In the summer of 2015, Delta was so proud of its operational performance that it put out a guarantee for corporate customers. If Delta failed to best American and United, then it would pay out. That didn’t seem like much of a stretch at the time, but when United followed in December? That was a shocker. United had improved its operation somewhat from the hole it was operating in for years, but this still seemed like a shaky bet… until I saw the terms. As expected, United crafted this in a way that meant it was highly unlikely it would have to pay out. And looking at the 2016 data, that appears to be the case.
You’re probably wondering why I’m not talking about Delta’s results here. That’s because Delta has created some very specific rules that use the Department of Transportation Air Travel Consumer Report. It’ll be later in February before those numbers are released, but there’s little doubt that Delta is going to be fine here. United, however, opted out of the many carve-outs and so it’s much easier to review using real-time data.
As a reminder, here’s how United put this together. It said it would meet one of the two following goals for 2016:
- United would have more flights arriving at the gate exactly on-time (A0) (or earlier) than American OR (not “and”) Delta
- United promised to cancel a lower percentage of flights than either American OR (not “and”) Delta
If it failed to meet either one of those goals, it would have to pay out.
Unlike Delta, [Update: I’m told that in 2016, Delta updated its guarantee to include international and regional flying] United used all flights worldwide, and it even included regional flights. Looking at those rules, it wasn’t exactly a huge stretch to think United could pull this off. And it did.
I took a look at masFlight to pull operational details for United, American, and Delta flights worldwide for all of 2016. To be clear, I don’t know what source United is using to get the official data, so this is just an approximation.
Let’s start with the cancellations. By my count, Delta had the fewest canceled flights at under 1 percent. United and American, however, were virtually tied at just over 2 percent being canceled. It looks like United may have squeaked ahead of American by a tiny bit, but really this doesn’t matter. Because remember, United only had to achieve one of its goals to avoid a payout.
If we look at on-time performance, United did indeed beat American pretty handily. I show Delta with just over 71 percent of flights arriving on-time or earlier. United was near 68 percent with American just above 62 percent. As I figured, United was quite deliberate in the metrics it chose. Take a look at this chart.
The bar on the right for each airline is the A0 metric that United used. As you can see, and as I’ve mentioned, it beat American and so it won’t have to pay out. But the bar on the left is D0, departures that go exactly on-time. In that case, you can see United did come in last. Why would that be the case? Because United pads its schedules more than American does. As has long been the case, it looks like the happy medium is somewhere in between.
American has really pushed hard on getting its airplanes out on time, but it doesn’t leave as much time for the airplane to get to its destination. So you can see the arrival rate drops. United has the opposite issue. It doesn’t get planes out on time as often, but it has plenty of time to make up for it along the way.
With that in mind, United knew it could likely control the numbers enough to avoid paying out on its operational guarantee. Sure enough, it did. And really, that’s a good thing. While agencies and corporates might have appreciated the payouts, there’s nothing they want more than an airline that gets their people where they need to be when they need to be there. United has stepped it up, so kudos to United for that. But there’s still plenty more work to be done in that operation.
United’s problem is that they chose to redefine the metrics instead of working with established metrics which the DOT already uses and which EVERY customer can find because they are published in the DOT’s air travel consumer report.
If you drill down in the data, Chicago, Denver, and San Francisco are problematic for United even if Newark has shown some improvement. In order to keep the mainline operation running when bad weather events happen in those cities, United’s regionals aggressively cancel. Where American overlaps at Chicago, the same thing happens.
And as much as American mgmt. talked about how rehubbing would help them run a more reliable airline, that hasn’t happened. Given that Kirby is at United now and says he will use the same strategy to increase connectivity and revenues, there is a good chance that United’s operational performance will not improve and may deteriorate in 2017. Heavily banked hubs in very busy airports simply are more prone to delays.
Delta has just NYC where regularly weather/ATC issues are problematic and they just don’t cancel as high of a percentage of flights during bad weather events and their on-time is still better than their peers in key markets.
The chances are good that Delta will be heading into some new corporate accounts this week to gain business based on its operational performance precisely because corporate customers want the most reliable service at the lowest cost. Delta operates its system measurably better than United and America as well as Southwest which you did not highlight here.
What metrics did United invent?
D0 and A0 are not stats the DOT uses. The only stat that matters in their reports is A15.
Tim – Minor nitpick, but it’s A14. DOT stats require flights to arrive less than 15 minutes late (so, 14 or less) to be counted as on time.
Yes you are right. Valid nitpick
I certainly understand the politics that can be involved with metrics, but UA didn’t redefine anything. While the DOT doesn’t track A0 and D0, they’re not convoluted things that are calculable only by some hocus pocus at UA and not reproducible. Anybody can calculate them from the full DOT data sets that are publicly available.
If those metrics are how UA wants to write their guarantee, I have no problem with that. In reality, those two things are what passengers notice anyway — the average Joe has no clue about this A14 statistic. They just want to know if they took off on time and landed on time. (In fact, the DOT has some skin in this game… airlines do code delays, and one of the codes is “ATC”. So, DOT buys themselves some political cover by measuring A14.)
A0 and D0 are realistically internal airline stats. Certainly customers look at their watch but the vast majority of customers have no idea how long a flight should take other than how an airline schedules a flight. If it is on-time but left very late, they assume the schedule was padded unless the pilot says something about how they made up time.
More significantly, most customers can’t query the DOT’s database; perhaps some large corporate accounts do but they rely on system level on-time data the way the DOT has been reporting it for years.
While airlines code the reasons for delays, popular media and even trade media doesn’t do the analysis to know why one airline has more delays than another, particularly in one city.
When UA reports its on-time performance using A0 and D0 compared to A14 as other airlines do, customers are even more confused because UA’s numbers “seem” so much worse.
As for Delta’s “overnight delays” they are following the DOT regulations but it was reported that they learned a lot in their relationship with Alaska. Further, Flghtstats.com does report average delays and Delta fares well even on that statistic so if they are intentionally taking lengthy delays to avoid a cancellation, it is not hurting them relative to other airlines “standard” delays.
Those stats are more than just internal figures, when I worked for a UAX carrier, part of the company’s financial compensation (and, at times, our profit sharing checks) was based on our D0 stats. To your point, UA makes a huge deal internally about D0, and has done so for years. If you were to say that UA places an extra emphasis on that D0 stat in ways that other airlines don’t, I wouldn’t disagree, but I’d ask what the big deal is. Their house, their rules.
However, it’s not like A14 is intrinsically better. Hell, of the three we’re talking about, it’s the most arbitrary of the three, and I’d concede any argument that says just because the government declares something, doesn’t make it right.
Again, I take no issue with which of these yardsticks UA chooses to measure their performance internally, or even against the competition. It’s not as if they’re picking a stat that only they can win with, or at least not lose.
Anyway, to contribute something a little more useful to this discussion: I don’t know what you mean when you say that most customers can’t query the DOT’s database. Well, ok, what you say is literally true but misleading. The underlying data is provided in a text file on their website, google “dot airline on time performance data” and you’ll get straight to it. Any high school kid can grab and crunch it. That’s why I take no issue with what UA is doing.
I agree with you regarding the metrics involved. Right or wrong, the DOT has established A14 as the most commonly used on-time metric; any airline that publicly uses a different standard creates a certain amount of confusion. Internally, I am well aware that airlines use lots of data and metrics and that data is available to the public but very well travelers know how to use that data outside of a few very large corporate accounts.
and as CF noted, the real issue is not the metric but that United chose to challenge Delta’s on-time dominance among the big 4 and appears that it did not succeed in that challenge regardless of the metric used. Even if it did beat American and did not pay out, United really did not leap past Delta in the process.
If the DOT would call a flight that is delayed over 6 hours or is delayed overnight when it was scheduled to leave prior to 9pm a ‘cancel’…. Deltas rankings would plummet. They are masters in delaying a 4PM fly until 9am when they can get another plane in.
I don’t take issue with that measurement. Several years ago (early 2000-2001, IIRC) UA was having an issue with their European flights, and often delaying them well into the next day.
This does become a question of “how/what do you measure” or “how do you define a cancellation?” If the capacity disappears, as in truly goes “poof” and the passengers have to be reaccomodated on already scheduled flights, then it’s a cancellation.
But if you cancel Flight #1, and replace it with Flight #10000, how do you measure the performance of the second flight? It doesn’t have a published departure time, so what do you measure?
TBH, if Delta truly cancelled the flight, in the real world, I don’t think you’d like that. You’d then have to be reaccomodated on whatever capacity is available, and you could wait a day or two to get out.
Delta’s dong it right. If they delayed you 6 hours, and you were never reaccomodated on a “new” flight, you were delayed 6 hours.
The key for me is that these guarantees are not for you and me but for corporate contracts.
If that is the case, the metric they use is only relevant to those departments, since we are talking about travel agencies or departments I am sure they have reviewed the metrics UA used and provided feedback.
This is not something that the average UA customer even knows about or even the individual customers of those corporate contracts.
I work for a large company and our primary carrier is not UA, but another major US carrier, and I can tell you that I don’t know what the details are and nor are they shared with me as part of my travel. Any breaks that might happen because of those guarantees will go to the company not me or even to my department expense.
To quote my Stat Teacher in College: “figures don’t lie, but liars figure”. Great article Cranky
Good analysis, Brett,
Shows how airlines never fail to fool the common traveler.
Thanks for all your effort.
Do you think that AA will start padding their schedules to improve their A14 numbers? What would an extra 5-10 minutes schedule time do to their overall operations?
Richard – Well, there’s some thought that they might pad more now that Kirby is gone. But we’ll just have to watch and see.
From a customer experience prospect, I kind of like the way United pad their flight. Imagine you are onborad a delayed flight, before departure the captain announcs, “folks, I am sorry that we are couple minutes behind schedule. But we will be able to make it up in the air and still arrive on time.” Meanwhile, onboard American, “folks, in order to depart on time we closed the door and left 5 connecting passengers behind…”
Does anyone know what the payouts are under the Delta and United “guarantees?”
Bill – The payouts are specific to each client, but Delta pays in the form of credit for future flights while United pays in the form of funds to be used for waivers and favors (waiving fees, assigning E+ seats, name changes, etc.)
While acknowledging that BTS measures A14, isn’t is better that we move to A0 as the measure? If you’re a corporate buyer, it is a number that means a lot more for reliability purposes–frankly (more than other reasons) for MCT purposes. Think of what it means if AA/DL/UA got to the point where A0 happens 85% of the time. That would mean a standard. able-bodied elite-member business traveler (meaning someone sitting in the first 12 or so rows of a plane and not reliant on accommodations) can reasonably make a connection in 30 minutes. I saw that because right now, I have found that from the front rows (essentially F or Y+) I can get to any other connecting flight at all the domestic hubs with possible exception of PHL in under 30 min, although MSP, IAH and DFW can push the limits of this.
For the airline, this probably opens ups several more connections to boost revenue. For the corporate buyer this means they can push their employees into tight connections that probably only add 65 to 80 minutes on travel time and have confidence that things will work. In a situation where (and I’ll lived this) where LGA-XYZ is $1000 but LGA-HUB-XYZ is $300 it can mean a lot to the corporate travel budget but only if the connection becomes less risky.
while it is, of course, ideal to run such an on-time operation, doing so is costly and more importantly it is a level of reliability that exists in virtually no other form of travel over comparable distances. I can’t even reliably plan travel time above 20 miles in large US cities within 15 minutes (perhaps other than via subways/train) so it is unrealistic to think that airlines should operate to such a high degree of reliability over much greater distances.
Airlines do know how long it takes to deplane aircraft and have to build connections for the person in the last row of the aircraft to any flight they sell as a connection. Airlines can’t operate a hub that is built entirely on minimum connecting times because there are a significant amount of flights that will be even a few minutes late in addition to the people who are at the back of the aircraft.
There are structural considerations such as the capacity of airports relative to airline schedules, design of terminals, and general weather patterns that give some cities advantages as hubs compared to others.
The market, not government, should incentivize carriers to operate more reliably and I strongly suspect that part of Delta’s revenue premium to the industry is precisely because there are corporate customers that choose Delta over other carriers because the chances of getting where and when their passengers are scheduled is higher. I’m certain that Delta’s on-time performance could be improved but as long as other carriers rank lower, those carriers could improve even more. United clearly believes on many levels that operating more like Delta will be good for their financials – they have said that. Given that they didn’t achieve that goal, it is hard to argue that tightening the standards will lead to a re-ranking of operational reliability or revenue.
I agree with you Tim. My point, generally, was that if operations run closer to D0 (as DL claims its trying to accomplish) then MCT (or MCT +15) can become more of a norm. That makes connecting itineraries more competitive. For all the talk about strong core markets and fortress hubs with strong O&D, the big three airlines have built a business that is really about connections; DL and AA (at least as far as the old US network was built) in particular. I’m arguing that with a more reliable operation (or a reliable but generous pad to the schedule) MCT can be made tighter, that in turn makes more connections open up and tighter connections can be priced to complete better with other airlines nonstop offerings.