AbstractAcceptance sampling techniques are used to monitor the accuracy of gas meters. Random samples of meters are taken from homogeneous lots, and two accuracy measurements are recorded for each meter. In the past, the two measurements were averaged, and an acceptance sampling test applied to the sample of averages. In 1987, the plan was modified so that virtually the same test is applied to the two measurements individually. This new procedure is more stringent than the old procedure. In a study of data sampled over three years, the new plan rejects more lots than does the old plan, leading to greatly increased costs to the gas industry and therefore to the consumer. Theoretical reasons are given for why this occurs, and an alternative plan is proposed.