Cycling UK Forum

Si wrote:I will repeat my oft used anecdote.

In Birmingham we used to have lots of BC sky/HSBC rides - perhaps typically the sort of people likely to use strava?

Surely not!

These are for people who've never been on a bike before, from what I can see.

Strava is for competitive/keen cyclists who want to go up a 10% hill at 25mph to beat the other chap who did it at 24.9mph. It's for the most lycraed of the MAMILs, those who've spent £10k on a bike, those who do 5000 miles a month, and so on.

Hardly what HSBC ride is going for at all.

It only takes a cursory glance at the heatmap to see that it is completely unrepresentative of everyday cycling. The people who use strava tend to be competitive riders measuring their performance. Even if they also happen to be utility cyclists I very much doubt they will be recording their local trip to nip to the shops - so the heatmap won't even be representative of strava users.

To take some examples:
The lab where I work probably has 50 people riding to it every day of the year, yet is completely dark on the heatmap; the woods behind it see occasional unauthorised use by mountainbikers, who leave a trace on the heatmap.

Look at the hotspots on the heatmap - the things that stand out are forests with single track courses, military bases, mountain passes in the Lake District, Regents Park in London, but not the area surrounding Cambridge.

Zoom wider and compare northern Holland with Belgium

Pete Owens wrote:It only takes a cursory glance at the heatmap to see that it is completely unrepresentative of everyday cycling. The people who use strava tend to be competitive riders measuring their performance. Even if they also happen to be utility cyclists I very much doubt they will be recording their local trip to nip to the shops - so the heatmap won't even be representative of strava users.

I have never argued that Strava is a perfect model, why would it be, but I do argue that based upon the Glasgow demographic, users seem pretty typical of the "normal" cycling demographic 30's to mid 50's with a similar male:female profile as one might expect. I have been able to show that on the datasets available to me, roadside cycle surveys and Strava activity show a high degree of correlation, That is a fact.

There are of course places that will logically fall off these correlations. Mountain bike tracks typically show high levels of activity, urban centres may show less, but even in Glasgow city centre I found the correlation reasonably good.

It very easy to dismiss Strava as unrepresentative. If you think so, I would be pleased to see some concrete evidence rather than just "it stands to reason".

https://glasgowcycleman.wordpress.com/2 ... e-traffic/

geomannie wrote:[...] I have been able to show that on the datasets available to me, roadside cycle surveys and Strava activity show a high degree of correlation, That is a fact. [...]
It very easy to dismiss Strava as unrepresentative. If you think so, I would be pleased to see some concrete evidence rather than just "it stands to reason".

You've got one big bit of evidence: roadside cycle surveys and Strava activity show a high degree of correlation. Vorpal has mentioned above how roadside cycle surveys are fundamentally flawed back on page 1 (and I would strongly agree with that and then some), but you ignored that - Strava correlating highly with something we know is very flawed is a red flag, isn't it?

The rest of us can show point disproofs, like me with National Route 1 compared to the A148 being the opposite ways round on counters compared to the heatmap, or the heatmap suggesting that the A149 Lynn eastern bypass is a popular route, or the huge hotspots in the forests, or I've a suspicion that the latest heatmap has been distorted by a couple of sportives using another section of the A148... but obviously we can't respond like-for-like with something as detailed as yours because we've not signed terms with Strava.

geomannie wrote:I have never argued that Strava is a perfect model, why would it be, but I do argue that based upon the Glasgow demographic, users seem pretty typical of the "normal" cycling demographic 30's to mid 50's with a similar male:female profile as one might expect.

I think you're right, but what you've proved is that "in a city where no-one cycles apart from MAMILs, then a website favoured by MAMILs shows a good correlation", which doesn't seem that surprising. I haven't seen any evidence that it's a useful model for everyday cycling in cities which have a modal share greater than Glasgow's 1.6%.

Richard Fairhurst wrote:

geomannie wrote:I have never argued that Strava is a perfect model, why would it be, but I do argue that based upon the Glasgow demographic, users seem pretty typical of the "normal" cycling demographic 30's to mid 50's with a similar male:female profile as one might expect.

I think you're right, but what you've proved is that "in a city where no-one cycles apart from MAMILs, then a website favoured by MAMILs shows a good correlation", which doesn't seem that surprising. I haven't seen any evidence that it's a useful model for everyday cycling in cities which have a modal share greater than Glasgow's 1.6%.

You don't really know Glasgow do you? Its much the same cycling demographic as elsewhere. I will tell my only friend that that to my knowledge uses Strava, a 65 year old female utility cyclist, that she is a Mamil, she may care to have words with you.

Box wrote:All models are wrong but some are useful

The suggestion seems to be that Strava heat map type data is a useful model of overall cycle activity, and so far I haven't seen any evidence that would present a serious challenge to that view. I would certainly expect it to be better somewhere like Scotland where cycling has basically been marginalised to a small bunch of extremists. That said, I have seen some suggestions that it is also considered to be a useful model of overall cycle activity in The Netherlands and that would be the one place Strava ought to be a poor model of overall cycle activity because cycling is normal and virtually everybody does it.

geomannie wrote:You don't really know Glasgow do you? Its much the same cycling demographic as elsewhere. I will tell my only friend that that to my knowledge uses Strava, a 65 year old female utility cyclist, that she is a Mamil, she may care to have words with you.

I don't know it, no. But a 1.6% modal share is half the (UK) national average. Given that cycling's modal share is usually higher in cities than in towns or rural areas, Glasgow is below what you would typically expect for a city of its size. That suggests that Glasgow cyclists are much more "committed" to cycling than the average for the UK, and therefore more likely to be using Strava.

Wanlock Dod put it well above: "I would certainly expect it to be better somewhere like Scotland where cycling has basically been marginalised to a small bunch of extremists".

Wanlock Dod wrote:

Box wrote:All models are wrong but some are useful

The suggestion seems to be that Strava heat map type data is a useful model of overall cycle activity, and so far I haven't seen any evidence that would present a serious challenge to that view.

OK, if you're not willing to take our words for it, fire up https://www.strava.com/heatmap#12.06/0. ... 4/hot/ride and look at somewhere you know and ask yourself whether it reflects where people cycle. The link I've put there is centred on Newmarket and the wavy warm line from bottom left to top right is the 3-lane 70mph A11/A14 multiplex which I've never seen a cyclist on (and my eyesight is fine, thanks).

Then if you go northeast to https://www.strava.com/heatmap#11.25/0. ... 1/hot/ride you've got far more people cycling around the forest than to/from it - OK, some are probably arriving by car, but many more will be only activating Strava for the MTB ride and leaving it switched off for the transport cycling. There's a similar false hotspot at https://www.strava.com/heatmap#12.55/0. ... 6/hot/ride which is the Snetterton Motor Circuit which holds cycling days.

I could go on picking fault with the heatmaps for plenty of local towns and cities that overemphasise fast-but-low-cycle-traffic bypasses and undervalue many popular commuting routes, plus hotspots on the local club MTB, TT and weekly training loops but the above are some of the more obvious anomalies IMO.

Wanlock Dod wrote:That said, I have seen some suggestions that it is also considered to be a useful model of overall cycle activity in The Netherlands and that would be the one place Strava ought to be a poor model of overall cycle activity because cycling is normal and virtually everybody does it.

Not really because even the roadies use cycle routes in the Netherlands.

mjr wrote:

Box wrote:All models are wrong...

I'm certainly not trying to suggest that there won't be errors, but the presence of some errors does not necessarily mean that Strava data is not useful.

Wanlock Dod wrote:I'm certainly not trying to suggest that there won't be errors, but the presence of some errors does not necessarily mean that Strava data is not useful.

I'm suggesting that the errors seem sufficiently widespread that it's not a useful model of overall cycle activity. It may still be a useful model of some places or some types of cycle activity.