Every rule behind your plan, and where it comes from.

Your temperature minimum sits about two and three quarter hours before your natural wake time

What we do
We take your no-alarm wake time and subtract 165 minutes, then show the result as a window rather than a single minute. Everything else in the plan is positioned relative to that point.

What the research says
Measured, not assumed. Lee, Smith and Eastman held subjects on a 23:00 to 07:00 sleep schedule for three weeks and found a temperature minimum of 04:24 in one group and 04:00 in the other, so 2.6 and 3.0 hours before waking. Baehr, Revelle and Eastman tracked 172 adults over six days and found the minimum near the middle of an 8-hour sleep, at 03:50 for morning types and 06:01 for evening types, and state it can be estimated from a person's weekend wake time. That is what this asks you for.

Our judgement, not a published constant
This was two hours flat until we went looking for the source, and it had none. It is now 2 hours 45, and it is still an estimate of a typical reader rather than a measurement of you. The two groups above differed by 24 minutes and individuals within them by nearer an hour, and Baehr found the minimum falls about half an hour later in men than women. That is why your plan shows a window rather than a single minute, and why the nudge control is there: if it does not match when you actually hit the wall, move it and everything else re-times with it.

Light before that point delays your clock. Light after it advances your clock

What we do
This is why the bright-light window is placed by the direction your plan moves your clock, not by what your roster is called. A plan moving you later puts it before your minimum, which on nights is usually early in the shift. A plan moving you earlier puts it after, which on an early start is usually after your alarm rather than at it. The light-block window then covers whichever side would undo that, and on nights that is the drive home. Get them the wrong way round and you push your clock backwards.

What the research says
The photic phase response curve: light in the late subjective day or early subjective night produces delays, light in the late subjective night or early subjective day produces advances.

Every light window stops short of your low point, by exactly the width of our own uncertainty about it

What we do
Your low point is an estimate, and the plan shows it as a window rather than a minute because that is all the estimate supports. A light window aimed at the middle of that window is therefore on the wrong side of your real low point about half the time, and there is no soft landing at the crossover: a little before it, light delays you; a little after it, light advances you. So a window that has to sit before your low point ends at the earliest that point could be, and a window that has to sit after it starts at the latest it could be. Whatever you do inside the window is then on the intended side of your real low point, wherever inside our estimate it actually falls.

What the research says
The margin is not a number we picked. It is the same half-hour that sets the width of the trough window shown on your plan, taken from the spread between the two Lee groups. Re-source the anchor and every light window moves with it, because the margin is that spread rather than a constant of its own. It replaced a hardcoded two hours in the night-by-night rows and a margin of zero in the main window: two different numbers for one idea, neither tied to the anchor.

Our judgement, not a published constant
This buys direction, not precision. It keeps light on the correct side of your low point across the range our estimate can support; it does not make the estimate itself sharper, and the spread between individuals is wider than the spread in our estimate of a typical reader. The nudge control is still the thing to use if the timings do not match when you actually hit the wall.

If you cannot say when you would wake up with no alarm, five published questions can place you instead

What we do
Everything below hangs on where your clock naturally sits, and we used to ask that one way: what time would you get up if you were entirely free? That is a fair question for a day worker and close to unanswerable if you have been on nights for years, because your honest answer has already been moved by the job. Get it wrong and every window in your plan is mistimed with nothing on screen to show it. So the direct question stays for people who know their answer, and there is now a five-question alternative that asks how you actually feel rather than what you would hypothetically do: the reduced Morningness-Eveningness Questionnaire. It scores 4 to 25 and places you in one of three bands.

What the research says
The full questionnaire was validated against the circadian variation of oral temperature: morning types had a significantly earlier temperature peak than evening types, with intermediate types between. That is the reason we use it rather than something we wrote, because body temperature phase is the exact quantity everything on your plan is anchored to.

Delays are easier than advances

What we do
We move your trough later at 50 minutes per day and earlier at 30 minutes per day. That is what produces your days-to-adapt number.

What the research says
Both numbers come from the protocol this plan actually prescribes. Smith and Eastman ran timed bright light during the shift, dark glasses outdoors and scheduled dark sleep, and moved the melatonin onset 7.5 hours across nine days, which is 50 minutes a day. Burgess and colleagues advanced 28 adults with morning light: 1.5 hours over three days on intermittent bright light, so 30 minutes a day. The same paper notes an advance is harder to win than a delay.

Our judgement, not a published constant
These were 90 and 60 until we checked the literature, and both were roughly double what the studies report. The numbers here are sustained rates across a whole block. The first night or two of a delay move faster, up to about 96 minutes, so if you feel further along than the count suggests early on, you probably are. We would rather be slow and right: someone who believes they are adapted when they are not is the person driving home at their worst hour.

Night workers should aim to move the trough out of the shift, not fully invert

What we do
We target your trough landing a third of the way into your day sleep, rather than trying to make you nocturnal.

What the research says
Smith, Fogg and Eastman call this the compromise phase position: delay the sleepiest circadian time out of the shift and into the day sleep, instead of inverting completely, which leaves a worker stranded on their days off. For a 23:00 to 07:00 shift they targeted a sleepiest time of about 10:00, inside the first half of the sleep episode.

Last caffeine six hours before sleep

What we do
We work backwards from your sleep onset. Your first cup is timed to the start of your shift, not avoided.

What the research says
Caffeine has a half-life of roughly five hours. Expert consensus for shift workers endorses strategic caffeine before and during the shift, and explicitly rejects the standard sleep-hygiene advice to avoid it.

Naps are a countermeasure, not a weakness

What we do
If your schedule leaves room, we place a 90 minute nap ending an hour before you clock on, and tell you to set an alarm.

What the research says
Short naps of 15 to 20 minutes boost alertness; naps around 90 minutes reduce sleep debt; naps over 20 minutes carry a risk of sleep inertia on waking. Expert consensus rejects the standard advice against daytime napping for shift workers.

The days-off reset protocol

What we do
If you choose to get your days off back, we give you a short morning sleep, daylight on waking, and an early bedtime, rather than a full daytime sleep.

What the research says
Expert consensus endorses a short sleep in the morning after the last night shift, going back to bed earlier that night, with morning light helping realign to the day-night cycle.

Your crash window is the hour either side of your temperature minimum

What we do
We shade two hours on the clock face and tell you not to schedule judgement calls inside it.

What the research says
Alertness and reaction time fall to their lowest around the temperature minimum, and drowsy-driving crashes cluster in the pre-dawn window. The precise two-hour boundary is a display convention we chose.

Our judgement, not a published constant
There is no published constant that says the risk window is exactly two hours wide. We picked a readable boundary around a real low point. Treat the edges as soft.