A test of non-null hypothesis on proportion is proposed. It is derived from the classical one-, two- and multi-sample test on binomial proportion using the non-null hypothesis in place of the null hypothesis. The required sample size and power function are obtained by the same way. Its performance was demonstrated through simulation. It is quite simple in calculation. It reduces to its classical counterpart when the minimal clinically relevant difference equals zero. Its observed power coincides with the prescribed power. The test is appropriate with the clinical trials for the clinical significance of treatment-control difference or bio-equivalence. A worked example illustrates the methodology.