A smartwatch has a small set of sensors: heart rate, blood oxygen in some models, motion, skin temperature, and sometimes an electrode for a single-lead trace. From those, software infers sleep stages, stress, recovery and readiness. The gap between measurement and inference is where most marketing lives.
What is measured directly
- Heart rate, via optical sensing, reliable at rest and less so during intense movement.
- Motion and orientation, which give step count and activity type.
- Skin temperature, useful as a trend rather than an absolute.
- Blood oxygen, which varies with fit and motion and is not a medical measurement.
What is inferred
Sleep stages are inferred from motion and heart rate, and they disagree with lab equipment more often than the marketing suggests. Stress scores are a transformation of heart-rate variability, which is real but sensitive to caffeine, illness and posture. Readiness scores are a model, not a measurement, and every vendor weights the inputs differently.
Why trends beat single readings
A single night of poor sleep data means little. Four weeks of the same pattern is worth acting on. The useful output of a wearable is your own baseline, and any score that does not show you the baseline is asking for trust it has not earned.
Questions worth asking before buying
- Can I export the raw data, or only the scores?
- Does the device work without an account?
- Which features require a subscription?
- What happens to my history if I switch brands?
The privacy trade
Sleep, heart rate and location together describe a life unusually well. That is why the account requirement matters more than the sensor list. A watch that stores data locally and syncs to your own machine is a different product category from one that uploads continuously.
Bottom line: trust the measurements, question the inferences, and judge the device on whether it lets you see the underlying data.
