Thread by @AlexandraErin

Quick little thread on designing systems for human use: any attempt to design a system for efficiency or speed has to take into account the people using it as a component of the system.

Let's talk boarding planes.

There's a theoretical fastest way to get passengers on a plane, created by an astrophysicist: Right hand window seat in the last row first, then the seat two rows up from that, and so on, every other right hand window seat all the way up...

...and then you switch sides, and you repeat it for the middle, and aisle seats, and then you start over and do the same thing with the rows you skipped. On paper, this is the fastest system because no one is ever in anybody else's way, not for the bins or to dig out a seat belt.

You might recognize this approach from absolutely no airline, because nobody uses it, because it's actually horrible and does not work at all.

That's the fastest theoretical approach. In practice, trying to get passengers to participate in it... well, you obviously don't leave it to the passengers to figure out the instructions. You assign them numbers on their ticket and have them line up according to that.

But your system throws an exception now if two passengers trade places in line, deliberately or accidentally, or if a group decides they're all going to board together, or someone makes it to the gate after their normal boarding order.

If you're just adding people who arrive late to the back of the line then it doesn't really hurt much, but chances are you're not having all the passengers line up at once, and someone who arrived too late for their group will *expect* to join the current one.

So now you've got a small delay at the end of each boarding group, however many there are. And you've got another delay for each person who gets to the ticket scanner and is told no, you can't board with your partner, please wait for your number.

There's already a problem with people who decide they're too important to take their assigned seat, they're going to plunk down wherever. This system's extra sensitive to that.

And maybe you're thinking, "Well, the fastest system on paper plus all the things that cause delays in other systems = still the fastest system in practice", but that's still not accounting for the fact that people are people.

When your system is based on orderly and efficient movement of people, and it's already bewildering and arbitrary to most of them, then each delay is going to cause further delays as people try to assert control over a process they feel adrift in.

Honestly, the "optimal" approach is based on the same logic as a zipper merge but whereas a zipper merge is relatively straightforward, there are like three layers of complications to the airline seating.

https://twitter.com/JoannaDeLaune/status/1143183007550828545

That's the fastest theoretical approach.

Then there's the fastest approach in a laboratory setting, which is basically a simplified version of that: outside in, or fancifully dubbed WilMA: Window seats, then Middle seats, then Aisle seats.

United does a version of this.

Mythbusters managed to get the fastest time with this method in a trial, and other tests have given similar results. But United has not noticeably improved its boarding times with it, because the real world isn't a test, and they do have to make exceptions for families.

The fastest system in actual real world conditions?

Some of you hate it, some of you love it: Southwest's open seating. Let people, the chaotic, uncontrollable element in the system, solve the problem of chaos.

In Mythbuster's test, Southwest's method apparently had the lowest satisfaction level for participants. But people who fly Southwest regularly tend to come to love it. Is that survivor bias? It could be. Repeat customers are self-selecting.

But I think there are two different levels of satisfaction. "Logical" scheme like United's are satisfying because they make sense, and you can cling to that even when you get stuck behind someone in the aisle. You're thinking it could be worse, if they didn't do it right.

In Southwest's scheme, you're unsatisfied with the set-up if it doesn't make logical sense to you, but you encounter fewer actual frustrations along the way. So you may spend all day griping about the bizarre boarding process, while still having a better experience.

Southwest's set-up works for the same reason airports as a whole work, which is that most of the human beings moving through them just want to get where they're going.

Southwest's process injects order at the start, to make sure that people aren't fighting over who boards first and are allowed on the plane in manageable groups, and no one is standing in line for too long at a time, but lets passengers figure the rest out once they're aboard.

It has the advantage of requiring very little explanation for new flyers. You get visuals for the line-up process, the instructions are simple and repeated. This is key. Any system designed for the masses needs to expect new users constantly.

A lot of people don't understand Southwest's process from hearing an explanation, but going through it demands nothing of them that they can't figure out in the moment. Which can be frustrating ,again, if they don't see the logic.

But in practice, it's faster and *causes* fewer frustrations, even if the existence of the process frustrates unfamiliar users.

I think one of the reasons Southwest works as well as it does is that it makes passengers into participants in the boarding process; when you have an assigned seat, you have no obligation to the other passengers. Southwest gives you agency, which carries a responsibility.

Southwest talks to its passengers as a group participating in the activity of loading the plane and asks everyone to help out with the task of getting the plane off the ground so we can all get where we're going, and it works. Most people try to make room for the next passenger.

Again, human nature: we are gregarious, in an airport we're all naturally focused on getting where we're going. A boarding process that gives us an assigned seat just directs that tendency inward, towards our own seat. A boarding process that emphasizes the group aids efficiency.

So here are your takeaways for designing a system that will involve people:

1. Front-facing intricacy is an enemy of efficiency, even if it's efficient on paper.
2. Give users a sense of agency, ownership, and collective participation and they'll solve a lot of problems for you.

and

3. Design your system as if it's going to *constantly* be used by newbies; any public-facing system that works fine "once people get to know it" is going to be constantly not working fine.

I will add that Southwest's system *only* works because the gate attendants/flight attendants are part of it. Their cheery dispositions, simple and formulaic repeated instructions, and use of humor are all part of it.

On paper or according to "common sense" it might seem like it would be more ~*efficient*~ for them to stick to the instructions, emphasize the importance of speed and orderilness, etc., but that would just increase frustration, stress, and resentment.

It's important to the boarding process that their passengers find seats and get out of the way of the next group of passengers, but that goal is better served, again, by treating this as a communal endeavor to get to the destination quicker than by making people feel pressure.

The instructions are simple so that you always know what's expected of you. The humorous interjections before/between them are there to help keep you from being harangued by them.

I guess this leads to our fourth takeaway, which is: the user interface is part of the system, too.

Southwest's system, absent Southwest's personnel, isn't Southwest's system and would not fare as well in the wild.

Now, disclaimer: I know people who've had terrible interactions with Southwest personnel (same as any other airline), but 1:1 interactions are a different aspect of the system than the heavily scripted and A/B tested boarding announcements.

So, to par this thread down to its absolute essence, I guess I'd say: if people interact with your system, you will gain more in efficiency by expecting them to be people and treating them like people than you will by expecting them to play the part of cogs in a machine.