It was a snowy January morning. I knew that all methods of transportation would be a mess and that my commute, if I decided to drive, would take well over two hours. So, I did what I always do. I bundled up, put some boots on, and started my quarter-mile hike through the snow to the Alameda Station. I check the Transit app on the way to the station and see that the train is going to be on time.
As I’m waiting for the train at the station, the E line that was supposed to arrive on time ended up to be a C line that I couldn’t take. Then a D line, H line, and then nothing for over 10 minutes. Silence. My app didn’t have an alert, and the signs at the station posted the scheduled times with no announcements. By the time the E line came, my backpack was soaked, lunch was wet, I was very late to work, and there were many unhappy passengers. This gave me the motivation to start tracking the reliability of RTD’s rail system.
I wish I could start this post by stating, “After tracking 102 train rides, RTD’s rail system is, for the most part, on time and serving its passengers without frustrations or delays.” Unfortunately, that statement is far from the truth. At best, from the sample data set I am presenting to you, RTD’s rail reliability clocks in at 36%. Sounds a bit outlandish, right?
Throughout this post, we will be exploring the data that backs up the above statement along with the methodologies used for tracking the rail system’s on-time performance. Let’s begin with a detailed explanation of how I collected the data and came up with the final reliability result. You can follow along with the screenshots below, which show just a small sample (my first 11 trips) of the overall data set. You can access the full spreadsheet of data here. In these tables, I’ve highlighted in yellow the columns being discussed in that section.
Scope of Data
Living right off the Alameda Station, I mainly commute to the office via the E or F line and also use the rail system to go Downtown. I tracked every single ride I took from February 20, 2019 to May 9, 2019, totaling 102 rides. Every tracked ride was on light rail; no commuter rail or bus rides were tracked.
Pickup and Drop-off
These columns are pretty straightforward. RTD trains run on a schedule so the Scheduled Pick-up and Scheduled Drop-off is the actual scheduled time that is posted at the stations and on RTD’s website.
Actual Pick-up and Actual Drop-off is when the train actually arrived at the station. Train arrival is a bit open to interpretation. Is it when the train crosses the back of the station, or when the doors open? The times documented in these columns were when the train came to a full stop and the doors opened. As a typical passenger, this is when you would expect your pick-up time to be. But, what if that’s not the standard definition of “pick-up” and RTD uses something else? This possible discrepancy is built into the result—more on that later.
Time Differences
These columns give us the numbers that contribute to the reliability calculation. Pick-up Difference and Drop-off Difference are pretty self-explanatory. This is the difference between the scheduled and actual pick-up times. Total Difference (Reference Only) is the addition of these two differences. This column is for reference only and doesn’t contribute to any reliability calculations.
Total Difference (Impact), highlighted in gray, is the most important column on the entire spreadsheet as it is the actual calculated time impact a passenger experiences. This is taken from the greatest of differences from either the Pick-up Difference or Drop-off Difference columns. This can also be calculated with various interpretations of “impact”, so here are a few examples of my methodology:
- The train picks me up 5 minutes late and drops me off 5 minutes late. The total impact would be 5 minutes as I would arrive at my destination 5 minutes late.
- The train picks me up on time but drops me off 2 minutes late. The total impact would be 2 minutes as I would arrive at my destination 2 minutes late.
- The train picks me up 3 minutes late but drops me off on time. The total impact would still be 3 minutes because it was originally 3 minutes late in picking me up. Even though I arrived at my destination on time, I was still waiting an extra 3 minutes. With adverse weather, 3 extra minutes can be the difference between being dry or drenched with rain or snow.
Miscellaneous Data
The Announcements column reports announcements made at the stations, over the train PA, or via real-time data apps. Real-Time Data Accurate is straightforward: tracking if the real-time data shown in the apps was accurate. I would check the real-time information 5 to 10 minutes before the scheduled pick-up or drop-off time. Fare Checked is if my fare was checked on the train.
Real-time data was accessed using both the Transit App and RTD’s Next Ride service.
Special Circumstances
When a train doesn’t show at all, or when a train departs the station before its scheduled departure time, riders are very negatively impacted. Unfortunately, I experienced these situations and included them in the data set, which decreases the reliability percentage. In a perfect transit system, when trains are running late or not coming at all, an announcement would be made at the station and through transit apps. In this case, there was not a single announcement from RTD about a late-running or cancelled train.
I also experienced a single ride that I recorded in the data set as Didn’t Take. Here’s what happened. I was waiting for an F train. It didn’t arrive on time, and while waiting for this late-running F train, the next-scheduled E train pulled into the station, so I took it instead. This counts as a no-show for the F train because I left on the E train before I could determine if the F train arrived or not.
Notes
You will want to access the spreadsheet to see the data explained in this section. The Notes column in the spreadsheet contains a variety of observations I made while riding the train during this assessment. Examples of these observations include abrupt stops between stations, non-functioning digital signs at stations, notes about weather (bomb cyclones!) and the like.
On multiple occasions, the PA announcements were incorrect and inconsistent with what was actually happening on the train. For example, the “doors-closing” announcement was repeated, alarmingly, in between stations while the train was going full speed. On one ride in particular, the “next-station” announcements were off by a station for the entire journey. I hope no one from out of town was on that train and got off at the wrong station.
More notes include trains leaving the station early and more of me complaining about the weather; it was a rough winter for public transit riders. However, the most exciting observation is that RTD is replacing their light-rail fleet with new cars! These new cars are great as they have a more modern feel and no more awkward face-to-face seating. Here are some interior pictures of the new cars.
Tolerances
As mentioned above, I have built in a tolerance to help mitigate any discrepancy between my data collection methods and ones that RTD may use to track when a train arrives or departs from a station.
- I considered the real time data to be “accurate” if it was within 1 minute (plus or minus) of reality.
- I considered trains that departed the station 1 minute before the scheduled time to be “on time”, but if the train departed 2 or more minutes before the scheduled time, I considered it to be “early”.
- I considered a Total Difference (Impact) time of 1 minute to be “on time”. Anything 2 minutes or greater was categorized as “late” which negatively impacted the reliability percentage.
Final Results
Out of 102 rides:
- 36% of the rides were on time
- Real-time data had an accuracy of 84%
- Fare was checked during 12% of my rides
- 4% of the trains didn’t show up at all
- 2% of the trains left the station more than 1 minute before the scheduled time
- 20% of the rides were late 4 minutes or more
- Not a single ride (0%) had an on-train or station announcement regarding delays or issues with the system
So what am I trying to say? As a growing metropolis, we need more reliable transit. With fare increases, route cuts, and new lines opening, it is completely unacceptable to have a transit system that does not reliably move, or notify people. Personally, I would never switch to driving to work as it would double my commute time; however, those with shorter commutes might give up on RTD and drive instead. This is not the result we would like to see.
If we want to solve our traffic issues and gain ridership on our transit system, reliability needs to be the foundation and, based on my survey, RTD is doing a poor job delivering on it. I hope this assessment will encourage RTD to do better.
We don’t just deserve better transit, we needed it yesterday.
Ryan, I have had the same experience as you about “no show” trains. I commute from Alameda to Dry Creek as well. One morning, the 7:48 F never arrived, and then three C lines came through in a row (one of them must have replaced the next E line), and I didn’t get on the train until the next F line arrived at 8:03 (like you, I can’t take the H line either). And you are also right that there are never any announcements about what is going on. I am concerned too that unreliability will drive commuters away. I am actually driving my car this month to “take a break” from it before I start commuting by train again in June.
I did 10th and Osage to County Line for years and had similar experiences but I did want to point out that when there’s bad weather I would jump on the H to get out of it and then wait for the E/F. I once was so ill prepared for a cold front “out of nowhere” (I may or may not have checked the forecast that day) and the delay was so bad that I even took a train in the opposite direction.
I think the line you are taking also affects reliability.
I used to live in Westminster and take the B line all the time. Those trains ran on time to the minute every day no matter the time of day. It certainly helps that it is just a point to point line with no stops so there are rarely delays. It also probably helps that the train sits at either station for an extended period of time for passengers to load at their convenience, and also that loading for the train requires no stairs or ramps for the disabled. Overall, I found the B line to be incredibly reliable.
I now live in Denver and commute to work on the L line. I would still say this line is quite reliable. I have never experienced a train leaving the station at 30th and Downing late (but again, it’s an end-of-the-line station). Even picking up the train in the evening at 16th Street Mall is usually on time, and that is affected by the other lines that come into downtown from the south as well.
Even when I’ve taken the A line to the airport (as it’s the only way I’ll ever go to the airport now) the trains were always on time. And that line was infamously publicized for delays and issues.
Obviously weather can affect anything, and perhaps I’m just lucky, but I would say overall I find RTD’s transit system pretty reliable and worthwhile. Perhaps some lines are just more reliable than others.
Ryan, I think your point about the lack of passenger notifications is well taken, but don’t you think it’s a little harsh to say that RTD is doing a poor job because some of their trains are running 2-3 minutes late? Especially if they still get you to your destination on time? During the busiest time of day and the snowiest months of the past 10 years?
It’s standard practice in the transit industry to consider any arrival within 5 mins of the scheduled time as “on-time.” That’s simply because transit networks are complicated and sh*t happens sometimes that the transit agency can’t control. For example, some stops have passengers with wheel chairs that require extra time to board, some passengers hold the doors and delay the departure of trains, sometimes cars block the tracks, and on and on and on. I think if you’re going to criticize RTD for being a few minutes late on some trains, you should also propose how they can deal with any of these variables. I bet you don’t have a good recommendation.
If you look at your data and apply a 5 minute threshold for lateness, the reliability figure becomes a lot better. But really, a better way to analyze your data would be to look at the variance of delays across all the trips that you recorded. What your data tells me is that if you ride RTD, you can count on being at your destination within 4-5 minutes of your scheduled time, about 90% of the time… something like that. How many folks that commute by car during rush hour can say the same thing? It seems to me that RTD offers an immensely more reliable trip than traveling by car, and objectively, their doing a pretty good job.
Thank You for the interesting article, You brought up some good issues. I commend you for the time it took to put it all together. However, we need to point out first that all schedules indicate that “All times are approximate.” There is no way that any transit rail or bus can run consistently exactly on time.
I am a frequent user of RTD public transit with much experience as I grew up in New York with living working in NYC and in a European city with many years of using bus, trains and
subways.There are many obstacles that impact time for transit; I want to point just one that is more personal to me as I am now a senior disabled.
RTD light rail has one big issue that will impact times is the ramp to access the trains for the disabled. The operator has to get out of the cab and opens the door for the disabled with wheelchairs, walker to enter and exit. Also, there are parents with strollers. Many of the disabled walk slower; some parents have trouble with the strollers in entering/exiting. Also many times the operator has to help secure the mobility devices and strollers to the cars. This all takes time and how many may need this access cannot be predicted and will change from stop to stop.
These same problems are even more evident with when the disabled or a parent with stroller enters a bus as a ramp has to be lowered and the driver has to make sure these devices are secured to the bus. There is some issues of entering the mall shuttle for the disabled as the driver sometimes has to engage the ramp and come out of the cab to secure the device–all takes time.
The commuter heavy rail trains are much easier to access because the cars are level with the platform; requiring no ramp. The operator has no need to leave the cab and the securement within the cars are much easier. However, the operator has to be aware of those using wheelchairs or walkers or strollers that they have entered and exited the cars safely and are clear of the doors before they are closed–and, of course, these riders are slower to board and exit. This impacts the time the train to leave the platform and again it cannot be predicted how many of these riders will be at each stop.
So what is the answer for this obstacle to keep transit on time? I do not know if there is anything specific that can totally solve the problem. Access to heavy rail is easier, quicker and securement is simpler, so I prefer to use this mode of transit.
What I have learned is using public transit is to have patience and to allow more time. I am retired so using public transit is much more relaxing as I no longer have the strict schedule of commuting to work. Even then when I allowed more time for going to work, the journey of public transit was much more relaxing than commuting by car.
one factor you didn’t mention is the impact on connections; this can be much more significant than a minute here or there
a person close to me often walks 1/2 mile to Alameda Station, rides to Auraria West, then connects to the W all the way to Golden; much too frequently, the arrival at Aurora West is just late enough to watch the W depart with no chance of boarding; at this person’s times of day, this has a 15-minute impact
i talked about this this with someone in-the-know at RTD and heard there has been a known delaying factor on the trips leading north toward Auraria West, but i got the impression there were too few making the connection to the W for RTD to care; so, instead of walking to Alameda station, my close person will often recruit me to drive them to Auraria West; traffic still makes it chancy …
p.s. on the subject of tardiness, your blog has an RSS feed, however it only seems to update every few weeks; if you could make it regular i would read more of the great articles here
Great article, with great data. I think I’m a little more positive than you on this, because your data shows that 80% of the time, you will be at your destination within 4 minutes of the arrival time. (Assuming I’m reading that summary table correctly). But that’s still not great. If you think about a person who is using the light rail to commute every day, that means they’ll be subject to a 4+ minute delay 2 times a week. If you are someone with a job where you have to be exactly on time, you’re going to have to give yourself a buffer to get to work on time. This is exacerbated because some of the light rail corridors don’t have frequent service. So you might have to leave 20 minutes earlier than you should to be on time when using the light rail. That’s not great.
The other thing is that this data is a lot better than if you had to transfer to a bus or another light rail line. If your train is even a few minutes late, that delay could cascade and make you even more late. This is also why early departures are infuriating. Some stations have bus and light rail timetables that are matched up, but when there’s a delay on one end, you’re going to have a long delay.
Parts of the light rail lines are on streets, and there’s only so much you can do to when a train is stuck in street traffic. But, a lot of the light rail system is not on streets, and there really shouldn’t be delays that frequently. I take the W line – the only thing that line has to deal with is some street crossings. Yet, my experience is very similar to yours. The trains frequently wait at stations where there’s no crossing and no apparent reason for waiting. Honestly, I would guess that the W line is more likely to be 5 minutes late than it is to be exactly on time. If trains cannot consistently meet the published schedule, what’s the point of the schedule in the first place?
Communication on all of this is a great point. Why aren’t the electronic boards at stations using real-time data instead of the published schedule? Why aren’t station and train announcements common when there’s a major delay? I shouldn’t have to go on twitter at the station to figure out there’s a major delay.
This isn’t all bad, but I don’t think this is good, either. The system needs to be more reliable.