The standard membrane feeding assay (SMFA) is currently considered to be the 'gold standard' for assessing the effectiveness of malaria transmission blocking interventions (TBIs) in vivo. The operation and analysis of SMFAs has varied between laboratories: field scientists often measure TBI efficacy as a reduction in the prevalence of infected mosquitoes whilst laboratory scientists are more likely to quote efficacy as a change in the number of oocysts within the mosquito. These metrics give outputs that differ widely, resulting in a need for greater understanding of how the SMFA informs TBI assessment. Using data from 536 different assays (conducted on Plasmodium falciparum and Plasmodium berghei, in either Anopheles gambiae or Anopheles stephensi) it is shown that the relationship between these metrics is complex, yet predictable. Results demonstrate that the distribution of oocysts between mosquitoes is highly aggregated, making efficacy estimates based on reductions in intensity highly uncertain. Analysis of 30 SMFAs carried out on the same TBI confirms that the observed reduction in prevalence depends upon the parasite exposure (as measured by oocyst intensity in the control group), with assays which have lower exposure appearing more effective. By contrast, if efficacy is estimated as a reduction in oocyst intensity, then this candidate demonstrated constant efficacy, irrespective of the exposure level. To report transmission-blockade efficacy accurately, the results of SMFAs should give both the prevalence and intensity of oocysts in both the control and intervention group. Candidates should be assessed against a range of parasite exposures to allow laboratory results to be extrapolated to different field situations. Currently, many studies assessing TBIs are underpowered and uncertainties in efficacy estimates rarely reported. Statistical techniques that account for oocyst over-dispersion can reduce the number of mosquitoes that need to be dissected and allow TBI candidates from different laboratories to be accurately compared.