What Is a Good Sleep Score?

Last updated: September 3, 2026

There is no such thing as a good sleep score in general, only a good score on your specific device, and the gap between devices is larger than most people realise. A score of 80 is "Good" on Google Health, "Good" on Oura, and only "OK" on an Apple Watch. Garmin's own published data puts the average user at 71, with just 5% averaging in the 90 to 100 band. Google states plainly that most of its users sit between 72 and 83, which falls inside Google's own "Fair" tier. So if your score feels stuck in the seventies, you are not sleeping badly. You are sleeping normally, on a scale that was calibrated to make the seventies look mediocre.

Want to see exactly which metrics your device feeds into that number and which it ignores? The sleep metrics comparison tool lays all four brands side by side.

I went looking for a straight answer to this for my own data and found that the four major brands do not agree on the bands, do not agree on the inputs, and in one case changed the bands after launch without the underlying measurement changing at all. That last part is the most useful thing in this post.

The bands, side by side

Every one of these is published by the manufacturer. Note that they do not line up.

Device	Top tier	Next tier	Middle	Bottom
Oura	Optimal 85-100	Good 70-84	Fair 60-69	Pay Attention 0-59
Google Health / Fitbit	Excellent 90-100	Good 80-89	Fair 60-79	Poor under 60
Apple Watch	Very High 96-100	High 81-95	OK 61-80	Low 41-60, Very Low 0-40
Garmin	Excellent 90-100	Good	Fair (user average is 71)	Poor

Read the 80 row across. On Google Health an 80 is the bottom of "Good." On Oura an 80 is comfortably inside "Good." On an Apple Watch an 80 is the top of "OK", one point short of "High." Same number, three different verdicts, and none of them wrong, because each brand drew its own lines.

Garmin does not publish a full band table, but it does publish something more useful, which is covered further down.

Why an 80 is not an 80

The reason the bands do not line up is that the scores are not measuring the same thing. They share a name and a 0-to-100 range and almost nothing else.

Device	What actually goes into the score
Oura	Seven contributors: Total Sleep, Efficiency, Restfulness, REM Sleep, Deep Sleep, Latency, Timing
Google Health / Fitbit	Five metrics: Sleep duration, Time to sound sleep, Sound sleep, Restlessness, Full awakenings
Apple Watch	Three weighted components: sleep duration 50 points, bedtime consistency 30 points, interruptions 20 points
Garmin	Duration plus time in light, deep and REM, the patterns between stages, awake time, restlessness, and average stress score during sleep
WHOOP	Not a 0-100 quality score at all. WHOOP works from Sleep Need, Sleep Debt, Sleep Consistency and Sleep Efficiency

Oura scores you on how much REM and deep sleep you got. Garmin folds in your overnight stress. Apple counts neither. WHOOP does not produce a comparable number in the first place, because it measures what you got against what it calculated you needed, which is a different question entirely.

If two devices on the same wrist and finger disagree by fifteen points, that is not a malfunction. They were asked different questions.

The single biggest source of cross-device disagreement

Apple's Sleep Score contains no sleep-stage input whatsoever. It is duration, bedtime consistency and interruptions. That is the whole formula, and Apple publishes the weights.

Oura's score contains two stage contributors out of seven. Garmin's contains three stage-related inputs plus stage transition patterns.

So consider a night where you slept seven and a half hours, at your usual bedtime, without waking, but with unusually little deep sleep. Apple sees a near-perfect night, because nothing it measures went wrong. Oura and Garmin both mark you down, because the thing that went wrong is exactly what they weight. Neither device is broken and neither is lying. The Apple score genuinely does not know.

The reverse case is just as common and catches more people out. Bedtime consistency is 30% of the Apple score, and it looks at your last 13 nights. A physiologically excellent night at an unusual hour gets capped on Apple and passes cleanly on Oura, whose Timing contributor is one of seven rather than nearly a third of the total. If your Apple score is stubbornly stuck in the OK band despite good sleep, an irregular schedule is the first thing to check, and it is a factor no amount of deep sleep will offset.

For the full metric-by-metric breakdown of what each device tracks and what it misses entirely, we compared every sleep metric your wearable tracks across all four brands.

The bands are an editorial decision, not a measurement

This is the part that should change how you read your score, and there is a clean documented example of it.

In watchOS 26.2, released November 2025, Apple changed the Sleep Score bands. The sensors did not change. The formula did not change. The 50/30/20 weighting did not change. Apple simply redrew the lines and renamed the tiers, and the stated reason was to better match how people might be feeling after a restless night. It also renamed its top tier from "Excellent" to "Very High" for consistency with its other labels.

The same 78 that was one label on a Tuesday was a different label after the update.

That is worth sitting with, because it means the word attached to your number is a product decision about how a company wants you to feel, layered on top of an estimate. The number itself has some information in it. The judgement attached to it is editorial.

What the average person actually scores

This is the number almost nobody publishes, and it is the one that actually answers the question.

Garmin, from its own user data: the average sleep score in 2024 was 71, which Garmin itself labels "fair." Only 5% of Garmin sleep users averaged in the "excellent" 90 to 100 range.

Google, from its own help documentation: "Most users see an average Sleep Score between 72 and 83." Google's own "Good" tier starts at 80. So the majority of Google Health users sit largely inside the "Fair" band.

Put those together and a picture emerges that no marketing page will show you. On two of the four major platforms, the typical user is told their sleep is fair, most nights, indefinitely. A score in the seventies is not a problem to be solved. It is the middle of the distribution.

That reframes the whole question. If you are chasing 90, you are chasing something 95% of Garmin users do not reach on average, on a scale where the median is 71.

The number underneath is only moderately accurate anyway

Before optimising a composite, it is worth knowing how good its inputs are. For the stage-based scores, the honest answer is moderate at best.

Measured against polysomnography, the clinical reference standard, four-stage sleep staging agreement in healthy adults runs from Cohen's kappa 0.21 to 0.65 depending on the device. Human sleep technicians scoring the same recordings agree with each other at roughly 0.75. No consumer device is accurate in an absolute sense, and none of them reach the human ceiling.

Accuracy also drops sharply in the people most likely to be worried about their sleep. The same devices score far worse in older adults and in genuine sleep-disorder populations than in the screened healthy young cohorts most validation studies use. Apple Watch four-stage agreement falls from kappa 0.53 in healthy adults to 0.30 in a sleep-apnea population. Oura falls from 0.65 to 0.35. Fitbit is the most population-robust of the group, holding around 0.42 in both.

So a stage-weighted score like Oura's or Garmin's is a weighted composite of estimates that are themselves moderately reliable. That is not an argument for ignoring it. It is an argument for reading the trend rather than the digit. Our most accurate sleep tracker comparison has the per-device breakdown, and the sleep tracker accuracy tool lets you check accuracy by stage.

So what should you actually look at

The score is a summary, and summaries hide the thing you can act on. Three shifts make it more useful.

Read your own baseline, not the label. Pull two or three weeks of scores and find your personal median. That number, not 85 and not 90, is your normal. A night five points under your median is a real signal. A night that is merely under the manufacturer's "Good" threshold may just be Tuesday.

Open the contributors instead of the total. Every platform except Apple exposes the components. A 74 driven by low efficiency is a completely different problem from a 74 driven by short duration, and they have opposite fixes. The composite deliberately blurs that distinction.

Compare like with like. Never benchmark your score against a friend's on a different device, and be careful even on the same device across a major software update. Apple's rebanding is the documented case, but any platform can redraw its lines.

Once you have your own baseline, the useful question stops being "is this a good score" and becomes "what did I do differently on my bad nights." That is a question about inputs the wearable never sees: what you ate, when you ate it, what you drank, how late. Kygo pulls your Oura, Apple Health, Garmin, Fitbit and WHOOP data in alongside a food log that takes 20 to 30 seconds by voice, photo or plain description, then runs the correlation over 12 to 36 hour windows so the pattern surfaces instead of you trying to remember it. Get it on iOS or Android.

If your score is low and you want the causes rather than the benchmark, we cover why your sleep score is low even after eight hours separately, ranked by evidence.

Common questions

What is a good sleep score?

It depends entirely on your device, because the bands differ. On Oura, 85 and above is Optimal and 70 to 84 is Good. On Google Health and Fitbit, 90 and above is Excellent and 80 to 89 is Good. On an Apple Watch, 81 to 95 is High and 61 to 80 is only OK. For context, Garmin's average user scores 71 and most Google Health users average between 72 and 83.

Is a sleep score of 80 good?

On Google Health and on Oura, yes, an 80 sits in the Good band. On an Apple Watch an 80 is the top of the OK band, one point below High. On Garmin an 80 is well above the 71 user average. The number is the same and the verdict is not, which is why comparing scores across devices does not work.

Is 71 a bad sleep score?

No. Garmin published that 71 was the average score across its users in 2024, and Google reports most of its users averaging 72 to 83. A score in the low seventies is the middle of the distribution on the two platforms that publish their data, even though both label that range "fair."

Why is my Apple Watch sleep score different from my Oura score?

Because they measure different things. Apple's score is duration (50 points), bedtime consistency (30 points) and interruptions (20 points), with no sleep-stage input at all. Oura's uses seven contributors including REM sleep and deep sleep. A night with poor deep sleep at a normal bedtime scores well on Apple and poorly on Oura, and both are behaving correctly.

What is a good Oura sleep score?

Oura defines 85 to 100 as Optimal, 70 to 84 as Good, 60 to 69 as Fair, and 0 to 59 as Pay Attention. Oura's own guidance is that 85 or higher suggests you are ready to take on new challenges, and that scores below 70 suggest prioritising rest and recovery.

Can you get a 100 sleep score?

It is possible but rare, and it requires maxing every contributor at once, which on Oura means seven of them including both stage metrics. On Garmin, only 5% of users average even in the 90 to 100 band. Treat 100 as a curiosity rather than a target.

Are sleep scores accurate?

The composite is only as good as its inputs, and for stage-based scores those are moderate. Four-stage agreement against clinical polysomnography ranges from kappa 0.21 to 0.65 across devices in healthy adults, while human scorers agree at about 0.75. Accuracy drops further in older adults and in people with sleep disorders. Read the trend, not the individual digit.

How do I improve my sleep score?

Start by opening the contributors rather than the total, since a score dragged down by efficiency needs a different fix from one dragged down by duration. On Apple specifically, bedtime consistency is 30% of the score and is measured across your last 13 nights, so a regular schedule moves it more than anything else. Beyond that, the inputs your wearable cannot see (alcohol, late meals, caffeine timing) are usually where the answer is.

The bottom line

A good sleep score is one that is good for you, on your device. The bands are drawn by manufacturers, they do not agree with each other, they contain different inputs, and at least one of them has been redrawn after launch with no change to the underlying measurement. Meanwhile the two platforms that publish user data both show the typical person parked in the seventies, in a tier labelled "fair."

Find your own median, watch the contributors rather than the composite, and stop comparing your number to anyone else's. The score is a decent trend line and a poor verdict.

Ready to find out what is actually moving your number? Kygo connects your food log to your wearable data and surfaces the correlations instead of leaving you guessing. Get it on iOS or Android.

---

Key sources

Manufacturer documentation, all read September 2026: Oura's Sleep Score page and sleep score blog post for the seven contributors and the 85/70/60 bands; Google Health's Sleep Score help page for the five component metrics, the Excellent/Good/Fair/Poor bands and the "most users see an average Sleep Score between 72 and 83" figure; Apple's Sleep Score support page for the 50/30/20 weighting, the 13-night consistency window and the current five bands; Garmin's sleep score explainer for the Firstbeat inputs, the 71 user average and the 5% figure for the excellent band; WHOOP's sleep guide for Sleep Need, Sleep Debt and Sleep Consistency. Staging accuracy figures are from Schyvens et al. and Robbins et al. for healthy adults and Lee et al. 2023 for the clinical population, compiled in Kygo's sleep tracker accuracy research. The watchOS 26.2 band change was reported by MacRumors on 4 November 2025.

---

A note on what this post leaves out

Several outlets published Apple's pre-watchOS-26.2 bands as a numbered list, and the version in circulation is internally inconsistent, pairing the lowest score range with the most positive label. We could not reconcile it against an Apple source, so the old Apple bands are deliberately not stated here. What is stated is only the change itself and Apple's reason for it, both of which are documented, plus the current bands, which come from Apple directly.

We have also not assigned a numeric band table to Garmin beyond the Excellent 90 to 100 range, because Garmin does not publish full band boundaries the way the other three do. The 71 average and the 5% figure are Garmin's own published user data and are more informative than an inferred table would be.

---

Disclaimer: Kygo is a personal data aggregation and insights platform designed for informational purposes only. The information provided by Kygo, including correlations, patterns, and trends identified in your data, does not constitute medical advice, diagnosis, or treatment. Always consult a licensed healthcare provider with any questions regarding medical conditions.

If you wear two devices and have watched them disagree on the same night, I would like to know how wide the gap runs and which one matches how you actually felt.