I have been beta testing Farecast, and I just recently used it to book a trip to Atlanta. The service claims to predict whether or not airfare is going to go up or down, helping you decide whether you should buy your tickets now or hold off. The idea is cool. The implementation, when completed, will probably be cool too. What I’ve experienced with the beta is that it feels like more of a novelty than a useful tool.

When I first started playing with it, I couldn’t find a combination of dates and locations that would produce anything but a statement that fares would be increasing, so I should buy now. That had me skeptical – after all, why would they tell me not to buy when they make money only when I do?

When I used it to book the Atlanta trip, I found a case where it did tell me to wait. I need to be in Atlanta by noon on a Friday, which means that I’ll have to leave Seattle late Thursday night. Farecast told me a Thursday->Sunday roundtrip will be decreasing by an average of $20 over the next 7 days.

But… it then occurred to me that the information they were providing me isn’t useful. I needed a late night flight on Thursday, so I could go to work during the day. So, even if the average ticket price dropped $20 over the next week, what if all the late night tickets sold out? Or, what if morning tickets drop $40 and late night tickets increase $20? The averages aren’t useful when you care about specific times. That said – according to the site they are working on predicting how airfares will move within time ranges. The ranges will be broken up into 6 groups, which seems like sufficient granularity for most cases.

They are not guaranteeing anything with the price predictions (which is probably for the best). They tell you that they are “80% confident” in their prediction. So, there’s still some element of gambling here. If they’re 80% sure the prices will drop for your trip home for Christmas, and you wait, and the flights end up doubling in price or selling out (which is a possibility 1 in 5 times), then you’re pretty much up a creek. It seems like they’d be well served to build in some monitoring tools for the case when they blew their prediction, and they can let people know as soon as possible.

I do like what they’re trying to do though – if they work out some of these details, it could be a very useful service. I double-checked other price searches and didn’t find any better deals, so if nothing else they seem to have a top-notch flight search.