You know about travel search engines that display the lowest fares airlines
are offering today. Now there's a travel engine that can read the airlines' minds.
Farecast.com not only shows you which fares are currently the cheapest, it uses
advanced mathematics to predict whether those fares are likely to go up or down
in the next few days. If ticket prices are likely to drop, you can save money
by waiting to buy until they do.
After years of development and private beta testing starting in 2003, Farecast opened its site to the public
for the first time today. The service is far from complete at this early stage.
But the features that are already working offer a tempting look at tomorrow's computer programs, which will give you
guidance and advice rather than mere data.
Using Farecast to Buy with Confidence
Before Farecast existed, you might start planning a trip using the "flexible dates" features of
Orbitz or Travelocity.
For example, you might want to fly from Boston to Los Angeles between July 1 and
July 30, but you don't have specific dates you must travel on. In a recent search,
Orbitz showed prices as low as $367 if you departed on July 18 while Travelocity
offered a price of $371 on the same date.
Farecast adds a new and improved wrinkle to your flight planning. Your search at
Farecast might proceed like this:
• A month's worth of fares. Farecast shows you the lowest fares
currently offered by airlines on each day within any 30-day period you specify.
You first select the "Plan Trips" tab on Farecast's home page. If you choose
Boston and Los Angeles as your airports -- departing from Boston between July 1
and 31 -- Farecast charts the lowest fares that are available on each date.
Since this period includes the busy Independence Day holiday in the U.S., prices
soar to a high of $577 on July 1 before dropping to $362 on July 15 and $361 on July 18.
• When to fly, and when to buy. Having decided that July 18 would
be a good day to depart, the question becomes: Will the airlines lower their fares
before then? When you provide a return date of July 25, one week after your
departure, Farecast informs
you -- with a "confidence level more than 80 percent" -- that the lowest fares
will rise $50 or more within the next 7 days. That's a good reason to buy your
tickets now rather than waiting. If the price was predicted to decline, by
contrast, waiting a few days could save you money.
• Airline fares vs. travel-engine fares. Once you've decided on
your plans, you can select an itinerary and jump directly from Farecast to a
site where you can purchase the tickets. Farecast shows four flights for $362 on
July 18, for example, operated by America West Airlines.
Farecast is unusual in
that it shows prices from both airline Web sites and travel-engine sites. In
some cases, the travel engines buy blocks of seats that give them prices or
times you may prefer. Farecast's business model is to collect the commissions
paid by travel engines for referrals (which the airlines generally don't pay).
Farecast offers a surprising number of options in addition to merely selecting
your cities and dates. Down the left side of Farecast's results page, a set of
check boxes allow you to narrow your desired departure and return times, number
of plane changes, and acceptable airlines. This is a trip planner's dream come
The Technology Behind Farecast
In an interview, Farecast CEO Hugh Crean said the company's methodology has
evolved out of studies of air fares conducted by University of Washington
computer science professor Oren Etzioni. That spark was enough to attract $8.5
million in venture capital from Silicon Valley's Greylock Partners and
Washington State's Madrona Venture Group and WRF Capital.
Farecast now has 22 employees, according to marketing VP Mike Fridgen, who adds
that about one-quarter of those are Ph.D.'s The rest include a few professionals
who formerly worked in the airline industry in fare-setting capacities.
One such employee is David Pelter. He's the former managing director of revenue
management for Alaska Airlines, a major carrier operating about 1,000 flights a
day, largely serving the West Coast of the U.S., Canada, and Mexico. Pelter is now
Farecast's VP of supplier development and revenue management.
He points out that airfares are typically available for purchase from the major
airlines as much as 330 days (11 months) into the future, but that isn't when
prices are lowest.
"As airlines see how demand is materializing," Pelter explains, "they may launch sales
to stimulate travel on weaker routes or during weaker times during the year to
'fine tune' their expected passenger loads. Most of this sale activity occurs inside
of a 90-day window prior to departure."
Fridgen opines that, "15 to 35 days [before a flight] is where we see a lot of activity, volatility."
Farecast has obtained millions of fare records, requiring terabytes of
storage on the company's servers. The analysis of this database is what allows
Farecast to predict whether fares will rise or fall in the next few days.
It's Called a Beta for a Reason
While Farecast exhibits a remarkable technology that actually improves on
human decision-making, it's barely out of the beta stage. The service suffers
from a few maddening limitations, which may frustrate you until they've
been worked out by the company's developers:
• Two airports to start from.
At this early date, Farecast predicts fares from only two airports: Boston and
Seattle. Crean says most major airports in the U.S. will be added to the
service by the end of 2006.
• Domestic flights only.
fares only to other U.S. cities, a limitation that may exist for quite some time. The system can predict only domestic
pricing at this point, omitting international destinations.
• 3-month horizon.
Because airlines mainly lower prices only within 90 days before
a flight, Farecast doesn't make predictions if your travel dates are more than
three months out.
• 2- to 8-night stays.
To get predictions of the next week's rate hikes and discounts, you must currently
be planning a stay of 2 to 8 nights. You can get prices for trips of longer
duration, but not predictions of dips and hikes.
In addition to the above hassles, users of Farecast are stuck with a dilemma if the
service predicts that fares will drop. How are you supposed to know the best date
to actually buy the ticket? At present, you pretty much have no choice but to return
to Farecast every 24 hours to run your search again.
An ideal solution would be for Farecast to obtain a "hold" of your itinerary.
Several carriers, including Alaska Airlines, allow a hold for a 24-hour period
without the purchase of a ticket. If the price rose the next day, Farecast could
buy the ticket before the hold expired. If not, the hold could be released and a
new hold established until the price did change, either up or down.
That procedure would probably violate all kinds of airline policies, although I
hardly think it's illegal, immoral, or unheard of. As an alternative, Fridgen
tells me Farecast is planning a notification service, via e-mail or other means,
that will inform you when a fare has changed and when to take action.
Despite its limitations, Farecast is a radical step toward greater control of the
buying process by the public. If its predictive ability is as good as it
appears, Farecast can shift some "revenue management" power back to consumers
that was lost when carriers perfected computerized fare-setting.
Buying airline tickets may seem more like an art than a science sometimes. That's
why I plan to take advantage of Farecast's computerized tools to even the odds
when I book my next flight.
To try it yourself, visit