Airfares are as unpredictable as the weather and can change even more quickly. So I was dubious of claims by online travel tool Farecast.com that it could not only find cheap airline tickets but also forecast whether a fare would go up or down.
So I put Farecast to the test.
I started by searching a dozen travel websites for fares from Burbank to Dallas.
Farecast not only came up with the lowest fare, but it also boldly advised me to wait before buying the ticket -- saying it was 76% confident that the fare would drop.
Over the next five days, I checked prices and the "farecast" each day for three round-trip flights -- Los Angeles to New York, Orange County to Honolulu and Burbank to Dallas -- for departure two weeks in the future and a month in the future.
How can Farecast predict airline pricing? A computer algorithm created by a University of Washington professor culls through millions of pieces of historical fare data and looks for patterns in pricing.
Exactly what patterns and data it looks for is proprietary, but a recent independent audit of Farecast's 44,000 predictions found it was right about 75% of the time.
Initially, Farecast seemed to do better in our limited test. Its prediction for fares from Los Angeles to New York two weeks hence was correct every single time. It advised me to buy when the fare was at $399 and continued to do so as it kept rising. On the fifth day, the fare had reached $449.
That means I would have saved $50 by buying the ticket on the day Farecast recommended.
It was right about 75% of the time for the rest of the flights. But it flopped on the Burbank-to-Dallas flight. It was wrong three-fourths of the time, predicting that fares would remain steady or drop, when in fact the price kept getting higher. By the fifth day the fare was $40 more.
I should have known better: Experience and research told me that the $279 fare was going to be hard to beat. It was as much as $100 better than fares on other websites.
The lesson: Farecast is a good tool, but it's not perfect.