DURHAM, N.C.--Duke engineers have shown that intentionally imprecise rules of thinking called "fuzzy logic" can help hotel computers sell the right room to the right customer at the right time, thus boosting income.
In a pilot study at two North Carolina hotels, researchers at the Duke School of Engineering's Machine Intelligence Laboratory found that the Bass Hotels and Resorts chain could achieve a "measurable" revenue increase by adding a fuzzy logic expert system to its computerized revenue management network.
"This is a very nice test bed to show the generality of fuzzy logic," said Mounir Ben Ghalia, a research assistant professor at the Duke lab. "Fuzzy logic is a theory that handles vagueness and uncertainty. It doesn't come with specific application."
Paul Wang, who directs the Duke laboratory and has helped organize a number of U.S. conferences on fuzzy technology, added: "This research has broken new ground in fuzzy logic's applications to economic and social systems."
Craig Eister, a Duke alumnus who directs Bass Hotels and Resorts' revenue management system, applauded the work, which is being funded by his Atlanta-based corporation. "For us, revenue management is an art," Eister said. "This project is extremely exciting and valuable because it combines the latest technology with the insight of the human mind."
Originating in the United States, fuzzy logic has so far commanded the most attention in Japan. It has been used there in subway locomotive controls, water purifiers, washing machines, vacuum cleaners, air conditioners and cameras, Wang and other authors noted in the journal IEEE Transactions on Fuzzy Systems.
Fuzzy logic, the researchers said, is useful in problems for which solutions are not clear-cut, and hotel revenue management is a good case in point.
Hotel chains have come to rely on elaborate software programs to predict customer trends well into the future, Ben Ghalia added. "They use historical data from past years to forecast what the demand is going to be on a particular day. The decision on whether to accept a room reservation at a certain rate is based mainly on what the computer forecasts."
"But there are a lot of uncertainties about the forecasting," he said. For instance, an annual convention may be held in the hotel's neighborhood for five consecutive years only to move elsewhere on the sixth. And bad weather, or even the mood swings of travelers, may cause bookings to be suddenly canceled.
And revenue management is about more than just booking rooms; it's also about deciding how much to charge for them, he added. Business travelers, government employees, conventioneers and tourists are all charged at different rates. If higher-paying customers get turned away because rooms for a lower-paying tour group are overbooked, the error costs the hotel money.
Amid all these uncertainties, computerized revenue management systems often begin making bad decisions, the Duke researchers found. Hotel managers must thus constantly monitor the program's decision making. And they must be prepared to override it, drawing on their years of personal knowledge, experience and intuition.
"They spend a lot of time doing that," Ben Ghalia said. "But even hotel managers go on vacation. And when they do, usually the manager's assistant is not as careful. New managers can also become overwhelmed with the system's complexity. It is very complicated to understand, because it is not written to respond to human experience and thinking."
Computers ordinarily run on "binary" logic, which calculates only in ones and zeroes or expresses problems as statements that can be answered "yes" or "no." Binary logic-based programs called "expert systems" can mimic the decision-making powers of real human experts by quickly scanning information stored in computer memory.
But neither binary logic nor expert systems can adequately handle "imprecise and uncertain" knowledge, the Duke researchers said. Examples range from whether a room seems "hot," "cold" or "comfortable," to whether advance booking rates have the right "feel" to a hotel manager.
"When human beings pass on information, we often use vague knowledge that is neither black nor white but instead occupies a gray scale between the two," Ben Ghalia added. "Fuzzy logic is able to model such uncertainty much better than binary logic."
Ben Ghalia and Wang said they have accordingly built a special prototype "fuzzy logic expert system," based largely on the insights of two seasoned managers at Holiday Inn hotels in Fayetteville and the Research Triangle Park area between Raleigh and Durham.
"We used two managers because sometimes different managers have different ideas for looking at problems," Ben Ghalia said. "We had informal interviews to learn the way they operate on a daily basis. We also attended 'user group' meetings for the East Coast, where all the hotel managers bring their questions and complaints about the current system."
The Duke investigators then wrote up formal questionnaires for both managers and followed that up with more question-and-answer sessions. Many questions dealt with how, when and why managers chose to override the computerized revenue management system, he added.
"It was a very interesting experience," Ben Ghalia said. "A person who has been a manager for 20 years knows how to do things, but he may encounter difficulty relating the hows and whys of his knowledge to others."
Their still-experimental fuzzy expert system could free managers from the constant job of watching over the hotel computer. "The system works nicely, and it does not get tired," Ben Ghalia added. "Instead of wasting time on monitoring the system, managers can use that time and energy for other things."
In later stages of the project, the Duke investigators also plan to evaluate fuzzy logic as efficiency boosters for the hotels' catering services and heating and air conditioning systems.
"We are building models for complex systems in the laboratory," Ben Ghalia said. "A complex system can be a robotics system, a hydraulic system, or a revenue management system.
"The common element among them is uncertainty."