May 11, 2026 · 5 min read
Personalization Is Required
Ablation testing across 4 subjects shows each person requires a different model configuration. Generic travel adaptation protocols are insufficient.
- 4 subjects tested, 4 different optimal model configurations. No single config works for all
- Athlete A adapts to jet lag 4x faster than average (0.6 days vs 2.5 day population norm)
- Athlete B shows zero travel-day response but delayed +7 bpm the next day, simplest model wins
- 94-97% accuracy across 265 scored days and 10+ international trips
We ran 7 model configurations on each of our 4 subjects and selected the one with the lowest prediction error on disruption days (travel + post-travel). The result was unambiguous: no single configuration was optimal for all individuals.
Athlete A adapts to jet lag 4x faster than the literature average. Their circadian time constant is 0.6 days versus the population norm of 2.5 days. A generic protocol would have them adjusting sleep for a week when they only need two days.
Athlete B shows essentially no acute travel-day response. Their heart rate doesn't spike on the day of travel. Instead, they show a delayed +7 bpm elevation the day after, returning to baseline by day 4. The best model for Athlete B is the simplest one: just a weighted moving average. Adding travel features makes their predictions worse, not better.
Athlete C has a muted travel-day response but a pronounced post-travel recovery pattern. The recovery model captures her physiology better than the full model because the travel-day signal is noise for her.the real information is in how she recovers.
Athlete D is the opposite: immediate travel-day spike with multi-day recovery decay. They need both the travel and recovery components. With only one trip in the system, the model missed their travel day by 8.8 bpm; accuracy improves with more exposure data.
What this means
A coach giving the same travel protocol to all four of these people would be wrong for at least three of them. The advice "hydrate more, sleep earlier, get morning light" is directionally correct but operationally useless when one person needs two days of adjustment and another needs seven.
Personalization is not a feature. It is a requirement. The system must learn the individual.and the ablation study is how we prove it did.
Validated across 265 scored days, 10+ international trips, 94–97% accuracy. Every person responded differently. The engine detected and adapted to each pattern automatically.