissions and setup. "It was clear that a redeployment of our priorities was necessary," says Frank Fiorille, director of enterprise risk management at Rochester, N.Y.-based Paychex, with $1.9 billion in 2006 service revenues.

That redeployment spawned the Paychex Attrition Model or PAM, a data-intensive tool to identify clients most likely to leave. The tool utilizes data modeling and prediction intelligence using a set of 200 characteristrics from empirical data on about 561,000 clients. "We felt that through actuarial science and analytics we could look at the reams of data we have on clients to predict which ones might sever ties and then do something to prevent it," Fiorille says. The names of the most likely to go are then given to field offices.

Rolled out in 2004, the tool has helped cut attrition below 20%, and Fiorille estimates that more than 2,600 clients were saved using PAM this year alone. "Had we lost them, our revenues would have been reduced by nearly $5 million," he says.Given the success, Fiorille anticipates a series of Son of PAM projects. Paychex is even considering bringing the tool to market.