Bikeshare operators, governments, and corporate owners should pay close attention to how they could incorporate the discoveries of hackers into improving their systems.
Hack nights centering on Capital Bikeshare are among the best-attended of Transportation Techies’ monthly events, drawing a mix of transit data, programming, and biking infrastructure enthusiasts
The group, sponsored by Mobility Lab, hosted its fourth CaBi Hack Night last Monday at 1776’s new Crystal City offices, where many presenters addressed some of the daily, unavoidable issues users and operators encounter as the bikeshare system grows in size and use.
Two dominant themes arose this time: the tricky issue of bike rebalancing and the likelihood of nabbing that last bike or dock.
The great rebalancing act
The sight of an empty Capital Bikeshare dock on your way to work is a common one, but it might be alleviated with better-optimized rebalancing, according to research by Anna Petrone, a transportation engineering graduate student at the University of Maryland.
As with all bikeshare systems, every weekday the bike stations empty out in the edges of the system in the morning and fill up the downtown. Petrone said that even with 10 rebalancing vans directed from an operations center, it’s hard to provide a consistent bike supply that matches the morning commute. Her proposal? A new algorithm to compute the optimal use of rebalancing vans.
Petrone also noted that rebalancing is the largest single cost to the system, a statistic confirmed by Kim Lucas, D.C.’s Capital Bikeshare manager at the District Department of Transportation, who attended the event. Approximately 55 percent of operating costs are due to rebalancing, she said, with most of that going to personnel expenses. In order to be effective, the vans have to be on the road consistently from 5 a.m to 1 a.m.
Lucas said that because rebalancing is contracted to operator Motivate, it’s up to the company to determine the most effective route. So far, operators have opted for the real-time dispatch model versus scheduling, but do rely on an algorithm for improving the process.
Henry Dunbar, BikeArlington program manager, said that new bike corrals, which can accept unlimited bikes, have helped deal with full docks downtown but have created the equally thorny problem of outer stations emptying out faster than ever.
Petrone noted that it’s also hard to gauge what the real demand is for bikeshare when looking at empty stations – how many more bikes would people use if they were available? As bikeshare continues to expand, capturing this lost usage could boost profits, she added.
Should I stay or should I go?
Matt Caywood, founder and CEO of TransitScreen, is an avid bikeshare user and has occasionally found himself faced with an empty bike dock, wondering whether he should wait or call an Uber.
Looking at Capital Bikeshare data from 2012 to 2014, Caywood zeroed in on the popular Thomas Circle dock and calculated the probability of a bike arriving within five minutes. Around the peak activity times – 9 a.m. and 5 p.m. – the probability was at least 50 percent that a bike would arrive within five minutes. However, at 1 p.m., the probability was less than 20 percent, so it would probably be best to do something else instead of wait.
While he hasn’t run the data for other stations, Caywood theorized that for most bikeshare docks surrounded by a similar mix of residential and office buildings, he expected the trend would replicate.
When a station shows one bike left, he found that even at the busiest hours there is a 60 percent chance that the bike would still be there within five minutes.
Matthew Ficke presents his bikeshare probability tool, Station Hero.
StationHero rescues riders from empty docks
Predicting exactly how many bikes will be in a bikeshare dock is an imperfect science but one to which Matthew Ficke has also tried to add a bit more certainty. Using Capital Bikeshare’s open data, Ficke created a webpage where you can get live data on the odds of a bikeshare station running out or filling up by the time you get there.
The interface is simple. You click on a station and it notes the number of bikes and docks currently available. It then calculates the probabilities of all bikes being taken and the dock becoming full in the next 15 minutes.
CaBiBrags makes bikeshare miles competitive
Malynda Chizeck first fell in love with bikeshare on Chicago’s Divvy bikes. After becoming a member, she regularly racked up more than 100 miles per month, including a memorable July 2014, when she hit 225 miles.
A part of her zeal for clocking serious bikeshare distance came from DivvyBrags, a chrome extension that allowed Divvy bikeshare users to upload and compare their monthly stats. Originally a stand-alone website, it became a Chrome extension after Divvy cried foul over having users plug their credentials into an unaffiliated party’s site.
Originally just a fan of the page, she eventually took over the project in June 2014 from creator Alex Soble after he lost interest in the project. When she moved to D.C. earlier this year, Chizeck brought the idea with her, and on October 15, she launched CaBiBrags for the Capital Bikeshare system..
The CaBiBrags Chrome extension includes stats for the number of trips, time spent biking, and the approximate distance traveled. The leaderboard is calculated by cumulative and monthly miles. With only four users as of October 26, Chizeck’s top spot was secure for the time being – but she hopes it will grow into the hundreds of users that DivvyBrags has accumulated.
… And a humble request: More consistent data
Mike Azar of the blog chart-it is a huge fan of Capital Bikeshare’s open-data policies but noted that the inconsistent terminology, column order, and even ride duration format makes it very time consuming for civic hackers to comb through and tease out the most interesting information.
“If you spend time, there really is a treasure of information there,” he said of Capital Bikeshare’s statistics.
His approach is also not that of a programmer, but of a data analyst charting trends. Based on his analysis of bikeshare trips from 2012 to 2014, Azar identified a clear seasonality to the system’s popularity. While usage by unregistered, or casual, and registered users has grown annually in the high-use seasons (spring through fall), usage has stayed almost exactly flat from winter to winter. He said this raises the important question of how to get riders to use bikeshare more in the winter – perhaps lobster-style mittens on each bike?
Looking at future challenges for the system, Azar noted that 25 percent of the current bike fleet is four years old, with bikes having an average lifespan of seven years – how will the need to replace bikes impact service down the road?
He also found that there was a clear difference in the riding habits of casual versus registered users. Registered users make up 80 percent of bikeshare riders and the average ride time is nine minutes. For casual users, however, the average ride time is 24 minutes.
And, of seasonal interest, Azar found that the most-used bikeshare stations on Halloween night are in Rosslyn. Spooky.
For more details on presenters and future Transportation Techies events, see the group’s Meetup page.
Photos: At top, Anna Petrone discusses rebalancing. (M.V. Jantzen, Flickr)