Airport Queue Management: Is More Information Better?

Two weeks ago I wrote about how Philadelphia managed to re-build I-95 in record time. I then wrote about queues and social norms. So it’s only natural that I finish this short trilogy with an article on Philadelphia [checks notes] being at the forefront of speed once again. This time, it's about the new queue management system installed in Philadelphia’s International Airport:

“PHL has been installing sensors that track how fast security lines are moving and calculate approximate wait times. Travelers can see those estimates, which are updated every 30 seconds, on screens at the airport or online.”

How do they do it?

“PHL’s QMS works by measuring “passenger dwell time” using a “fusion of sensors,” according to the airport’s website. The sensors are placed throughout the security lines, from the beginning of the queue to the end of the screening process.”

This is what it looks like:

And this is whatt the monitors look like.

Sounds like a great idea. And it is. But … beyond pure innovation and providing more information, is it really helpful?

I’m not sure at this point. 

Why?

Information is only valuable if we can act on it. 

In fact, the value of information is measured by the benefits that can be derived from it, considering the new actions you can take, based on this information. 

Let’s analyze this further.

What actions can passengers take? 

“The estimates can give passengers ‘an opportunity to decide which [checkpoint] to go through ahead of time, since our terminals are all connected on the secure side,’ airport spokesperson Heather Redfern told Billy Penn.”

But how many times do we really decide on which checkpoint to go through? Most of the time, moving between terminals takes approximately 10 minutes, in which case, the difference in time will be completely nullified. In fact, when there are significant delays, it’s the airport’s responsibility to allocate security officers (which are easier to mobilize) and ensure smooth operations rather than to expect this from people. 

So what other decisions can passengers make?

The most important decision is how early to leave for the airport. But note that this system only tells you what the current waiting time is. It doesn’t tell you what the waiting time is going to be upon your arrival (my hope is that over time it will). The system could have had all the information about past performance, and actually predict what the load at any point in time during the day would be. But this is where I fear things may become worse for quite some time before it settles. 

And let me explain why.

Arrival Time at the Airport as a Newsvendor 

My regular readers could see this coming from a mile away: Almost everything in life is a newsvendor problem. 

Rather than explain to new readers what a newsvendor is, let me describe the airport arrival time as such. 

When deciding when to leave for the airport, it’s important to consider the uncertainty regarding the time it will take to reach the gate. While you may be aware of the boarding time (although it can be delayed), there’s always the possibility of congestion on the way. It’s common for security lines to vary in length, and God forbid you have luggage to check in! In some airports, it’s almost impossible to calculate how long it takes to reach the gate (Atlanta, for example, has a train within the airport). So you must factor all this in, to make an informed decision. 

When you leave for the airport, it’s highly unlikely that you’ll be there exactly at the right time for boarding (for me, that’s about 15 minutes before the official boarding time, so I have enough time to stand aimlessly in line, but ensure space in the overhead bins), so you balance two different costs: the cost of overage and the cost of underage. Both costs are subjective and depend on your time, the value of time, and the activities available at the airport. 

The cost of overage is the cost of being at the airport too early. This is wasted time and since you’re not at home, having your preferred meal or coffee, you have to settle for Starbucks. Luckily, at PHL we have La Colombe Coffee. It’s clear that airports aren’t generally a location you would choose for important meetings or to be productive, but for me, this overage cost is low. I don’t mind sitting at the airport catching up on work. In fact, some of my newsletter articles have been written or edited while at the airport. But there’s no doubt this wouldn’t be my first choice, unconstrained. Some airports are better at reducing this cost with better seating, and some airlines control this cost by having nicer lounges. Either way, it’s a cost that needs to be factored in. 

The cost of underage is the cost of arriving a little too late for the flight. For simplicity, let’s assume that this means arriving at the gate after the doors have closed. The cost includes the cost of having to rebook the flight, missing a meeting, or a day of your vacation (or at least a few hours). I’m operating under the assumption that you want to be at your destination. I’ve definitely had trips where I wished the trip didn’t happen and prayed for a cancellation, so I have an excuse, but I digress. 

For me this overage cost is very high. I optimize my schedule, so most of the time, I plan for things immediately when I land at my destination, and the idea of not only rebooking my flight but also reshuffling my schedule (or missing time at home with the family on the way back from a business trip) makes me view this cost as extremely high. 

The implication: even though I have TSA-Pre and Clear and every queue basting method available to travelers, you will see me very early at the airport. My Service Level (which is the CU/CU+CO) should be close to 100%. I don’t want to miss flights, as long as I have control over it. 

But you can compute your own service level (which is the fraction of flights you need to be on time for), and I hope you are going to leave a comment below with your own math. 

How Does Information Help?

So, imagine that not only do you have information on the current queue length (which is how the system operates now), but you can also forecast the waiting time when you arrive at the airport. In the end, it shouldn’t be that hard: the system knows what the waiting time was at exactly the same time last week (assuming all flights are the same schedule, not accounting for seasonal changes), and can add a little bit of factor if flights were canceled or delayed the day before, adding a bit more inflow into today’s queues. 

But here comes the issue: people are rational. 

And with that I mean people use the queue information to build beliefs on the waiting time. Seyed Emadi and Jay Swaminathan from UNC have a very nice paper, Customer Learning in Call Centers from Previous Waiting Experiences in which:

“We show that in this call center, new callers who do not have any prior experience with the call center are optimistic about their delay in the system and underestimate its length irrespective of their priority classes. We also show that our Bayesian learning model not only has a better fit to the data set compared to the rational expectation equilibrium model but also outperforms the rational expectation equilibrium model”

The results were further extended in a paper I co-authored with Qiuping Yu and Achal Bassamboo to deal with delay announcements. The conclusion of both papers was that customers who visit a service provider multiple times “learn” the true waiting time underlining the announcement. 

So let’s see where the problem will occur with this new system: 

Passengers will respond to this information. So if the queue is projected to be short at a specific point in time, more people are going to arrive, making it longer than expected. If the queue is projected to be long, fewer people are going to arrive, making it shorter than expected. 

I don’t expect these swings to be massive since the queue is only a small part of the overall journey from home to the gate, and to a large extent, the queues are determined by the number of flights departing, rather than any choices of arriving 5 minutes earlier by any single passenger, but it may take the system some time to balance. And once it does, the equilibrium waiting times are going to be the ones we see now. So those who travel a lot, already “know” the equilibrium. The system is going to be an improvement for those who don’t, or for airports passengers are not very familiar with. 

Following the paper by Seyed and Jay, which showed that people who didn’t frequent the service tended to be too optimistic, it’s evident that after receiving information about the queue length, people will arrive a bit earlier than before. 

Other Informational Effects

While I was somewhat skeptical of how the system can help (at least in its current configuration), one has to acknowledge that it has some benefits. 

For example, it’s clear we hate waiting times. It’s been well-researched that we hate uncertain waiting times even more, which explains why it has already boosted passengers’ satisfaction:

 “Once the QMS pilot launched in Terminal D/E two years ago, Redfern said passengers there were more likely to be satisfied with their experience at airport security: 73% and 72% of surveyed D and E passengers, respectively, said the security process went ‘faster than expected’ after the screens were installed — 10 and 14 points higher than those figures before the pilot began.”

Information can also help in reducing our cost of waiting time, and this can happen even if the information is given right before we enter the queue. In our paper How Do Delay Announcements Shape Customer Behavior? An Empirical Study with Qiuping Yu and Achal Basamboo, we show that customers that received announcements about longer delays(in a call center) had a lower cost of waiting time per unit of time. 

What’s the mechanism?

Our time has no actual value other than the opportunity cost of not doing something else. Giving information to customers allows them to choose from a wider set of activities (e.g., answer one more email, read a longer article, etc.), but admittedly, this is hardly the case here since the queues are visible. I can see the progress, but it does help if I need to make a quick phone call (or potentially ask to cut the line if I can see that I’m going to miss the flight). 

So overall, this new system is a nice innovation, and my hope is that it will be the starting shot to make it an even better experience. I just hope other airports adopt it too (it’s not that necessary in Philadelphia where it’s easy to forecast waiting times), and I hope airlines start adopting and incorporating this information into their apps so we don’t need to guess where to find it in other airports. But sadly, (non-pricing related) innovation and airlines are two words that don’t really go together.