Changes coming to Strava

[Alistair]

Originally Posted by Philster

Maybe Strava woke up to the potential for privacy leaks from these other apps. Maybe they see the potential for profit leaks also ...

My vote is for the latter. Strava was a convenient way for another service to skip direct integrations with device services. Just build the integration with Strava and save money. I'm surprised the APIs remained open as long as they have.

TrainerRoad is/was a good example - they don't have direct integration to pull data from Wahoo's ecosystem - instead relying on Strava to middle-man those rides back to TR. By the new Strava ToS, they'll need to stop that and build their own integration to Wahoo...

https://support.trainerroad.com/hc/e...to-trainerroad

[glepore]

I don't disagree with their decision. They had become a free data transfer service. l synch all of my data to Dropbox so that I don't need to rely on Strava for synch to Golden Cheetah etc.
The bigger issue is ML and the data. Companies like TR that use ML to create training platform and virtual FTP estimation had huge amounts of data available thru Strava. Now, TR may be responsible with it an not use it to train ML in violation of existing prohibitions, but not everyone is.

[fourflys]

Originally Posted by Alistair

By the new Strava ToS, they'll need to stop that and build their own integration to Wahoo...

or start paying Strava? The video didn't get into that, but I assume Strava has at least considered letting platforms like TR enter into a contract with them to use their data.. I'm guessing I signed away my right to my data I give to Strava in some user agreement somewhere..

[unterhausen]

Originally Posted by glepore

Companies like TR that use ML to create training platform and virtual FTP estimation had huge amounts of data available thru Strava. Now, TR may be responsible with it an not use it to train ML in violation of existing prohibitions, but not everyone is.

I feel sorry for trainerroad right now because making it so they avoid running their ml on just strava data sounds like a pain.

TR has my data from strava because I authorized TR to get it. It's my data. I want tr to analyze it. AFAIK, all the data that TR ever got from Strava was from a TR user's account. Strava just thinks they can get money from TR users by withholding it from TR. In my case, they are quite mistaken.

At this point, Strava doesn't even have the capability to do what other platforms are doing with the data. Just cutting off the stream isn't magically going to give them that capability, so they are just screwing their users. Pretty typical for them.

There are issues that strava users have been complaining about for years that the company doesn't address. Now they are going into an area where they have no expertise and nobody is going to trust their results. This is a serious endeavor, they aren't going to get it right immediately, if they are even trying right now. It's demonstrating an incredible level of hubris to think they can replicate overnight what other companies have been working on for years.

I'm not really upset, but sometimes I wish Strava would just go away. Like I said in a previous comment, the only thing that's this is going to affect for me is my zwift workouts being transferred to TR. I like that TR tracks my training stress. They've been doing that with data they got from Strava for years, but it looks like now it's a TOS violation.

[glepore]

Originally Posted by unterhausen

I feel sorry for trainerroad right now because making it so they avoid running their ml on just strava data sounds like a pain.

TR has my data from strava because I authorized TR to get it. It's my data. I want tr to analyze it. AFAIK, all the data that TR ever got from Strava was from a TR user's account. Strava just thinks they can get money from TR users by withholding it from TR. In my case, they are quite mistaken.

A quote from a clairification message strava sent out'"Our previous terms already disallowed the use of Strava user data in model training and development but we’ve made this more explicit in light of the increasing activity in this space. ".

It may be that TR has honored this. I agree that the data that you uploaded to TR is yours, and that you're knowledgeable about how TR uses it, and faik TR is completely responsible. Who knows about other players in the market.

I'm not in the garmin universe so idk how their transfer works. But wahoo allows you to synch just about any service (RWGPS, Training Peaks, Dropbox) so you could import to TR from somewhere other than Strava.

[Likes2ridefar]

Never considered a file drop location for syncing. Would it work to a NAS on my home network?

[benb]

Originally Posted by marciero

Sometimes you dont need a shotgun to kill a flea. For some riders who have a few years worth of data - ride and race data, biometric and health data, coaching feedback in the form of simple metrics, even diet and sleep- it would be hard to beat conventional modeling approaches. By conventional I would include a lot of "traditional" machine learning approaches, many of which have been around for decades (or even hundreds of years in one case). I agree that for full blown AI to work would require a huge amount of rider data, ideally including personal data, I would think. Maybe some public data can be a proxy for some of the private data but thats going to be a challenge. But I think we are a long way from being able to have a chat gpt type back and forth on a Sunday to decide what your training regime for the week will be.

I basically agree with you.

My gut feeling is Strava doesn't really have the resources to build something like that.

The current "AI/LLM describe my ride" feature is really low hanging fruit that a couple of people could have built in a week.

A lot of the games they are playing that annoy customers seem to be a result of their business model not exactly bringing in massive money for a tech company.

[glepore]

Originally Posted by Likes2ridefar

Never considered a file drop location for syncing. Would it work to a NAS on my home network?

Not unless you were a programmer, I suspect. The dropbox thing works for me only because its an option in Wahoo's firmware. You'd need to send the .fit file from your bike computer to a server that was addressable by the computer or its companion app, and then move it to your NAS. But the more difficult issue is synching the receiving service-their API needs to be able to pull data from that source. I don't know what platforms allow synch with each other-as an example I use Golden Cheetah and Xert-GC allows either automatic on startup or manual synch with Xert and Dropbox (and a bunch of others), but Xert only auto synchs with Wahoo, Garmin, Strava and Zwift. So Wahoo loads to Dropbox and Xert (also strava, rwgps and a free trainingpeaks acct), and GC synchs with Dropbox.

[EB]

Originally Posted by benb

A lot of the games they are playing that annoy customers seem to be a result of their business model not exactly bringing in massive money for a tech company.

This is the entire story, coupled with fairly frequent leadership and strategy changes over their long history - Strava was founded in 2009.

Based on news sources, Strava has raised a total investment of $152 million dollars over 7 rounds of VC funding. Those investors - at minimum - require a 10x return for their investment to make sense - so $1.52 billion dollars. And if you assume a 30% ownership stake for those investors, which might be a very conservative assumption, the implied minimum required valuation is $4.5 billion. For a subscription service, you're looking at 12x recurring revenue best case. Strava's subscription gross revenue is $144 per user, so that gets you to 3 million paying subscribers, assuming a 0% churn rate, which is probably a pretty bad assumption.

Their reported revenue in 2003 was $275 million. If you make some wildly wrong assumptions about that number - 100% from subscription revenue, 0% churn rate - that gets you to 2 million paying subscribers, which is far short of even the minimum of where they need to be. And the likely number is probably quite a bit lower given the generous assumptions that I've made.

Not surprisingly, they are extremely cagey about the subscriber numbers - those numbers aren't released anywhere, since they reveal far too much about their business and how much traction it has.

[benb]

APIs don't pull... you would need to script something to actually POST/PUT your data file to their API.

That's not hard to write but it does basically require a programmer. It is certainly a task you can learn with some tutorials though if you're motivated.

[Alistair]

Originally Posted by glepore

I'm not in the garmin universe so idk how their transfer works. But wahoo allows you to synch just about any service (RWGPS, Training Peaks, Dropbox) so you could import to TR from somewhere other than Strava.

That's assuming TR has an integration with RWGPS, TP, etc. Currently, their documented way to get data from Wahoo devices is via Strava. They really should build a direct interface with Wahoo's Cloud API.

[Alistair]

Originally Posted by unterhausen

I feel sorry for trainerroad right now because making it so they avoid running their ml on just strava data sounds like a pain.

It would be trivial to code imported workouts by source (if they don't already do it - I'd be shocked if they don't - most apps show a "recorded by Apple" or "recorded by Strava" or similar messages per workout). And then they limit their AI process to non-Strava workouts.

[EB]

A postscript on the VC story around Strava is that some of the VCs investing in Strava might not be acting entirely rationally around this investment, given how many of them are also roadies, and have built a professional network around bike riding - see "cycling is the new golf."