Sleep measurements
Each measurement needs three answers: what it represents, how well it is measured, and what improvement would establish. A physiological value can be useful to observe without being something to maximize or minimize.
Benchmark means a defined performance challenge. Range means more or less is not indefinitely better. Monitor means observe a signal before deciding whether to change it; Trend emphasizes its repeated pattern. Investigate means measurement or outcome evidence is insufficient for a numerical target. Outcome means something valuable that sleep is intended to improve. These are project roles, not clinical diagnoses.
| Measurement | Role | Current decision |
|---|---|---|
| Sleep score | Benchmark | Proposed milestones |
| Duration and sleep opportunity | Range | Establish personal sufficiency |
| Timing, regularity and circadian alignment | Investigate | Separate the three questions |
| Continuity, latency and efficiency | Range | Clarify disruption and definitions |
| Sleep stages and brain activity | Investigate | Validate before setting stage targets |
| Heart rate and heart-rate variability | Trend | Match the measurement window |
| Breathing rate and oxygen signals | Monitor | Clarify the index and validate persistent changes |
| Temperature: skin, core and environment | Trend | Name the temperature being measured |
| Alertness and waking performance | Outcome | Select and validate the assessment |
Four kinds of information
- Measured signals: photoplethysmography, movement and peripheral temperature; a signal can contain noise or missing intervals.
- Estimated states: heart rate, breathing, sleep/wake and stages inferred from signals using an algorithm.
- Composite judgments: Sleep Score and Readiness combine estimates through provider scoring rules.
- Outcomes: vigilance, sleepiness, performance, clinical assessments and longer-term health. A better estimate or composite does not itself prove an outcome improved.
What makes a measurement useful
Record units, episode boundaries, missingness, source and hardware/algorithm where exposed. Repeated nights matter more than excessive decimal places. A population-average error, correlation coefficient, sleep/wake accuracy and four-stage agreement answer different questions.
The first paired night illustrates this: the devices broadly selected the same episode but assigned different scores and awake time. The validation investigation examines the reference measurements and the evidence for Oura Ring 5 specifically.