The aim of this paper is to set out criteria for defining trend and seasonal components in a time series. The criteria are set up primarily in terms of properties involving prediction. Because a structural time series model is set up in terms of components of interest, the relevant information on these components is given directly. It is shown that the Basic Structural Model has statistical properties, which are not dissimilar to the ARIMA model used by other authors, but the B. S. M. is only one model within a range of models all of which satisfy our proposed criteria. This methodology is applied to two series: US Investment and Industrial Production in Brazil.