What your sleep tracker actually measures
Movement and pulse, mostly. The rest is inference — here's how good it is.
Consumer sleep trackers measure two things well: movement, through an accelerometer, and pulse, through optical sensors at the skin. Everything else on the morning screen — sleep stages, "deep sleep", readiness — is inferred from those two signals by statistical models.
How good is the inference?
Validation studies comparing wearables against polysomnography, the laboratory reference standard, consistently find the same pattern: total sleep time is estimated reasonably well; the division into sleep stages is much weaker, because stages are defined by brain activity that a wrist sensor cannot see.
What this means in practice
Trends are more trustworthy than absolute numbers. If your tracked sleep shortens over weeks, that is worth attention. If the app says you had 42 minutes of deep sleep instead of 55, that difference is well inside the error of the method.
And a tracker cannot diagnose anything. If it flags irregular breathing or you wake unrefreshed for months, the next step is a clinical assessment, not a different wearable.