The hardware building blocks of all contemporary wavefront sensors for astronomical adaptive optics can be described as 'some front end optics, one or more CCD cameras, and a processing computer'. Selection of one of these sensors for a new installation should be based on a comparative cost/performance/risk evaluation. There are four levels of evaluation. First, we can calculate a 'phase error per photon' figure of merit inherent to the optical transformation. Second, we can evaluate the effect of various effects on the noise (precision) and accuracy of the sensors. Third, we can examine the complexity of the optical transformation from non-detectable wavefront to detectable intensity pattern and the concomitant processing complexity to extract the phase from the detected intensity. Finally, we can estimate the engineering difficulties in implementing the desired optical transformation. We suggest that the first level of examination, while providing an important, quantitative performance discriminator, does not provide a basis for sensor selection. The second level of evaluation, often approached qualitatively, suggests possible operational limits for the sensors. The third level suggests hidden difficulties, while the fourth level is perhaps a cost or risk discriminator.