Pausing protects your estimate
The expenditure algorithm only updates when it has enough evidence to update correctly. When your recent logging can't support a conclusion, the estimate holds at its last good value instead of drifting on bad data. Pausing is never a penalty — it's what keeps the number trustworthy.
The Expenditure screen shows a banner explaining exactly which condition isn't met, with a shortcut to fix it.
The reasons, and what fixes them
The algorithm checks these in order and shows you the first one that fails, so if two things are missing at once you will see the earlier one until you fix it.
| Banner | What it means | What fixes it |
|---|---|---|
| We're learning from your data | Not enough history yet. The algorithm needs 7 complete logged days inside its 21-day window before it can produce an estimate at all. | Keep logging. Logging from day one, this clears in your first week. |
| Estimate paused | Missing nutrition: fewer than 4 days of food logged in the last 7. Below that, a week's average intake isn't meaningful. | Log the days you missed, or mark the ones you can't reconstruct as incomplete. |
| Weigh in to resume your estimate | No weigh-in on record at all. Without a weight, intake can't be checked against anything. | One step on the scale. |
| Weigh in to refresh your estimate | Your last weigh-in is a week old or older. | One step on the scale. |
The first condition is the one everybody meets, because everybody starts there — your first three weeks covers it as a period rather than as a fault. It is also the one people meet again later without realising it: it is not about how long you have had the app, but about how many complete days sit inside the last three weeks.
Days that don't count
Two day-states interact with this deliberately:
- A day marked intentional fasting counts as a fully logged day of zero calories — fasting doesn't pause anything.
- A day marked incomplete logging is excluded from the calculation entirely — better to exclude a half-logged day than to let it read as a low-calorie day. Marking many days incomplete, though, can leave too few valid days and trigger a pause.
What a pause does and doesn't cost you
While paused, your targets stay where they are, your confidence level is preserved underneath, and the shaded Flux Range carries over from the last good reading. Updates resume automatically as soon as the missing data shows up — you don't have to do anything to restart them.
One thing worth knowing: a pause does not rewind the algorithm's maturity. The estimate does not go back to learning fast once you return. Whatever history it had accumulated before the pause is still there, so it picks up with the same careful step size it had when it stopped.