Timers & measurement

Recura keeps two kinds of numbers and never lets them masquerade as each other:

• Estimates are plans — the minutes you think a task takes, the budget you gave a routine.
• Measurements are history — what the timer actually clocked.

Surfaces that show history label which is which: a day’s log says “12m timed” for measured work and “~15 min” for a plan; heatmap totals count only timed minutes. A checkbox tap records that you did a thing, not how long — so it honestly contributes nothing to time totals.

Heatmaps

Every level has one — your year on Today and Stats, each routine’s run history, each task’s own record. The cells encode three honest states:

• filled — things were completed that day (intensity = how many, or measured minutes if you flip the toggle);
• dashed — something was skipped that day (an observation, not a stain);
• hollow — things were due and nothing happened. Stated once, never nagged about.

Days credit when the work happened: catch-up backdates fill the past, live runs fill today.

Recalibration

Estimates drift — you get faster at routines, tasks grow steps. Recura watches the drift and offers corrections, never applies them:

• Task estimates: after enough timed runs, a suggestion appears — “runs at ~25m (p90 of 9 runs), estimate says 15.” One tap to adopt.
• Routine budgets: same statistic over whole-session wall time — “runs at ~50m, budget is 30.”
• Cadences: the weekly report compares the rhythm you declared with the rhythm you actually keep (“set every 7 days, done every ~11”) and offers to match reality.

The p90 statistic is chosen deliberately: “fits 9 times out of 10” is what a budget should mean, and it shrugs off the one run where you forgot to stop the timer.

The weekly report

Sundays and Mondays, Today offers last week’s report: completions against schedule, cadence drift, per-person breakdown. It states what happened. What to do about it stays your call.