One of the striking aspects of the US bike share schemes thus far is that they have generally been dominated by weekend use. There has been some weekday use but not a large commuter morning/evening surge, like has been seen consistently in London. However Washington DC at least seems to have reached a turning point, with the characteristic commuter spikes starting to appear, and a post-pm-peak distribution of bikes that had a distinctly London-esque blue in the middle (few bikes in the centre of Washington DC) and red on the outside (presumably more residential areas).
As Washington DC is joining London’s characteristic commuter “tidal flow”, London itself seems to be moving away from that. “Casual” use combined with unseasonably hot and sunny weather here, has meant a “tourist” afternoon surge, always seen at weekends, is present on weekdays now too. This somewhat dilutes the evening commuter use, although the system still ends up quite unbalanced at the end of the day.
I’ve made a minor adjustment to my bike share map statistics (see the “graph” link) – namely the one on the current number of bikes predicted to be in use. I’ve removed this statistic now. Previously, this assumed that the highest number of bikes available in the preceding 24 hours indicated the lowest moment of use. This is the case if no bikes are marked as faulty (or no docks are added to/removed from to the system) – however it turns out this is a significant number, at least in London. When a bike is marked as faulty, at least on the London system, then both the bike and the dock are removed from the availability numbers. By plotting this on a graph, estimates can be made of the number of bikes being marked as faulty each day. At the moment, this seems to be about 5-10% each day for London. Such bikes then get taken to the workshop, fixed, and appear to be replaced en masse just before the morning rush hour:
So now I look at the minimum number of free docking spaces in the last 24 hours instead. This should avoid the poor numbers after a heavy day of usage, where several hundred still appear to be in use at around 2am, whereas actually they are just bikes that were marked as broken on that day, on the stands. I have instead replaced it with a statistic showing the proportion of docks that don’t have bikes in them. This is effectively the same statistic as before, but now I don’t make assumptions about the “baseline” value, i.e. when no bikes are supposedly being used. In other words, previously I was effectively substracting a percentage from this value, based on the baseline percentage. I’m no longer doing that subtraction.
Finally, Toronto and Tel Aviv have just gone live with their bike share schemes, and have been duly added to the map. Ottawa, Boston and Antwerp are all launching in the next month or so.
[Update – Those missing the # number of bikes measure for Washington DC/Arlington can find it here. These measurements are carried out a different way – by looking at changes in individual docks.]
5 replies on “Washington DC Cycle Commuters Suddenly Appear”
These stats are great to see. I recently picked up a folding bike which made by bike commute possible. I couldnt do the whole distance before but now I take my bike on the train and ride the rest. It’s so much better than sitting in traffic
Mitch – thanks, that’s good to know there was a recent software change that is now removing the broken bikes from the counts. This is much better for the users and app producers, and I definitely think it is a good change. I think it has been incorporated in London’s scheme. It does mean I can no longer show the % bikes in use, as the “total docks” number now varies, and if I assume that is entirely due to broken bikes, it can make the numbers for schemes being built out negative – not good.
Hi Joey,
My worry is more that people are taking these wrong figures and putting them in presentations, doing analysis with them. Now that I know that they are wrong, as the systems are no longer closed, I would prefer not to show knowingly incorrect data. The near-correct figures can still be derived, with care, and I hope to be able to that at some point in the future.
People will generally ignore asterisks and still report the figures as the gospel truth.
Your method is pretty much what I was doing, but is flawed – e.g. Toronto is building out its scheme day by day so will be consistently underreporting by 10-20% at the moment. Conversely London’s scheme, as it is large and bedded in (and because I think Londoners are a bit rougher with their bikes) is seeing quite a high number of bikes marked as faulty, so my metric was overreporting by about 5-10%. There were other issues with Miami Beach, and with cities where the data feed regularly breaks and broadcasts a few, but not all, of the docks. This seems to frequently happen in Milan, Mexico City and Barcelona to name but three.
I don’t like the change. I really liked seeing the utilization rate, even if it isn’t perfect. It at least allowed me in DC to compare usage day to day when I looked at it, even if it’s off by some small percentage.
I hope that you’ll bring it back and just include an asterisk caveat.
Now, I find myself doing the numbers myself by estimating similarly to how you had done it (in my case, subtracting the current bikes-in-docks # from the highest bikes-in-docks # for the day, and then dividing it by that highest figure). It’s just annoying to have to do this every time myself.
Regarding your adjustments to the map statistics, you’re correct that bikes marked as needing repair are not included in the available bike count. There’s another piece to the puzzle which may explain some of the fluctuations you’re seeing. Recent changes allow operators to take docks needing repair out of the available dock count. In the past, if a dock was broken and awaiting service, it still showed on the maps as available. As far as I know, this should apply to all PBSC schemes, although it’s up to the individual operators to make use of it.