The increasingly widespread use of climate services for decision-making has highlighted the need for service developers to more clearly establish the benefits and limitations of the information they provide. Using a simple cost/loss framework applied to an idealised forecast system, we explore the critical level of accuracy required for the expected utility of a predictive service to exceed a benchmark based on the climatological frequency of the hazard. In this simplest case, the critical accuracy is a function only of the cost/loss ratio for the decision and the climatological frequency. Although the utility of climate services depends on a number of non-climate-related factors, comparing current forecasting capabilities to an estimate of the critical accuracy can provide a guide to the expected marginal benefit of the climate service. More generally, to ensure that climate service evaluation is relevant to users it must account for the finite duration over which a climate service is used, since the realised performance can differ significantly from its expectation. Here, the statistical distribution of realised performance is used to quantify the prior likelihood that the service will outperform a benchmark as a function of accuracy, climatological frequency, cost/loss ratio and duration of usage. This information can help users assess the potential risks associated with adopting a climate service for a given period, and may help identify geographical regions and meteorological parameters for which a forecasting tool could provide a worthwhile investment with respect to the next-best alternative.
