Posted by: jdantos | February 7, 2012

Capital Bikeshare Data, Part 6

Next in a series of posts mining crowd-sourced Capital Bikeshare data. This one focuses on net “balanced-ness” across the system. See also parts 1, 2, 3, 4, 5, 6, 7, 8, 9.

Usually, the biggest cost in a transportation network is labor. It takes non-travelers to operate and maintain cars, buses, and trains, and there’s little room for mechanical substitution. The majority of most transit agencies’ operating resources in America go to people, not capital. Intrigued by the idea that bike-sharing might reduce operating costs by letting other travelers move the system back towards equilibrium, I wanted to delve further into the role of “rebalancing” in the bikeshare system. I’ve heard a rumor that Montreal’s bike share system barely needs re-balancing, which intrigues me. Are the “peaked” demand patterns (temporally and spatially) in Washington driving the need for re-balancing and its costs?

We can look at the trip-making patterns in the system and try to understand how far off the system is from being inherently self-balancing. When a station shows big net differences between trips supplied and received, we can guess it’s re-balancing.

So, are certain times of the day more “senders” to the next hour than “receivers”? Between 8:00 and 8:30am, how many bikes are taken out vs. returned? In other words, what are the most popular hour-boundary lines for rentals to cross? For each half-hour chunk of the day, do more trips begin (net “sender”) or end (net “receiver”)? It’s a bit arbitrary, but since most trips are 30 minutes or less, I divided the day into half-hour chunks, 2011 only:

Which half-hour chunks are net Senders/(Receivers) of trips? It’s a little arbitrary, but it does shed light on usage patterns over the day.

This shows that, for example, between 8:00 and 8:30 am, 7,000 more trips began than ended in 2011. In fact, from 7:00 to 8:30am, the system is still getting going – the rate of change is going up, and more bikes are out than were before. Between 8:30 and 9:30 am, the trend reverses and the system has started to trend back towards more bikes docked than active, and it stays that way ’til 10:00 am. Midday is inconsistent, but shows some lopsided activity in the 10-11:30am timeframe – maybe this is tourists taking out long rentals? Remember that systemwide there’s a “mid-morning lull,” but this chart suggests there’s a net out-flow of bikes leading up to noontime.

Again, this is a bit arbitrary since a trip from 8:50am and 8:59am would show up as net zero in the above graph, but a trip from 8:50 am to 9:01 am would show up as +1 net sender for the 8:30 time and -1 net receiver for the 9:00 am timeframe. But, these raw numbers end up being 10-30% of all trips begun in a 30-minute window, so we’re not talking noise here, and it gives you a sense of the net inflows and outflows of bikes.

A huge chunk of people take out bikes in the 4:30-5:00pm hour, and return them after 5:00pm. Then, in the evenings, the system is on net trending slightly downwards in activity from 5:00pm onwards. (BTW math geeks – is this kind of the derivative (rate of change) of overall trip-making? Calculus was many moons ago for me).

I imagine this pattern is markedly different on weekends (note the smaller scale on this graph):

Capital Bikeshare Net Trips Supplied or Received by Half-Hour, Weekends Only, 2011

More interesting would be to see where the net in-flows and out-flows are occurring. So, I took the graph in Part 5 and broke it down by net trips sent/received by zone (zones defined in Part 4), and by time of day – before noon and after noon.

Capital Bikeshare Net Trips Supplied or Received, by Zone

This one is pretty interesting – for example, you can see that Downtown is a huge net importer of trips in the morning, and exporter of trips in the afternoon/evening. On the whole, it’s a net receiver of trips, which we saw in Part 4, this shows the imbalance over time as well. Mid-City, Mid-City North, and Capitol Hill are all net “senders” of trips in the morning, and the trend reverses in the afternoon. The Mall is more like downtown. Other places like Rosslyn-Ballston, Crystal/Pentagon City, SW/Near SE, and upper NE and NW look more in balance, although their overall level of activity is lower (Rosslyn-Ballston, remember, is a relative newcomer in 2011).

But how does this metric of “balanced-ness” compare to the total trips in and out of these zones? I mean, downtown looks pretty imbalanced over time in the above graph, but there’s also a huge number of docks and stations in downtown, so it’s not necessarily a problem. Here’s total in-flows and out-flows before and after noon, plus the total number of originating trips in each zone, to put the sender-vs-receiver numbers in context:

Viewed in context of overall trip activity, the relative “balancedness” of Bikeshare stations doesn’t seem so off-kilter.

In percentage terms, actually Mid-City North (Petwork, Columbia Heights, etc.) and upper Northeast is more off-balance than downtown. Capitol Hill isn’t far behind. These neighborhoods generate significantly more trips in the morning, but don’t receive an equal amount in the afternoon to compensate.

Here are the top 10 “supplier” or “sender” stations, before and after noon. Interestingly, Lincoln Park is the single biggest net “sender” of trips in the morning, meaning it is the single most lopsided station in the mornings in 2011. This stations is followed quickly by several stations in the Mid-City neighborhoods – U Street, Logan Circle, Adams Morgan, etc. In the afternoon, the most lopsided stations are downtown, especially in federal office areas where there’s little activity at night. However, I was surprised to see the uneven-ness at Foggy Bottom and 14th and G Streets NW – any idea what’s going on here?

Capital Bikeshare Top 10 Sender Stations Before and After Noon

From a financial sustainability perspective, it’s probably okay for a station(s) to show an imbalance over the course of the day. The question is whether the station or area gets full or empty before the tide begins to change back the other way, and balance is restored without the cost of a re-balancing team.

What do you think? What’s the end-game for financial sustainability, given the clearly “peaked” patterns of demand in Washington? How peaked is the demand really, unconstrained? Is that a useful thing to measure?



  1. […] Next in a series of posts mining crowd-sourced Capital Bikeshare data. This one maps a bunch of data by station. See also parts 1, 2, 3, 4, 5, and 6. […]

  2. […] between temperature and usage. If you’re really bored, see also parts 1, 2, 3, 4, 5, 6,7, and […]

  3. […] data. This post focuses on system-level usage by a few dimensions. Check out parts 2, 3, 4, 5, 6, 7 (maps of travel patterns), 8, […]

  4. […] usage data. This post focuses on system-level usage by trip duration. See parts 1, 3, and 4, 5, 6, 7 (maps of travel patterns), 8, […]

  5. […] at peak times/places –The weekday work commute trips are high-demand times for CaBi, and the highest variation in peak-period usage is very spatially concentrated.  We can continue to add stations in peak-demand origins and […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s


%d bloggers like this: