Cohen's h measures the distance between two proportions on the arcsine scale: h = 2·arcsin(√p₂) − 2·arcsin(√p₁). Unlike a p-value or a Wilson interval, it does not depend on sample size — it only depends on how far apart the two rates actually are.
|h| < 0.2: negligible — treat as noise, even if "significant."
0.2 ≤ |h| < 0.5: small effect — worth a note, rarely urgent alone.
0.5 ≤ |h| < 0.8: medium effect — usually worth investigating.
|h| ≥ 0.8: large effect — material drift, prioritise recalibration.