Which risk factor, at which threshold. Live.

What this rests on

• The sealed public record. releve-public.json in the repository, fingerprint f51ff169dbc4fd15, measured at commit 4469413 on 2026-09-05. The extractor that feeds this page verifies the seal, then recomposes witness cells with the tool’s own interval code, and refuses to emit if a single figure disagrees.
• Files the repository wrote. The labelled half is authored: archetypes of risk (shell layers, a PEP relative, cash-intensive trade, an offshore structure, undeclared turnover) and benign look-alikes (a local shop, a salaried resident, a domestic SME, a retiree) that resemble them. The labels ship with the files, debatable ones with their reasons.
• Synthetic variants, kept apart. Generated from the written cases, nature by nature, and never merged with the authored half: the toggle above switches the whole grid, it never blends.
• The tool’s selection rule. The slider holds your recall floor at the Wilson lower bound, exactly as npm run optimise does: a point estimate never clears a floor here.
• The declared tables, sealed in. Every weight the factors read (the country list, the activity and product tables, the exposure, structure, behaviour and tenure settings) is photographed inside the record this page reads: a figure never floats free of the tables it was measured under.

What this refuses

• Your data. This page cannot read it: no network requests leave it (connect-src ‘none’), no third-party resource is loaded, and there is no input field to paste a customer file into.
• Bare rates. Every cell carries its n and its 95 % interval; every factor of the registry is present tonight; an absent one would be a named row, never a guessed column.

Measure your own review history

The instrument shows our record. Yours is measured at home, by the tool, and nothing about your files leaves your machine.