MBTA Crowding: COVID-19

In 2020, I helped Boston's public transportation system tackle a tricky new communication problem for the Agency. During a pandemic, when we are all meant to socially distance - how do we keep riders safe and informed about a new standard for 'crowded' vehicles?

Roles: Principal UX/UI Designer (current), Senior UX/UI Designer
Responsibilities: UX/UI, Research, Prototyping, Usability-Testing, Department-wide Design Standards

April - June 2020

Tools

Sketch
Illustrator
InVision
Zeroheight
Abstract
Zeplin
FullStory
Userlytics

MBTA Customer Technology Department

The crowding project is quite different from my previous work at the transit agency since it lacked a single interface.  Primarily it was a real-world customer communications problem created by the global pandemic that then became a set of data, design, and software solutions.

Crowding data represented in 3rd party application Transit App and interconnected organizations and software tools

Before the Pandemic

Before COVID-19, most customers tolerated crowding. They would board vehicles even if all the seats were filled and aisles full of riders. And internally, the organization had other methods of measuring rider demand that they used when planning seasonal service. As a result, published crowding information was not a significant priority, but for a long time considered just a nice-to-have.

Once there was a communicable virus in the air, data that hadn't been particularly important suddenly became a matter of public health and safety.

Fortunately, our modern buses came equipped with automatic passenger counters (APCs). These sensors could measure the number of people on board a bus at any given time, monitoring doorways interrupted by passengers boarding and exiting a vehicle. Until this project, that data had been available but left dormant and unused.

Where we Started

Once we realized we had this powerful data onboard a good number of our buses our Transit Data team began manually testing and observing APC data throughout service, ensuring accuracy and reliability on a number of important routes. For the most part, they did so by spot-checking security footage aboard vehicles to ensure that the count sent by the APCs matched reality.

Onboard footage used to validate crowding data


While they painstakingly worked on that for priority bus routes, the design portion of the project was to figure out how we should represent and communicate crowdedness on vehicles. As the Principal Designer working on both our rider tools on MBTA.com and in connection with my work on our internal bus operations tools, I was in charge of leading the design across multiple teams in the department.

The resulting data and design was then deployed across the many places in our system where we represented realtime vehicle information:

  • On our website (on maps, schedules, trip planners, and more)
  • On digital station and street signage
  • In data feeds sent to other apps and technologies
Example of crowding on an MBTA e-ink bus sign


Re-Defining the Crowding Standard

There were already standards for crowding information before the pandemic that needed revisiting in this new context. 

The MBTA, like many public transit agencies around the world, shares its data on a feed used by apps like Google Maps and Transit. It's called the General Transit Feed Specification (GTFS). Unfortunately, this original standard had a few problems preventing us from using it as-is.

GTFS crowding standard (pre-pandemic)

These existing labels to define crowding in GTFS all took on a different meaning in the context of social distancing:

MANY_SEATS_AVAILABLE
FEW_SEATS_AVAILABLE
STANDING_ROOM_ONLY
CRUSHED_STANDING_ROOM_ONLY
FULL

Whether 'seats were available' wasn't sufficient to say whether there was space for passengers to board when filling all seats would lead to dangerous public health conditions.

Whenever a new set of data would be published, it was tough to deviate from this international standard. But in extraordinary circumstances, a deviation was necessary and our team was the first agency worldwide to reissue this data.

Communication Conflict

It was simple enough to calculate the numbers of passengers required to classify a crowded vehicle.

What is 'crowded'?
Those raw numbers were a matter of policy, decided at the official level in consultation with our team. For example, buses used to be considered crowded once 50 people were on board; during the pandemic, but during the pandemic the MBTA determined this would be reduced to 20.

When is a bus almost crowded?
While we could define crowding at greater than 20 people. it was hard to say what we should call the bus as it approached that number. For example, words or colors that implied "safe" and "unsafe" sounded helpful to some but alarming and off-putting to others.

Is a bus that is not yet crowded safe?
We needed some nuance as well since crowding also depended on how folks arranged themselves. For example, four strangers on a bus is not a big deal in the abstract. But if they all sit next to each other, we couldn't comfortably call that 'safe' either.

And that's only a slice of the discussion generated around crowding.

Like many design processes, the most challenging part is the sheer number of opinions involved. Of course, it was enough to have input from the department or MBTA leaders, but even the Massachusetts State office also had their say.

In these pressurized situations and decision points, I helped my team urge our stakeholders to allow us to test the labels for each bucket. Specifically, to let us test it with our riders rather than make those calls on their behalf.

Testing the Visuals and Language

A few other transit agencies used very disparate visual representations for crowding. I audited several of those examples to select an enhanced set of cues we would use throughout our applications when talking about crowdedness. Critiquing these other examples helped us get a feel for what to start testing with:

Short and simple wording allows riders to spend less effort decoding meaning.

Warning colors help communicate urgency to those who can see colorization

People icons help convey information visually to those with limited English proficiency or color-blindness

Partnering with our lead user researcher, we showed current riders these schematics in Userlytics (an unmoderated user testing and recruitment tool) and asked:

Which of the following options best represents how you would describe the condition aboard this bus? 

  • Not crowded 
  • Some crowding 
  • Crowded 
  • Very crowded 
  • Full 

The results from these user tests gave us concrete information to counter the flurry of opinions, but also told us what members of the public (often 'essential' workers who still relied on public transportation during the pandemic) would interpret, and personally tolerate onboard our buses.

Examples of crowding data on MBTA.com and interpreted by 3rd party application 'Transit' app

Reception

At the MBTA, it wasn't rare for our work to become the subject of news. Once crowding information was launched to the public it also got some coverage. Beyond this however, it's remained a part of our applications and data feeds.

In just three months, our teams took the Real-Time Crowding Data Project from an idea to a product that made the MBTA the largest American transit agency to publish real-time crowding information on local buses, helping riders get around more safely.

Excerpts from the Boston Globe article about the crowding project:

As part of the live tracking service in the popular smartphone app Transit, and on the T’s website as well, riders can now see one of three labels alongside bus locations: “not crowded,” “some crowding,” or “crowded.”
Back before the virus, the T considered a rush-hour bus to be overcrowded at 56 riders, so the 20-rider threshold is a marked change. The T is not, however, kicking riders off or preventing them from boarding once that threshold is met, but instead will rush extra buses into service to try keeping levels down. The agency is also assuming ridership will remain low for many months, though buses have had a higher rate of riders stick around throughout the pandemic than trains.

Maybe it’s the heightened awareness from the pandemic, but overcrowding is now a much more relative measure; that is, it’s surprising how much more crowded a bus feels with 11 riders than nine. Six months ago most riders would have rejoiced at that level.