A new biometric model is presented for the rate of thermal disinfection of a microbial population. Unlike the usual approach, this model incorporates random errors arising at two stages into exponential kinetics. The first such stage is a result of sampling at the initial time of deposition of the microorganisms. The second stage corresponds to the variation arising from the subsequent experimentation. Maximum likelihood estimation of the 'decimal reduction time' parameter, tau, is described together with a numerical application and simulation study of the efficiency of estimation of tau under the usual model and the new model.