How our data-quality findings are compiled

Our published findings — such as the annual OpenDQV Data Quality Report — are built from counts we derive from how workspaces use the Service: each workspace's sector (as given at signup), its pass rates expressed as banded ranges, and how many checks ran by rule family. They never include the records a workspace validates, record identifiers, rule or contract names, or anything that names a company, a workspace or a person. Every published figure rests on at least ten workspaces, on every dimension of every breakdown; anything ranked rests on at least twenty. Where a sector has fewer than ten contributing workspaces in a period, its figures are combined into a wider period or reported under “other” — never published on their own. A figure is also withheld where a single workspace would account for more than 40% of it. Workspaces operated by us for testing are excluded from every count. A workspace may exclude itself at any time using the report setting in its profile, and a named contact may object by emailing privacy@opendqv.com; both take effect for every edition whose cut-off has not passed.

This edition publishes one overall pass-rate band and, where a sector meets the floor, a pass-rate band by sector. It publishes no cross-tabulations (for example sector by size or sector by region) and no rankings. Because no two published figures share a subset of workspaces, nothing in this edition can be subtracted from anything else to reveal a smaller group. The observation window and the objection cut-off date for each edition are stated in the report itself. Methodology, thresholds and cadence may change between editions; the floor only ever tightens, and each edition's methodology page stays published at its own address so that every report can be read against the rules it was compiled under.