Sleep timing

The short version

How regular your sleep is, not only how long it lasts, is linked with survival. In 60,977 UK Biobank adults who wore wrist sensors, how regular their sleep was predicted their risk of dying better than how long they slept.

But “regular” hides four different questions, and each needs a different amount of data before it can answer:

• Bedtime average: where the middle of your bedtimes sits.
• Drift: whether that middle is moving.
• Spread: how much your bedtime moves, measured two ways.
• Sleep Regularity Index: whether your whole day, asleep and awake, repeats.

Four diaries

Picture four people writing down the time they fall asleep, every night for two weeks.

• Person A is steady. Asleep within about ten minutes of 23:00, every night.
• Person B slides. Five nights around 23:00, then each night lands 12 to 15 minutes after the last. No single night looks unusual.
• Person C is scattered. 23:00, 02:00, 22:00, 01:00, 23:30. Late and early, with no direction.
• Person D sleeps Person A’s nights to the minute, but naps for two hours on some afternoons and not others.

Here is what each measure makes of them:

No single measure catches all three. Only drift catches Person B’s slide. Spread and the regularity index both catch Person C. Only the regularity index catches Person D. The average alone flags no one.

1. Bedtime average: where the middle is

It answers what time do I usually fall asleep?

Clock times sit on a circle, so they have to be averaged as angles. With plain arithmetic, 23:50 and 00:10 are 23 hours 40 minutes apart and average to noon. On a clock face they are 20 minutes apart and average to midnight.

The average settles fast. In 10,412 adults who wore a WHOOP band for a year, three nights gave a reliable average time of falling asleep, and seven a reliable average total sleep time (Leota, Sleep 2026).

It can’t tell you how consistent you are. Person C’s average, 23:47, looks perfectly ordinary.

2. Drift: whether the middle is moving

Compare one week’s average with the next. Person B’s moved from 23:08 to 00:20: an hour and 12 minutes later in a single week, though no night was more than 15 minutes from the one before.

Drift is the only one of the four with a direction. It needs two windows to compare, so two weeks of nights.

3. Spread: how much your bedtime moves

It answers how much does my bedtime bounce around? Verve measures it two ways, because each one sees something the other doesn’t.

Night to night is the average size of the move from one night to the next, in minutes, whichever way it goes: 8 for Person A, 11 for Person B, 2 hours 30 for Person C.

Around your average is the standard deviation: how widely your bedtimes scatter around their own average. It is one of the measures of regularity listed in the American Heart Association’s 2025 statement on sleep health, which gives 60 minutes or less as a commonly used cut-point while noting there is no agreement on the best threshold (St-Onge, Circ Cardiovasc Qual Outcomes 2025). Person A reads ±5 minutes and Person C ±83.

Read the two together. Person B’s nightly moves are small, 11 minutes, but around the average Person B reads ±41, because measured from the middle the slide looks like scatter. When around-your-average is much wider than night to night, the gap is drift.

Both are slow to trust. In the same 10,412 adults, a reliable standard deviation took 41 to 65 nights depending on the measure, and 44 for the time of falling asleep. Estimates from 7 to 14 nights, the length of most sleep studies, were poor, and a single week could misjudge the variability of total sleep time by as much as 50 minutes either way (Leota, Sleep 2026). No study has measured this for night-to-night moves, so Verve applies the same 44 nights to both.

4. Sleep Regularity Index: whether your whole day repeats

It answers am I asleep and awake at the same clock minutes today as yesterday? It checks every minute of the 24 hours against the same minute a day earlier, so waking up, naps and time awake in the night all count, not just bedtime. 100 is a day that repeats exactly. 0 is no better than chance.

That is why Person D scores 80 to Person A’s 91 despite identical nights. It is also why Person B scores 91: a slow slide changes each day only a little from the day before, and the index compares only neighbouring days (Fischer, Sleep 2021).

This is the measure the outcome data rest on. Across 60,977 UK Biobank adults followed for an average of 6.3 years, the four most regular fifths had a 20 to 48 per cent lower risk of death from any cause than the least regular fifth, and regularity predicted it better than duration did (Windred, Sleep 2024). A second analysis, of 88,975 participants, found the risk concentrated at the irregular end: a hazard ratio of 1.53 at the 5th percentile against 0.90 at the 95th, both relative to the median (Cribb, eLife 2023). The extra risk sat mostly in the irregular tail.

Why Verve doesn’t put your score on that scale.Those two medians come from the same biobank. When two widely used SRI calculators were run over the same 73,794 adults, their scores differed enough to change what the outcome models concluded (Czeisler,Sleep 2026). A score only compares with scores from the same calculation, so Verve reads yours against your own past months, and the study figures stay here as context.

It needs time. It takes at least five to seven days of continuous sleep data to calculate. Related regularity scores built from fewer than seven days carry measurable bias and don’t compare with longer recordings (Lok, J Sleep Res 2025). How long one person’s score takes to settle has no agreed answer yet.

How many nights each one needs

So the average can be trusted within days, and drift within a fortnight. Spread and the regularity index stay provisional much longer. Verve counts both as settled at 44 nights, about six weeks; for the regularity index, where there is no agreed figure, that is Verve’s own choice.

What these measures are not

• Not hours. A steady six-hour sleeper can look more regular than an erratic eight-hour one. Timing and duration are separate markers, each linked with health in its own right.
• Not sleep quality. The regularity index sees only asleep or awake, not how deep the sleep was.
• Not travel-proof. They compare clock times, so crossing time zones reads as irregular even when your body’s rhythm is steady.

In Verve

Verve reads the sleep your watch or ring writes to Apple Health and works out all four measures the way this page describes.

• The Sleep page keeps to numbers, under three headings. Bedtime average with drift: your average bedtime over the last seven nights, and how far it moved from the seven before, in amber once the move passes 20 minutes. Bedtime spread: night to night, and around your average, in green or amber against the 60-minute mark. Sleep regularity index (SRI): the last 30 days, beside your own median. Until 44 nights are in, spread and the index stay grey with a count toward 44.
• Trends shows the same three cards. The Snapshot sets this week’s average bedtime against last week’s, with spread and the index as on the Sleep page. Over Time shows all three for the period you pick, each with a thin line of how its number moved: night by night, week by week, month by month or year by year. On the spread and index lines, points resting on fewer than 44 nights are hollow; on the months view, the index line also carries your median for the year.