Discovery limits in prospective studies

We discuss the procedures used in studies of a prospective nature to estimate the minimum sample size (or luminosity) needed to achieve a certain significance of signal above background, i.e. a so-called discovery limit. We propose a general unambiguous scheme to computer such a limit and compare its results to the different procedures used de facto in prospective studies. This allows us to point out the ambiguities and limited scope of application of some of the other methods.