Can models fitted to small data samples lead to maintenance policies with near-optimum cost?

In a general context of maintenance modelling, the excess cost of operating at the estimated optimum value of some decision variable (as opposed to the true unknown optimum) is related to the sample size of the dataset available for modelling. Although large sample sizes are required to estimate the optimum value of a decision variable to a high accuracy, it is shown that the cost which is optimised can be reduced to near its true minimum value even for small sample sizes. In fact, the expectation of the excess cost is inversely proportional to sample size. This implies that even modest sample sizes would be sufficient for the practical use of sensible models for determining cost-based maintenance policies. An example relating to a simple model for determining the optimum inspection interval for a repairable system illustrates these ideas. In particular, some results on the maintenance of medical equipment are presented.