A first order irreversible reaction is investigated using a stochastic model in order to account for the perturbations inherent to any chemical process. The main conclusion is that the concentration is distributed according to a lognormal law rather than to a normal one. This approach leads to a maximum likelihood estimate of the kinetic constant which is different with respect to the usual least squares estimate. Moreover, the stochastic model allows us to predict not only the expected value of the concentration, but to get a confidence interval, too. Statistical tests are provided in order to detect outliers, high-leverage and influential points.