She's Got Their Number

The word is out: The math team can help. Dietrich fields a few dozen requests a month, half of which she turns down because the problem has already been solved or is not challenging enough. "We want to push the frontiers of what's solvable," she says. "Otherwise, what's the point?"

In a sense, Dietrich is doing what she enjoyed as a young math whiz--solving word problems. Here's a doozy: After IBM's sales team signs a consulting contract, the company often has to assemble the project team on deadline--say, 50 Java developers in Chicago by the following Monday. It can choose from 190,000 consultants around the world with various skills, personalities, and availability. It must do this for thousands of projects a year for clients of all sizes in every imaginable industry. Meanwhile, the mix of projects and available consultants is constantly changing.

"When we first started asking what resources consultants use on projects, they said every project was different," says Dietrich. "That just drove me crazy." By poring over two years of project data, the mathematicians identified which skills were most often applied in certain types of assignments. "You may not know exactly what the customer wants, but now you have a rough idea who you need for a $5 million project versus a $50 million project," says Dan Connors, optimization manager for the Workforce Management program. That staffing-analysis tool helped managers anticipate demand and schedule accordingly, boosting the consultants' productivity 7% and reducing travel expenses and the use of outside contractors. The savings exceeded $500 million. So do the math: Add in sales from the OnTarget forecasting tool, and that's a $1 billion contribution by Dietrich's math whizzes.

The brainiacs are tackling another problem whose solution could be just as valuable: how to pick the best teams. Project managers tend to select the most talented developers and engineers available, or the ones they already know. That may work well for the project at hand, but in the long run, it doesn't necessarily benefit IBM as a whole; better to spread the talent around. Researchers are also creating a social- networking analysis that would assess trails of email, instant messaging, and phone calls to identify which teams operate as flat organizations and which ones are hierarchical--who works well together and who doesn't.

But the problem that's really grabbing Dietrich involves predicting the workforce of the future. By analyzing population trends, employee demographics and skills, and demand for certain technologies, her researchers hope to identify labor shortages in various functions and professions before they happen.

That work, almost unthinkably complex and far-reaching, is nowhere near complete. Each answer generates new questions, and that's fine. That's good. Even mathematicians don't have all the answers. Dietrich won't get bored, and she'll turn out some lovely knitting. Eventually, she'll have numbers that help us think differently about the world and where it's headed--and IBM and its customers will hire or train employees accordingly.

It may well turn out, of course, that what they need are more mathematicians.