A night in pieces is still a record.

The morning comes before the number.

The first thing Mama Bloom asks is how restored you feel. Three answers, none of them correct: restored, somewhere in between, drained. The check-in screen states its own limit—“Choose the closest answer. This is not a score.”

The order is deliberate. Mama Bloom asks first so that no number can arrive before your own account of the morning. The site puts it plainly: “Before you see a number, Mama Bloom listens to you.”

When sleep comes in pieces, this is the one entry that never breaks. A morning you can describe still exists on a night the sensor half missed.

Read the source before the shape.

Every night carries where it came from and how complete the sensed intervals are. Apple Health, a compatible band that writes sleep into Apple Health, or a manual chapter you added yourself. The app's own line: “Coverage and provenance are shown so confidence comes from visible boundaries, not an invented score.”

This matters more on broken nights than on smooth ones, because broken nights are where coverage thins. A watch left on the charger through the middle stretch did not produce a short night—it produced an unrecorded one, and the source line is what tells those two apart.

Stage estimates deserve the same caution. REM, Core, and Deep are inferred, they can be incomplete, and a manual entry has none of them: “Sleep stages remain unavailable because this interval was entered manually.” Nothing is invented to fill the gap.

Naps and split rest are rest.

The split sleep and naps view lays out a full 24 hours—noon, midnight, noon—and draws every recorded interval where it actually fell. Its heading counts the intervals rather than grading them: rest appeared in this many recorded intervals, totalling this much time.

The position is stated inside the view: “Mama Bloom keeps naps and split sleep visible. A different rest shape is context—not a consistency failure.” Two hours, an interruption, three more hours, and an hour after lunch is a shape. It is not a broken version of one ideal block.

Shifting clocks get the same treatment. “Travel nights are kept in their recorded local time so a time-zone change is not mistaken for poor routine.” Across Patterns, every night stays on the same clock, and nothing is stretched to look more regular.

A pattern is something that repeats.

One night cannot be read. Night Accord will not attempt a relationship until there are at least 3 prior recorded nights, and below that line it says so: “One night is a moment, not a pattern.”

Above it, the pair resolves into exactly one of three readings. Similar stories: “The two observations point in a similar direction, but they do not prove why.” Different stories: “The watch and your morning tell different stories. Both can be true.” Still learning, when there is not enough agreement yet. The default action is that no change is needed tonight; when the two disagree it becomes “No change tonight. Give this difference another morning before acting.”

Duration is described as longer or shorter only when it differs from the median of your recent nights by 25 minutes or more. Inside that band, the night is reported as close to your recent duration. Each reading carries the count it was compared against—your previous nights, named—so the boundary of the claim stays visible.

The practice is allowed to conclude nothing.

A practice is one small change tried across 14 nights, with each night marked by you as tried or not tried before. “Only nights you mark as tried can enter the comparison.” The comparison runs against your own preceding fortnight, the 14 days immediately before the practice began, and against nobody else's average.

Three thresholds must all be met before any conclusion appears: 3 prior nights, 3 tried nights, and 3 check-ins. Below any one of them the result reads “There is not enough tried-and-recorded data yet.”, under the eyebrow “AN HONEST RESULT”, with the reason attached: Mama Bloom “will not turn unconfirmed nights—or a few observations—into a verdict.” That is a designed outcome, not a shortfall. The evidence footer is shown either way, naming tried nights, recorded nights, prior nights, and check-ins.

Movement is only reported at 15 minutes or more of change in rest duration, or 25 percentage points or more of change in the share of restored mornings. Below both, the result is steady: “This change traveled quietly.” A neutral result is kept as it is, without being dressed up as a breakthrough.

Every result ends the same way. “This is an association in your own record—not proof that the practice caused it.” An association is a relationship between two things you recorded. It is not a cause, and no amount of repetition inside your own record converts it into one.

You are allowed to stop reading.

Some weeks the record is the loudest thing in the room. Quiet Week lets the numbers disappear without deleting your history. No broken streak. No guilt.

The boundary stays where it has always been. Mama Bloom is a wellness reflection, not a diagnosis. It does not treat a sleep disorder, detect disease, or replace professional care, and persistent or worrying sleep concerns belong with a qualified clinician.

Reading a record this way is slower than reading a score, and it deliberately gives you less. No verdict on last night. No number to beat. What it gives instead is a record that stays true when your nights refuse to be tidy: your own mornings in your own words, beside evidence that shows its edges. Some fortnights that adds up to one association worth carrying into tomorrow. Some fortnights it adds up to nothing, honestly labelled. Both are readings.