The confidence score, what 100 actually means

This page is a stub. The full guide is written against the shipped app version and is on its way.

The number in the top bar is the app grading its own certainty about your figures. It starts at 100 and findings subtract from it, recomputed live on every row change, fix, and import.

Weighting is by distortion, not by count. A finding costs points according to how badly it bends the tax figure while it stands. Missing cost basis and negative wallet balances cost 20 each, the number is wrong today while those exist. Other blocking or corrupting data errors cost 10. Warnings, the probably-wrongs like an unpaired transfer, cost 6. Informational findings cost nothing.

One problem is one problem, even five hundred times. Findings sharing a code form a family, and a family is capped, 30 points for errors, 18 for warnings, with the visible deductions rescaled so they still sum true. Five hundred copies of the same issue read as one issue, not as a destroyed score.

Dust is listed, not charged. Unpaired movements worth under £1 are shown so nothing is hidden and cost nothing, a receipt that small cannot distort any figure. A year whose tax rates are not built in also costs nothing, the position is computed and only the tax estimate is withheld.

Every lost point is navigable. Deductions are attributed per tax year and per wallet, and each names its row and its fix, so raising the score is a to-do list, not a mystery.

What 100 means, precisely. No finding stands that carries points, no missing cost basis, no wallet spending more than it received, no unpaired transfers, no blocking errors. It certifies that nothing the engine can detect is distorting the number. It does not certify that your history is complete, an exchange you never imported is invisible to any software, which is why record keeping matters alongside the score.