Methodology

How we measured 38,114 nights.

A consumer sleep ring, a two-line morning diary and a 31-question survey, asked twice. Here is exactly what we did, who we missed, and how to use the data yourself.

Methodology
This report draws from
1,412Volunteers who finished the study
38,114Nights recorded by the rings
31Survey questions, asked twice
12Countries, from Norway to Chile

Five things to know before you quote a number.

Rings are good at when you sleep and how often you wake. They are less sure about sleep stages, so we report stages only as shares of a night.

Method notes

Every volunteer wore the same consumer sleep ring on the index finger for 27 consecutive nights between 13 January and 30 March 2026. The ring estimates sleep onset, wake-ups and stages from movement, heart rate and skin temperature. We counted a night when the ring recorded at least three hours in bed.

1,812 people applied through partner clinics, universities and employers; 1,412 completed at least 20 nights. Night-shift workers and new parents are over-represented on purpose. People without a smartphone, and anyone under 18, could not take part, so the report says nothing about them.

31 questions, asked in week one and again in week four: bedtime habits, screens, caffeine, work hours, how rested people felt and what they believed about their own sleep. The survey vs ring comparison uses the week-one answers.

Rings were registered to a random ID, never to a name. Diaries were written in the same app and stripped of place names. The published dataset rounds bedtimes to five minutes and ages to the year. The six diarists in this report agreed to be quoted and chose their own pseudonyms.

Stillhour Institute (2026). The Night Report 2026: how 1,412 people slept, scrolled and woke. Figures may be quoted with credit and a link to this page.

The dataset

Swap in your own data

Every chart in this report is drawn from one CSV with a row per respondent. Upload yours to a chart’s “Dataset (CSV)” control and the dots, axes, medians and every {token} in the copy recompute.

Download the sample CSV ↓
ColumnStatusWhat it holdsExample
idoptionalAny unique id412
grouprequiredCategory label. Colours follow the order groups first appearNight-shift workers
labeloptionalWhat a respondent is, shown on hoverICU nurse
ageoptionalNumber41
arequiredFirst measure: a number, or a clock time HH:MM (before noon counts as after midnight)23:40
brequiredSecond measure: a number6.5
flagoptionalYes/no split: yes, y, true or 1yes
idgrouplabelageabflag
500FoundersTrader3601:346.06yes
980Office workersDesigner2722:575.63yes
430FoundersProduct lead4000:086.94yes
1002Office workersDesigner2922:597.80no
12New parentsNew parent2923:015.85yes
993Office workersDesigner2822:486.66no
549StudentsUndergraduate2101:304.28yes
969Office workersEngineer4800:516.91yes
First 8 of 1,412 rows · built-in demo data

Check our working

The same data, hours asleep for every volunteer.

Median night per volunteer. The dashed line is seven hours.

How to cite

Stillhour Institute (2026). The Night Report 2026: how 1,412 people slept, scrolled and woke.

Figures may be quoted with credit and a link. For corrections, write to data@stillhour.org; we publish every correction with a date on the About page.

Create a free website with Framer, the website builder loved by startups, designers and agencies.