Metro’s new stat shows how many 8-car trains may be missing

WMATA just unveiled its official stat for measuring how many rail cars are available. The number helps to show how the Silver Line brought down performance, why on-time performance isn’t where it should be, and why the new 7000-series cars are so important.

An entire day’s operations hinges on how many railcars WMATA can use:

Railcar availability is a key driver of on-time performance and supports the ability to meet the Board standard for crowding. When the availability target is met, scheduled departures of all 8- and 6-car trains from end of line stations is possible. When not enough railcars are available, train lengths are first shortened to six cars, which can contribute to crowding.

When railcar availability dips further and there are not enough trains to depart from end-of-line stations, headways (time between trains) increase, lowering OTP for customers. - WMATA Vital Signs Q3, 2015

The statistic isn’t new in the sense that it’s never been computed before, but it’s now public and gives us a new way to analyze the rail fleet. WMATA computes the availability statistic using two numbers.

First is the number of rail cars that were put into service by 7am, and second is the total number of rail cars in the entire fleet. Dividing the first by the second results in the percentage recorded. Data on this is only available for 2013 to the present, but it’s an extra way to take a look into how system performance is partially dependent on how many cars are available for use.

If everything was perfect, WMATA would have around 20-25% of its rail cars available as “spares:” cars out for scheduled and unscheduled maintenance, as well as larger “engineering campaigns” that might be needed to fix an issue on all cars of a specific series.

I’ve written before about how everything is not perfect, and this spare ratio doesn’t exist. Recently it has dropped to around 11% due to more cars out of service and a general lack of railcars entirely. WMATA has 1134 cars in the fleet and has a maximum requirement of 954 in use at any one time, which leaves 180 rail cars left (16%).

We also know that over 60 (62 as of Thursday November 5th) rail cars were out of service waiting for parts, some of which could have been out for a month or two already. The number of disabled cars works to cut the number further to around 120, or 11% of the fleet left for maintenance.

Reliability, performance, and availability are intertwined

I took a look at some relevant data collected about the rail system and graphed a few pieces. The chart below shows three of them— rail car reliability, the overall system on-time performance, and the rail car availability statistic— and starts to get into the nitty gritty about what’s gone on the past couple of years.

The daily railcar requirement of 954 cars for morning and evening service includes the 64 cars required for the first phase of the Silver Line. When the Silver Line opened in mid-2014, fleet availability rose to a maximum of 88% as 64 extra cars from the fleet were put into the system for morning and evening peak hours. However, as the fleet availability rose, the overall system on-time performance dropped and still appears to be dropping, according to the latest data released from WMATA:

Railcar availability, on-time performance, and fleet reliability are all correlated with each other, although a number of other items contribute as well. A cold winter for instance, like the one we had in 2014-2015, can cause more train car issues, causing fewer to be available for service, thus decreasing on-time performance.

When digging a little further, on-time performance exhibits a loose correlation with three of the main classifications of reported train car issues - door, brake, and general “equipment” problems, however it is far from perfect. As the delays shown would affect vehicle reliability, it shares similar peaks and valleys as the reliability (aka MDBD) metric from the first chart.

Prolonged, increased use of rail cars can lead to issues

There was a distinct drop in on-time performance starting around July/September 2014 when the Silver Line opened, shown in the first chart above. As 64 cars of the rail fleet were put into service at that time, the overall on-time performance appears to have dropped. Since the Silver Line opened before WMATA was able to receive and accept 64 new 7000-series cars that would be needed for service, the ones needed were pulled from the “spare” pool.

Rail car MDBD also drops drastically from October 2014 to February 2015, but this dataset also varies much more than the others and is more affected by things like weather than the others. The drop around February 2015 closely corresponds with the sub-freezing temperatures we experienced last winter, which caused various railcar issues.

The drastic drop of availability in June correlates to when the 100 4000-series rail cars were pulled from service due to door issues. Fleet availability dropped to 79% in June in part due to the lack of 4000’s, and then gradually increased to 82% in September. With a total fleet of 1126 railcars at the time, 80% availability would mean that around 900 cars were available for service on average, well below the 954 cars required for full service. This matches with WMATA’s statement that “availability was below the minimum 954 car threshold every weekday in July and August and most of September.”

By it’s nature, fleet availability changes when WMATA adds and removes rail cars

The number of cars in the fleet will increase as more 7000-series ones are added, which changes how the fleet availability is calculated. 80% availability in June 2015 resulting in 900 available cars is not going to be the same as 80% availability in June 2016, which (theoretically) could equate to 952 or more available cars. However, the way that availability is calculated means it is not a direct relation to how many total rail cars WMATA has. 88% or greater of the fleet could be made available like when the Pope visited in September, but this does not mean WMATA received more rail cars — just that more were put into service. Fleet availability, like many things, is relative.

When fleet availability falls, on-time performance generally decreases as well. All the way back to the beginning of the data in 2013, the on-time performance metric appears to parallel and trend in the same direction as the fleet availability, but experiences a significant separate drop starting in April 2015 and continuing through to September.

The drop in on-time performance and fleet availability within the past year or so is likely being caused by several factors (not to mention the delays caused by single-tracking and other maintenance activities), but one hypothesis is that the lack of rail cars is wearing on maintenance crews. With fewer spare cars that can be out for maintenance, employees are being hard-pressed - officially or not - to fix and return cars to service as quickly as they can. WMATA dropped the number of cars they would run on Mondays and Fridays during the summer to help with the maintenance process. But with less time available in the shop, issues are being forgotten or not fully fixed, leading to more customer-facing breakdowns.

Fleet availability correlates with the number of canceled trains

As the fleet availability drops, the number of trains that don’t run increases. This makes sense in theory, but is neat to also see play out in real life. The graph below plots fleet availability in orange vs the number of monthly canceled trains in blue. Both measures are relatively steady from January 2013 through to May 2014, at which point both start to increase. Cancellations drop in the fall, but then continue to rise more through to a maximum of 159 in February 2015.

In addition, while WMATA does not release public data on it, a decrease in fleet availability is very likely related to how many eight-car trains run. A story WAMU ran earlier this year showed the number of eight-car trains was not matching up with WMATA’s service level goals. With the principle that WMATA generally tries to decrease the number of eight-car trains before cancelling them, this lack of eight-car trains is explained by the data.

Fleet availability takes a quick fall in May 2015 and ends at 79% in June. This drop in availability could be when new, stricter purchasing requirements were put in place, causing more cars to be taken out of service waiting for parts. The removal of the 100 4000-series cars did not occur until June when cancellations spiked to 233, then 248 and 226 for July and August.

The data suggest that railcar availability is yet again on the rise possibly due in part to the slow addition of the 7000-series cars, but has a ways yet to go to stabilize comfortably back around 80%. You can find this data and much more in the quarterly Vital Signs document released by WMATA, along with other performance indicators tracked by the agency. It hopefully helps to show the relationship between the number of rail cars available for service, the total number of rail cars in the fleet, and the resulting performance that customers experience.