A great piece by Michael Perkins in Greater Greater Washington highlights a perennial problem with on-time performance measures for urban buses. He cites the policy of the Washington area transit agency, WMATA, which says that a bus is considered on-time if it’s no more than two minutes early and no more than seven minutes late. Perkins explains, with diagrams, that under this policy you could wait 19 minutes for a bus that supposedly ran every ten minutes, and yet the bus (and the one 19 minutes in front of it) would both be considered on-time.
(By the way, the WMATA standard sounds lax to me, though I haven’t done a survey. Few agencies I’ve worked with accept anything more than five minutes late, or one minute early.)
But Perkins argues, as I often have, that when we’re dealing with high frequency services, say every 10 minutes or better, earliness and lateness are the wrong measures. Earliness and lateness matter if somebody is really going to expect the bus at 7:43. But on high-frequency services nobody does that. They just go out to wait for the next bus and trust (or hope) that it will be along soon. Many transit agencies don’t even publish exact times for very frequent services. So the thing the customer experiences is wait time, not earliness or lateness. If you care about the customer, it follows that wait time is the thing you should measure.
Suppose you went out to catch a bus that’s supposed to come every 10 minutes, but every bus on the line was exactly 10 minutes late. By any lateness standard, that would count as total failure. But by any appropriate standard, it would be perfection. You wouldn’t know anything was wrong, and in a well-managed system, nothing would be wrong.
So at high frequencies, “on time” shouldn’t be about the time the bus arrives, but the actual frequency, i.e. the elapsed time between consecutive buses. A standard might say that 90% of the time, the next bus will come in no more than 150% of the published headway — i.e. 90% of the time you won’t wait more than 15 minutes if the published frequency is 10 minutes.
Now the interesting thing about this policy is that the individual driver can’t be held responsible for it. She’s responsible for when her bus gets to the stop, but she obviously doesn’t control the buses in front of her and behind her. But an aggressive operations management, armed with current tools such as GPS to tell them where every bus is, can monitor these things, giving the necessary real-time direction to drivers about how to adjust their operations to keep buses more evenly spaced.
So I suspect that the reason this change hasn’t caught on more widely is that it shifts the onus from drivers to management. I suspect some operations managers like to have their service judged on earliness and lateness because they use the same measures to assess their drivers. This means the operations management job is easy: just monitor the drivers’ performance and deal with disruptions. But as soon as you measure actual frequency, you’re measuring how well management and labor are doing the job together. Bus operations managers don’t all have the training, or the tools, to do it that way.
Of course, most urban rail transit operations have long been managed for frequency rather than specific arrival times. A train driver in a subway system is part of a large interconnected system that is actively managed by dispatchers all the time. I think we should have the same demand for operations management on frequent, high-volume bus routes. But it’s a big cultural transition, and it won’t come overnight.
In the Moscow subway, every station has a clock that counts how long it’s been since the previous train left the station, and the train drivers use that in order to maintain headway. Also, normally there’s a little bit of slack in the schedule, so that if a train is late, it can catch up, which is important when the headway is below 100 seconds. I suspect that with GPS, similar things would be possible for buses, along the lines of the indicator light that tells the driver whether he’s late. And centralized monitoring would be possible in order to manage the service as a whole and manage things like short turns in times of serious disruption. For the less frequent bus operations (like most of the VTA, for example), centralized monitoring can also be used to hold buses for connections, so as to lessen the impact of delays on passengers (a bus being 4 minutes late can make you 30 minutes late if you miss a connection), and allow for tighter connections when everything is running on time.
A 90% target really isn’t good enough: presuming that you make two trips per day, that’s one a week messed up. And typically 90% would be achieved by making the same intervals messed up every weekday, so for some people that would be equivalent to 0% punctuality.
And 150% would be difficult to manage at headways of under 5 minutes and annoying for passengers at much over that. It’s probably better to go for a simple 3-minute over headway, 99% of the time rule.
I formed a kinda snarky rule about on-time percentage. I call it Wad’s Law.
I hypothesize: Reliability is inversely proportional to frequency.
For the nonmathematical: The more frequent a bus service is scheduled, the less reliable it’s going to be.
Thanks for the link. I enjoyed your article and the insight that the change in measurement puts the onus on management.
Interesting choice of photo — that’s Ottawa, Canada!
____San Francisco has electronic signs that tell you when the next vehicle is due. Two problems :  when there is a missed run the gap looks very ugly (there are frequency charts all over the city on the system maps) and  they DON’T use those signs to tell their patrons WHY there is a gap. And those signs seem to be very rigid in their line info – I’ve seen an unlisted yardie (K -> Balboa Park) roll from the L’s outer terminal. They do use those signs for canned announcements about info up on the web and fare increases.
____Also, SF’s Muni is doubly trapped on performance – culture and statute. Getting them to switch over to consistent spacing would probably take a major miracle. A line from a 1960’s song about London’s omnibus’s applies to several Muni lines :
“We like to travel in convoy, we’re most gregarious”.
P.S. My apologies – I double-checked the lyrics :
We like to drive in convoys, we’re most gregarious
[Flanders + Swann, “A Transport Of Delight”]. As a kid I often saw three or four #38 Geary bus’s headed West from Masonic + Geary during the evening rush “hour”. Last night at ~ 1830 (Sat., 1 Aug.) at Geneva + Mission I saw three #14 Mission trolleys headed South almost nose-to-tail. The four decades in between have had countless single-line packs. The weird thing is how even the multi-line nodes (e.g. Balboa Park Stn.) have things in clumps or pulses – several bus’s in five minutes then nothing for fifteen minutes. Go figure.
One other disadvantage, I guess, of mixed-traffic frequent service streetcars (as opposed to busses) is they have slightly fewer options for headway management. Busses (who pull off to the curb) can stay longer at stops without disrupting other traffic; whereas streetcars who slow down (or stop longer) to increase the distance between them and whatever they are following, block cars and such.
Just to throw more logs on that particular fire. 🙂
First of all, what a refreshing blog. Basic common sense ideas explained concisely in plain language. The problem with US transit is that it has to operate inside of a particular cultural construct. When an American goes to Japan, Germany, Switzerland and experiences punctual, frequent, easy to understand service, its like stepping into heaven. But really its just what should be standard everywhere. In the 1980s I witnessed the decades old mechanical time recorders in the CTA control center that had very slowly rotating disks of paper with an EKG-like pen that jumped everytime a train passed a particular place. So as long as the tick marks were evenly spaced, the line supervisor AND his superiors and peers could see in real time the performance of his line. But back to the question of ‘why can’t we have a Japanese or German level of service in the US?’ I beleive that much of the reason that we don’t has less to do with funding levels, age of infrastructure, accountability or experience of management. I think it is a less tangible and harder to deal with social/cultural problem. I hope others here can expand upon that idea. It should not be acceptable that some US commuter trains that are 5 minutes 58 seconds late are considered on-time. Or that notices explaining scheduled outages are not posted at stations or in the stations of connecting services. Yet the public has long given up demanding better in a meaningful and effective way.
When I worked as a city bus driver in Christchurch (NZ) in the 1970’s, Leo one of my fellow drivers in his fifties, told me he had been a bus driver in Los Angeles in the 1950s. He said that on some the busy routes – multiple route corridors the transit authority employed timekeepers on corners. There primary task was to stop buses serving the city centre bunching up – as he drove past they would make a hand signal to him. If they pointed upwards he needed to speed up, if downwards slow down. As usual many of things now being done by computer have traditional methods! This speed management is of course within the legal and sensible boundaries, but every bus driver has a certain amount of plasticity in this area, at least most of the time. Unlike the normal motorist they keep a register of the colour of traffic lights ahead in some part of their mind, and take a little bit longer loading passengers, or let a vehicle pull out of a carpark etc, micro-management of time specifically to miss a favourable light phase that will put them too far ahead of scheduled time. I agree that any high frequency service should have some sort of GPS indicator, not only for in the moment but for transit operators to identify and remove sticking points, such as large dumps of schoolkids at a single stop, over filling and slowing buses for kilometres downstream, a problem that might need extra school only buses on the same route at key points. I have waited 25 minutes for a morning peak “Orbiter” bus – not a normal route for me but ostensibly every 10 minutes – and the driver referred to schools bout 8km away holding things up. Of course I find it incredibly embarassing arriving for work 15 minutes late – the sort of thing that can very quickly kill most people’s interest in switching from car to bus.
The “harshest” standards for buses I’ve seen have been one minute early to three minuets late. In rail terms, trains in the UK legally cannot depart from stations early, and anything over five minutes behind (or ten on long distance routes) counts as “late”.
The important point is that if the maximum wait time taht counts as “on-time” is substantially higher than the headway, then it’s a bad measure. Specifcally, if on-time means less than x early or y late, then buses on a headway way of h minutes can be “on-time” even a customer has to wait x+y+h minutes. Give miost cities have a mix of frequent and infequent routes, the correct meaure would be x early, y late (if x+y+h is less than 150% of h) *or* 150% of h (if x+y+h is greater than 150% of h.