The Cycling Industry and Software - A Commentary

Back in 2017, I did some research into how cyclists use training software and what they use it for. The research supported a user experience initiative at GoldenCheetah, and we wanted to better understand the user experience athletes had with cycling software as a whole. Since the research didn't deal with GoldenCheetah specifically, but aimed at the overall user experience,I published a report and a three-part series of articles that tried to make sense of the results.

Since the Internet doesn't forget, I was contacted recently by Eric DeGolier of Body Rocket, a company developing aerodynamics sensors for cyclists, and we had a great conversation about the research, the findings, and how the software landscape has changed in the eight years between the report and now.

Much to my dismay, I have to say that not much has changed at all.

Of course, if you look at the market players, there were some earth-shattering movements. Zwift and Wahoo came onto the scene with force, bringing along a couple of companies offering something similar, big players like TrainingPeaks have established themselves more firmly by extending their moats, some companies left the market or were absorbed by others (for example, Sufferfest). Yet others who were once dominant are now just a shadow of themselves (who is still using Garmin Connect?).

But apart from new offerings, the fundamental user experience has not changed. On the contrary, the new offerings even contributed to the feeling of fragmentation, because they added functionality and platforms, without rethinking integration beyond Strava.

Strava as a Data-Hub

One curious development was that Strava established itself as the platform for cycling. It also has a great developer experience. As a result, practically everyone integrates with Strava, turning it into a data-hub for exchanging all kinds of ride information.

Capabilities-Thinking as Dominant Product-shaping Force

In the original article, I wrote:

"It (...) means that all software tools are overspecialized into a set of features that may or may not serve a specific audience (...). (I)t looks like most vendors created their feature set based on a certain technology or a certain idea, but not on needs people actually have."

Source

I think this is still more or less the case. One example I can think of that operates within a different paradigm is Zwift. Yes, it was built around the idea of a gamified indoor-cycling platform. But since its inception, it integrated many other elements of the ecosystem, branching into hardware and racing, and is now able to offer a mostly complete user experience. They followed option 1 I described. And where this is not possible or desirable from their perspective, they started integrating (option 2) with other software vendors (for example, the recent integration of TrainerRoad).

But that's the only product I can think of that seems to really put the user at the center and not start out with a set of capabilities and asking "hmmm what can we do with that?".

Adaptive Training Plans

In The Planning Conundrum, I talked about how athletes understand the importance of planning, yet do not use software to do so.

My conclusion was that software needs to be able to predict changes in form in relation to training goals, and adapt a laid-out training plan to those changes. Basically, what a trainer or coach does.

Back in 2017, such tools did not exist. Fast-forward to today, and both features are a staple of training planning software. For example, I use TrainerRoad, and it allows me to select training plans according to my target events, and will automatically adapt the plan according to my performance, and my body's response to training load.

Sense-Making

I think this is where the cycling software industry as a whole really missed out.

Everyone is still clinging to the raw data (and as such, a training plan is also just raw data), neglecting the need for a guide that helps athletes make sense of all that data.

And if there is a sense-making layer added to the data (think Recovery Score for Whoop), it is just as fragmented as the raw data was eight years ago. Both TrainerRoad and Whoop try to predict when I am rested enough for a tough workout, and when I should recover and take it easy. But do they have access to each others data? No. Do they come to the same result? Also no. Is this a useful state of affairs? Again, no.

One important thing to realise: Most athletes have goals. And if your software doesn't help them understand how they are progressing towards these goals, it is next to useless.

Software companies all to often are still focused on creating neat little graphics of raw data, sometimes accompanied by some sort of interpretation of said data that is on the level of "this was a hard ride". I know. I was there. I don't need a computer to tell me how hard my ride was.

I want to know why it was hard. I want software to help me understand my training, and my body's response to it. I want it to find the patterns that I am missing.

We're swimming in a sea of data, yet the cycling industry still struggles to make anything useful out of it.

I want software to contextualise my training experience with the training experience of all the other athletes out there. I just have a single data point: Me. I want someone make sense of what is happening who has many more data points. Traditionally, this is what a coach is for.

Back in 2017, I wrote:

"Can a software algorithm successfully coach an athlete with average goals and enough discipline to follow a simple training plan? I don’t know. But if it can, the internet coaching business is in for a rough ride."

I don't know if that really is true. It might also be the case that those who chose to pay a coach back then also choose to pay a coach now, regardless of the feature set of training software. But with the help of those basic coaching capabilities, more people have access to this sort of support. People who would never get a coach, but are willing to invest in a software subscription.

And if that is true, and automated training planning expanded the user base for such features from human-coached athletes to software-coached athletes, the same might be true for such advanced sense-making features.

The big question in 2025 is of course what role AI will play. Is it the enabling technology for creating such a sense-making layer?

Summary

Overall, it is my impression that many results of the 2017 research still are valid today. Things might look different, but the basic pattern of fragmentation and lack of data integration is still recognisable.

I think this is a shame. The hardware experience radically changed with the introduction of Bluetooth Low Energy. The software part of cycling is still looking for its Bluetooth-moment.