This thought occurred to me. Perhaps someone can explain why it hasn't been tried before....

It would seem to me that Southwest Airlines, with its open seating policy, has the ability to switch aircraft at the last minute between the various size 737's that they have to decrease the chance that they have an overbooked flight. Further, they can match the airplane to the passenger load. For example, say they have 2 different kinds of 737's (in practice, they have more). One holds 110 pax and the other holds 130 pax. As they are selling tickets, the computer tracks this and when the pax load passes 110, they assign the bigger 737 to the route, when possible. By doing this, they get the right size plane for the pax load more often.

Why don't the other major airlines who assign seats try the same sort of thing? Instead of promising that flight #1234 will be a 737 or A-320, just promise us a "jet" aircraft. If they only sell 50 seats for a particular day, put an RJ on the route. (I know some flyer talkers won't like that, but if it would make the airline more profitable, keep them in business and maybe one day lower fares, it might not be so bad).

Another example might be to put a 737 on a flight, but upgrade it to a 757 if they sell enough tickets. That would reduce the number of "bumps" they would have to give out. There would be a time window, 24 to 48 hours before the flight when the airline would have to commit to a certain aircaft for a certain flight. They would want to leave a few seats for last minute walk-ups.

Seems like it would be a great computer scheduling program, but the airlines are known for their ability to create such programs. Obviously, it can get complex scheduling crews and planning the entire day's flight, but to compare the revenue generated against the cost of operation seems like a way to squeeze out some profit.

The airlines have a lot of airplanes on the ground right now in mothballs. Does it really cost more to have a few extra airplanes ready to go, even if you don't fly them, compared to having them parked? Having those few extra would help with maintenance problems as well as ensuring that every flight few near full, thus ensuring a profit on every flight -- rather than flying some routes 1/2 full.

Maybe I am missing something here. Certainly, it would not be an easy computer program to write to schedule all of that. While there could be *some* assigned seats (every airplane has a row 5), when it got back to the highest numbered row on the smallest aircraft, no additional seat numbers could be assigned. If an airline uses 3-2 seating, it should number the seats ACDEF so that the seat assignments translate smoothly to 3-3 seating.

Anybody have any thoughts?

It's a logistical nightmare. Different plane types need different crews, for one thing. For another, the bigger plane is porbably needed down the line. If the fleet has many extra aircraft, this can work, but the larger fleet is in itself inefficient.

nsx --

Definately having a lot of airplanes sitting on the ground not being used is inefficient. Yet, the airlines are currently faced with that, except that they are sitting in Arizona in mothballs instead of at the airline's hub locations. That was one of my questions -- does it cost significantly more to have an aircraft at a hub, ready to go, compared to sealed up tight in Arizona? It would seem to me that the biggest portion of the cost is interest or lease payments. The cost to make a plane operational has got to be pretty small compared to the lease payments.

As business rebounds, it would become more and more difficult to have any spare airplanes, but, by then the airlines would be profitable under the current model.

Also, I would think, with good scheduling, it might be possible to have extra flights just on Monday and Friday, the two busiest days. Yet, the flights would be added just "as needed" rather than promising increased service. Thus, if flight 123 goes from DEN to ORD at 3pm, on Monday and Friday, there would be a flight 8123 added at the same time. The airline would have to get that worked out with the FAA and the ATC system, but I would think that FAA/ATC would be very helpful given the difficult times the airlines currently are in.

I will agree that the scheduling is exceptionally complex and that you have to have the right aircraft positioned in the right place all the way down the line for the entire day. However, "down the line" usually is a hub city every stop or two, when aircraft can be substituted. Some hub cities (e.g. DEN) might be better suited to storing extra aircraft than others (e.g. SFO). Indeed complex, but, aren't the complex problems the main way that business today is squeezing out extra profit? Consider Wal-Mart, who makes its suppliers work with Wal-Mart's computers to ensure the proper inventory and stock is available at each of their stores. Using this type of technology, they squeeze out extra profit by keeping their costs down. They build customer loyality by having the lowest price most/all of the time. Computer time is pretty cheap compared to flying an airplane across the country only 25% full.

Maybe it is too complex to do... Maybe it is because of the complexities added by some of the union contracts. Maybe instead of big pay cuts, the unions take smaller pay cuts and give some more flexibility?

I personally love complex programming challenges, and this would certainly be one of the most complex. If a computer could sit there and "hum" on the problem and crank out $100,000 or $1m per day in cost savings, it would be worth it.

Load factor based dynamic rescheduling? Not in this decade.

I think the airlines have a hard enough time with creating a new schedule every N months. Beyond basic routing, there's making sure the equipment is in the right maintenance stations when they need various checks.

I'm looking into this for my PhD thesis. This is actually one of the open problems out there. If you find any good solution, feel free to contact me and share your ideas.

- Pat

Believe me, and I've looked into this, if they could do it they would. Inability to assign seats wouldn't stop them - there are ways to deal with that issue.

Another factor not addressed yet is that they also have to schedule aircraft for service after certain numbers of flight hours. Some of the more frequent maintenance checks can be done at any major airport, but the biggies need a maintenance base. They try to schedule aircraft to end up there at the right time in order to avoid an empty flight to get it there. AA has an artificial intelligence application that handles this for them. (Wiirachay: They published a paper on this application - IIRC, in Interfaces or somewhere like that. If you can't find it and want the reference for your dissertation, I may be able to dig it up.)

The idea is to complex for the average (airline) employee to be implemented. Several difficulties will come up, however I still like the idea very much. E.g. UA needs more profitability immediately!

This idea would work if all the fares were the same for a city pair. Of course, the fares aren't the same. Different fares and booking class inventory is the reason why airlines don't swap aircraft all the time.

With yield management, you can control the number of tickets sold on each flight with better precision than any aircraft-swapping system. Unless you have 10 different types of aircraft, your choice of number of physical seats available is going to be less than your choice of number of seats available for sale at a particular fare.

It costs a lot of money to have "live" aircraft rather than parking them in the desert. For one thing, you need to have crews available, and crew costs more money than the airplane itself.

Yield management, on the other hand, is free. You can increase or decrease the fare amount or the seat inventory at no cost to the airline.

I'm not sure I understand how yield management precludes Load Factor Dynamic Rebalancing (I love that name!).

Say an airline has 10 fare classes and 13 seats in each class for a total of 130 seats. The airline sells 80 seats in 7 fare classes (6x13+2=80). Fourty-eight hours before the flight, the past experience is that the airline sometimes sells up to 6 seats at a premium price. The airline had scheduled an aircraft that has 130 seats, but also has an aircraft that has 90 seats. The airline searches its databases to find a crew for the smaller aircraft. If it finds one, it substitutes the aircraft. The airline "commits" the smaller aircraft at that point, giving up the revenue if more than 10 people show up at the last minute (or, perhaps just flying them on a less than direct route) It still has exactly the same revenue assuming that 90 people or less want to fly, but they fly a smaller aircraft, thus saving money.

If the airline can schedule faster, perhaps it can commit at 24 hours prior to the flight. That would be even more efficient.

If the airline can't find a crew to fly the route, the airline sticks with the 130 seat aircraft. Or, if the aircraft is going from X to Y to Z and the Y to Z leg has more than 90 seats sold, they stay with the 130 seat aircraft.

Easy? No. More Profitable? I think it could be. To figure out if it is more profitable, I or the airline would have to run experiments on a database of past sales patterns combined with data about types of aircraft available, crew availability, etc.

Getting the aircraft in position for maintenance, or for the next day's flights obviously are critical parameters too. Just another parameter to make the problem more complex, but a necessary parameter none-the-less.

You have several issues -

1) Different airline types require different pilot certifications. So if you switch, say, a 737, with a 757, you have to switch pilots as well. Switching pilots is more problematic than simply switching airplanes as eventually the pilot has to end up at "home" and also the pilot can't fly over a certain amount per day or month. So you can't switch a pilot dynamically from IAH-SAN to an IAH-MCI routing with the plane.

If you ran an airline with a single type (like Southwest does), you could potentially fly pilots on a fixed schedule and dynamically allocate the actual aircraft since it doesn't really matter to the pilot if they are flying a 737-500 or a 737-800.

2) A single aircraft change affects multiple flights. Let's say a particular schedule calls for a flying BNA-IAH-BNA round-trips all day long (stuff like this happens with
hub-and-spoke airlines). In this simplest case, if I remove 40 seats from IAH-BNA to give to another route, I also have removed 40 seats from the subsequent BNA-IAH flight.

In the case of Southwest, it's potentially much worse. They might have an aircraft flying PVD-BWI-MDW-MCI-OKC-MSY and then on to who knows where. Switching a plane in this case, would make it very difficult to get back to the right size of plane later. Hub-and-spoke model better allows for getting back to the expected state.

3) Big differences in seating requires changed in flight attendant staffing. I switch from 125 seats to 175 seats and I need another FA. Like pilots, I can't simply keep the FA's with the aircraft because than the people won't end up in the right place, or they end up working too many or too few hours.

4) Seating assignments - you basically couldn't pre-assign anything in the back of the plane or risk re-assigning these seats later. Of course airlines already don't pre-assign all seats, so this wouldn't necessarily be a huge problem.

I think this could be done if your airline was single aircraft type (let's say RJ-50, RJ-70, RJ-90 or a variety of different 737 models), but you would make the changes before the day of the flight. So if sales were particularly good on IAH-BNA (perhaps a convention is in BNA or a Titans-Texans football game) two weeks ahead of the flight, you switch that plane with a larger plane from a routing that is getting lower than expected passengers. You would tend to lose some of the high-yield "must buy the last ticket on the plane" market because your planes would tend to always be less than full, but you would have an overall higher load factor as you turn fewer purchasers away.

If you made the changes multiple days in advance, it could all be made seamless to everyone involved (airport, staff, catering, passengers).

Definitely would be an interesting computer program to optimize this.

They are one and the same.

There are two ways to manage the load factor:

1) re-schedule aircraft and crew all the time

2) adjust the fares


When you see an aircraft flying with few people, you can bet that re-scheduling is not going to help, because the aircraft will be needed down the line anyway, either to carry more passengers or for maintenance.

Let me give you three examples (happens to be on Delta, but all the airlines are basically the same in this respect).

1) A 757 flying LGA to CVG at 5:30 PM on a Wednesday. You would think the plane would be packed, but it's less than half full. Substituting it for a RJ or 737-200 won't work because that plane continues to LAX.

2) An MD-90 flying DFW to MSY at 7:00 AM. Not too many people on this flight. I took the "continuing" flight (same plane, different flight number) to CVG and it was full from MSY to CVG. There would have been no point in downgauging the DFW-MSY flight.

3) A 757 flying CVG to LGA late at night. Only 33 people on board! No continuing flight that night, but the plane is needed somewhere the next morning, such as ATL.


So, what it boils down to is that airlines already schedule aircraft as efficiently as possible within the constraints of the hub-and-spoke system. There is no point in constantly looking for flights to downgauge because the plane of that size will be needed later on.

When airlines see a particular route that has too many seats, they downguage for the next schedule (e.g., 757 to 737). Other routes that could use more seats get the larger plane.

You don't need to do this every single day, because the frequency with which airlines change aircraft types on a routing does not change that often.