Wearable reference · Documentation checked November 16, 2026
Two sleep trackers can disagree because they detected different sleep, counted different time windows, or turned similar observations into different scores. Before deciding which device is wrong, find out which of those disagreements you actually have.
An Oura score of 85 and an Apple score of 85 are not interchangeable measurements. A Garmin sleep score is not its Body Battery. WHOOP Sleep Performance is not WHOOP Recovery. The number becomes useful only when its definition stays attached.
We compared the manufacturers' current explanations and built the reference below. This is documentation research, not a claim that we wore these devices together. The worked examples are hypothetical; the downloadable comparison log starts empty.
Compare score definitions · Work through a disagreement · Download the reference and log
Four scores, four different questions
| Product and metric | What its documentation includes | Comparison caution |
|---|---|---|
| Apple Watch: Sleep Score | Sleep duration (50 points); bedtime consistency (30); interruptions (20) | Do not translate the same number into another brand score. |
| Oura: Sleep Score | Total sleep; efficiency; restfulness; REM sleep; deep sleep; latency; timing | Confirm main sleep versus total including naps before comparing durations. |
| WHOOP: Sleep Performance | Sleep sufficiency; sleep consistency; sleep efficiency; sleep stress | Sleep Performance and Recovery are different metrics. |
| Garmin: Sleep score | Sleep duration; stress; sleep stages; additional quality factors | Sleep score is not Body Battery or Training Readiness. |
These are disclosed contributors, not a reconstruction of each company's complete algorithm. Features depend on device and software support. The linked source for each row is the authority for that product's current definition.
Apple's published split makes the point clearly: duration can contribute 50 points, bedtime consistency 30 and interruptions 20. Matching total sleep duration therefore cannot guarantee matching Apple scores. Apple's explanation also says bedtime consistency looks at the preceding 13 nights.
Oura's contributor documentation lists seven inputs. It includes naps in total sleep, while latency concerns the longest sleep period. An overnight-only number and a total that includes a nap can differ even when neither is missing that overnight period.
WHOOP's updated Sleep Performance description includes sufficiency, consistency, efficiency and sleep stress. An older explanation that reduces the score to sleep obtained divided by sleep needed is incomplete for this documented version. Always attach a check date to an explanation of a changing algorithm.
Start with minutes, not colors
Imagine two apps showing the same night. Tracker A says 7 hours 10 minutes asleep; tracker B says 6 hours 35 minutes. Their scores are 84 and 73. The 11-point gap is hard to interpret because the scores have different definitions. The 35-minute duration gap is a more concrete place to begin.
| Observation | Tracker A | Tracker B |
|---|---|---|
| Detected start | 11:10 p.m. | 11:35 p.m. |
| Detected end | 6:50 a.m. | 6:45 a.m. |
| Elapsed interval | 7 h 40 min | 7 h 10 min |
| Minutes marked awake within it | 30 min | 35 min |
| Calculated time asleep | 7 h 10 min | 6 h 35 min |
Here, 30 of the 35 minutes arise from the different start and end boundaries. The remaining five arise from the awake estimates. That directs the investigation toward the beginning and end of the night, rather than into an argument about whether 84 is a better score.
The example does not establish which tracker is correct. A note that you were reading until 11:30 p.m. could help explain the first boundary, but your recollection is not a clinical reference for every minute. Keep the original records before making any manual edits.
Now imagine both apps agree on 7 hours overnight, but one dashboard also includes a 30-minute nap. A daily total of 7 hours 30 minutes and an overnight total of 7 hours answer different questions. Changing sensor accuracy would not fix that mismatch in scope.
A practical order for checking a disagreement
- Match the date and time zone. Compare the same night, especially after travel, a daytime sleep or a sleep period crossing midnight. Record the actual start and end, not just the date label.
- Name the field. Write down time in bed, time asleep, main sleep, total sleep including naps, or the exact score name. Do not compare sleep opportunity with detected sleep as if they were the same.
- Look for missing recording time. Check battery gaps, charging, wear time and a loose fit. Apple specifically notes that battery loss stops tracking and that fit affects movement readings. Follow the manufacturer's setup instructions.
- Check the native app before an aggregator. If the native app has the night but another app does not, investigate permissions, source selection and last sync before concluding that the sensor failed. Preserve screenshots or exports of the mismatch.
- Compare the timeline. Find the interval responsible for the difference. Did one device begin earlier, finish later, include a nap or label more time awake?
- Then inspect score contributors. If durations broadly match, differences in consistency, timing or other inputs can still produce different scores. The manufacturers do not publish a universal conversion between them.
Garmin's sleep-tracking support page also distinguishes detected sleep data from manually entered sleep times. Entering a duration is not equivalent to recovering the sensor data needed to generate a sleep score. Check device-specific requirements before trying to repair a blank night by hand.
What validation research can actually tell us
A 2023 multicenter validation study recruited 75 adults with subjective sleep discomfort and compared 11 consumer systems with laboratory polysomnography. Its devices included Apple Watch 8, Oura Ring 3, Fitbit Sense 2 and Galaxy Watch 5. The number of usable observations differed by device; 75 was not the sample size for every comparison.
The investigators found different patterns of sleep-stage misclassification, including wearables classifying wake as light sleep. That helps explain how a tracker can produce a plausible-looking night while differing in particular intervals. It does not tell you which device correctly labeled your night.
There are two especially important limits: software was fixed after updates on March 1, 2023, and several authors disclosed financial ties to Asleep, whose product was included. This is useful evidence about those tested systems and that study population, not a current ranking of every model on sale.
When reading a new accuracy claim, look for the exact model, software, population, reference measurement and metric. “Accurate sleep tracking” could mean detecting sleep versus wake, estimating total minutes, or classifying stages. A result for one task does not settle the others. Nor does a good average establish a small error for every person or night.
What to do with the disagreement
For ordinary tracking, choose a consistent source for a particular trend and keep its definition visible. Avoid switching between whichever app gives the most reassuring score. If you switch devices or an algorithm changes, mark the change instead of treating the series as perfectly continuous.
Do not average two brands' scores and call the result more accurate. An average of 84 and 73 is 78.5, but there is no validated shared scale behind that arithmetic. Agreement between devices also cannot prove correctness: both could differ from an appropriate reference.
Repeated observations can help you describe a problem to a manufacturer's support team. They can show that one device repeatedly misses the start of your night, or that data appears in its own app hours before another service. They cannot, by themselves, determine your true sleep stages.
The AASM's 2018 position statement distinguishes consumer sleep information from medical evaluation. A general nightly score should not be used to diagnose a sleep disorder or dismiss symptoms. A product's separately regulated detection feature, where available, has its own instructions and limits; it should not be confused with the general nightly score.
Download the definitions and a blank comparison log
Score definitions CSV · Definitions JSON with sources · Blank night-by-night log.
The log records models, software, time zone, detected boundaries, time asleep, naps, score names, missing data and your own notes. Keep it private: it is for organizing your observations, and you do not need to send Vora your personal sleep data to use it. Several comparable nights are more informative about a recurring mismatch than a single screenshot, but there is no magic number of home observations that turns this into an accuracy trial.
This reference was checked manually on November 16, 2026. We will change its date when we substantively recheck it, rather than letting an automated date imply fresh research. Report a changed definition through support@askvora.com. Our editorial policy explains the sourcing and correction process.
Keep the context, not just the score
Publisher disclosure: we make Vora. Vora's proactive AI is built around connected health context, including wearables, workouts, nutrition and recovery. Bringing those together does not make every source interchangeable or repair a missing measurement. If you want that broader context, check Vora's supported integrations for your setup. The comparison worksheet above works independently of Vora.