Night shifts and sleep trackers: why most apps misread shift workers

What shift work does to sleep

Sleep after a night shift is usually shorter than night sleep, often by one to four hours (Åkerstedt 2003). The body clock is the reason: it keeps signalling day while you try to sleep. Very few people fully adapt, even on permanent nights. In Folkard's review of melatonin studies, fewer than 3 % of permanent night workers showed full adjustment of the body clock (Folkard 2008). Rotating schedules add another layer: the clock is pulled one way, then the other, every few days (Boivin and Boudreau 2014).

For a shift worker the useful questions are not "did you go to bed at 10?" but "did you get your main sleep, whenever it was?" and "how is this rotation treating you over weeks?"

Where sleep apps go wrong

• Daytime sleep filed as a nap. Six hours from 9:30 to 15:30 after a night shift is your night. An app that looks only at the evening and the small hours sees almost no sleep and a long nap.
• A sleep debt you do not have. Once the main sleep is missing, every after-night day looks like a short night.
• Averages that fit no day. Average a 4:45 wake before a morning shift and a 14:00 wake after a night shift and you get a time you never wake up. Bedtime advice built on that average fits none of your days.
• Training suggestions in the middle of your night. A "best time to train" counted from yesterday's bedtime can land at 1 am on the day after a night shift.
• Months to notice a new job. Long averages keep your old schedule alive long after you have moved to days.

What to look for if you work shifts

1. The main sleep read from when you slept, not from the clock.
2. Your rotation learned from your own nights: two mornings, two nights, three off, or a short and a long week.
3. Quick adaptation when the pattern changes.
4. Predicted times labeled as estimates, not shown as measurements.
5. No forms to fill in every week.

How MyBodyAI reads shift workers

• Your sleep counts as your night. On a day when you did not sleep between 21:00 and 9:00, your daytime sleep is your night, so you are not handed a sleep debt for sleep you had.
• Several kinds of day. From your wake times the app finds your day types, for example a day-shift wake, a free-day wake and an after-night wake.
• The cycle behind them. It then looks for the period, from 1 to 28 days, that best predicts which kind of day comes next: an ordinary week, a 14-day short and long week, or a rotation. When you change jobs, four nights on a clock three or more hours away (a week for a smaller shift) are enough for the app to start learning the new schedule fresh instead of averaging it with the old one; within it, recent nights weigh more (a half-life of 21 days).
• Training windows from what you do. Your usual training time is the densest window of your real sessions, not an average of morning and evening workouts that lands at noon.
• No settings needed. You can still enter shifts in the day plan, but you do not have to. With too few nights or no repeating pattern, the app says it is estimating and uses plain averages.

The app needs a few weeks of nights to learn a cycle with confidence: at least eight nights and two full rounds of the pattern.

References

• Åkerstedt T (2003). Shift work and disturbed sleep/wakefulness. Occupational Medicine, 53(2), 89 to 94.
• Folkard S (2008). Do permanent night workers show circadian adjustment? A review based on the endogenous melatonin rhythm. Chronobiology International, 25(2), 215 to 224.
• Boivin DB, Boudreau P (2014). Impacts of shift work on sleep and circadian rhythms. Pathologie Biologie, 62(5), 292 to 301.