Editor's note: An abridged version of this Q&A ran in the May issue of Bicycle Retailer & Industry News.
RENO, Nev. (BRAIN) — Even before the pandemic, the intersection of cycling and technology has been where all the action is. Companies that offer high-tech hardware or software have generated press coverage, excitement, investment, and participant growth faster than traditional suppliers.
Think Zwift, Peloton, Wahoo, and Strava, not to mention the many new e-bike brands that continue to bring in multi-million dollar investments.
Although it's stayed out of the headlines, TrainerRoad is a perfect example. It has grown steadily since its launch in 2011. The privately held firm doesn't release revenue figures, but will say that it's grown to 82 employees with another 15 job openings and plans for a South African office.
We can say with confidence there are many well-known brands on the hardware side of the bike industry with many fewer employees and correspondingly lower revenues.
TrainerRoad launched when smart trainers and power meters were rare and expensive. It calculated power from "dumb" trainers, allowing riders who lacked power meters to train with power estimates indoors.
Now TrainerRoad links smart trainers, power meters and other gear with specific workouts and training plans. It connects to hardware in a way similar to the better-known Zwift application, but doesn't simulate outdoor riding or racing and has little social aspect beyond a video group workout feature that is something like a sweaty Zoom call.
This spring TrainerRoad is rolling out a major update with a new feature called Adaptive Training. AT uses machine learning to tailor specific workouts for athletes, based on their recent performance, indoors or out. Instead of following a static calendar of workouts, athletes get updated suggestions for the day's best workout.
BRAIN spoke with CEO and co-founder Nate Pearson as the company was just beginning to roll out Adaptive Training in beta tests. This interview has been edited for length and clarity.
BRAIN: We've been publishing a lot of stories about the industry's supply chain woes, but it occurred to me that TrainerRoad is completely isolated from those issues.
NP: I was just talking about that, too. At-home fitness is the perfect product for a pandemic. That or a grocery store or if you made masks I guess. There are few businesses that did better in the pandemic than those.
BRAIN: How has your company been operating during the pandemic, is everyone working remotely?
NP: A little over half our company is remote anyway. So when the pandemic hit we just shut down the office. Now I'm here and the office manager is here, but it's 9,000 square feet so we can stay away from each other. And everyone else is home. I think everyone is ready to come back but they still like to be at home.
BRAIN: So you were already hiring remotely before the pandemic?
NP: We increased it recently. Before, the test team and support would be in Reno and some marketing people were there. But engineering was mostly outside because there's such great talent out there. There aren't enough people in Reno but there's so much talent in the Midwest, the East, and the South. If you pay them well they get over-market wages and it's great.
We also are going to have an office in South Africa for both support and testing. We can get better testing hours early in the morning on East Coast time. And for support it's good because (the time zone) is the opposite of the U.S. We have some international support people already but we want to fill it up so you can get an answer at 3 in the morning right away.
BRAIN: Is that a pretty typical tech company strategy?
NP: Yeah, South Africa and Lisbon, Portugal, are the big hot spots right now for outsourcing tech talent. But in South Africa the culture is so similar to here and we are going to be in Stellenbosch and we are going to sublease from a bike company. Specialized Africa is right there and there's a little cafe and you'll see (Swiss mountain bike world champion) Nino Schurter down there because they all come from Europe to train in their winter. So there's the culture aspect and amazing mountains and cycling is huge there so it fits really well.
BRAIN: About a year ago TrainerRoad prioritized pushing out a Group Workout feature because of the pandemic. Is the new Adaptive Training feature also a response to the pandemic?
NP: No, actually Group Workouts was supposed to come out after Adaptive Training because what we want to do when we put people together is dynamically adjust the workout so it's the correct difficulty for everybody. So if you are doing five-by-3-minute intervals at 120%, I might not be ready for that. I might need those at 115%. So (Group Workouts) was supposed to be made afterwards, with a better team functionality so you can better find and join workouts and discover them more easily.
(as for) Adaptive Training — since 2011 when we launched it's been in the plan. We started working on it about three years ago on some of the low-level ML (machine learning) pieces in order to get those pieces in place. Before that could happen, we knew we had to have a calendar system and we needed to have a way to lay workouts on there. We needed to have a way to bring in outside workouts and we needed outside structured workouts like we have today. We also knew we had to rebuild our mobile and desktop apps so we could build things more quickly on that system. So we've had backend data science people working and then our front-end people have been building different parts to connect so they can meet right now. So it's all coming together inside this time frame.
BRAIN: So is this the biggest update to TR that you've done?
NP: By far, it's been huge.
BRAIN: Did you think about calling it Trainer Road 2.0?
NP: Internally yes, we definitely think this launch is a pivotal moment for TrainerRoad. This sounds like marketing but I really do think we are going to look back at this in 5 or 10 years and see this (launch) was the start in a major way for our adaptive training and AI (artificial intelligence) stuff.
BRAIN: Does it expand your potential customer base, or does it reach deeper into the same customer type that you already have?
NP: I think it will totally reach different audiences. I think in the training (software) space there are going to be people who have this and people who don't. And the people who don't have this will either have to build it very quickly and build a database or they will have to license it. If this goes where I think it's going to go, in 10 or 20 years it will be silly to have a coach pick a workout for you. We'll look back at that and say that was antiquated for someone who has coached maybe 10 athletes or 20 and has access to maybe five studies that had 10 people in each. It seems kind of weird. It's like self-driving cars. I think in 20 years none of us will be driving our own cars. So this is a turning point.
To be clear, we don't intend or expect for Adaptive Training to eliminate or replace coaches. We do believe that ML/AI-based technology will be an essential tool that all good coaches will use to more effectively train their athletes, but not a replacement for a coach and their wisdom.
"You can have a great life, a profitable company, and impact the world without trying to be a billion-dollar company." — Nate Pearson
BRAIN: You mentioned licensing, is there a part of this new technology that could be licensed out?
NP: I definitely think this could be licensed out to people to help people deliver the best workouts on their platform. The next step for us is to go into running with the same kind of idea. We already have a whole bunch of run data brought in from our triathlete people. That's something we haven't started but that's a growth area for us, too.
BRAIN: TrainerRoad obviously has already gathered a lot of data from athletes' workouts. With Adaptive Training you are taking this to another level. So what are you doing with all this data?
NP: If you have an ML model, you have to tell it what you want to optimize for and that has to be some kind of performance improvement. That (improvement is) probably not just FTP (functional threshold power) but the performance in workouts, because that gets into repeatability, which is critical. It's not just how hard you can go for one minute, but how many times can you go hard for one minute. So you train on that and then you try to run it against the dataset where you know the outcome to see how well it performs. Then you add new data input, let's say HRV (heart rate variability). So with HRV, can we predict the outcome of workouts better than if we don't have HRV? It's all about ML and statistics.
So if some data improves predictability, that gets put into the model and we move forward. If it doesn't, then you don't need it in the model. So we want to keep bringing in more and more data to increase the accuracy of the model and how well it predicts things that would be best for you. That's a never-ending thing. It's not like you just bring in heart rate; your engineers have to quantify what that heart rate is: Are we looking at heart rate drift in a workout? Are we looking at heart rate relative to certain power outputs, are we looking at resting heart rate, or how many heartbeats you've done in the last week? There are all these different ways to describe heart rate.
With AI or ML, it's like cooking a meal. You have the ingredients, which is the data, and then you have the chef, which is the engineers. You need great ingredients and a great chef. You can't just put the data in and get results.
BRAIN: It's an interesting ask of your customers, though. You are asking them to record and share more data even though that data may not directly affect what they are doing today or even in the longer term. You are asking them to help you build a better TrainerRoad in the long term.
NP: Yes. The other thing we are adding is post-workout surveys to see how difficult the workout was for you and if you ended a workout early, why? That's a big missing piece of data. If you ended a workout, was it because it was too hard, you were too tired, or did you have a work meeting so you had to stop. If we don't know that we can't use that information to train the ML.
That data might actually affect that athlete immediately depending on the choices but also it helps them train their cohorts later on and maybe even help train that particular athlete in the future.
BRAIN: I've never seen any reports of TR receiving any outside investments, are you completely self-funded?
NP: Yes. Self-funded, boot-strapped.
BRAIN: How long can you keep doing that?
NP: Forever.
BRAIN: That's the plan?
NP: I'm not sure. There has to be a reason to take an investment. Usually when you get funding you "buy customers." That's when the marketing channel is so good that the more money you spend the customer acquisition cost doesn't inflate, it stays steady. So let's say you can buy a whole bunch of customers for x dollars, that's where getting an investment to spend on marketing makes a lot of sense. The other way is you can also grow your engineering team but unless you can grow your customers at the same time, it's hard to predict the kind of revenue you can get from that kind of investment.
I'm not completely opposed to it and I talk to investors but without getting extremely inefficient in the market I don't see a super clear path to spend tens of millions of dollars very quickly and get that kind of revenue. You could also do a play to grow the company and sell it or something, but then it gets out of control. Some of these long-term things, like Adaptive Training that took three years, that's hard to sell to an investor. They want faster growth. So it's good to have a long-term vision that you can work to, without having pressure from the outside.
BRAIN: Is there another company that's inspired you?
NP: When I first started, the guys at 37signals, which is now Basecamp, had a book called REWORK that influenced me. I used to work at a Fortune 500 company, and I loved the people but the processes and BS around the company was just insane. REWORK said, "Why can't you have more remote workers? Why can't you trust people? Why do you need an investment? Why do you need to be the biggest company in the world?"
You can have a great life, a profitable company, and impact the world without trying to be a billion-dollar company. With our mission of making people faster and, actually, making the world faster, we can do all that without having to be a huge company. So that message really resonated with me: That there's a different way than just the standard "invest-pump-sell" cycle
