Your sleep score sums up the night that ended this morning, out of 100, from four weighted parts: duration, consistency, restfulness and stages. Each part shows its working, so you can always see what moved the number. A night files under the day you woke up - the Apple Health convention - so today's score describes the night that set up today.
The four parts
Duration carries much the most weight, close to half the score. Wearables measure time asleep well, and it's the thing that matters most. It's scored as how close you came to a full night's need - full marks arrive a little under eight hours, and sleeping longer than your need is never penalised, because a long night is usually debt being repaid.
Consistency compares last night's bedtime with your own median over the previous fortnight. A quarter of an hour either side of your usual costs nothing; beyond that, going to bed later is penalised gradually, while turning in early barely registers. An early night is rarely a problem.
Restfulness is how undisturbed the night was: mostly sleep efficiency, the share of your time in bed you were actually asleep, plus a penalty for a fragmented night. The first few awakenings are free - everyone surfaces briefly, usually without remembering it.
Stages is the smallest part. More on that below.
Why stages count for so little
Wrist and finger devices estimate deep and REM sleep, they don't measure it. Against a sleep lab, consumer wearables identify deep sleep correctly roughly half the time. Healthy stage architecture also varies widely between people.
So tonight's deep and REM are compared only with your own typical levels, never a population target - you won't be told you're "short of the deep sleep you need" - and only downwards: more than your usual is never penalised. Light sleep isn't scored at all.
A night well below your own usual is worth noticing. A night below someone's idea of a healthy percentage is close to meaningless.
When parts are missing
Some nights won't have all four parts. Stages need about five nights of your history before your baseline means anything, consistency needs three prior bedtimes, and some devices don't report stages or efficiency at all.
NOYU drops whatever is missing and rebalances the remaining weights, so the score is built from what your device reported. Naps appear in your data but don't feed the score - it describes your main night.
What to do with it
Read the week, not the night. One poor score after a late dinner or a hard session is Tuesday; a fortnight of scores drifting down, usually led by the duration part, is a pattern worth acting on.
And don't chase the number. The score exists to point you at whichever part is costing you - a bedtime that's drifted, a fragmented stretch in the small hours - so you know where to look.
