So what's going on? Simply put, these are adjustments to the durations (transit times) of certain trains. It is admitting that the May 23 schedules were too aggressive and too optimistic about the durations of those trips.
The May 23 schedules were developed using computer modeling for the durations of the trips. Once the schedules were implemented, real world experience revealed that there were some flaws with the assumptions used in the model. I've heard that the flaws included:
1) Schedules assumed a generic 'dwell time' for how long a train needed to be in a station for loading or unloading. Unloading 200-300 passengers at Framingham on the busiest rush hour train takes more time than unloading 50 passengers at Wellesley Farms. But both stations were assigned the same amount of 'dwell time' - and it wasn't enough for the real world.
2) Schedules did not account for the "Delay in Block" (DIB) rule. This is a pretty technical issue, and you can read about what DIB is in the glossary. Suffice it to say that this rule means that trains approaching some signals are required to go slower than the speed limit would suggest they could operate at. The computer model used for simulating trip durations took into account acceleration and stopping times at station stops, but assumed that trains would operate at or near the speed limit for most of the rest of their journey.
3) Schedules did not account for the slow CP 4 switch at Boston Landing construction. There is a 30 mph speed limit adjacent to the construction zone, and that was accounted for in the model. But due to the way the switch on the west end of Boston Landing had to be built, it has an effective 15 mph limit for trains switching from track 2 to the single track (or from the single track to track 2). This is the move that most rush hour (both AM & PM) trains make. Most of you are familiar with this really slow spot.
The computer modeling is described around page 26 of this PowerPoint.
The three issues above would presumably affect every rush hour train, and perhaps they do. It's possible that they chose to modify the schedules for the worst performing trains only - even if other trains are experiencing consistent ~2 minute delays.
Some important thoughts on these schedule changes:
1) The MBTA creates the schedules and Keolis implements them. Keolis might be consulted and have some input and advice, but the final implementation of any schedule is solely by the MBTA (even the 7/27 schedule adjustments). Minor point, but it is important to remember the client / contractor relationship between the MBTA and Keolis.
2) The 'customer facing' schedules (printed schedule flyers, PDF schedules on the internet, and schedule timetables within apps) will NOT be changing. This is probably because it would be too confusing and expensive to make this change for the minor changes being implemented. I was actually somewhat surprised that they even announced it to the public - they could have kept the change internal to employee timetables and not announced it. I'm glad they did announce it.
3) The announcement only includes the final cumulative time adjustment for the final destination. However, every station stop between departure and destination is actually changing - although that's not documented anywhere and the actual departure time at intermediate stations was probably not even calculated. But it's NOT like the delay is ONLY happening between the second to last stop and last stop. For example, train P523 has had 10 minutes added to it's arrival time in Worcester. This probably means that the arrival time in Framingham will be about 5 minutes later than the 5/23 schedule, with the 'delay' building from 5 to 10 minutes between Framingham and Worcester. Or maybe Framingham is only 2 minutes later and the 8 additional minutes accumulate from Framingham to Worcester.
4) Most of the train durations on the 5/23 schedules were shorter than previous schedules. So the 5/23 schedule was predicting / planning for better performance vs. the old schedules. The changes implemented on 7/27 essentially negate those optimistic predictions / plans for the trains involved. I created this table to illustrate this point (click on the image for a larger version with the backup data included):
"Old" means pre-5/23 schedule
"New" means the 5/23 schedule
"Revised" means the 7/27 schedule adjustments
"Difference (negative value = min faster)" compares the 5/23 schedule to the pre-5/23 schedule
"Difference to 5/23 schedule" compares the 7/27 revisions to the 5/23 schedule
"Difference to original" compares the 7/27 revisions to the pre-5/23 schedule
Notice that the morning train durations were all reduced. The only trip that didn't get a faster duration was P582/P504 which added West Natick as a station stop - but that only cost 1 extra minute of duration. There were no adjustments made to these durations at the 7/27 adjustment announcement, so they must be doing OK.
The afternoon commute is a different and more complicated story. The new P519 train gained one additional stop when compared to the closest 'old' express train (P523). But it operates 20 minutes later (attracting more passengers) and was predicted to operate 2 minutes faster even with those two factors which should have resulted in longer durations! The schedule adjustment for P519 adds 7 minutes vs. the 5/23 schedule but only 5 minutes when compared with pre-5/23 schedules.
The new P521 (5 PM express) was predicted to operate 5 minutes faster on the 5/23 schedule. Those 5 minutes have been 'taken away' and the revised schedule is now 2 minutes longer than the pre-5/23 schedule (no change in number of station stops).
The new P523 (~5:40 PM express) has been a source of much complaining. On the old schedule, this train leapfrogged past the local. The new schedule eliminates this leapfrog move - and now there isn't enough time between the preceeding local (P593) and P523. The addition of 10 minutes to the revised duration is probably meant to account for these issues.
So, with the exception of P523, when you compare the 7/27 revisions to the pre-5/23 schedule, the differences are not that much. In fact, the only train (except P523) that runs more than a few minutes longer than the pre-5/23 schedule is P519 - and that train has one additional station stop.
Finally, there are two factors affecting current trip durations which require mentioning: equipment shortages (and problems) and crew staffing. There is a serious shortage of coaches at the present time. Apparently this is due to a combination of routine maintenance and inspection backlogs along with equipment problems. Shorter trains lead to extended dwell times since more passengers have to squeeze through less doors - or even worse, extended dwell times due to overcrowding and slow loading / unloading as a result. Crew shortages lead to the same problem - not enough doors open = longer dwell times to unload passengers. The crew shortage should be mitigated with a new graduating class of assistant conductors that are scheduled to start in mid-August.
The en-route or 'real-time' equipment failures don't seem to be any more or less severe than summers past (other than the lack of large enough sets due to the out-of-service coaches). But that's just my very unscientific assessment for our line. [Equipment failures on the North side have been epic over the last few weeks.] But it must be mentioned that any equipment failure results in delays that can easily cascade into the PM rush hour - especially the peak 5 PM to 6 PM timeframe.
Permanent schedule changes are ideally aligned with the 6 month period that train crew staffing assignments last. The train crews are allowed to 'bid' on jobs based on seniority - and this bid process happens every six months. More than minor changes to the schedules would require reworking the crew schedules and re-bidding the jobs, which is a time consuming and difficult process. Therefore, there will be two upcoming schedule changes that will result in completely new 'customer facing' schedules - both aligned with the crew staffing assignment 6-month period:
October or November 2016: Expect minor tweaks that will memorialize the 7/27 schedule adjustments. We will also probably see somewhat more than a minor tweak for the 5 PM to 6 PM outbound departing trains to try and fix the conflicts with P593 and P523. Don't expect to see the #HeartToHub train removed from this schedule...but I'll be posting a separate blog post about that soon.
May 2017: Based on construction schedules, the Boston Landing station will not be ready in time for the November schedule change. The actual station construction may be complete, but associated track work may lag into the winter. Therefore, expect Boston Landing station stops to be added to the schedule in May 2017.
Two other items I'll mention here for those of you smart enough to read this whole blog post:
Track 2 tie replacement: Scheduled to start around the third week of September. Will start at CP 4 (near Boston Landing) and progress west from there. I don't know the duration, but expect 2-3 months. And yes, that will mean similar delays to what we saw this spring with the track 1 tie replacement.
P513 / P520: Astute readers of the 5/23 schedule will notice these two trains don't exist - those numbers are skipped. Those trains are actually planned to exist - and you can see where they will fit in the schedule. They have been left off the schedule due to the rail destressing and tie replacement projects. Once those projects are entirely complete, that train pair (which is one set of equipment that goes outbound and then inbound) will be put back on the schedule. Not sure if it will happen in November 2016 or May 2017.