Demand calculation questions (to AS)

I have two brief demand calculation questions:

1) Is there any preference between O/D and Connection flight when flights are booked during demand calculation:

Explanation: If i fly A->HUB->B

Which would get booked first (at times of demand calculation): A-> HUB or A-> B ?

If A->HUB gets booked first, then maybe A->B flights will not be booked on HUB->B leg because A->HUB gets booked full by O/D pax on A->HUB route, leaving no seats for A->HUB->B  pax.

2) Is the passenger assignment equal for the same flight among all 3 days the flight gets booked for?

Explanation: Many times I find out that flight on day 3 gets less passengers booked than the flight on day 1 before the flight (or on day of the flight when flight departs after DCT*).

*DCT Demand Calculation Time.

I have two brief demand calculation questions:

1) Is there any preference between O/D and Connection flight when flights are booked during demand calculation:

Explanation: If i fly A->HUB->B

Which would get booked first (at times of demand calculation): A-> HUB or A-> B ?

If A->HUB gets booked first, then maybe A->B flights will not be booked on HUB->B leg because A->HUB gets booked full by O/D pax on A->HUB route, leaving no seats for A->HUB->B  pax.

 Good question :)

waiting for the answer...

2) Is the passenger assignment equal for the same flight among all 3 days the flight gets booked for?

Explanation: Many times I find out that flight on day 3 gets less passengers booked than the flight on day 1 before the flight (or on day of the flight when flight departs after DCT*).

*DCT Demand Calculation Time.

As far as I know, the demand is equal across the days.

Maybe on some particular days, there is a competitor offering the route with a better rating (or there is a connecting flight to get from A to B with better ratings than your direct flight).

I once tried looking into what would be the best time to have a departure wave, relative to the DCT:

If you're flight departs just after the demand calculation time, you may have more empty seats as pax may find alternatives that close to time of travel. Just a thought - didn't progress much on it.

From the perspective of a larger airline, it seems to me the last day gets more bookings, but there's a simple reason why. For example, if we are talking Monday, Tuesday, and Wednesday bookings... since Monday's flights have had more calculation times than Wednesday's flights, they are going to be closer to capacity, and many will be at capacity. Because of this, and because most flights are not direct, the tendency will be for more Wednesday bookings. I've seen this time and time again on my flights. I'll get, say, 120 bookings on a flight on the first day, 20 on the second, and maybe 5 on the third. This, of course, doesn't mean there are more people booking three days out. Just means there are more options, and my airline can accommodate more passengers going to different places.

As to what gets booked first, connections or direct flights, my guess is that it's completely random. For the calculation of (for example) LHR, I would think it would have a number of passengers leaving LHR (the number depending on several factors including AGEX) that would each go through a "random destination generator". The server takes each passenger and randomly selects a destination, the chances for each airport weighted by traffic data and other factors. Then the passenger picks the route and "books" the flights. No, this has no basis in fact, or is speculated on with any programming knowledge. It's just a stab in the dark.

The bookings are done with the aim to travel soon. So flight within the next hours are more requested then flight in 2 days. About the explicite relationship between direct and part of transfer-connection bookings, I can't tell you anything, sorry. Have to check this with the development.

...

As to what gets booked first, connections or direct flights, my guess is that it's completely random. For the calculation of (for example) LHR, I would think it would have a number of passengers leaving LHR (the number depending on several factors including AGEX) that would each go through a "random destination generator". The server takes each passenger and randomly selects a destination, the chances for each airport weighted by traffic data and other factors. Then the passenger picks the route and "books" the flights. No, this has no basis in fact, or is speculated on with any programming knowledge. It's just a stab in the dark.

Hi,

I think you're stabbing in the wrong direction  ;-)

Why would Sascha go through the trouble of gathering real life traffic data, and then tell the software to randomise destinations ?

1) Is there any preference between O/D and Connection flight when flights are booked during demand calculation:

Explanation: If i fly A->HUB->B

Which would get booked first (at times of demand calculation): A-> HUB or A-> B ?

If A->HUB gets booked first, then maybe A->B flights will not be booked on HUB->B leg because A->HUB gets booked full by O/D pax on A->HUB route, leaving no seats for A->HUB->B  pax.

...

Computers are boringly punctual. Let's say the program goes down it's list of destinations when it calculates demand for an airport. If the destination list would be alphabetical (airport names) and your hub is Frankfurt, the flight to your hub will get bookings after the connecting flight origin-Frankfurt-Amsterdam gets bookings. But passengers bound for Zürich would check possible routes after the passengers for Frankfurt have booked their flight. So passengers for Zürich may find the flight to your hub already fully booked before they can book the two connecting flights.

The same would happen when the ORS calculates demand from other airports. In the end, passengers for Zürich may find it very difficult to find a plane that can take them to their destination  :-)

But let's say the program goes down it's list of destinations, with the airports in numerical order (the number being the airport ID). You would however get the same problem. Just replace Zürich by the airport with a high ID number.

It is possible that the program randomises the list of destination airports. But that means an extra split second for every airport that needs to calculate its demand. It is more plausible that (for example) airport 1 first checks possible routes to airport 2 (if there is demand for that airport) and continues until it reaches its own number. Airport 1000 would then first check possible routes to airport 1001 (if it has demand for that airport) and continue until it reaches airport 999.

But I could be completely wrong :-)

Jan

Hi,

I think you're stabbing in the wrong direction  ;-)

Why would Sascha go through the trouble of gathering real life traffic data, and then tell the software to randomise destinations ?

I was meaning a randomization based on the real life traffic data.

For example, let's say 20% of all passengers originating at LHR want to go to CDG. 15% want to go to ORY. 10% want to go to AMS. For each passenger, a random number is generated, maybe between 1 and 10,000. 20% of those numbers point to CDG. 15% to ORY and so on down the line. There would be some slight fluctuation that would appear random, but for the most part, the data is consistent. The real reason for doing it like this would be to force the passengers to book flights in a random order of destinations, as opposed to having a predictable and computationally possible order that could be "broke" by a player with enough time and patience -- which you basically said in the second part of your post.

I think the correct way would be to book destinations form lowest demand airports all the way up to the highest demand airports, and O/D traffic should be calculated as the last in line.

That way passengers booking from BIG Airport A going to 1-bar airport could be assured that their seats on BIG Airport A to HUB would be available and not taken up by the trunk traffic going between BIG Airport A and HUB.

I think if the demand calculation is (would be) set up to go from lowest demand airports all the way up to the highest demand airports, it would be big boost to connecting traffic.I am sure there are lots of connecting flight to/from slot constrained airports that do not get sufficient traffic (sufficient to the extent as they really could) because lots of seats on trunk routes is taken up by O/D pax... and when you cannot add more flights to the trunk route because of slots, there is practically not much you can do, unless you can fit a bigger plane on the route which is not always possible.