If A and B are values of type decimal, with scales scaleof(A) and scaleof(B), then: scaleof(A add B) = max(scaleof(A), scaleof(B)) scaleof(A sub B) = max(scaleof(A), scaleof(B)) scaleof(A mul B) = scaleof(A) + scaleof(B) scaleof(A mod B) = max(scaleof(A), scaleof(B)) scaleof(A div B) = computed with max decimal scale; how this is reported is service defined. client can cast to force scale. For decimal values with "floating" scale, result is decimal with floating scale.