Managing Uncertainty

How can managers make smart decisions about staffing, inventory, production scheduling and pricing? Too often such decisions are made haphazardly, says Hyun-Soo Ahn, who believes that insights from analytical models can help businesspeople make better choices. The assistant professor of operations and management science at the Ross School develops and analyzes mathematical models that guide resource allocation and pricing.

Hyun-Soo Ahn

"My research can be applied in manufacturing, supply chain design and management, and retail and service operations such as call centers and healthcare delivery systems," explains Ahn, an industrial engineer by training. He has worked with several companies, including 3M, GE Healthcare, Intel, Medtronic and Powerlight.

"The beauty of analytical models is that they can transcend biases and limitations present in real-time human decision-making processes. Carefully analyzed models create insights to support decision makers when they are faced with uncertainties," explains Ahn, who uses stochastic models, mathematical models that contain random variables that represent random factors, to prove his theories about optimal resource allocation and decision making.

In managing workflow, no simple rule works in every case, he admits. Even in the simplest case, some workers may be faster or slower than others, which complicates how resources should be allocated. In a highly flexible environment, natural workload balancing can be achieved by a simple "bucket brigade," named for the way firefighters once lined up and passed buckets of water to extinguish fires. Tasks are completed as they move down the line. When employees finish their tasks, they walk "upstream" to take over their predecessor's task. To avoid bottlenecks and for maximum efficiency, Ahn says, employees must be arranged so that the fastest person is at the end of the line.

Another practice that Ahn says works well in a flexible environment is "last-buffer, first-served" (LBFS), in which employees finish tasks that are closest to completion even if the task is primarily another employee's responsibility.

"Each activity along the way adds value," says Ahn. "So it is natural to think that the company should not pile up valuable work in process (WIP). Following LBFS minimizes the WIP and generates a smooth flow of finished output, both of which are consistent with the just-in-time production associated with a successful pull-based manufacturing system, most notably Toyota's."

Another common workflow approach used in retail and service operations, he says, is the "pick and run," in which one employee handles a task from beginning to end. It is used in IBM's order processing department where "it is much more desirable to have one person who can perform multiple duties associated with a single client or job," says Ahn, rather than have one person take the order, another check the customer's credit and a third provide a price quote. "Coordinating information takes time. When individuals are trained to do multiple tasks, it eliminates time spent communicating and simplifies the process."

However, Ahn's research shows that assigning one employee to complete a job from start to finish doesn't always work. In a less flexible environment, the manager needs to evaluate each production resource — what task each resource can do and at what speed — and the importance of each job and adjust the resource allocation accordingly.

Using cross-trained employees works well in environments where employees can quickly change from one task to another and the tasks are simple enough that it is easy to train workers, like a sandwich shop or order-fulfillment center. However, cross-training is not necessarily beneficial when tasks are so complicated that it takes a long time for a worker to become proficient at each task.

As part of his research portfolio, Ahn also studies dynamic pricing and issues at the interface of operations and marketing. Traditionally, marketing has set prices, and the operations department has managed production schedules, says Ahn. Both decisions were made in isolation. Because activities that increase revenue do not necessarily improve profits, he adds, it is valuable to look at pricing and production decisions together.

One model he works with assumes that consumers make purchasing decisions based not only on prices in the period in which they enter the market, but also on past and future prices. For example, suppose a person plans to buy a computer and spend $3,000 or less. If the computer costs more than the consumer is willing to pay, the consumer can wait until the price drops, change preferences or buy a competitor's computer.

"Consumers look at price history as well as current prices," Ahn says. "This kind of behavior can change the manufacturing strategy for firms whose capacity at times is not sufficient to meet demand. In order to maximize sales, firms need to realize that while they are selling at a high price, more and more low price customers are waiting patiently. To make sure all this demand is met, the firm might need to build up inventory for its low-price sales. For retailers using periodic sales to price discriminate, having enough inventory is crucial to the company's profit picture."

Ahn says the healthcare delivery system is another field ripe for study using mathematical models. "In healthcare, the division of labor has limits, as doctors and nurses can spend a significant portion of their day simply transferring information to one another. In such environments, cross-training makes sense.

"Determining what tasks to cross-train and how many doctors and nurses a healthcare system needs is a natural application of this kind of research. A great thing about being at the University of Michigan is the opportunity to do this kind of multidisciplinary research," says Ahn, who last spring was named the Sanford R. Robertson Assistant Professor in Business Administration in recognition of early career development and excellence in scholarly research.

In January, Ahn and BBA students who have taken his statistics and management science course will work on a project with the U-M Hospital's emergency medicine department. "The students like analytics and decision making and want more experience. This will be a test bed of action-based learning for BBAs in statistics, management science and operations."

Ahn, who also received the Ross School's 2006 BBA Teaching Excellence Award, frequently uses mathematical models when advising Multidisciplinary Action Project (MAP) teams and Tauber Manufacturing Institute (TMI) summer internship team projects.

"Through TMI and MAP, I am helping future managers gain analytical skills that can be applied to solve problems. Analysts who appreciate data have better tools to make decisions."

Written by Mary Jo Frank