This paper introduces a new verification technique designed for, but not limited to, quantitative precipitation forecasts. It is tested on, and examples are given for, an intercomparison of very short-period nowcasting schemes. One of these nowcasters is a Bayesian scheme that is used in an extensive ensemble formulation, and the verification scheme is uniquely capable of treating both the ensemble members and the mean forecast. The verification scheme uses Procrustes shape analysis methods that are well established in statistics but have not, to date, been applied to meteorological forecast assessment. The Procrustes methodology allows for a decomposition of the forecast error into any number of components such as location (displacement), shape, size, orientation and intensity. Each error component can be afforded a separate weighting such that a cost or value of the forecast can be calculated that accounts for different error types. For example, a forecaster who is concerned with the location of a storm would place greater emphasis on correct location in the forecast than other attributes. This ability to apply weights makes the system particularly suited to real-time verification applications where confidence in the performance of the forecast translates into improved dissemination to users. In addition, the decomposition of the error into parts enables diagnosis of the error sources that can lead to model adjustment and improvement.