Boulder evaluates right-sizing streets to encourage biking and walking

The City of Boulder’s “Go Boulder” organization is seeking input on right-sizing car-centric corridors to improve biking, walking, and motorist safety. I ask the transportation advisory board and city council to support these changes.

Here are the “Complete Street” corridors the city will evaluate:

Map of Boulder potential right-sizing streets

Map of Boulder potential right-sizing streets

Iris Avenue Conditions

I want to analyze the current conditions of one of these roads I bike on ~500 times per year (approximately twice every workday), Iris Avenue in North Boulder. On Saturday, May 2nd, I spent an hour photographing the corridor.

Iris’ bike lanes are in disrepair where vehicles routinely drive over the bike lanes.


The current bike lane design includes areas that dangerously squeeze cyclists, as demonstrated by how vehicles have removed the bike lane paint from the road.

Car drivers routinely drive in the bike lanes.

The width of the lanes vary, from as wide as 4.5 feet in some places, to just over three feet (39”) on the northwest side of the corner.


This portion of the bike lane is only 39″ wide.

Right-Sizing Our Streets

Here’s a comparison image between the current street and a mock-up provided by Go Boulder:

Here is a street-view comparison of how the corridor might change. The bike lanes could increase from the current 3-4 feet to 5-6 feet with 2-3 feet of buffer zone.


iris-ave-right-size (3)

Iris is one of the few East/West bike corridors but it’s used significantly less by bikes than other thoroughfares in the city. A Strava heat map (based on recreational cyclists who use Strava) shows its use versus surrounding corridors:

strava heatmap iris

The Urgency of Improving Our Streets

Boulder has had a strong history of biking and walking, and now has the opportunity to make those amenities accessible to a much larger part of the population. Here are a few reasons to do so:

  • To give residents and their families safe, economical options to move around the city
  • To reduce the need or desire to drive when taking short trips through town or to downtown
  • To improve safety for motorists by reducing the speed in these corridors and adopt a safer lane configuration
  • To design streets that work better for more people at all times of day rather than focusing on only rush-hour or peak traffic periods
  • To align with the vision set forth in the Transportation Master Plan relating to mode share, energy, and vehicle use

The demand for bike infrastructure exists. We need to provide safe spaces that are equitable for all users of the road.

A woman rides with her child on the sidewalk because the bicycle lanes are not safe.


Posted in Uncategorized | Leave a comment

How Strava Could Embrace Bike Commuting and Electric Bikes

Strava, the popular GPS-based activity tracker for athletes, has developed a strong following among cyclists and runners worldwide. The service offers competitive aspects on “segments” of road or trail, tracking best times or performance to compare against oneself or others.

Even though Strava’s team designs with athletes in mind, it’s shown a potential to serve a larger and more general audience with the global heatmap and activity playback using data aggregation in urban areas.

Strava Global Heatmap (Boulder, Colorado)

Strava Global Heatmap (Boulder, Colorado)



Why might Strava invest resources into such features? By embracing bike commuters and e-bikes, both the company and end-users would benefit by:

1. Growing a huge untapped market of new users

Bike commuters are the fastest growing group of cyclists in the US and a substantial percentage of bike riders around the world. E-bike users are a small but growing group of riders that will likely increase as tech advances improve price and availability of these bikes. 

2. Increasing the number of paying users

While commuters and more e-bike riders may want to start riding with Strava, the service doesn’t tailor to non-athletic uses. Providing these users a similar value proposition as for athletes could lead to higher use and willingness to pay for a membership. Many of Strava’s users commute to full time jobs and would benefit from improved ways to track active transportation.

3. Leveraging excellence from athletics into e-bikes

While electric bikes do provide motorized assist, they also provide an opportunity to analyze fitness. E-bikes measure power output contributed by the rider, a feature that a vast majority of ordinary bikes do not provide. By highlighting the fitness elements of e-bikes while downplaying the speed/time elements, Strava could also give a great user experience for e-bike riders that fits in the context of health and athletic performance.

4. Getting better data

Because Strava is athlete-focused, heatmaps and routes have a likely bias toward younger, fitter, often-male cyclists. By also catering to commuters and more casual riders, Strava could better balance its data set and provide more value to cities that may be interested in using the data for planning purposes.


So, how?

Without prying into Strava’s business model, it’s hard to make a case that the company should pursue these features. I’m going to focus on the ‘how’ – if their priorities aligned, what are a few cool things their team could do to meet these ends?

1. Interface enhancements

One simple change would be to add an option in the post-ride screen to mark a ride as a commute. In the right picture, you see the Desktop interface which allows a user to mark a ride as a commute (so Strava already supports the feature on the back-end), but unfortunately there’s no button in the mobile app (left) mark commutes (I’ve added one).

Strava Mobile (added commute button)


Strava Desktop commute tagging (existing layout)










Other enhancements might include:

  • Optimizing the mobile app startup process to allow quicker ride starts. For long recreational rides, startup time matters less since it’s a low proportion of total ride time, but matters more when a total commute ride may be 5-20 minutes.

*** Update 1/11 – Matt Laroche informed me there’s now a quick start option from mobile home screens:

Strava fast record

Strava fast record

  • Use previously recorded user patterns (or any rides under a certain distance) to auto-populate the commute checkbox post-ride. For instance, if I ride point-to-point and previously select a route as a commute, Strava could check the starting/ending locations compared to previous commutes and pre-mark the commute check-box.

2. Highlight data that commuters care about

Once collecting reliability commute data, Strava could provide metrics more interesting to people who bike commute, such as:

  1. Use average end-to-end time instead of just riding time
  2. Summarize information about commutes (number, time, total distance, average distance) per week, month, year
  3. Gamify The Commute – make it social, and connect people who live nearby or ride the same routes
  4. Provide a “City View” option to enhance the Activity Feed: instead of a chronological feed of mostly text, display the day’s rides as a map where routes of people you follow overlay onto your city. In fact, some of this tech could be adapted from the “activity playback” feature, with a mock-up below:

A city-based Activity Feed

As a commuter in your city, you may develop connections with friends who also use Strava to track their biking or running trips around town. A view of the city will be much more interesting for that type of user than a series of short, disjointed rides.

3. Make electric-assist bikes a feature like power meters

Here’s a screenshot from data captured at a recent test event for the Copenhagen Wheel, an add-on for regular bikes that converts them into electric-assist bikes. Newer e-bikes have sophisticated power measurement techniques that can isolate the human-powered component, which Strava could use just as if the rider had a built-in power meter.

MyCopenhagenWheelDemoRide crop

The fundamentals of athletic training carry over to e-bikes, and could become more prevalent among bike riders than power meters (the latter being a $600+ investment). This is a basis for some of James Peterman’s work at CU-Boulder who’s analyzing the fitness benefits of sedentary individuals who begin using e-bikes. (Contact Jim if you’re interested in this study)

The same concepts of power-based training still apply: duration (time), functional threshold (sustainable power),  intensity factor (power output relative to sustainable power). All can be used to improve athletic performance for e-bike riders in the same manner used for elite athletes.



Wrapping up

To finish where we started: Strava aims to be the best GPS-based tracking software in athletics—but I hope to have shown some compelling reasons to see the service more deeply. Strava could become a nexus for all active transportation, not only the athletically-based kind. Athletes often commute, and people who start to use the service as commuters may transition into using it as athletes.

Posted in Uncategorized | Leave a comment

Geography, movements, and other tidbits from Boulder’s B-cycle bike sharing system

I put together an analysis of the 2013 Boulder B-cycle data and wanted to share a few interesting tidbits! Here’s a link to a PDF of the full presentation for anyone who wants dive right in: Boulder B-cycle 2013 Analysis


as a person in a city

I want to understand the layout/flows of bikeshare systems

so I can move around efficiently

System Layout

In 2013, the Boulder B-cycle system had:

  • 22 Stations
  • 276 Docks
  • 138 Bikes

The largest station had seventeen docks while the smallest had nine. Note the large grey dot which marks the geographic center of the system, which did not have a station in 2013 (but does now!)

2013 station layout

Elevation and distance from the center:

2013 Station distance from center and elevation

System Usage

The basic 2013 Boulder B-cycle usage statistics:

  • Total 2013 rides: 28,256
  • Median duration: 14 min
  • Median distance*: .77 mi

Note that distance is calculated “as the crow flies” and understates actual distance traveled, but can be useful as a proxy for distance.

Here’s a map showing each station, with the largest bubbles having the most in/out trips while the smallest have the least:

2013 most active stations map


Right away, we notice that the two most likely indicators of trip count are both proximity to downtown Boulder and proximity to the geographic center of the bike-sharing system.

The majority of trips are under 60 minutes and travel less than 1.5 miles.

2013 Distance and Duration


On an hourly basis, total bicycles checked out peaks at 11AM and has a second, lesser peak at 5PM. Casual users peak gradually in the mid-afternoon while annual users patterns are less uniform. Also note that Boulder B-cycle was only available between 5AM and midnight in 2013.

2013 24-hour usage profile


When accounting for day of week, we see more and later-in-the-day usage toward the end of the workweek. And total usage on weekends for all users more closely mimics that of casual users in general.

2013 24-hour usage profile per day

But another interesting question – how does this usage change in different seasons, with sunlight and temperature likely being big factors?

In the colder/darker months, September through April, we see a much sharper usage pattern that focuses on warm daylight hours:

2013 24-hour September through April

In the summer, we see a much wider usage pattern, extending more heavily into the evening, with many more total riders:

2013 24-hour May through August

We see total ridership reflected in the next graph, with significant increases in the summer months (much of the increase reflecting many tourists in Boulder):

2013 Daily Rides


Station Usage Factors

Since we’ve looked at total rides, now let’s look at where people ride to/from the most. Bike-sharing is point-to-point instead of out-and-back, the system allows for net increases or decreases in the number of bikes available at any given station (which requires system rebalancing).

Stations with large bubbles marked in green are net positive in bikes, and small bubbles in yellow/orange are net negative in bikes:

2013 flows map


Let’s look at some of the factors that may be in play to explain why stations are net positive or net negative.

We see there’s a correlation between distance from center and a net decrease in bikes at a station.

2013 station net change vs distance from center


Another attribute we’d like to test is station elevation. This chart shows an inverse correlation between station elevation and net change in bikes. There may be several factors at play, but generally people have an easier time riding downhill than up.

2013 station net change vs elevation

Other Usage Factors

How sunlight trends with ridership.

2013 Daily Rides vs sunlight

How high temperature affects ridership.

2013 Daily Rides vs temperature

How precipitation affects ridership. Yes, Boulder did get 9 inches of rain on Thursday, September 12th, 2013, during the history flooding event.

2013 Rides vs precipitation

A closer look at ridership during the 2013 flooding:

2013 September Rides vs precipitation

Hopefully this helps understand a bit more about the bikesharing patterns in Boulder. The system has since added many more stations so I’ll hope to update some of the graphs in the future to see how things change. Please send me a comment or note if you have any questions!

@ericmbudd on Twitter.


Posted in Uncategorized | Leave a comment

Facebook is the suburbs while Twitter is the city

Facebook headquarters

Social networks enable varied forms of interaction between their users, through spectrums of openness, hierarchy, and discovery. Facebook and Twitter are the most used services to connect people socially, but bring people together in surprisingly different ways. Facebook’s strengths rely on easily connecting with established networks, showing highlights, and sharing meta-actions (like posting pictures or events). Twitter’s advantages lie in instant communication, building ad-hoc networks, and providing public and widely accessible information.

Why compare social networks to geographical networks? I’ll argue that the same openness, hierarchy, and discovery also applies to suburbs and the city, greatly affecting our modes of relationship. Suburbs span a large area, creating both silos of community and a greater privacy. Suburbs aren’t great for meeting people, but they do provide a framework for connecting disjointed entities to a center.

Twitter headquarters

In comparison, cities move quickly and connect people through greater density. People gather at the local events or meetups, making new friendships from shared interests. Rather than surrounding a center, the city is the center.

The Chasm of Public versus Private

Probably the most often used word in a sentence concerning “Facebook” is privacy, a concept almost never discussed about Twitter. Many Facebook posts are marked ‘for friends only’, and some are concerned over the privacy issues with ‘mutual friends’. Conversations happen in response to a post and comments—very much in a stimulus/response methodology. While commenters sometimes talk among themselves in a post, a trend emerges: the commenters are either friends already, or there will likely be no lasting relationship between them. Compared to Twitter, responses to an original tweet may invoke conversation between responders (sometimes facilitated by the original tweeter), and at its conclusion, responders may choose to follow one another, building a network link between them where one previously didn’t exist and being a part of future cross-talk. Choosing to follow someone on Twitter denotes a more casual relationship than befriending someone on Facebook.

Facebook interactions tend to center around the originator, making only weak links between the other parties.

These human interactions mirror those of a suburb versus a city. Imagine the conversation on the Facebook post as a house party in the ‘burbs. The friend of a friend that you met has context per your conversation at the party, but you may have little connection afterward. And without much way to build a relationship between the two, you may never cross paths again perhaps unless you happen to be at your friend’s house. But Twitter is like meeting that friend of a friend at your local coffee shop that you three frequent. You can establish a few things in common, and if you go there in any frequency, you’ll meet up again and potentially create that new relationship. Not only that, but because your conversations are public, like a group of 10-15 people meeting at that coffeeshop, new organic connections will be made between individuals with shared interests much more easily than in the private confines of a home.

Screen Shot 2013-06-03 at 9.36.26 PM

Twitter interactions have fewer barriers to growing beyond the initiator.

Crosstalk is a feature of Twitter where, if two people you follow converse with messages to each other, these notes will show up in your timeline. Facebook has a similar feature—if one posts to a mutual friend’s wall, you’ll often see this in your Facebook timeline. But on Twitter, conversations are natural—encouraging others to pipe in—as they are clearly public, after all. Twitter serves as an equivalent of a public forum, rather than overhearing someone’s conversation, as there is no expectation of privacy.

The Destination versus The Journey

Suburbs classically keep their gems along huge arterial roads, surrounded by seas of parking lots. Drivers often complain about traffic while traveling on unsafe roads, having to walk across a large expanses of blacktop, traveling five or ten miles from their homes to a destination. City-living people often walk, bike or bus a few minutes from their homes, to a small and livable place, built at human scale. The experience that can be enjoyable—or terrifying—like our interface through Facebook or Twitter.

Facebook posts are often post-facto—reflections, pictures, collection of thoughts—in which the things you see are selected by an algorithm, based on likes, comments, or how much a company has paid for you to see it. It’s prepared. Sometimes stale. May be non-chronological. Packaged for sitting on shelves in a big box store, staying ‘fresh’ for months. Twitter, on the other hand, has fresh vegetables from the farmer’s market downtown. The fish from the sea this morning (sometimes with a less-than-pleasant smell). Rather than buying a five pound box, you get just what you need for breakfast this morning.

Tweets are often not fully-processed, mass-produced packages. They have defects, or they don’t sufficiently capture an idea. But sometimes they capture someone’s first thought in the morning, or an observation walking down the street not significant enough to warrant a post to Facebook. And sometimes that thought comes in a series of three, four, or five tweets—the timing, organic nature matters, and gives a unique flavor that big-box chain restaurants will consistently fail to deliver.

Choosing your friends versus Befriending your neighbors

When someone retweets someone else’s tweet into your timeline, it’s introducing something new into your system. The tweet is reproduced whole, with no comment, and including the original poster’s avatar. At first, retweets can feel almost like a breach of friendship—similar to a friend letting an unknown stranger into your home. It’s not uncommon for new Twitter users to ask how to turn off retweets in their timeline.

This similar occurrence doesn’t wholly exist on Facebook, but people can share an item (usually with comment). Instead, Facebook’s filtering isolates you from unwanted intrusions. Your Facebook timeline’s stories are likely influenced from a friend’s likes or comments, or Facebook’s algorithms show you things it determines you may want to see. Getting that semi-random retweet from a person that’s not expected, or sometimes not particularly wanted, doesn’t exist. You can move further out of town to stop this (i.e. disable retweets from the retweeter) . You can put up a fence (blocking the retweeted person). But you live in a dense neighborhood which by default accepts diversity over homogeny.

Mapping existing networks versus Building new networks

Friends don’t let friends move to the suburbs. Why? Because often it’s a precursor to devolving friendships rather than building them. Partly the lack of density strains friendships, requiring travel to visit old friends (suburban or urban). People keep their friendships when moving out of the city, but the friendship structure is now virtually out of balance with reality. The same is often true with Facebook.

Facebook communities and friendships are descriptive—they exist to mirror an external friendship/relationship (school, family, acquaintances . Twitter communities and friendships are conscriptive—they’re often building a relationship that wasn’t there prior to Twitter (location based, interests, learning). The problem with Facebook networks is that they quickly get out of date, stale, and siloed.

Living versus Deteriorating

Twitter is the ever-changing city. There are always new people to meet (follow) and old friendships may lapse when it’s time (unfollowing). Time is a constraint. There is a limit to how many people, things, and ideas a single person can care about. Twitter rarely has the nostalgia factor that keeps so many dead Facebook friendships alive. Yet Facebook allows relations to persist like an old, dead mall that’s long past its relevance.

@ericmbudd on twitter

Posted in Uncategorized | 69 Comments

bikeshare: analyzing incentives to increase annual members

With any system, every potential member has a price he or she is willing to pay for the service provided. When deciding whether to purchase an annual bikeshare membership, customers often have only one option: the list price, unless a special discount applies. In essence, system operators try to balance two variables: appropriate revenue for the system to cover its costs, while serving the general public with an affordable transit option.

When setting price, system operators must have an understanding of what potential customers will pay for their service. Those who value the service highly would often pay more than the list price, while those with better alternatives would pay significantly less for the same service. Because we have limited data on customer preferences (without issuing a survey or having other transit data), we can start by analyzing the general needs of particular transit groups, and estimate an average cost change over the current mode of transit. The end goal will be to understand how variable pricing can increase the number of annual bikeshare members.

Each mode will be ranked individually (grouped by transit mode) with these parameters:

1. Willingness to pay (demand) – low, medium, high
2. Cost change (-/+) over current mode

The chart below is the end result, after the analysis has been complete. Later in the article, I’ll discuss each transit mode to justify the rankings.

With this ranking, we can now try to plot this to desired price-points along this spectrum. Again, we don’t have hard data yet, so let’s simply try to anchor the upper and lower bounds.

On the low end, imagine a single parent with several young kids whom all need transportation to school; she may like the idea of bike sharing, but her needs are simply not fulfilled by the service – we’ll set her willingness to pay at $20/year. On the high end, imagine a young male who lives downtown and his primary mode of transit is walking/busing and doesn’t want to own a car or his own bicycle – we’ll set his willingness at $130/year.

The rest of the points in the range will now be interpolated through our rankings:

Another reference point we add in the average full price, based on prices from Boulder – $55, Denver – $80, and DC – $75.

Now that our framework is built, I’ll provide further justification for these rankings. Moving from the least ideal demographics to the most ideal, we’ll discuss each transit group’s incentives to use bikeshare and their potential costs.

Least-ideal demographics

family vehicle, L++

Families usually need flexibility, including a method to transport younger children – an automobile is the most common choice. If these families use bicycles, they will often purchase bikes that are capable of transporting children instead of using bikeshare, which can only transport one person each. Bikeshare will be an added cost over a car and/or other commuter bicycles, and may not meet their needs.

bike commuters (non-bikeshare), L+

This group is paradoxical: while bike enthusiasts are often the most vocal and supportive of bike sharing, they have a low potential to pay for or use the system since many already commute by bike. Only the most earnest supporters will buy an annual membership – more as a gesture of community support rather than for direct personal benefit. One positive benefit bike-sharing provides is against theft. With traditional bike theft is so hard to stop, bike sharing is a win for both the user and the system. Anti-theft is a good selling-point, but often not a sole reason bike-commuters would opt for bike share. Bikeshare will be an added cost over their commuter bike, with limited upside.

carpool/vanpool riders, M+

Carpool and vanpool riders often use this mode of transit to save money – particularly on the capital cost of owning an additional vehicle. This group may positively gain daytime flexibility of bike sharing when individuals might not have access to a vehicle or as an alternative to using a bus/metro. However, carpool/vanpool riders are more commonly a small demographic, and often commuting from suburban areas, which would limit their use to mainly work/weekday hours. This group of people may be willing to pay a reduced price for bike-sharing, but likely won’t have enough benefit to justify a full price.

Key demographics

single-occupancy drivers, M-

Single-occupancy drivers are numerous and potentially the most desirable customers: they use significant road and parking resources, are less fuel-efficient, and spend a great deal of money to own and operate their vehicles. However, the barriers to entry are harsh; often these drivers are often not comfortable with biking/alternate transportation, and typically have the means to pay for the convenience of car use, and thus choose not to give up the flexibility of the car.

Reducing single car use is an important goal for environmentally sustainable communities, but one that’s difficult without significant assistance from local and regional governments and shifting culture. Other strategies may involve promoting bus/carshare along with bikeshare to fully replace the convenience of owning a car. Many single-occupancy drivers have already paid the fixed cost of car, and may be reticent to invest in other options while the vehicle still has many years of life.

walking, H++

Walkers (and those particularly without bikes) are ideal candidates for bike sharing. According to research from Walkscore, the ideal range for walkers tends to be one mile or less. Using bike sharing, a walker could easily extend to longer trips. This Capital Bikeshare data analyzed by JD Antos shows that 83% of trips are under three miles, and up to seven miles – an order of magnitude increase over the typical walking range (and likely faster). Another bonus: walkers will have fewer barriers in walking to a bikeshare station.

However, walking is a thrifty activity. While some may choose to walk for enjoyment or relaxation, others may necessarily walk due to the expense of other forms of transportation. Fortunately, bike sharing is an inexpensive option for consistent commuting and likely worth the price/time savings. Bike sharing can serve as important service for those who cannot afford other options.

bus/metro regionally, H+

Regional busing provides important longer distance travel, but often has limited transit options outward from the designated bus stops. Similar to car/van sharing, regional buses often travel a significant distance, but often leave the rider still some distance from where he or she needs to go. Bike sharing can help greatly to close this last mile gap in reaching a destination.

One newer innovation in bike sharing from Bcycle allows for free sharing of annual passes between two regions – Boulder and Denver, Colorado, for example. This is an exciting development as it can close the last mile gap on both ends, while also providing greater value for using the service in either area. Connecting bike sharing programs is a huge win for residents of these cities, and adds a tremendous value to the region. Customers who can take advantage of multiple regions will also be willing to pay more for this service, raising the average revenue.

bus/metro locally, H=

Those who often ride bus/mass transit should be prime targets for bike sharing as a more flexible and functional option. Bikeshare’s main advantage is its asynchronous operation, allowing a user to check out a bike at nearly any time, while buses have specific time schedules or limited service hours. In the case of light rail/subway, the same rules apply – and these services may be a relatively expensive option for some riders, while a bikeshare membership will be only a fixed cost for the customer. Bikeshare may also provide quicker and easier access to areas that a person may need to go. And lastly, bus/metro users want to take one leg of a trip using bikeshare, and return using the bus/metro.

When adding up all the benefits of pairing bikeshare with multiple alternative transit options, the sum is much greater than its parts, making a bike sharing annual membership a welcome addition.

Bridging the price gap

My hope from this article is to provide a method of viewing potential customers in a more descriptive and useful way, using some of the incentives we know about each population. More data will be needed to make judgments on each group’s willingness to buy an annual pass from changes in price or offering targeted discounts.

When we combine this with the concepts from my previous article on optimizing usage,  the next logical step will be determining what amount of discounts may be offered, how much the average rider will use the system, and how discounts can be justified relative to the fixed and variable costs of running a bike sharing program.


Posted in Uncategorized | 5 Comments

bike sharing: optimizing system usage/revenue while keeping a great user experience

Great user experience for a bike sharing system has many components, including station location, ease of use, bike quality and availability. Specifically to bike operation and availability, two real factors apply:

  1. the start station with adequate bikes (influenced by a. total bikes in use, and b. bike distribution)
  2. the end station has adequate docks (influenced by a. total bike-to-dock ratio, and b. bike distribution)

The focus of today’s analysis will be on how bike sharing user experience changes with increases in total bikes in use (up to peak usage), while also considering the bike distribution among stations (as some stations tend to have a net-positive or net-negative flow which requires re-balancing).

Using data from Capital Bikeshare in Washington, DC (a large, urban bike sharing system with over 1500 bikes and 165 stations), we can begin to identify a case for maximum bike usage – both under an equal distribution of bikes, as well as under a more likely distribution of bikes (considering the balance under a natural station flow). The dataset used will be used from 1222 bikes (2443 docks) with 147 stations in the final two weeks of Q1 2012. The average number of bikes is determined by halving the number of total docks available, as we know that Capital Bikeshare uses a 50% bike-to-dock ratio. Based on this information, we can generate a model that simulates how many bikes the average station contains when x bikes are in use (i.e. not at any station).

Nominal operation

Here’s how the graph looks when no bikes are in use – the minimum number of docks at a station is 10, while the maximum is 39 (and will first model them at half capacity):

Next, we’ll use this as a baseline and steadily increase the number of bikes in use, decreasing bikes from stations at an equal rate. By maintaining our first goal of user experience, that the start station has adequate bikes, we must not let the peak rise beyond a level where the smallest stations no longer have bikes. For Capital Bikeshare, the smallest station is 10 docks; as it is an outlier, we’ll optimize for the many stations that have 11 docks:

Once we reach 662 bikes in use, we now see that stations with 10 or 11 bike capacity have less than 1 bike per station, on average – a total of 44 stations without bikes. Although the system still has 560 bikes available, there’s now a possibility that a new user will come to a station with no bikes available.

Station outflow/inflow

The above “average” case needs to also be viewed as the best-case scenario: bikes are evenly distributed, and new users check out bikes at an equal rate among all stations. However, bike sharing system rebalancers know better – some stations tend to always have a surplus, and some that tend to have few bikes, if any. We can model this by showing the total in/out data from the Capital Bikeshare system for the same period (knowing that any rebalancing does not reflect in this data):

As you can see, the graph of station flows are roughly 30% on each end that require rebalancing (+/-), and about 40% that mostly rebalance themselves. Interestingly, some stations have opposite trends on weekdays vs. weekends.

Typical operation

The modeling difficulty now: how can we apply this flow data to our static model of averages? My method is to take these measured flows (two week period) and divide them into a 12-hour average (which could be tweaked for more/less effect), which I then add to my baseline. Here’s the new graph, sorted by bikes available under typical use, starting at zero bikes:

Even with no bikes in use, some stations are empty or near empty. Using this new baseline, now we’ll assume the same previous number of bikes in use (our potential optimal maximum), 661:

With 661 bikes in use, we now show 40 stations without bikes – as compared to the nominal model, which only found one station without bikes. Some stations are deeply negative, which suggests even more demand for bikes (i.e. unhappy users who can’t check out a bike at the station they want).

A system operator must understand the potential trade-offs between increasing levels of ridership and the cost/ability to rebalance. This last graph helps answer that question, with its output being the number of stations that have less than one bike as the concurrent riders increase:

Using this data, a system operator can then make educated decisions for a target peak ridership (in order to maximize revenue) and what level of service customers should expect at peak system use. The cost of having peak bike usage less-than-optimal will be to forgo revenue or reduce community benefit; the cost of higher-than-optimal usage being increased or unmanageable rebalancing, or degraded user experience for the typical customer.



Read the first article in this series: bike sharing usage patterns in Washington DC’s Capital Bikeshare

Posted in Uncategorized | 1 Comment

bike sharing usage patterns in Washington DC’s Capital Bikeshare

Using data from Capital Bikeshare in Washington, DC (a large, urban bike sharing system with over 1500 bikes and 165 stations), we can analyze system usage and time distribution patterns for:

  1. All users of the system for each 24 hours of the day.
  2. By weekday, a distribution of all, casual, and annual users.

The first look – all days, all users [and subdivided between annual/casual users] uses data from 2012 – January first through March 31st:

To begin, note the peak average utilization occurs at 5PM (165 bikes – 13.6%) while minimum utilization at 5AM (5 bikes – 0.4%).

However, we can find more interesting data points when separating between registered users (an annual and monthly pass) and casual users (three-day and daily pass).

Peaks for registered users occur at 8AM and 5/6PM, coinciding with typical working hours, with sharp changes during these periods.

Casual users, more likely visitors to the city and less likely workers or students, begin a gradual rise in usage about two hours later than the registered users, showing a steady increase in the number of bikes in use until 4PM, when the number begins a significant but more gradual decline.

The next three graphs use this data and subdivide it by days of the week, first showing all users, second just registered users, and third showing just casual users:

All users

Capital Bikeshare - all users - average number of bikes per hour per day 2012 Q1This graph features the same data as the blue line from the first graph above, divided by weekday. A few notes:

  1. The heaviest average day (most total bike-hours) occurs on Saturday; the lightest on Wednesday.
  2. The afternoon peak users shifts left (earlier) gradually, starting on Monday (latest) to Sunday (earliest)
  3. Late night use (midnight to 4 AM) is significantly higher on Saturday/Sunday mornings, though still a low number of total bikes used.

Registered Users

Capital Bikeshare registered users - average number of bikes per hour per day - 2012 Q1

This graph features the same data as the red line from the first graph above, divided by weekday. A few notes:

  1. With the casual users removed from the graph, we can see a more significant/deliberate usage during the lunch hours (noon to 1PM)
  2. Weekdays have a distinct pattern (largely centered around the work week) while on weekends, registered users behavior in a manner very similarly to casual users.

Casual users

Capital Bikeshare - casual users - average number of bikes per hour per day - 2012 Q1

This graph features the same data as the green line from the first graph above, divided by weekday. A few notes:

  1. Bike sharing system usage is similarly distributed throughout the day for each day of the week.
  2. As noted above, he heaviest average day (most total bike-hours) occurs on Saturday; the lightest on Wednesday – but the difference is even larger here.
  3. Peak system usage is highest at approximately 3:30PM for casual users.

Some caveats to this data: 1. Due to the time frame involved (winter), it’s possible some data could be skewed as compared to summer months, due to changes in temperature and sunlight (increasing day use, and decreasing night use). 2. The system itself and its usage is in its infancy, and we could see a higher usage during less standard times as the system continues to mature and users check out the bikes for more non-standard trips.

Given the data’s somewhat logical divide between visitors to the city and residents, perhaps it’s most helpful to see what further questions we can ask:

  1. Can we show how changes in pricing for either casual or registered users would affect total system utilization?
  2. What percentage of registered users are using the system, on average?
  3. How can we estimate the total available market for casual users? Registered users?
  4. How could we increase revenue by getting new riders to ride during more off-peak times?
  5. How can we protect the user experience by ensuring bikes/stations are available, if we increase average system use?

I look forward to exploring these (and many more) topics relevant to bike sharing systems.


Posted in Uncategorized | 5 Comments