American Airlines has new technology that’s enabled them to avoid cancelling 1,000 flights in just the past year. It helps them recover quickly during weather events, replacing manual processes with computers. And it changes how we should think about who’s at fault when a flight does cancel.
Last summer I shared details of American’s new internal Hub Efficiency Analytics tool or “HEAT” as they were calling it. They’ve automated the process of recovering after a storm – looking at available planes, gates and crew and figuring out which flights to delay, and which ones to cancel, in order to have the smallest impact to their overall operations.
This used to be a manual operation, but now computers are able to calculation millions of potential scenarios in the worst storms that last for days. American is now publicly sharing details of HEAT after a July 4th weekend where they performed well even in the face of weather.
Last May, American’s Chief Operating Officer told employees “you gotta own the airport to make it worthwhile otherwise you’re not making much of a change.” The focus of this tool is at their big hubs.
One thing that strikes me, though, is that when American announces that “[s]ince its initial deployment last year, HEAT has prevented nearly 1,000 flight cancellations across our network” they are saying that many weather-related flight cancellations are really controllable cancellations since they controlled 1,000 of them!
The entire airline industry model of what is due to a customer when a flight is delayed or cancelled hinges on whether or not what happened to the flight was ‘controllable’ – whether or not it’s the airline’s fault. They generally take the position that,
- Weather and air traffic control delays and cancellations are not their fault, and so not their responsibility to take care of customers
- Mechanical and crew-related delays and cancellations are their fault, so their responsibility.
This has led airlines to err on the side of blaming weather somewhere. American Airlines, though, now admits that they are not, in fact, powerless in the face of weather.
- When they’ve cancelled flights for weather in the past, they could have done something about it with better planning and technology. They just hadn’t done so yet.
- They’re making proactive choices about what is best for them in the event of weather, it’s their decision which flights get flushed.
This suggests a re-thinking of when an airline is ‘responsible’ for a cancellation might be in order. And as better technology and AI is developed and deployed, the current bifurcated model of controllable versus non-controllable may be on a collision course with itself.
I don’t think it’s fair to say these 1,000 “saves” were controllable cancellations. Clearly they were never controllable before but American has moved to a new realm at a higher level than ever before. They deserve to be congratulated and their reward is more completed flights and fewer food and hotel Comps.
We saw this working 2-3 weeks ago in the Northeast. United had a complete operational meltdown while AA was only mildly impacted with bad weather.
The event causing the cancellations is still the uncontrollable factor of weather. The effects are mitigated with better technology here but it’s still triage choices based on available information and the ability to process that.
It’s because this technology has only faced baby tests. If the same tech was used by United at EWR, how much better would United have fared compared to what actually happened? The main reason why AA didn’t have a meltdown like United did was because AA is hubbed ag JFK and LGA instead of EWR, and AA has a small presence at JFK compared to Delta and Jetblue.
AA and all airlines have had some bad news about them lately. Maybe they’re just trying to promote some good news.
You’re way overthinking “controllable cancellations.” If an airline can use technology to reduce cancellations, that’s great. If the economics don’t make sense to that airline to use these tools, the delays are still “uncontrollable.”
@Chopsticks – part of the economics is the cost to accommodate customers, which if they can call it uncontrollable they can squeeze to zero.
And here the economics is better IT to better simulate their decisions.
Great. Before we had airline employees lying about the weather and now we will have artificial intelligence lying about the weather.
@Justin do you know how much bad weather affects AA’s DFW hub every month? Their presence in DFW is double what United is at EWR and they’re able to recover within a day.
@Junipee DFW has many more runways and taxiways than EWR. Besides airports, everything is bigger in Texas. There is more airspace to work with in and around Dallas than the NY airspace. Not to mention saturation caused by LGA and JFK. DAL had minimal to no impact on DFW.
It seems to be a leap to say that mitigate the impact of weather is equivalent to removing the impact of weather.
Preventing over 1000 flights over the last year is great but shouldn’t you look at the overall percentage? https://www.newsweek.com/flight-cancellations-soared-past-last-years-total-1720888 says “Among the 10 busiest airlines this year, American Airlines has had the most cancellations with 17,138 through June 29”. So even if we say that all 1000 saves were only this year then that’s around 5% (and sure, not all the cancels were weather related but I don’t think that’s going to shift the percentage such that you can even claim that a majority of cancellations were avoided. Maybe someone with access to the data could do some analysis?)
UA has always had a program similar to this – at least PM UA did when I was a Zone Controller (aka Ops Manager/Ramp Tower Manager). The software made recommendations on which flights to cancel / delay during ATC flow programs & weather events based on load factors, crew pairings, # of downline connections, and aircraft routing. End goal was to keep the network flowing and minimize cascading disruptions. I think the scale of the EWR meltdown was beyond the capacity of said software… When crews can’t reach crew scheduling, and there’s airport gridlock with no place to move aircraft, it’s beyond the scope of any recovery tool.
End goal was to keep the network flowing and minimize cascading disruptions.
If this tool makes a significant difference then AA should open-source it to the other carriers. After all, better hub ops for the “owner” might also benefit the other carriers at the same hub. Might even benefit customers god forbid.
It’s absurd to argue that these 1,000 cancellations are evidence that weather events are controllable cancellations because the tool used to control the problem simply didn’t exist before. The airlines could control all weather related system disruptions if they could control the weather but we don’t have that technology yet
This may have been at work a month ago when a large group of storms wreaked havoc at PHL for hours. We were in MCO waiting for our plane from PHL to take us back. The 5 PM departure kept getting pushed back as the plane sat in PHL longer and longer. The next round trip flight also started to sit. We had to be back for the next morning and were concerned about crew timeouts. Our flight was getting less full as people rebooked. An Admirals Club agent came to us saying they had heard it was chaos in PHL and were booking passengers on our flight on a 6 AM flight, which we took (along with a hotel voucher). What ended up happening is the second flight was canceled and the remaining passengers were rebooked on our original flight. That flight ended up arriving back in PHL at 2:30 AM. This system probably canceled one flight and tried to clear space on the other.
Where are decisions made? The computer or does the computer provide options to operations or the airport staff? How is communication to the crews handled? This was one of the UA problems where crew members had to call in and were on hold forever.
This paradigm may help others but in June, they delayed my flight for 2.5 hours at 3.5 hours prior to departure. Then delayed it again. They loaded and taxied at 7 hours delayed knowing the runway had a long wait and then taxied right back (crew timed out) and posted another 9 hour delay and then 2 hours later added another hour for good measure. 4pm took off at 8am. But it wasn’t canceled!!!
On the way back posted a 4 hour delay 2.5 hours prior to scheduled departure. That’s when I ran to UA. A walk up fare was still $5 less than a fare booked may weeks prior.