Last week, Strava, the company that provides tools for cyclists to track their rides via GPS, was making the news for their Global Heatmap A project of Strava Labs, the Global Heatmap compiles GPS tracking data from 100 million rides and runs by their members (about 80% are rides). Routes that Strava users travel more frequently light up brightly on the Heatmap. In this map red is highest use, then bright green, then faint green.
The Heatmap for Palo Alto is what I’d expect. The most popular routes are a pair of roads with bike lanes that offer direct, fast routes, and a bike boulevard designed with few stop signs and diverters that reduce car traffic.
Bike advocates, who often struggle to get ridership data, jumped on the Heatmap with glee. With this data, they could refute claims that people don’t bicycle on certain roads or streets. One example: El Camino in Mountain View shows green sections, meaning significant use, something city officials and the public are quick to deny.
I see that value, and yet I’m concerned about how Heatmap data is used, namely because it’s gathered from a select subset of cyclists. Strava doesn’t reveal demographics on who uses its service, but I’ve seen estimates in reports that it’s about 90% men. And I know that within my circle of cycling friends, Strava is more heavily used by those who are training to race or complete an endurance event. They’re mostly road riders and mountain bike racers who have raced at some point, if not currently training to race.
That mirrors Strava’s goal-oriented “prove it” marketing messages highlighted on their web site: “Prove It: Track your progress and challenge your friends” and “Prove Your Story”, “Prove Your Efforts”, “Prove To Others”, “Prove To Yourself.” These are all messages aimed at the Type A folks in the bike world.
I’m sure there are significant numbers of Strava users who never race (unless you count the Cat 6 racing on the trail), and others who track both their training rides and their casual rides. But I agree with a friend who wrote: “The data is skewed to longer rides done be people who take the effort to log their ride on Strava using a smart phone or GPS. A 4 block ride for shopping or to the library is not likely to be logged.”
A bigger problem is what he wrote in the next sentence: “Since [Strava users] are likely to be more experienced riders, they tend to know the easiest, safest, fastest routes.” I accept that Strava users are more experienced riders. But are they choosing routes that are easiest and safest? Or just popular with Strava users, who likely prefer straighter routes with fewer stops, even if that means it’s on the shoulder of a 50 mph expressway with heavy vehicle traffic. That’s hardly a representative set of riders to use for bike infrastructure decisions.
Without demographic data, it’s hard to counter or defend these assessments. But I got new insight from examining “hot spots” on the Heatmap where cyclists stop on their rides, based on a story by Cyclelicious.
Here’s the Town & County shopping center in Palo Alto near Stanford University. Car parking is painfully crowded so many visitors arrive by bike. Where do they go? According to the Heatmap, these two locations: a Peet’s Coffee Shop and a bike rack around the corner by Gott’s Roadhouse, a popular restaurant.
What’s missing? The mass of people who park their bikes to shop at Trader Joe’s on the other side of the center. The main rack holds a dozen bikes and has frequent turnover. It’s often full or crowded so bikes spill over to a nearby rack at Calafia. This was the situation last Sunday at around 12:30 pm: 13 bikes near Trader Joe’s, five at Peets and four at the Roadhouse. Why don’t the bikes at TJs create a hot spot?
Sunday at 12:30 pm isn’t a peak time for groceries, nor a slow time for a coffee shop. It’s clear that people who shop by bike at Trader Joe’s are not tracking their trips on Strava. That’s a lot of everyday bike trips to ignore.
So before you’re tempted to use Strava data to support bicycle policy or infrastructure changes, think carefully. If Strava data included trips by the average Joe, Jane, José or Jin-Wei biking across town to grocery shop instead of training data from cyclists tracking achievements, how would your recommendation change?
SJ Rides
April 29, 2014 at 5:27 pm
Another “goal oriented” aspect of Strava is the fact that they refer to their users as “athletes”. I enjoy looking at the data but as you point out, it’s definitely skewed toward a certain demographic, specifically those people who are able and want to record their trips. That being said, it is a pretty comprehensive dataset… and seeing it makes me want to record even more of my trips to try to “skew” the data towards the kind of riding I do.
ladyfleur
April 29, 2014 at 10:23 pm
Yeah, definitely speaks to athletes in their messaging. Would feel weird to me now too. And kudos to you for skewing the data!
Bob Fisch
April 29, 2014 at 7:45 pm
I actually disagree that Strava users are likely to be more experienced riders. You might want to think more about this before conceding this point.
I don’t use a GPS and I never even heard of Strava until this heatmap news made waves. I just ride my bike places I need to go everyday about 8 months of the year. (I walk during the snowy season.) And I probably know every street in our urban area better than anyone else who lives in my small WIsconsin town ( pop. ~ 40K).
I know all the easiest, safest, and fastest routes for bike riding in my urban area. I also know the most comfortable routes and all the alternative choices for anyone who wants to avoid any specific hazard (like a difficult crossing). In fact, I probably know these better than any of our local Strava rural bike riders who don’t feel discomfort easily and have little reason to seek out such alternatives in the city.
Not being a racer or a fast-paced rural bike rider does not equate to being inexperienced.
I’m a very experienced bicycle rider – I just have no interest in using gizmos when I ride my bike. They don’t make riding more fun. They would pointlessly distract me from all the other great things I sense when I ride. I suspect there are many daily bike riders who avoid gizmos as well. And many of these folks are experienced bike riders also.
I know this agrees with the main point of your post. I suppose that one sentence you wrote caught my attention. The only attribute I’m willing to concede about Strava users is that they are more knowledgeable about using gizmos. Anything else and you’re gonna need to show me quality data to prove it.
ladyfleur
April 29, 2014 at 10:26 pm
Oh, there are plenty of experienced riders not on Strava, including my husband and me. But I don’t think there are many very casual riders–the kind that ride 2 miles to the train station or grocery store or escorting their kids to school–that use Strava.
In others words, not all experienced riders use Strava, but Strava users are more experienced than the average Joe or Jane who rides a bike.
BC
April 30, 2014 at 11:18 am
Riding to the train station and escorting your kids to school (maybe even every day) is casual/ not experienced? Huh?
ladyfleur
April 30, 2014 at 11:40 am
Oh gosh. No offense intended. Once again, there are plenty of experienced riders who grocery shop or ride to the train station, including me. What I’m saying is that less experienced and/or less frequent riders rarely use Strava, not that people doing short rides are inexperienced.
Ken
April 29, 2014 at 8:48 pm
I believe that San Francisco (and probably other cities) has its own smartphone app that cyclists can use to send their commuting routes to city traffic engineers. Maybe you can interview city officials to see if/how their data differs from Strava.
ladyfleur
April 29, 2014 at 10:29 pm
There are other apps for people to track their rides and help planners. The downside is that they don’t offer as much to the users as Strava does. The gaming aspect keeps people so inclined coming back for more.
mtnbkr1
April 29, 2014 at 10:59 pm
Another great blog entry! I’ve been been trying to point city officials towards this great resource of data for a few years now. (I’ve been able to use the data at least on one occasion to help make the case that a certain commuting segment – the ‘missing’ link between the 237 trail and coyote creek trail – needed maintenance as a substantial numbers of folks used that particular segment). Understandably, the data isn’t complete and shouldn’t be relied on exclusively, however, in the absence of any other data on cyclists routes and activities, I’ll gladly use Strava’s data – and couple together Garmin Connects’ data – to gain insight and help drive the conversation on where improvements need to be made and refute the generic arguements that there simply aren’t many cycling commuters.
Andrew Boone
April 30, 2014 at 4:11 am
I’m guessing Strava heat maps will first be misused by planners to prioritize bicycle improvements for the most popular routes, which is exactly the opposite of what they should do.
The heat maps show which streets are already the BEST for cycling, not which streets are the WORST and therefore most in need of fixing. El Camino Real isn’t popular for cycling because in most places it’s a six-lane traffic sewer – completely horrible and awful for bikes (but fine for cars). Because it doesn’t show up on the heat maps as a popular cycling route, it isn’t clear that El Camino is actually the most important street to fix.
San Francisco, which is actually using such data (collected using their own custom app), does not seem to be making this mistake. They’ve used it “to quantify in a statistically valid way the benefit of bike infrastructure on congestion, mode share, and active travel,” which is extremely helpful since we’re still making transportation spending decisions as if there are no benefits to bicycling. http://www.sfcta.org/modeling-and-travel-forecasting/cycletracks-iphone-and-android
Yes of course Strava data is biased by racer wanna-be’s and hard-core roadie types. It’s also biased by being a smartphone app, so poor people are also excluded. But any such app will suffer from self-selection biases, including the custom app San Francisco uses. To get representative data, a city would have to randomly select residents to participate in recording their bicycle trips, just as is done for traditional transportation surveys.
ladyfleur
April 30, 2014 at 9:45 am
I dispute that Heatmap indicates which streets are already best as well. They’re just best for the sub-set of riders on Strava. I would say that if all cyclists with smartphone were polled, Central Expressway would not light up as brightly and Evelyn and Arques would heat up more. And while it’s true that people without smartphones are not included in other models, smartphone adoption it way way above Strava adoption among people who bike.
You’re right about how to get the best data, though, and it’s cities using methods they already do for other forms of transportation.
Ron BeLieu
November 29, 2015 at 12:17 pm
That is why Strava data can help also, because it can help show areas that people don’t travel and how to connect them. I constantly look at google maps for bike trails to go to different areas of our city. Strava is another tool to help with that, but as said it can show where we need improvement. It all takes, people, planning, data, and tools. Once a weakness is shown in access to travel you can act. Strava is just a tool, it give data but not necessarily complete data.
Cycling Bob
April 30, 2014 at 7:12 am
It is a given that Strava is designed for riders who are training or competitive because provides data that shows progress.
However to say that only people who like “gizmos” should use Strava is wrong. Strava has a free iPhone app (not sure about Android) that will track your ride and automatically upload to their database. You can turn it on and put it in your pocket and not look at it until the ride is over. If you want to contribute data and potentially show city planners and developers where you ride just down load the app.
ladyfleur
May 7, 2014 at 12:05 pm
I actually signed up with Strava and have been tracking since May 1. So far, I’ve tracked 24 rides in less than a week and many of those rides involved stopping and starting again. That’s four rides a day.
Needless to say, It’s been a royal pain in the rear. Most rides I take are multi-stop errand trips and I have a poor track record of starting and stopping with any accuracy. That’s in sharp contrast to the more typical weekend warrior who logs about 2-3 rides a week and doesn’t make as many stops.
I cannot see myself continuing to track much longer. There’s little upside for me and a whole lot of trouble. I’ll give it Bike Month just because and that’s it.
Andrew Boone
April 30, 2014 at 12:22 pm
Janet, just because Central Expressway appears to be more widely-used than alternative routes on Strava heat maps doesn’t mean that the maps don’t indicate which streets are most used by cyclists. It does. In many cases the Strava data backs up the (very little) real bicycle count data that has been collected. For example, in Palo Alto, the most popular “north-south” bicycle routes are Bryant Street and Park Blvd. Both the city’s bicycle count data and the Strava data show this.
Also, how do you know that bicycling is more common on Evelyn and Arques than on Central Expressway? What data do you have that shows this? We don’t know what Strava’s self-selection bias errors are because no one has yet compared this data with any other source of data.
This would be really interesting to do and it could be done for some cities, including San Francisco, where both Strava and an alternative cycling tracking app are being used to record cycling trips.
Also, smartphone use being higher than Strava use among bicyclists (if that is indeed true) doesn’t necessarily mean that the error caused by excluding people who can’t afford smartphones is less than the error caused by Strava’s bias toward bike racer wanna-be’s. The exclusion of the poor from Strava’s data might result in large errors because poor people are segregated into small geographic areas through the widespread practice of exclusionary zoning.
ladyfleur
April 30, 2014 at 12:43 pm
I can’t say how Arques and Evelyn would fare if counted. It was a hypothesis. But what I can say is that Strava users probably represent *at best* 5% of people riding bikes on any given day.
In contrast, smartphone users probably represent at least 60% of people on bikes, given that last year Nielson reported over 60% of Americans own smartphones. http://www.nielsen.com/us/en/newswire/2013/mobile-majority–u-s–smartphone-ownership-tops-60-.html
I’m not saying smartphone apps are the way to go. I’d lean toward some kind of bike counting system, either mechanical or human. I can say for sure I’d choose data derived from 60% of the user base vs the 5% any day of the week.
Andrew Boone
May 1, 2014 at 9:34 am
Menlo Park counts bicyclists and pedestrians as part of its annual traffic data collection program, the purpose of which is to make sure that car traffic congestion isn’t increasing. They’re only counting bikes and peds at intersections where they care about car traffic congestion, and they don’t actually analyze that data, but having it exist is still better than have no bike/ped traffic data at all.
ladyfleur
May 1, 2014 at 9:46 am
It’s always about the cars, isn’t it?
hoghopper
May 7, 2014 at 10:58 am
I realize this was intended to be a cautionary article directed to people who might try to do too much with incomplete information. Bravo to you for that, and perhaps Strava should publish blind demographic information about its users to help understand what the data means. However, the article’s tone seems to criticize Strava’s Heatmaps for what it isn’t, or for what others might do with it. I’m not sure that’s fair.
But on another note, a lot of people are riding a part of Skeggs I never knew existed. Must check it out!
ladyfleur
May 7, 2014 at 11:11 am
I’m not criticizing the Heatmap or Strava at all here. I am primarily cautioning everyone who is considering using Heatmap to plan bike infrastructure. And I’m criticizing the people who were quick to say: “This is great! Now we know where people are riding!” To which I’m saying, “Not so fast, you’re only getting a very select group of riders on a very select type of rides.”
I will admit that underneath it all is a strong emotional reaction that once again, the largely-male, hardcore sport-oriented riders are being allowed to speak for all of us. I won’t stand quietly by while that happens. It’s already caused way to much damage here in the US. It’s part of the reason why we have far fewer women riding here than in Europe. More on that here: https://ladyfleur.wordpress.com/2013/08/12/five-things-i-already-knew-about-women-bikes/
Ken
May 7, 2014 at 11:15 am
City planning departments are buying Strava data: http://blogs.wsj.com/digits/2014/05/07/strava-popular-with-cyclists-and-runners-wants-to-sell-its-data-to-urban-planners/
ladyfleur
May 7, 2014 at 11:21 am
Yes, Ken, I saw that. It was part of why I felt I should write this post. So far, the post has gotten over 500 views so I’m glad I wrote it.
Karl
September 16, 2014 at 2:17 pm
I agree with you, one shouldn’t use this data uncritically. And I liked your detective approach, a bit like Freakonomics!
Personally I’m hooked on Strava. It added a dimension to my riding and made me ride more. Silly? Probably. But whatever works! I also got new riding friends through Strava. I know this post wasn’t about whether Strava is good or bad, but at least my case is along your analysis of the user base. I’m male, I love to compete and I have a lot to prove.
ladyfleur
October 6, 2014 at 11:08 am
Thanks, Karl!
Sean Reilly Vienna
October 6, 2014 at 11:06 am
The data is good as a supplement. As you said, people are looking for data on actual ridership numbers. To date the efforts that have been made for bike counting have either been peak hour manual counts (a few locations on specific days) or automated counts (like Eco-Counters). Problems with these counts is that you do not get route information or “Origin and Destination points”,
With the Strava Data, it is possible to select specific time periods (i.e. peak commuting hours) and you can get specific route information for trips that are being made during those times (which like in traffic engineering can be assumed to be commuting trips).
Active transportation monitoring is still in it’s infancy, but what can be seen from the data collection techniques that have been performed in the past, is that it is going to have to be a collaborative effort of all the data to get a better picture of what the actual numbers are and the route choices these riders are using.
ladyfleur
October 6, 2014 at 11:19 am
I’m not saying that other methods of counting are better, I’m just cautioning people to be very very aware of its limitations because its user base skews it toward the habits of more competitive male riders on longer rides.
Here’s another example. For the urban bike trail I take every day, trail counts show 22% of riders were women. On Strava, only 10% on the same segment are women. That shows how women are severely underrepresented in Strava data.
Esther Lumsdon (@el_biker)
October 6, 2014 at 11:25 am
I don’t use Strava. I started with Endomondo, and switched to RideWithGPS. It would be very useful to map the routes used by low income folks who bike for transportation – it would certainly inform advocates who want to reach that population. How to map the bike routes used by people and count the riders might be an interesting set of projects for research in a variety of disciplines.
I remember reading about use of aggregate cell phone signals on routes to map motor traffic congestion. I wonder if using the moving cell phone data, but only speeds between 4-30mph might provide a rough idea of bicycle traffic? There might be a way to adjust for uphill/downhill direction (30mph uphill is a motor vehicle). And 20 people in close proximity and identical motion with frequent stops is a bus – that sort of data mangling.
Ron BeLieu
November 29, 2015 at 12:10 pm
I agree the data from Strava is geared toward riders that do distance rides. That being said it can still be useful to city planners on routes. It requires using more than just the Strava data, and using biking advocates, and manual data collection. It is a very good tool even though it may point to more affluent riders.
Cycling Bob
November 29, 2015 at 12:54 pm
Wow I want pleasantly surprised to see this thread resurrected 18 months after I originally commented.
Last week I was at a transportation meet-up at Mobility Lab in Arlington VA (check them out) where the following was quoted:
All models are wrong; some are useful. — George Box
To me that exactly sums up Strava’s usefulness to bike infrastructure planning. It provides a good but imperfect picture of where cyclists ride and should be used that way. Other data added to Strava ( or vice versa) helps to make a more complete picture leading to a better decision. Versace)