Model deficiencies limit a subseasonal or seasonal forecast system's ability to produce accurate predictions. In this paper, we present an approach for transforming the output of a forecast system into a revised forecast, an approach designed to correct for some of the deficiencies in the system and thereby increase forecast skill. The approach, based on the joint consideration of the correlation structures present in the observational record and the inherent potential predictability of the model, is tested on a pre-existing subseasonal forecast experiment. It is shown to produce modest but significant increases in the accuracy of forecasted precipitation and near-surface air temperature at monthly timescales.