Using Data to Prepare for Big Events

The following is a guest post from the MBTA Customer Experience Department.

Photo of Copley Square crowded with people during the Boston Marathon, 2014
Copley Square during the Boston Marathon 2014. Photo credit: Susana Hey, MBTA, 2014

The MBTA sees thousands of visitors during major events, like the Patriots Victory Parade, St. Patrick’s Day Parade, and Boston Marathon. When these big events come to town, it’s all hands on deck: many lines run rush-hour level service all day, extra staff are deployed across stations, and MBTA employees from non-operating departments volunteer to help people who may not be familiar with the T.

We know during these events we will have higher than normal ridership, but in order to prepare we also want to know where and when it will be busiest. We know that stations along a parade route will have high numbers of exits during the event, and stations with large parking garages will have high numbers of boardings before an event, but using data can help us make better decisions for how to deploy staff and service.

Fortunately, most of these events have happened before, so we have data to work with. Ridership data can be very complicated, as discussed in our post Ridership and the National Transit Database and in this Ridership Trends presentation, but we can get a good approximation of which stations will be busiest by looking at past data from the fare gates at subway stations.

Here’s a graph of fare gate taps at Braintree on the day of St. Patrick’s Day Parade last year (Sunday, March 20, 2016), compared to a normal Sunday. (All graphs in the blog post use a 3 point moving average.)

A line graph showing average entries at Braintree by half hour on an average Sunday vs St. Patrick's Day. St. Patrick's Day entries sharply peak between 9 AM and 2PM; an average Sunday's entries remain flat through the day.
Based on this information, we knew that we needed extra staff at Braintree around 9 AM to 2 PM for this year’s St Patrick’s Day Parade on Sunday, March 19.

The situation is a little more complicated when a major event occurs on a holiday, like the Boston Marathon. We know that the MBTA is going to be very busy, but we also know that many regular commuters won’t be using the T to get to work because many businesses are closed.

A line graph showing Back Bay Station entries by 30 minute period for Marathon Monday vs. an average weekday in October 2016. Both graphs are two-humped and have smaller AM peaks between 7-9 AM and larger PM peaks between 2 and 7 PM, but an average weekday has a larger AM peak than Marathon Monday, and Marathon Monday has a larger PM peak than an average weekday.
Looking at just fare gate taps, we can see that Back Bay Station has fewer entries than usual during the morning rush hour, but more entries than usual in the afternoon.

Not all riders use the T in the same way, however. Commuters and other frequent T users know the system well and don’t need much help navigating it. Visitors, on the other hand, may need directions to their final destination or help buying a ticket at a Fare Vending Machine. Because so many visitors come to Boston for these large events, we need different data to help us figure out where we need extra staff. Fortunately, in the same database as fare gate information, we also have Fare Vending Machine (FVM) data. Most commuters have a monthly pass and only use the FVM once a month (or not at all if they are in the Corporate Pass program), while visitors are much more likely to use an FVM to buy a ticket on any given trip. Therefore, using FVM data gives us a better sense of which stations will be crowded, especially in the entrance areas where the FVMs are located.

A line graph of Back Bay Station Fare Vending Machine transactions by 30 minute period for Marathon Monday vs an average weekday in October 2016. Marathon Monday has more taps overall and its peaks are more defined.

So even while fare gate taps are lower in the morning and only slightly higher in the afternoon, FVM transactions are significantly higher all day at Back Bay because of all the visitors for the Marathon. By looking at FVM data across the system, we were able to predict the busiest stations during the 2017 Boston Marathon and make sure we had extra staff posted there. Finally, we followed up by comparing station data from Marathon Monday 2016 with data from this year. We can see that both entries and FVM transactions tracked very closely from one year to the next. It turns out that when it comes to large events on the MBTA, the past is the best predictor of the future.

A line graph showing Back Bay Fare Vending Machine transactions by 30 minute period for Marathon Monday 2016 vs Marathon Monday 2017. The patterns are almost identical.
A line graph showing Back Bay Station Entries by 30-minute periods for Marathon Monday 2016 vs. Marathon Monday 2017. The patterns are almost identical.