How accurate is Farecast's forecast of airline fares?

Farecast launched this summer with a witty name and a clever premise: Using data mining and a huge library of historical prices, it would be able to predict a rise and fall in airfares.

By telling you whether to buy now or wait, the Seattle startup offered a beacon of light to people navigating the confusing and unfamiliar territory of airline pricing.

Even better, the company said it was 70 to 75 percent accurate on fares so far.

A Seattle Times review of nearly 30 plane trips, however, found Farecast's accuracy rate about 61 percent. And the vast majority of the time, Farecast has one piece of advice for travelers wondering whether to pull out their credit cards or hold off: Buy.

By offering essentially a second opinion on what amounts to a major purchase for some consumers, Farecast has won a significant amount of praise.

The online arm of BusinessWeek named the site one of the best in the Web for 2006. Time magazine called it one of its 50 "coolest Web sites." Frommer's added it to its "editor's choice" list.

Airlines have taken notice, too. Last month, American Airlines announced a marketing relationship in which Farecast would refer customers directly to the airline's Web site to book travel.

Farecast won't say if all the attention has led to an increase in users. The company doesn't give traffic numbers for its Web site but said the number of visitors has exceeded expectations.

Here's how the site works:

Once you tell Farecast where you're flying and when you want to go, the company pulls current fares and predicts whether the lowest fares will rise or fall in the next week.

Farecast offers predictions for about 75 departure cities in the U.S. It also tells you how confident it is about that prediction and shows the average fare for that trip going back two or three months.

Links to airline sites

If you want to make a purchase, Farecast often sends you directly to the airline's Web site.

Some airlines pay Farecast a fee for referring customers, though the company hopes to make most of its money from advertising sales.

For example, if you search for a flight to Philadelphia, Farecast shows ads for hotels and car rentals in that city.

Can Farecast be completely unbiased, given its tight relationships with airlines? The company says customers can track the information it provides and make their own purchasing decisions based on that.

Analysts generally say the information is neutral because it's based on aggregated search results.

A potential traveler could spend a long time on Farecast's page, with all the research options the company offers.

When asked for information recently on a roundtrip flight from Seattle to Dallas in early December, Farecast predicted the lowest fare of $221 would hold steady or rise within the next seven days.

Farecast said its confidence level in that prediction was 68 percent. Its advice? Buy.

Farecast also showed the fare history going back some 60 days for that route.

The Seattle Times tested its predictions in August and September and found that most of the time, Farecast said a price would hold steady or rise. Confidence levels ranged from 50 to more than 80 percent.

Though prices would very often stay steady, sometimes they would rise or fall by large amounts.

A trip from New Orleans to Washington, D.C., for example, rose $60 even though Farecast said it would hold steady.

Another time, Farecast said it was 80 percent confident that a trip from Omaha, Neb., to Atlanta would hold steady or drop. It rose $43.

Executive reacts

When shown a chart of the findings, Farecast executive Mike Fridgen generally did not dispute them. He took issue with just one, in which Farecast said a flight from Cincinnati to Seattle would hold steady or rise.

The fare went up by $186, which Fridgen said was technically accurate in that it did rise.

The Times gave that trip a "C" grade in its informal evaluation system, which assigned a letter grade to each forecast based on how far off the mark it was. Although the fare did indeed increase, it rose by a much larger amount than one would expect with Farecast's prediction.

Fridgen stood by Farecast's self-calculated 70 to 75 percent accuracy rating, saying it was based on all predictions on all of its routes on a daily basis. The company makes nearly 38 million predictions a month, he said.

"We're not clairvoyant"

"It reinforces what we're telling customers," he said. "We're not clairvoyant. We are as transparent as possible with what we're seeing."

Farecast has observed more than 100 billion airfare prices to date, Fridgen said, and has a team of computer scientists working on its data-mining technology.

But some people familiar with the travel industry wonder how Farecast could make predictions in a business where airfare prices are volatile and sometimes guided by emotion over logic.

"There are times when airline pricing managers behave like petty children," said Henry Harteveldt, an industry analyst with Forrester Research

"One airline gets mad at another for something they're doing in an unrelated way, and they retaliate through a price action," he said.

Rival taunts

George Hobica, the creator of travel site AirfareWatchdog.com, said he's heard stories about industry conventions where pricing managers meet at the hotel bar and shout taunts like, "See what I did to you last year in Phoenix? How'd you like that?"

"These are not rational people," Hobica said. They're passionate about it, but it's not rational."

Airfares can be very unpredictable, Hobica said, which flies in the face of a prediction site. You can't anticipate human nature, he said.

"The real question is do you believe that you can predict an airfare?" he asked. "Personally, I'm somewhat doubtful, but I think the jury's still out."

Airfares can be influenced by other factors, such as the entry of a new competitor into a market.

Also, analysts say, airlines are keeping a lid on capacity by using smaller aircraft and are being more aggressive with the way they manage inventory. That can affect prices as well.

Fridgen disputed the idea of emotional managers setting erratic fares in an industry so tightly focused on making money.

Rational patterns

The airlines go through a rational process, and there are patterns to be gleaned from that, he said.

"Obviously they have a goal, and it's profitability," he said. "It would be hard for me to believe that they're putting that at risk for emotional reasons."

Farecast is confident in some markets, because the pricing patterns there are recognizable, Fridgen said. In other markets, it's less confident.

The company makes sure customers know that, he added.

Kim Peterson: 206-464-2360 or kpeterson@seattletimes.com