We optimize the design of future spectroscopic redshift surveys for constraining the dark energy via precision measurements of the baryon acoustic oscillations (BAO), with particular emphasis on the design of the Wide-Field Multi-Object Spectrograph (WFMOS). We develop a model that predicts the number density of possible target galaxies as a function of exposure time and redshift. We use this number counts model together with fitting formulae for the accuracy of the BAO measurements to determine the effectiveness of different surveys and instrument designs. We search through the available survey parameter space to find the optimal survey with respect to the dark energy equation-of-state parameters according to the Dark Energy Task Force Figure-of-Merit, including predictions of future measurements from the Planck satellite. We optimize the survey to test the LambdaCDM model, assuming that galaxies are pre-selected using photometric redshifts to have a constant number density with redshift, and using a non-linear cut-off for the matter power spectrum that evolves with redshift. We find that line-emission galaxies are strongly preferred as targets over continuum emission galaxies. The optimal survey covers a redshift range 0.8 < z < 1.4, over the widest possible area (6000 sq. degs from 1500 hours observing time). The most efficient number of fibres for the spectrograph is 2,000, and the survey performance continues to improve with the addition of extra fibres until a plateau is reached at 10,000 fibres. The optimal point in the survey parameter space is not highly peaked and is not significantly affected by including constraints from upcoming supernovae surveys and other BAO experiments.