In Quantitative Information Flow, refinement expresses the strong property that one channel never leaks more than another. Since two channels are then typically incomparable, here we explore a family of refinement quasimetrics offering greater flexibility. We show these quasimetrics let us unify refinement and capacity, we show that some of them can be computed efficiently via linear programming, and we establish upper bounds via the Earth Mover's distance. We illustrate our techniques on the Crowds protocol.