How data helps bike share operators keep the good times rolling

DC is home to one of the many bike share systems that dot the US. Having launched in September 2010, Capital Bike Share (CaBi for short) is now everywhere in the Capital. Tourists grab the friendly red bikes and use them to jet around the National Mall. Commuters use them to bike to work downtown, forgoing the hassle of the Bus/Train system and the worry of stashing a personal bike for the day.

Joining CaBi was one of the first things I did after moving back to DC after a brief stint up north. I absolutely love it and on mornings when I have my affairs in order, CaBi rewards me with a killer commute that descends from the National Cathedral, skirts the Potomac, and cruises up the National Mall. Not bad, right?

More often than not however, my mornings are anything but orderly. And on days when the sun is shining and the humidity is in check, I often stumble out to an empty bike share station and am forced onto the bus, helmet in hand.

Fortunately for folks like me, there are some good data efforts under way to help manage the problem for individuals and the system at large. Last week, a group of civic hackers from Code for DC released this CaBi Bikeshare Odds app allowing users to estimate the probability of getting a bike based on the time of day.

Code for DC drew data and much of their inspiration from the Data Science for Social Good Bikeshare project. The project works all 5 steps of the Data to Insight chain to help bike share companies re-balance their bike fleets so that people like me stand a decent chance at grabbing a rig when and where we need it.

At a basic level, the Code for DC product works by pulling real-time bike and weather data in through APIs (Get the Data), using the Python programming language to prepare the data for modeling (Ready the Data), running the data through a model built to predict the number of bikes at a station based on things like weather, time of day, number of bikes currently at the station, and historical usage (Analyze the Data), and sharing the results at frequent intervals by way of a simple, map-based web app that shows the current and predicted number of bikes around the city (Explain the Data). For a more detailed description of tools and methods used, check out the team's awesome documentation here!

While fantastic, I think both projects could do more to Explain the Data. By that I mean take the interface development several steps further by making it available to people on the fly. Time-challenged people like me aren't going to power up a computer and navigate to a website. Why not flash relevant information about nearby bike stations and the likely flows in and out of them? That way, your users seamlessly pick up pieces of information that they can use to plan their mornings. Getting information to users seamlessly is the problem that emerging wearables technologies like Google Glass are trying to solve, and something I will explore in a later post.

It is also worth checking out some of the other projects the Data Science for Social Good folks are churning out. As government agencies continue to make more and better data available on issues ranging from health, to transportation, to employment, we'll see more projects like the one above.