Before, air travel had certain rhythms. Business travelers flew out on Monday mornings and back on Thursday evenings, filling pricier seats. Come summer, price conscious leisure travelers took to the skies. Crowds flew for Thanksgiving, Labor Day, and Christmas, and to specific destinations for events—sports championships, music festivals, fashion weeks. Decades of historical data plugged into complex mathematical models helped airlines determine schedules and prices.
Then came the pandemic. “All of the history, all of the old practices that airlines used to follow to decide what was scheduled to fly and what prices to charge, had to be thrown out the window,” says Jim Barlow, vice president of strategic consulting at Amadeus, which builds software for airlines.
Now, as more passengers are vaccinated and willing to travel, the airline industry is seeing green shoots. More than 2.1 million people traveled through US airport security checkpoints on July 5, nearly twice as many as last year; but that was still 20 percent fewer than in 2019.
That doesn’t mean that the pictures created by airlines’ algorithms have gotten any clearer. Airlines are operating with less data, and more uncertainty, than usual, creating a complicated math problem. It’s not just figuring out where people want to go, and how much they’ll pay. It’s also making sure that the right-sized aircraft and full, rested crew are in the right place for takeoff. The number crunchers who run their systems have found other ways to cope.
For about six months at the outset of the pandemic, many airlines leaned less on their algorithms and more on their human scheduling and pricing teams who used hunches about where people wanted to go, says Barlow. They froze hiring and laid off thousands of workers. Some put aircraft in storage, and photos of Delta and Southwest planes parked in the California desert became a creepy, pandemic-era sign of the times.
Part of the problem was that their customers had changed—and continue to change. The airfare-setting process is one of the most complicated in the business world. Passengers on the same flight, and even in very similar seats, often pay different prices, depending on where they bought their tickets and when. In-house teams create pricing structures and schedules based on when passengers are likely to buy tickets. Vacationers, seeking deals, tend to buy early, which is why airlines tend to offer the lowest prices on tickets bought far in advance. Business travelers, meanwhile, buy closer to flight time, and are willing to pay more.
Since the pandemic hit in early 2020, most people flying tend to be leisurers. And they were booking closer than usual to their travel times, probably because they weren’t sure how the coronavirus would affect their plans.
The influx of vacation flyers has changed airlines’ schedules—and made them more willing to experiment with routes less traveled. In the past year, JetBlue added routes to the Carribean. United premiered nonstop flights to Florida, and its popular domestic vacation spots. As business travel continued to sag, airlines subtly pivoted away from the big, traditional hubs to quainter routes: Milwaukee to Las Vegas; Boise, Idaho, to New York; Des Moines to Portland, Oregon.
As the routing experiments continue, airlines and the people that build their pricing systems are testing other data sources to make better operational decisions. They’re using customers’ web searches and requests for online notifications to suss out what’s in demand. Did a bunch of people sign up for notifications for cheap flights to Vegas in November? Maybe airlines should schedule a few extra flights that month. In the future, Barlow says, airlines are hoping to integrate other sources of information into their operations, like cellphone data that tells them how full competitors’ flights are, in real time.
“Dynamic pricing”—targeting specific fares to specific people, based on their flight history and real-time market conditions—has also picked up during the pandemic, with airlines imitating e-commerce companies in changing prices based on live demand. Since the 1980’s, airlines have varied seat prices according to tightly prescribed schemes, selling blocks of tickets at predetermined prices. But dynamically priced tickets can be changed all the time. For airlines, it’s a holy grail because it promises to near-perfectly predict the prices customers are willing to pay. Research suggests that more accurate pricing, not just of seats but also goodies like meals and extra legroom, could boost revenue anywhere from from 5 to 15 percent.
United, Delta, and Spirit Airlines did not respond to requests for comment, and Southwest and American Airlines declined to discuss future pricing. But in a recent call with investors, American’s chief revenue officer said that the airline had become “materially more efficient through the pandemic,” noting that the airline was flying 150 fewer planes than normal, but operating as if it were 80 to 85 airplanes down.
Of course, other industry-related dysfunction has complicated the picture. Bad weather, the occasional IT-related meltdown (Southwest Airlines delayed 1,400 flights in mid-June because of problems at a weather data provider), and now, a shortage of pilots and flight attendants add to the airlines’ headaches. After the big layoffs, airlines laid off large numbers of employees are finding that some employees are retiring and others are reluctant to come back to work.
Airlines didn’t use to work this quickly, says Bryan Terry, who leads the global aviation practice at the consulting firm Deloitte. Before, they would have worked out schedules months in advance; this spring, many quickly added new flights to what would have once been thought of as middle-of-nowhere destinations like Kalispell, Montana—the gateway to Glacier National Park. As a result, industry-watchers say, a pandemic that broke airlines’ internal systems might end up making them more dynamic and nimble. “Data analysts and data scientists are going to be in high demand and working overtime in most airlines,” Terry says.