flynnibus
Premium Member
People keep calling it FP is just flat out wrong and is just inflammatory.
FP is more about shifting demand by dispursing a percentage of it. This model is about overflow. Push excess demand into the future.
Where this is so different in queueing theory is it tries to siphon off EXCESS demand. The consequence though is by deferring those queue entries... if you don't want the queue to grow later... you have to shrink the queue you allow to build later so it can absorb those deferred entries.
Imagine... *** are your desired standby line.. +++ are excess demand $ are returning customers and | is your line cap.
You start with something like this..
0900 - ************................|
0920 - ******************..........|
1000 - ****************************|
1010 - ****************************|++++
1020 - ****************************|++++
assume | is a 60 minute wait. Now the **s will decrease as the attraction eats up the queue. If the ++s are a reasonable level... they can be absorbed into the line left of the | when they return (similar to FP).
But if you have this situation...
1100 - ************************$$$|+++++++++++++++++++++++
1130 - **********************$$$$$|+++++++++++++++++++++++
1200 - ********************$$$$$$|+++++++++++++++++++++++
The amount of +++ will be large.. and as this accumulates (becuase you are handing out passes at a rate faster than the attraction eats the line) you must do one of three things
1) limit the number ++ you will accept or
2) limit the rate you hand out passes or
3) move | to the left to be able to absorb the returning +s without causing the wait to balloon
So you end up with something like
1200 - **************|$$$$$$$$++++++++++++++++++++++++++++++++++
1400 - *******|$$$$$$$$$$$$$+++++++++++++++++++++++++++++++++++++++++
According to the reports.. they ended up here
1730 -|$$$$$$$$$$$$$$$$$$$+++++++++++++++++++++++++++++++++++++++++
You can hold less and less 'walk up' capacity as the day goes on because of all the return people.
What would be really interesting to know about the test is how they were advertising the wait time. The wait time is what influences the +'s in the illustration above.
Queues work through a feedback system. People make a decision of "is X minutes worth waiting for this attraction".. if yes, they get in line. That decision repeats itself and if the inflow of people exceeds the attraction's capacity, the line grows. Eventually the line grows enough that people see a new wait time.. and decide that wait is too much and they pass.
This process happens constantly as people evaluate getting line or heading to the attraction.
If they artificially cap the line at say.. 60 minutes. For many people, 60mins is still considered a 'low' wait time for Soarin' so they happily will queue up. So now Disney is giving out return times at a pretty significant rate... a posted wait time of 60mins could really draw a lot of people (high rate of +'s above)
In the old system, the posted wait time would be increasing during this rush of people, deterring people from queuing up. If they didn't mimic this type of behavior in the wait time... they would effectively overload the demand for the attraction with no negative feedback to fight it.
What I think is difficult to achieve in this type of model is
1) how to keep toggling between return passes and feeding the queue. The queue will shrink, so you need to fill it, and you don't necessarily want/expect return users to be the only way to fill it
2) How they advertised wait times to account for the return users. The physical line may look like a 30min line, but that's because a ton of poeple will be showing up shortly, so you need to advertise a LONGER wait.. to discourage people thinking there is a short wait, and again increasing demand
The hard part about this kind of model is influencing the ++ rate and how to practically hand out passes.
FP did this by using the machines, rationing who could get a pass, and using the return time to influence behavior.
If all you have is people handing out paper passes as fast as people ask for them... and don't replace the feedback loop... the demand will overload the queue easily.
FP is more about shifting demand by dispursing a percentage of it. This model is about overflow. Push excess demand into the future.
Where this is so different in queueing theory is it tries to siphon off EXCESS demand. The consequence though is by deferring those queue entries... if you don't want the queue to grow later... you have to shrink the queue you allow to build later so it can absorb those deferred entries.
Imagine... *** are your desired standby line.. +++ are excess demand $ are returning customers and | is your line cap.
You start with something like this..
0900 - ************................|
0920 - ******************..........|
1000 - ****************************|
1010 - ****************************|++++
1020 - ****************************|++++
assume | is a 60 minute wait. Now the **s will decrease as the attraction eats up the queue. If the ++s are a reasonable level... they can be absorbed into the line left of the | when they return (similar to FP).
But if you have this situation...
1100 - ************************$$$|+++++++++++++++++++++++
1130 - **********************$$$$$|+++++++++++++++++++++++
1200 - ********************$$$$$$|+++++++++++++++++++++++
The amount of +++ will be large.. and as this accumulates (becuase you are handing out passes at a rate faster than the attraction eats the line) you must do one of three things
1) limit the number ++ you will accept or
2) limit the rate you hand out passes or
3) move | to the left to be able to absorb the returning +s without causing the wait to balloon
So you end up with something like
1200 - **************|$$$$$$$$++++++++++++++++++++++++++++++++++
1400 - *******|$$$$$$$$$$$$$+++++++++++++++++++++++++++++++++++++++++
According to the reports.. they ended up here
1730 -|$$$$$$$$$$$$$$$$$$$+++++++++++++++++++++++++++++++++++++++++
You can hold less and less 'walk up' capacity as the day goes on because of all the return people.
What would be really interesting to know about the test is how they were advertising the wait time. The wait time is what influences the +'s in the illustration above.
Queues work through a feedback system. People make a decision of "is X minutes worth waiting for this attraction".. if yes, they get in line. That decision repeats itself and if the inflow of people exceeds the attraction's capacity, the line grows. Eventually the line grows enough that people see a new wait time.. and decide that wait is too much and they pass.
This process happens constantly as people evaluate getting line or heading to the attraction.
If they artificially cap the line at say.. 60 minutes. For many people, 60mins is still considered a 'low' wait time for Soarin' so they happily will queue up. So now Disney is giving out return times at a pretty significant rate... a posted wait time of 60mins could really draw a lot of people (high rate of +'s above)
In the old system, the posted wait time would be increasing during this rush of people, deterring people from queuing up. If they didn't mimic this type of behavior in the wait time... they would effectively overload the demand for the attraction with no negative feedback to fight it.
What I think is difficult to achieve in this type of model is
1) how to keep toggling between return passes and feeding the queue. The queue will shrink, so you need to fill it, and you don't necessarily want/expect return users to be the only way to fill it
2) How they advertised wait times to account for the return users. The physical line may look like a 30min line, but that's because a ton of poeple will be showing up shortly, so you need to advertise a LONGER wait.. to discourage people thinking there is a short wait, and again increasing demand
The hard part about this kind of model is influencing the ++ rate and how to practically hand out passes.
FP did this by using the machines, rationing who could get a pass, and using the return time to influence behavior.
If all you have is people handing out paper passes as fast as people ask for them... and don't replace the feedback loop... the demand will overload the queue easily.