As a follow-up to the alternately captivating and disturbing transcription of my 12-minute interrogatory of an invisible seated Terry Bellamy yesterday, I’ve decided that a post on the super-sexy subject of bikeshare performance measurement is just the thing to seal your lifelong readership. “Whattaya mean, ‘shut up’?”
So, in a comment on a prior bikeshare post, Richard Layman noted in the context of a discussion about a particular bikeshare system characterized as ‘successful’, “…if you define everything as successful (through constantly changing definitions) then you don’t learn and you don’t improve.” For the time being, I am not going to touch on the full basket-load of different areas where a bikeshare system should measure and benchmark, such as finances and customer service metrics, and will concentrate solely on a preferred ridership performance metric of mine.
A frequently-discussed metric is “rides per bike, per day.” Which makes sense for news stories, dinner conversation, and to be frank, easy understandable interpretation. But for a variety of reasons, I would propose that all of us interested bikeshare bystanders begin talking about “rides per dock per day.” In measuring performance, I think that we should cede a little on intuition for a lot more precision, and here’s why:
- The bike quantity is too inconsistent as a measurement basis – How many bicycles a bikeshare system actually has deployed in use varies greatly, even intraday, and for reasons that can relate to operational performance. For example, bikes undergoing shop maintenance are out of circulation. Systems requiring large-scale intraday redistributions require taking bikes out of service routinely. Systems with relatively lower ridership rates may also put fewer bikes in docks to reduce “dockblocking.” The more bikes out of service, for whatever purpose and for whatever means, is bikes unavailable for riding.
- This rewards low bike/dock ratio systems – A system with a lower bike deployment level by design to ensure system balance is using both bike and dock assets inefficiently by comparison. A system with a low bike/dock ratio due to maintenance or rebalancing time is rewarded with a skewed ridership picture stemming from operational inefficiency.
- The dock is universally comparable – Unlike bikes, the supply of dock spaces is, on a day-to-day basis, relatively stable.
- The dock is the true capital asset on which to measure operational performance – The dock makes up the majority of durable capital cost of a bikeshare system. Not only in the direct cost of purchase, but also the overhead labor necessary to locate and install a station, the rental or opportunity cost of the space on which the station is placed, and the longer service life of the station. All of these argue for measuring performance by how many rides the system managers can squeeze out of each dock space, rather than each bike, which is an asset with a much shorter service life and is laden with variable costs associated with use.
- The dock is the thing that matters most – The scope of a bikeshare system is defined by the number of docks. The number of bikes really doesn’t matter, as long as there is at least one bike at their origin. The availability of docks at as many of their desired destinations as possible is what drives ridership. The more rides generated by each of those dock spaces is the best measure of how a system performs in creating bike rides.
System A has 12,000 docks, around 8,000 bikes deployed on any given day, and cranks out about 60,000 rides per day. Yikes!
System B, meanwhile, only has about 3,000 docks, has about 1,500 bikes deployed on an average day, and does about 11,000 rides per day. Not too shabby either.
If we look at rides per bike, per day, these systems look about the same. Every bike in System A is getting ridden 7.5 times per day, while every bike in System B gets ridden about 7.3 times. We would applaud System B for wringing as many rides out of each bike as the titans at System A.
But as I argued above, it is the number of docks that is the better basis on which to judge performance. And each dock space installed as a part of System A is generating 5 rides per day. Meanwhile, each dock in System B, which deploys fewer bikes per dock space, is only generating about 3.7 rides per day. System B either has excess dock assets, or is operating at less efficiency in generating ridership, than System A, as revealed when we measure rides per dock per day.
Well, I’ve convinced myself, which is always Job 1. So, in the weeks to come, I will try to start compiling rides per dock per day metrics for bikeshare systems that publicly release their data, and hopefully, we’ll have a more uniform method for measuring ridership performance.