This translation has not been editorially reviewed yet. The German version is authoritative. Deutsche Fassung →
Measuring twice: why a varying finding isn't a finding
Every layout finding is collected twice per device profile. Only what both runs see gets reported. If a result varies between the measurements, that's exactly what the report states — instead of a finding nobody can reproduce.
- Layout measurements are time-dependent: fonts arrive, images get swapped in, scripts rebuild areas. Anyone who measures once occasionally gets a result that's gone on the next visit.
- The conservative evaluation is deliberate: a false positive costs the operator time searching for an error that doesn't exist, and damages trust in every other finding. A missed finding is annoying; a fabricated one is worse.
- For the same reason, undersized controls count by the smaller of the two measured numbers — what gets reported is what's certain.
