What does the sleep score measure?

Duration, efficiency and consistency, scored against your own thirty-day baseline rather than a population norm. That last part matters: a score built against a population would tell a habitual seven-hour sleeper they are failing every night, which is not useful information about whether last night was good for them.

What is sleep efficiency?

Time asleep divided by time in bed. Someone in bed for eight hours who slept seven has 87.5% efficiency. It is the number that distinguishes a short night from a disturbed one, which matters because the two have different causes and different fixes.

How much deep sleep should I get?

The commonly cited healthy-adult range is roughly 13 to 23 percent of total sleep, and Base rates each stage against its own reference band. Those are consensus ranges from sleep-medicine summaries rather than a diagnosis, and one night below band is noise rather than a verdict.

Why does my sleep data show nothing some nights?

Because the watch was not worn or was charging. Base states that there is no data rather than rendering a zero, which matters more than it sounds — a zero would be averaged into your thirty-day baseline and would quietly deflate every recovery score for the following month.

Does Base need sleep data to work?

No. Sleep feeds the recovery score and the readiness banner, and without it those state that they cannot be computed. Food, weight, trend weight, adaptive maintenance and all of the training features work without any wearable at all.

Read Your Sleep Score and Stages in Base

The headline numbers

Number	What it is
Score	Duration, efficiency and consistency against your own 30-day baseline
Asleep	Actual sleep time
In bed	Total time in bed
Efficiency	Asleep ÷ in bed
vs 30-day average	Whether last night was good for you

Efficiency is the one people overlook and the one that diagnoses. Eight hours in bed and six asleep is 75% — a disturbed night, not a short one. Six hours in bed and 5.8 asleep is 97% — a short night, not a disturbed one. Those have different causes and different fixes, and total sleep time alone cannot tell them apart.

in bed 8h 10m  = 490 min
asleep 6h 55m  = 415 min
efficiency     = 415 ÷ 490 = 84.7%

Why your own baseline

A score built against a population reference would tell a habitual seven-hour sleeper that they fail every night, which is not information about whether last night was good.

The useful question is whether this night was better or worse than your normal, because that is the one that predicts how today will go. Every score in Base's health section works this way, which is also why they take a couple of weeks to become meaningful.

The stages, and what each is for

One night with the trend behind it. Each stage is rated against its own reference band, and every rating opens an explanation of what that stage does and where the band came from.

Deep sleep (slow-wave) is when the body does most of its physical repair — tissue and muscle protein synthesis, immune consolidation — and when growth hormone release peaks. For anyone training hard this is the stage that most directly maps to recovery.

REM is when the brain does memory consolidation and emotional processing. It also carries a disproportionate share of the night's testosterone release, which is concentrated in the back half of sleep where REM periods lengthen. That is one mechanism by which cutting a night short at the end is worse than starting it late.

Core (light) sleep is the connective tissue between the two. Necessary, but not where the night's work gets done.

Awake has no target band and is rated separately by absolute minutes.

The reference bands are commonly cited healthy-adult ranges from sleep-medicine consensus summaries, computed as a percentage of total time asleep with awake excluded from the denominator. They are not a diagnosis, and one night below band is noise.

None of this is unique to you — it is the standard account of the stages rather than a personalised measurement.

What to actually do about it

Honestly: stage percentages are not very actionable. You cannot decide to get more deep sleep.

What is actionable is total sleep time and consistency, and those two drive the stages. Sleep longer and you get more of everything, with REM gaining disproportionately because it lengthens through the night.

So the useful reading is:

Total sleep, against your own normal
Efficiency, to distinguish short from disturbed
Consistency of timing across the week
Stages, as context rather than a target

Missing nights

A night with no data says so and shows nothing.

This is not a cosmetic choice. A zero would be averaged into the thirty-day baseline, deflating it, and every recovery score for the following month would be computed against a corrupted reference. A missing night is unknown, and unknown is not zero.

Feeding recovery

Sleep is one of the three inputs to the recovery score, alongside heart-rate variability and resting heart rate, and it carries the smallest weight of the three. Poor sleep also contributes a fatigue signal to the readiness banner.

Common mistakes

Chasing deep sleep percentage. It is not directly controllable. Total sleep is.

Reading one night as a trend. The bands are population ranges and the score is against your own average. One night is noise in both frames.

Ignoring efficiency. It is what separates a short night from a broken one.

Treating a missing night as zero. Base does not, and neither should you.

Comparing your stage split to someone else's. Individual variation is large and mostly not meaningful.

Assuming the watch is precise about stages. Wrist-based staging is inferred rather than measured. Directionally useful, not polysomnography.

What to do next

Look at your last fortnight's efficiency rather than your total sleep. If it is consistently under about 85%, the problem is sleep quality rather than time in bed — and that is a much more tractable thing to work on.