A Garmin sleep score runs from 0 to 100 and combines how long you slept, how well you slept, and a stress reading taken from heart rate variability during the night. Garmin has not published how those parts are weighted, so nobody outside the company can tell you why a given night scored 68 rather than 74.
The bands, and what is typical
Garmin’s own blog gives the labels as excellent (90 to 100), good (80 to 89), fair (60 to 79) and poor (0 to 60). The 60 appears at the edge of both fair and poor in Garmin’s wording, so treat anything below 60 as poor.
The more useful number is where real users land. Garmin’s published data from February 2024 to January 2025 puts the average score at 72, up from 70 the year before. By band, 7% of users averaged excellent, 31% good, 43% fair and 19% poor.
So if your score is stuck in the 60s and 70s, you are looking at a label (“fair”) that describes the largest group of Garmin owners. An average of 90 is rare by Garmin’s own account. “Always low” often means “ordinary”.
Garmin’s blog also gives two examples that show the spread: 8.5 hours of good-quality sleep scoring 82, and 7 hours of poor-quality sleep scoring 49.
What goes into it
Garmin’s description of the score names these ingredients:
- Duration and quality. How long you slept and how well. Quantity and quality are compared against age-based standards that Garmin says were agreed with sleep experts.
- Sleep stages. Time in light, deep and REM, and the pattern between them. The watch estimates these from heart rate, HRV and body movement.
- Awake time and restlessness.
- Stress during sleep. An average stress score, based on HRV, that Garmin says shows how well your body is recovering.
- Age and baseline readings. Added as context to improve reliability.
- Respiration and blood oxygen, on devices that track them.
What the sources do not say matters as much. There is no table of weights, no HRV threshold, and no description of how age changes the targets. Anyone who tells you “deep sleep is worth 25% of the score” is guessing. Garmin’s own page says only that the score is built on how long and how well you slept.
Why a good night can score badly
Because the weights are hidden, you can only reason from the ingredients.
- Short sleep. The duration component is checked against an age-based standard. Garmin notes that sleep needs depend partly on age, generally 7 to 9 hours for adults.
- Stress during sleep. If the HRV-based stress average is high overnight, the score can drop even if you were in bed for eight hours. Alcohol, a late meal, illness or a hard evening workout are common reasons heart rate stays up overnight, and an elevated heart rate with lower HRV is what a stress reading picks up. Garmin’s manuals describe the stress measure as derived from HRV and say that training, sleep, nutrition and general life stress all affect it.
- Restlessness and awake time. A watch that is loose, or a partner or pet that moves you, adds restlessness.
- Stage estimates. Light, deep and REM come from a wrist sensor, not from a brain-wave recording. They are estimates, and a score that leans on them inherits the error.
- Wearing it to bed. If the watch is on the wrong side of a strap or is not worn all night, the stress reading has less to work with.
What to do about it
Change one thing at a time and watch the trend over two weeks rather than a single night. One bad night is a data point, not a verdict.
- Hold a steady bedtime and wake time. It is the one change that tends to help every ingredient at once.
- Move alcohol and big meals well before bed, and see whether the stress component settles.
- Wear the watch snugly enough that the sensor stays against the skin all night; the movement and heart rate readings both depend on it.
- Treat a score below about 60 for several nights, with daytime exhaustion or loud snoring, as something to mention to a doctor. A watch cannot diagnose a sleep disorder.
Reading it as a trend, not a grade
The sleep score is most useful against your own history. Open the seven-day and 30-day views in Garmin Connect and look for the usual range, then for the nights that fall well outside it. A single 55 after a late flight tells you nothing; four 55s in a row after you changed your routine tells you something.
It also helps to separate the two things the score is mixing. Duration is something you control with a bedtime. Quality, including the stress reading, is partly outside your control and partly a measurement artefact. If your score is low but you woke refreshed and your hours were fine, the quality side is probably the cause, and the quality side is the part with the least visible method. If you feel terrible and the score is a comfortable 80, believe the feeling. The score is an estimate from a wrist sensor and has no view into how you feel.
A published alternative
If you would prefer a score where every weight is visible, Pulse, a free app you host yourself, computes Sleep Performance from Google Health sleep data: hours against your own sleep need (50%), efficiency (20%), restorative sleep as deep plus REM (20%) and consistency over the last 7 days (10%). It scores 85 to 100% as optimal, 70 to 84% as sufficient and anything below as poor (how Sleep Performance works). Pulse scores the main sleep only, once Fitbit has processed it, and a night without deep and REM totals scores its restorative part as zero. That does not make it more accurate than Garmin’s. The stage data still comes from a wrist sensor, and Pulse is tested with the Fitbit Air only, not with Garmin watches. The difference is that you can check the arithmetic.
What is not known
Garmin does not publish the weights, the exact HRV measure, or how the thresholds change by age. The labels have a small overlap at 60. How the score handles naps is not stated in the pages read. If a score looks wrong, the Garmin Connect app shows the night’s stage chart and stress line, which is the nearest you can get to seeing what the algorithm saw.