In December 2016, while I was still at Reonomy, I was trying Aaptiv and talking with Ethan about joining the company.
One of the classes I tried was a walking meditation. I laughed a little while doing it because being guided to notice different parts of my body and the things around me was unfamiliar. Then I realized that I liked it. The walking classes were, as I wrote at the time, “crazy relaxing.”
That surprised me. I was a serious endurance athlete who spent a lot of time on running, cycling, training schedules, and data. A guided walk was not the part of the product I expected to care about.
I sent Ethan notes on the app. The updated mobile navigation made more sense to me. I thought the icons could help people understand where they were as the layout changed. I also thought training plans needed a more prominent and more complete place in the product because following a plan was different from choosing a single class.
The class catalog would need better metadata as it grew. Duration and intensity were useful, but they were not enough for good search or recommendations. We needed a clearer way to describe what a class was for, how it fit into a program, and who it might be right for.
Training plans also changed the navigation problem. A member following a calendar should not have to search the full catalog every day and remember where she left off. The app needed to make the next planned workout easy to find while still allowing somebody to choose another class.
Aaptiv already did a good job after someone pressed play. A trainer explained the pace, effort, recovery, and structure of the workout. The music and coaching removed a lot of uncertainty while the person was exercising.
The larger product question was what happened before that. Which workout should someone choose today?
I told Ethan that taking the guesswork out of the workout itself was a $100 million idea. Taking the guesswork out of choosing the workout could be a $1 billion business. Those were my own rough opinions in a private product conversation, not company valuations.
I thought the second problem was much harder. A large catalog gives people options, but it does not necessarily help them decide. Someone might be choosing among a hard run, an easy ride, strength, recovery, meditation, or a rest day. A new athlete may not know how to compare those options. An experienced athlete can still spend too much time arguing with himself about them.
I knew that from my own training. I liked calendars, spreadsheets, pace data, and detailed plans. I was willing to spend time building them. Even then, I regularly had to answer the basic question: what do I do today?
I thought Aaptiv could use information from the member, workout history, and better class descriptions to recommend something appropriate. Training plans also needed to be treated as a sequence rather than a loose collection of classes. A recommendation for one day should account for what the person had already done and what was coming next.
I also asked whether members could help label workouts. People using the classes had useful language and judgments about them. We would still need editorial control and clean data, but member behavior could help us understand how classes were actually being used.
Any recommendation system would have to earn trust. Exercise costs time and energy, and sometimes it hurts. If the app repeatedly suggested the wrong class, people would stop following it. If somebody only had twenty minutes, a technically ideal forty-five-minute workout was not useful.
We would need better class data, user feedback, workout history, and a simple interface around all of it. That was a bigger technical and product problem than reorganizing the catalog.
I was not writing this as an outside advisor. I was using the product and deciding whether I wanted the job. The walking meditation helped. So did the fact that the hardest product question was still open.