Air travelers are accustomed to long fl ight delays andcancellations for any number of reasons. Few customers realize thatairlines themselves are not particularly accurate at predictingwhen a fl ight will arrive at its destination even when it is readyto leave the gate. Making a pinpoint-accurate prediction on gatearrival times is notoriously tricky, because many factors alter flight times. Weather and wind are the most common, but there arealso ground issues, such as the passenger who neglects to board hisfl ight on time, forcing the airline to delay departure while theyoffl oad his luggage. As a result, airline predictions are off byan average of seven minutes across the industry. Flights normallyoperate according to a fl ight plan put together a few hours inadvance of a fl ight’s scheduled departure. After the fl ight takesoff it is tracked by a dispatcher, who may be monitoring 15 flights simultaneously. So, for example, if headwinds increase, thenthe pilot must talk to a dispatcher, who may decide to reprogramthe “cost index” of the fl ight, revise the fl ight plan, and givepermission to the pilot to pick up speed (and therefore use morefuel) to arrive on time. Airlines have been looking to automatethese kinds of processes to save costs and also to providetravelers with a better fl ying experience. Gary Beck, vicepresident of fl ight operations for Alaska Airlines, maintains thatairlines need to eliminate the human part of these communicationsin favor of automation. To encourage this process, Alaska Airlines(www.alaskaair .com) and General Electric (GE) (www.ge.com)sponsored a Flight Quest contest aimed at developing an algorithmthat could help airlines better predict fl ight arrival times andreduce passenger delays. The contest, which was set up on thecontest Web site Kaggle (www.kaggle.com), provided contestants withtwo months of fl ight data, such as arrivals, departures, weather,and latitudes and longitudes along the routes. Such data aretypically not available to the public because they are owned by theairlines and manufacturers. A team from Singapore won the contestand the $100,000 prize. The winning algorithm produced fl ightarrival estimates that were nearly 40 percent more accurate thanexisting estimates. The algorithm could help airlines reduce gatecongestion, manage crews more effi ciently, and save travelers upto fi ve minutes at the gate. Each minute saved in a fl ight saves$1.2 million in annual crew costs and $5 million in annual fuelsavings for a midsized airline. A second Flight Quest contest, witha $250,000 prize, challenged data scientists to determine the mosteffi cient fl ight routes, speeds, and altitudes at any moment,taking into account variables such as weather, wind, and airspaceconstraints. The winning model proved to be up to 12 percent moreeffi cient when compared with data from past actual fl ights. GEplans to develop software and services that incorporate the resultsof the two Flight Quest contests. It is important to note that GE’sgoal is not to replace pilot decisions, but to create smartassistants for pilots. It may take some time, however, beforesoftware can be used to fundamentally change how commercial flights operate. For example, Alaska Airlines, which frequently landsplanes under diffi cult weather conditions, has pioneered the useof satellite navigation, as opposed to relying on ground-basedinstruments. The use of satellite navigation lowers the standardminimum elevations for a plane’s approach upon landing. The airlineis working with the U.S. Federal Aviation Administration to spreadthe technique, which also saves fuel, to the lower 48 states. Beckclaims that the greatest challenge has been to alter the practicesand offi - cial handbooks of air traffi c controllers, who would nolonger need to tell planes where and when to turn. He maintainsthat satellite navigation systems essentially transform air traffic controllers into air traffi c monitors.
1. Do you think that satellite-based navigation willmeet resistance among air traffi c controllers? Why or whynot?
2. Do you think that pilots will object to having “smartassistants” help them make decisions? Why or why not?
3. Do you think the overall response of the airlines tosatellitebased navigation and smart assistants for pilots will bepositive or negative? Support your answer.
4. What is the relationship between analytics and smartassistants for pilots?
