MD Exclusive: Q&A with Michael Smith of NextBus
We can all thank Alameda-based NextBus for those snazzy little marquees we can’t stop staring at in the bus shelters. Not only does it provide a flashing update about where our bus might be, it provides us, at the very least, with something blinky to hold our attention while we brave the chills (hey, lay off, it gets cold here, sometimes).
Turns out the technology is available in about 60 different areas of the country, though its hometown SF is still by far the largest user base. NextBus Director of Engineering Michael Smith chatted with Muni Diaries editor Tara Ramroop about the ins and outs of the system, why it’s not always 100-percent accurate, and what’s in store in the very near future.
Muni Diaries: Tell me how long NextBus has been around.
Michael Smith: NextBus actually started in 1997, so it’s been quite awhile now. It actually started in San Francisco, by someone who rides Muni. They were just frustrated waiting for the cable cars, actually. And thought, Gee with all this technology around, shouldn’t there be a better way to deal with this? So they came up with this idea of having something that tells you when the bus is going to arrive. But his friends said, “What’s an idea if you don’t actually implement it?” So he ended up starting this company, and now we have NextBus all over San Francisco.
MD: And it’s still most popular in San Francisco, it looks like.
Smith: We’ve been working with San Francisco for quite a long time and because of that, more people know about us here. San Francisco also has, by far, the largest number of signs. I think San Francisco is equipped with about 800 now.
MD: I thought I noticed more cropping up around town lately. Is there a continuing initiative to get more and more of these?
Smith: We installed our last batch of signs, and we just finished a few weeks ago. Over the last three or four months, we installed about 300 more. Muni had additional funding that they could use for that, but I don’t expect that you’ll see a lot of extra signs in the near future; we’re pretty much at what Muni’s going to get for now.
MD: Why’s that? Is it fiscal or does Muni think it has as much as it needs for now?
Smith: You have to have electrical power at the stop in order to put up a sign. So we’ve only been able to put them up at the bus shelters that have power, and those are usually the ones that have advertising in them. That’s really just a fraction of the bus shelters. There’s probably only a few hundred left that have power, that don’t have signs. So I think we’ve covered most of the bus shelters that do have power now. And also, there’s a cost issue, too. Muni wants to have as many signs as possible, but they can’t put up a sign at every stop. I think Muni has 5,000 stops, but I’m not sure if that’s the right number. The signs are great for people who don’t have cellphones, but now a huge number of people can get their information that way, too.
MD: So tell me a little bit about how it works. I understand it’s GPS-related, but how exactly do we get the information that’s flashing at us on the little marquee?
Smith: We have GPS trackers installed on all the vehicles: the cable cars, streetcars and all of the buses. And they use a cellphone type of modem to transmit the information back to our central computer. So they can communicate wirelessly. We get a new position report from the buses about once a minute.
MD: It seems like one of the big-ticket items that many of our readers, and some of us editors, complain about is when it says the bus will be here in three minutes…oh wait, no, it’s six minutes…OK maybe it’s closer to two…OK it’s actually 10. So there’s a little bit of back and forth that shows up on the marquee sometimes; why does that happen?
Smith: There’s a couple of different reasons for that. One of the biggest things you can see is that we might be making a prediction for a vehicle, and then it turns out that something happened. Either it gets called back to the yard, or a supervisor tells them to do a short turn, or it could get stuck in traffic, it could do a detour, or there could be some kind of electrical problem with the tracker itself. So there’s actually a variety of reasons the predictions might actually disappear. All of a sudden, you’re actually seeing predictions for the other buses. So a prediction might say there’s one in two and then 20 minutes. Then the two-minutes one disappears, so you only see the one for 20 minutes. And you’re thinking, wait a second, is it two minutes or is it 20 minutes? It turns out that most of the time when we make a prediction, the vehicle will show up. But there are times when the vehicle won’t show up just because of problems with the tracker, or that bus has been reassigned into doing something else.
MD: Would you say the data that gets to the marquee is just a little bit confused then?
Smith: I think the main issue is that there is so much variability. If the bus is making good progress, then you see the predictions count down as expected. But in a real-world situation, there are just so many things that could happen. And that’s the whole reason we have a NextBus system, because of all these things that can happen. And most of those we can take into account. But there are some things that are just unexpected, that you can’t really predict for.
So those are the kinds of things we work with Muni on. We don’t want to just tell people when the bus is going to arrive; we provide a whole suite of tools so they can more easily manage their system. For example, one thing you’ve probably seen is when no bus comes for 20 minutes or something, and then all of a sudden three buses come by.
MD: Exactly. It’s kind of nice if you’ve been waiting for a bus, but it’s very confusing, especially when it doesn’t match up with what’s on the marquee.
Smith: One of the things we’re trying to do is provide tools and information so Muni can better manage that system. Speaking, for example, of reducing bus bunching, where you make sure the buses are spaced apart properly. And also make sure the drivers are leaving the terminals on time. For example, we actually make the predictions based on GPS information. But when a bus is still at a terminal, we’re also basing it on when the bus is actually supposed to leave. Now if there’s some kind of issue and the driver doesn’t leave as scheduled, then of course the predictions can be off; and that’s the reason you can see fluctuations.
MD: OK. But even the GPS on the buses can’t correct that?
Smith: We know where the bus is. But if the bus is still at the terminal, then let’s say it’s supposed to leave at 10:00; then we start making predictions expecting the bus to leave at 10:00. But if, all of a sudden, the bus leaves early, then the predictions will jump down a bit. Or if the bus leaves late, then the predictions will jump up. But we don’t know that until the bus has actually left the terminal and we get a new GPS position.
MD: How often do you get updates, again?
Smith: Once a minute on average. Some buses report more frequently because they might be traveling faster. In that case, we need more information. But on average, it’s about once a minute.
MD: And this is automatically sent to your servers and all that?
MD: Does somebody just sit there and manage this information, as it comes in? How does that part work?
Smith: The system is pretty much all automated, and that’s how it can run 24 hours a day. Basically all of the generated information is based on the GPS information, and that comes in real time. And also, we use historic information. We know from gathering data how long it typically takes a bus to go from one stop to another, and that’s based on time of day, and all sorts of different conditions. So the system really runs on its own. The main thing that we do is monitor to make sure it’s working and deal with configuration changes. Muni and other transit agencies, they typically make changes once every three months; sometimes there are changes in between that. Those configuration changes are always a good amount of work.
MD: You mentioned that it’s a partnership between NextBus and Muni to ensure this works properly. How often do you chat and what are some of the latest things you’ve discussed to make sure it does indeed run smoothly?
Smith: We definitely talk to them every day. It takes quite a bit of effort to make sure the trackers are working and that the configuration is correct. We also get feedback from passengers, and sometimes it’s issues we can deal with, sometimes it’s issues we need to pass on to Muni, and so we’re always doing that. And, of course, Muni passes us information, too. It’s not like you can buy a NextBus system, plug it in and expect it to work. There’s just an amazing number of complexities that happen every day. And so we always have to be monitoring to make sure the system is working.
MD: So Muni is kind of at its maximum then, as far as the number of NextBus signs it can have for now, but are there any new and exciting things happening for the future?
Smith: We’re always evolving. Like I said, Muni hasn’t just bought this black box; we actually continue to provide service. We try to make the interface better for the passenger and provide more ways passengers can get their information. For example, you can actually get the information now from the MTC’s telephone system. You can call up 511 and actually get the information, so that’s kind of a nice feature. You don’t have to even have an internet-based phone or anything like that. You can use a regular phone or even a pay phone to get that information.
MD: You can also text NextBus, though, yes?
Smith: Yes, that’s actually just an experiment that we started, but it turns out a lot of people are using it now. You can use text messaging [text to 41411 and including “nbus muni” in the text field] to get the predictions by texting in a command and specifying an address. And then our system will determine what bus stops are near that address, and give you the predictions for the buses running by there. It’s kind of different in that you can specify an address.
In the future, near future, actually, you’ll actually be able to specify a stop by entering a code. Each stop will eventually have a five-digit code associated with it. And you’ll just be able to text that number to us, and then we’ll reply with the predictions.
MD: Ah. So the five-digit number goes in the “to” field?
Smith: Yeah, then you get a text back in just a couple of seconds. It actually works quite nice. One other new feature with the SMS system is when you get a text prediction back, you might find out that, oh, the next bus is in 15 minutes, but you’re only three minutes away from the bus stop. So you can tell it you want a reminder when the bus is really three minutes away. So you just reply with a number three, and you put your phone back in your pocket. When it’s three minutes away, your phone will buzz, you realize it’s time to go, and the bus will be there.
MD: Can you do that now, or is it still evolving tech?
Smith: You actually can do this now. We don’t really advertise it in the San Francisco system because it’s basically a new feature, and anytime we do something in SF, we get so many users, so we try to make sure it’s working well before we actually advertise it. But it is available right now.
MD: I think that takes care of some of my burning questions. Anything else you wanted to add?
Smith: This is really a continuing process. So it takes a lot of use of our management tools. We all want the system to run better: I’m a Muni rider myself.
MD: Oh? What lines?
Smith: I live off the 12-Folsom and the 19-Polk, so those are the ones I usually use.
I really see that there’s a bright future here. Muni has so much more information on where the buses are, how they’re doing, and where the problems are, and things like that. They will really be able to react to problems and fix them, and so what we really hope is that people see a better level of service. Another big thing that’s plaguing Muni is that the trips can take a long time. The buses really aren’t that fast, and a lot of that’s due to traffic, or due to there being so many bus stops, so we really hope they can move forward with their transit effectiveness project (TEP), which really should help the transit system.
MD: Yeah, there’s a lot of good stuff in that TEP that we’re all really excited about.
Smith: Now it’s just a matter of looking at what’s most important and implementing that. And of course, coming up with the money to make the changes.
The last thing I’d like to say is that we’re always looking for feedback. Email is best (email@example.com). We always understand the system can be improved. [NextBus] employees use it, of course, but we really want more information about what works and what doesn’t work. Visit NextMuni.com for more information.
MD: I’m sure our readers will be excited to give you as much feedback as they can.
Smith: Yeah, we found out San Francisco people are pretty good at that.
MD: Oh yeah, we’ve got plenty to say.
Smith: But we can always use more, and we like to encourage that.
Photo by Flickr user jm3
So many mysteries unraveled here! Great interview and feedback from Michael.
On an aside, I used the send-me-a-reminder-text feature the other day on the 43-masonic. I asked for NextBus to send me a reminder when the 43 line was 4-minutes away from my stop, and it worked like gangbusters.
By the way, if you spot glitches with the system we really sincerely want feedback. And when you send us info to firstname.lastname@example.org please include the date, time, route, direction, and stop, along with a short description of what happened. By providing us this information we can investigate what happened and further improve the system.
I can add a little more information to Michael’s explanation about powering more signs at stops. As a member of the SFMTA Citizen Advisory Council, I’ve been meeting regularly with staff members who’ve been working on the design of replacement shelters that will be installed over the next several years as part of the new bus shelter advertising contract.
The money isn’t there to buy signs, but everyone hopes we can find funding someday to install more of them, they’ve planned ahead for them. The roofs of the new shelters will have solar panels, so even the shelters without electrical conduit will have the power to run NextBus signs, and more immediately, add lighting to the shelters that don’t. For the shelters that do have conduit, the solar panels will allow them to add power back into the grid. I don’t think they were figuring on them generating anything significant in revenue, but they could at least cover their own costs.
Thanks for the interesting interview. I find that NextBus is working amazingly well given the unreliability of Muni.
My only feedback so far is that the “1 minute” predictions are usually too optimistic. Often it says “1 minute” or “Arriving” when the bus is not even in sight. 5 NextBus minutes >> 5 real minutes. I understand the need to overshoot though. It’s better than the alternative.
It needs to take into account that a bus can travel a lot further down Market Street in five minutes at 11pm than it can at 6pm.
Wow – what a comprehensive interview. I feel like I got a great education about NextBus from reading this post. Good stuff.
Jamison, I’m so excited to hear about the solar panels being used for the shelters! Thanks for the details – knowing the technology behind what we use every day is really great.
Erik – NextBus does take into account day and time of day, i.e. NextBus predictions are based upon a proprietary algorithm – whenever a new system is implemented or a schedule/route change made, the algorithm is updated. Thus NextBus knows to predict a “faster” arrival time at 11 p.m. than at 6 p.m.
Jamison, would you happen to know when the designs for the shelters will be released to the public?
Does anyone know when the new muni subway displays, like the one at Van Ness Station, will be out of beta/testing?