UW professor's software may help consumers get better airline prices

Oren Etzioni was headed to his brother's wedding in Los Angeles last summer when the thought came to him: Everyone on his flight probably paid different amounts for their 17 inches of space.

Etzioni's technical mind began beeping and blurping. Is there a rationale behind how airlines price their flights? Is there a way to track the best times to buy a cheap ticket? Could he create software to help consumers battle the airline companies' reservation systems?

Yes, he decided. To all three.

The associate professor of computer science at the University of Washington claims he's discovered certain patterns that can save consumers hundreds of thousands of dollars.

The new software he's helped to develop to do that is called Hamlet, as in "To buy or not to buy."

On a simulated run of the program, 607 would-be passengers saved almost $300,000 based on advice on when to buy a ticket.

"Airlines have the benefit of using sophisticated software that does yield management, as it's called, which attempts to move ticket prices to maximize their profits," said Etzioni. "I would like to make the playing field a little more level and allow the consumers to have more sophisticated software for their buying decisions."

Etzioni applied for a patent for the approach and formula he and UW graduate student Alexander Yates developed, and they predict a workable program for the average consumer can be produced within a year.

But the skeptics have already lined up.

"To me it sounds like a pretty far-fetched idea," said Jack Walsh, spokesman for Alaska Airlines.

Walsh said airlines determine prices by numerous factors including the economy, competition, time of the day and unpredictable events like the Sept. 11 terrorist attacks and war in the Middle East.

"It's hard enough for us to predict what fares we're going to have on down the road, let alone anyone else," he said.

Although there are some general rules to cheaper fares, such as including a Saturday stay and flying midweek, there really is little predictability when it comes to airfares, said Tom Parsons, an airline-fare-watcher for the past 20 years.

"Airlines put a guy in a padded room and throw him around for about 15 minutes and whatever comes out of his mouth, they spit out," said Parsons, chief executive officer of the online discount-travel company BestFares.com. "It's a roll of the dice."

BestFares monitors about 7 million airfares a day, Parsons said.

"Airfares have highs and low. And those can pop at any single time, and they can be the cheapest fare in months or years," he said. "Sometimes there's logic to it, but for someone to predict he knows when the cheapest airfare is, I wouldn't buy that."

Etzioni recruited a professor and graduate student at the University of Southern California, which has specialized data-gathering infrastructure, to record the prices of more than 12,000 flights over 41 days last fall. They looked at nonstop flights from Seattle to Dulles International Airport near Washington, D.C., and from Los Angeles to Boston in January 2003.

Using the lowest fares published on the Orbitz discount-travel Web site for six airlines, they found a price of a ticket for a flight can change as often as seven times in a day.

They discovered a fare from Los Angeles to Boston can vary by more than $2,200 and a flight from Seattle to Washington, D.C., can vary by almost $1,400.

More important, the researchers found patterns to the price fluctuations.

While some of the general rules still apply — prices increase two weeks before departure dates and they're at a maximum the day of the flight, they also found that big airlines like American and United airlines alter their prices more often than smaller ones, such as Air Trans and Southwest.

And when one airline drops its fares, the others often follow. There are also airline-specific rules the researchers discovered.

"The rules are actually pretty complex," said Etzioni.

But the research is flawed because flights need to be monitored for at least six months or a year for there to be any true value in its conclusions, said Parsons of BestFares. And flights between Seattle and Washington, D.C., and Los Angeles and Boston would not be indicative of the entire industry, because those routes are loaded with competitors and low-cost carriers, he said.

Etzioni admitted more work needs to be done and said they are continuing to collect data on more flights and will expand their research to include multi-leg flights.

The ultimate goal is for an average consumer to be able to type a flight into the system — which would perhaps be available over the Internet — and the system would return an optimal date to purchase the ticket.

But wouldn't that change buying habits, and thus the airlines' pricing techniques?

"One of the things that I've learned over time is that people's behavior changes slowly," said Etzioni. Even if just 5 percent of airline ticket buyers purchased their flights using Hamlet technology, that wouldn't be enough for airlines to change their fare structure, he said.

And if they did, Hamlet would detect the changes and alter its model as well.

"It's like an arms race," said Etzioni. "When two people are deploying technology, both sides continually update their technology."

Gina Kim: 206-464-2761 or gkim@seattletimes.com