When you walked into the supermarket recently, you may not have noticed right away that you were in a Markov process with a 3-D state space.
But you surely noticed those long lines at the checkout counters, and so have operations research (OR) specialists. They use queuing theory, Monte Carlo simulations, linear programming and other esoterica to shorten customer lines and delivery times, minimize inventories and wring more revenue out of airplane seats, rental cars and hotel rooms. And they depend on information technology to do it.
Today more than ever, OR techniques are impacting the corporate bottom line. And IT managers can deliver the technology to make it happen.
OR Software Sources
Just where do you get operations research software?
A big company might just roll its own. In fact, IBM developed a system that optimizes inventory levels in its PC manufacturing supply chain, saving the company $750 million in 1998.
IBM now sells the system as its Asset Management Tool. But implementing that product is no slam dunk, says Brenda Dietrich, a senior manager at the IBM Research Division's Optimization Center. "You need some expertise in OR. . . . I'd say someone with a master's degree in OR could use it very effectively."
Supply-chain software from the major enterprise resource planning (ERP) vendors has some OR capabilities. "Some are integrated, so you don't even see it," says John Birge, president of the Institute for Operations Research and the Management Sciences in Linthicum, Md. "It might just tell you, 'Route your trucks this way,' and you have no idea it's optimizing in the background."
But Birge, who is also dean of the McCormick School of Engineering and Applied Science at Northwestern University in Evanston, Ill., says standard ERP packages can make unrealistic assumptions, such as unlimited parts access. So users may want to use an OR software specialty company, he says.
For example, Ilog Inc., Manugistics Group Inc. and Aspen Technology Inc. have supply-chain optimization packages, some tailored for specific industries.
"They have some general-purpose tools that you have to be fairly expert in to use," Birge says. "But they . . . can develop more application-specific tools for you.""IT managers should know what OR can do for them," Birge says. "They should know what applications have been developed by the specialized software houses and then decide if they need those capabilities." If they do, he says, they should consider having an OR-trained consultant install them and train users.
The term operations research dates to the late 1930s, when British and U.S. mathematicians developed ways to conduct research on military operations. For decades afterward, these powerful OR techniques saw little use at most firms.
But that's changing, as more companies use OR to improve operational efficiencies and increase revenue yields. Two forces are driving the surge in use: OR practitioners have gotten smarter at applying their tools, and corporate IT is now giving OR specialists the computational resources and databases they need. Indeed, experts say, the most successful companies will be those where OR and IT have established strong bonds of collaboration.
Recently, OR specialists at IBM worked with the IT department at Fingerhut Cos. to optimize the mailing of 480 million catalogs to 7 million Fingerhut customers.
Fingerhut decided to whom it would send each mailing, without regard to previous or future mailings. But the company wanted to map out a suite of mailings for each customer over time. That change, enabled by OR in a $3 million project, reduced mailing costs by 6% while boosting profit by $3.5 million annually, says Deb Campbell, business intelligence manager at Fingerhut in Minnetonka, Minn.
Despite the financial success of the project, Fingerhut wasn't able to push the optimization as far as it wanted. With 1 billion possible combinations of mailings to each customer, the company's computers weren't powerful enough to evaluate every possible outcome. Instead, the model considered groups of similar customers.
"The problem was too huge," says Campbell. "We can't solve the sizing issue today, but maybe we'll have the solution in a couple of years."
Cues from Queues
Queuing theory - the use of probability principles and modeling to understand how lines form and behave - was developed in 1915 for analyzing telephone switching systems. It's now used in systems that not only analyze queues but also control them. Richard C. Larson, an MIT professor who goes by the nickname Dr. Queue, has developed a tool that can help retail outlets deploy workers in a way that both shortens lines and reduces personnel costs.
Larson says the math that underlies queuing theory, such as Markov processes, is now being augmented with models of human behavior. "For example," he says, "the Disney Co. is world-class at managing the psychology of queues, and every facility they build, they model first." Indeed, the injection of human factors into models is one of the forces moving OR from academia to corporate data centers, Larson says.
And thanks to IT support, OR now can go where it never could before. "A confluence of factors - data warehouses with gigabytes of data and the cheapening of computation - have facilitated a new wave of OR applications," Larson says.
Supporting OR is both an opportunity and a challenge for IT, he adds. IT managers should know something about OR, Larson says, so they can explain to users and senior managers where the "hidden value" lies in all that data they're collecting.
IT managers "are looking for job security and growth and should structure things so the OR people know about the data sets and how to get more profitability out of them," Larson says.
Warren Powell is a professor of OR and financial engineering at Princeton University in Princeton, N.J. As director of the school's Castle Laboratory, he helps transportation companies optimize the placement of their assets. Castle has just two application programmers who write the optimization models for its clients, but "the heavy lifting is on the data side, and that's all done by in-house IT groups," Powell says. "That's the hard part of the project."
IT plays a strategic role in modern OR projects, Powell says. In the past, OR models often failed because they erroneously assumed they had complete and correct input data. "The new way of thinking is, I've got to model information the way it exists in the real world," he says. "This is where we are bringing the IT world into the OR world."
Modeling for Dollars
Michael Schrage, a research associate at the MIT Media Laboratory in Cambridge, Mass., urges companies to make extensive use of simulations, prototypes and models. He argues in his book Serious Play: How the World's Best Companies Simulate to Innovate (Harvard Business School Press, 2000) that the benefit comes less from the models' output than from the very act of modeling. The management insights gained though playing with models will shape the most successful companies in the future, he predicts.
Schrage argues that IT and OR should join forces to promote the use of OR, which, he says, often scares off general managers. "I think we'll see an explosion of OR, but you have to make it more accessible," he says. It should be possible, for example, to construct an intranet Web site at which general managers could run their own simulations.
Some companies are already at the state of modeling awareness and appreciation that Schrage advocates. All employees of Level 3 Communications Inc., a telecommunications company in Broomfield, Colo., are expected to have basic knowledge of three disciplines: economics, finance and optimization. Lorraine Lotosky, vice president of business optimization, says Level 3 has saved tens of millions of dollars by applying OR techniques to network design.
Lotosky, whose 15-person OR group is outside of Level 3's IT organization, says she feels it's so important to have tight links between the two disciplines that she employs her own application programmers. "OR could be in IT, or it could be anywhere there is a sponsor," she says. "But I like to have an IT function within the OR unit itself."
Fingerhut's Campbell says her OR group was recently put inside IT. "This latest move . . . will hopefully bring the development of an idea and the implementation of the idea closer together," she says.