This paper is aimed at clarifying the statistics of single molecule (SM) surface enhanced Raman scattering (SERS) signals. The argument of the possible existence of a Poisson distribution in the statistics of intensities in SM-SERS has been used many times in the last decade as a proof of single molecule detection. We show theoretically and experimentally that the conditions under which a Poisson distribution would be present are so unlikely to exist in a real system that there is no other option but to attribute the claims to poor statistical sampling. We believe the argument based on Poisson statistics should be dropped as a proof of single molecule detection in SERS.