What Should ‘Reliability’ Mean?
I am a huge fan of Amtrak Cascades, and as a car-free resident of Vancouver BC who must frequently travel to Seattle, the two daily trains to Seattle are a lifeline. I’ve made 8 round-trips in the past 9 months, and I find it to be a generally enjoyable service with unparalleled scenery. People may rightfully drool over the West Highland Line in Scotland or The Canadian into Jasper, but for my money Cascades just south of Bellingham is as scenic as they come.
Last Saturday, however, while stuck in a siding in the agricultural hinterlands of Delta BC, waiting interminably to proceed toward Blaine, I got to thinking about just how subjective and equivocal “on-time performance” is as a metric of reliability. Of the 16 one-way trips I’ve taken in the past year, not a single one has arrived on-time. Seven trips have been late by 10-40 minutes , and nine have been 5-20 minutes early. By Amtrak’s on-time performance definitions, in which trains can be 10-30 minutes late and be considered “on-time”, my trips would have earned an 81% on-time rating and garnered them significant press for ostensibly improving their service. But shouldn’t “reliability” mean more than merely “not being late by X minutes”? Shouldn’t it mean consistency of journey time, whether early or late? If so, Cascades has a very long way to go, and much of the blame lies north of the 49th Parallel.
Check out the following chart. Using data from the invaluable Amtrak Train Status Archives, I analyzed the performance of the two morning trains (#510 Seattle to Vancouver BC, and #513 Vancouver BC to Seattle) for every day so far this year.
The chart depicts actual deviance from scheduled arrival time in minutes. Each red-green data pair indicate one day. Green lines denote arrival into Vancouver, while red lines denote arrival into Seattle. Gaps indicate days in January and March in which service was shut down due to mudslides. A negative bar indicates an early arrival. Notice two things:
(1. EXTREME VARIANCE. Trains arrive anywhere from 25 minutes early to 4 hours late, but journey times from one day to the next are incredibly volatile. On a chart like this, one should ideally see a straight line.
(2. COUPLING. Notice how delays on #510 cause reciprocal delays on #513, and vice versa. I chose these two trains for a reason; they run on single-track rails and must pass each other in a siding near Mt. Vernon. If one is late, the other train will automatically be late as well, and the effects ripple throughout the system. Very rarely on the chart is there a day in which one train arrives early and the other very late.
Thus if there is one word to categorize Amtrak Cascades service, it is inconsistency. Cascades is not slow. Cascades is not fast. Its defining characteristic is its variability…it is slow, then fast, then slow, then fast. And not just from day to day, as in the chart above. Within each run there is incredible variability as well. So check out a second chart. Using real-time data I gathered from my trip last Saturday, it shows average speed between selected segments of the route. I generally dislike using anecdotal trip reports as an inductive measure of general performance, but I’m confident that this chart illustrates a structural truth.
The Canadian section is embarrassingly slow, and there are frequent calls for British Columbia to invest in upgrading its decrepit infrastructure. While in this post I don’t wish to discuss the many technical fixes and reroutes that are needed, you may read them in the Amtrak Cascades Long-Range Plan. My broader point is merely that Amtrak operates its service as a match an inch from the fuse. Things are always a minor error away from causing major delays. Thus I suppose they deserve a certain bit of praise for offering the service they do on deficient trackage. Cascades frequently runs close to on-time, offers a comfortable and even luxurious service on comfortable Talgo trains, and offers a gorgeous alternative to I-5. But in the most robust sense of the word, let’s not call it reliable.