How trace started: One platform instead of four tools
Generic AI summaries, cluttered interfaces, four tools for one training decision. The wish to unite all of that in one platform was the beginning of trace.
Anyone who trains seriously knows the routine: after every session, an AI summary pops up somewhere. „Solid endurance ride, watch your recovery." Three sentences, interchangeable, with no real connection to my last load week, my CP development, or my season goals. Generic text that would look the same for every athlete.
On one platform the power analytics were solid, but the interface looked like a database mask from the early 2000s. On another the design was sleek, but for an honest training-load picture I had to export and continue in a second tool. A third tool for the CP curve. A fourth for training planning. Every app solves one piece well and falls short on the rest.
Out of that grew the wish for a platform that brings it together: an interface that doesn't overwhelm, AI summaries that actually look at your data, and every relevant metric in one place. That is where trace began.
Phase 1: The foundation
The first prototype was deliberately minimal: login, manual FIT-file imports, and a simple activity list. From the very first test it was obvious that „showing activities" can't be the point. The point is: what does this activity tell me in relation to every activity before it, and what do I take away for the next one? An isolated session is statistical noise. Only the sequence tells a story.
Phase 2: The power curve
The CP curve was the first feature that was genuinely fun to build. From every activity we extract the best mean-maximum-power values for 5 s, 15 s, 30 s, 1 min, 5 min, 20 min and longer, and overlay them across your entire history. From those we calculate Critical Power and W'. Suddenly it was visible: where are you actually strong, where do you have a gap?
Phase 3: Understanding load, not just measuring it
Next came training load. The classic metrics ATL, CTL, TSB have been standard for decades, but most apps spit them out as bare numbers. We wanted more: what does a TSB of -25 mean? When is a recovery day worth it, when is it time for a hard session? The answer belongs on the same screen as the number itself.
Phase 4: The second dimension – efficiency
Watts and heart rate alone aren't enough. Only their ratio reveals whether you're getting fitter or just more tired. A 200-watt ride at 145 bpm is a different training stimulus than 200 watts at 165 bpm. Aerobic Efficiency was the first metric where several users said: „No other app showed me this so clearly."
Phase 5: Durability
Pro teams have been talking intensely about durability for the past two or three years, meaning the ability to maintain power after sustained load. We took the concept from the studies by Leo et al. and Mateo-March et al. and integrated it into the CP curve. The „Fatigued Overlay" shows your best values after 20 kJ/kg of work done. The gap between the fresh and fatigued curve is the most honest diagnosis an endurance athlete can get.
Phase 6: Run training
Many cyclists run in winter. Many runners ride for recovery. So we added pace-based run zones and a separate run training load: accounted for separately, but unified in the overall picture.
Phase 7: The trace Coach
The last big piece: the trace Coach, an AI coach that interprets all the numbers. Not as a generic chatbot, but with full context on your values like CP, TSB, recent sessions, and season goals. It speaks both German and English, knows the terminology, and can suggest a plan for next week that fits your current form.
Where trace is today
Direct API integrations are live for Wahoo, Polar, Suunto and Hammerhead; Garmin and Strava data flow in through intervals.icu. And the outlook that originally stood here has since become reality: training plans that periodize toward your target race, a daily readiness value built from load and health data, and automatic insights that analyze relevant sessions right after the sync. On top of that, the trace app for iPhone has launched.
trace isn't finished. But it's already the platform that was missing back then: an interface that works, AI analysis with real substance, every metric in one place. That was the goal from the start.