When a forward forecast appears

Before 8 weeks corrected age, Cocoon shows elapsed time, the typical range for the age, and caregiver-observed cues only — never a predicted time.

After that, the “Next up” card shows a usual window from the first day: until there is enough of your baby's own history it uses the typical range for their age and says so (“Typical for age · tuning to…”). A feed forecast is tuned to your baby once there are eight comparable milk-feed gaps across three days; a sleep forecast once there are eight comparable sleep intervals across five days.

What the calculation uses

The model looks back 14 days, gives recent observations more weight, and removes clear outliers. Feed patterns are separated by breast or bottle and by daytime or overnight. Sleep patterns compare a similar time-of-day or nap slot. Days you mark as illness, travel, childcare disruption, or bad data are excluded.

initial usual window =
  weighted 10th–90th percentile
  of comparable recent intervals

centre = weighted median

after 10 resolved forecasts =
  centre ± local 80th-percentile error

Why it is a range

Babies vary from day to day, so Cocoon says “usual feed window” or “usual sleep window,” never “hungry,” “late,” or “due.” Feeding cues, tired cues, safe-sleep guidance, and a clinician's plan always take precedence.

Learning from you

You can tell Cocoon whether a forecast's timing was too early, about right or too late. Together with when the next nap or feed actually came, those answers show whether the forecast leans early or late; once there are at least five and they agree, the centre of the window moves — by at most 30 minutes either way.

The app compares each displayed range with the next event and keeps a rolling 90-day history on this device. Edits and deletions are reconciled. After ten resolved forecasts, the interval uses the local 80th-percentile error around its centre. This history and your timing feedback are never uploaded.

Full-day schedules

“Today's rhythm” projects the rest of the day from the last wake. With fewer than five usable days of sleep logs it uses the typical wake windows and nap count for the age and is labelled “Typical for age”; from five days it is tuned to your baby's own pattern, and from fourteen it is marked “Tuned”. Each later interval is wider because uncertainty compounds. It is a planning aid, not a rule or a claim about the cause of a sleep pattern.

Evidence and review status

Population guidance supports total sleep ranges, safe sleep, and responsive feeding; it does not establish an exact timetable for an individual baby. The sources below were read and recorded by the Cocoon product team (September 2026). That is a review of published guidance, not a clinical endorsement of the forecast, and no clinician has signed off on an individual timetable — which is why every forecast is a range, shows its working, and defers to your baby's cues and your clinician's plan.

Limitations

Logging gaps, a new routine, illness, travel, and rapid developmental change can make recent history a poor guide. When there is not enough of your baby's own history, Cocoon says so and shows the typical range for their age instead, clearly labelled, rather than presenting a guess as a tuned forecast. Safe-sleep advice remains separate from forecasts.

Sources the team reviewed