A systematic strategy for design of optimum experiments in multidimensional coherent spectroscopy is described and applied to nuclear magnetic resonance (NMR). First an appropriate bound on the quantum dynamics is employed to determine a theoretical upper limit on coherence transfer efficiency which translates into the sensitivity of the experiment. Next it is investigated whether the theoretical limit can be achieved by a series of selective rotations. Then the final step serves to convert the selective rotations found into a nonselective pulse experiment which has a more general applicability. The example chosen for illustration is double- to single-quantum coherence transfer in INADEQUATE which is one of the most popular NMR experiments. Using the design strategy presented we show how the sensitivity of INADEQUATE can be doubled.