Cycling Clydeside: The Glasgow Bike Share Scheme
March 09, 2020 by Philip Lowney

Overview

In this post, I study the performance of the Glasgow bike sharing scheme based on data collected for three months in 2019/2020. We will:

  • Attempt to visualise the behaviour of the scheme through animation
  • Present daily profiles of the scheme in terms of key metrics
  • Identify the daily geographic flow trends of the scheme
  • Identify which stations which are performing best and worst with respect to availability of bikes

I hope users of the scheme, authorities responsible, and advocacy groups will use this data to better understand the provision of bike sharing in Glasgow and inform discussion around future developments. Thanks is due to Nextbike, the operators of the scheme, for providing an API which we query to gather the data herein.

Most of the data presented here is freely available via the data-browsing console I have built here on SchemeStats.bike. Please feel free to explore the data for yourselves using the Dashboard, Live Map, Day Profiles and Metrics Map – and if you like what you see, please share it.

Contents

Introduction

The Glasgow Bike Share Scheme was introduced in 2014, and now consists of approximately 490 bikes distributed across 68 stations. Operated by Nextbike, the bicycle fleet is composed of both standard bikes and e-bikes which assist riders through a battery-powered electric motor. All bikes come with an on-board interface with a screen and keypad through which users release and return bikes. The scheme is station based, with docks composed of metal tubing to which the bike is secured by riders. Subject to an additional fee, bikes may left locked outside of a dock for a short time.

Data Gathering

Time Period

The data presented herein was gathered over a three month period from December 1st 2020 to February 29th 2020. The API provided by Nextbike was polled at one minute intervals.

Bikes in Transit

Figures provided for bikes in transit at any one time are estimated values calculated by SchemeStats.bike and are based on the number of bikes in stands/undocked at any point in time, versus the maximum figure for the same value over the preceding 24 hours. This yields an approximate result only: the figure for bikes in circulation necessarily includes those taken out of circulation in the previous 24 hour period, and those out for redistribution on the back of a truck etc.

Limitations

SchemeStats does not collect data on the type of bike in stations, and for this reason no breakdown is provided on e-bikes versus standard bikes. This will be revisited in future.

Animating Glasgow

What follows is a time-lapse animation of the Dashboard tool over a one-week period from February 10th to 16th 2020. The video is composed of frames separated by 5 minutes of elapsed time. There are several panels captured, each of which expresses a different view on the data:

  • ‘Live Bike Status’ Illustrates the estimated division of the bikes into three states – docked, in use, and undocked (i.e. not in use but not docked in a station).
  • ‘Live Station Status’ divides stations up by those which are open, closed, full, and empty. Closed stations do not count towards the other three values. Full stations aren’t a big deal for the Glasgow scheme, as stations are allowed to ‘overflow’ their capacity.
  • The ‘Last 30 Minutes’ panel is a map of the net change over the preceding half hour, with red stations gaining bikes and blue stations losing them.
  • The ‘Station Population’ is a histogram showing the distribution of station populations, with bars on the left representing emptier stations. A wellbalanced scheme has bars clustered towards the centre of the histogram. More area on the left of the graph indicates lots of stations with few or no bikes, and more area under the right of the graph indicates full stations.
  • ‘Utilisation History’ and ‘Bikes History’ are related graphs, with the former plotting the change in the number of bikes in circulation as a percentage of the total stock of bikes, and the latter recording the raw value. In this view, the last six hours are rendered in a sliding window.
  • ‘Station History’ plots a sixhour sliding window of the number of empty/full stations.
  • The ‘Live Map’ panel shows the population of each station, with blue for bikes and white for space in stations. Smaller green circles indicate undocked bikes which are available for hire.

The video is not a systematic analysis by any means, but we can draw some observations suggestive for further analysis:

  • There seems to be a reliable morning increase in activity during weekdays, suggesting commuting activity.
  • Very few stations become empty throughout the week.
  • Morning and evening times appear to show clusters of areas emptying out and filling up each day, further suggesting commuting activity.
  • Undocked bikes don’t see that much mobility, often staying in the same place for a long time. It is unlikely these are out on hire, as undocked bikes incur a cost to users. These bikes may be undocked due to a mechanical issue or flat tyre.

Day Profiles

A ‘Day Profile’ is a plot of the average values of a given metric on a given day of the week – e.g. ‘Available Bikes on Tuesdays’. I’m also overlaying the trend with the standard deviation – a measure of the spread of the data - to get an idea of how noisy the signal is. In the charts below, the solid lines are averages, and the shadow on either side of them is the standard deviation range.

Bikes in Circulation

This estimated figure yields the following day profiles for 2019.

Day Profiles - Bikes in Circulation

Empty Station Profiles

We can also form a picture of the measures of the number of stations which are empty per day:

Day Profiles - Empty Stations

Undocked Bike Profiles

Finally, here is the picture of the average number of bikes parked outside a docking station:

Day Profiles - Undocked Bikes

Analysis

The trend for bikes in circulation shows a similar shape to that which we observed in Dublin previously: Weekdays show a peak in the morning and evening, implying commuting activity, and weekends show a more gradual pattern with less use. Weekday mornings see a peak of around 30 bikes in circulation on average, representing a fleet utilisation of approximately 6% at peak times.

The scheme does not appear to be under a lot of strain – most of the time between 3 and 7 stations are empty, and these figures do not appear strongly correlated to the commutes. Users stand an excellent chance of finding bikes at stations throughout the day.

Figures for undocked correlate with what can be seen in the time lapse above – they make up a small proportion of bikes available.

Weekends are typical of other schemes insofar as they are quieter than weekdays, however Sundays are markedly busier than Saturdays. Perhaps this is due to leisure use. Sunday peak usage reaches an average of about 20 bikes and stays steady from around 10AM.

Tuesdays see the strongest evening usage – I can’t explain this, but conversely Friday evenings see the lowest, shortest duration period of increased use. We saw this earlier in Dublin too, and it makes intuitive sense: Friday evenings are different for people socially, they stay out after work, or they head out of the city for the weekend, not using the same modes of transport as the rest of the week.

Mondays see the greatest number of empty stations, with numbers rising from 8AM. This can likely be ascribed to the pattern of redistribution of bikes over the weekend: Bikes are left back in commuter's places of residence/origin in the evenings during the week, and get more randomly distributed during the weekend. This means that on Monday mornings there are fewer bikes available in stations from which users are departing. This was also the case when Dublin was examined.


A Note from the Author

I really hope you find the tools and blog posts provided here useful. This website is a hobby I undertake in my spare time. If you have enjoyed using this site, please consider sharing it on a platform of your choice. Knowing this project is useful to people spurs me on to provide further functionality and content. If you have any recommendations, please don’t hesitate to get in touch via the feedback link above.

Thanks,

Philip


Weekday Flows

Using SchemeStats’ Metrics Map tool, it’s possible to plot the geography of a given station measure over a given time period. That time period can either be a simple from- and to- value, or a slice of a day. We can plot things like ‘The average net change of all stations between the hours of 7AM and 10AM, Monday to Friday’. This is what I’ve done below for two times of the day during the week when the scheme may be used by commuters as we see elsewhere. If we see the same stations filling up in the evenings that also empty in the mornings, it is likely people are using them to get to and from their place of work or study. Based on the day profile data for bikes in circulation, I’m using 7AM-10AM as the morning period, and 4PM-7AM for the evening period. These time periods correlate with the two peaks we see for weekdays on the graph.

Morning Period

Net Change per Station, 7AM to 10AM, Monday to Friday:

Net Flow per Station, Weekday Mornings

Average values across all stations:

Top 5 Stations Gaining Bikes - Weekday Mornings 7AM-10AM

Station

Mean Change 

Min

Max

Std. Deviation

Glasgow Science Centre 8.48 0 17 4.75
ELECTRIC - Broomielaw 7.88 -1 23 5.39
Central Station 6.74 0 16 4.08
Argyle Street Railway Station 6.43 -4 14 3.51
Waterloo Street 5.91 0 13 3.14

Top 5 Stations Losing Bikes - Weekday Mornings 7AM-10AM

Station

Mean Change 

Min

Max

Std. Deviation

City of Glasgow College (Riverside Campus) -7.94 -16 0 4.41
Glasgow Green -6.91 -16 0 3.54
Partick Interchange -4.54 -12 0 3.25
ELECTRIC - Battlefield -4.45 -15 0 3.25
ELECTRIC - Queens Park Railway Station -4.25 -12 0 3.54

Evening Period

Net Change per Station, 6PM to 7PM, Monday to Friday:

Net Flow per Station - Weekday Evenings

Average values across all stations:

Top 5 Stations Gaining Bikes - Weekday Evenings 4PM-7PM

Station

Mean Change 

Min

Max

Std. Deviation

City of Glasgow College (Riverside Campus) 6.55 -1 16 4.09
Glasgow Green 4.43 -2 13 3.15
ELECTRIC - Langside Hall 3.31 -2 16 3.15
Partick Interchange 2.88 -17 9 4.64
Paisley Road Toll 2.58 -3 8 2.39

Top 5 Stations Losing Bikes - Weekday Evenings 4PM-7PM

Station

Mean Change 

Min

Max

Std. Deviation

Glasgow Science Centre -8.2 -22 0 4.59
ELECTRIC - Broomielaw -6.28 -18 3 5.08
Waterloo Street -5.12 -19 3 3.98
Argyle Street Railway Station -4.83 -18 2 4.31
Central Station -3.94 -23 2 4.12

Analysis

The maps illustrate clearly that the Glasgow bike sharing scheme is being used to get people to work in the mornings and back home in the evenings. With that being said, the tabulated data above indicate a weak flow, as net change figures for individual stations are not that large mostly – e.g. maximum/minimum net change of +6.5 to -8.2 bikes in stations in the evenings, with the majority of stations showing smaller changes. Further, standard deviations are relatively large, suggesting the trend is only somewhat regular. We do see the same station names pop up in both the mornings as the biggest gainers and evenings as losers of stations, indicating clearly these are targets of commuting activity.

Station Performance

One important measure for evaluating the performance of a given station is how much of the time it spends empty. This value is most relevant during commuting periods, when the scheme is most in demand. Therefore, I have chosen to plot this measure for the two commuting periods.

Weekday Mornings

Here is a map of the percentage of time each station spends empty on average, from 7AM to 10AM, Monday-Friday:

Percentage of Time Empty - Weekday Mornings

Note that the Citizen’s Theatre station is still appearing – I’ve been meaning to remove this as it is no longer in commission, but please ignore it henceforth.

Plotting the actual dataset yields the following:

Time Spent Empty - Weekday Mornings

The top 5 stations most likely to be empty are as follows:

Top 5 Stations by % of Time Empty, Weekday Mornings 7AM-10AM

Station

Mean % Time Empty

Min

Max

Std. Deviation

University of Strathclyde (North) 36.15 0 100 40.86
Sauchiehall Street 24.85 0 100 36.56
ELECTRIC - Cambridge Street 24.05 0 100 36.7
Barrowlands 19.94 0 100 35.71
Mount Florida Railway Station 19.8 0 100 33.77
Springfield Quay 2 15.78 0 100 36.33

Weekday Evenings

Here is a map of the percentage of time each station spends empty on average, from 4PM to 7PM, Monday-Friday:

Percentage of Time Empty - Weekday Evenings

Let’s plot these values from highest to lowest:

Time Spent Empty, Weekday Evenings

The top 5 stations most likely to be empty in the evenings are as follows:

Top 5 Stations by % of Time Empty, Weekday Evenings 4PM-7PM

Station

Mean % Time Empty

Min

Max

Std. Deviation

University of Strathclyde (North) 25.23 0 100 38.57
Barrowlands 22.88 0 100 36.64
University of Strathclyde (South) 18.49 0 100 30.52
Duke Street Railway Station 15.48 0 100 32.85
ELECTRIC - Cambridge Street 15.16 0 100 29.76

Station Performance Analysis

Performance by amount of time spent empty is good – the majority of stations have bikes available the majority of the time. Specifically, during the morning commute all but three stations are empty more than a fifth of the time, and in the evenings only two stations are empty more than a fifth of the time. There is a correlation here with the stations experiencing the strongest net flow during the commuting periods, indicating a greater effort at populating these stations with bikes in the mornings/afternoons is warranted in these areas.

Summary

  • The Glasgow bike sharing scheme experiences a weekday trend of increased activity at two points Monday to Friday, with the morning peak of estimated bikes in transit at approximately 9AM and the evening peak occurring around 5:30PM.
  • We see a trend of certain areas draining of bikes and filling with them each weekday morning, and the converse occurring each evening. Combined with the daily doublepeak of activity, this strongly indicates the scheme is used for commuting to places of work and study.
  • Peak usage sees a utilisation of approximately 6% of the bikes during the morning commute. A much greater percentage of bikes will be used over the course of the day however, as this is just a snapshot of how many bikes are in use at one time.
  • Weekend activity is more subdued, although Sundays are markedly busier than Saturdays, with traffic picking up from 10AM onwards, staying steady with an average of 20 bikes in circulation throughout the afternoon, before dropping off again in the evenings.
  • Bike availability is typically good, with most stations having bikes available most of the time during the morning and evening commuting periods. Combined with the figures for utilisation and net flow, there does appear to be additional headroom in the scheme to allow for greater usage with the current infrastructure.