It’s no secret the Fall weather (short of an errant hurricane) makes for the best flying weather. And this September was no exception. Every big airline in the US had more than 80 percent of flights arrive on time (except Spirit, which was close). Delta’s mainline operation passed 90 percent and canceled almost no flights. But digging into the data, we can see that not all airlines are running efficiently. Some (*cough* United *cough*) are padding to improve arrival times.
I dug into masFlight’s on-time performance data to get the details. Let’s start by looking at marketing carrier statistics. In other words, we’ll look at arrivals within 14 minutes (DOT’s metric of choice) by brand, including all regional carriers. (I’ve included US Airways in American’s numbers.)
Take a look at those numbers. Very impressive. What stands out? With the exception of American, nobody canceled more than 1 percent of flights in the month. That is fantastic. And wait, is that Southwest at number one for on time arrivals?!? We haven’t seen that in ages.
While we’re here, take a look at Frontier. It canceled very few flights and was toward the top in on-time performance. Even Allegiant did well for itself. Spirit? Well those numbers would be good in any other month, but as usual, the on-time performance lagged everyone else.
Now this may be what matters most to people who choose an airline, but it’s fun to dig into the details further to see what kind of patterns might emerge. Let’s look at the big guys.
American
We saw that American canceled a higher percentage of flights than anyone, but where did that come from? Mainline ran just above 99.5 percent, so it was the regionals that brought things down with only 98.1 percent completed. Republic with all of its staffing problems brought the average down, but so did Piedmont and Trans States. This wasn’t just due to a single operator. It could reflect a scheduling issue.
On-time performance for mainline was just above 85 percent. Most of the regionals did a respectable job… except for Trans States which ran only 68.7 percent on time. That is awful. Everyone else was in the realm of respectability.
Delta
It’s hard to explain just how good of an operation Delta is running right now. Its completion percentage was good, but when you look at just mainline? It was nearly 100 percent. Seriously. Of the more than 83,000 mainline flights in September, less than 50 canceled. That’s a 99.95 percent completion. Amazing. Oh, and on-time performance? Mainline alone ran 90.3 percent of flights on time, blowing away anyone else.
What’s interesting is that Delta’s regionals stepped up as well, running more than 85 percent on time. GoJet, owned by Trans States, didn’t crack 70 percent. On the other end, Republic’s Shuttle America subsidiary actually did well. It operated 99.2 percent of flights and 84.6 percent were on time. Considering Delta just sued Republic for not living up to operational requirements, this seems puzzling. I’m guessing that Republic said it couldn’t operate some flights far in advance so they were removed from the schedule.
United
Looking at the regionals, Republic is the standout screw-up here. Only 92.8 percent of its flights operated. Yikes. Perennial cellar-dweller CommutAir surprised me. It had 98.6 percent of flights go, with 81.9 percent of those arriving on time. But let’s dig in to mainline.
United actually did well on the surface last month. Mainline had 99.6 percent of flights operate with 85.9 percent running on time. But something didn’t smell quite right.
Take a look at this chart comparing on-time departures to on-time arrivals (exactly on-time, not within 14 minutes).
As you can see, both Delta and US Airways get airplanes off the gate on-time quite often. But that’s where things diverge. Delta has an even higher percentage of flights arriving on-time while US Airways sees a big drop. That tells me that US Airways may be scheduling its block times too tightly. Loosen those up and US Airways could be back at the top. Airplanes are ready to go on time, and that’s the basis for a good operation.
Notably, American’s departure performance is coming up significantly and passed 70 percent. It saw a higher on-time arrival rate. This makes me think that the management team is putting more slack into the American system until it can perform on the same level as the US Airways system, though they may need to meet somewhere in between. But the trends look pretty good.
Southwest, however, looks different. It had a harder time getting airplanes off the gate but was able to make up a lot of time in the air. Southwest appears to be putting slack into its system in order to improve its performance. That works, but it’s costly.
Then look at United. It was absolutely awful at getting airplanes off the gate on-time, a full 10 points even behind American and 5 behind Southwest. But look how much time it makes up in the air. It looks like United has put a ton of slack into its system. That’s inefficient and despite all that effort, United has only been able to get into the middle of the pack.
The end result? From an operational perspective, Delta and American are doing a good job of having their airplanes ready to go on time more often, though American probably needs to adjust its schedules to improve arrival performance. United and Southwest, on the other hand, aren’t ready to go as often and have added a bunch of padding to mask that. As a traveler, that’s fine since you’re still getting there on time. But it’s not efficient and it is costly. They both need to improve.
35 comments on “September’s Operational Performance Was Great, Though Some Airlines Are Padding To Make That Happen”
Why is adding slack into the system inefficient and costly?
More slack/padding means the planes are actually in the air less, which costs United money.
After all, the actual airtime for a city pair remains the same amongst carriers, so if United’s schedule makes up for that by the flights being “longer,” they’re covering for slower baggage handling, ground operations, boarding, etc.
Sure you don’t want to schedule a flight to leave too much before it needs to, so that it can land at the scheduled time. However, if I ran an airline, I would want to minimize my scheduled turn-time and maximize my scheduled block time. Does anyone know if United has shortened their scheduled turn-time to offset their longer block time to maintain the same aircraft utilization rate? Things can get delayed before or after departure. With a shorter scheduled turn-time, you have the chance at departing earlier most of the time to mitigate a post departure delay. And most of your passengers arrive early anyway.
Jeremy – As other have stated, it’s inefficient and costly because your airplanes are expensive assets. You want to keep them flying as much as possible, not sitting on the ground. If you have to pad, then you aren’t going to get the utilization you want on those airplanes because you’re overscheduling the amount of time it’ll take on each flight. It also makes it tougher from a hub perspective. You may need to keep your banks further apart than is ideal because you’re scheduling your flights to take longer.
What hasn’t been said much is that flight crews are generally paid on a “block or better” basis, so if you pad too much, you’re paying your crews to sit around. That costs money.
I had an interview at US Air back in the day, and one of the ops guys told me that airlines essentially buy their on time ranking. As we know, airlines can run on time with padded schedules, but padded schedules are costly. So depending on what the airline’s operational goals are for the quarter/year whatever, they can pay for the spot that they want.
“””””As you can see, both Delta and US Airways get airplanes off the gate on-time quite often”””””
Isn’t it if you close the door and pull back from the jetway what counts as departing within the time limit? All that means is they could have backed up three feet and sat there for whatever reason and not been considered late, right?
Well, except that would reflect in the arrival statistics. (Which, maybe, it does, given US Airways’ poorer on-time arrival performance.)
It’s still better than not even being ready to pushback, though.
David SF – For most airlines (the ones that use ACARS), it’s when the brakes are released that the clock starts. It’s true that they can kick the brakes and then just sit around, but that’s not a great strategy. For the most part, if you’re ready to close the doors and push back, you’re ready to get going.
Back in the stone ages of the 90’s I was an international CSR at ORD for AA. If we were running into boarding issues and we had a decent pilot it would help our performance. The pilot would often simply pop the breaks to send the message to ACARS that the flight was under way. It was a way to game the system. it didn’t regularly happen but it sure helped. Besides if I would have to take responsibility for a 15 minute plus delay I had to complete paperwork explaining what happened.
Regards this and the previous CF post: Oscar has been on CAL’s board since 2004 then UCH’s after the merger. As a member of the BOD, he has been a party to the strategy of the merger, the integration plan, the execution of the integration plan, all the many issues with the integration (still no flight attendant contract!), the adverse publicity, and the many customer negative comments. I could go on and on. The point is, how can the board not be aware of all the issues? The BOD had to be aware of and approve the use of consultants. This should have been a discussion topic with the number one issue being why there is no inhouse expertise ? I find it ironic that Oscar, with his board tenure, is now questioning the use of said consultants and just about everything else wrong with the airline. You can bet the consultants, who probably have little to no operational experience, recommended adding block time which can mask a lot of operational problems. Not only does block time increase costs (crew, maintenance, etc.) it can make the hub and spoke system inefficient as the connect times are compromised to the point of flights not being able to connect at the hubs with legal connect times, thus creating revenue issues and longer connections. One of Oscar’s biggest challenges will be to find the operational expertise on his team. The team seems to be composed of a lot of finance people in operational roles. The Board is also responsible for the CEO succession plan. I cannot believe this was there succession plan, what no internal candidates, big red flag.
Do you guys think AAs arrival metrics are taking a hit as they teeth on banking DFW,ORD & MIA?
Eric – American’s arrival metrics haven’t historically been good. I need to take a look, but my guess is AA is still doing better than it has been. The banking stuff doesn’t seem to have caused big problems, but now that it’s been a few months, I should dive in more.
Brett, is it possible to look further into Delta’s numbers? While very impressive and they are running a great operation, Delta pads its flights heavily. I fly DL out of IND and it is not uncommon to arrive at your destination almost 30 minutes early. This is great but doesn’t the over padding of numbers have it’s own downside as well? Especially when connecting or arriving at one of the busy NYC airports where gates are not always available.
Jonathan, I don’t think this is case – Brett’s numbers above demonstrate that Delta’s gap between its departure and arrival on-time rates is smaller than United, Southwest and American. With Delta’s departure rates, which are the best in the industry, there is no need to take the costly step of padding the schedule like United is clearly doing. While there certainly will be cases of early arrivals on occasion, it just doesn’t look to me like they are building in a lot of extra time into their schedule the way United does.
Jonathan – What other info do you want to see? This makes it look like Delta isn’t padding its numbers all that much.
That is not true CF. If someone was padding the schedule it would not be represented in the data you provide. It is the example provided by the previous poster that plays this out.
Your simply looking at those that depart late and arrive at A0. Those are the obvious padded schedules.
What could also happen is if a flight is 100% of the time at A0 and then is always at D0 then you would say it is a perfect flight, by your data. If we take the posters example, that does not mean a flight is not padded it just means they can operate better to A0.
Now I will admit my research is limited but on the EWR-ATL route, UA has a block time average (across all planned flights for today 10/8) is 154mins (or 2:34), but DL’s is 151 (or 2:31). Not a big padding but one of the few head to head examples I could quickly pull up. In this case UA is longer. Maybe this is not the best example, but it proves that DL’s ability to get to D0 is much better and should result in a much better hit to A0 no matter if padded or not. The spread you identify is true that UA pads it’s schedule, but it doesn’t prove that DL doesn’t either. Those are 2 separate questions that need to be asked. I would suggest you would need to see a blocktime average compared to stage length or something similar.
Kevin – Right, the metric that really answers this is B0, but I don’t have access to that in my tool. I can get that from masFlight, and have used it before. But let’s think about this rationally. Arriving early is not something that airlines ever want to happen. It’s operationally-disruptive. It’s more likely to happen for airlines that don’t depart on time consistently and want buffer. On the occasion that they do depart on time, they’ll arrive early. That’s what I’d expect to see happen at United.
But from Delta’s perspective, they’ve been running a great operation for some time now. For that reason it would be strange for Delta to have a chronic early arrival problem. You’re right that this metric won’t show that, and if I can get B0 pulled, I’ll let you know.
Shocked to see F9 so high on the completion list. Even at $9 fares I wouldn’t consider them because reliability is very important to me. May have to reconsider now.
I am very curious as to what the correlation is between padding and airport. Consider…
UAL hubs DEN, ORD, EWR, IAD, and SFO consistently rank among the worst US airports for flight delays. Conversely, SLC, SEA, and MSP – Delta hubs – rank among the best. The question I can’t answer is how much of a chicken and egg scenario this is: does the airport beget the airline delays and increase the need for padding or does the airline’s padding because of poor operational performance beget the airport delays? Or both?
Doug – Well, certain airports have operational issues, but then that wouldn’t be considered padding. That’s what’s required to run a good operation. If United was scheduling right, it would have about the same percentage of flights departing the gate on time and arriving the gate on time in the next city. If it’s Newark, it probably means you need longer block times than if it’s Des Moines, but that shouldn’t matter.
It makes sense that location of an airlines hubs can help contribute to good performance. Both MSP and DTW are both far better (operationally) than ORD and relatively speaking in the same neighborhood…but…the O/D traffic in Chicago probably is far more desirable than getting a couple % better on-time stats. Airline business is more than just one metric.
I had a mechanical going on AA yesterday from DFW to DTW. Left a half hour late and arrived one minute early. This morning I had another mechanical, another half hour late and arrived right on time..but then had to taxi forever. I have seen this time after time on AA. Only problem with padding like that is arriving 40 minutes early and then no gate available. Had that happen the other evening where we actually ended up being late after having sat there 40 minutes. Very interesting article
There are other ways to improve ontime statistics that cost money without padding the schedules: Buy more aircraft and hire more pilots. While airlines aren’t known for having too many “spare” aircraft sitting around, a much more subtle way to improve your D0 at the hub has to do with how many reserve flight crews you have sitting around. When your flight is running late because of “incoming crew”, you just call up that reserve guy and get the flight out on time, presuming the aircraft is there.
I’ve been working on the analytics side of this business for long enough, and I generally get a kick out of the accusations that an airline “pads” its schedules. Fact is, it’s hard to establish an emperical number of how long a flight is “supposed” to take.
The times listed in the OAG (known as “block” times) are gate to gate times. So you have the taxi out, taxi in, and airborne components. An emperical taxi out time is a bitch, particularly if you are scheduling for your hub. You own gates all over the airport, and are pushing out 50-100 flights at a time. If the first flight you scheduled out is delayed ever so slightly, he could get stuck at the end of the queue
It’s even worse for non-hub operators. Take a DL flight out of IAD. if it departs at 5pm, in the middle of the United push, his taxi out time is going to be longer than if he departs at 3pm when the airport is dead.
Enroute, airport configuration and congestion can have an impact as well. Take flights coming from ATL to IAD. When IAD is in a north configuration, ATL arrivals enjoy a time advantage. When IAD is in a south configuration, the flights have to fly way past the airport in order to do a 180 and come back to land. This can add a lot of time compare to a north configuration. When you’re doing your scheduling months in advance, what time do you pick? You can split the difference and be wrong by a certain percentage all of the time, or you could pick one configuration, be right when the winds favor you and be way wrong when the winds don’t.
Airline scheduling is hard to do right.
Those are good points. The amount of time that it takes to travel by any kind of transportation from one place to another is always going to be subject to some degree of variability.
Let’s take 13 recent QF93 flights, from Melbourne to LAX, as an example. If a ballpark estimate of when that flight will touch down, based on a small sample of FlightAware data, is good enough, I’d say 7:26 a.m., which is based on the median. But if someone planned to make a connection and wanted to be confident of making it, I’d suggest starting with the 90th percentile landing time of 8:31 a.m., and allowing a few hours for immigration, customs, security, etc.
Must be hell being in a business where the marketing side makes a strong case for basing your calculations on medians, and the pressure to keep passengers moving without having to rebook them pulls you toward the 90th or 95th percentile.
Dan – Of course when you look at a single flight, it’s impossible to know if a schedule is padded because of all the variability in the system. But when you look at a month in aggregate, then you can see those trends.
A September without significant weather events is about as good a month as you can imagine for getting good operational results.
Domestic ASMs are down about 10% relative to, say, July, so there’s simply less flying. That will tend to be even more true on some key markets like Florida, which is truly dead in September — and that, in turn, will have a noticeable impact on congested east-coast corridors. In other words, some of the most congested operational areas get the most relief in September.
Further, there are a lot fewer thunderstorms in September than there are in July, so there’s less delay-causing weather. The only exception would be when there’s a hurricane, but there were no hurricanes in September that affected the US.
Point being, this kind of September is almost optimal in terms of flattering airline operational capabilities. July is a much more meaningful test.
Is Hawaiian getting worse, or is everyone else just getting better?
For a while they were consistently on top in the on-time rankings, but they were running a stellar interisland operation (not the easiest to do even with the generally good weather due to a high utilization schedule with short turn times) but a sub-par long-haul operation. But the sheer number of on-time interisland flights overwhelmed the handful of delayed long-haul flights.
David M – Good catch there. Apparently Hawaiian knows this are bad because it put out a release explaining what’s going on today: https://newsroom.hawaiianairlines.com/releases/multiple-factors-diminish-hawaiian-airlines-punctuality-in-august
There is actually a major benefit to padding schedules. It enables the pilot to say “look, we got you here early!”
I have heard comments like “Southwest usually gets you there ahead of schedule”.
This helps their brand perception.
The data isn’t there to tell how padded a schedule is for the airlines with good d:0 times (like DL & US) you’d need early arrival time data instead of a:0 to make that comparison.
While Dan is correct about the many at airport variables are difficult to be exact for on ground time, there is lots of historical data for most hubs at various levels of activity to make a planning decision on. An accurate block time which is an estimate by definition just needs to know estimated taxi time at departure point, est flight speed and historical wind information for flight time, and then estimated taxi in arrival city. All is data that is available and all are subject to a variety of factors that could significantly effect an individual flight but assume the estimates are close to accurate will provide good times in the aggregate
I rode high speed trains all over western Europe earlier this year. They are literally the definition of reliability. People schedule sub-10 minute connections between two hour legs and don’t even bat an eye. Once, the conductor actually apologied over the PA for a four minute arrival delay…
Not totally on topic, but I take it you didn’t take the train in England @AW… Trains are far from reliable here: http://www.theguardian.com/uk-news/shortcuts/2015/jan/13/late-arrival-brighton-to-london-victoria-southern-britains-worst-train
I’m a little bitter; my own train journeys have not on time for the last three weeks in either direction… And they’ve been this unreliable for years.
Closer to “on topic” – I like the padding on the schedules – I’d much rather arrive a bit early on a padded schedule than a bit late on an optimistic schedule.
RyanAir seems to have figured this out, and they regularly tout being “Europe’s most punctual airline.” On every flight I’ve taken with them (a fair few now), I’ve arrived into the destination on time or a bit early, which they make sure to announce with a little trumpet fanfare!
As for the AA on-time departures and arrivals at least out of Seattle we usually have been getting them early (At least on my night shift) which gives us more time but due to our new location the upper D gates we have trouble getting them out on time. And don’t get me started on mornings I have heard its a mixed bag of leaving way early or leaving a whole lot of gate checks behind.
Back in the early 00’s I flew IAD->SFO. It was a five hour flight and always late. Then mysteriously one day it became a 6 hour flight and started arriving on time and sometimes early. Science or Magic ???