Missing the Express: Designing In Train Connecting Train Arrivals for the NYC Subway

A product case study for a PM internship application

TL;DR

Riders on the NYC subway can’t see when trains on other lines will arrive at the same station they’re passing through. With no phone signal underground and no screen telling them otherwise, they routinely miss faster connections that are sitting one platform away. This case study proposes Connecting Trains: a feature on the in car and platform display screens that shows when trains on intersecting lines will arrive at upcoming shared stations, so riders can make smarter transfer decisions in the moment, without a phone or internet connection.

The 7:50 AM problem

Every morning I take the F train into Manhattan to get to work near 42nd Street. The F is a local, so it stops everywhere, and the ride feels longer than it should. A few stops in, the train pulls into Broadway Lafayette Street. This is one of the most useful stations on my commute because the B and D trains stop on the platform right across from me, and the D and B is an express. If a D train happens to be pulling in as I arrive, I can step outside the platform, hop on, and shave several minutes off my trip.

The problem is that I never know whether that D train is coming. There is no cell signal in the tunnel, so I can’t check an app. There is nothing on the train telling me what is happening on the other line. So I make a blind guess. Sometimes I jump off to wait for the express and watch three local F trains go by before a D or a B shows up. Other times I stay on the F, only to see a D train sitting on the opposite platform as my doors close or comes after a minute. Either way, I lose time, and the decision feels like a coin flip rather than a choice.

Problem statement

Subway riders cannot see real time arrival information for connecting lines at the stations they are about to reach. The information they need to make a good transfer decision exists in the MTA’s system, but it never reaches them at the moment and place where the decision has to be made: inside a moving train with no internet access. As a result, riders miss faster routes that are physically within reach, and they make transfer choices based on habit and luck instead of information.

How big is this?

The subway carries millions of riders on a typical weekday. Per the MTA’s own ridership tables, average weekday subway ridership was about 3.7 million in 2024 (3,735,571), down from roughly 5.5 million in 2019 (5,493,875) but still firmly in the millions. Many of those trips involve at least one transfer. Because the system uses a flat fare with free transfers and riders only tap or swipe on entry, the MTA does not directly measure how many trips include a train to train transfer, so the exact share is an estimate rather than a published figure. What matters for this feature is the subset of stations where a rider can quickly switch between a local and an express, sometimes across the same platform. A clear example of that cross platform local/express arrangement is 59th Street-Columbus Circle, which was the seventh busiest station complex in 2024 (about 16.9 million entries); other genuine cross platform cases include 96th Street on the 1/2/3 and Roosevelt Avenue-Jackson Heights on the Queens Boulevard Line. Broadway-Lafayette, the station in my own commute, is a slightly different case: it is a quick connection between the 6 and the Sixth Avenue lines (B/D/F/M) rather than a single shared local/express platform, but the rider’s decision is the same. Even a modest time saving per affected rider, multiplied across many such transfers, would add up to a meaningful amount of collective time returned to commuters (a hypothetical worth validating with MTA ridership and travel-survey data before committing to the project).

Who is this for?

The primary user is the time sensitive commuter: someone with a regular route who knows the system well enough to recognize a good transfer but lacks the information to act on it confidently. A secondary user is the occasional or visiting rider who does not know which transfers are possible at all, and who would benefit from simply being told that a connection exists. Both share the same core need: to know, before the doors open, whether getting off here gets them where they are going faster.

The core insight

A transfer decision is only useful at a specific moment, the few seconds before the doors open at a shared station, and in a specific place, the train itself. Solving this with a phone app fails on its own terms, because the moment of decision happens exactly where phones do not work. The right place to deliver the information is therefore the hardware the rider is already looking at: the digital screens inside newer train cars and the countdown clocks on the platform. That hardware is connected to the MTA’s train control and arrival systems, so it does not depend on the rider having signal at all.

The solution: Connecting Trains

Connecting Trains is a panel on the in car digital screen and platform displays that surfaces, for each upcoming station, the next arrivals of any lines that connect there. As the F train approaches Broadway Lafayette, the screen above the doors shows a line such as: “Next stop: Broadway Lafayette St. Connections: D (express) in 2 min, same platform; B (Express) in 3 min.” A rider who wants the express now has a real choice rather than a guess.

Three details make the feature trustworthy rather than just decorative. First, it labels whether the connecting train is express or local, because that is the entire reason the transfer is worth making. Second, it indicates the transfer effort, whether the connection is across the same platform or requires walking to another level, so the rider can judge if the connection is realistically catchable. Third, it only shows connections the rider can plausibly make given the arrival timing, rather than listing every line at the station, which keeps the display honest and uncluttered.

How it works without signal

The rider never needs a connection. Each train’s onboard display receives arrival predictions for nearby stations through the same data backbone that already powers the platform countdown clocks. When a train is between stations, it knows its next stop and can request the predicted arrivals for the connecting lines at that stop, then render them on the in car screen. Because the data flows to MTA hardware rather than to the rider’s phone, the dead zone in the tunnel stops being a blocker. As a fallback for cars without screens, the same information appears on the platform displays at the shared station and can be echoed in optional onboard audio announcements.

What it looks like

The in car screen already shows the route strip map and the next stop. Connecting Trains adds a compact band beneath the next-stop label. Each connecting line is shown as its colored bullet, a one word tag for express or local, a countdown in minutes, and a small icon for same platform versus walking transfer. The design goal is that a rider can read and act on it in under three seconds, because that is roughly how long they have before deciding whether to stand up. Color, the express or local tag, and the minutes carry the message; everything else is secondary.

Success metrics

The north star metric is rider time saved per transfer eligible trip. Supporting metrics would include the transfer conversion rate (how often riders actually switch when a faster connection is shown), the accuracy of the displayed countdowns against actual arrivals, and the catch rate (how often riders who choose to transfer successfully board the connecting train). On the experience side, I would track rider reported confidence in transfer decisions through station exit surveys, and watch crowding and dwell time metrics to ensure the feature does not create unsafe platform rushes. All targets would be set against a baseline measured before launch. These are proposed metrics, not measured results.

Risks and constraints

The most important risk is trust: if the countdown is wrong even occasionally, riders will jump off for a train that does not come and will stop believing the feature entirely. The product has to be conservative, showing a connection only when confidence is high and degrading gracefully to “check platform” when it is not. A second constraint is hardware coverage, since older cars lack screens; this argues for leaning on platform displays during a phased rollout and treating in car screens as the data improves. A third risk is safety, because surfacing a two minute connection could encourage riders to dash across platforms; the timing logic and copy should never imply a connection is catchable if it realistically is not. Finally, the feature depends on the reliability of the underlying arrival predictions, so I would scope an early phase to the highest confidence shared stations rather than the whole system.

Rollout plan

I would start with a pilot at a handful of high traffic stations where riders frequently choose between a local and an express, beginning with true cross platform transfers like 59th Street Columbus Circle, and using platform displays first because the hardware and data are most mature there. I would also include Broadway Lafayette, the multi line transfer at the heart of my own commute, since the F to D or B decision it enables is exactly the behavior this feature targets. After validating accuracy and measuring the transfer conversion and catch rates, I would expand to in car screens on the equipped fleet, then widen station coverage. Each expansion would be gated on the arrival prediction accuracy clearing a trust threshold for that station.

Why this matters

This feature does not require new trains, new track, or new schedules. It takes information the MTA already has and puts it in front of riders at the one moment it can change a decision. For me that means catching the express at Broadway Lafayette on purpose instead of by accident. Across millions of daily riders, it means turning a coin flip into a choice, and giving people back small pieces of their day, a few minutes at a time.