Why Divvy Needs to Densify as It Expands

Divvy is shown as a “low performance” (red) bike-share system in this chart from the Institute for Transportation and Development Policy. While it is a young system and still growing, Divvy needs to increase its station density.

For a very new American bike-share system, Divvy is doing well, but it has a lot of room to improve, according to the Institute for Transportation and Development Policy’s new Bike Share Planning Guide [PDF]. The guide includes best practices for designing, distributing, and marketing this new form of transit. While Divvy is still growing and hasn’t gone through a full peak season yet, the guide illustrates one area where Chicago should focus on improving its bike-share system: station density.

ITDP evaluated Divvy and more than two dozen other bike-share systems according to two basic metrics: daily trips per bike (a.k.a. “system efficiency”) and trips per capita (or “market penetration”). The guide says between four and eight trips per bike is a good range. Fewer than four daily trips per bike, and there’s too much slack in the system. More than eight, and the availability of bikes and docks will suffer. (This is a problem New York City has seen, even outside rush hour.)

In Divvy’s peak 30-day period, the system saw just over three daily trips per bike, below ITDP’s recommended range. Sean Wiedel, assistant commissioner at the Chicago Department of Transportation, pointed out that Divvy continued to expand through the summer and fall. Because ITDP’s guide was written while Divvy was still rolling out stations, he said, it gives an incomplete picture of system performance. Looking at American cities that launched bike-share networks a year or more before Chicago, he said, Divvy “is on par with systems that have been in operation considerably longer.”

Divvy will probably gain steam in the year ahead. To reach the level of use of the most successful systems, though, planners will have to increase station density as the system expands.

Colin Hughes, one of the guide’s authors and ITDP’s policy director in D.C., clarified that the 30-day period they analyzed in Chicago occurred before all current Divvy stations were online. He then recommended that to improve Divvy’s “mobility performance” planners should focus first on “increasing station density significantly within the coverage zone.” He added that Divvy could improve performance by also increasing the number of bikes or by having more stations with fewer docks.

The ITDP guide recommends that stations should be spaced about 300 meters, or 984 feet, apart from each other. This is the same station density that Chicago’s bike-share RFP sought for “most of the implementation zone.” But most Divvy stations — especially ones outside downtown — don’t meet this standard. Using data from Alex Soble’s Divvy Brags application, I found that only 34 out of 300 Divvy stations have at least one station within a 300 meter bike trip.

The RFP said stations in outlying areas “or areas of lower demand” should have other stations located between 300 and 500 meters apart. Of all the current Divvy stations, 183 — or 61 percent — are within 500 meters of another station.

Divvy stations are densest in downtown, West Town/River West, West Loop, and Cabrini Green. There are especially high concentrations of Divvys near Union Station and Ogilvie Transportation Center, reflecting the importance of using bike-share extend the reach of transit.

Stations near Washington Park have the lowest connectivity, with the station at Prairie Avenue & Garfield Boulevard a full 1.2 miles from the nearest other station. Additionally, the stations in Washington Park and Hyde Park have 22-37 stations within 5 miles (what you can bike in 30 minutes at 10 mph) while downtown stations have over 200 within 30 minutes travel time. Granted, the number of destinations in Washington Park is different than in the neighborhoods with a greater concentration of stations, but the population density doesn’t change much, and bike-share users need to be able to “bike and park anywhere…easily and conveniently,” ITDP’s guide notes.

Riding Divvy downtown, where station density exceeds 100 stations within the 30-minute limit. Photo: Kyle Wiberg

Divvy will grow by 175 stations this year, adding 1,750 bicycles. Stations will be added to expand the coverage area, but also to add more stations in the existing coverage area. Wiedel says that infill stations will comprise approximately 20 percent of the system expansion. The coverage area, Wiedel described, would expand to Howard Street in the north, 79th Street in the south, and Kedzie Avenue on the west, and reach closer to the lakefront as you move north and south.

In addition to distance between stations, bike-share systems should be planned to have “a more or less uniform station density throughout the coverage area,” according to the ITDP guide. Divvy’s station density varies greatly, and Hughes said the system has “half of the minimum station density that ITDP would recommend.”

When I tweeted that the station at Prairie/Garfield was the “most disconnected” one, 3rd Ward Alderman Pat Dowell agreed, tweeting back, “You’re right. A new station will be at 51st between Prairie and King (tbd) this year. The Garfield CTA station is a transit hub.” However, the 3rd Ward, with nine stations, has a density of less than one per square kilometer, and even with a 20 percent increase would remain below ITDP’s minimum recommendation.

With 175 more stations set to go online in 2014, and possibly 75 more after that (for a total of 550), between 35 and 50 bike-share stations should be added to the existing zone, improving connectivity for many neighborhoods. However, according to Hughes, Divvy needs to double or triple station density from the current four stations per square kilometer to reach the top-tier performance of the systems in New York City, Montréal, or Mexico City. Even 50 stations to fill out the existing system may not be enough.

  • BlueFairlane

    I agree that increasing station density is essential, but I don’t understand why increasing the number of bikes would help. Wouldn’t this just decrease the number of trips per bike?

  • Jakub Muszynski

    I’m going to speculate there are at least two reasons for the increase of both stations and bikes; the first reason being to able to handle future demand, the second being contractual.

    Either way, as a user I care less hitting a recommended range by manipulating the total station count while not increasing the total number of bikes, and more about have addition station options while using the system.

  • Fedor Manin

    > 2.5 miles (what you can bike in 30 minutes at 12 mph)

    Wait, what? I agree that 2.5 miles is a good benchmark, but at 12 mph it takes 12.5 minutes.

  • Fedor Manin

    Adding more stations while keeping the number of bikes constant would make the system more useful because it would make more destinations reachable. If your destination is more than a block or so from a station there’s not really a point in Divvying. That said, I’m not sure how the cost of planning and building a station depends on the number of bikes at the station — I’m sure it’s not linearly. We’ll just have to wait.

  • Aaron M. Renn

    What is the “capita” in this study? Did they capture the true addressable market?

  • Jeff Wegerson

    I wondered that as well. From reading the pdf I paraphrase (badly):

    Capita is calculated by extending the boundary created from a border formed by the outermost stations out 500 meters more. Then they use the average density of the creating entity times the area calculated. They understand that it likely understates the actual “capita” as stations tend to be placed in the denser areas.

  • Jeff Wegerson

    I know we are dealing with a sophisticated audience here but as a reminder, 200 meters is about a Chicago block (660 feet).

    I hope you, Steven, are working on a version of the comparison graphic of the 19 bike sharing systems station density maps to the same scale that will include Chicago’s. Here is an overlaid Paris metro on Chicago’s CTA el lines. The comparisons are interesting.

  • BlueFairlane

    You may not care whether usage of an individual Divvy bike falls within a prescribed range, but somebody cares, or else there wouldn’t be a statistic for it, and the guide Steven’s citing wouldn’t have a preferred range.

    I don’t know why it matters (Steven could likely explain it), but I have theories. For one thing, efficiency demands a balance between the cost of bike maintenance with the miles ridden. I know these are supposed to be superbikes, but simply sitting out in the weather will wear on them. If nothing else, tires rot. If you have an excess number of bikes sitting out and not being used, then you’re building unnecessary maintenance costs which ultimately harm the system and make it less sustainable.

    For another thing, an excess number of bikes means they take up an excess number of docks. The whole point of the system is to encourage turnover, but that’s hard to do when you have dock space occupied by a bike that might only move once a day. Personally, I’ve had trouble finding an available dock downtown every time I’ve used Divvy. Increasing the number of docks would help this. Increasing the number of bikes at the same time would only maintain the status quo.

  • Colin Hughes

    The most critical thing Chicago needs to do to drive up performance is increase station density so bikes are more covenantally available. ITDP’s analysis of 20+ systems globally shows that no system with less than 8 stations per km2 has high performance (see Fig. 3 of the Bike Share Planning Guide). Chicago only has 4 stations/km2, while NYC has 10, Paris has 12, and Mexico City has 15.

    But Chicago should also add more bikes because that will drive up the number of uses per capita (market penetration) because more people take more trips since more bikes are available. Further, as Divvy adds more stations, they will still want to retain a bike-to-dock ratio around 2 or 2.2, so more docks means more bikes (and/or using smaller stations). Adding more bikes does mean the existing users would have less trips per bike, research shows that improved bike availability is likely to attract more users and usage as well improving market penetration, and canceling out the negative impact on trips per bike.

  • BlueFairlane

    This ignores the information in the article above:

    The guide says between four and eight trips per bike is a good range. Fewer than four daily trips per bike, and there’s too much slack in the system. … In Divvy’s peak 30-day period, the system saw just over three daily trips per bike, below ITDP’s recommended range.

    There is no shortage of bikes available in the system. That’s one of the two things measured in this evaluation. In personal experience, there are always bikes waiting at virtually any station, even in mid-morning when the downtown migration has passed. I’ve never seen a station without an available bike. I’ve rarely seen a station with fewer than two.

    Edit: I just noticed the name. Okay … you wrote the guide. Why is “system efficiency” a rubric when you choose to then ignore it and call for more bikes?

  • Colin Hughes

    Increasing station density and the number of bikes is not needed so much to respond to an expected increase of demand in the future, it is necessary to improve the system in order to induce or catalyze demand in the future. To optimize the system so its performance is better and it attracts more new users and facilitates their travel more efficiently.

    Station density primarily impacts how many origins and destinations the system can access. Low density systems have stations that are close to fewer origins and destinations, thus they are convenient to fewer trips and get fewer uses. Further, low station density is often associated with more severe problems with full/empty stations because peak demand is focused on fewer stations and secondary stations are further away for users. Low-density makes bike-sharing less convenient (spatially) and less reliable (in terms of availability at peak periods). Station density of 8 stations per km or higher is associated with higher performance
    (see fig 3, ITDP guide).

    The number of bikes per capita relates to availability. When a system does not have enough bikes, it’s trips per bike indicator may be high, but it will be serving a very small
    portion of the population. The system will suffer from empty/full stations at peak periods and low reliability (there isn’t a bike when you need one). Adding more bikes to a system has been shown to drive up market penetration (Fig 4, ITDP guide).

    The ranges for station density and bikes per capita are based on empirical analysis of 20+ systems globally. If Chicago wants to improve performance, the data makes it pretty clear these are two important means of doing so to consider.

  • WestLooper

    Interesting overlay but shouldn’t’ the city centers overlap?

  • WestLooper

    Partly it must be that under-used bikes suggests they are mis-deployed and you could be serving a denser or broader area with the same number of bikes.

  • Colin Hughes

    I’ve been pretty clear that in Chicago I think the primary issue is low station density, not the low
    number of bikes. Nonetheless, if Chicago expands its number of stations it will likely mean they have expand the number of docks by almost an equal factor since Bixi doesn’t make stations smaller than 10-11 docks and Divvy’s current average station size is 14. So then the question is do they let those docks sit empty with a system that has a crazy dock-to-bike ratio of 4 (double anywhere else in the world)? They could, but I wouldn’t advise it. ITDP’s data from 20+ systems globally is clear that adding more bikes catalyzes more trips (see Fig 4 of ITDP Guide). Chicago currently just meets the bare minimum with 10 bikes/capita – not bad, but not great. I think the marginal benefit in terms of trips/per capita from adding more bikes to the system (once there is a decent station density and a consistent or higher dock-bike ratio) would easily exceed their marginal cost.

    And remember, just because it seems easy to find a bike at 11 am or 8 pm in January, doesn’t mean a bike is easy to find at 8 am in May when far more people want to ride one to work in the nice weather. These systems have to be balanced so the cost,
    supply, and distribution strategies are optimized for both PEAK and off-peak demand.

  • J

    Jeff Wegerson — Do you have, or know where one can find, identically scaled maps of the rail lines AND stations (as opposed to just the rail lines) of various cities’ rail mass rapid transit systems — i.e. subway systems. I would like to see Paris, Berlin, London, Moscow, Tokyo, Barcelona, and other big city systems in comparison to Chicago’s (and each others’).

  • BlueFairlane

    Filling docks seems a silly concern, when it seems like it would save money on maintenance to not have unused bikes sitting around in bad weather … snow in summer, rain in winter, etc. Remember, the numbers listed above for a bike’s usage was below your guide’s target when measured during PEAK. It’s probably much less now. So I have to ask again, why does this “system efficiency” rubric exist?

  • Perhaps my next article should look at “how to get people to use the darn thing”, which the guide covers.

    Chicago’s usage per bike is relatively low. We can use this comparison to understand that perhaps our system is not as good (well-distributed) as we thought it was and then start moving stations around (which is an expensive maneuver).

    Gabe was quoted several times in news media saying something along the lines of “Divvy usage is meeting our expectations” without actually revealing the expectations. This guide provides *recommended* expectations where CDOT provided none.

    Example: ‘While on the one hand calling the public response to the Divvy program “beyond expectations,” city officials have set a high bar for ultimate success.’

    So our usage per bike is relatively low. Did we put stations in the right spot? Do enough and the right people know about Divvy? Shoot, are there enough bike lanes going to the right places? (That’s something no amount of marketing can resolve.)

  • This is correct. Here’s a KML file of the coverage area that ITDP’s analysis used.


  • I can provide you the data for Chicago, but I don’t know how to create a to-scale graphic. For the curious, the original graphic is on Quartz.

    Neil Freeman created a subway graphic that inspire the bike-share graphic David Yanofsky created and placed in ITDP’s guide.

  • Divvy is only a few months old! Even though it launched end of June, they didn’t get all the initial 300 stations in until beginning of October, which is past the time of year that most people currently think of as viable “bicycling season”. So, we definitely have to let a little more time pass and let the system develop before worrying about the usage per bike being low. I know from experience that Divvy isn’t that useful unless there are stations 1/ close to your home and 2/ close to where you need to go. Both have to be in place, and for many people, they weren’t in place until late in the season (and many are still waiting, of course).

    Another factor I think will make an enormous difference will be when protected east-west bike lanes are installed in the Loop this year in conjunction with Loop BRT. I know many of my coworkers are planning to join Divvy just to ride to/from the train stations to our office on Michigan Avenue, just as soon as they don’t have to dodge buses and taxis for the one-mile stretch.

  • There are differences in our systems, of course, but the whole exercise of building bicycle infrastructure, designing bike networks, increasing ridership, etc. is intimately based on comparisons and experiences elsewhere.

    Citibike is a month older and I think that Chicago should be closer to the usage path New York City’s Citibike has shown than D.C.

  • Rebecca

    Important note about Chicago’s system is that, unlike NYC, our full 300 stations were not installed until several months later. NYC’s stations activated all at once. I think 2014 will give us a clearer picture.

  • jeff wegerson

    I created the map for a post about why Chicago needs BRT. I wanted to illustrate how many more rapid routes we would need in Chicago to begin to get a comparable level of rapid service Paris has.

    Basically my conclusion was that on the north side we might want all of the four lane arterials to have BRT. So yes Ashland and Western, but also Peterson and Irving Park etc.

  • jeff wegerson

    Thanks for the links. From the Oliver O’Brien’s interactive maps, ( http://bikes.oobrien.com/ ) link I did this rough start of a Chicago/Paris comparison. That Paris looks denser is partly dot size differences and rescaling issues. Remember that the Chicago map is just from Edgewater on the north to Hyde Park on the south and however west Divvy goes.

  • jeff wegerson

    Steven provided all that I had: http://fakeisthenewreal.org/subway/

  • I agree we should compare with other systems, of course. I’m just saying that New York’s system was activated all at once, as Rebecca points out below, whereas ours was rolled out slowly, and then fall/winter arrived. That makes a difference in the number of users, both daily and even more so annual memberships. So I think we need to give Divvy a little more time before making more of an apples to apples comparison. I agree with Rebecca that 2014 will give a clearer picture.

    On a related note, I’m looking forward to Divvy releasing data more regularly at some point, so that you and all your hacking friends can do cool things with it, especially colorful visuals and of course fun and useful apps.

  • Thanks for the reminder, Rebecca. I believe that difference in our rollout makes a difference in activity levels.

    I can’t help but look at their stats during January winter, though, and see 9,000+ trips yesterday!

  • There will be a special Divvy-focused Open Government Hack Night on Tuesday, February 11, where locals will be presenting their apps and visualizations using only the station data.

    Divvy and CDOT will be there, too, to introduce newbies to the transit system and talk about data available and analysis/app needs. I’ve invited them to release the trip/user data like Alta has in Washington, D.C. so we can make those colorful visuals!

  • anonymous

    Data question about DivvyBrags: Isn’t that a chrome extension that tracks an individual’s mileage on Divvy? I’m not sure how you used that to find station-to-station proximity.

  • Pingback: Divvy Surveying Members About Different Pricing Options | Streetsblog Chicago()

  • Pingback: As Divvy Grows, Station Placement Should Work for Pedestrians | Streetsblog Chicago()