Well, it’s Monday, so here’s another great guest post from my friend Courtney!
–
…Well, it’s prestigious to me anyway. This week we award the first ever “Earlier than the Other Guys” performance award.
I want to take a look at the big beautiful on-time database the DOT puts out monthly. For data nerds such as myself, the 550,000 record database is a gold-mine of operational information. I won’t spill all my secrets on what I do with the data, (mostly because my company pays me to do those things and to specifically not tell you) but I do want to show you some of the recent fun I’ve had with it.
We’re usually made aware of the reports each month in the form of a press release issued by the DOT. May, 2010 was the latest, and as usual, Hawaiian was at the top and Comair at the bottom. Since the island airline would get the top accolades every month, it always drove me nuts to see that once again, the carrier operating in the best weather and least traffic posted the best on-time performance. The East Coast carriers are stuck in airports like JFK, ORD, and PHL, while Hawaiian is in…well…Hawaii.
Of course we all know that JFK sees far more delays than HNL, and the graph above shows just that. As the shadows get long in the day, the delays grow at JFK, while in HNL there’s barely a blip. My quest this week was to strip away all of the location-dependent issues such as weather and ATC delays, and find out who really runs the sharpest operation.
My methodology was rather simple. I simply calculated the average delay at each airport for each hour of the day, and graded the airlines’ performance for each flight against that average. To dumb it down even further, I wanted to see how each airline performed against all of the other airlines regardless of where they flew.
The results are in, and I was surprised to say the least. The winner? United Airlines. In fact on average, each of United Airlines’ flights had shorter arrival delays than the industry average by 8 minutes. Compare this with the bottom airline, Comair, whose flights arrived an average of 7 minutes after the industry average. That’s a 15 minute swing. Other airlines to note are Hawaiian, which still performed above average, but slipped to number 8, and Southwest, typically towards the top of the on-time statistics, slipping to number 13.
While this does go a long way in leveling the playing field as we compare airlines’ performance, it is subject to some inaccuracies. I do take the effects of regionalization out by comparing to the average airlines operating at each airport, but the average delays are influenced by the airline we’re studying; such as with Delta in Atlanta. Nor does this account for the schedule padding we hear of in the industry.
To the passenger, this tells you that with United, you’ll arrive an average of 8 minutes before the rest of the industry. You should be in baggage claim by the time Comair shows up. Whether it’s schedule padding or not, it’s nice to see United most living up to their time commitments. Congratulations United on winning the first ever “Earlier than the Other Guys Award”. Might I suggest taking the non-stop 777 from ORD to HNL to rub it in?
Courtney is the co-creator of the Airplane Geeks Podcast, founder of AirlineEmpires.net, currently works for a commercial aircraft OEM, and is a self-proclaimed stud muffin. You can contact him through the Things in the Sky contact link.
You factored in average delay, but did you look at the average arrival time diff from scheduled for all flights, or just the average delay for those flights that were delayed?
I took the average time difference from all flights from scheduled arrival time, including those that arrived early. To illustrate the concept, if the average delay into an airport for the given hour was 25 minutes, United averaged only a 17 minute delay.
Likewise, if the average delay into an airport for a given hour was 0 minutes (or on time), United averaged system-wide an 8 minute early arrival.
It got a little difficult to transfer the concept from the numbers to descriptive words, but that example should help.
Do you have some numbers available to split out excessive schedule padding? For example, UA has long block times into congested airports, which times become even longer at peak times. I don’t know about other airlines’ practices but I know they differ widely. The results above are relevant only to reasonable certainty of arrival time vs schedule. However, likely flight time is different. It would be interesting to see if UA’s -8 minutes comes out nearer the level if it turns out that they schedule an extra 8 minutes for the flight.
To put it another way, if we want to see who runs an efficient operation (to be defined by having a serviceable aircraft ready to go and having passengers ready to go at the scheduled time the most often, or with the fewest average delays), then you must factor in block times.
You miss LAN airline. It’s one of the biggest in America.
Where did you get the data for this? I’d like to take a shot at this and all I can find (at least so far) is summary data…
Brace for the Lifehacker effect!! Lifehacker posted this, so prepare for the effects!
NB makes a great point. Airlines routinely add extra time to the flight time to even out their arrival times. This is why it is so easy to depart 45 minutes late and arrive closer to scheduled or “make up time” in the air. United must be scheduling more time for their flights than competitors, falsely inflating their on-time/early arrivals.
All that padding can cut both ways. American Airlines lists Austin to DFW as 65-70 minutes. Actual time between pushing back in Austin and wheels-down in Dallas can be a little more than 30 minutes if the winds are right. On those sorts of days, you end up sitting on the tarmac in Dallas for as long as you were actually in the air, sometimes longer, because you were “early” and the previous flight using your gate isn’t ready yet.
Excellent overview and data crunching exercise. Thanks for sharing this info as I’ve always jumped from Delta to United and wondered just who was the best. My United experiences have been a little better but for the most part they’re about even. I will say the most efficient flight I ever took was on United’s non-stop 777 from ORD to HNL that you mentioned. Round-trip we were early going both ways.
I agree with NB. Now that the FAA is fining airlines for delays, airlines are padding their schedules. Until recently, I was flying Southwest at least once a month. Their flights would regularly arrive 10 minutes early unless there was a delay, then they would arrive “on time.”
@NB: It would be great to know, but for the general consumer/air traveller I’m not sure it’s too important. I mean, if United DOES add in that 8 extra minutes for the flight wouldn’t that just show an acknowledgment of responsibility toward their costumer to get them from point A to point B as quickly as possible?
It may be more efficient to be, say Southwest, and not put that extra padding in, which could allow them to fly more often with not too much of a delay. But if they do have that extra padding, it makes them more reliable.
In other words, in this example all we care about is arrival times. The efficiency of any given companies operation is nearly irrelevant, especially when you consider that the actual flight time will vary from plane to plane, model to model, time of day (I imagine flying a red-eye flight you deal with less traffic), and which direction you’re flying (DEN to LAX average about 2:20, LAX to DEN is around 2:15).
What I think would be interesting to see is if this 8 minutes applies to ALL flights, or if it’s a per hour or mile basis. My history of United with little puddle jumpers has been less than stellar…
While NB has a point that to truly figure out the most efficient operation, block time should be a factor.
But for me, and the majority of consumers, actual airline efficiency means very little to me.
The only thing that I really care about, is if the flight leaves when it says it will and if I arrive when it says I will.
If United Schedules 2.5 hours for a flight that takes 2 hours, and leaves a few minutes late, as long as they arrive on time, that’s all that matters.
RiLLy,
These numbers are only released by the U.S. Department of Transportation. Unfortnately, Lan falls outside of that.
Andrew,
There’s more data than you’d ever want at http://www.bts.gov
As far as schedule padding, it’s something I put a lot of thought into. Unfortunately, it makes it almost impossible to compare. In order to take out regionality and time of day, I would have to look at actual block times between two airports at the same times. While this does happen between some of the major airports, (LGA-DCA for example) over 99% of the routes have only one aircraft, from one airline flying during that hour. The sample size was too small.
That being said, I considered how bad it is to leave the effects of schedule padding within the results. What I deduced was that passengers have almost no qualms booking a United flight over a Delta flight if it was scheduled for 5 minutes longer. But if you sat on the airplane an extra 5 minutes, that’s a big deal. So I just went with the market and said “if passengers aren’t willing to pay for a marginally shorter flight, then I’m not willing to take it out.”
It’s just my way of justifying my own laziness.
What about cancelled flights? My wife and I recently took a trip where our layover flight was cancelled. We ended up renting a car and driving the remaining distance.
As a passenger what I would like to see is results for the express/regional brands of each of the majors that are operated under the major’s code.
For example, since United contracts out about 2,000 (or 60 percent) of its flights to United Express carriers such as SkyWest, Mesa, etc.) what would make sense is something like:
United: x percent (combined average between all UA/UAX flights)
- United Express (combined average between all UAX carriers/flights)
- United (mainline)
I like this analysis quite a bit.
Airlines should pad their schedules based on historical wait times at different airports.
If the airline knows that at 16:00 at LAX there will be a 10-15 minute wait before they can push back, they should add this into the total flight time. When a flight is delayed, it is more important to the passenger if they are able to make their connecting flight, vs someone who will arrive at their final destination a few minutes late.
Additionally if the flight pushes back early, the pilot can extend the flight time by 5 or 10 minutes to burn less fuel. Which is becoming a bigger and bigger concern for the airlines.
The additional step you could make in the analysis is comparing delays based on stage length. HA flies a bunch of really short flights, between the islands. The only have a handful of longer stage length flights between Hawaii and the mainland, and Hawaii and their international destinations. I think you would see HA fall like a stone in the rankings.
Most of the short haul flights for the legacy carriers are flown under contract and consumers don’t realize it isn’t actually UA, DL, or AA flying the short connecting flight.
This is interesting, but still doesn’t really paint a true picture. United, for example, is notorious for padding its flight times – so a 3:30 airtime ORD-LAS flight, for example, might have a scheduled duration of 4:30. If the flight leaves the gate 30 min. late but gets into the small LAS airport easily, voila, it’s 30 min. early! (Wow, we musta “made up time” in the air!).
Worse, they can push back the flight on time even though they know there’s a ground stop, leave you sitting on the tarmac for an hour, and that flight then posts as both departing and arriving on time!
Mark,
The schedule padding is kind of a touchy subject. People don’t like to see their airline score poorly (reference the Delta pilot in the LifeHacker comments). It’s easy to say United pads their schedule and that’s cheating.
But like you said, is it cheating? If United has a padded schedule, they have to fly that padded schedule 100% of the time. Pilots are paid based on that schedule which increases costs, and also limits your ability to schedule effectively. It’s a give-and-take game that I wouldn’t call cheating at all.
I like your idea of using stage length, and might consider that for the June numbers. You’re right in suggesting that it has a factor, but I’d be willing to guess it’s a negative factor. Long flights have more time in the air to make up delays, and are less affected by the cause of most delays, the airports. Just a guess, but I’d be interested in finding out.
What I don’t think is taken into account here is departure delays. Maybe it was, but wasn’t mentioned. Most flights I’m on that are late are due to departure delays at the departure airport itself.
So while we may board on time, but the plane may wait on the tarmac for an hour or more before clearance is given by the destination airport to depart, or if there is just a delay due to heavy traffic, etc.
This frequently happens to me leaving to go to Atlanta. So this may account for some of Comair’s poor performance.
I have to say that I have a really hard time believing that United comes out on top. I haven’t paid attention to your methods but maybe there is something that we’re missing that is inaccurately putting this airline on top.
Interesting analysis! Surprising results for me as I frequently fly the major Chicago airlines and United always seems to have more delays than Southwest, but perhaps ORD is more delayed than MDW…
Did you consider that an airline’s dominance at a particular airport could skew the results? E.g., United/AA may set the average at ORD or Southwest may set the average at MDW, reducing their normalized delay times. Perhaps in calculating the normalized delays for an airline at an airport, you could calculate the average delays excluding that airline?