Most Muni riders at one time or another have shared the experience of standing at a bus stop and glancing at the LED sign which predicts the next bus.
Mysteriously, the sign stays stuck on the same prediction for minutes at a time — a common occurrence, according to transit app company Swyft.
NextBus predictions can be inaccurate 40 percent of the time if a bus is 20 minutes or more away, according to a Swyft study, released Thursday. “NextBus accuracy plummets as it tries to predict arrivals further out in time,” Swyft wrote in a summary of its study findings.
NextBus sent the following statement in response to the study: “In urban environments the time between buses may only be 8 to 20 minutes, so people aren’t looking out 30 minutes for a bus. They want to see the next one or two.”
NextBus also added, “customers are focused on the zero to five minute window.”
For predictions under five minutes, accuracy is above 91 percent in San Francisco, according to NextBus.
The San Francisco Municipal Transportation Agency partners with NextBus to produce real-time arrival predictions. The bus arrival time data’s derived from GPS trackers on individual buses, according to the SFMTA. The data’s piped into LED signs with real-time arrival predictions at bus stops all over San Francisco.
Smartphone app makers like Swyft, Metro San Francisco, Routesy, Pocket Muni and others use the data as the backbone for their bus prediction services.
Swyft collected real-time August predictions from NextBus for “the entire city of San Francisco every minute,” the company said.
Above, Swyft animated its findings in a map showing real-time transit prediction accuracy from NextBus in San Francisco during a morning commute. Each dot represents a Muni stop at a particular time. Darker dots represent stops where the NextBus prediction accuracy was lowest.
It analyzed the predicted arrival time against the actual arrival time of the buses to measure accuracy. When a bus is within 30 minutes of a stop, NextBus predictions are accurate 70 percent of the time, Swyft found. Actual written schedules are only accurate 60 percent of the time, according to SFMTA.
When a vehicle’s within five minutes of a stop, NextBus is accurate 91 percent of the time. When predictions are 25-30 minutes out, NextBus accuracy drops to 59 percent, the report said.
Among the Muni routes with the least accurate predictions, from least to most, were the 82X-Levi Plaza Express, 28-19th Avenue, Muni Metro Bus Shuttle, 81X-Caltrain Express, and 39-Coit.
Swyft gave a margin of error for a bus to arrive 30 seconds early or four and a half minutes late to count as an “accurate prediction.”
To address prediction inaccuracies, Swyft’s smartphone app launched a more robust version of its “crowdsourcing” function, Nov. 30, Swyft CEO Jonny Simkin told the San Francisco Examiner.
To bolster those predictions, Swyft’s beta app now draws on complaints from users of its app to deliver a real-time “heads up” to its users about Muni delays, based on targeted locations of its app users.
The SFMTA has a similar service for its riders, but it sends Twitter, email and text message announcements of delays to users.
Simkin said “Those issues broadcast to everyone, but most people don’t care about the N-Judah (train). We serve the right information at the right time.”
Robert Lyles, an SFMTA spokesman, wrote in an email to the Examiner that “Muni routes are always the focus of improvements – included in that analysis are predictions and customer information. There is no greater way to better service than to improve reliability.
And, “Within Muni,” he wrote, “reliability and predictability work hand-in-hand.” To that end, SFMTA has improved reliability in numerous ways, he said, including the Muni Forward program.
The SFMTA has addressed the issue of inaccurate NextBus predictions and “ghost buses,” which are predicted buses which never show up, in a 2014 blog post titled “NextMuni, Ghost Buses & Trains, Oh My!” and again in a post “Not Your Ordinary GPS – Tracking Your Ride.”
To explain “ghost buses,” SFMTA wrote in its post, “Predictions from the terminal are based on scheduled departure times until the train starts moving, when predictions can be made on vehicle movement. If the train does not depart the terminal as scheduled, then the NextMuni system will drop predictions for the train after a few minutes and drop the subsequent predictions on the route.”
And as for inaccurate predictions, SFMTA wrote that GPS technology aboard buses “takes into account the actual position of the transit vehicle, the intended stops and anticipated traffic patterns. So, when traffic is snarled or your bus or train has a mechanical malfunction, NextMuni predictions often become inaccurate.”
So for instance, if a vehicle is two blocks away from your stop but stuck in traffic, it may read “2 Minutes” away until the bus is able to move – no matter how long it takes.
In that way, the “time” predicted is perhaps, more accurately, a measurement of distance.