"It doesn't matter necessarily what the subject is, the question is how do you give people incentives to make better decisions."
Assistant professor of management
Companies spent more than $60 billion on television advertising in the United States last year. Given the size of the market, perhaps it isn't surprising that the question of how much a network should charge for its advertising time is not simple to answer.
Mihai Banciu, an assistant professor of management who specializes in operations research, found part of the answer in bundling, a concept familiar to anyone who plans a vacation, from airline and auto reservations to hotel and tourist destinations, online. In the case of television advertising, the idea is to offer discounted packages of prime time and non-prime time slots together, since advertising time is perishable (just like an empty airline seat after takeoff), and therefore selling the entire advertising time inventory is critical to the TV networks.
"It's not simple to figure out exactly what bundles of prime and non-prime time slots to offer, in what proportions to mix them inside the bundle, and how much to charge," Banciu says. "I had to think about finding faster algorithms that could find prices in a reasonable time, given the number of possible bundling options, estimated demands and their perceived value." In addition, bundling is not always the optimal strategy, so he also has devised ways to determine when a network should switch from selling bundles to selling time slots individually.
Applying Banciu's methods can give both sides a better deal than the old ways, in which prices were set by "gut feeling," he says. Television networks can maximize revenue while advertisers get the best exposure given their ability to convert this exposure into purchases. "It doesn't necessarily follow that if a television network would charge a higher price, it would be at the expense of their corporate client, because there are also competition effects to consider" he says. "There are definitely situations in which everybody is better off, a win-win proposition."
Banciu also works on a problem called "survivable network design." As with a highway system that consists of cities connected by roads, a telecommunications network consists of hubs and other access points, connected by lines. The challenge is to design a network that is robust enough to resist failure and meets the desired quality of service levels, while still being affordable.
With too little redundancy, a problem at one node (or along a single path) can take down the entire system. With too much redundancy, companies waste money laying excess cables. Moreover, perfect redundancy usually is prohibitively expensive to attain, since it involves connecting all pairs of nodes in every possible way.
Figuring out exactly how to design the optimal network is a complex problem that is not quickly solved with current computing power, especially for systems that contain 10,000 or more hubs. Instead, Banciu is working out practical compromises. "You start thinking about solutions that are not necessarily the cheapest - they are reasonably cheap but they can be found very fast," he says.
Banciu was drawn to the field for two reasons: it's interdisciplinary and it's practical. Operations research is about solving problems, regardless of where they occur. "It doesn't matter necessarily what the subject is, the question is how do you give people incentives to make better decisions," he says.
The interdisciplinary nature of operations research, which draws on fields like mathematics, economics, computer science, psychology and industrial theory, makes someone with such a background a perfect fit for Bucknell. The practical aspect suits Banciu. He says, "I can be as theoretical as I want in my research, but at the end of the day, I have to show that someone out there in the real world benefits from my work."
Posted Sept. 22, 2009