This article originally appeared on TravelandLeisure.com.
Searching for airfare online is a special kind of torment. According to a recent study by Google, the average person visits 15 websites, does seven research sessions, and takes three weeks to book a single flight. Now a crop of apps and websites, led by Kayak and followed by newcomers Hopper and Flyr, are promising to remove the doubt by delivering customized fare forecasts—letting you know, once and for all, when and if that price is good.
Price forecasting is hardly an original idea. Note Microsoft Bing’s prediction tool, which folded last year. The trailblazing algorithm, originally developed by veteran airfare- intelligence company Farecast nearly a decade ago, used historical and active data to offer simple Buy or Wait recommendations for any given route. It was one of Bing’s signature features—until its primary data source was acquired by Google. Rather than pay monthly fees to its direct competitor, Bing abandoned the predictions.
But another major player was ready to take the spotlight: Kayak. In 2013, a company-sponsored hackathon inspired programmers to develop a feature for the site called the price forecast tool, which gauges whether prices will rise or fall for a route within the next seven days and offers a corresponding percentage-based “confidence rating” for its predictions. (The price forecast, now available only on the Web, will be added to Kayak’s app later this year.) The tool draws proprietary data from Kayak’s 1 billion annual user searches and historical trends from third-party provider ITA Software, and uses machine learning to identify patterns in airfare fluctuations. Currently, about half of Kayak’s routes offer predictions, depending on their popularity and whether the data is statistically sound.
While Kayak makes farecasting only a small part of its offerings, a handful of other services are focused exclusively on using Big Data to deliver even better airfare intelligence. Founded in Montreal in 2007 but launched only in January of this year, the Hopper app searches for patterns across a vast amount of information: up to 20 million daily searches from most online travel agencies, as well as data taken from eight years of archived analysis, all gathered during the app’s development stage. As a result, Hopper can deliver incredibly precise forecasts for hundreds of thousands of routes. It aims to tell you not just which prices will rise—but the exact date when they’ll start climbing.
Flyr, meanwhile, understands that the cheapest price does not always make for the best flight. Originally founded in 2012 as Hubskip, it began offering its predictions late last year, before launching a standalone app this spring. Unlike its competitors, Flyr allows consumers to track specific flights rather than entire routes (which can include non-direct flights with inconvenient layovers). Such targeted results make Flyr the most ambitious prediction tool of the bunch, and the challenges show: though it has managed to build a comprehensive database of domestic routes, Flyr’s international coverage is still thin.
Just how accurate are these services? In our tests, they performed fairly well, saving us nearly $200 on one flight, but occasionally steering us toward higher prices on other routes (see below). There’s definitely room for improvement—which should be coming. Kayak is considering broadening its forecaster to include hotel bookings. And more airfare predictors are on the horizon. “The potential for this technology is enormous,” says Azim Barodawala, CEO of Volantio, the parent company of flexible fare-search engine Adioso, which has been developing its own tool for complex itineraries. Farecast founder Oren Etzioni, who is now CEO of the Allen Institute for Artificial Intelligence, is not surprised to see so many new services picking up where he left off. “The pain point is still there: airlines’ prices are still fluctuating wildly—and consumers still want to save money.”
How the players stack up:
To measure the accuracy of each service, we tracked the same 10 trips every day for two weeks. Each afternoon, we recorded the lowest price on offer, along with the accompanying “buy” or “wait” recommendations. Here’s a snapshot of how the three services performed for one route—and our assessment of their overall success across all 10 routes.
How far in advance should you book?
Predictive tools can tell you when to buy, but you also need to know when to start your airfare search—and that depends on where you’re going. Kayak looked at a year’s worth of data to see when fares were at their lowest for different regions around the world.