Following up on Bruce’s post several weeks ago about the side effects of not having a schedule for RapidRide, I want to share information I collected for my masters research paper. The paper is structured around a survey of transit information in 24 cities, mostly in Germany, The Netherlands, Denmark, Sweden and Norway. While my research touches on all aspects of transit information, I wanted to pull out the relevant information about transit schedules to inform this discussion.
The table below contains relevant schedule and service information for the routes in the 24 cities I decided to include in the survey. I purposefully included a range of service types from local bus to subway. The table below is sorted by minimum service headway and includes other service related information that impacts service reliability.
Scanning the table a few things jump out:
- All 24 routes in the survey provides scheduled departure times when service headways are at or above every 10-minutes. Eight routes in the survey have scheduled departures times at or below 5-minutes.
- Of the four routes that use headway based schedules, two do so for headways at or below 10-minutes, one at or below 7-minutes and one at or below 4-minutes. All had real-time information available via smart phones or text message and real-time information at all stops, with the exception of Glasgow’s route 61 local bus which only had it at stops with shelters.
- Three of the four routes that used headways based schedules are bus routes, and operate in either mixed flow traffic (ROW C) or some combination of exclusive ROW and mixed flow (ROW B/C).
- Of the 23 routes with a schedule, 17 use a stem and leaf schedule design.
- Despite the availability of real-time information in some form for 21 of the routes, 20 routes provide exact departure times at all times and 23 routes provide exact departure times during off peak hours.
- All tram and subway routes, with the exception of Barcelona’s L3, provides scheduled departure times at all headways.
Using the data above I believe a summary of the schedule information practices in the 24 cities above can be laid out in a few guiding principles for schedule information. These can be applied regardless of mode but not necessarily route reliability.
- Stem and leaf schedules should be used. Go here to see an example.
- In general real-time information should be viewed as supplemental to schedule information, not a replacement for schedule information especially during off peak hours.
- Headway based schedules should primarily be used when time savings for riders from improved reliability due to active headway maintenance exceeds the added uncertainty or wait time due to lack of a set schedule.
- Schedules with departure times should be provided as long as service at that headway can be provided reliably. The minimum headway this occurs depends mostly on the ROW:
- above every 10 minutes for transit in ROW C (mixed flow)
- above every 7-10 minutes for transit in ROW B/C (exclusive and mixed flow)
- at all headways for transit in ROW A/B (grade separate and exclusive) unless headways are below every 5-minutes
- Service with headway based schedule and headways above ~7-minute should have fully developed, heavily marketed and easy to use real-time information at stops, but especially on cell phones via apps, website, text or voice.
Applying these principles to all routes and modes regardless of service brand or agency provides a consistent, rider concentric policy guide on how to provide schedule information to riders.
Hear hear! The idea that you don’t need a schedule for a bus that comes every 10 minutes at best is a bit of a joke.
For the record, Zurich uses stem-and-leaf schedules with exact times for all its tram and bus routes, and headways on most routes are 7 minutes during peak. They have real-time info also, but the running joke when I lived there was that the real-time info was less reliable than the printed schedules (it had teething problems, and the schedules are super-reliable thanks to dedicated ROW and the Swiss obsession with punctuality). Many people used an iPhone app that worked completely offline with a GPS database of stops and the published timetables.
It’s not a joke.
It’s an acknowledgement that buses are generally unreliable during the peaks. Printing a schedule only reinforces the riders opinion of that, making it too easy to see how late it really is.
At least without a schedule, the agency makes you work at figuring out how bad it was.
So they are also unreliable all day long and on weekends?
Oh yeah.
This makes it sound like intentional obfuscation to hide poor on-time performance. The reality is that a printed schedule is counterproductive when on-time performance is poor. It encourages people to attempt to plan trips based on timed boardings/transfers, when on many of our core routes, those trips have basically zero chance of going as “planned”. Remember, Metro’s official instructions to riders is to just show up at a stop 5 minutes before the scheduled departure time and everything will be ok! How much sense does this make when systemwide average on-time performance is 75%? Or on the core crosstown routes where it is much much worse? Show up at a 48 stop with a 2 or 3 leg trip-planner printout 5 minutes before the scheduled departure and see how well that goes for you.
They are setting up new/inexperienced riders to fail from the very beginning. Tell them “this bus comes about every 15 minutes or so”, give some estimated travel times and real-time arrival info, and those riders have a much higher chance of trip success.
If a bus runs every 15 minutes, the last thing you want to do is to show up 5 minutes before the schedule departure time – when the bus is 10 minutes late, you’re waiting the full 15 minutes. If you show up at a random time, the odds are you’ll do better than that.
At 12 minute headways the Swift is generally reliable and I don’t even look at the schedule. I don’t look at the 358 schedule on the other hand because it’s never on time with 15 minute headways.
Conclusion: Show up often and I don’t need a schedule. Be completely unreliable (CT 113) and I won’t use a schedule.
When I really want a schedule is when the bus headways are >= 20 minutes and the bus is reliable.
This must have taken a lot of work. Thanks Adam.
So, my understanding of this is that you believe Central Link should publish schedules for every train’s departure and arrival, and not just say “every 7.5 minutes” and so forth? Is that what you are suggesting? Or not?
You realize that Link gave up giving scheduled departure and arrival times because its “on-time” performance was embarrassingly bad, by that metric, right?
Unlike the D, Link is frequent enough that I’ve never worried about a schedule (aside from that one all-nighter I pulled in SeaTac where I needed the first train in, but that train’s on the schedule)
I am asking about Adams observations of other transit systems, not your experience.
The data shows that in most places with rail operating in ROW A/B, schedules are almost universally provided schedules as well as real-time information.
Link doesn’t have either and I think this is a huge problem and hurts ridership and is a big frustration for me. With that said the DSTT isn’t really “exclusive” ROW since bus volumes are so high as to delay trains. If the DSTT was more reliable ST could easily provide a schedule at all times of the day.
I think during peak times it’s just borderline okay, because headways are right at the break point of not mattering as ST doesn’t have a way to make real-time information happen without new technology and buses aren’t leaving the DSTT for a while. But off peak it is absolutely wrong to operate the service without a schedule.
I take Link every day, usually at peak in the morning and off-peak (8-10 pm) at night. I’m fine with no schedule at 10-minute headways, because Link is pretty reliable. It only gets frustrating at 15-minute headways (which you have before 6:30 a.m.).
I think some people certainly are okay with waiting when headways are 10 minutes but relying on headway based schedules adds a good amount of travel time unreliability. Essentially when you arrive at the station you have to wait somewhere between 0 and 10 minutes. If you have a schedule you can time your arrival down to a few minutes.
In my opinion unless an agency has a compelling operational purpose for not providing a schedule, which I think ST and Metro do during peak periods but not off peak, they should provide a schedule.
I’d love to -only- have to wait 10 minutes from the scheduled time with the 24 sometimes… meanwhile my 19’s almost always a minute early (not that I mind)
I think the southbound train travel time in the tunnel during PM peak is pretty reliably 1-3 minutes slower than off-peak, and mostly due to boarding and deboarding volume on the train.
Northbound is still messed up, with overbooking Bay A with outbound bus pick-ups as the primary culprit. I haven’t observed AM Peak.
Train travel time through the tunnel is slower due to buses, but a dependable amount slower most of the day, simply due to speed limits and short pile-ups getting in and out at IDS.
When Link is running every 10 minutes, I absolutely want the schedule. I know it is about a 5 minute walk from the skybridge at the north end of Seatac until I reach the platform. I want to know whether I need to rush or I can stroll leisurely. I don’t want to rush and just miss it, nor stroll and just miss it. And I have the same issue at the other end of my ride, where I am about 6 mins from the stop I use.
The amazing thing, of course, is that ST actually does operate a schedule – they just don’t tell riders. From Seatac, during 10 minute service it leaves on the 0’s: 00, 10, 20, 30, 40, 50
in 15 min service it leaves at 05 20 35 and 50
and in 7-8 min service it leave at 05 12 20 27 35 42 50 57
From Westlake in 10 min service it leaves on the 7’s at 07 17 27 37 47 57
In 15 min service it leaves at 07 22 37 52
In 7-8 min service it leaves at 07 14 22 29 37 44 52 59
Yes, it operates on a schedule, perhaps with poorer adherence during the 7-8 time, esp southbound, but it operates on that schedule
ST just doesn’t tell riders when it’s supposed to operate
It’s arrogant and poor practice that means only people who either observe it or use the planner to figure it out will know
@Carl thanks for your comment. I knew they have a “schedule” but didn’t know what it was. With this information and travel time information it wouldn’t be too hard to create a schedule for every station in excel. I could see that being a very handy app.
That’s what I used to create my Link schedule mock up. The data is there, just not easily accessible.
I notice that in your examples, you don’t have a single route — not even for subways — where the headways are under 4 minutes. ST claims they will be able to have headways of about 2 minutes, or at least 3 minutes, inside the downtown tunnel at some point. Do you think 2-minute headways are realistic for Link at any point of the line at any time in the future?
And why is it that none of your examples have headways under 2 minutes? Do some of these subways actually have 2-minute headways between trains, but alternating between 2 different routes on the same track? Or is it because it just is not practical? Or because that much capacity is not needed? Or, what?
Nuremberg gets 100s headways only with driverless trains.
http://en.wikipedia.org/wiki/Nuremberg_U-Bahn#ATC_and_driverless_trains
There are plenty of subway lines with headways under 4 minutes, but that wasn’t the point of this post, or his paper.
I am interested in Adam’s answers from his experiences. He is the one who is writing the paper for his Masters program.
Norman, all you’re interested in is trolling as hard as possible.
There are rail systems that operate in ROW A that have headways of 2-4 minutes. Stockholm interlines like color lines into a single tunnel in downtown, with the red line for example operating at 2.5 minute headways during peak periods. If you automate the system as they do in Vancouver for example you can get down to even 60-90 second headways.
This discussion is however off topic.
Paris gets down to 105-second headways on manually controlled lines. The newly deployed all-automated fleet on a few lines can operate at 85-second headways.
As for interlined light rail: Boston runs 38.5 trains per hour through the Green Line’s central subway at rush hour, with grade crossings, a grossly antiquated signalling system, and nothing that remotely resembles automation.
And yes, operations do break down at these volumes. But anecdotally, only when headways drop below 2 minutes on average does the constant stream of trolleys start to get clogged.
Mexico City does 60 second headways with manually operated trains. Toulouse has 60 second headways but I think the trains are not manually driven if I remember correctly.
The drivers in Mexico City are hard on the brakes and hard on the “gas” to keep those schedules. It’s also easy to keep a 60 second headway when there’s no room on the train for passengers. Makes the stop fairly quick. LOL
These is no point of studying information for a transit system that comes so frequently no one ever needs to look at a schedule. That is why.
There are lots of examples of systems with sub-4-minute headways, including in all three of the biggest cities on our own East Coast.
London
Adam,
You mention “active headway maintenance” above. If this is carried out for bus transit service, how is it executed? Who carries out these efforts? Radio? On-street inspectors? What are the elements? How effective are these efforts? How long are the routes where headway is actively maintained? Are any of these lines interlined?
Thanks for your insight.
Lots of questions. Here is some information I do know.
Metro is doing it manually with a radio operator calling drivers and having them slow down or leave a terminal early, probably more the former. There are automated system that do it as well and I think Metro has the ability to do it with their new on board computer system but they aren’t now.
Generally I think they are successful because bunching is cause by a leader being too far ahead of a follower. However ensuring that you minimized boarding/aligning delay gets at the core issues of bus bunching.
Here is some more info. http://ntl.bts.gov/lib/45000/45600/45630/OTREC-RR-12-09_Final.pdf
with a radio operator calling drivers and having them slow down or leave a terminal early, probably more the former.
Yup. Lots and lots and lots of the former.
First bus gets delayed. Metro slows down the second bus.
Second bus, now late for no reason, accumulates its own additional delays.
Metro slows down the third bus even more.
Rinse and repeat.
Nobody ever gets anywhere quickly.
#RapidRide
Agreed. Metro’s “old” culture of the three S’s (Safety, Service, Schedule) and well-meaning safety bulletins admonishing us to not “rush” are to blame. But some drivers take these messages a little too seriously. While safety should always be our number one priority, sometimes “Service” is being ruthless and leaving a straggler/runner behind. Metro is starting to issue bulletins about not waiting, but there is a lot of work to be done to undo decades of the three S’s.
Yeah that is the problem. Active headway management is a day to day operational technique to avoid bus bunching, but it doesn’t speed up buses because basically all Metro can do is slow buses down to avoid bus bunching. It’s a balancing point though, because if you can address the lead/following gap issue early you can yield improvements.
I guess I would add that most of the times I have heard of headway management being used was for a very frequent local route.
I guess I would add that most of the times I have heard of headway management being used was for a very frequent local route.
I’m extrapolating that you mean elsewhere in the world.
That makes sense, then. You’re only preventing bus bunching along a short, easy-to-visualize corridor.
The problem with its usage on RR — which has only been bad in my experience thus far — is not only that RR isn’t especially frequent, but also that it travels over multi-mile distances (Interbay, West Seattle Bridge) where the difference between a free-moving and an artificially-slowed bus can quickly become 5 or 10 minutes differential.
Start buses late, or slow them before they hit a Ballard Bridge opening, the Mercer Place light, or traffic at Denny, and you’re just setting yourself up for cascading delays. Especially if you don’t learn from your mistakes, and insist the next bus remain “15 minutes behind” at Market, which is sure to become “20 minutes behind” by LQA and “25 minutes behind” by Virginia.
Every time headway-management has played a role in my RR experience, it has turned a “bus due in 6 minutes” into a more-than-15-minute wait, followed by a 25-30 minute trip the rest of the way from Leary.
Total trip: nearly double the former 18.
This is not improvement!
Wow. You need a master’s degree in boring to enjoy this analysis. I would rather stand in the pouring rain waiting for a C/D RR bus than figure the above graph out.
Come on now, Adam put a lot of work into it. It may not be for everyone, but you could say something nice.
Some of us are data fanatics ;-)
I rather be boring and spot on than exuberant and dead wrong. The point of the post was to lay out a pretty cut and dry case for how transit agencies here are doing it wrong.
and I think you have done that quite well here. Metro is willfully brushing aside the many many requests for schedules they seem to be getting. Hopefully analysis like this will help drive the point home at last.
With RapidRide C line in particular (this is the one that I actually have frequent encounters with now), the coach will sometimes crawl along and spend extra time at some stops, just like it did when it was route 54 with published schedules.
It really feels like they are adhering to some sort of schedule-only when they are “on time”…
Along those lines, I also wonder why it is that when two inbound D Lines (continuing as C Line) are right behind each other, they stay that way. So, what happens is that the first one is completely and totally packed with passengers. Nobody says “hey, there’s an empty bus behind me”. Can’t they work it out so when there are two coaches in a row for them to communicate with one another and more evenly distribute passengers? I was really hoping that the introduction of RapidRide would cut down on the back-to-back coaches.
ALSO, route 116 seems to be underutilized. Now that it goes directly to the West Seattle Bridge from 1st Ave, it is so much faster than before, but I don’t think that passengers really know that. I was taking the RapidRide C line from DT to the Fauntleroy ferry because of how slow the 116 used to be. Then, after looking at the schedule I realized it was rerouted and much quicker. I won’t take RapidRide to get to Fauntleroy in the afternoon any more…It’s not “rapid” enough =)
Anyways, I don’t see why they need published schedules UNLESS the service has more than 15 minute headways…I can totally deal with 15 minutes. Anything above that and it starts to get annoying, much like seeing the expensive, fancy GPS connected sign say “***REFER TO SCHEDULE***”
The “Next bus” sign really could use some help. I wish I could get a good photo of the sign as it jumps all over the place. One seemed to be perpetually displaying “6 min” the other day. I’ve seen it start at 8 minutes and go down to 5, then up to 11…It seems to me that the system could use some fine tuning.
Speaking of fine-tuning, what about the Busview? It’s tracker only shows a few coaches/runs. No RapidRide anymore, no East, Bellevue or South Base routes????
With RapidRide C line in particular…the coach will sometimes crawl along and spend extra time at some stops…
That’s called “headway management”. It’s how Metro ensures that one slow-ass driver will screw up everybody else’s day. Pace-setting by slowest common denominator.
#RapidRide
Slowest common denominator, ha!
My comment above RE: the three S’s belongs here. Slowest common denominator – Well played…
Next time you’re at a meetup, corner me and I’ll tell you the story of the driver who personifies this as told by other drivers.
Curious!
Personally, I’ve always had the feeling that Busview must have enemies or something, because it was never really promoted – even though it was really useful. (Especially for folks who were getting ready to commute home from work).
I think about 10 people in seattle use busview.
Following on to d.p. and what Velo has said about RR-B signal pre-emption, I wonder if the signals respond to a set schedule for each bus, knowing it is behind, and give maximum green to them, or is the signal system just looking at the route in general to decide when an approaching bus will trigger the lights.
It seems that if bus A were late, then they would get lots of green time to get the total schedule back on track.
Anyone know for sure?
Thanks for this snippet from your thesis work, Adam! He probably did gloss over a lot of boring detail about the scientific methodology (i.e. I assume he didn’t just survey over a hundred cities, and then cherry-pick 24 that best supported the conclusions while tossing all the cities that did not).
But the question has been asked why Metro ignores all the calls for having schedules instead of just showing headways. I suspect one answer is that a huge chunk of their customer service calls are about buses being late. Statistics on types of customer complaint calls might be informative in understanding the politics of why a transit agency might opt to go schedule-free on some routes.
Maybe. I guess I would hope that a transit agency would never obfuscate schedule information just to reduce customer complaints. Maybe that is naive. I don’t know. Either way I think it is the wrong approach.
However I do think it is important to note that almost all of Metro’s core routes come at nearly or better headways than RapidRide during peak and off peak hours. Metro provides schedules for all those routes, I don’t see why they wouldn’t for RapidRide as well.
Thesis work may not be the best place to make moral judgements on a process.
I can see reasons to not have a schedule on RapidRide during the first pick or two of a line’s existence. After that, I don’t have reasons to offer not to do it.
Maybe I should rephrase the question. Might printing a schedule actually reduce customer complaints? (I realize this is not terribly germane to the data set in the post, but more on the local topic.)
I don’t know. I do however think it is wrong question to ask.
I think the primary question that needs to be asked is what produces a better users experience. A headway based schedule at all times of the day, a schedule with departures times at all times of the day or something in between?
I think whatever meets the needs to users will in a general sense maximize ridership and minimize complaints?
“almost all of Metro’s core routes come at nearly or better headways than RapidRide during peak and off peak hours”
This is a joke, right? There are very few routes that aren’t 30 or 60 minute headways evenings and Sundays, including all the routes that the A, B, C and D replaced. That has been the biggest accomplishment of RapidRide so far, and is something that Metro finds very difficult to achieve otherwise. Only the 7, 26/28, 36, 44, 49, 70/71/72/73, and maybe a couple other routes have managed to reach or almost reach frequent status until 10pm without RapidRide.
@Mike I wasn’t talking about late night service. I was really talking about mid-day and evening service when I said off peak. Regardless that isn’t really the point of the discussion Brent and I were having.
Boston’s light rail system is 4 different lines merging onto a single track operation with turn around spaced thru the trunk line. The headways are originated from the outermost points heading inbound by line. At no time are the headways regulated during the single track operation other by the antiquated light system and dispatch monitoring spacing in the tube. They arrive when the arrive and the supervisors have the authority to dead head outbound service in order to maintain headways and/or next trip on time performance from the outermost point. Once in the tube headway adjustments are made either by short turns or deadheads.