The Story of American's DFW Schedule Changes
To bolster on-time performance, American will shift from nine banks to 13 at its biggest hub. The airline is betting it can improve reliability without compromising revenue. But that's hard to do.
The Airline Observer offers discounted subscriptions for companies and airports. Contact me for pricing.
Dear readers,
I’m not sure I can sugarcoat it: American Airlines had a really tough summer at Dallas/Fort Worth. Only about 61 percent of flights arrived within 14 minutes of their scheduled time in July, and the airline struggled to transfer connecting bags.1 “We didn’t have the operation we wanted,” Brian Znotins, senior vice president of network planning, said in an interview.2
When summer ended, American’s top executives asked Znotins if he could build a DFW schedule that would improve operational reliability and still capture roughly the same amount of revenue. That’s really hard because airlines typically must push through as many departures as possible to spur lots of efficient (and lucrative) connections.
At least initially, Znotins wasn’t sure he could do it. Then he examined his plans for the 2030s, which is when American expects its Terminal F (with 31 new gates) to be fully open. With all those new gates, American will have well over 1,000 daily departures, and that was going to require a new way of thinking about schedule planning. So Znotins asked his team: could American implement that plan earlier at DFW — as soon as this coming summer, when it will have about 940 peak-day departures — and not bleed revenue?
“I was initially reluctant to do it, but when we dropped it into the models, we looked at it and we saw that we’re still building the same number of connections,” Znotins said.
So American is stealing from the future for this new DFW plan: beginning in April, it will maintain 13 banks, up from nine. American also is dropping its omni-directional banks, so airplanes will flow from west to east, or east to west, depending on the time of day — a move that should simplify operations and boost connections in each bank.
This all seems simple enough, but there’s a potential drawback. Until the extra gates come online (allowing American to increase departures), some connecting itineraries will come with longer layovers. Maybe that won’t matter in peak season, when American has more demand than it can handle, but it could push customers to other carriers at other times.
“We’ll get a look at bookings over the next couple of months and see how the bookings are coming in at DFW versus our competitors, and it will tell us ....do these longer connections, offset a little bit by shorter flight times, impact our revenue?” Znotins said, noting that this new regime starts in April so that American can start to gauge customer behavior in the shoulder season.
“What we haven’t decided yet is if we’re going to run this structure into the fall, when demand goes down and flight levels go down, because the operation will be better able to handle it,” he said. “Demand will be lower in the fall, which would say that maybe nine banks is the right answer for August-forward and 13 banks is the best structure for the summer.”
I enjoyed speaking with Znotins, a schedule and airline geek (in the best way). We chatted about the intricacies of running a massive connecting hub in a complex that opened five decades ago to serve a growing local market.
Let’s get into the details.
How American expects the connecting mix to change
Even if American loses no revenue by adding extra banks, its revenue mix will change, as schedules improve in some markets and worsen in others.
First, the good news. American will create 4 percent more O&Ds in the new schedule. Also, American is cutting cut the number of pre-6 a.m. departures into DFW by one-third, because it no longer needs so many incoming passengers to feed its first bank. The first flight of the day still exists — it just may depart later — which should encourage bookings at better prices.
“There’s a big psychological barrier when a flight itinerary starts with an 05-something versus an 06-something,” Znotins said. “We even see a 0559 versus 0600.”
On the other side, customers may find fewer connecting options for sale through DFW on existing city pairs, as the number of those will decrease by about 6 percent. “Imagine — and I am making this up — Jacksonville-to-Orange County, we may have built eight connecting itineraries in the old structure, and now we’re only building seven,” he said.
Then there’s longer layover times. When Znotins says that the schedule won’t affect connectivity, he is including what he called ‘skip-a-banks,’ or passengers who arrive inbound to DFW in one bank but depart in the next.
That’s a reasonable assumption because with 13 banks, travelers won’t have to wait very long until the next one. But it also goes against industry dogma. For years, airlines sought to build the shortest connections so their itineraries would appear atop flight search rankings that prioritized elapsed time, but it’s not clear how much that matters anymore.
“Twenty years years ago, I would have been a zealot about elapsed time and display on GDSs,“ Znotins said. “I would have been like, [it’s] true [that] nobody buys a ticket because one’s six hours and two minutes and the other one is six hours. But if I fall off that first page on the GDS, that was a really big deal.”
Sure, GDS placement still matters. But most modern search platforms employ proprietary algorithms to rank flights. Google Flights, my preferred search site, shows me the “best” options, not the fastest one. “By default, Google shows you the flights that offer the best trade-offs between price, convenience, and ease of booking,” Google explains.3
American’s website does something similar. “It comes up with nonstop and fare and time of day — and not elapsed time,” Znotins said. “Everywhere I look, elapsed time doesn’t seem to be the second-most important thing anymore. It’s somewhere in the mix. And so I’m not as hyper-focused on squeezing those one- or two-minute improvements out of my itineraries than I was before.”
Customers may also prefer longer connections. I’ve heard anecdotes about people who book long layovers on purpose, fearing lost bags and misconnections, but Znotins put some data to it, using predictions from American’s model, which looks at the last two decades of passenger data to predict what people will book.
That model calculates that American should have about 9,000 passengers who book a 40-54 minute connection at DFW on an average day. “In reality, when we look at the actual demand we got for our flying versus what the model thought we were going to get, it was 4 percent lower than that, in that window,” Znotins said.
The model also predicted that about 19,000 passengers would book connections between 55 and 84 minutes. That number is 8 percent higher. “So what it says is, when people are buying their tickets, something else is helping to make their decision, not just elapsed time.”
To Znotins, the takeaway is clear: customers want a mixture of short- and long-connection options, and the new banks offer that.
What’s wrong with omni-directional banks?
Perhaps it’s just because I once connected from Rochester, Minnesota, to Los Angeles, but I’ve long endorsed omni-directional banks. People don’t always take the most efficient routing, especially from smaller markets with limited choices.
Until this change, American had some omni-directional banks. But Znotins said omni-directional banks are complicated to operate, because if aircraft arrive from the east and the west at the same time (and then also depart to the east and west an hour later), the banks can get really big and unwieldy, as the airline must schedule enough flights to feed aircraft in both directions.4
In recent schedules, American has used its third bank of the day to connect passengers in all directions. “Everything came into Bank Three, and then everything went out of Bank Three,” Znotins said. “And that was one of the operational challenges we had: sending too many airplanes to the runway all at the same time. And so we split Bank Three in directions — east and west.”
Directional-only banks should not materially change revenue, Znotins said. While American could lose New Orleans-to-Indianapolis customers (who first fly west to DFW and then northeast to Indiana), Znotins said the majority of American’s domestic customers fly only east or west.5 Plus, it’s a cleaner way to operate the hub, with less complexity.
“In the Jacksonville-to-DFW schedule, I can fly it seven times a day, and it will connect to every single Orange County flight that I have, because all the Orange County [flights] go out of the west [bank], and all the Jacksonville will come into the west [bank],” he said. “And I build seven itineraries a day. But if I had omni in that structure, then maybe I’ve got Jacksonville in some banks where I don’t have Orange County, and I don’t build seven connections a day, and it becomes three or four.”
No, this is not a rolling hub
In industry circles, the rolling hub generally has a poor reputation. While there’s no official definition (sadly, there is no dictionary for airline nomenclature), aircraft at a rolling hub typically come and go at random intervals. If the hub has enough flights, customers can build connections, but the schedule is not designed for them. Some frugal CFOs might like these hubs, because they cost less to run.6 But they stink (comparatively) for revenue.
Before its US Airways merger, American ran a rolling hub at DFW with what Znotins called a “very consistent level” of flights throughout the day. After Scott Kirby became American’s president in 2013, he re-banked it.
I wanted to make sure this isn’t a back-to-the-future moment. Znotins said it isn’t, because each of the 13 banks facilitates connections. Had operations asked him to go back to the old way, he said he would have pushed back against them.
“I am a firm believer that rolling hubs generally don’t work from a financial point of view,” Znotins said. “If we wanted to solve entirely for operations, it would be a rolling hub. If we wanted to solve entirely for revenue, I’d have nine banks with flights leaving within five-minute windows and sending 60 airplanes out to the runway in a five-minute window, and that’s not going to work either. This is really just moving along the continuum a little bit more toward the operational solution and away from the revenue solution, but going to a rolling hub would be a step too far.”
This DFW change should be good (relatively) for block time
When it announced its DFW changes, American said it would add block time system-wide to boost on-time performance. Znotins wouldn’t tell me how much block time American added, but he said that like most airlines, American sets block time based on the performance of past flights, usually so it can meet or exceed its scheduled arrival time between 70 and 80 percent of the time. With the extra block time, American will be more conservative and move closer to 80 percent. “You need to shift to the right on the distribution,” Znotins said.
Adding block time is expensive, though American could recoup some of the extra costs if it pays less to rebook passengers and re-route bags from irregular operations.
Still, American is adding less block time at DFW than elsewhere, because with fewer airplanes departing or arriving at any time, the airport should be less congested. With nine banks right now, American has up to 69 airplanes pushing back during a 30-minute window; starting in April, it should have no more than 56. That’s fewer airplanes trying to take off at the same time, so each should reach the runway faster.
“What I’m giving up on connection time in the hub, I’m actually getting a lot back in elapsed time because the flight times are shorter than they otherwise would be,” Znotins said.
With the new terminal (the first phase, with 15 gates, is slated to open in 2027), Znotins said he can further slash block time at DFW. Today, American has more gates on the east side of airfield than the west, so an airplane coming from Los Angeles or San Francisco might land on a western runway but taxi to a gate on the opposite site of the complex. Then the same airplane might return to California, forcing a long taxi time on reverse. “We have to schedule for that,” Znotins said.
The new terminal solves that problem, giving American a roughly equal number of gates on each side of the airport complex.
“If an airplane is coming in from the west and going out to the west, you want to gate on the west,” Znotins said. “It’s going to be much faster for everybody. They’ll get off the airplane sooner, more likely to make their connections, and it’s better for labor and block and all those kinds of things.”
That’s all for today. Thanks for reading. Assuming Scott Kirby says something interesting on United’s fourth quarter earnings call (and why wouldn’t he?), I’ll be back soon with some thoughts about United.
As far as I know, airlines don’t publicly break out how many bags they lose in each hub. But American had a tough July overall, ranking ninth out of 10 larger U.S. airlines for mishandled bags, according to DOT data. And out of all of American’s hubs, DFW might be the most challenging for bags, because it opened 50 years ago as an airport for local passengers. “Connecting bags at DFW is actually harder than connecting people,” Znotins said. “There’s no centralized bag system. All of our bags have to be run between terminals using people and carts. A true built-from-scratch connecting hub would have a centralized bag system where you just ingest the bag and then it goes to the terminal that you want it to automatically.”
OK. Fine. I’d like to know how much money American intends to lose this year in Chicago, or why it is content with having the lamest international network of the Big 3. But no one from American has sat for an interview with me since I started this newsletter so when I secured a discussion with Znotins, I opted to focus on this less controversial (but still very interesting) topic. Perhaps next time I can come out swinging.
I asked Znotins if Google tells him everything that goes into the secret sauce. If it did, I wondered if he could optimize the schedule for that algorithm so American would show up on top for all of its itineraries. He said Google does not share, and that can be frustrating. Sometimes, he said, the issue comes up with one of his itineraries to the South Pacific. “I’m like, ‘come on, I’ve got the best connection and the best price on this itinerary, and I’m not in the top three. Like, why is that?’ Google won’t tell us why.”
I wondered how American treats markets very close to DFW, like Waco and Tyler. If it’s within 200 miles of the hub, Znotins said the feeder flights don’t follow the east or west designation. “We think of those as omni,” he said.
It’s different with flights to Latin America. Those are often north-south connections, but American doesn’t use a strict bank schedule with many Latin America flights because it has fewer international gates than it needs. “We have to better utilize those gates, and can’t leave too much time sitting just waiting for a bank,” Znotins said.
American will see some cost savings here. Znotins and I didn’t focus too much on that, which makes sense, because schedule planners often focus more on revenue opportunities. But about cost, he said: “The costs of failure, from an operational point of view, are expensive. You’re putting people in hotels, you’re having to pay them for their lost bags, [and] you’re having to re-accommodate them on other flights, [and] that takes up capacity.”
Great stuff from both Brians. You tackled multiple operational and economic angles here.