THE PRINCIPLE
Preserve what happened before interpreting what it means.
Source first: retain the recorded machine, workout, timestamps and original measurements before deriving a training conclusion.
Scope evidence: only let a session influence the kind of forecast it can genuinely inform.
Keep uncertainty: sparse, noisy or dissimilar evidence should reduce confidence rather than disappear behind a precise-looking number.
FROM RESULT TO EVIDENCE
Six checks before a row changes a future recommendation.
Provenance
Know whether the result came from Concept2, SmartRow or another supported source and retain the source identifiers needed to avoid duplicates.
Machine context
Do not assume watts are perfectly interchangeable across different ergometer and measurement systems.
Session structure
Identify what was actually rowed, continuous work, repetitions, recovery and breaks, rather than forcing the upload to match an old prescription.
Execution quality
Incomplete, interrupted or technically ambiguous sessions should carry less forecasting weight than clean comparable work.
Capability relevance
Zones learns steady-state response and drift. Intervals learns power-duration and interval execution. Target keeps comparable benchmark results and pacing. Evidence crosses over only where the underlying quantity genuinely does.
Confidence
Recommendations should expose when they rest on strong personal evidence, weaker analogues or generic starting assumptions.
WHY MACHINE CONTEXT MATTERS
A watt is useful only when you know how it was measured.
Rowing-ergometer research shows that mechanical design can change stroke mechanics and physiological demand, while separate validation work has found systematic differences between power displayed by a Concept2 system and power calculated from independent mechanical sensors. Those studies do not imply that one device is unusable; they show why provenance should not be thrown away.
Rowing Peak therefore keeps power-learning machine scoped. That is a conservative product decision: compare like with like unless there is strong evidence that two sources can safely be combined.
BOUNDARIES VS PRESCRIPTIONS
Training evidence can refine a session without proving a physiological threshold.
This separation is especially important in Zones. A completed steady row can show how heart rate settled, how it responded to the prescribed work and how it drifted later. That can improve the next session prescription. It does not, by itself, establish the athlete's lactate or ventilatory transition.
Intervals has a parallel distinction. Maximal efforts can support a power-duration model, while ordinary interval sets are better used to test execution and repeatability than to silently redefine CP or W′.
Target keeps another boundary clear: a completed benchmark is a measured result. Its achieved pace can anchor a later comparable attempt, but it is not converted into a promise about what the athlete will do next.
THE PRODUCT PROMISE
Explainable personalisation beats invisible adaptation.
A rower should be able to understand why a recommendation changed: which sessions informed it, which machine they came from and how confident the model is. Rowing Peak's long-term advantage should come from accumulating better evidence, not from hiding more of the arithmetic.