Heterogeneity of Samples as a Problem in Shelf-Life Prediction.

A study of the common assumption that a single package of milk represents the production lot was made. Sensory and microbial observations on shelf-life were of sample sets of sequentially produced cartons of milk from four different commercial operations. Neither spoilage rate nor nature of spoilage was uniform for a typical sample set of ten units. A single package, therefore, provided low probability for predicting behavior of the entire production lot. Observations on individual colonies from standard plate counts at the time of sensory spoilage indicated the microflora to be a pure culture in each spoiled unit. The extreme differences in spoilage rates of individual units within sample sets indicated sensory evaluation of multiple samples to be the most logical, simple criterion for evaluating shelf-life. The size of the sample set to be observed and frequency of sampling awaits further observations and application of statistical techniques to establish the accuracy of estimates desired.