IBM Research developed a series of 12 "micropractices" to focus its efforts (see chart, below). "A micropractice is really just a little services practice that we think could potentially grow some day into a big practice for our BCS [Business Consulting Services]," Horn said.
"One of my favorite examples is in this whole area of text analytics," Horn said. "One of the micropractices is called Information Mining and Management. I think of it as Google on steroids. Web Fountain is an example where you can get detailed business insights using technology that can be a huge lever for our services business."
Another micropractice is in optimization, which is "about taking hard optimization problems in the presence of a noisy environment and trying to find the best solution," Horn said. "So most mathematical techniques for doing an optimization on a certain problem, such as optimize your supply chain or flight scheduling, these are hard math problems. But most of them dont work well in the real world because the real world is noisy, so IBM applies a technique known as stochastic optimization."
So far, IBM Research has worked on dozens of services projects with clients and used as many as 200 researchers around the world.
"Probably our poster child effort is BostonCoach, Kennelly said. "BostonCoach is a limousine company, and their business model is based on the premise of providing very high customer satisfaction. So if they tell you theyre going to be at your house at 2 oclock, theyre going to be there. And if they anticipate therell be traffic or problems, they will dispatch two drivers.
"In fact, they average about 1.2 to 1.5 drivers per customer call. They had done this by localizing their operation and updating schedules often. But no one person can manage all those variables on their own, [variables] like Are you a VIP? Are you one of their best customers? Theres a storm coming."
Kennelly continued: "They thought they could solve this problem by putting GPS [Global Positioning System] systems in their cars, but one of the researchers concluded that that wasnt going to solve their problem. Ultimately, we worked with our consulting division to build them an optimization model. We were able to take the variables and trim them into a math problem."