Tracker apps tell you what changed since yesterday. This tells you whether it is inside your baby's own range.
Everything here is computed from your own child's history — nothing is compared to a population average.
A night is a chain of sleeps, not a clock window. It begins at the first sleep starting after 18:30 whose chain runs past 04:00, and it continues through every following sleep separated by a gap of an hour or less. The first gap longer than an hour ends the night. This keeps an early-evening catnap out of the front of the night and a morning nap out of the back of it — counting a 07:30 nap as night sleep after a 06:10 wake inflates the total by half an hour or more, which is enough to invent a trend that is not there.
Day sleep is everything from that morning's final wake until 18:30 the same day, so the day and the night that follows it are paired and nothing is counted twice. The anchor is the longest single block of a night. Time in bed runs from the first sleep of the night to the final wake, and sleep efficiency is the night total as a share of it.
Night sleep can rise for two unrelated reasons: your baby sleeps better, or your baby is in bed longer. Only efficiency separates them, so when the total moves, this reports the split rather than the total alone. A later recorded morning wake also lengthens the night — and once a baby starts waking quietly and playing, the recorded wake is only an upper bound, so a gain sitting at that end is the least trustworthy kind.
Sentences about the baseline describe a seven-night average against the fourteen before it, and always include last night in the recent window. Sentences about last night compare that one night to the spread of the twenty-one before it. The two are never mixed: a rolling average that has drifted does not mean last night drifted, and last night's outlier does not mean the baseline moved.
A trend is a third thing again, and it is the only one that can say the range itself has moved. It fires when the last seven nights clear the fourteen before them by more than 1.5 standard deviations of that earlier window and at least four of the seven fall outside it. Both conditions are required: the first alone fires on one extreme night, the second alone fires on drift too small to matter. When a trend is running it is reported first, because a single night outside a range that is itself moving means much less than the movement does.
The band is the middle 80% of the last 21 nights — the normal range. The thin line is the median. The dot is the most recent night. Feed metrics use only the last 16 days, since logging habits drift.
A run of nights outside the band only counts as a run if they fall on the same side of it. A long night does not extend a run of short ones. Measures that swing more than 12% from night to night can go outside range without that meaning anything, so they are reported without a cause suggested. A miss smaller than the measure's own resolution is not reported at all.
When something changes about how a number is produced rather than about your baby — a different diaper at night, a different way of stopping the feed timer — the old range no longer applies. Those measures are held back until enough nights accumulate under the new arrangement.
Your tracker app uses different day boundaries, so its totals will not match these. Neither is wrong; they answer different questions.
This is a descriptive tool, not medical advice. It describes what is usual for one child and cannot tell you whether that is healthy. Growth, feeding adequacy, and anything that worries you belong with your pediatrician.