Accuracy & Validation Master · As of July 2026

Accuracy you can trust,
at every layer.

Luna is a full-stack health system, and accuracy is engineered into every layer of it. This page doesn't just walk you through core biomarker testing, it walks you through how accuracy decisions taken at hardware, biomarker measurement, analysis, environmental conditions, and real-world scenarios are measured in Luna.

0
HR accuracy %
0
HRV accuracy %
0
Steps agreement %
0
Readings benchmarked
We don't grade ourselves against other wearables

We measure against instruments that set the standard.

It is tempting to say we are more accurate than this ring or that watch. We won't. Comparing to another consumer device assumes that device is right, and too often it isn't. Most public wearable accuracy data is marketing data. If we want to lead where health guidance goes next, our job is to raise the bar, not match it.

Heart ratePolar chest strap. Chest placement gives the highest signal quality and is the best consumer-grade reference to optimize toward.
HRVPolar H10, the accepted chest-strap standard used in sports-science labs.
StepsLiteral manual step counting, plus Apple Watch head-to-head, removing device-to-device bias.
RespiratoryInternal lab tests against industrial-grade equipment (CPAP-class references).
Consumer-device comparisons appear only as context, never as ground truth.
Accuracy is not one test. It is engineered at five layers.

Explore the stack

It starts with the signal

Better guidance begins with better raw data. Luna senses every second through multiple light channels and a six-axis motion sensor, so the signal we build on is among the richest any wearable captures. Nothing downstream can be more accurate than the signal it starts from.

ComponentSpecification
Optical sensor7-LED optical array (4 green + 1 multi-wavelength green/red/IR) with 4 high-sensitivity photodiodes for stronger signal across HR, SpO2, HRV, and stress.
Motion6-axis high-precision IMU (3-axis accelerometer + 3-axis gyroscope) for activity, workout recognition, and sleep analysis.
ProcessorUltra-low-power wearable processor (Apollo3 Blue Plus) tuned for continuous per-second sensing and extended battery life.
Skin temperatureHigh-accuracy skin temperature sensor for continuous wellness tracking and recovery insight.
Wear detectionSAR + PPG combined wear detection for reliable on-body sensing.
Water resistance5ATM rated for continuous sensing in and around water.
Battery160 mAh, managed by a dedicated charger IC for efficient, safe charging.
FeedbackVibration motor and RGB indicator for haptics and device status.

Then we make it true for you

Your body is not the average body. LifeOS Edge, our on-device firmware, turns raw signal into accurate data across the variables that trip up other wearables. It is validated in controlled conditions, then with real people, and gets better with every firmware release.

Where we stand
We are proud of these results, and we are not claiming perfection. No sensor is right for every body in every moment. What we commit to is this: we keep finding the edge cases, and we keep closing them, release after release. Accuracy at Luna is a direction of travel, not a finish line.
Validated acrossWhy it matters for accuracy
Skin toneOptical absorption varies with melanin; Edge normalizes so readings hold across all skin tones.
Body hairHair scatters light between sensor and skin; compensated so signal quality stays high.
Band placementSensor-to-skin position shifts the optical path; corrected across realistic wear positions.
Band tightnessLoose or tight fit changes contact pressure and perfusion; handled across a fit range.
Workout intensityMotion artifact is where optical HR usually fails; the 6-axis IMU separates it from true signal.
Sweat & moistureMoisture alters reflectance; validated wet and dry so readings stay stable mid-workout.
Age & genderPhysiological baselines differ; models validated across demographic groups, not one cohort.
Hand movementEveryday motion adds noise; filtered so resting and active readings both stay clean.
Firmware release log · LifeOS Edge 1.3.1 · full per-release notes pending internal export

We take it everywhere you go

Deserts, humidity, altitude, snow, a hard knock, and five atmospheres underwater. We test the signal in each, because your life doesn't happen in a lab and your data shouldn't only hold up in one. Every one of the 19 reliability and durability tests passed, with zero failures.

Luna test lab
Luna test lab
Luna test lab
Luna test lab
Luna test lab
Luna test lab
#TestWhat it provesResult

And we keep learning, in real time

Clinical research is a starting point, but it can't see how a stressful day changes the way magnesium works for you. Because Luna senses continuously and models your biology as it goes, LifeOS Cloud discovers how your habits move your own biomarkers, and turns that into guidance built for you, not for a population.

Go deeperThe full story of what LifeOS Cloud discovers lives in the LifeOS Research space: how habits move real biomarkers, tracked over 1.5 years of building.
See LifeOS Research →
CapabilityWhat it does
Per-second sensingContinuous capture creates the density of data needed to model an individual, not an average.
Personalized health modelA live per-user model (PHM) updated in a real-time feedback loop from that user's own signal.
Habit-to-biomarker discoveryAnonymized data reveals how specific habits move a person's biomarkers over time.
Research foundationBuilt over 1.5 years of research; guidance is grounded in real discovery, not fixed rules.
Research stats · LifeOS Cloud · headline discovery numbers pending from research space

Held to a responsible standard

Accuracy without responsibility is not accuracy at all. Everything Luna ships is held to the same standard as the data it produces: low-emission wireless that stays within strict radio limits, battery safety validated against thermal and charge-fault conditions, and restrictions on hazardous materials so the device is safe on your skin and safe at end of life. Twelve certifications are held today across the EU, UK, USA, UAE, Japan, and Australia and New Zealand, spanning radio conformity, electrical safety, EMC, hazardous-substance limits, and e-waste take-back. Each one is an independent check by a regulator, not a claim we make about ourselves, and the list grows with every new market Luna enters.

CertificationRegionWhat it covers
Heart rate · vs Polar chest strap and 4 reference devices
0
Mean accuracy %
0
Paired readings benchmarked
0
Mean absolute error, bpm
0
Mean Pearson correlation
Weekly trend · 225 sessions · May 04 – Jul 14 2026
MAE improved from 2.25 bpm (week of May 04) to 1.73 bpm (week of Jul 13). A MAE under 5 bpm is generally considered a good reference point.
By reference device
ReferenceAccuracy %MAE bpmPearson rPaired readings
Polar (standard)98.482.040.97209,352
Apple Watch98.721.800.979,342
Zepp98.532.120.9716,225
WHOOP97.403.610.962,505
Garmin97.982.790.9422,571
Polar chest strap carries the most weight by design: it is the reference standard. 259,995 paired comparisons in total.
Session accuracy distribution
Lowest single session: 95.9%. 168 of 225 sessions above 98%.
HRV · vs Polar H10 · the hardest metric to get right
95.4%
Agreement with the Polar H10 chest-strap standard across overnight and resting benchmarks. HRV is measured in milliseconds between beats.
4.6%
Mean absolute percentage error across sessions
+1.87 ms
Directional bias: under 2 ms, well inside noise for HRV
Luna vs reference · anonymized benchmark sessions · 27 to 117 ms range
Points on the line are perfect agreement. Accuracy holds across low and high HRV, where errors usually appear.
Sleep · vs Apple Watch and WHOOP
0
User types tested
0
Nights compared
0
Mean delta vs Apple, min
0
Mean delta vs WHOOP, min
Total sleep · Luna vs Apple Watch · 40 nights
Mean absolute delta 17.3 min per night.
Total sleep · Luna vs WHOOP · 32 nights
Mean absolute delta 12.7 min per night.
User types tested
#SexAge
01Male30
02Female36
03Male46
04Female29
Steps · literal manual counting + Apple Watch
0
Median agreement % · 308,371 steps recorded

Manual counting removes device-to-device bias entirely, and Apple Watch head-to-head sessions add real-world coverage. Median agreement is 97.2%, with best sessions above 99%.

Reference · Manual count Reference · Apple Watch
Head-to-head sessions
ReferenceLunaReferenceDiffAccuracy %
Representative sessions shown, not cherry-picked. Median across the program: 97.2%.
Luna vs reference steps · per session
Bars near equal height mean close agreement. Manual and Apple Watch references combined.
Certifications

Held to a responsible standard

Accuracy without responsibility is not accuracy at all. Everything Luna ships is held to the same standard as the data it produces: low-emission wireless that stays within strict radio limits, battery safety validated against thermal and charge-fault conditions, and restrictions on hazardous materials so the device is safe on your skin and safe at end of life. Twelve certifications are held today across the EU, UK, USA, UAE, Japan, and Australia and New Zealand, spanning radio conformity, electrical safety, EMC, hazardous-substance limits, and e-waste take-back. Each one is an independent check by a regulator, not a claim we make about ourselves, and the list grows with every new market Luna enters.

CertificationRegionWhat it covers
Luna plans your day. Accuracy is where that promise begins.

We are at some of the best results in the world today, and we are not stopping here. We will keep building and improving this, together with you and with the community's feedback.