TrainerRoad · 2021
Adaptive Training
Built on Plan Builder. Grades how each workout went and adjusts the rest of the plan. I ran the beta from 100 athletes to 20,000 before it went on for everyone. It now adjusts 72% of plan workouts before they're ridden.

Challenge
Athletes fail workouts at a high rate. Not because they lack motivation. Because static plans don't adapt. A workout prescribed three weeks ago has no idea about the cold you had last week or the extra stress at work. The plans treated every athlete the same, and athletes paid the price in failed sessions and lost motivation.
Approach
I led the ML/AI product team from concept through launch, on top of Plan Builder. We built Progression Levels, a 1-10 scale that tracks each athlete's fitness across every training zone. We built Workout Levels to rate every workout's difficulty within its zone. Then we layered on difficulty ratings (Recovery, Achievable, Productive, Stretch, Breakthrough, Not Recommended) so athletes always know where a workout sits relative to their ability. The system adjusts the plan after every completed workout and handles real life: missed sessions, time off, great weeks, bad weeks. We added Workout Alternates so athletes could swap for shorter, longer, easier, or harder options without breaking the plan. The tagline was 'The Right Workout. Every Time.' and we meant it.
Impact
I ran the beta from 105 athletes in March 2021 to 20,545 by October, a quarter of subscribers, then switched it on for everyone. Today 72% of plan workouts are adapted before they're ridden. The surprise in the data: adapted workouts run a little harder than the originals, because the system moves difficulty toward the athlete's real level instead of letting it drift easy. The right workout, not the easy one.
Reflection
The athletes who benefit most are not the most talented ones. They are the ones who show up consistently and let the system work. Same as anything else.