Your Devices Disagree.
Here's Why That Matters.

EEG electrodes on scalp measure brain waves directly. Delta waves indicate deep sleep. Mixed frequency with rapid eye movement indicates REM.

Only method that truly measures sleep stages. Everything else is inference from proxy signals.

Optical heart rate sensor on wrist plus motion detection. Algorithms classify stillness combined with low HR and specific HRV patterns into stages.

Validation studies show 78-82% agreement with PSG for sleep/wake detection, dropping to 60-70% for stage classification.

Finger arteries sit closer to the skin surface, producing a cleaner PPG signal. Skin temperature naturally drops during deep sleep, providing an additional classification signal.

Slight edge on some overnight metrics due to finger vascular anatomy, but carries its own biases in position detection.

Pressure sensors embedded in the mattress detect micro-movements from heartbeat and breathing patterns. No skin contact required.

Good at total sleep time estimation. Weaker on stage classification because it lacks direct cardiovascular measurement.

Detects when the phone is placed down and picked up. Uses ambient light sensor and touch screen inactivity. No biometric data whatsoever.

Least accurate by a wide margin. Cannot detect when you actually fell asleep, only when you stopped using your phone.

Apple Watch counts extended motionless periods as light sleep even when awake. Oura may miss sleep onset in unusual positions. iPhone has no idea when you actually fell asleep.

Most contentious metric. A 2023 validation study found Apple Watch overestimated deep sleep by ~18 min/night vs PSG, while Oura underestimated by ~12 min. Without EEG, both are guessing from proxy signals.

Heart rate becomes more variable during REM, which helps detection. But some devices misclassify light sleep with elevated HRV as REM, inflating the number.

Each device uses different thresholds for the transition from "lying still but awake" to "asleep." iPhone just tracks when you put the phone down, not when sleep actually began.

“Pick the most expensive device”

“Average across all devices”

“Let Apple Health sort it out”

Sensor-Aware Weighting

Each sensor excels at something specific. Oura Ring finger-based PPG produces cleaner overnight HRV signals. Wrist accelerometers catch micro-awakenings that rings miss. Manual sleep/wake times add context no sensor provides.

Context Normalization

A 3am wrist PPG reading carries different confidence than an 11pm reading when you are still moving. Vora weighs each data point by measurement context: time of night, motion artifacts, skin contact quality, and recent activity.

Timeline Construction

Rather than choosing one device, Vora builds a unified minute-by-minute sleep timeline. Where devices agree, confidence is high. Where they diverge, the system uses sensor-specific knowledge to resolve conflicts.

Multi-Source Amplification

The more data sources you connect, the more accurate your data becomes. Two devices are better than one. Three are better than two. Each additional source adds signal that helps resolve ambiguity in the others.

Why it matters: A genuine 4 bpm increase over 2 weeks is a classic overtraining signal. A spurious 4 bpm increase from switching measurement windows is noise. If your app cannot distinguish these, your trends are unreliable.

Vora's approach: Normalize by measurement method, time window, and device before comparing across time. When you connect a new device, Vora calibrates it against your existing baseline so trends reflect genuine physiological change, not device artifacts.