A record of Escherichia coli counts in the wash water of a commercial poultry plant was examined. The 3000 successive counts were “de-rounded” and then divided into 10 equal segments. Only occasionally were the counts weakly auto-correlated, primarily for lag one, indicating that successive counts are slightly more prone to represent a similar microbial load than counts taken days or weeks apart. The counts distribution in the segments was highly skewed. It could be approximated by the logLaplace distribution function, the parameters of which were determined by the method of moments (MM) and the maximum likelihood estimation (MLE). These parameters were used to estimate the frequencies of excessively high counts in the next segment of counts. The estimates calculated by the two methods (MM and MLE) were in reasonable agreement with the actual (rounded) observations. They were much closer to the observed values than estimates derived from the non-parametric approach, i.e. where the actual frequencies in one segment served as estimates of the frequencies in the next segment. These findings suggest that despite the crudeness of the procedure, the future frequencies of high counts, and the associated risks they may indicate can be estimated from the fluctuations pattern of past records not only in principle, but also in practice.