top of page

(1/3) Representation Strava commute cycling data, comparison with bicycle counters.



This article is a summary of the main findings; the detailed information/comparisons can be found in the PDF file above.


*THIS ARTICLE IS AUTOMATICALLY TRANSLATED, AND MIGHT INCLUDE TRANSLATION MISTAKES.


The activity tracking app ‘Strava’ was used in the Netherlands by approximately 700,000 cyclists in 2023; it is the largest activity tracker for cyclists. For most users, road cycling, mountain biking, or running are their main activities. However, once the app is installed, Strava also records many ‘commuting’ or utilitarian cycling trips. Since 2021, Strava has offered free access to the anonymized and aggregated spatial route data generated by its cycling and running users. Governments can access the data of their district under strict privacy conditions and structures. More information can be found at (https://www.track-landscapes.com/privacybeleid-1 and https://www.strava.com/legal/privacy).


In 2022, we conducted a series of Strava analyses for the province of Utrecht, showing the cycling route usage of both sportive and utilitarian cyclists (link). Strava cycling data is considered a highly valuable source of cycling data for regional/provincial cycling policies. The sportive cycling data represents a relatively clear/specific group of cyclists, namely road cyclists, gravel bikers, and mountain bikers. However, more insight into the representativeness of the utilitarian Strava cycling data is needed to better understand the overall picture of cycling usage.


In 2022, we already wrote a first series of blogs about the ‘representativeness’ of this Strava data (https://www.track-landscapes.com/blog), where we concluded as follows:

“You can conclude that the total/average Strava utilitarian cycling dataset does not provide a good representation of the total/average Dutch utilitarian cycling usage. However, based on the few initial comparisons we made, it appears that Strava data can provide a good representation of ‘long-distance,’ regional cycling traffic. It would be desirable to make more comparisons between Strava data and provincial counting points in more locations (in the Netherlands). This could provide an even better picture of who and what Strava does and does not represent.”


This article summarizes the latest insights on the representativeness of utilitarian Strava cycling data based on: (1) a comparison between the 70 permanent cycling counting points in the Province of Utrecht and Strava utilitarian cycling data,

(2) general characteristics of Strava utilitarian cycling trips and cyclists based on Strava Metro data and input from the Strava Metro team, and

(3) a comparison between the 70 permanent cycling counting points in the Province of Utrecht and Strava sportive cycling data.


Based on this, we provide advice on application and opportunities for improvement in the Strava data structures. This is a summary of the main findings; the detailed information/comparisons can be found (here, see pdf above).




1. Comparison of 70 Provincial Counting Points and Strava Metro Utilitarian Cycling Data


From the comparison between bicycle passages at the local counting points and Strava utilitarian bicycle passages at the same locations, it appears that on average approximately 1 in 200 bicycle passages is recorded as a Strava utilitarian cycling trip.

But what we particularly want to know is whether the provincial counting points with the most total bicycle passages are also the most frequently passed points in the Strava utilitarian cycling data, and vice versa. In other words, is there a (strong?) correlation between the usage rate in Strava utilitarian data and provincial counting points?

The comparison assumes that the vast majority of cycling movements in the Province of Utrecht (and generally in the Netherlands) consist of ‘utilitarian cycling trips.’ The CBS (Central Bureau of Statistics) states that approximately 96% of cycling trips have a utilitarian nature, making this comparison reasonably valid. A strong correlation would mean that Strava utilitarian cycling data strongly represents all utilitarian cycling traffic.


Result from the Comparison

The provincial counting points are divided into ‘located within the city’ and ‘located outside the city.’ Particularly outside urban areas, a clear/strong correlation is visible between the provincial counting points and Strava utilitarian cycling data (r² of 0.78 with 0 intercept). Within urban areas, the variation is significantly greater (r² of 0.57 with 0 intercept). Strava-utilitarian usage also scored relatively high outside urban areas; approximately 1 in 125 passages was recorded as a Strava-utilitarian trip, compared to within urban areas where this was about 1 in 300. The relative over-representation outside urban areas confirms the conclusion from our previous studies: Strava utilitarian cycling trips are relatively more often recorded during long cycling trips (which also more frequently occur outside the city) than during short cycling trips (which usually occur within the city).



A clarification of the characteristics of Strava-utilitarian cycling trips becomes clear by looking at which counting locations Strava-utilitarian represents a relatively high or low share of the total number of cycling passages.


Within urban areas, the representation of Strava-utilitarian is highest at counting points that logically connect to or pass through the cycling network outside the city, or on long continuous routes within the city (e.g., the counting points on bridges over the Amsterdam-Rhine Canal). The lowest representation within the city is found at all counting points around Utrecht Central Station (where mainly local cycling traffic is expected to converge). Outside urban areas, the highest representation is visible at counting points located along well-known long-distance routes, such as Utrecht-Amsterdam (Rhine Canal), Utrecht-Hilversum, Utrecht-Amersfoort(seweg). Therefore, Strava-utilitarian cycling trips represent a picture of ‘regional cycling traffic,’ but within this regional spectrum, Strava also has a relatively higher representation in cycling distances over 10 kilometers compared to the range of approximately 5-10 kilometers.




At our request, Strava Metro has now also added graphs of cycling distances to the Strava Metro dashboard. This confirms this outcome; the average cycling distance for utilitarian Strava trips is 20 kilometers.


An additional clarification of the characteristics of Strava utilitarian cycling usage emerges from the comparison with the number of passages at provincial counting points across different speed classes (every cyclist passing a counting point is divided into a speed class (0-5, 5-10, 10-15, etc., km/h)). At the counting points outside urban areas, the Strava utilitarian cycling data has the strongest correlation with local passages recorded at a speed of 25-30 kilometers per hour. At the counting points within urban areas, the strongest correlation was with the 20-25 km/h speed class.



2. Additional Characteristics of Strava Utilitarian Cycling Trips/Cyclists




Utilitarian cycling is determined by how these trips are recognized. Strava Metro has provided us with more explanation about this. First, the cyclist decides this themselves; Strava users can mark their cycling trips as 'commute'. Quite a few people do this, but not everyone takes the time to do so. Therefore, Strava has its own script that reclassifies 'sportive' cycling trips as utilitarian trips. This is mainly based on the starting and ending points of a cycling trip; if these are far apart, the trip is labeled as a 'commute'. The script also considers 'stopping behavior'. For instance, if someone cycles to a location, pauses the trip for several hours, and then cycles back (even if the endpoint is close to the starting point), the trip will also be labeled as a commute. Strava does not only consider home-work trips as 'commutes'; any cycling trip that seems to have a more functional purpose is labeled as a 'commute'. Starting a cycling trip between '7:00-9:00' is not a requirement for it to be classified as a commute.


In the 2022 Strava Metro study for the Province of Utrecht, Strava's ‘utilitarian’ cycling traffic mainly showed home-work movements; much of the usage occurs on weekdays between 7:00-9:00 and 16:00-19:00. Large employment locations are also clearly recognizable as the strongest destination areas.




In cases of high levels of home-work commuting, the use by one or a few individuals can pose a threat to representativeness. If one person records a home-work cycling trip five days a week, this could result in about 400 passages on a single route per year. However, we see that this hardly ever occurs in practice. In the detailed section ‘influence of one or a few individuals,’ this is mapped out, but it appears, for example, that there are virtually no routes where the number of unique individuals passing through is fewer than 10, and the total number of passages exceeds 400.



Between 2019 and 2023, the average cycling distance of utilitarian trips decreased; more ‘short’ cycling trips have been added. In 2019, the average utilitarian cycling distance was approximately 23 kilometers, in 2021 and 2022 this was approximately 21 kilometers, and in 2023 it was approximately 20 kilometers.





The picture of cycling speeds shows that Strava utilitarian cyclists are fairly 'fast riders': they cycled on most inner-city roads at speeds between 20 and 25 km/h, and on roads outside urban areas more often between 25 and 30 kilometers per hour (as seen in the speed maps; see pdf). Between 2018 and 2020, the median speed decreased from 24.2 to 23.2 km/h, and since 2021 it has been stable at around 22.5 km/h. This is directly related to the trend of decreasing cycling distance; a higher proportion of short cycling trips means relatively more cycling kilometers within urban areas (where the average speed is lower).


Notably, the ratio of men to women also increased significantly between 2019 and 2021, and has been stable since 2021. In terms of the number of unique individuals passing on roads, this ratio in 2023 (and also in 2021 and 2022) was 72% men - 28% women. In 2018, this ratio was 85% men - 15% women. The increase in shorter trips and the decrease in cycling speed may be related to the increase in the number of women using Strava for utilitarian cycling.



The share of e-bike utilitarian cycling kilometers within Strava is growing; in 2023, at least 10% of utilitarian cycling kilometers were covered using an e-bike. In 2022, this was still 6.8%. It is possible that some e-bike users do not mark their e-bike usage within the Strava app, although this is expected within the Strava community (because e-bikes are not desired in segment rankings). But even if, for example, half of e-bike users did not indicate this, the vast majority of cycling trips are still done under one's own power. The above two points show that Strava utilitarian cyclists are largely road cyclists or other 'active/sporty' people (such as runners or walkers) who also see their utilitarian cycling trips as a way to stay in good shape.


3. (Strava) sportive cycling traffic accounts for a significant portion of cycling movements at some counting points.


The comparison between the local cycling counting points and the Strava sportive cycling data shows that outside urban areas, the representation of road cyclists is significant on a considerable portion of cycling paths. Strava sportive cyclists represent more than 10% of total cycling traffic at 18 of the 47 out-of-town counting points, and at six points, the percentage is even higher than 20%, reaching a maximum of 35% along the N201 (Vinkeveen). This only concerns Strava-using road cyclists; the total share of road cyclists is likely even higher. It is reasonable to estimate that on some cycling paths, road cyclists make up at least half of the cycling traffic.




This also leads to another important realization: bicycle counters that do not measure carbon bikes (like those in the Province of Utrecht) will miss a significant portion of bicycle passages at certain points.

The previous assumption that the vast majority of regional cycling movements consist of utilitarian cycling traffic is therefore not reasonable at all counting points. Moreover, there is still a lack of visibility on road cyclists who do not use Strava, as well as non-sportive recreational cyclists (who rarely use Strava).




Application of Strava Cycling Data in Cycling Policy

The outcome that Strava utilitarian cycling data is ‘reasonably representative’ for ‘regional/longer’ cycling movements supports the idea that Strava utilitarian cycling data can be used for tasks in regional cycling networks. It can provide an overview of significant/coarse differences in regional cycling usage. For many applications, an initial look at general usage patterns is already very valuable.


However, this should be done with the awareness that Strava utilitarian cycling traffic more strongly represents the routes of longer regional cycling movements (>15km) than shorter regional cycling movements (5-15km). That representation is shifting a bit more towards the ‘shorter’ side each year.


Incidentally, Strava could easily address this by adding a few columns with distance categories to the heatmap datasets, allowing for the mapping of relative differences in cycling route usage between cycling trips of different distances. This is where the greatest potential of Strava utilitarian cycling data lies.


Strava utilitarian cycling data cannot simply be used as ‘indisputable’ numerical evidence on which general cycling policies/infrastructure interventions can be fully based (this applies to any form of data, by the way).

However, cycling data can also be valuable precisely because it is not representative (of average cycling). For some questions/tasks, you specifically want to focus on particular target groups, and Strava utilitarian cycling data offers the opportunity to examine the movement patterns of specific cycling groups in more detail. Which routes form the most important ‘utilitarian long-distance routes’ in the Province? And on which routes is the share of road cyclists substantial? These are highly relevant questions for regional cycling policy, where Strava provides good insight.


Thus, the value of Strava utilitarian cycling data may also lie within urban areas, precisely because it was found to be ‘least representative’ there. Strava mainly shows which routes longer cycling movements take in and out of the city, and these are completely different movement patterns than local urban cycling movements. In such cases, Strava utilitarian cycling data provides a valuable specific perspective that other data does not capture.


Even when (knowledge about) representativeness is not perfect or complete, insight into usage can raise good questions about certain cycling routes and uses. The data then primarily serves as a means of reflection and discussion. From clear maps of cycling usage, many realizations, insights, follow-up questions, and suggestions arise, which in themselves are already very valuable. Other data sources, local measurements, observations, and inventories, together with Strava data, can then provide a more complete basis for decisions on cycling investments.

Comments


bottom of page