Too short or too long: sleep data from 80,000 workers in Japan

What we know

Most sleep research has relied on surveys that ask a few hundred people, “How many hours do you usually sleep?” It is surprisingly hard to answer that accurately, and the samples are small.

A team led by Masashi Yanagisawa at the University of Tsukuba took a different route (2025, npj Digital Medicine). They used data from sleep-tracking smartphone apps: about 80,000 workers in Japan and more than 2 million nights of records. Every night, the apps logged how long people slept, how long it took them to fall asleep, how much of the night they spent awake, and how much their bedtime shifted between weekdays and weekends. The researchers matched this with a questionnaire measure of presenteeism: being at work but not working at full capacity because you feel unwell. It is a common way to measure lost productivity.

Three findings stand out.

1. Sleep length and presenteeism had a U-shaped relationship. People who slept very little and people who slept a lot were both on the high-loss side.
2. Taking long to fall asleep, waking often in the night, and a big gap between weekday and weekend sleep times were all linked to lower performance.
3. When the team used AI to group sleep patterns, five types appeared: healthy sleepers, long sleepers, fragmented sleepers, insomnia-prone sleepers and “social jetlaggers”. Productivity loss was greatest in the social jetlag and insomnia-prone groups, for both men and women.

“Social jetlag” is what happens when you get up early on weekdays and sleep in on weekends. Your body clock shifts every weekend as if you had flown across time zones.

How to read it

This is a correlational study that compares people at one point in time. It cannot tell whether long sleep lowers productivity or whether people who feel unwell simply sleep longer. Illness or built-up fatigue could explain part of the link.

The participants were also people who chose to use a sleep app, so they may not represent all workers in Japan. Presenteeism was self-reported. It is not a manager’s rating, and it is not income.

Still, a Japanese dataset this large showing that “too short and too long are both linked to worse outcomes”, and that timing matters as well as length, is worth taking seriously. RAND’s five-country report points the same way at the short end: people sleeping under six hours had a 13% higher risk of death than those sleeping seven to nine hours.

Why the app says “about 7 hours”

UPSHIFT’s sleep habit used to read “slept 7 hours or more”. But “or more” suggests that nine or ten hours would be even better. This Japanese study linked very long sleep to lower productivity too, so we changed the habit to “about 7 hours”, a middle target that avoids both extremes.

In the published materials, the study does not name a single best number of hours. Seven is not an exact right answer. It is a rough middle point, and the sleep each person needs varies.

A small step for today

• Log “slept about 7 hours”. You will start to notice both the short nights and the very long ones.
• Try to keep your weekend wake-up time within an hour of your weekday time. It is one easy way to shrink social jetlag.
• Put your phone down an hour before bed. That removes one thing that can delay falling asleep.

If you still feel tired after long nights, or you are very sleepy during the day, please talk to a doctor.

Sources

• [B3] Seol, Iwagami & Yanagisawa (2025) Association of Sleep Patterns Assessed by a Smartphone Application with Work Productivity Loss Among Japanese Employees. npj Digital Medicine.
• [B2] Hafner et al. (2016) Why sleep matters — the economic costs of insufficient sleep: A cross-country comparative analysis. RAND Corporation.