How it decides the hours
The idea is that the system regulates itself and that nobody has to tune anything. Underneath there are eight rules and a table; the numbers on this page are read from the model itself, so it cannot fall behind what the app actually does.
1. Where it starts from
Age gives a reference wake window: how long they stay awake between sleeps. It is the only thing there is on day one, and it is at the end of this page.
2. What it learns
Every time a sleep of at least 20 minutes follows an awake stretch, that window counts as evidence. Five-minute catnaps do not count: they say nothing about how long they last and they would only muddy the average. A nap that stops at a single cycle —under 45 minutes— teaches nothing about the window before it either: it says it was cut where naps get cut.
3. How it blends them
The reference for their age weighs as much as 3 observations, so within a few days it is your baby in charge and not the table. Recent counts for more: the average is exponential with a half-life of 5 days. Anything older than 21 days is dropped, and so is anything far outside the reference — a day in hospital, a button nobody tapped. At around 8 useful windows the confidence is full, and the app says at every moment how much of the forecast is theirs by now.
4. The shape of the day
A window is not worth the same in the morning as it is before bed: the first is shorter and the last is longer. Observations are normalised by their slot before being averaged, and the prediction applies the factor for the slot in question again. And after a short nap the next window shortens, but not by half: sleeping half as long does not leave a baby with half the wind in them, it leaves them with a little less.
5. How many naps
The other number the table only gets right at the beginning. It is learned the same way as the window but more slowly —half the weight every 7 days, forgotten after 28— and counting only days that are closed and have something recorded: a blank day almost never means they did not sleep, it means nobody tapped the button.
6. The night
It is not a window, it is the household's hour: it is anchored to the clock —the median of the last 14 nights— and it does not drag when a nap runs long. And how long it lasts is measured with the median of the last 10, not the mean: where the four o'clock waking sometimes gets recorded as two sleeps, half a dozen split nights drag a mean and do not move a median.
7. The whole day, not just the next one
The same sum repeated until the night: nap, window, nap, and when the day's naps run out, bed. It comes from the server and not from the browser, because two ways of working out the same hour end up saying different things on the same screen. And only what is left gets projected: what has already been slept is recorded and is shown exactly as it happened.
8. The days that have not arrived
The same sum over an empty day. As there is no last waking to start from, it starts from the time they usually get up, and never before 5 in the morning, which is where this model cuts one day off from the next. Every hour carries its margin, added in quadrature, because the morning moves and every window adds its own. 14 days is the limit: beyond that this repeats the same pattern, and offering it would be faking a precision that is not there.
The table by age
The starting point, and nothing more. As soon as there are sleeps recorded, what actually happens counts for more than this.
| Up to | Awake | Naps | Night |
|---|---|---|---|
| 1 months | 50 min | 5 | 8 h |
| 2 months | 65 min | 4 | 9 h |
| 3 months | 80 min | 4 | 10 h |
| 4 months | 95 min | 3 | 10 h |
| 5 months | 110 min | 3 | 10 h |
| 6 months | 125 min | 3 | 11 h |
| 8 months | 150 min | 3 | 11 h |
| 10 months | 170 min | 2 | 11 h |
| 12 months | 195 min | 2 | 11 h |
| 15 months | 225 min | 2 | 11 h |
| 18 months | 255 min | 1 | 11 h |
| 24 months | 285 min | 1 | 11 h |
| 36 months | 315 min | 1 | 11 h |
| beyond that | 330 min | 1 | 11 h |
What appears, and when
A newborn and a one-year-old do not need the same app, so napnap does not show the same thing to both. This is the whole list: nothing is hidden behind any other condition.
- From day one · The button, the feeds and whatever gets recorded. There is no rhythm to find yet, so napnap suggests no times and sends no alerts to the phone — unless this house has already taught it enough not to be guessing.
- From 2 months · A shape begins to appear. The day's forecast and the alerts are worth having by now, always with their margin in front of them.
- From 5 months · Solids appear, a month before the first spoonful, which is when it gets prepared. It can be brought forward in settings; the plan itself never starts before four months.
- From 6 months · The plan under way: a new food every two or three days, iron from the second week and every allergen with its own wait.
- From 10 months · Lumps, pieces and three meals. This is the stretch with a date on it: past ten months, carrying on with everything blended gets flagged once.
- From 12 months · They eat what the household eats and the naps are heading for one. Weight gets looked at every few months, not every month.
- From 24 months · One nap or none. From here on napnap is mostly a history and a report to take to the check-up.
And the food
Solids appear from 5 months onwards. 3 new foods go in each week, one every two or three days, with at most 2 major allergens so that a reaction has an owner. From the second week there is a source of iron every week. It takes 5 refusals with no acceptance at all to say something is not to their taste, and by 10 months the window for textures has closed.
There is no medical advice here. The order and the waits come from what has been published (WHO, AEP, AEPap, ESPGHAN, AECOSAN) and the decisions are yours, with your doctor.